EPA/635/R-11/002F
^ ^mf^mi www.epa.gov/iris
TOXICOLOGICAL REVIEW
OF
LIBBY AMPHIBOLE ASBESTOS
In Support of Summary Information on the
Integrated Risk Information System (IRIS)
December 2014
(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 has been reviewed in accordance with U.S. Environmental Protection
Agency policy and approved for publication. Mention of trade names or commercial products
does not constitute endorsement or recommendation for use.
11
-------
CONTENTS—TOXICOLOGICAL REVIEW OF LIBBY AMPHIBOLE ASBESTOS
LIST OF TABLES ix
LIST OF FIGURES xvi
FOREWORD xxii
AUTHORS, CONTRIBUTORS, AND REVIEWERS xxiii
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. EPAHealth 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 2-11
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. Libby Amphibole Asbestos 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
3.3.2.1. Inflammation and Reactive Oxygen Species (ROS)
Production 3-16
in
-------
CONTENTS (continued)
3.3.2.2. Genotoxicity 3-16
3.3.2.3. Carcinogen!city 3-16
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
4.1.3. Other Effects, Noncancer 4-37
4.1.3.1. Cardiovascular Disease 4-37
4.1.3.2. Autoimmune Disease and Autoantibodies 4-38
4.1.4. Cancer Effects 4-41
4.1.4.1. Lung Cancer 4-41
4.1.4.2. Mesothelioma 4-46
4.1.4.3. Other Cancers 4-50
4.1.4.4. Summary of Cancer Mortality Risk in Populations Exposed
to Libby Amphibole Asbestos 4-50
4.1.5. Comparison With Other Asbestos Studies—Environmental Exposure
Settings 4-51
4.2. SUBCHRONIC- AND CHONIC-DURATION STUDIES AND CANCER
BIO AS SAYS IN ANIMALS—ORAL, INHALATION, AND OTHER
ROUTES OF EXPOSURE 4-53
4.2.1. Inhalation 4-62
4.2.2. Intratracheal Instillation Studies 4-63
4.2.3. Injection/Implantation Studies 4-65
4.2.4. Oral 4-66
4.2.5. Summary of Animal Studies for Libby Amphibole and Tremolite
Asbestos 4-67
4.3. OTHER DURATION- OR ENDPOINT-SPECIFIC STUDIES 4-69
4.3.1. Immunological 4-69
4.4. MECHANISTIC DATA AND OTHER STUDIES IN SUPPORT OF THE
MODE OF ACTION 4-71
4.4.1. Inflammation and Immune Function 4-78
4.4.2. Genotoxicity 4-81
4.4.3. Cytotoxicity and Cellular Proliferation 4-83
4.5. SYNTHESIS OF MAJOR NONCANCER EFFECTS 4-84
4.5.1. Pulmonary Effects 4-85
4.5.1.1. Pulmonary Fibrosis (Asbestosis) 4-85
IV
-------
CONTENTS (continued)
4.5.1.2. Other Nonmalignant Respiratory Diseases 4-86
4.5.2. Pleural Effects 4-86
4.5.3. Other Noncancer Health Effects (Cardiovascular Toxicity,
Autoimmune Effects) 4-87
4.5.4. Summary of Noncancer Health Effects of Exposure to Libby
Amphibole Asbestos 4-88
4.5.5. Mode-of-Action Information (Noncancer) 4-88
4.6. EVALUATION OF CARCINOGENICITY 4-90
4.6.1. Summary of Overall Weight of Evidence 4-90
4.6.1.1. Synthesis of Human, Animal, and Other Supporting Evidence.... 4-90
4.6.2. Mode-of-Action Information (Cancer) 4-92
4.6.2.1. Description of the Mode-of-Action Information 4-92
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 4-101
4.7. SUSCEPTIBLE POPULATIONS 4-102
4.7.1. Influence of Different Life Stages on Susceptibility 4-102
4.7.1.1. Life-Stage Susceptibility 4-103
4.7.2. Influence of Gender on Susceptibility 4-107
4.7.3. Influence of Race or Ethnicity on Susceptibility 4-107
4.7.4. Influence of Genetic Polymorphisms on Susceptibility 4-108
4.7.5. Influence of Health Status on Susceptibility 4-109
4.7.6. Influence of Lifestyle Factors on Susceptibility 4-110
4.7.7. Susceptible Populations Summary 4-110
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
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) 5-40
-------
CONTENTS (continued)
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 5-46
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) 5-51
5.3.1. Uncertainty in the Exposure Reconstruction 5-51
5.3.2. Uncertainty in the Radiographic Assessment of Localized Pleural
Thickening (LPT) 5-56
5.3.3. Uncertainty Due to Potential Confounding 5-57
5.3.4. Uncertainty Due to Time Since First Exposure (TSFE) 5-61
5.3.5. Uncertainty in the Endpoint Definition 5-65
5.3.6. Summary of Sensitivity Analyses 5-69
5.4. CANCER EXPOSURE-RESPONSE ASSESSMENT 5-70
5.4.1. Overview of Methodological Approach 5-70
5.4.2. Choice of Study/Data—with Rationale and Justification 5-72
5.4.2.1. Descripti on of the Libby Worker Cohort 5-73
5.4.2.2. Description of Cancer Endpoints 5-75
5.4.2.3. Description of Libby Worker Cohort Work Histories 5-77
5.4.2.4. Description of Libby Amphibole Asbestos Exposures 5-78
5.4.2.5. Estimated Exposures Based on Job-Exposure Matrix (JEM)
and Work Histories 5-85
5.4.3. Exposure-Response Modeling 5-90
5.4.3.1. Modeling of Mesothelioma Exposure Response in the Libby
Worker Cohort 5-91
5.4.3.2. Results of the Analysis of Mesothelioma Mortality in the Full
Libby Worker Cohort 5-93
5.4.3.3. Modeling and Results of Lung Cancer Exposure Response in
the Full Libby Worker Cohort 5-96
5.4.3.4. Rationale for Analyzing the Subcohort of Libby Workers
After 1959 5-101
VI
-------
CONTENTS (continued)
5.4.3.5. Results of the Analysis of Mesothelioma Mortality in the
Subcohort 5-103
5.4.3.6. Results of the Analysis of the Lung Cancer Mortality in the
Subcohort 5-113
5.4.3.7. Sensitivity Analysis of the Influence of High Exposures in
Early 1960s on the Model Fit in the Subcohort 5-124
5.4.3.8. Additional Analysis of the Potential for Confounding of
Lung Cancer Results by Smoking in the Subcohort 5-126
5.4.4. Exposure Adjustments and Extrapolation Methods 5-127
5.4.5. Inhalation Unit Risk (IUR) of Cancer Mortality 5-128
5.4.5.1. Unit Risk Estimates for Mesothelioma Mortality 5-128
5.4.5.2. Unit Risk Estimates for Lung Cancer Mortality 5-131
5.4.5.3. Inhalation Unit Risk (IUR) Derivation for Combined
Mesothelioma and Lung Cancer Mortality 5-132
5.4.6. Uncertainties in the Cancer Risk Values 5-140
5.4.6.1. Sources of Uncertainty 5-140
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
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
vn
-------
CONTENTS (continued)
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
APPENDIX J: DOCUMENTATION OF IMPLEMENTATION OF THE 2011
NATIONAL RESEARCH COUNCIL RECOMMENDATIONS J-l
Vlll
-------
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) at 0.01 fiber/cc 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 4-22
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
IX
-------
LIST OF TABLES (continued)
4-10. Prevalence of pleural thickening in 280 participants according to various
cofactors 4-26
4-11. Pathological alterations of lung parenchyma and pleura in community-based
studies 4-30
4-12. Pulmonary function and respiratory symptoms and conditions 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 4-36
4-15. Autoimmune-related studies in the Libby, MT community 4-40
4-16. Respiratory (lung) cancer mortality and exposure-response analyses based
on related studies of the vermiculite mining and milling workers in Libby,
MT 4-42
4-17. Mesothelioma mortality risk based on studies of the vermiculite mine
workers in Libby, MT 4-48
4-18. Exposure levels and health effects observed in communities exposed to
tremolite, chrysotile, and crocidolite asbestos 4-52
4-19. In vivo data following exposure to Libby Amphibole asbestos 4-55
4-20. In vivo data following exposure to tremolite asbestos 4-61
4-21. In vitro data following exposure to Libby Amphibole asbestos 4-74
4-22. In vitro data following exposure to tremolite asbestos 4-76
4-23. Hypothesized modes of action for carcinogenicity of Libby Amphibole
asbestos in specific organs 4-100
5-1. Summary of candidate principal studies on LAA for reference concentration
(RfC) derivation 5-6
-------
LIST OF TABLES (continued)
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 who did not report any previous
occupational exposure to asbestos 5-18
5-5. Modelsa 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
5-7. Model features considered in exposure-response modeling to develop a
point of departure (POD) 5-28
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) 5-31
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% 5-38
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 5-48
5-11. Multiple derivations of a reference concentration from the Marysville, OH
cohort 5-49
5-12. Exposure distribution among workers at the O.M. Scott plant in Marysville,
OH 5-53
XI
-------
LIST OF TABLES (continued)
5-13. Effect of truncating exposures after 1980 and of using arithmetic or
geometric mean to summarize multiple fiber measurements 5-55
5-14. Effect of including covariates into the final model 5-60
5-15. Effect of different assumptions for the plateau parameter 5-62
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 5-64
5-17. Effect of using different case/noncase definitions 5-66
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 5-68
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) 5-70
5-20. Demographic, mortality, and exposure characteristics of the Libby worker
cohort 5-74
5-21. Exposure intensity (fibers/cc) for each location operation from the beginning
of operations through 1982 5-79
5-22. Demographic, mortality, and exposure characteristics of the subset of the
Libby worker subcohort hired after 1959 5-83
5-23. Mesothelioma mortality rate shown by duration of exposure (yr) in the full
Libby worker cohort including all hires (n = 1,871) 5-93
5-24. Mesothelioma mortality rate shown by age at first exposure in the full Libby
worker cohort including all hires (n = 1,871) 5-93
xn
-------
LIST OF TABLES (continued)
5-25. Mesothelioma mortality rate shown by time since first exposure (TSFE) in
the full Libby worker cohort including all hires (n = 1,871) 5-94
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) 5-95
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) 5-97
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) 5-97
5-30. Mesothelioma mortality rate in the subcohort of employees hired after 1959
shown by duration of exposure (yr) 5-104
5-31. Mesothelioma mortality rate in the subcohort of employees hired after 1959
shown by age at first exposure 5-104
5-32. Mesothelioma mortality rate in the subcohort of employees hired after 1959
shown by time since first exposure (TSFE) 5-104
5-33. Comparison of model fit of exposure metrics for mesothelioma mortality in
the subcohort hired after 1959 5-105
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 5-107
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 5-107
5-36. Mesothelioma mortality rate in the subcohort of employees hired after 1959
for the Peto model 5-107
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 5-107
xin
-------
LIST OF TABLES (continued)
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 5-108
5-39. Mesothelioma mortality exposure metrics fits, slopes per day, and credible
intervals in the subcohort of employees hired after 1959 5-112
5-40. Peto model and Peto model with clearance fits, slopes per year, and credible
intervals in the subcohort of employees hired after 1959 5-113
5-41. Lung cancer mortality rate in the subcohort of employees hired after 1959
shown by duration of exposure (yr) 5-114
5-42. Lung cancer mortality rate in the subcohort of employees hired after 1959
shown by age at first exposure 5-114
5-43. Lung cancer mortality rate in the subcohort of employees hired after 1959
shown by time since first exposure (TSFE) 5-114
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 5-116
5-45. Lung cancer mortality exposure metrics fits, slopes, and confidence intervals
(CI) for all retained metrics from Table 5-44 5-120
5-46. Sensitivity analysis of model fit comparison for different exposure metrics
and mesothelioma mortality associated with LAA 5-125
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 5-126
5-48. Unit risks for the Peto model and Peto model with clearance 5-128
5-49. Mesothelioma mortality exposure metrics unit risks for the subcohort hired
after 1959 5-129
5-50. Mesothelioma unit risks for the subcohort hired after 1959 adjusted for
underascertainment 5-130
xiv
-------
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 5-130
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 5-131
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 LAA) for
different combination of mesothelioma and lung cancer models 5-133
5-54. Lung cancer regression results from different analyses of cumulative
exposure in the cohort of workers in Libby, MT 5-136
5-55. Mesothelioma analysis results from different analyses of cumulative
exposure in the Libby workers cohort 5-140
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
xv
-------
LIST OF FIGURES
2-1. Vermiculite mining operation onZonolite 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 (shown
along the fiber axis) 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
xvi
-------
LIST OF FIGURES (continued)
4-2. A (left). Gross appearance at autopsy of asbestos-associated pleural plaques
overlying the lateral thoracic wall [(ATS, 2004) Figure 12]. Figure 4-2. B
(right). Gross appearance of large asbestos-related pleural plaque over the
dome of the diaphragm [(ATS, 2004) Figure 13] 4-19
4-3. Lung cancer mortality risk among workers in the Libby, MT vermiculite
mine and mill workers 4-45
4-4. Proposed mechanistic events for carcinogenicity of asbestos fibers 4-72
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
5-2. Radiographic outcomes among Marysville, OH workers 5-13
5-3. Plot of exposure-response models for probability of localized pleural
thickening (LPT) as a function of mean concentration of occupational
exposure in the subcohort 5-34
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% 5-40
5-5. Plot of the National Institute for Occupational Safety and Health (NIOSH)
job-exposure matrix for different job categories overtime 5-84
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 5-109
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 5-110
xvn
-------
LIST OF FIGURES (continued)
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 5-111
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 5-122
xvin
-------
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 radiograph!c change
ATS American Thoracic Society
ATSDR Agency for Toxic Substances and Disease Registry
BALF bronchoalveolar lavage fluid
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
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
xix
-------
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 ceramic fibers
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
xx
-------
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/La autoantibody marker
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
xxi
-------
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 mixture of amphibole
fibers identified in the Rainy Creek complex and present in ore from the vermiculite mine near
Libby, MT. It is not intended to be an assessment of the toxicity of asbestos generally (nor 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.
The intent of Appendix J, Documentation of Implementation of the 2011 National
Research Council Recommendations, is to present the IRIS Program's implementation of the
NRC recommendations. Implementation is following a phased approach that is consistent with
the NRC's "Roadmap for Revision" as described in Chapter 7 of the formaldehyde review
report.
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).
xxn
-------
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
Formerly with the 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
Formerly with the National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
Maureen R. Gwinn, PhD, DABT, ATS
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
xxin
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS (continued)
CONTRIBUTING AUTHORS
Rebecca Dzubow, MPH, MEM
Formerly with the National Center for Environmental Assessment
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
xxiv
-------
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
Lisa Walker
ECFlex, Inc., Fairborn, OH
Heidi Glick
Crystal Lewis
Carman Parker-Lawler
Lana Wood
IntelliTech Systems, Inc., Fairborn, OH
Cris Broyles
xxv
-------
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
xxvi
-------
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
xxvn
-------
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
XXVlll
-------
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
XXIX
-------
1. INTRODUCTION
This document presents background information and justification for the Integrated Risk
Information System (IRIS) summary of the hazard and exposure-response assessment of Libby
Amphibole asbestos (LAA),l a mixture of amphibole fibers identified in the Rainy Creek
complex and present in ore from the vermiculite mine near Libby, MT. IRIS summaries may
include oral reference dose (RfD) and inhalation reference concentration (RfC) values for
chronic exposure durations, and a carcinogenicity assessment. This assessment reviews the
potential hazards, both cancer and noncancer health effects, from exposure to LAA and provides
quantitative information for use in risk assessments: an RfC for noncancer health effects and an
inhalation unit risk (IUR) addressing cancer risk. LAA-specific data are not available to support
RfD or cancer slope factor derivations for oral exposures.
An RfC is defined as "an estimate (with uncertainty spanning perhaps an order of
magnitude) of an exposure (including sensitive subgroups) that is likely to be without an
appreciable risk of adverse health effects over a lifetime." (U.S. EPA, 2002). In the case of
LAA, the RfC is expressed in terms of the lifetime exposure in units of fibers per cubic
centimeter of air (fibers/cc) in units of the fibers as measured by phase contrast microscopy
(PCM). The inhalation RfC for LAA considers toxic effects for both the respiratory system
(portal of entry) and for effects peripheral to the respiratory system (extrarespiratory or systemic
effects) that may arise after inhalation of LAA.
The carcinogenicity assessment provides information on the carcinogenic hazard
potential of the substance in question, and quantitative estimates of risk from inhalation
exposures are derived. The information includes a weight-of-evidence judgment of the
likelihood that the agent is a human carcinogen and the conditions under which the carcinogenic
effects may be expressed. Quantitative risk estimates are derived from the application of a
low-dose extrapolation procedure from human data. An inhalation unit risk (IUR) is typically
defined as a plausible upper bound on the estimate of cancer risk per ug/m3 air breathed for
70 years. For LAA, the RfC is expressed as a lifetime daily exposure in fibers/cc (in units of the
fibers as measured by PCM), and the IUR is expressed as cancer risk per fibers/cc (in units of the
fibers as measured by PCM).
Development of these hazard identification and exposure-response assessments for LAA
has followed the general guidelines for risk assessment as set forth by the National Research
Council (NRC, 1983). U.S. Environmental Protection Agency (EPA) Guidelines and Risk
Assessment Forum technical panel reports that may have been used in the development of this
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.
1-1
-------
Mixtures (U.S. EPA. 1986c), Guidelines for Mutagenicity Risk Assessment (U.S. EPA. 1986b),
Recommendations for and Documentation of Biological Values for Use in Risk Assessment (U.S.
EPA, 1988b), Guidelines for Developmental Toxicity Risk Assessment (U.S. EPA, 199 la),
Interim Policy for Particle Size and Limit Concentration Issues in Inhalation Toxicity (U.S. EPA,
1994a), Methods for Derivation of Inhalation Reference Concentrations and Application of
Inhalation Dosimetry (U.S. EPA, 1994b), Use of the Benchmark Dose Approach in Health Risk
Assessment {U.S. EPA, 1995), Guidelines for Reproductive Toxicity Risk Assessment (U.S. EPA,
1996), Guidelines for Neurotoxicity Risk Assessment (U.S. EPA, 1998), Science Policy Council
Handbook: Risk Characterization (U.S. EPA, 2000b), Benchmark Dose Technical Guidance
Document (U.S. EPA, 2012), Supplementary Guidance for Conducting Health Risk Assessment
of Chemical Mixtures (U.S. EPA, 2000c), A Review of the Reference Dose and Reference
Concentration Processes (U.S. EPA, 2002), Guidelines for Carcinogen Risk Assessment (U.S.
EPA, 2005a), Supplemental Guidance for Assessing Susceptibility from Early-Life Exposure to
Carcinogens (U.S. EPA, 2005b), Science Policy Council Handbook: Peer Review (U.S. EPA,
), and A Framework for Assessing Health Risks of Environmental Exposures to Children
(U.S. EPA, 2006b).
The literature search strategy employed for this assessment is based on EPA's National
Center for Environmental Assessment's Health and Environmental Research Online database
tool (which includes PubMed, MEDLINE, Web of Science, JSTOR, and other literature
sources). The key search terms included the following: Libby Amphibole, tremolite, asbestos,
richterite, winchite, amphibole, and Libby, MT. The relevant literature was reviewed through
July 2011. Any pertinent scientific information submitted by the public to the IRIS Submission
Desk was also considered in the development of this document. Note that references have been
added to the Toxicological Review after the external peer review SAB (2013) in response to peer
reviewers' comments and for the sake of completeness.
1.1. RELATED ASSESSMENTS
1.1.1. Integrated Risk Information System (IRIS) Assessment for Asbestos (U.S. EPA,
1988a)
The IRIS assessment for asbestos was posted online in IRIS in 1988 and includes an IUR
of 0.23 excess cancers per 1 fiber/cc (U.S. EPA, 1988a): this unit risk is given in units of the
fibers as measured by PCM. The IRIS IUR2 for general asbestos (CAS Number 1332-21-4) is
derived by estimating excess cancers for a continuous lifetime exposure and is based on the
central tendency—not the upper bound—of the risk estimates (U.S. EPA, 1988a) and is
applicable to exposures across a range of exposure environments and types of asbestos.
Although other cancers have been associated with asbestos [e.g., laryngeal, stomach, ovarian
2For purposes of this document, termed "IRIS IUR."
1-2
-------
(Straif etal., 2009)1, the IRIS IUR for asbestos accounts for only lung cancer and mesothelioma.
Additionally, pleural and pulmonary effects from asbestos exposure (e.g., pleural thickening,
asbestosis, and reduced lung function) are well documented, although there is no RfC for these
noncancer health effects on the IRIS database (U.S. EPA. 1988a).
The derivation of the unit risk for general asbestos is based on the Airborne Asbestos
Health Assessment Update [AAHAU (U.S. EPA, 1986a)]. The AAHAU provides various cancer
potency factors and mathematical models of lung cancer and mesothelioma mortality based on
synthesis of data from occupational studies and presents estimates of lifetime cancer risk for
continuous environmental exposures [0.0001 fiber/cc and 0.01 fiber/cc; (see Table 6-3 of (U.S.
EPA, 1986a)]. For both lung cancer and mesothelioma, life-table analysis was used to generate
risk estimates based on the number of years of exposure and the age at onset of exposure.
Although various exposure scenarios were presented, the unit risk is based on a lifetime
continuous exposure from birth. The final asbestos IUR is 0.23 excess cancers per 1 fiber/cc
continuous exposure3 and was posted on the IRIS database in 1988 [(U.S. EPA, 1988a) see Table
1-1 below].
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) at 0.01 fiber/cc
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 lO'1
2.42 x 10-1
2.30 x 1Q-1
Unit risk
(per fiber/cc)
0.23
aData are for exposure at 0.01 fiber/cc for a lifetime.
Source: U.S. EPA(1988a).
The IRIS database has an IUR 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)1. Some uncertainty remains in applying the
resulting IUR for asbestos to exposure environments and minerals different from those analyzed
3An IUR of 0.23 for general asbestos can be interpreted as 0.23 excess risk of death from mesothelioma or lung
cancer per person for each 1 fiber/cc increase in continuous lifetime exposure. Thus, as shown in Table 1-1, for
100,000 people exposed at a concentration of 0.01 fiber/cc, 230 excess deaths would be expected
[IUR x Concentration x Number of people = (0.23 excess cancer deaths per fiber/cc per person) x (0.01 fiber/cc) x
(100,000 people) = 230 excess cancer deaths]. "Fiber/cc" is a commonly used measure; it is equivalent to 1 million
fibers per cubic meter of air while 0.01 fiber/cc is 10,000 fibers per cubic meter of air.
1-2
-------
in the AAHAU (U.S. EPA. 1986a). No RfC, RfD, or oral slope factor are derived for asbestos
on the IRIS database (U.S. EPA. 1988a).
1.1.2. EPA Health Assessment for Vermiculite (U.S. EPA, 1991b)
An EPA health assessment for vermiculite reviewed available health data, including
studies on workers who mined and processed ore with no significant amphibole fiber content.
The cancer and noncancer health effects observed in the Libby, MT worker cohort were not seen
in studies of workers exposed to mines with similar exposure to vermiculite but much lower
exposures to asbestos fibers. Therefore, it was concluded that the health effects observed from
the materials mined from Zonolite Mountain near Libby, MT, were most likely due to amphibole
fibers and not the vermiculite itself (U.S. EPA, 1991b). At the time, EPA recommended the
application of the IRIS IUR for asbestos fibers (0.23 per fiber/cc) in addressing potential risk of
the amphibole fibers entrained in vermiculite mined in Libby, MT.
1.2. LIBBY AMPHIBOLE ASBESTOS-SPECIFIC HUMAN HEALTH ASSESSMENT
LAA is a complex mixture of amphibole fibers—both mineralogically and
morphologically (see Section 2.3). The mixture primarily includes winchite, richterite, and
tremolite fibers with trace amounts of magnesio-riebeckite, edenite, and magnesio-arfvedsonite.
These fibers exhibit a complete range of morphologies from prismatic crystals to asbestiform
fibers (Meeker et al., 2003). Epidemiologic studies of workers exposed to LAA fibers indicate
increased lung cancer and mesothelioma, as well as asbestosis and other nonmalignant
respiratory diseases (Larson et al., 201 Ob; Larson et al., 2010a; Moolgavkar et al., 2010; Rohs et
al., 2008; Sullivan, 2007; McDonald et al., 2004, 2002; Amandus et al., 1988; Amandus et al.,
1987b; Amandus and Wheeler, 1987; Amandus et al., 1987a; McDonald et al., 1986a; McDonald
et al., 1986b; Lockev et al., 1984).
1-4
-------
2. LIBBY AMPHIBOLE ASBESTOS: GEOLOGY AND EXPOSURE POTENTIAL
2.1. INTRODUCTION
Libby is a community in northwestern Montana that is located near a large open-pit
vermiculite mine that operated from the mid 1920s to 1990 (see Figure 2-1). Vermiculite is a
silicate mineral that exhibits a sheet-like structure similar to mica (see Figure 2-2, Panel A).
When heated to approximately 870°C, water molecules between the sheets change to vapor and
cause the vermiculite to expand like popcorn into a light, porous material (see Figure 2-2,
Panel B). This process of expanding vermiculite is termed "exfoliation" or "popping." Both
unexpanded and expanded vermiculite have found a range of commercial applications, the most
common of which include packing material, attic and wall insulation, various garden and
agricultural products, and various cement and building products.
Figure 2-1. Vermiculite mining operation on Zonolite Mountain, Libby, MT.
2-1
-------
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.
The primary product from the mine was vermiculite concentrate, which was produced by
milling, screening, and grading the raw ore to enrich for the vermiculite mineral. In general,
mining practices sought to exclude nonvermiculite material when harvesting the ore, and
subsequent processing steps were designed to eliminate nonvermiculite materials from the
2-2
-------
finished product. Nevertheless, small amounts of other minerals from the ore body tended to
remain in the vermiculite (Zonolite) product. This included a form of asbestos referred to as
Libby Amphibole asbestos (LAA).
This chapter provides a brief description of the mineralogical characteristics of asbestos
(see Section 2.2), an overview of methods used to identify and measure asbestos in air and solid
materials (see Section 2.3), a review of the mineralogical characteristics of LAA in particular
(see Section 2.4), and an overview of the potential for current human exposures to LAA (see
Section 2.5).
2.2. GEOLOGY AND MINERALOGY OF AMPHIBOLES
2.2.1. Overview
Asbestos is the generic name for a group of naturally-occurring silicate minerals that
crystallize in long thin fibers. The basic chemical unit of asbestos and other silicate minerals is
[SiO/t]4 . This basic unit consists of four oxygen atoms at the apices of a regular tetrahedron
surrounding and coordinated with one silicon ion (Si4+) at the center (see Figure 2-3, Panel A).
The silicate tetrahedra can bond to one another through the oxygen atoms, leading to a variety of
crystal structures (see Figure 2-3, Panels B, C, and D).
There are two main classes of asbestos: serpentine and amphibole. The only member of
the serpentine class is chrysotile, which is the form of asbestos that was most commonly used in
the past in various man-made asbestos-containing materials (insulation, brake linings, floor tiles,
etc.). Chrysotile is a phyllosilicate (see Figure 2-3, Panel D), occurring in sheets that curl into a
fibrous form.
There are many different types of amphibole asbestos. This includes five types that were
previously used in commerce (actinolite, tremolite, amosite, crocidolite, and anthophyllite), and
these forms of asbestos are now regulated. Numerous other asbestiform amphiboles exist, even
though they were never used as commercial products and are not currently named in regulations
(Gunter et al., 2007). All forms of amphibole asbestos are inosilicates (see Figure 2-3, Panel C)
in which the long axis of the fiber (crystallographically called the c-axis) is parallel to the
direction of the chain of silicon tetrahedra.
2.2.2. Mineralogy of Amphibole Asbestos and Related Amphibole Minerals
Different types of amphiboles differ from each other primarily in the identity and
amounts of monovalent and divalent cations that bind to sites (referred to as A, B, or C sites)
along the silicate chains (see Figure 2-4). The specific cations between the two double-chain
plates define the elemental composition of the mineral, while the ratio of these cations in each
location is used to classify amphiboles within a solid-solution series. The general chemical
formula for double-chain inosilicate amphiboles is shown below:
2-3
-------
(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 [SiCb]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
V
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).
2-4
-------
• »X* •
/~
-, t -_.. -°^^ ?J
A/fcH^^H^^
X^wX*/^ ^x/
-4 Oxygen! Hydroxyl
- 7 Cotion
-4 Oxygen 1 Hydroxyl
— 4 Silicon
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).
, F, Cl)2
where:
(2-1)
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 [also known as "crocidolite"], arfvedsonite)
• Iron-magnesium-manganese-lithium amphiboles (anthophyllite,
cummingtonite-grunerite [also known as "amosite"])
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
2-5
-------
members. For example, a solid solution series for the cation Site A will have one end member
with 100% sodium ions and one end member with 100% potassium ions. This series would
include all intervening ratios.
Because each cation site has multiple possibilities, the elemental composition of the
amphibole silicates can be quite complex. It is the complexity of the amphiboles that has
historically given rise to a proliferation of mineral names with little systematic basis (Hawthorne,
1981). Currently, amphiboles are identified by a clear classification scheme based on crystal
chemistry that uses well-established names based on the basic mineralogy, with prefixes and
adjective modifiers indicating the presence of substantial substitutions that are not essential
constituents of the end members (Leake et al., 1997). As implemented, this mineral
classification system does not designate certain amphibole minerals as asbestos. However, some
mineral designations have traditionally been considered asbestos (in the asbestiform habit; e.g.,
tremolite, actinolite). Other commercial forms of asbestos were known by trade names (e.g.,
Amosite) rather than mineralogical terminology (cummingtonite-grunerite).
2.2.3. Morphology of Amphibole Minerals
Most amphibole minerals occur in a variety of growth habits, depending on the
temperature, pressure, local stress field, and solution chemistry conditions during crystallization.
The nomenclature used to describe the crystal forms varies between disciplines [field geologist,
microscopist; e.g., see Lowers and Meeker (2002)1. Text Box 2-1 provides definitions for
common terms used to describe the morphology of asbestos and other related minerals.
2-6
-------
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
length greater than or equal to 5 um and aspect ratio greater than or equal to 3:1 (by PCM) or
5:1 (by transmission electron microscopy [TEM]).
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: A single fiber which cannot be separated into smaller components without losing its
fibrous properties or appearance. 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.
2-7
-------
Asbestiform morphology is present where the conditions of formation allow crystals to
form very long individual flexible fibers which are parallel and easily separable and may become
visible to the naked eye when crushed (see Figure 2-5). Under the microscope, individual
amphibole structures may be described as asbestiform, acicular, prismatic, or fibrous. Typically,
a fiber is defined as a highly elongated crystal with parallel sides. The definitions for acicular
crystals are "needlelike" in appearance while prismatic crystals may have several parallel faces
with a low 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).
Where conditions are not conducive to the formation of individual fibers and particles,
the amphibole is described as massive—appearing as a solid contiguous sample. Mechanical
forces that break amphibole crystals along the cleavage planes create smaller pieces or cleavage
fragments. These fragments may be elongated and have a morphology that is generally similar
to amphibole asbestos, but differ from the regulated minerals in that they did not grow in an
asbestiform habit.
2-8
-------
2.3. METHODS FOR ANALYSIS OF ASBESTOS
Because asbestos is a solid that does not dissolve in water or other solvents, methods for
the analysis of asbestos are somewhat different than for most other chemical substances. This
section provides a brief overview of the most common methods for the analysis of asbestos.
2.3.1. Methods for Air Samples
The exposure pathway of primary health concern for humans is inhalation of asbestos.
Air is evaluated for the presence of asbestos by drawing a known volume of air through a filter
that traps the solid particles in the air on the filter surface, and the number of asbestos particles
are then determined. The concentration is generally expressed as fibers4 per cubic centimeter of
air (fibers/cc) and is computed by dividing the number of asbestos fibers on the filter by the
volume of air drawn through the filter.
In all cases, the evaluation of the particles that are collected on an air filter is performed
using a microscope. All methods begin with the basic shape (morphology) of a particle to
classify it as a possible asbestos particle or not. In general, particles that are clearly fibrous
(substantially longer than they are thick) are considered to be potential asbestos. However, other
minerals besides asbestos may occur in long thin particles, and a number of nonmineral fibers
may be present in a sample as well. Consequently, some techniques rely on other physical or
optical properties of the particles to help distinguish asbestos from nonasbestos and to classify
the type of asbestos. These differences in the ability to visualize and distinguish asbestos
particles are the most important differences between the various microscopic techniques.
The most common technique in the past for analyzing asbestos in air samples was PCM,
and this method remains the current industrial hygiene (IH) standard methodology, usually using
National Institute for Occupational Safety and Health (NIOSH) Method 7400
(http://www.cdc.gov/niosh/docs/2003-154/pdfs/7400.pdf). Under this method, a fiber is defined
as any particle greater than 5 um in length with an aspect ratio greater than or equal to 3:1.
Depending on the microscope set-up and refractive index of the mounting medium and fibers,
the limit of detection of PCM can be up to a fiber width of 0.25 jim. Fibers thinner than the limit
of detection are not observable. A key attribute of PCM is that identification of countable fibers
is based only on morphology, and does not consider mineralogy or crystal structure. Because of
this, it is not possible to classify asbestos fibers by mineral type, or even to reliably distinguish
between asbestos and nonasbestos fibers. This is not usually a significant concern when applied
to air samples collected in a workplace where asbestos is present, but can become an issue in
4Most techniques for analyzing air samples do not 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.
2-9
-------
nonworkplace settings where asbestos concentrations tend to be lower and other types of fibers
are more common.
Transmission electron microscopy (TEM) has also been developed for analysis of air
samples for asbestos. TEM uses a high-energy electron beam rather than a beam of light to
irradiate the sample, and this allows visualization of structures much smaller than can been seen
under light microscopy. In addition, most TEM instruments used for asbestos analysis have
equipment that allows a more detailed characterization of a particle than is possible by PCM:
• EDS (energy-dispersive spectroscopy) provides data on the atomic composition of
each particle being examined. This makes it possible to distinguish organic fibers
from mineral fibers, and also allows for distinguishing between different types of
mineral fibers.
• SAED (selected area electron diffraction) provides a diffraction pattern for crystalline
particles that is helpful in distinguishing organic from mineral fibers, and in
classifying the nature of the crystalline structure (serpentine, amphibole, pyroxene,
etc.).
• WDS (wavelength-dispersive x-ray spectroscopy) provides x-ray spectral data from a
single wavelength at a time, providing detailed atomic composition of a particle.
Generally, WDS is a more precise measure of the atomic composition of a particle
than EDS and is often used with an electron microprobe attached to a scanning
electron microscope.
Several different standard methods have been developed for TEM analyses of air
samples, the most common of which is ISO 10312 (ISO 10312:1995). ISO 10312 defines a fiber
as an elongated particle with parallel or stepped sides, an aspect ratio of 5:1, and a minimum
length of 0.5 jim, and further defines a PCM-equivalent fiber (PCMe) as having an aspect ratio
greater than or equal to 3:1, longer than 5 jim, and a diameter between 0.2 jim and 3.0 jim. The
counting rules require both types of fiber to be reported.
2.3.2. Methods for Solid Materials
Measurement of asbestos in solid samples (vermiculite, building materials, soil, etc.)
usually employs polarized light microscopy (PLM). There are several standard PLM methods
for the analysis of asbestos in bulk materials, including NIOSH 9002, EPA/600/R-93/116, and
CARB 435. PLM uses the optical properties of asbestos to identify and classify different types
of asbestos fibers. In general, these methods are most reliable for materials that contain
relatively high concentrations of asbestos, and results tend to become more variable as
concentrations decrease below about 1% by mass. At present, the use of TEM for the analysis of
bulk materials is not a well-developed procedure.
2-10
-------
2.4. CHARACTERISTICS OF LIBBY AMPHIBOLE ASBESTOS
Amphibole asbestos occurs in the Libby vermiculite ore body both in high concentration
veins (>80%), as well as in lower concentrations (0.1 to 3%) within the layers of the vermiculite
ore itself (Lowers et al., 2012; U.S. EPA, 2000a: Boettcher, 1967; Pardee and Larsen, 1928).
Analysis of historical ore samples from the Harvard and Smithsonian Museums (circa 1920s),
the Butte Museum (circa 1960), and recent ore samples from the mine (circa 1999) indicate that
the amphibole content of vermiculite ore from the mine has remained approximately constant
over the 70-year mining history at the Rainy Creek complex (Sanchez et al., 2008; Meeker et al.,
2003).
2.4.1. Mineralogy of Libby Amphibole Asbestos
Historically, the amphibole mineral fibers that occur in the Libby ore body were
described as a sodium-rich tremolite (Amandus et al., 1987b: McDonald et al., 1986a: Leake,
1978: Boettcher. 1966: Larsen. 1942). although McDonald et al. (1986a) noted the sodium
content was too high to allow classification as tremolite, and suggested that at least some fibers
might be better classified as magnesio-riebeckite or richterite.
More recently, various research groups (Gunter and Sanchez, 2009: Sanchez et al., 2008:
Meeker etal.. 2003: Wvlie and Verkouteren, 2000: Rossetal.. 1993: Moatamed et al.. 1986)
have recharacterized the mineralogical composition of amphiboles from the Libby mine using
the modern classification scheme developed by Leake etal. (1997).
The most extensive investigation was reported by the U.S. Geological Survey [USGS
(Meeker et al., 2003)1. In this investigation, USGS personnel collected amphibole samples from
different areas of the mine to identify the range of materials present. The mineral composition of
individual fiber structures was determined using EDS and electron probe microanalysis (EPMA).
The results, which are presented in Figure 2-6, show that most amphibole structures were
classified as winchite (84%), with lesser amounts classified as richterite (11%) and tremolite
(6%). Trace amounts of magnesio-riebeckite, magnesio-arfvedsonite, and edenite are also
present. Sanchez et al. (2008) found a similar distribution of amphibole mineral types in a
sample of ore collected from the mine in 2009. Wylie and Verkouteren (2000) reported the
presence of asbestiform winchite and richterite in ore samples from the mine, which was
consistent with the alteration of alkali igneous rocks.
The relationship between the cationic compositions of the three primary minerals is
illustrated in Figure 2-7. In some fibers, the composition differed within the length of a single
fiber (e.g., winchite at one end and richterite at the other end). All of these minerals are within
the solid solution series for tremolite-richterite-magnesio-riebeckite. Magnesio-riebeckite and
magnesio-arfvedsonite fall within the range of sodic amphiboles, winchite and richterite fall
within the range of sodic-calcic amphiboles, and tremolite and edenite are considered calcic
2-11
-------
amphiboles. Structural formulae and optical and crystallographic data are presented in
Table 2-1.
Edenite
Meeker et al fia. 6
4-
EDS EPMA
Mean 11MI ili-.j
;-:
x Magnesio-
arfvedsonite
/ x
X O<
Tremolite
Magnesio-
riebeckite
0.5
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)
2-12
-------
Trcmolite
CajMgsSigOjjIOH)?
Figure 2-7. Solution series linking tremolite, winchite, and richterite
amphibole fibers.
2-13
-------
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
1.600- .628
1.604- .612
1.599- .612
1.6063
1.600- .628
1.612- .668
1.613- .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
1.625- .655
1.627- .635
1.625- .637
1.6343
1.625- .655
1.635- .688
1.638- .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)h
2-14
-------
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
0.012-0.02611
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 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 and Verkouteren (2000) Amphibole asbestos from Libbv. 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 and Zussman (1997) Rock Forming Minerals Volume 2B: Double Chain Silicates, 2nd Edition The Geological Society, London.
1www.mindat.org/min-1351.html.
2-15
-------
2.4.2. Morphology of Libby Amphibole Asbestos
A number of investigators have reported on the morphology of LAA structures in
samples from the mine site, as well as in air samples from the former mine and mill or from
present-day town of Libby. McDonald et al. (1986a) used TEM to examine particles collected
on air filters from the mine and mill in Libby. The authors reported that fibers on the filters
included a range of morphologies, including straight with uniform diameter, lath- or
needle-shaped, or curved.
Brown and Gunter (2003) used PLM to examine structures obtained from three different
mineral samples collected at the mine in Libby. Each of the three samples was crushed and
sieved through a 250 um screen. Based on aspect ratio, 95% of the structures were considered
possible asbestos. Based on a more detailed evaluation of crystal structure, about one-third were
judged to be asbestos, about one-third were judged to be cleavage fragments, and about one-third
could not be classified with confidence.
Meeker et al. (2003) reported that all of the amphiboles found at the mine site, with the
possible exception of magnesio-riebeckite, can occur in fibrous habit. It was observed these
amphibole materials—even when originally present as massive material—can produce abundant,
extremely fine fibers by gentle abrasion or crushing.
Figure 2-8 shows a scanning electron microscope image of amphibole mineral collected
from the mine in Libby (Meeker et al., 2003). This image illustrates the broad range of size and
morphologies that can occur in this material. As individual structures are viewed under greater
magnification, the range of morphologies can be more clearly seen (see Figure 2-9). The USGS
has observed structures that are fibrous, acicular, and prismatic, all within the minerals from the
mine (Meeker et al., 2003).
2-16
-------
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).
2-17
-------
MBEKF.R f.l A1-: TIIEC'UMrosrilCH ()!• AMFIIIBUI.IIS IROM IU1: RAIN* CKtliK COMPLEX
Figure 2-9. Fiber morphology of amphibole asbestos from the Libby, MT
mine viewed under a scanning electron microscope.
Source: Meeker et al. (2003).
Sanchez et al. (2008) evaluated fiber morphology using an electron microscopy and
EMPA and reported that most structures could be classified as either prismatic or fibrous. There
was no difference in the mineralogy between the two morphologies, and the authors concluded
the different habits were formed at the same time.
Figure 2-10 shows cumulative particle-size-distribution frequencies (CDFs) for LAA
fibers (aspect ratio >3:1) observed using TEM in Libby ore Grade 3, expanded Libby ore
Grade 3, and ambient air samples collected in Libby. The data used to construct this plot are
described in Appendices B and C. In general, most fibers identified as LAA have widths that
range from about 0.1 urn to 1 um, with an average of about 0.6 um. Fiber lengths vary greatly,
ranging from <1 um to >100 um. Aspect ratios also range widely, from 3:1 to >100:1.
2-18
-------
Particle Size Distributions of LA Particles - Libby #3 Ore (N = 320),
Libby #3 Ore Expanded (N = 108)
Libby #3 Ore
Libby #3 Ore Expanded
Libby Air
Libby #3 Ore
Libby #3 Ore Expanded
Libby Air
1.0
0.9
0.8
0.7
0.6
Q 0.5
u
0.4
0.3
0.2
0.1
0.0
Aspect Ratio
Libby #3 Ore
Libby #3 Ore Expanded
Libby Air
10
Aspect Ratio
100
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.
2-19
-------
An important question is whether the mineralogy and morphology of LAA fibers
observed in geological samples of amphibole material collected at the mine are similar to that
observed for airborne fibers collected on filters in Libby or other locations where vermiculite
was used or processed. As shown in Figure 2-10, the size distributions for fibers observed in the
unexpanded and expanded Libby Grade 3 ores are very similar to each other, while the LAA
fibers observed in air monitoring samples from Libby tend to be slightly thinner and shorter than
in the ore samples. However, the differences are relatively minor. Mineralogical
characterization by EDS and SAED of the fibers from the Libby ore Grade 3 and the expanded
product provided additional confirmation of the similarity between the fibers from the Libby
Grade 3 ore and Libby air samples (methodology described in Section 2.3; see also Appendix B).
EDS spectra yielded an elemental fingerprint with sodium and potassium peaks that were highly
consistent with values reported for the winchite-richterite solution series described for the Libby
ores (Meeker et al.. 2003).
2.5. HUMAN EXPOSURE POTENTIAL
Several different populations have the potential for exposure to vermiculite (Zonolite)
from the mine in Libby, MT, and hence the potential for exposure to the LAA associated with
this material. This includes not only the former workers at the mine and mill site, but also
residents in the community of Libby, MT, as well as workers at other locations who processed
the vermiculite product. A brief description of these potentially exposed populations is presented
below.
2.5.1. Exposures Pathways in the Libby Community
When the mine in Libby, MT was active, miners, mill workers, and those working in the
processing plants were exposed to vermiculite, silica dust, and amphibole structures released to
air from the ore during the mining and processing operations (Meeker et al., 2003; Amandus et
al., 1987b: McDonald et al., 1986a). In some cases, workers may have inadvertently transported,
typically on their clothing, shoes, and hair, contaminated materials from the workplace to
vehicles, homes, and other establishments. This transported material may have resulted in
"take-home exposure" for the workers, their families, and other coresidents. The magnitude of
these historic take-home exposures was not measured, so the levels to which individuals in the
home might have been exposed are not known.
The Agency for Toxic Substances and Disease Registry (ATSDR) performed an exposure
survey in Libby to identify activities that may have led to the exposure of residents to vermiculite
and LAA. Based on the responses of survey participants, it was found that men were more likely
than women to have had both occupational and nonoccupational exposures, while women were
more likely to have had only household contact with exposed workers (Peipins et al., 2003;
ATSDR, 200Ib). Expanded vermiculite, as a finished product (Zonolite), was used for
2-20
-------
insulation in attics and walls in homes in Libby, and was also used as a soil amendment in home
gardens and recreational areas. Community members may have been exposed, and are possibly
still exposed, to these consumer (Zonolite) products. In a survey of Libby residents conducted
by ATSDR in 2000-2001, almost 52% reported using vermiculite for gardening, 8.8% used
vermiculite around the home, and 51% reported handling vermiculite attic insulation [VAI;
(Peipins etal., 2003)]. Because vermiculite ore, vermiculite product, and waste stoner rock (the
waste material from exfoliation) were present in the community, numerous other activities may
also have resulted in exposure. Individuals also reported exposures from the following activities:
participating in recreational activities along Rainy Creek Road, which is the road leading to the
mine (67%); playing at the ball field near the expansion plant (66%); playing in vermiculite piles
(34%); heating the vermiculite to make it expand/pop (38%); or other activities in which contact
with vermiculite occurred [31%; (Peipins et al., 2003)1.
Because a number of different activities may be associated with exposure to LAA in
Libby, it is important to recognize that the overall health hazard to an individual is related to the
sum of the exposures across all scenarios that apply to that individual.
2.5.2. Exposure Pathways in Communities with Vermiculite Expansion and Processing
Plants
While some vermiculite concentrate was exfoliated and used in Libby, MT, most of the
concentrate was transported to expansion plants at other locations across the country where it
was exfoliated and distributed. A review of company records from 1964-1990 indicates that
more than 6 million tons of vermiculite concentrate was shipped to over 200 facilities outside of
Libby (ATSDR, 2008). Figure 2-11 shows the locations of facilities that received and processed
vermiculite from the mine in Libby.
2-21
-------
Total tonnage by county
^f 3DD.ODD or more
^ 20D.ODD 10 2B9.WBO
0 1DD.ODD1D 1BB,iBB
O 1 to 6B.95S
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).
Workers in these expansion and processing facilities likely were exposed to LAA
released during processing operations. The 2008 ATSDR Summary Report (ATSDR, 2008)
on28 Libby vermiculite expansion and processing facilities stated that in some cases household
residents may also have been exposed by contact with vermiculite from the workers' clothes,
shoes, and hair. Workers' personal vehicles likely contained vermiculite dust from facility
emissions and from vermiculite that fell from their clothing and hair on the drive home after
work.
Other residents living in communities near the expansion plants may also have been
subjected to some of the same LAA exposure pathways as for the Libby community. The 2008
ATSDR Summary Report observed that individuals in a community with a vermiculite
expansion and processing plant could have been exposed to LAA by breathing airborne
2-22
-------
emissions from the facility or by inhalation exposure to contaminants brought into the home on
workers' clothing or from outdoor sources (ATSDR, 2008).
2.5.3. Libby Amphibole Asbestos 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)1, 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).
2-23
-------
3. FIBER TOXICOKINETICS
There are no published data on the toxicokinetics of Libby Amphibole asbestos (LAA).5
However, to help inform the reader as to the expected toxicokinetics of LAA, this section
contains a general summary description of the toxicokinetics of inhaled particles, with specific
discussion of dosimetry differences for fibers. A more detailed discussion of fiber dosimetry is
beyond the scope of this document and is reviewed elsewhere (NIOSH, 2011; ICRP, 1994).
LAA includes fibers with a range of mineral compositions, including amphibole fibers
primarily identified as winchite, richterite, and tremolite, along with magnesio-riebeckite,
magnesio-arfvedsonite, and edenite (see Section 2.2). Although the fiber size varies somewhat
from sample to sample for LAA, a large percentage (-45%) is less than 5 um long in bulk
samples examined from the Libby mine site (Meeker et al., 2003). Limited data from air
samples taken in the mill and screening plant at the Libby mine site also document a large
percentage of fibers (including both respirable6 fibers as well as fibers >3 um long; see
Section 4.1.1.2 and Table 4-3). Laboratory animal studies have examined the biologic response
to LAA fibers from both raw samples (Blake et al., 2007; Pfau et al., 2005) and samples of
particles respirable by rats (<2.5 um) following water elutriation [(Cyphert et al., 2012b; Cyphert
etal.. 2012a: Shannahan etal.. 2012a: Shannahan et al.. 2012c: Shannahan et al.. 2012b:
Shannahan et al., 2012d; Padilla-Carlin et al., 2011; Shannahan et al., 201 la; Shannahan et al.,
201 Ib; Shannahan et al., 2010); see Section 4.2; Appendix D]. The mean fiber dimensions in the
rat-respirable fractions are in the range of length = 4.99 um; width = 0.26 um; aspect ratio >5:1
(as measured by TEM)7. The importance of the dimensions and density of fibers to their
inhalation dosimetry—how they deposit and are subsequently cleared—is described below. Due
to a lack of toxicokinetic data specific to LAA, these dosimetry mechanisms are discussed for
inhaled fibers in general, with a specific focus on amphibole asbestos.
The main route of human exposure to mineral fibers is through inhalation. Inhaled dose
of fibers to the respiratory tract tissue depends on the fiber concentration in the breathing zone,
the physical (aerodynamic) characteristics of the fibers, the breathing mode (nose only or also
oronasal), anatomical and physiological features of the respiratory tract (e.g., airway branching
pattern and ventilation rate), and clearance mechanisms (Oberdorster et al., 2002; U.S. EPA,
1994a; Oberdorster, 1991). Ingestion is another pathway of human exposure and occurs mainly
through the swallowing of material removed from the respiratory tract via mucociliary clearance
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 reach the alveolar regions when inhaled and are defined by their aerodynamic
diameter [da < 3 urn:(NIOSH. 201111.
7 When available, detailed fiber dimension information for each study can be found in Appendix D.
3-1
-------
asbestos-contaminated work environments (Condie, 1983). Handling asbestos can result in
heavy dermal contact and exposure. Asbestos fibers can become lodged in the skin, producing a
callus or corn—but generally with no serious health effects (Lockey et al., 1984). Because few
studies have examined the deposition and clearance of fibers following ingestion or dermal
exposure to fibers, the focus of this section is on the main route of exposure: inhalation.
Studies useful for assessing the relationship between airborne fiber concentrations and
respiratory disease must involve meaningful measurements of environmental exposure and an
understanding of how to apply these measurements to the target tissue dose. Tissue dose is a
more specific measure associated with disease development than is external dose. Many studies
have examined the role of the physical and chemical characteristics of fibers in asbestos-induced
disease in the lung and are reviewed in more depth elsewhere (NIOSH, 2011; AT SDR, 200 la:
Mvoio and Takava. 2001: Witschi and Last 1996: Lippmann. 1990: Merchant 1990: Yu et al..
1986: Griffis et al.. 1983: Harris and Fraser. 1976: Harris and Timbrell. 1975). Factors
influencing dose to other tissues in the body (e.g., pleura, peritoneum, stomach, and ovaries) are
not as well known, but they are discussed below where data are available.
The principal components of inhaled fiber dosimetry in mammalian respiratory tract
systems are (1) inhalability, (2) deposition on the epithelial surface, (3) clearance from the lung
due to both physical (e.g., dissolution) and biological mechanisms (including mucociliary
transport, phagocytosis, and translocation from the lung to other tissues [including the pleura]),
and (4) elimination from the body (see Figure 3-1).
3.1. DEPOSITION OF FIBERS IN THE RESPIRATORY TRACT
The respiratory tract encompasses the extrathoracic region (nasal passages, pharynx, and
larynx), tracheobronchial region (the conducting airways [trachea, bronchi, bronchioles]), and
the gas-exchange or pulmonary region of the lung (respiratory bronchioles, alveolar ducts, and
alveoli). Each region has unique anatomic and functional features, including dramatically
different architecture, cell types, and defense mechanisms, that determine the dosimetry of
inhaled agents in each region (U.S. EPA, 1994a). A full review of the anatomy and architecture
of the respiratory tract is beyond the scope of this document but has been reviewed by the
International Commission on Radiological Protection for its reference human respiratory tract
model (ICRP. 1994). Figure 3-2 illustrates the major anatomical features of the human
respiratory tract and mechanisms of fiber deposition.
5-2
-------
Dust in inspired and expired air
sputum
nasopharyngeal compartment
tracneo-broncnial compartment
pulmonary lymph
vessels and nodes
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).
-------
bronchi olus
rtrminalhj
bronchiolu*
regpiratoriu*
acinus
(B)
random path
due to collision
with air molecules
particle ctpturtd
It c»nnj sit«
trajectory du*
to particle
inertia
respiratory bronchioles
gravitatioi
V
ton—F
\
AC
* ,
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).
Four major mechanisms determine fiber deposition: impaction, interception,
sedimentation, and diffusion. Some authors also suggest electrostatic precipitation plays a role
in fiber deposition, but no experimental data exist to verify this (Sturm and Hofmann, 2009).
The relative contribution to deposition in each region of the respiratory tract depends on fiber
dimension and density, breathing mode and ventilation rate, and the airway architecture of the
species in question (e.g., rat vs. human). The deposition mechanisms and where these
mechanisms are typically dominant in the human respiratory tract are described below (see
Table 3-1).
J-4
-------
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. 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).
5-5
-------
3. Sedimentation: Gravitational forces and air resistance cause fibers to settle out of
the convective air stream onto the airway surface. For sedimentation to occur, air
flow velocities must be low to allow the fiber to settle, so this is a predominant
mechanism in the smaller conducting airways.
4. Diffusion: This method of deposition is predominant in the alveolar region where air
movement is negligible. Diffusion occurs from interactions of the fibers with the
movement of air molecules, and becomes important for particles <0.5 um in physical
diameter.
The aerodynamic properties of particulate aerosols, including fibers, are characterized by
the particle's aerodynamic diameter and its distribution, usually as the mass median aerodynamic
diameter and geometric standard deviation (GSD) because aerosols tend to be log normally
distributed. The aerodynamic diameter is the diameter of a unit density (1 g/cm3) sphere that has
the same gravitational settling velocity as the fiber of interest and the aerodynamic diameter
derivation is based on fundamental laws governing fluid dynamics. However, characterizing a
fiber by its aerodynamic diameter is dubious due to the multiple factors that define a fiber's
aerodynamic properties [e.g., density, length, width, and its orientation with respect to the
convective airflow (Asgharian and Anjilvel, 1998; Cheng, 1986)]. Vincent (2005) has proposed
that fibers should be described by criteria that address both the aerodynamic properties that
govern their regional deposition after inhalation and the biological effects and responses
following deposition.
Computational models or algorithms to address fiber dosimetry typically derive an
aerodynamic equivalent diameter (deq) to remain consistent with the concept of aerodynamic
diameter and provide some comparative context, and are based as well on fundamental fluid
dynamics. Such formulae are mechanism specific (e.g., for impaction or sedimentation) and
describe fibers as cylinders characterized by their density and length-to-width aspect ratio (beta),
although the explicit bivariate distribution for a fiber aerosol can be described by the means and
variances of the natural logarithms for length and width with correlations for their joint
distribution (Moss et al., 1994; Cheng, 1986). The latter is unfortunately not often done due to
the lack of bivariate data when aerosols are sized in various experiments. The formulae to derive
deq must additionally account for the orientation of the fibers with respect to the convective
airflow in the Stokes flow regime where it is necessary to describe the frictional forces
encountered by an object (i.e., fiber or particle) in a fluid (i.e., air). For example, dynamic shape
factors (x) that relate the drag force of the cylindrical object to a sphere are derived for either
perpendicular or parallel orientation with respect to the flow. Likewise, adjustments in these
formulae are made for fiber orientation to the Cunningham slip correction factor which accounts
for the relative velocity (or "slip") of gas molecules in air at the surface of objects entrained in
that airflow. Additional assumptions regarding the orientation are typically used for each region
of the respiratory tract, for example, random orientation for fibers in the upper airway subject to
5-6
-------
impaction or parallel orientation in the peripheral airways. Fibers enter the respiratory tract
through the nasal and oral passages.
Deposition in the nasal and oral passages is mainly by impaction. The nasal passage,
from the nostril to the pharynx, serves as a filter for some fibers with diameters 5-30 um.
Clumps of fibers could deposit in these regions. Many animal species, including rats and mice,
are obligate nose breathers so fibers pass only through the nasal passages and are always subject
to nasopharyngeal filtering. Humans, monkeys, and dogs breathe both orally and nasally
(oronasal). During oral respiration, larger fibers and clumps of fibers can bypass the filtering of
the upper respiratory tract and be inhaled directly into the larynx/trachea, especially during
exertion (e.g., exercise or work), thereby altering deposition. This distinction is important when
comparing results of inhalation studies conducted in different species.
Fibers in the lower respiratory tract deposit by combined mechanisms of impaction,
interception, sedimentation, and diffusion. The relative contribution of each mechanism depends
on the fiber characteristics and region-specific airway anatomy, and respiratory flow rates (air
velocities). Interception is heavily influenced by fiber length. Where the physical length of the
fiber greatly exceeds the aerodynamic diameter, interception can be underpredicted by modeling
the center of gravity of the fiber because the length of the fiber will determine its propensity to
intersect with the airway. Sedimentation is related to the mass of the fiber, as well as the
aerodynamic diameter, but generally occurs at lower velocities in smaller airways. Diffusion
occurs from interactions of the fibers with the movement of air molecules; this Brownian motion
increases with decreasing fiber size (<0.5 um diameter).
The conducting airways beyond the nasopharyngeal region serially bifurcate into airways
of decreasing internal diameters that restrict the size of fibers deposited in these regions. Fiber
length enhances bronchial deposition via interception, especially fibers exceeding lengths of
10 um (Sussman et al., 1991a, b). The aerodynamic diameter of fibers that are more likely to
deposit in the tracheobronchial region is in the range of 1-5 um. Fibers with an aerodynamic
diameter of <1 um are more likely to deposit in the bronchioles and the alveoli (ICRP, 1994). As
reviewed in Austet al. (2011), some studies have demonstrated that short fibers (<5 um) are
present in substantially greater numbers than long fibers (>5 um) when examining the whole
lung (Churg, 1982). Whether this is due directly to deposition of smaller fibers or the breakage
of larger fibers over time is not known (Bernstein et al., 1994; Davis, 1994). Although
information is limited on how fibers get to the pleura, fibers observed in pleural tissue from
mesothelioma cases are more likely to be short [<5 um; (Suzuki etal., 2005)1.
Fibers with aerodynamic characteristics conducive to depositing in the respiratory
bronchioles and alveoli may cause pulmonary fibrosis and other associated diseases. Regardless
of shape, mineralogy, or concentration, the majority of fibers that are small enough to reach the
alveoli are deposited at the alveolar duct bifurcations (Brody and Roe, 1983). Deposition is
controlled by air flow characteristics and is greatest at the bifurcations that are closest to the
3-7
-------
terminal bronchioles (Brody et al., 1981). Furthermore, deposition in the bifurcations is
consistent across laboratory animal species (Warheit and Hartsky, 1990). Alveolar deposition is
limited when fiber length approaches 40 um (Morgan et al., 1978). However, alveolar
deposition of fibers can occur with high aspect ratios and lengths ranging from <1 um to
>200 um long (Morgan et al., 1978). All fibers having an aerodynamic diameter less than
approximately 2 um, which includes LAA, meet the physical criteria necessary for deposition in
the deeper regions of the respiratory tract at the level of the terminal bronchioles or alveoli.
3.2. CLEARANCE MECHANISMS
Once fibers deposit on the surface of the respiratory tract, they may be removed (cleared)
in several ways—including physical clearance, dissolution, phagocytosis, encapsulation, or
transcytosis. Some of these mechanisms, such as dissolution of the fibers or removal via the
mucociliary apparatus, can result in the fibers being cleared from the body (see Figure 3-1).
Other clearance mechanisms may remove fibers from the surface of the respiratory tract but
result in transport of the fibers to different locations or tissues by translocation. Translocation of
fibers from the terminal bronchioles and alveoli into the peribronchiolar space, lymph nodes, and
pleura has been implicated in disease causation (e.g., pleural plaques, mesothelioma) (Dodson et
al., 2001). In human studies, the translocation of asbestos fibers following inhalation has been
observed to varying degrees throughout the pulmonary and extrapulmonary tissues of the
respiratory system (Dodson et al., 2005; Dodson et al., 2001; Kohvama and Suzuki, 1991;
Suzuki and Kohyama, 1991; Armstrong et al., 1988), as well as to other organs, including the
brain, kidney, liver (Miserocchi et al., 2008), and ovaries (Langseth et al., 2007). In many cases,
the type of fiber is not defined, and the individual exposure information not available. Fibers
that are not cleared can remain at the epithelial surface or enter the parenchymal tissue of the
lung. Retention of fibers in the human thoracic region generally shows two distinct phases. The
first, on the order of 24 hours, is considered to represent mucociliary clearance to the
gastrointestinal tract from the conducting airways and bronchioles; the second represents
clearance from the alveolar region (ICRP, 1994).
Berry (1999) provided a review of the animal toxicity literature specifically for fiber
clearance. There are limited data on clearance patterns based on autopsy studies in humans.
Two studies estimated clearance half-life for amphibole asbestos (-20 years) as compared with
chrysotile asbestos [-10 years (Finkelstein and Dufresne, 1999; Churg and Vedal, 1994)1: in
evaluating the data on lung fiber burden, Berry et al. (2009) estimated the range of the half-life
for crocidolite to be between 5 and 10 years. Generally, studies have focused on determining the
size and type of asbestos retained in specific tissues (Suzuki et al., 2005; McDonald et al., 2001;
Suzuki and Yuen, 2001: Dumortier et al., 1998: Gibbsetal., 1991: Dodson etal., 1990) and do
not discuss changes in fiber content since exposure due to possible clearance. Sebastien et al.
-------
(1980) concluded that lung fiber burden could not be used as an accurate reflection of pleural
fiber burden.
3.2.1. Physical and Physicochemical Clearance of Fibers
Mechanisms of physical and physicochemical clearance of fibers depend on the fiber
size, physicochemical characteristics, and site of deposition (IOM, 2006). Physical clearance
includes mechanical mechanisms, including transport via the mucociliary apparatus, macrophage
uptake, and translocation. Fibers can also translocate due to physical forces associated with the
mechanics of respiration [e.g., expansion, contraction of the rib cage (Davis, 1989)].
Physicochemical clearance of fibers includes dissolution and breakage of fibers.
3.2.1.1. Mechanical Reflex Mechanisms
Fibers deposited in the nasal passages can be removed by all clearance mechanisms.
When breathing occurs through the nose, many fibers are filtered by the turbulent airflow in the
nasal passages, impacting against the hairs and nasal turbinates, as well as becoming entrained in
mucus in the upper respiratory tract where they can be subsequently removed by mucociliary
action (described below) or reflexive mechanical actions such as coughing or sneezing.
Dissolution can also occur in this region, especially for soluble fibers.
3.2.1.2. Mucociliary Clearance
Physiological mechanisms include mucociliary escalator movement and how specific
cells or mechanisms in various regions of the respiratory tract respond and attempt to detoxify or
remove deposited fibers.
The mucociliary escalator removes fibers through ciliary movement of the sticky mucus
lining much of the respiratory tract (Wanner et al., 1996; Churg et al., 1989). Fibers removed
from the conducting airways through this mechanism are coughed out or swallowed and enter the
digestive tract where they may adversely affect the gastrointestinal tissue, enter the blood stream,
or be excreted. Clearance of fibers via mucociliary action is usually complete within hours or
days (Albert et al., 1969).
The mucociliary escalator extends only down to the level of the terminal bronchioles and
not to the alveoli. Thus, the particles deposited in the alveolar region of the lung cannot be
cleared through this process. Particles can reach the mucociliary escalator from the alveoli either
by way of surface fluids that are drawn onto the mucociliary escalator by surface tension or by
travelling through lymphatic channels that empty onto the escalator at bronchial bifurcations.
Although ingestion is a potential route of exposure due to swallowing of material from
the mucociliary escalator, limited research has examined clearance (e.g., translocation) of fibers
following ingestion, and no clearance studies are available specific to LAA. An early study to
examine gastrointestinal tissue response to asbestos fibers is not truly representative of a natural
3-9
-------
ingestion exposure, as the researchers directly injected a suspension of amosite fibers into the
duodenal wall (Meek and Grasso, 1983). This study, however, also examined oral ingestion of
amosite in healthy animals and those with gastrointestinal ulcers to determine whether
translocation of fibers occurs through ulcers. Following injection of amosite, granulomatous
lesions were observed. Ingestion of the same material resulted in no such lesions or in any other
histopathological changes in either healthy or rats compromised with ulcers. Thus, no
translocation was observed from either the healthy or the compromised rat gastrointestinal tracts
in this study. Truhaut and Chouroulinkov (1989) examined the effects of chrysotile and
crocidolite ingestion in Wistar rats. No translocation was observed. No further studies have
been found on clearance or translocation of fibers from the gastrointestinal tract.
Some fibers are not cleared via the mucociliary escalator from the respiratory tract,
leading to an accumulation with time (Case et al., 2000; Finkel stein and Dufresne, 1999; Jones et
al., 1988). The fibers that remain in the conducting airways and alveolar regions may undergo a
number of processes including translocation, dissolution, fragmentation, splitting along the
longitudinal axis, or encapsulation with protein and iron. Available data indicate prolonged
clearance from the thoracic region of long (>5 um) or short amphibole fibers (Coin et al., 1994;
Tossavainen et al., 1994).
The prolonged clearance times for long amphibole fibers have led some investigators to
conclude that long fibers (>5 um) rather than short amphibole fibers (i.e., LAA) are predominant
in the cause of disease due to their persistence in the lung (Mossman et al., 2011; AT SDR,
2003). However, others argue that fibers of all lengths induce pathological responses and urge
caution in excluding, based on length, any population of fibers from consideration as possibly
contributing to the disease process (Aust et al., 2011; Dodson et al., 2003). Respirable-sized
fibers of LAA have been identified in air samples from Libby, MT and in airborne fibers
suspended from both Libby vermiculite ore and in the exfoliated product from that ore (length
range from 1 um to 20-30 um, with average length of 7 um; width range from 0.1-2 um, with
average of 0.5 um; see for details Appendix B and Appendix C). Based on fibers counted by the
TEM analytical method (ISO 10312), the majority of counted fibers are respirable (see
Figure 2-10).
3.2.1.3. Phagocytosis by Alveolar Macrophages
The principal clearance pathway for short, insoluble fibers deposited in the alveoli is
through phagocytosis by macrophages. Impaction of durable fibers in the deeper region of the
respiratory tract stimulates activation of alveolar macrophages. In vitro and in vivo studies
clearly indicate that macrophage cells play a role in the translocation of fibers (Dodson et al.,
2000a: Castranova et al.. 1996: Brodv et al.. 1981: Bignon et al.. 1979). These studies
demonstrated the presence of asbestos fibers in cell cytoplasm where the fibers can be
transported in association with cytoskeletal elements to the proximity of the cell nucleus.
3-10
-------
Alveolar macrophages that have phagocytized insoluble fibers migrate to the bronchoalveolar
junctions along epithelial surfaces to ciliate bronchioles where they are removed (Green, 1973).
Alternatively, alveolar macrophages that have phagocytized insoluble fibers can also migrate
through the epithelial wall into the interstitial space and enter the lymphatic system (Green,
1973).
A number of processes can disrupt the normal phagocytic function of alveolar
macrophages, such as the overwhelming of phagocytosis and the mucociliary escalator by an
excessive number of particles (often termed "overload"), or the attempted phagocytosis of fibers
with lengths that exceed the dimensional capacity of the macrophage [> 15-20 um depending on
species; often termed "frustrated phagocytosis"; (NIOSH, 2011)]. Any of these processes can
induce inflammatory and fibrogenic responses. Limited inhalational laboratory animal studies
exist at concentrations of fibers insufficient to induce overload; therefore information is
insufficient to determine mechanisms of inflammation at lower doses (Mossman et al., 2011).
Fibers that are too long to be easily engulfed by the alveolar macrophage can stimulate
the formation of "asbestos bodies." Asbestos bodies are fibers that become coated with proteins,
iron, and calcium oxalate as a result of prolonged residence in the lung where they can remain
throughout an individual's lifetime. Due to their iron content, histological stains for iron have
long been used to identify them in tissue; thus, they are sometimes called "ferruginous bodies."
The mechanisms that result in the formation of asbestos bodies are poorly understood, although
most appear to be formed around amosite fibers (Dodson et al., 1996). The iron in the coating is
derived from the asbestos fiber, cells, or medium surrounding the fiber and can remain highly
reactive (Lund et al., 1994; Ohio etal., 1992). Asbestos bodies comprise a minor portion of the
overall fiber burden of the lung. Once fully coated, fibers within asbestos bodies may or may not
participate directly in asbestos disease. The presence of iron in the coating could provide a
source for catalysis of reactive oxygen species (ROS).
3.2.1.4. Epithelial Transcytosis
In addition to phagocytosis by alveolar macrophages, fibers deposited on type I alveolar
epithelial cells may also be subjected to transcytosis with subsequent sequestration to the
alveolar interstitium (Sturm, 2011). Fiber length may play a key role in this aspect of clearance.
3.2.1.5. Translocation
Translocation represents the movement of intact fibers along the alveolar epithelial
surface towards the terminal bronchiole, or into and through the epithelium. Translocation
typically occurs via drainage of the alveolar macrophages to the lymphatics, but transcytosis of
fibers by type I alveolar epithelial cells can also result in transport to the interstitium. The
relative contribution of a specific mechanism and translocation route depends both on fiber
characteristics and the tissue upon which the fibers deposit. Fiber translocation depends on the
3-11
-------
physicochemical characteristics of the deposited fibers, including two-dimensional size (length
and width), durability, solubility, and reactivity. This translocation is aided by high durability
and an inflammation-induced increase in permeability, but could be hindered by fibrosis.
Translocation of fibers to extrapulmonary tissues has been reported in multiple studies;
however, the precise mechanism is still unknown. This process was recently reviewed by
Miserocchi et al. (2008). Fibers have been identified in all of the analyzed locations, including
pleural plaques and mesothelial tissue (i.e., pleural or peritoneal) in miners, brake workers,
insulation workers, and shipyard workers (Roggli et al., 2002; Dodson et al., 2000b; Churg,
1994; Kohyama and Suzuki, 1991). However, amphibole fibers were less prevalent than
chrysotile fibers in the pleura and mesothelial tissues (Kohyama and Suzuki, 1991; Sebastien et
al.. 1989: Armstrong et al.. 1988: Churg. 1988: Bignonetal.. 1979). Confocal microscopic
observations of rats inhalationally exposed to amosite fibers showed that fibers were present on
the parietal pleural surface 7 days postexposure and more than twofold thickening of the pleura
was noted (Bernstein et al., 2011). Bignon etal. (1979) also reported increased amphibole fibers
in the thoracic lymph nodes. Conflicting results from another inhalational rat study do not
indicate any evidence of fiber translocation from the central to peripheral compartments,
although this could be due to the short duration of the study [29 days postexposure; (Coin et al.,
1992)1.
Few studies have examined the size distribution of fibers translocated to specific tissues.
For example, one early study suggested that longer amphibole fibers predominate in the lung
(Sebastien et al., 1980); other studies showed that the fiber-length distribution was the same by
fiber type regardless of location (Kohyama and Suzuki, 1991; Bignonetal., 1979). Dodson et al.
(1990) observed that the average length of fibers found in the lung (regardless of type) was
longer than that of fibers found in the lymph nodes or plaques. Most fibers at all three sites were
short (<5 um). Similar results were observed in a later study by this group [i.e., Dodson et al.
(2000b)] which examined tissue from 20 individuals with mesotheliomas, most with known
nonoccupational asbestos exposures.
Transplacental transfer of both asbestos (tremolite, actinolite, and anthophyllite) and
nonasbestos fibers occurs in humans, as measured in the placenta and in the lungs of stillborn
infants (Hague et al.. 1998: Hague et al.. 1996: Hague et al.. 1992: Hague and Kanz, 1988). It is
hypothesized that maternal health might influence the translocation of fibers, as some of the
mothers had preexisting health conditions [e.g., hypertension, diabetes, or asthma; (Hague et al.,
1992)1. This group also measured transplacental translocation in a mouse study and observed
early translocation of crocidolite fibers through the placenta in animals exposed via tail-vein
injection (Hague etal., 1998). These transplacental migration studies did not evaluate the source
or levels of exposure, only the presence of fibers in the body during early life stages in mice and
humans.
5-12
-------
3.2.1.6. Dissolution and Fiber Breakage
Dissolution, or the chemical breakdown of fibers, is another method of physical removal
of fibers from the respiratory tract. This process varies, depending on the solubility and
chemical composition of the fibers, as well as the physiological environment. Dissolution can
occur in the extracellular lung fluids or in the macrophage phagolysosome; the former can make
the breakdown products available for uptake into the blood. Studies performed in vitro to
determine dissolution rate of fibers attempt to mimic the extracellular lung fluids and
macrophage-phagolysosome system to understand the length of time that fibers remain in the
system (Rendall and Du Toit, 1994). Fibers can also be physically degraded through splitting or
breakage. These smaller fragments are then more easily removed by phagocytosis or
translocation.
3.3. DETERMINANTS OF TOXICITY
Multiple determinants of fiber toxicity, including dimension (length, width, aspect ratio,
and surface area), chemical characteristics (solubility, charge, and surface chemistry) and
durability (dissolution, breakage) have been studied relative to specific biological responses to
fibers and recently reviewed (Aust etal., 2011; Broaddus etal., 2011; Bunderson-Schelvan et al.,
2011; Case etal.. 2011; Huang et al.. 2011; Mossman et al.. 2011).
3.3.1. Dosimetry and Biopersistence
The dosimetry factors discussed in the previous sections are major factors influencing
toxicity, as the initial deposition sites in the respiratory tract determine the subsequent clearance
mechanisms (Brain and Mensah, 1983); solubility and composition also influence the
biopersistence of deposited fibers (Maxim et al., 2006; Bernstein et al., 2005a). Thus, fiber
toxicity has been associated with dose, density, dimensions, and durability, and likely involves a
combination of these and other factors. To the extent that a fiber is resistant to the clearance
mechanisms described in Section 3.1., biopersistence becomes a determinant of toxic response.
Fiber durability is a determinant of retained dose at the site of deposition and likely plays a role
in chronic inflammation, fibrosis, and lung burden following chronic exposure to fibers.
Biopersistence is influenced by fiber characteristics such as size (length, width) and chemistry.
Hesterberg et al. (1998a) and Hesterberg et al. (1998b) observed that, in general, increased in
vitro solubility decreases in vivo biopersistence. Several supporting studies reported increased
levels of crocidolite, tremolite, and amosite in respiratory diseases (asbestosis, mesothelioma)
compared to chrysotile and controls (Churg and Vedal, 1994; Churg et al., 1993) and found that
chrysotile has a lower biopersistence than amphibole fibers (Churg and Vedal, 1994; Wagner et
al., 1974). The role of fiber size in biopersistence was examined by Bernstein et al. (2004) who
found that the clearance half-time of longer fibers (>20 um; 1.3 days) was less than that of
shorter fibers (<5 um; 23 days) for one form of chrysotile. Biopersistence is the basis of
O 1O
3-13
-------
short-term in vitro testing required for new fibers introduced to the market (Maxim et al., 2006;
Bernstein et al.. 2005a).
Fiber burden analysis of human tissues is frequently employed to determine the presence
of specific fiber type and size in disease (Aust etal., 2011; Case et al., 2011). The majority of
these studies have focused on lung tissue, but studies have also examined fiber burden in other
tissues of interest, including lymph nodes and pleural tissues (Bunderson-Schelvan et al., 2011;
Dodson and Atkinson. 2006: Dodson etal.. 2001: Dodson et al.. 2000b: Boutin etal.. 1996).
While informative, analysis of tissue fiber burden has some limitations, including differences in
methodologies that hinder comparisons between laboratories, as well as potential cross
contamination with other tissues. A further limitation of fiber burden analysis is that it is
generally performed on tissue digests, making it difficult to show fiber dimensions at specific
tissue locations. The use of TEM analysis can determine length and width of fibers found in
tissues from exposed individuals.
3.3.2. Biological Response Mechanisms
Although numerous studies have examined specific mechanisms of toxicity for many
different fiber types, the results often are contradictory or do not account for dosimetry, and thus,
only limited conclusions can be made for fibers in general. Research has focused mainly on the
role of length, width, and durability in fiber toxicity. The relative contribution of fiber
dimensions and composition that drive the toxicity of fibers remains poorly understood due to
the difficulty in experimentally evaluating each determinant independently. Further, as can be
appreciated from an evaluation of Table 3-2, the determinants of toxicity induce various toxic
responses that represent interrelated biological activities (e.g., chronic inflammation, oxidative
stress, and genotoxicity), making a clear causal relationship or relative contribution of any
individual endpoint difficult to determine. The information described below focuses on in vivo
studies that have examined determinants of amphibole fiber toxicity for some major specific
biological responses. A more detailed discussion of potential tissue response mechanisms can be
found in the mode of action (MOA) section (see Section 4). Table 3-2 summarizes the major
determinants of toxicity as fiber-host interactions along the continuum of
source-exposure-dose-response used for risk assessment.
5-14
-------
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).
5-15
-------
Limited studies have examined the role of specific fiber determinants in autoimmune
disease or pulmonary function, and therefore these endpoints are not discussed in this section.
3.3.2.1. Inflammation and Reactive Oxygen Species (ROS) Production
Inflammation is an important biologic response to fibers and is related to multiple
pathways following exposure. Fiber exposure leads to ROS production, which in turn has been
shown to increase the activation of inflammatory and immune signaling pathways (Mossman et
al., 2011). Inflammation often occurs at the site of fiber deposition; therefore those fiber
characteristics that play a role in fiber deposition (e.g., length and width) will also play a role in
chronic inflammation. Further, those additional characteristics that lead to ROS production (e.g.,
surface chemistry) may also contribute to induction of chronic inflammation. Acute
inflammation in response to asbestos further contributes to chronic inflammation with the
activation of signaling pathways (e.g., mitogen-activated protein kinase [MAPK]) that lead to the
release of proinflammatory cytokines.
Increased ROS production is hypothesized to result from frustrated phagocytosis and
activation of signaling pathways in various cell types or through iron catalysis, which may
account for the differential induction of ROS due to variable intrinsic or acquired iron by
different fibers (Aust et al., 2011). Either indirect or direct ROS release following exposure to
fibers may in turn lead to increased damage to DNA or other biological molecules.
3.3.2.2. Genotoxicity
Genotoxicity (including mutagenicity) from exposure to fibers likely involves multiple
pathways, and the role of specific fiber determinants is not completely understood. This
genotoxicity is generally described as direct (e.g., fiber interference with spindle formation) or
indirect (e.g., ROS production). A recent review by Huang etal. (2011) examines the role of
fiber determinants in genotoxicity and mutagenicity in detail. Briefly, research studies designed
to examine the role of fiber dimensions or surface characteristics in genotoxicity are limited, and
are mainly in vitro work. In general, fiber dimensions are expected to play a role in genotoxicity.
Long thin fibers are associated with interference with the spindle apparatus during mitosis, as
well as increased ROS/reactive nitrogen species (RNS) production through frustrated
phagocytosis, which in turn may lead to increased genotoxicity. Similarly, increased iron
associated with fibers may also lead to increased ROS/RNS production and increased
genotoxicity.
3.3.2.3. Carcinogenicity
The work by Stanton et al. (1981) examined fiber type and dimension in relation to
carcinogenicity in an animal model resulting in the "Stanton Hypothesis" that identifies fiber size
as a major determinant of toxicity. This study focused on amphibole asbestos due to difficulties
3-16
-------
in measurements of chrysotile (Stanton et al., 1981). However, this hypothesis was formulated
from the results of studies where fibers were imbedded in agar and implanted against the pleura,
thereby inducing sarcomas in rats (Stanton etal., 1981; Stanton and Wrench, 1972). The results
of these studies led Stanton and colleagues to state that "carcinogen!city of fibers depended on
dimension and durability rather than on physicochemical properties" (Stanton et al., 1981).
However, the study design did not allow for the influence of pulmonary clearance mechanism, as
the implant was made to the outer pleural tissue. Additionally, it is unknown how the dissolution
of fibers in the agar influenced findings. All fibers tested (including mineral wool and fiber
glass) induced sarcomas. While these studies showed high correlation between disease and
longer (>8 um), thinner (<0.25 um) fibers, high correlations were noted in other size categories
as well. The ability of these studies to define the dimensional aspects of fibers (length and width
cutoffs) that determine toxicity is clearly limited because major aspects of toxicokinetics and
biological activity in the lung tissue are not included in the experimental design. Additionally,
these studies do not rule out the potential role of shorter (<4 um) and wider (>1.5 um) fibers
(Stanton et al., 1981). This latter point was further confirmed by Pott et al. (1987) and Pott et al.
(1974) who showed that shorter fibers (<10 um in length) could also induce tumors in rats
following intraperitoneal injection. Although informative, both of these study designs bypass
normal physiological deposition and clearance mechanisms that would be observed following
inhalation of fibers, an important consideration when comparing these types of studies.
Suzuki et al. (2005) also examined the role of fiber dimensions in mesothelioma, but
through fiber burden analysis of human mesothelioma tissue. Fibers were identified by
high-resolution analytical electron microscopy from digested or ashed lung and mesothelial
tissues samples taken from 168 cases of malignant mesothelioma. Their results were that the
majority of fibers (89%) were shorter than or equal to 5 um in length, and generally (92.7%)
smaller than or equal to 0.25 um in width, which on initial consideration might seem contrary to
the "Stanton Hypothesis." However, this study is also not without interpretation challenges, as,
longer fibers that had been translocated to the mesothelial tissue may have broken down by
dissolution or fiber breakage during life, or the digestion and ashing process may itself have
degraded longer fibers to shorter fibers.
Analyses of fiber dimensions in exposed humans have not led to any clear determinants
of toxicity for fibers in general. Lippmann (1990) correlated fiber length with disease status in
exposed humans and concluded that asbestosis was associated with shorter (>2 um), thicker
(>0.15 um) fibers; mesothelioma with longer (>5 um), thinner (<0.1 um) fibers; and lung cancer
with longer (>10 um), thicker (>0.15 um) fibers. Throughout the years, some laboratory animal
studies have demonstrated a role for longer (>20 um) and thinner (<0.3 um) fibers in lung cancer
(Berman et al., 1995) or mesothelioma (Miller et al., 1999), while yet other laboratory animal
studies have suggested a role for shorter structures (<0.5-5 um) in disease based in part on
increased numbers in dust clouds and in lung retention (Dodson et al., 2003). Some human
3-17
-------
epidemiology studies have supported the role of longer (>20 um), thinner (<0.3 um) fibers in
lung cancer (Loomis et al., 2010; Berman and Crump, 2008; Dement et al., 2003). However,
these results have not been confirmed in other studies and are in some cases contradicted. For
instance, Churg and Vedal (1994) did not find an association between fiber length and cancer.
However, McDonald et al. (2001) observed that shorter fibers (<6 um) were more abundant in
diseased tissues than longer fibers (>10 um), and Dodsonet al. (1997) concluded that fibers of
all sizes are associated with increased mesothelioma risk. More recently, Berman (2011)
performed a quantitative analysis of previous studies and demonstrated that differences in
biological potency among various amphibole fiber types may be due to differences in their
dimensions, particularly fiber length. In all cases, the analytical methods used need to be
carefully described in order to draw any conclusions across studies.
3.4. FIBER DOSIMETRY MODELS
Modeling of fiber deposition has been examined for various fiber types [e.g., refractory
ceramic fibers, chrysotile asbestos (Sturm, 2009; Zhou et al., 2007; Lentz et al., 2003; Dai and
Yu, 1998; Yu et al., 1997; Coin et al., 1992)1, but not for LAA. In general, the pattern of
deposition for fibers is expected to have some similarities to the well-studied deposition pattern
for particles that are essentially spherical [reviewed in (ICRP, 1994)]. For example, the
multipath particle dosimetry model (Brown et al., 2005; Jarabek et al., 2005) uses information on
the physical properties of the particles (length, width [also called bivariate distribution] and
density), the anatomy and architectural features of the airways, airflow patterns that influence the
amount and the location of the deposition of the particles, and the dissolution and clearance
mechanisms that are operative to estimate the retained dose in the target tissue. The site of fiber
deposition within the respiratory tract has implications related to lung retention and surface dose
of fibers. It should be noted that differences in airway structure and breathing patterns across life
stages (i.e., children, adults) change the depositional pattern of differently sized fibers, possibly
altering the site of action and causing differential clearance and health effects (see Section 4.7).
3.5. SUMMARY
Although oral and dermal exposure to fibers does occur, inhalation is considered the main
route of human exposure to mineral fibers, and therefore, it has been the focus of more fiber
toxicokinetic analyses in the literature. Similar to other forms of asbestos, exposure to LAA is
presumed to be through all three routes of exposure; however, this assessment specifically
focuses on the inhalation pathway of exposure. Generally, fiber deposition in the respiratory
tract is fairly well defined based on fiber dimensions and density, although the same cannot be
said for fiber translocation to extrapulmonary sites (e.g., pleura). The deposition location within
the pulmonary and extrapulmonary tissues plays a role in the clearance of the fibers from the
organism.
3-18
-------
Fiber clearance from the respiratory tract can occur through physical and biological
mechanisms. Limited mechanistic information is available on fiber clearance mechanisms in
general, and no information specific to clearance of LAA fibers is available. Fibers have been
observed in various pulmonary and extrapulmonary tissues following exposure, suggesting
translocation occurs to a variety of tissues. Studies have also demonstrated fibers may be cleared
through physical mechanisms (coughing, sneezing) or through dissolution of fibers.
Multiple fiber characteristics (e.g., dimensions, density, and durability) play a role in the
toxicokinetics and toxicity of fibers. There is extensive literature examining a variety of fiber
determinants and their role in disease, with a focus on fiber length, width, and durability;
however, these studies are often contradictory, making conclusions difficult for fibers in general.
This is in part due to the variety of fibers analyzed, inadequate study design, and/or lack of
information on fiber dimensions in earlier studies. However, due to the importance in
understanding the role of these fiber determinants in the biological response, careful attention has
been paid to these fiber characteristics when analyzing research studies on LAA and asbestiform
tremolite, an amphibole fiber that comprises part of LAA (see Appendix D). No toxicokinetic
data are currently available specific to LAA, winchite, richterite, or tremolite. When available,
fiber characteristic data are presented in the discussion of each study in relation to the toxic
endpoints described.
5-19
-------
4. HAZARD IDENTIFICATION OF LIBBY AMPHIBOLE ASBESTOS
This section discusses the available data derived from studies of people exposed to Libby
Amphibole asbestos (LAA),8 either at work or in the community, and from various laboratory
studies. The effects in humans (e.g., parenchymal damage and pleural thickening, lung cancer,
and mesothelioma) are supported by the available LAA experimental animal in vivo and
laboratory in vitro studies. The health effects from asbestiform tremolite exposure, one of the
constituent minerals of LAA, reported in both human communities and laboratory animals, are
consistent with the human health effects reported for LAA. Studies examining the health effects
of exposure to winchite or richterite alone were not available in the published literature. The
review presents noncancer and cancer health effects observed from exposures to LAA.
4.1. STUDIES IN HUMANS—EPIDEMIOLOGY
The LAA epidemiologic database includes studies based in occupational settings and
community-based studies of workers, family members of workers, and others in the general
population. Occupational epidemiology studies have been conducted at two worksites where
workers were exposed to LAA: the vermiculite mine and mill at the Zonolite Mountain
operations near Libby, MT, and a manufacturing plant using the vermiculite ore in Marysville,
OH. Community-based studies have also been conducted among residents around Libby, MT,
(ATSDR, 200Ib, 2000) and in an area around a manufacturing plant producing vermiculite
insulation in Minneapolis, MN (Alexander et al., 2012).
The epidemiology studies of people exposed to LAA were primarily identified through
EPA's specific knowledge of the research endeavors that have taken place since recognition in
the 1970s of the asbestos contamination from the vermiculite mined around Libby, MT. These
studies were conducted by NIOSH, McGill University, University of Cincinnati, and the
ATSDR. Analyses by other researchers using the data collected through these studies as well as
other studies of people exposed to LAA were also identified through contacts with these research
groups and through "forward searching" through Web of Science for references citing the key
publications describing the initial studies [i.e., Amandus et al. (1987a), Amandus et al. (1987b),
McDonald et al. (1986a). Lockevetal. (1984). and Peipinsetal. (2003)1.
Figure 4-1 depicts the sets of studies conducted by different groups of researchers in
Libby, MT and in two areas with plants that used Libby vermiculite in various production
processes (fertilizer and other lawn products in Marysville, OH and insulation materials in
Minneapolis, MN). These studies have examined cancer and noncancer mortality, pulmonary
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.
4-1
-------
effects detected through x-ray examinations, pulmonary function tests or respiratory symptoms,
autoimmune diseases, and prevalence of autoantibodies.
Libby, MT
Marysville, OH
Occupation-based
(vermiculite mining and milling workers;
WR Grace)
NIOSH
Exposure
Assessment: Amandus et al., 1987b
Mortality: Amandus and Wheeler, 1987
Sullivan, 2007 (follow-up)
Moolgavkar et al., 2010 (reanalysis)
Herman and Crump, 2008 (reanalysis)
Pulmonary: Amandus et al., 1987a
\X-fays)
'McQill University
Exposure
Assessment: McDonald et al., 1985a
Mortality: McDonald et al., 1986a
McDonald et al., 2004, 2002 (follow-
up)
Pulmonary: McDonald et al., 1986b
Community-based
(Libby, MT and surrounding areas)
ATSDR Community Health Screening
Mortality: ATSDR, 2000
Pulmonary: Peipins et al., 2003
(X-rays, ATSDR, 2001b
PFT, Weill et ai., 2011
Symptoms) Larson on et al., 2012b
Vinikoor et al., 2010
Autoimmune: Noonan et al., 2005
Occupation-based
(fertilizer and other lawn products
production; OM Scott)
Exposure
Assessment: Borton et al., 2012
Mortality: Dunning et al., 2012
Pulmonary: Lockey et al., 1934
(X-rays) Rohset al., 2008
Clinic-based
Pulmonary: Winters et al., 2012
(X-rays, PFT, symptoms)
Minneapolis, MN
ATSDR
Mortality: Larson et al., 2010b
Pulmonary: Larson et al., 2012a, 2010a
(X-rays, PFT)
Other
Autoantibodies: Marchand et al., 2012
Pfau et al., 2005
Community-based
(area around Western Minerals
insu.aticTi nate-jg.5 p.art
Other (Clinic-based)
Mesothelioma: Wriitehouse, 2008
Pulmonary: Whitehouse, 2004
(X-rays, PFT)
Minnesota Dept of Health
and ATSDR
Methods: Adgate et al., 2011
Kelly et al.,2006
Pulmonary: Alexander et 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.
The various populations and study designs are summarized in Section 4.1.1, and the
results of these studies are presented in subsequent sections: respiratory effects other than cancer
(see Section 4.1.2), other noncancer effects (see Section 4.1.3), and cancer (see Section 4.1.4). A
brief summary of the epidemiology studies of environmental or residential exposure to
asbestiform tremolite or asbestiform tremolite-chrysotile mixtures and to crocidolite is presented
in Section 4.1.5.
4-2
-------
4.1.1. Overview of Primary Studies
4.1.1.1. Studies ofLibby, MT Vermiculite Mining and Milling Operations Workers
4.1.1.1.1. Description ofvermiculite mining and milling operations in Libby, MT. The
vermiculite mining and milling operations have been described in considerable detail (ATSDR,
2000; Amandus et al., 1987b). Briefly, an open-pit vermiculite mine, located several miles east
ofLibby, began limited operations in 1923, and production increased rapidly between 1940 and
1950. The mining and milling operations continued until 1990 (ATSDR, 2008. 2000). Some of
the important features of the operations that affected exposure to workers or community
members are described below.
The drilling and blasting procedures used in the open-pit mining operations generated
considerable dust exposures, including silica dust, although the mining operations were
associated with lower intensity exposures compared to the milling operations. Amandus et al.
(1987b) noted that in 1970, a new drill with a dust-control bagging system aimed at limiting
workplace exposure was introduced to the mining operations.
Another aspect of the operations was the loading of ore for railroad shipment. From
1935-1950, railroad box cars were loaded at a station in Libby. In 1950, the loading station was
moved to a loading dock on the Kootenai River, 7 miles east of town. Tank cars were used from
1950-1959 and then switched to enclosed hopper cars in 1960.
The milling operations used a screening or sifting procedure to separate vermiculite
flakes from other particles and increase the concentration ofvermiculite ore from approximately
20% in the bulk ore to 80-95% in the resulting product. A dry mill began operating in 1935, and
a wet mill began operating in the 1950s in the same building as the dry mill. One of the primary
changes in the conditions in the dry mill was the installation of a ventilation fan in 1964.
Exposure to LAA inside the mill was estimated to be 4.6 times higher preceding this installation
(McDonald et al., 1986a). This ventilation fan resulted in higher amphibole fiber exposures in
the mill yard until 1968, when the exhaust stack for the fan was moved to route the exhaust away
from the mill area (Amandus et al., 1987b): although these changes reduced the exposure
potential within the mill operations, they also resulted in spreading the LAA through the
surrounding areas. Other changes to the milling operations in the 1970s included replacement of
hand bagging and sewing with an automatic bagging machine (1972), pressurization of the
skipper control room used for transferring the ore concentrate from the mill to a storage site
(1972), and construction of a new wet mill (1974). Closing of the old dry and wet mills in 1976
had a substantial impact on exposures at the worksite. In 1974, a new screening plant used to
size-sort the ore concentrate was constructed at the loading dock near the river.
Two processing plants operated within the town ofLibby (AT SDR, 200 Ib). These
expansion or exfoliation plants heated the ore concentrate, resulting in additional release of the
LAA fibers in the area.
-------
4.1.1.1.2. Descriptions of cohorts ofLibby, MT vermiculite mining and milling operations
workers. The two cohort studies conducted in the 1980s (by NIOSH and McGill University)
were similar in terms of populations and study design. For example, both studies included
workers who had worked for at least 1 year. Amandus and Wheeler (1987) included men hired
before 1970 (n = 575), with follow-up through December 31, 1981, while McDonald et al.
(1986a) included men hired before 1963 (n = 406) with follow-up through 1983. A subsequent
analysis extended this follow-up through 1999 (McDonald et al., 2004). Another analysis of the
Libby, MT workers expanded the NIOSH cohort to include all workers, regardless of duration of
employment (Sullivan, 2007). The total sample (n = 1,672 white men) included 808 workers
who had worked for less than 1 year. These short-term workers had been excluded from the
previous studies. Analyses presented in the report were based on follow-up from 1960-2001.
Larson et al. (201 Ob) reconstructed a worker cohort based on company records and analyzed
mortality risks through 2006. This study included 1,862 workers (including a small number of
women). Other differences between these two most recent studies include a lower cumulative
exposure estimate in Larson et al. (201 Ob) (median 4.3 fibers/cc-yrs) compared with Sullivan
(2007) (median 8.7 fibers/cc-yrs). Part of this difference could reflect the increase of
approximately 12% in the total sample size, most of whom are likely to have had low exposures,
in Larson et al. (201 Ob), as well as the use of a higher estimated exposure intensity for the
workers with "common laborer" or unknown job categories in Sullivan (2007). A more in depth
examination of the reasons for this difference is not possible based on the available information
in these publications.
4.1.1.1.3. Fiber exposure estimation in Libby, MT mining and milling operations. The
exposure assessment procedures used in the NIOSH and McGill University investigations of the
Libby, MT mining and milling operations relied on the same exposure measurements and used
similar assumptions in creating exposure estimates for specific job activities and time periods
(see Table 4-1). In brief, available air sampling data were used to construct a job-exposure
matrix assigning daily exposures (8-hour time-weighted average) for identified job codes based
on sampling data for specific locations and activities. Various job codes and air exposures were
used for different time periods as appropriate to describe plant operations. Individual exposure
metrics (e.g., cumulative exposure [CE]) were calculated using the work history of each
individual in conjunction with the mine and mill job-exposure matrix.
4-4
-------
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.
TV = 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.
TV = 406 men, hired
before 1963, worked at
least 1 yr
McDonald etal.(1986a)
Libby, MT mining and
milling operations;
ATSDR investigation.
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 estimated using
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)
McDonald et al. (2002)
McDonald et al.
(1986a)
Pulmonary (x-rays):
McDonald et al.
(1986b)
Mortality:
Larson etal. (20 IQb)
Pulmonary:
Larson etal. (2010a)
Larson etal. (2012a)
""Cumulative exposure reported in units of fibers-yr (equivalent to the unit of fibers/cc-yr EPA is using for all
studies).
Mppcf = million particles per cubic foot.
4-5
-------
4.1.1.1.3.1. Asbestos fiber quantification and development of job-exposure matrices.
Before 1970, exposure estimates were based on midget impinger samples taken primarily
in the dry mill by state and federal inspectors (n = 336). Total dust samples were measured as
million particles per cubic foot (mppcf); these samples included mineral dust and vermiculite
particles as well as asbestos fibers. Membrane-filter air samples for fibers, taken at various
locations within the operations, began in 1967, and data are available from company records as
well as state and federal agencies (see Table 4-2). Stationary and short-term (i.e., 20-minute to
less than 4-hour) measurements were primarily used before 1974. Air samples collected through
membrane filters (n = 4,116) were analyzed by PCM to visually count fibers greater than
5-um long and having an aspect ratio >3:1 (Amandus et al., 1987b).9 PCM methods from the
1960s allowed reliable characterization of fibers with widths greater than approximately 0.4 um
(Amandus et al., 1987b: Rendall and Skikne, 1980). Further standardization of the PCM method
and improved quality of microscopes provided better visualization of thinner fibers; a
0.25-um width was considered the limit of resolution for fiber width in the 1980s (IPCS, 1986),
with subsequent improvements in resolution to 0.20 um in width.
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
mppcf3
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; in 1977, MSHA took over MESA's
membrane filter collection activities.
dMSHA: U.S. Mine Safety and Health Administration.
Source: Amandus et al. (1987b)
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.
4-6
-------
The samples taken from specific work locations within the operations were used to
estimate exposures in specific jobs and time periods based on industrial hygienist consideration
of temporal changes in facilities, equipment, and job activities. These were defined to categorize
tasks and locations across the mining, milling, and shipping operations to group like tasks with
respect to exposure potential. Both research groups established similar location operations for
the Libby cohort. Because measures from sample filters were not available before 1967,
different procedures had to be used to estimate exposures at the various locations for this earlier
period. Amandus et al. (1987b) and McDonald et al. (1986a) provide high and low estimates for
several locations to address the uncertainties in assumptions used in these estimates. Both
researchers also applied a conversion factor to estimate asbestos exposures at the dry mill before
1967: the conversion factor was 4.0 in Amandus et al. (1987b) and 4.6 in McDonald et al.
(1986a).10
Jobs were mapped to operation/location based on estimated time spent in different job
tasks, thus estimating an 8-hour time-weighted average exposure for each job during several
calendar time periods. Job histories from date of first employment to 1982 were used with the
job-exposure matrix to develop cumulative exposure estimates for each worker.
Additional considerations
The resulting exposure estimates presented by both research groups, and the job-exposure
matrices used in calculating cumulative exposure for the cohort, were based on fiber counts by
PCM analysis of air filters. As discussed in Section 2 (see Section 2.4.4), PCM analysis does not
distinguish fiber mineralogy or morphology, and all fibers >5 um in length with an aspect ratio
of 3:1 or greater are included. Both researcher groups analyzed fibers available at the facility in
order to identify the mineral fibers in the air samples.
TEM11 analysis of airborne asbestos fibers indicated a range of fiber morphologies—
including long fibers with parallel sides, needlelike fibers, and curved fibers (McDonald et al..
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.
4-7
-------
1986a). Of the fibers examined by TEM, >62% were >5 um in length and a wide range of
dimensional characteristics were noted: length (1-70 um), width (0.1-2 um), and aspect ratios
from 3:1-100:1. Energy dispersive spectroscopy used to determine the mineral analysis
indicated that the fibers were in the actinolite-tremolite solid solution series, but sodium-rich;
some fibers could be classified as magnesio-riebeckite and some as richterite (McDonald et al.,
1986a). This analysis is consistent with the current understanding of amphibole asbestos found
in the Libby mine (see Section 2.2.3).
At the time of their study, when exposure concentrations were reduced to generally less
than 1 fiber/cc, Amandus et al. (1987b) obtained eight air filters from area air samples collected
in the new wet mill and screening plant (provided by the mining company). These samples were
analyzed by PCM using the appropriate analytical method for the time (NIOSH Physical and
Chemical Analytical Method No. 239). From early method development through current PCM
analytical techniques, the Public Health Service, Occupational Safety and Health Administration,
and NIOSH methods have defined a fiber by PCM analysis as having an aspect ratio >3:1
(NIOSH. 1994: Edwards and Lynch. 1968). Amandus etal. Q987b) reported the dimensional
characteristics of the fibers from these filters including aspect ratio, width, and length (see
Table 4-3). Data for 599 fibers from the eight area air samples collected in the wet mill and
screening plant are provided. These data are limited in one sense by the minimum width and
length cutoffs (>4.98-um long, >0.44-um wide, aspect ratio >3:1);12 16% had an aspect ratio
>50:1. Only 7% of the fibers had a width greater than 0.88 um, with one fiber reported of the
559 with a width greater than 1.76 um. Note that these data do not give the full fiber-size
distribution of LAA fibers because NIOSH was examining only PCM-visible fibers (see
Section 2.3.1).
12See footnote 9, page 4-6.
4-8
-------
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 (urn)
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 PCM.
Source: Amandus et al. (1987bX
4.1.1.2. Studies ofO.M. Scott, Marysville, OH Plant Workers
4.1.1.2.1. Descriptions of cohorts ofO.M. Scott, Marysville, OH plant workers. The first study
of pulmonary effects in the Ohio plant workers was conducted in 1980 and involved 512 workers
(97% of the 530 workers previously identified with past vermiculite exposure; (Lockey et al.,
1984): see Tables 4-4 and 4-6). The Rohs et al. (2008) study is a follow-up of respiratory effects
in this cohort conducted approximately 25 years later; chest x-rays and interview data were
collected from 280 of the 431 workers known to be alive at this time. Dunning et al. (2012)
examined mortality rates in the cohort, using an updated exposure analysis described by Borton
et al. (2012). In this analysis, vital status through June 2011 was ascertained.
4-9
-------
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 = 512 (from 530 identified
employees with past vermiculite exposure;
nonparticipants included 9 refusals and
9 unavailable due to illness or vacation).
Mean age: 37.5yr
Mean duration: 10.2yrb
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
(n)
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 pleura! lesions)
(294)
Rohs et al.
(2008)
n = 280 with interviews (2004) and
readable chest x-rays (2002-2005) (from
513 workers in the 1980 study group, 431
were alive in2004d; 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: 59.1yr
Ever smoked: 58.6%
Mean (range) cumulative exposure: 2.48
(0.01-19.03) fibers/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 etal. (1984); see first row of this
table]. Limited to n = 465 white men.
Follow-up through June 2011; 136 deaths
Mean duration: ll.Oyr
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
"Lockev 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-hr 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.
4.1.1.2.1.1. Exposure estimation: O.M. Scott, Marysville, OH plant. The plant that processed
vermiculite ore in Marysville, OH had eight main departments in the processing facility,
employing approximately 530 workers, with 232 employed in production and packaging of the
commercial products and 99 in maintenance; other divisions included research, the front office,
and the polyform plant (Lockev, 1985). Six departments were located at the main facility
4-10
-------
(trionizing, packaging, warehouse, plant maintenance, central maintenance, and front offices).
Research and development and a polyform plant were located separately, approximately
one-quarter mile from the main facility. In the trionizing section of the plant, the vermiculite ore
was received by rail or truck, unloaded into a hopper, and transported to the expansion furnaces.
After expansion, the vermiculite was blended with other materials (e.g., urea, potash, herbicides),
packaged, and stored. Changes to the expander type and dust-control measures began in 1967,
with substantial improvement in dust control occurring throughout the 1970s.
Industrial hygiene monitoring at the plant began in 1972, with measurements based on
fibers >5-um long, diameter <3 um, aspect ratio >3:1. Lockey etal. (1984) noted that the limited
availability of data that would allow for extrapolation of exposures for earlier time periods
possibly resulted in the underestimation of exposures before 1974.13 Breathing-zone samples
were used after 1976, with fiber analysis by PCM.
Cumulative fiber exposure indexes, expressed as fibers/cc-yr, were derived for each
worker from available industrial hygiene data and individual work histories; three categories of
exposure levels were defined for the 512 workers previously identified with past vermiculite
exposure [see Table 4-4 (Lockey et al., 1984)1. Group I was considered to be the nonexposed
group and consisted of the chemical processing, research, and front office workers, as well as
other workers with an estimated cumulative exposure <1 fiber/cc-yr. Group II was the
low-exposure category and included central maintenance, packing, and warehouse workers. The
8-hour time-weighted average fiber exposure in this group was estimated at approximately
0.1-0.4 fiber/cc before 1974 and 0.03-0.13 fiber/cc in and after 1974. Group III was the
high-exposure category and included expander, plant maintenance, and pilot plant workers. The
8-hour time-weighted average fiber exposures in this group were approximately 1.2-1.5 fibers/cc
before 1974 and 0.2-0.375 fiber/cc in and after 1974. The estimated cumulative exposure for the
work force, including Group I workers, ranged from 0.01 to 28.1 fibers/cc-yr using an 8-hour
workday and an assumed 365 days of exposure per year.14 Exposure was assumed to occur from
1957 to 1980 in this study. Exposure outside of work hours was assumed to be zero.
Additional exposure information was identified in 2009, and exposure estimates were
updated and refined to reflect information (including fiber measurements) from company reports
and other written materials (Borton etal., 2012). In addition, worker focus groups provided
insight into plant processes—including industrial hygiene measures—and work patterns and
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).
4-11
-------
4.1.1.3. Community-Based Studies Around Libby, MT Conducted by Agency for Toxic
Substances and Disease Registry (ATSDR)
Analyses using data from community-based studies in Libby, MT conducted by ATSDR
are summarized in Table 4-5. ATSDR (2000) includes a mortality analysis based on death
certificate data from 1979-1998, with residence at time of death geocoded to areas
corresponding to Libby, MT and its surroundings. The estimated population size in 1991 for the
areas used in the standardized mortality ratio (SMR) calculations ranged from 2,531 in the Libby
city limits to 9,512 for the central Lincoln County area (based on a 10-mile radius around
downtown Libby). Cause-specific standardized mortality ratios were computed based on
Montana and United States comparison rates; asbestosis SMRs were somewhat higher using the
U.S. referent group, but the choice of referent group had little difference on SMRs for most
diseases.
ATSDR also conducted a community health screening from July-November 2000 and
July-September 2001, with 7,307 total participants (ATSDR. 200Ib). Eligibility was based on
residence, work, or other presence in Libby for at least 6 months before 1991. The total
population eligible for screening is not known, although the population of Libby, MT in 2000
was approximately 10,000. Other studies (Larson et al., 2012b; Weill et al., 2011; Vinikoor et
al., 2010; Noonan et al., 2006) used data collected during this community health screening.
Two additional community-based studies, using data sources other than the ATSDR
community health screening (Marchand et al., 2012; Pfau et al., 2005) are discussed in
Section 4.1.3.2. (Autoimmune disease and autoantibodies), and two clinic-based studies from
Libby, MT (Winters et al., 2012) are discussed in Section 4.1.2.2.4 (Clinic-based reports and
case reports of respiratory disease [noncancer]).
4-12
-------
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, 200Ib) 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, 200lb)1.
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, 200lb)1.
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; pulmonary
function in relation to
chest x-ray findings
Noonan et al.
(2006)
ATSDR community health screening
[see (ATSDR, 200lb)1.
Nested case-control study of
rheumatoid arthritis, scleroderma, and
systemic lupus erythematosus cases
(« = 161 cases, 1,482 controls); initial
self-report confirmed in second
interview.
Exposure pathways as described in
Peipins et al. (2003)
Systemic autoimmune
diseases
4-13
-------
4.1.2. Respiratory Effects, Noncancer
4.1.2.1. Asbestosis and Other Nonmalignant Respiratory Disease Mortality
Several studies described previously reported noncancer respiratory disease mortality
data. Nonmalignant respiratory disease is a broad category (International Classification of
Diseases [ICDJ-9 codes 460-519) that includes chronic obstructive pulmonary disease, asthma,
pneumonia and respiratory infections, asbestosis (ICD-9 code 501), and various forms of
pneumoconiosis. A greater specificity of effects due to asbestos would be expected using the
narrower category of asbestosis compared with nonmalignant respiratory disease.
The initial studies of the Libby, MT vermiculite mining and milling worker cohorts were
based on a relatively small number of nonmalignant respiratory-related deaths (<25); more than
50 deaths in this category were seen in later studies (see Table 4-6). The analytic strategy (e.g.,
use of a latency period to exclude cases that occurred before the effect of exposure would be
expected to be manifested, or use of a lag period to exclude exposures that occurred after the
onset of disease) and the cutpoints for exposure categories varied among the studies, but a
pattern of increasing risk with increasing cumulative exposure is seen, with more than a 10-fold
increased risk of death due to asbestosis and a 1.5- to 3-fold increased risk of nonmalignant
respiratory disease in the analyses using an internal referent group (Larson et al., 201 Ob:
Sullivan. 2007: McDonald et al.. 2004). Larson etal. (201 Ob) used a Monte Carlo simulation to
estimate the potential bias in nonmalignant respiratory disease risk that could have been
introduced by differences in smoking patterns between exposed and unexposed workers in the
cohort. The bias-adjustment factor (relative risk [RR]Unadjusted/RRadjusted = 1.2) reduced the overall
RR estimate for nonmalignant respiratory mortality from 2.1 to 1.8. Asbestosis risk was also
increased in the ATSDR geographic-based analysis, with SMRs of approximately 40 based on
Montana rates and 65 based on U.S. comparison rates (ATSDR, 2000). Only one
asbestosis-related death was observed in the Marysville, OH worker cohort, resulting in a very
imprecise risk estimate (Dunning et al., 2012).
4-14
-------
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)
McDonald et
al. (1986a)
(McGill)
Sullivan
(20071
(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
o
5
10
SMR (95% CI)b
2.2 (not reported)
1.7 (not reported)
1.8 (not 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
o
J
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
25.2-1 13.7 fibers/cc-yr
>1 13. 8 fibers/cc-yr
per 100 fibers/cc-yr
n
5
13
14
19
-
RR (95% CI)d
1.0 (referent)
2.5 (0.88, 7.2)
2.6(0.93,7.3)
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
SRR (95% CI)C
1.0 (referent)
7.3(1.9,28.5)
25.3 (6.6, 96.3)
(pO.OOl)
15-yr exposure lag: nonmalignant respiratory diseases
Cumulative Exposure
0.0-4.49 fibers/cc-yr
4.5-19.9 fibers/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)
linear trend test
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)
(p<0.01)
4-15
-------
Table 4-6. Nonmalignant respiratory mortality studies of populations
exposed to Libby Amphibole asbestos" (continued)
Reference(s)
Larson et al.
(2010b)
Respiratory disease
(SMR, 95% CI)
Asbestosis (n = 69)
SMR: 143(111, 181)
Nonmalignant
respiratory diseases
(n = 425)
SMR: 2.4(2.2,2.6)
Chronic obstructive
pulmonary disease
(n = 152)
SMR: 2.2(1.9,2.6)
Other nonmalignant
respiratory (n 120)
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-<8.6fibers/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)
(/?<0.0001)
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)
Libby city limits
Extended Libby
boundary
Air modeling
Medical screening
Libby valley
Central Lincoln County
SMR
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)
(95% CI)
Comparison area
(Montana reference rates):
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)
SMR (95% CI)
Comparison area
(U.S. 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)
4-16
-------
Table 4-6. Nonmalignant respiratory mortality studies of populations
exposed to Libby Amphibole asbestos" (continued)
Occupational studies ofO.M. Scott, Marysville, OH plant workers
Dunning et
al. (2012)
Asbestosis (n =
SMR: 15.4(0.4
1)
,86)
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.
4.1.2.1.1. Pathological alterations of the lung parenchyma and pleura, pulmonary function,
and respiratory symptoms
4.1.2.1.1.1. Definition of outcomes
Text Box 4-1. Pathological Alterations of the Lung Parenchyma and Pleura According
to ILO (2002)
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(s)
of the lung.
Obliteration of the costophrenic angle: The costophrenic angle is the angle between the
ribcage and the diaphragm on a standard posterior-anterior 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 localized pleural thickening
(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." LPT may also be viewed
in-profile or face-on and is generally a pleural plaque (parietal). Calcification is noted
where present.
4-17
-------
Respiratory disease risk can be evidenced by pleural and parenchymal abnormalities
(pathological, structural alterations) detected through radiographic or other types of imaging (see
Text Box 4-1). These types of effects are usually classified using criteria developed by the
International Labour Organization (ILO) of the United Nations to standardize descriptions of
effects and improve inter-rater agreement and accuracy for reading chest radiographs in
pneumoconiosis. The guidelines were initially developed in 1950 with several subsequent
revisions. A key component of the guidelines is the use of a set of standard films illustrating
different types of findings; these films are used by "B Readers" as a reference for comparison to
films collected in a research or clinical setting. Findings that were considered to be definitely or
probably not pneumoconiosis were to be noted by the readers to allow resolution of
discrepancies (ILO, 1980). The B Reader program was initiated in 1974 to reduce variability in
readings; B Readers are physicians who pass an examination, recertifying every 4 years, in the
adherence to detailed criteria when reading radiographs in individuals with pneumoconiosis.
Parenchymal (the inner structure of the lungs) abnormalities include opacities; these
abnormalities are defined as small (<10 mm diameter) or large (>10 mm diameter). Small
opacities are assigned a score based on the concentration of opacities in a given area (profusion),
zone(s) of the lung(s) affected, shape, and size. Small opacity profusion is graded on a 4-point
scale (0 = absence of small opacities or the presence of small opacities less profuse than
Category 1; 3 = highest level of profusion). Two ratings are given (e.g., 0/1 or 2/2), with the
second number allowing an indication of a category that was seriously considered as an
alternative to the first grade. Large opacities are scored based on their (aggregate) dimension(s).
The scarring of the parenchymal tissue of the lung contributes to measurable decrements in
pulmonary function, including obstructive pulmonary deficits from narrowing and/or distortion
of airways, restrictive pulmonary deficits from the decreased elasticity of the lung or
displacement of lung tissue by mass lesions, and decrements in gas exchange (ATS, 2004).
According to 2000 ILO guidelines (ILO, 2002), pleural abnormalities are classified as
(a) localized pleural thickening (LPT) or (b) diffuse pleural thickening (DPT). LPT can be
present on the pleural membrane of the chest wall, diaphragm, and other sites (medinstinal
pleura, para-spinal pleura, and para-cardiac pleura). Plaques on the chest wall can be viewed
either face-on or in profile. A minimum width of about 3 mm is required for an in-profile plaque
to be recorded as present according to the 2000 ILO guidance. This requirement was added to
reduce the number of false positives. The terms "LPT" and "pleural plaque" are the same in the
2000 ILO guidelines, and are different from the definition of pleural plaques used prior to 2000
ILO guidelines. Different researchers implementing the earlier ILO guidelines variously used
terms such as "discrete pleural thickening," "circumscribed pleural thickening," or "pleural
plaques" to describe what is currently called LPT. DPT is now defined as diffuse pleural
thickening on the chest wall that is present "only in the presence and in continuity with an
obliterated costophrenic angle." The previous ILO guideline (ILO, 1980) did not include the
4-18
-------
requirement for involvement of the costophrenic angle. The definition of DPT now also includes
a requirement that the thickening on the chest wall have a minimum width of 3 mm when viewed
in profile. The changes in the definition of DPT were added to reduce the number of false
positives. Both LPT and DPT are scored based on their location, extent, and whether
calcification is seen and are recorded separately for the left and right side. LPT is a change in
tissue structure and is not known to be an adaptive response to toxicity generally or to asbestos
specifically. Examples of what is now called LPT visualized on autopsy are shown in Figures 4-
2A and 4-2B (ATS, 2004). Additional discussion of the adversity of LPT is included in
Section 5.2.2.3 (Selection of Critical Effect) and Appendix I.
Figure 4-2. A (left). Gross appearance at autopsy of asbestos-associated
pleural plaques overlying the lateral thoracic wall [(ATS, 2004) Figure 12].
Figure 4-2. B (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.
The latency period for the initial detection of pleural or parenchymal abnormalities varies
by type of lesion. Larson et al. (2010a) examined x-rays of 84 workers from the Libby, MT
mining and milling operations for whom pleural and/or parenchymal abnormalities were seen
and who had one or more previous x-rays covering a span of at least 4 years available for
comparison. Circumscribed pleural plaques [a term that corresponds to the term "circumscribed
4-19
-------
pleural thickening" in ILO (1980)1 was seen in 83 of these 84 workers at a median latency of
8.6 years. Any pleural calcification was seen in 37 workers, with a median latency of 17.5 years,
and DPT was seen in 12 workers (median latency: 27.0 years). The latency period for small
opacities indicating parenchymal changes (e.g., asbestosis) increased with increasing profusion
categories, from a median of 18.9 years for>1/0, 33.3 years for progression to >2/l, and
36.9 years for progression to >3/2.
Pulmonary function
Pulmonary function, commonly measured by spirometry, is used as an indicator of
respiratory health and lung disease. Spirometric measurements involve assessment of lung
volume and air flow (Pellegrino et al., 2005). Forced vital capacity (FVC) is a measure of the
maximum amount of air that can be exhaled forcefully after a full inspiration. Forced expiratory
volume (FEV) is the maximum amount of air exhaled forcefully after a full inspiration in a given
time period; for example, FEVi refers to the amount of air exhaled in the first second of the test
procedure. Standardization of test procedures is very important in these tests, and measurements
of multiple forced expiration (>3) are typically needed. Values are compared to "reference
values" based on age, gender, height (and sometimes race).
Combinations of various functional measurements may be indicative of specific types of
abnormalities affecting lung function. For example, restrictive lung function (or restrictive
ventilatory defect) refers to reduced lung volume. Both FEVi and FVC would be reduced, but
the reduction in FVC would typically be greater than that for FEVi (e.g., FEVi/FVC ratio >0.8).
Restrictive lung function can result from interstitial lung disease (including inflammatory or
fibrotic disease) or other conditions that restrict the ability of the lungs to expand. Obstructive
lung function (or obstructive ventilatory defect) refers to reduced airflow, and is most commonly
characterized by narrowing of the airways. It is indicated by a reduction in FEVi without a
proportionate reduction in FVC (e.g., the ratio of FEVi/FVC <0.7, or FEVi/FVC <5th percentile).
Both restrictive and obstructive conditions can result in dyspnea (shortness of breath), and many
of the various underlying diseases that are associated with restrictive and obstructive lung
function cause cough, and chest pain.
4.1.2.1.1.2. Results: pathological alterations of lung parenchyma and pleura: occupational
studies
Libby, MTvermiculite mine and mill workers
Studies examining pleural and parenchymal abnormalities in the Libby, MT worker
cohorts are shown in Table 4-7. In the McDonald et al. (1986b) and Amandus et al. (1987a)
studies, x-ray films for each worker, which NIOSH obtained from the Libby hospital that
performed the screening, were independently read by three qualified readers using the 1980 ILO
classification system. For the analysis, classification indicating pleural abnormalities by at least
4-20
-------
two of the three readers was used to determine the presence of pleural abnormalities, while the
median reading was used to determine the profusion category of small opacities. Although both
research groups used the ILO 1980 guidelines, McDonald et al. (1986b) reported pleural
thickening on the chest wall (both pleural plaques and diffuse) but not in other sites. Amandus et
al. (1987a) examined "any unilateral or bilateral pleural change", which included ".. .pleural
plaque, diffuse pleural thickening of the chest wall, diaphragm or other site, but excluded
costophrenic angle obliteration."
Amandus et al. (1987a) reported pleural thickening of the chest wall in 13% and small
opacities (>1/0) in 9.8% of employees. The analysis reported by McDonald et al. (1986b) was
stratified by employment status. Among current workers, pleural thickening of the chest wall
and small opacities were observed in 15.9% and 9.1%, respectively; corresponding figures
among former workers were 52.5% and 37.5%, respectively. In both studies, prevalence of these
abnormalities increased with increasing cumulative exposure. McDonald et al. (1986b) also
included 80 former employees in their study. The prevalence of pleural thickening of the chest
wall (52.5%) and small opacities (37.5%) was higher in former employees compared with
current workers. These groups differed by age, with only one of the 80 former workers being less
than 40 years of age while 80 of 164 current workers were under 40 years of age. Both overall
and within age categories, however, the prevalence is higher among former employees, and this
is attributed to higher cumulative exposure in this group.
Both Amandus et al. Q987a) and McDonald et al. Q986b) provided categorical
exposure-response data as well as logistic models for various endpoints (e.g., small opacities,
pleural calcification, pleural thickening of the chest wall, and "any pleural change"). In
McDonald et al. (1986b), exposure and age were both predictive of pleural thickening along the
chest wall; the regression coefficient for cumulative exposure (fibers/cc-yr) was 0.0024 per unit
increase in cumulative exposure for the log odds of the presence of pleural thickening, adjusting
for age and smoking. Cumulative exposure, age, and smoking status were all predictive of small
opacities; the parameter for cumulative exposure had a regression coefficient of 0.0035 per unit
increase in cumulative exposure. In contrast, although the categorical analysis reported by
Amandus et al. (1987a) indicated a positive exposure response relationship for both "any pleural
change" and pleural thickening along the chest wall, cumulative exposure was not a significant
predictor in regression analysis adjusting for age (regardless of smoking status). The lack of
statistical significance in these models may reflect a nonlinearity reflected in the observation of
the lowest prevalence in the second of four exposure categories rather than in the lowest
category. The estimated relationship between exposure and prevalence of small opacities in
Amandus et al. (1987a) was similar to that reported by McDonald et al. (1986b).
4-21
-------
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 (w = 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
3.6fibers/cc-yr
Restrictive spirometry defined as
FVC < lower limit of normal and
FEVi/FVC > lower limit of normal.
Results
Prevalence (%)
Pleural thickening
Small opacities
£1/0)
Current
workers
15.9
9.1
Former Comparison
workers group
52.5 8.5
37.5 2.1
Both abnormalities increased with age, and with increasing
cumulative exposure in age-adjusted and stratified (>60 yr
old) analyses.
Pleural thickening
of the chest wall observed in 13%.
Small opacities (>1/0) observed
in 10%.
Beta (/?-value), cumulative exposure in relation to:
Small opacities
Any pleural change
Pleural calcification
Pleural change on wall
0.0026
0.0008
-0.0010
0.0008
(p<0.05)
(p>0.05)
(p>0.05)
(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 (5)
18 (5)
117 (35)
45 (16)
Association with cumulative
exposure (cc/fiber-yr)a
Starting at
5
1
0.5
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.
Larson et al. (2012a) used data collected as part of the community screening program
conducted in 2001 [(ATSDR, 200lb): see Section 4.1.1.3] to examine the pleural and pulmonary
outcomes based on chest radiographs, spirometry results, and self-reported symptoms in relation
to cumulative exposure among 336 workers. Diffuse pleural thickening (in the presence of
4-22
-------
costophrenic angle obliteration) and parenchymal small opacities (profusion >1/0) were each
detected in 5% of the workers. Risk increased monotonically with increasing cumulative
exposure for each of these outcomes; however, the slope was shallower for diffuse pleural
thickening and was not statistically significant. LPT (only) was found in 35% of the workers
with an elevated risk associated with cumulative exposures as low as 1 fiber/cc-yr. For a
diagnosis of restrictive spirometry (excluding mixed restrictive and obstructive spirometry;
prevalence = 16%), risk began to increase at 26 fibers/cc-yr and reached statistical significance at
166 fibers/cc-yr. Chronic bronchitis defined as coughing up phlegm "for at least 3 months of the
year for the past 2 years" was reported in 8% of the workers, and a statistically significant
increased risk was calculated at 24 fibers/cc-yr.
O.M. Scott, Marysville, OH plant workers
The first study of the O.M. Scott, Marysville, OH plant workers was conducted by
(Lockey et al., 1984): see Table 4-8. Physical examination (for detection of pulmonary rales and
nail clubbing), pulmonary function (spirometry and DLco), and chest x-rays were performed, and
information pertaining to smoking history, work history at the plant, and other relevant work
exposures was collected using a trained interviewer. Approximately 44% of the 512 workers in
the study were current smokers, 20% former smokers, and 35% lifetime nonsmokers, but
smoking history (i.e., smoking status, pack-years) did not differ by exposure group. An
increased risk of costophrenic angle blunting (n = 11), other pleural and parenchymal
abnormalities (n = 11), or any of these outcomes (n = 22) was observed in relation to exposure
assessed by job title and area (see description of exposure groups in Section 4.1.1.2.2) and
categorized into groups based on the cumulative fiber estimates. The prevalence of any
radiographic change was 2.8% in Group I, 3.9% in Group II, and 5.8% in Group III. Using the
cumulative fiber metric, the prevalence of any radiographic change was 2.4% in the
<1 fiber/cc-yr group, 5.0% in 1-10 fibers/cc-yr group, and 12.5% in the >10 fibers/cc-yr group.
Lockey et al. (1984) used a modification of the ILO 1971 guidelines; one modification was that
costophrenic angle blunting was considered a category separate from other pleural lesions.
A follow-up study of this cohort [(Rohs et al., 2008); see Table 4.4] included
298 workers, of whom 280 completed the study interview (with work history and smoking
history) and chest x-ray. The evaluation of each worker included an interview to determine work
and health history, pulmonary examination, and chest x-ray. Exposure was estimated using the
procedure described (Lockey etal., 1984). Exposure was assumed to occur from 1963 to 1980 in
this study, assuming an 8-hour workday and 365 days of exposure per year (Benson, 2014).
Each worker supplied a detailed work history (start and end date for each area within the
facility). The exposure reconstruction resulted in a cumulative exposure estimate for each
individual. The estimated cumulative exposure for this follow-up study ranged from 0.01 to
4-23
-------
19.03 fibers/cc-yr (mean = 2.48). The time from first exposure ranged from 23 to 47 years.
Exposure outside of work was assumed to be zero.
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
Lockev (19851
Lockev et al.
(1984)a
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
pleural thickening or plaques in (n = 10);
bilateral, small opacities (n= 1).
Abnormality (combined outcomes) increased
with increasing cumulative exposure.
Rohs et al.
(2008)
n = 280 with interviews (2004) and readable
chest x-rays (2002-2005)
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 pleural 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).
Three board-certified radiologists, blinded to all identifiers, independently classified the
radiographs using the 2000 ILO classification system (ILO. 2002). Rohs et al. (2008)
determined that diffuse pleural thickening was present when at least two of the three readers
recorded pleural thickening with blunting of the costophrenic angle, localized pleural thickening
was present when at least two of the three readers recorded thickening, with or without
calcification, excluding solitary costophrenic angle blunting, and that interstitial abnormalities
indicative of asbestosis were present if at least two of the three readers identified small irregular
opacities of profusion 1/0 or greater. Radiographs classified as unreadable (n not reported) were
not used in the analysis.
Pleural thickening was observed in 80 workers (28.7%), and small opacities (>1/0) were
observed in 8 (2.9%). The 80 workers with pleural thickening included 68 with LPT only (85%)
and 12 with DPT (15%). Six of the eight participants with small opacities also had pleural
4-24
-------
thickening (four as LPT, two as DPT). The prevalence of pleural thickening increased across
exposure quartiles from 7.1% in the first quartile to 24.6, 29.4, and 54.3% in the second, third,
and fourth quartiles, respectively [see Table 4-9 (Rohs etal., 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,
fibers/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)
"Two 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. (2008X 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).
Pleural thickening was strongly associated with hire on or before 1973 and age at time of
interview, but not with body mass index (BMI) or smoking history (ever smoked; see
Table 4-10); BMI is a potentially important confounder because fat pads can sometimes be
misclassified on chest x-rays as localized pleural thickening. A hire date of on or before 1973
and age at time of interview are each highly correlated with cumulative exposure to fibers. The
small number of females (n = 16) in the cohort limits the analysis of the association with gender.
4-25
-------
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
^-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).
Odds ratios (ORs) for quartiles of cumulative fiber exposure were also estimated
including various cofactors (age, hired before 1973, or BMI). Each model demonstrated the
same trend: increased prevalence of pleural thickening with increasing cumulative exposure to
fibers. Adjusting for age, date of hire, and BMI resulted in odds ratios of 2.7, 3.5, and 6.9 for the
second, third, and fourth quartiles, respectively. There was no evidence of significant
interactions using this modeling.
4-26
-------
There was potential coexposure to a number of fertilizers, herbicides, pesticides, and
other chemicals in the facility (Smith, 2014).15 No quantitative information on exposure to these
chemicals is available. However, the addition of the other chemicals to the vermiculite carrier
occurred in a different part of the facility after expansion of the vermiculite ore. Industrial
hygiene monitoring in these areas showed very low levels of fibers in the air. It is unlikely that
workers would be coexposed by inhalation to these other chemicals. In addition, EPA has no
information indicating that exposure to any of these individual chemicals causes pleural
thickening or small opacities typical of those found in workers employed in the Marysville
facility, and thus EPA does not consider the presence of these coexposures likely to produce any
confounding in the observed associations between LAA exposure and the pulmonary effects seen
in this cohort.
The Rohs et al. (2008) study demonstrates that exposure to LAA can cause radiographic
evidence of pleural thickening and parenchymal abnormalities (small opacities) in exposed
workers. The prevalence of pleural abnormalities was 28.7% in 2004 (80/280), compared to a
2% prevalence observed in 1980 (10/501). In addition, the prevalence of small opacities
increased from 0.2% (n = 1) in 1980 to 2.3% (n = 8) in the 2004 study. These increases in
prevalence are most likely due to the length of follow-up in the later study, giving additional
time for the abnormalities to become apparent in conventional x-rays; little additional exposure
occurred after 1980. The follow-up study also shows an increasing prevalence of pleural
thickening with increasing cumulative exposure to LAA.
The influence of some potential sources of selection bias in Rohs et al. (2008) is difficult
to qualitatively or quantitatively assess. One type of conceivable selection bias is the loss of
participants due to the death of 84 of the 513 (16%) workers in the first study; this group may
represent a less healthy or more susceptible population. Exclusion of the very sick or susceptible
may imply that the population of eligible participants was somewhat healthier than the whole
population of workers; this exclusion may result in an underestimation of risk. Another type of
selection is the loss due to nonparticipation among the 431 individuals identified as alive in 2004
(n = 135 refusals and nonresponders; 31%). Participation rates in epidemiologic studies can be
associated with better health status, and participation is often higher among nonsmokers
compared with smokers. This type of selection of a relatively healthier group (among the living)
could also result in an underestimation of the risk of observed abnormalities within the whole
exposed population. However, if participation was differentially related to exposure and
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.
4-27
-------
outcome (i.e., if workers experiencing pulmonary effects and who were more highly exposed
were more likely to participate than the highly exposed workers who were not experiencing
pulmonary effects), the result would be to overestimate the exposure response relationship. This
latter scenario is less likely to occur for asymptomatic effects (i.e., abnormalities detected by
chest x-ray), such as those that are the focus of this study, than for symptoms such as shortness
of breath or chest pain.
Some information is available on differences by participation status in the Rohs et al.
(2008) study. Although current age was similar (mean: 59.1 and 59.4 years, respectively, in
participants and living nonparticipant groups,/? = 0.53), participants were more likely to have
been hired before or during 1973 (66.4 and 49.7%, respectively,/* = 0.001) and were also
somewhat less likely to ever be smokers (58.6%) compared with the living nonparticipants
(66.2%). Participants had higher mean exposure levels (mean cumulative exposure: 2.48 and
1.76 fibers/cc-yr, respectively, in participants and nonparticipants, p = 0.06), but when
combining living and deceased nonparticipants, there is no evidence of major differences in
exposure distribution in participants compared with the original full population.
4.1.2.1.1.3. Results: pathological alterations of lung parenchyma and pleura, pulmonary
function, and respiratory symptoms—community-based studies
Pathological alterations of parenchyma and pleura
In the ATSDR community health screening IYATSDR, 200Ib) see Table 4-15], two
board-certified radiologists (B Readers) examined each radiograph, and a third reader was used
in cases of disagreement (see Tables 4-5 and 4-11). Readers were aware that the radiographs
were from participants in the Libby, MT health screening but were not made aware of exposure
histories and other participant characteristics (Peipins et al., 2004a: Price, 2004; Peipins et al.,
2003). The radiographs revealed pleural abnormalities in 17.9% of participants, with prevalence
increasing with increasing number of "exposure pathways" (defined on the basis of potential
work-related and residential exposure to asbestos within Libby and from other sources). The
authors noted that the relationship between number of exposure pathways and increasing
prevalence of pleural abnormalities was somewhat attenuated after excluding those who had
formerly worked at the vermiculite mining and milling operations. The prevalence of pleural
abnormalities decreased from approximately 35 to 30% in individuals with 12 or more exposure
pathways when these workers were excluded from the analysis. Among individuals with no
defined exposure pathways, the prevalence of pleural anomalies was 6.7%, which the authors
report is higher than reported in other population studies (Peipins et al., 2004a: Price, 2004). The
direct comparability between study estimates is difficult to make; the possibility of over- or
under ascertainment of findings from the x-rays based on knowledge of conditions in Libby was
not assessed in this study. No information is provided regarding analyses excluding all potential
work-related asbestos exposures.
4-28
-------
Weill et al. (2011) used the ATSDR community health screening data to analyze the
prevalence of x-ray abnormalities in relation to age, smoking history, and types of exposures (see
Tables 4-5 and 4-11). Analysis was based on five exposure categories in n = 4,397 participants
ages 25 to 90 years. The prevalence of x-ray abnormalities (plaques, or diffuse pleural
thickening, and/or costophrenic angle obliteration) also generally increased with age (divided
into 25-40, 41-50, 51-60, and 61-90 years) within each of the exposure categories, with the
highest prevalence seen among former workers in the vermiculite mining and milling operations.
Among those with environmental exposure only (i.e., no household or occupational exposures),
the prevalence increased from approximately 2% at ages 41-50 years to 12% at ages
61-90 years.
The community-based study by Alexander et al. (2012) was conducted in an area other
than Libby, MT. The Western Minerals plant in Minneapolis, MN processed Libby vermiculite
ore to produce insulation material from 1939 to 1989. The plant was surrounded by residential
neighborhoods, and the waste material from the plant was offered to community residents for use
as filler in their yards and driveways. The Minnesota Department of Health and ATSDR
initiated a study of community exposures in 2000, including a baseline survey of
>6,400 residents. Residential history information was combined with period-specific air
dispersion models and data on facility emissions to classify the level of background exposure
(Kelly et al., 2006): details pertaining to the input parameters and modeling assumptions are
limited and result in considerable uncertainty in exposure estimates. Intermittent high exposures
were estimated for specific activities (e.g., playing on waste piles, moving waste from the plant)
based on experiments reconstructing exposure occurring during these activities (Adgate et al.,
2011). In a follow-up study of people who had not worked in the plant (or lived with a worker
employed at the plant), measures of background exposure and activity-based (intermittent)
exposure were associated with increased prevalence of pleural abnormalities [(Alexander et al.,
2012) see Table 4-111.
4-29
-------
Table 4-11. Pathological alterations of lung parenchyma and pleura in
community-based studies
Reference(s)
Inclusion criteria and design
details
Results
Libby, MT community
Peipins et al.
(2003)
ATSDR
(200 Ib)
Weill et al.
(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.
Pleura! abnormality: (a) any unilateral
or bilateral pleura! 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
categories; n = 880), and
(5) environment ("no" to work and
household exposures in
Categories 1-4; n = 1,894).
Outcomes: (1) Small lung opacities
defined as "any two readers reporting
any profusion >1/0"; (2) Plaque,
defined as "any two readers reporting
any diaphragm or wall, or other site
plaques, even if the readers did not
agree on specifics"; (3) 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%
prevalence with increasing m
(6.7% among those with no s
among those with 12 or more
ATSDR (200 Ib): Moderate-
<70% predicted): 2.2% of m
>17yrold
Exposure Source
of participants; increasing
amber of exposure pathways
pecific pathways, 34.6%
pathways).
to-severe restriction (FVC
en >17 yr old; 1.6% of women
Prevalence (%)
Small
opacities
Plaque
DPT/CAO
Age 25-40 (n = 1,075)
Vermiculite worker by
company
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
company
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
4-30
-------
Table 4-11. Pathological alterations of lung parenchyma and pleura in
community-based studies (continued)
Reference(s)
Weilletal.
(20111
(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);
Clinical examination, chest x-rays
read by 2000 ILO classification
guidelines.
Participants more likely than
nonparticipants to report
exposure-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. (201 1).
Pleura! abnormality (any)
DPT
LPT
Prevalence
49 (10.6%)
5 (1.1%)
45 (9.9%)
Regression analysis:
Exposure type Beta(±SE)
Background 0.322 (±0.125)
Intermittent 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.
Respiratory symptoms
Vinikoor et al. (2010) used the 2000-2001 health screening data to examine respiratory
symptoms and pulmonary function results among 1,003 adolescents and young adults (<18 years
in 1990 when the mining/milling operations closed), excluding individuals with a work history
4-31
-------
that could result in vermiculite or dust exposure (see Tables 4-5 and 4-12). The potential for
vermiculite exposure outside of the workplace was classified based on responses to questions
about six activities (e.g., handling vermiculite insulation, playing in vermiculite piles, "popping"
vermiculite by heating it to make it expand). The medical history questionnaire included
information on three respiratory symptoms: (1) usually have a cough (n = 108, 10.8%);
(2) troubled by shortness of breath when walking up a slight hill or when hurrying on level
ground (n = 145, 14.5%); or (3) coughed up phlegm that was bloody in the past year
(n = 59, 5.9%). A question on history of physician-diagnosed lung disease (n = 51, 5.1%) was
also included. The pulmonary function results were classified as normal in 896 (90.5%),
obstructive in 62 (6.3%), restrictive in 30 (3.0%), and mixed in 2 (0.2%). There was little
variation in prevalence of shortness of breath, physician-diagnosed lung disease, or abnormal
spirometry across the exposure categories; for two symptoms, the highest relative risk was seen
in the highest exposure group, but neither of these estimates was statistically significant (OR
2.93, 95% confidence interval (CI) 0.93, 9.25 for usually having a cough and OR 1.49, 95% CI:
0.41, 5.43 for coughing up bloody phlegm).
Table 4-12. Pulmonary function and respiratory symptoms and conditions
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) by exposure category3
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.
Obstructive (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).
4-32
-------
Pathological alterations of lung parenchyma and pleura in relation to pulmonary function
Two studies that examined LAA specifically provide insight into the question of the relation
between specific types pleural or parenchymal lesions and pulmonary function (see Table 4-13).
These studies, based on data from the community health screening conducted by ATSDR,
reported an association between the presence of pleural plaques and a reduced mean FVC
(approximately 5% below predicted) (Weill et al., 2011), and of an increased risk of restrictive
pulmonary function (Larson et al., 2012b). The authors of the first study (Weill et al., 2011)
concluded that the mean reduction in FVC associated with pleural plaques (in the absence of
other pleural abnormality or small opacities in the lung parenchyma) was "probably clinically
insignificant." The second study (Larson et al., 2012b) focused on the likelihood of an
"abnormal" pulmonary function test (restrictive lung function), rather than on a difference in
population means. Although the association with a restrictive pulmonary function was weaker
for circumscribed pleural plaques (OR = 1.4) than for DPT (OR = 4.1), the risk of restrictive
lung function increased with increasing index score based on plaque width and extent. Among
those with restrictive impairment, risk of functionally significant pulmonary impairment was
associated with the presence of plaques with a high index score (OR 1.7, 2.1, and 2.3 for
outcomes of mild, moderate, and severe levels of impairment, respectively). These two analyses,
using essentially the same data set, illustrate that what may appear on first thought to be an
"insignificant" impact on lung function (i.e., a relatively small mean decrement in lung function
among an affected population) is entirely compatible with the presence of clinically important
individual lung function decrements for some individuals within that population. An additional
point that can be drawn from both studies is the relative strength of the influence of DPT on
FVC, for example, with a >20% decrease in percent predicted associated with DPT (in the
absence of small opacities in the lung parenchyma) in Weill et al. (2011).
4-33
-------
Table 4-13. Analyses of pulmonary changes seen on radiographs in relation
to pulmonary function in the Libby, MT community
Reference(s)
Weill et al.
(2011)
Larson et al.
(2012b)
Inclusion criteria and design details
Participants in ATSDR 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 (Hankinson et 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 (median
DPT, 2.5 for LPT)
(95% CI)
(2.1,7.8)
(1.4,6.0)
(1.2,2.4)
(1.1,1.8)
.0 JOT
DPT Only
median
78
57
2.1
5.6
(1.1,3.7)
(2.7, 11.6)
Index of plaque size
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%
-------
radiographic evidence of interstitial changes at profusion category 0/1 or 1/0 evident on their
initial chest x-ray. No evidence of interstitial changes were found in the other 67 (55%), but all
67 had evidence of pleural disease (either pleural plaques or diffuse pleural thickening). For the
entire group of 123, the average yearly loss of pulmonary function over a mean follow-up time
of 35 months was 2.2% for FVC, 2.3% for total lung capacity, and 3.0% for DLco. For the
subset of 67 patients with pleural disease (in the absence of interstitial abnormality), the
corresponding mean declines were 2.2%, 2.3%, and 2.9%, respectively. Although the details of
the classification system used for the radiographic readings was not described, the analysis of the
pulmonary function testing incorporated appropriate analytic procedures.
A study by Winters et al. (2012) was conducted among patients seen for annual
examinations at a clinic in Libby, MT specializing in the diagnosis and treatment of
asbestos-related disease (see Table 4-14). The x-rays (posterior-anterior and lateral views) were
read by one radiologist. In this clinic sample, 60 individuals were considered to have no
abnormality, 182 had pleural abnormality only, 18 had interstitial abnormality, and 69 had both
pleural and interstitial abnormalities. FVC was lower among those with pleural abnormalities
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.
4-35
-------
Table 4-14. Pulmonary function and respiratory system changes in the
Libby, MT community: clinic-based study
Reference(s)
Winters et al.
(20121
Inclusion criteria and
design details
Patients seen at Center
for Asbestos-Related
Disease clinic in Libby,
MT.
Af= 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 %predicted (SD)
Normal
(n = 60)
Abnormalities
Pleural only
(n = 182)
Interstitial
only (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.
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
with the development of asbestosis and fatal lung cancer (Srebro and Roggli, 1994). Although
these case reports do not provide quantitative exposure measures, they do illustrate the potential
for demonstrable health effects from exposures of short duration and from exposures in a
community setting.
4.1.2.1.1.5. Summary of respiratory effects, other than cancer.
Epidemiology studies demonstrate consistent results pertaining to the association between LAA
exposure and various adverse forms of respiratory effects, with effects seen in both
4-36
-------
occupationally exposed worker populations and in community populations with nonoccupational
exposure. The risk of mortality related to asbestosis and other forms of nonmalignant respiratory
disease is elevated in the Libby vermiculite mining and processing operations workers, with a
pattern of increasing risk with increasing cumulative exposure (more than a 10-fold increased
risk of asbestosis and a 1.5- to 3-fold increased risk of nonmalignant respiratory disease) in the
analyses using internal, referent groups in McDonald et al. (2004), Sullivan (2007), and Larson
et al. (201 Ob). Radiographic evidence of small opacities (evidence of parenchymal damage) and
pleural thickening has also been shown in studies of Libby workers (Larson et al., 2012a; Larson
etal., 2010a: Whitehouse, 2004; Amandus et al., 1987a: McDonald et al., 1986b), and in the
studies of workers in the Marysville, OH plant (Rohs et al., 2008; Lockey et al., 1984). In the
Marysville cohort, the prevalence of small opacities (interstitial changes in the lung) increased
from 0.2% in the original study to 2.9% in the follow-up study, and the prevalence of pleural
thickening increased from 2 to 28.6%. No effects on lung function were found in the original
study (Lockey etal., 1984), and lung function was not reported for the Rohs et al. (2008)
analysis of the cohort follow-up. Data from the ATSDR community health screening study in
Libby, MT indicate that the prevalence of pleural abnormalities, identified by radiographic
examination, increases substantially with increasing number of exposure pathways (Peipins et
al., 2003). The presence of pleural plaques is associated with a small decrement in lung function
(approximately 5%) when evaluated based on mean values (Weill etal., 2011), and is associated
with an increased risk of restrictive lung function (Larson et al., 2012b). Stronger associations
are seen with the presence of DPT in both of these studies. Additional evidence of respiratory
effects of LAA exposure comes from the study of residents in an area surrounding a processing
plant in Minneapolis, MN (Alexander et al., 2012).
4.1.3. Other Effects, Noncancer
4.1.3.1. Cardiovascular Disease
Larson et al. (201 Ob) present data on mortality due to cardiovascular diseases (CVDs)
among the Libby cohort of vermiculite workers, with SMRs of 0.9 (95% CI: 0.9, 1.0) for heart
disease (n = 552) and 1.4 (95% CI: 1.2, 1.6) for circulatory system diseases (n = 258). Similar
results were seen in the analysis by Sullivan (2007) of this cohort, with an SMR for heart disease
of 0.9 (95% CI 0.8-1.1), and SMR for diseases of the circulatory system (specifically, diseases
of circulatory diseases involving the arteries, veins, and lymphatic vessels) of 1.8 (95% CI,
1.2-2.6). In the study by Larson et al. (201 Ob), deaths due to heart diseases were further
categorized into ischemic heart disease (n = 247) and other heart disease (n = 120, for
pericarditis, endocarditis, heart failure, and ill-defined descriptions and complications of heart
disease). The SMR for ischemic heart disease was 0.7 (95% CI: 0.6, 0.8), and the SMR for
other heart disease was 1.5 (95% CI: 1.2, 1.8). Circulatory diseases included hypertension
without heart disease (n = 42), with an SMR of 1.7 (95% CI: 1.2, 2.4) and diseases of arteries,
4-37
-------
veins, or lymphatic vessels (n = 136), SMR =1.6 (95% CI: 1.4, 2.0). The combined category of
cardiovascular mortality resulted in modestly increased risks across quartiles of exposure, with
RRs of 1.0 (referent), 1.3 (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 <1.4, 1.4 to <8.6, 8.6 to <44.0, and >44.0 fibers/cc-yr, respectively. In
the Monte Carlo simulation used to estimate the potential bias in cardiovascular disease risk that
could have been introduced by differences in smoking patterns between exposed and unexposed
workers in the cohort, the bias adjustment factor was relatively small
(RRunadjusted/RRadjusted = 1.1), reducing the overall RR estimate from 1.6 to 1.5. The observed
association between asbestos exposure and cardiovascular disease mortality may reflect, at least
in part, a consequence of an underlying respiratory disease.
4.1.3.2. Autoimmune Disease and Autoantibodies
Three epidemiology studies have examined the potential role of LAA and autoimmunity.
Noonan et al. (2006) used the data from the community health screening to examine
self-reported histories of autoimmune diseases (rheumatoid arthritis, scleroderma, or lupus) in
relation to the asbestos exposure pathways described above (see Tables 4-5 and 4-15). To
provide more specificity in the self-reported history of these diseases, a follow-up questionnaire
was mailed to participants to confirm the initial report and obtain clarifying information
regarding the type of disease, whether the condition had been diagnosed by a physician, and
whether the participant was currently taking medication for the disease. Responses were
obtained from 208 (42%) of the 494 individuals who had reported these conditions. Of these
208 responses, 129 repeated the initial report of the diagnosis of rheumatoid arthritis, and
161 repeated the initial report of the diagnosis of one of the three diseases (rheumatoid arthritis,
scleroderma, or lupus); approximately 70% of those confirming the diagnosis also reported
taking medication for the condition. Among people aged 65 and over (n = 34 rheumatoid
arthritis cases, determined using responses from the follow-up questionnaire), a twofold to
threefold increase in risk was observed in association with several measures reflecting potential
exposure to asbestos (e.g., asbestos exposure in the military) or specifically to LAA (e.g., past
work in mining and milling operations, use of vermiculite in gardening, and frequent playing on
vermiculite piles when young). Restricted forced vital capacity (defined as FVC <80% predicted
and a ratio of FEVi to FVC >70% predicted), presence of parenchymal abnormalities, playing
on vermiculite piles, and other dust or vermiculite exposures were also associated with
rheumatoid arthritis in the group younger than 65 years (n = 95 cases). For all participants, an
increased risk of rheumatoid arthritis was observed with increasing number of exposure
pathways. Although the information gathered in the follow-up questionnaire and repeated
reports of certain diagnoses decreased the false-positive reports of disease, the reliance of
self-reported data is a limitation of this study. Considerable misclassification (over-reporting
and under-reporting) is likely, given the relatively low confirmation rate of self-reports of
4-38
-------
physician-diagnosed rheumatoid arthritis (and other autoimmune diseases) seen in other studies
(Karlson et al.. 2003: Raschetal.. 2003: Ling et al.. 2000).
Another study examined serological measures of autoantibodies in 50 residents of Libby,
MT, and a comparison group of residents of Missoula, MT [(Pfau et al., 2005) see Table 4-15]
The Libby residents were recruited for a study of genetic susceptibility to asbestos-related lung
disease, and the Missoula residents were participants in a study of immune function. None of the
50 Missoula residents and three of the Libby participants reported a history of a rheumatoid
arthritis, systemic lupus erythematosus, or other systemic autoimmune disease (SAID). Libby
residents exhibited an increased prevalence (22%) of high-titer (>1:320) antinuclear antibodies
when compared to Missoula residents (6%), and similar increases were seen in the Libby
samples for rheumatoid factor, antiribonucleoprotein (RNP), anti-Scl-60, anti-Sm, anti-Ro
(SSA), and anti-La (SSB) antibodies. Although neither sampling approach was based on a
random selection from the community residents, an individual's interest in participating in a gene
and lung-disease study would not likely be influenced by the presence of autoimmune disease or
autoantibodies in that individual. Thus selection bias would not be considered likely in this
study.
In a follow-up study, Marchand et al. (2012) examined the association of autoantibodies
with asbestos-related lung disease in 124 Libby residents (65 female, 59 male). Serum samples
were tested for the presence of antimesothelial cell antibodies (MCAA) to determine if the
mesothelial cells of the pleural lining are targets for autoimmune responses. Mean
concentrations of MCAA were increased in Libby residents, particularly in those with lung or
pleural lesions compared to a reference population of Missoula residents. In addition, Libby
residents positive for antinuclear antibodies or MCAA had an increased odds ratio of also
presenting with pleural abnormalities (odds ratio 3.60 and 4.88, respectively). However, the
cross-sectional nature of this study design makes it difficult to determine if these autoantibodies
were a principal mechanism for inducing pleural disease, or if their presence is an indication of
tissue damage.
4-39
-------
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 serf-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.
Outcome: self-reported cases of physician-
diagnosed rheumatoid arthritis, lupus, or
scleroderma
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)
(trendy <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.
Outcome: serum samples obtained for
measurement of IgA levels and autoantibody
prevalence (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 + 3 SD of Missoula samples.
Outcome: 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:
Pleura! abnormalities 25%
Interstitial abnormalities only 52%
No abnormalities 23%
Association with pulmonary abnormality:
ANA
MCAA
76
23
(61.3)
(18.5)
OR
5.6
ANA = antinuclear antibody; dsDNA = double-stranded DNA; MCAA = antimesothelial cell autoantibodies;
RNP = ribonucleoprotein.
4-40
-------
4.1.4. Cancer Effects
4.1.4.1. Lung Cancer
Several analyses of the mortality experience of Libby vermiculite workers have been
conducted (see Section 4.1.1 for a summary of exposure measures and cohort descriptions for
these studies). The studies of the Libby worker cohort by Amandus and Wheeler (1987),
Sullivan (2007), and Larson et al. (201 Ob) defined lung cancer mortality based on a more
specific cause of death codes (e.g., cancers of the trachea, bronchus, and lung) compared to the
broader classification of "all respiratory cancer" used by McDonald et al. (2004) and McDonald
et al. (1986a), which would include laryngeal and "other" respiratory cancers. In the national
Surveillance, Epidemiology, and End Results cancer data from 2003-2007, the age-adjusted
mortality rate for cancer of the larynx was 1.2 per 100,000 person-year, compared to 52.5 per
100,000 person-year for lung and bronchial cancer (NCI, 2011). Thus, these additional
categories (larynx and "other" respiratory cancers) represent a relatively small proportion of
respiratory cancers. Although they could also be a source of some misclassification of the
outcome if these other cancers are not related to asbestos exposure, the magnitude of this bias
would be small.
In the more recent study by McDonald et al. (2004), 44 respiratory cancers were observed
among 406 men who had worked at least 1 year in the vermiculite mining and milling facilities.
Sullivan (2007) and Larson et al. (201 Ob) included workers with less than 1 year of work,
resulting in a larger sample size (approximately 1,700) and more than 80 lung cancer deaths.
Each of these studies observed an increased overall risk, with SMRs of 1.4, 1.6, and 2.4,
respectively in Sullivan (2007), Larson et al. (201 Ob), and McDonald et al. (2004).
Exposure-response analyses from these studies demonstrated increasing mortality with
increasing exposure, using categorical and continuous measures of exposure, different lag
periods, and different exposure metrics, with approximately a twofold to threefold increased risk
in the highest exposure group (see Table 4-16 and Figure 4-3). Larson etal. (201 Ob) uses a
referent category of 0 to <1.4 fibers/cc-yrs for all analyses (based on the distribution among all
deaths); this is a lower level compared with the other studies. In addition, as described in
Section 4.1.1.1.2, the cumulative exposure distribution in (Larson et al., 201 Ob) is lower than in
(Sullivan, 2007).
4-41
-------
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 (1987)
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
Meanfiber-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 = U)
SMR: 2.3 (p O.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 0.11 (0.04). All estimates are statistically
significant (p <0.05).
4-42
-------
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.
(2004) McDonald
etal. (1986a)
Sullivan (2007)
Inclusion criteria and design details
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 July 1983-1998
coded by nosologists using ICD-8
classifications; cause of death for deaths
trom iyoj tyys outaineu trom 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 (2.2% loss to
follow-up) (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 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
5
9
10
16
RR (95% CI)d
1.0 (referent)
1.7 (0.58, 5.2)
1.9 (0.63, 5.5)
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
<1 yr
1-9.9 yr
n
41
34
SMR (95% CI)b
1.6(1.1,2.1)
1
.7(1.1,2.3)
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)
(/?<0.001)
SRR (95% CI)C
1.0 (referent)
1.1(0.7, 1.8)
4-43
-------
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
(10% loss to follow-up); 952 deaths (87%
with known cause of death).
Median duration: 0.8 yr
Median fibers/cc-yr =4.3
Immediate and underlying causes of
death data from death certificates or
National Death Index-Plus
(In a follow-up of the Sullivan (2007)
cohort through 2006, 1,009 deaths were
identified, indicating an
underascertainment of approximately
5.6% of the deaths by Larson et al.
(2010b); this difference could be related
to the higher loss to follow-up in this
study).
Standardized mortality
ratio (95% CI)
Lung (n = 104)
SMR: 1.6(1.3,
2.0)
Exposure-response analyses — lung cancer
>10yr 14 2.5(1.4,4.3) 1.8(0.9,3.4)
Linear trend test p O.05
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 et al. (2010bX 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.
4-44
-------
0>
+••
re
"iC
(A
UJ
^C
(A
'on
0)
_>
"•8
to
0}
5 -
1 -
o McDonald etal., 2004
RR: 10 year latency
Referent: 0.0-11.6 fibers/cc-yrs
[n = 44]
• Sullivan 2007
SRR: 15 year latency
Referent: 0.0-4.49 fibers/cc-yrs
In = 89]
O Larson etal., 201 Ob
RR: 20 year latency
Referent: 0.0-<1.4 fibers/cc-yrs
[n = 104]
Cumulative Exposure
(fibers/cc-years)
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.
Two of these studies included data addressing the question of the extent to which the
results could be confounded by smoking (Larson etal., 201 Ob: Amandus and Wheeler, 1987).
Amandus and Wheeler (1987) provide some information on the smoking history of a sample of
161 male workers employed during 1975-1982 with at least 5 years of employment in the Libby
cohort study and comparison data based on surveys conducted in the United States from
1955-1978. Among the workers, 35% were current smokers and 49% were former smokers.
This smoking information was obtained from questionnaires the company administered to
workers after 1975 (Amandus et al., 1987a). The prevalence of current smokers was similar in
the worker cohort compared to the U.S. white male population data (ranging from 37.5-41.9%
current smokers between 1975 and 1978). The only year in this range with data on former
smokers in the national survey is 1975, and at that time, the prevalence of former smokers in the
population data was 29.2%, about 20% lower than among the workers. Using an estimated RR
of lung cancer of 14 among smokers, Amandus and Wheeler (1987) estimated that the difference
in smoking rates between workers and the comparison population could have resulted in a 23%
increase in the observed risk ratio and commented that the increased risk observed in the lower
dose range (<50 fiber-year) could be the result of confounding by smoking status.
Smoking patterns in the U.S. population changed considerably over the period
corresponding to the data reported by Amandus and Wheeler (1987). W. R. Grace and Co.
instituted a smoking ban on the property in 1979 (Peacock, 2003), which may have affected
4-45
-------
smoking habits, reporting of smoking habits, or both. In the National Health Interview Surveys
conducted between 1974 and 1983, the prevalence of smoking in males age 20 and older
decreased from 42.1 to 35.5% (HHS, 1990). Based on 1986 survey data, the percentage of adults
age 17 and older classified as former smokers varied between 14.7% and 25.8% using different
definitions for time since last smoked [e.g., from quitting 5 or more years ago to quitting within
the past 3 months; (HHS, 1990)]. Thus, given the lack of information pertaining to the period in
which smoking information was collected and the specifics of the sources that were used, EPA
concludes there is considerable uncertainty regarding the evidence for differences in smoking
rates between the workers and the external comparison population.
Larson et al. (201 Ob) used data from the ATSDR community health screening in Libby
(described in Section 4.1.1.3) pertaining to smoking history to estimate that the proportion of
smokers ranged from 50 to 66% in the unexposed group (defined as exposure <8.6 fibers/cc-yr)
and between 66 and 85% among the exposed (defined as >8.6 fibers/cc-yr). Larson et al.
(201 Ob) used these estimates in a Monte Carlo simulation to estimate the potential bias in lung
cancer risks that could have been introduced by differences in smoking patterns. The bias
adjustment factor (RRunadjusted/RRadjusted =1.3) reduced the overall RR estimate for lung cancer
from 2.4 to 2.0.
O.M. Scott, Marysville, OH plant workers
There was no evidence of an increased risk of lung cancer in the analysis of mortality
among the 465 Marysville, OH plant workers (Dunning et al., 2012). The SMR was 0.9 (95%
CI: 0.5-1.5), based on 16 observed lung cancer deaths, and there was no indication of an
increased risk in analyses stratifying by tertiles of cumulative exposure (SMRs varying between
0.8 and 1.0, and standardized rate ratios (SRRs) varying between 0.9 and 1.0).
Geographic mortality analysis
In the geographic mortality analysis (1979-1998) conducted by ATSDR (2000). the
SMR for lung cancer ranged from 0.9-1.1 and 0.8-1.0 for each of the six geographic boundaries
using Montana and U.S. reference rates, respectively. These analyses did not distinguish
between deaths among workers and deaths among other community members.
4.1.4.2. Mesothelioma
Prior to the 10th revision of the ICD, which was implemented in the United States in
1999, there was no unique ICD code for mesothelioma. The updated NIOSH study by Sullivan
(2007) identified 15 deaths for which mesothelioma was mentioned on the death certificate.
Only two deaths occurring between 1999 and 2001 were coded to mesothelioma under ICD-10
(Code C45). Larson et al. (201 Ob) classified all death certificates listing mesothelioma as
ICD-10 code C45. The updated McGill study [(McDonald et al.. 2004): with analysis through
4-46
-------
1998] noted that the classification of mesothelioma was based on a nosologist's review of death
certificates; 5 of the 12 cases classified as mesothelioma had a cause of death listed as pleural
cancer (ICD-9 code 163).
Data pertaining to mesothelioma risk from the available occupational studies are
summarized in Table 4-17. McDonald et al. (2004) presented dose-response modeling using
Poisson regression of mesothelioma risk based on 12 cases. Note that the referent group was
also at excess risk of dying from mesothelioma; that is, one to three cases of mesothelioma were
observed in the referent group, depending on the exposure index. Three exposure indices were
used in the analysis: average intensity over the first 5 years of employment, cumulative
exposure, and residence-weighted cumulative exposure. Because of the requirement for 5 years
of employment data, 199 individuals (including three mesothelioma cases) were excluded from
the analysis of average intensity. The residence-weighted cumulative exposure was based on the
summation of exposure by year, weighted by years since the exposure. This metric gives greater
weight to exposures that occurred a longer time ago. Although evidence of an excess risk of
dying from mesothelioma was seen in all groups, only the residence-weighted cumulative
exposure metric exhibited a monotonically increasing pattern, with an RR of 1.57 among those
with 500.1-1,826.8 fibers/cc-yr exposure, and an RR of 1.95 among workers with higher
residence-weighted cumulative exposure. In the study by Sullivan (2007), which identified
15 deaths from mesothelioma through a manual review of death certificates, the SMR for
mesothelioma was 14.1 (95% CI: 1.8, 54.4), based on the two mesothelioma deaths occurring
between 1999 and 2001, the period for which comparison data using the ICD-10 classification
criteria were available.
4-47
-------
Table 4-17. Mesothelioma mortality risk based on studies of the vermiculite
mine workers in Libby, MTa
Reference(s)
Amandus and
Wheeler (1987)
McDonald et al.
(20041
McDonald et al.
(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 et al. 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
0.0-1 1.6 fibers/cc-yr
1 1.7-25. 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
1 1.7-25. 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 (<0, 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)
4-48
-------
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 (2.2%
loss to follow-up) (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
(10% loss to follow-up); 952 deaths
(87% with known cause of death).
Median duration: 0.8 yr
Median fibers/cc-yr = 4.3
Immediate and underlying causes of
death data 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
Pleura! (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)
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.
A more descriptive presentation of a collection of mesothelioma cases was reported by
Whitehouse et al. (2008). This report reviewed 11 cases of mesothelioma diagnosed between
1993 and 2006 in residents in or around Libby, MT (n = 9) and in family members of workers in
the mining operations (n = 2). Three cases were men who might have had occupational asbestos
exposure through construction work (Case 1), working in the U.S. Coast Guard and as a
carpenter (Case 5), or through railroad work involving sealing railcars in Libby (Case 7). One
case was a woman whose father had worked at the mine for 2 years; although the family lived
100 miles east of Libby, her exposure may have come through her work doing the family
laundry, which included laundering her father's work clothes. The other seven cases (four
women, three men) had lived or worked in Libby for 6-54 years and had no known occupational
or family-related exposure to asbestos. Medical records were obtained for all 11 patients;
4-49
-------
pathology reports were obtained for 10 of the 11 patients. The Centers for Disease Control and
Prevention estimated the death rate from mesothelioma, using 1999 to 2005 data, as
approximately 23.2 per million per year in males and 5.1 per million per year in females (CDC,
2009). The rate in females is more likely to reflect a population that is not occupationally
exposed to asbestos, and is closer to the background incidence of mesothelioma estimated by as
about 1 per million in a population with little or no asbestos exposure (Hillerdal, 1983).
Although not calculated by Whitehouse et al. (2008), the observation of seven cases without
known personal or familial occupational exposure, over a 15-year observation period and an
estimated Libby area (Lincoln Country) population of 9,500 (142,000 person-year) would result
in an incidence rate of approximately 49 per million per year. Whitehouse et al. (2008) stated
that a W.R. Grace unpublished report of measures taken in 1975 indicated that exposure levels of
1.1 fibers/cc were found in Libby, and 1.5 fibers/cc were found near the mill and railroad
facilities. Because the mining and milling operations continued to 1990, and because of the
expected latency period for mesothelioma, Whitehouse et al. (2008) suggested that additional
cases can be expected to occur within this population, as well as in transitory workers and in
workers who had left the area.
4.1.4.2.1. O.M. Scott, Marysville, OH plant workers. In the analysis of mortality through
June 30, 2011 among the Marysville, OH plant workers, 2 of the 465 workers died of
mesothelioma compared to an expected 0.2 cases (SMR 10.5, 95% CI 1.3, 38) (Dunning et al.,
2012). The cumulative exposure for each of these two cases was approximately 45 fibers/cc-yr.
One other incident mesothelioma case (diagnosed in 2010 and alive through the defined
mortality follow-up period in the study) was identified in the cohort. This case is not included in
the mortality analysis, but the cumulative exposure of the individual was 5.73 fibers/cc-yr.
4.1.4.3. Other Cancers
Larson et al. (201 Ob) presented data on cancers other than respiratory tract and
mesothelioma. The category of malignant neoplasms of digestive organs and peritoneum
included 39 observed deaths, for an SMR of 0.8 (95% CI: 0.6, 1.1). No risk in relation to
asbestos exposure was seen with a 20-year lag.
4.1.4.4. Summary of Cancer Mortality Risk in Populations Exposed to Libby Amphibole
Asbestos
The studies conducted in the 1980s (Amandus and Wheeler, 1987; McDonald et al.,
1986a) as well as the extended follow-up studies published in more recent years (Larson et al.,
201 Ob: Sullivan, 2007; McDonald et al., 2004) provide consistent evidence of an increased risk
of lung cancer mortality and of mesothelioma mortality among the workers in the Libby
vermiculite mining and processing operations. The lung cancer analyses using an internal
4-50
-------
referent group in the larger follow-up studies (Larson etal., 201 Ob: Sullivan, 2007; McDonald et
al., 2004) observed increasing risks with increasing cumulative exposure when analyzed using
quartiles or as a continuous measure. Increased risks are also seen in the studies reporting
analyses using an external referent group [i.e., standardized mortality ratios (Sullivan, 2007;
Amandus and Wheeler, 1987; McDonald et al., 1986a)1. Although an increased lung cancer risk
was not observed in the Marysville, OH plant workers, three cases of mesothelioma (two of
whom were included in the mortality analysis) have been identified as of June, 2011. These
observations further support the identification of cancer (specifically, lung cancer and
mesothelioma) as a hazard of LAA.
4.1.5. Comparison With Other Asbestos Studies—Environmental Exposure Settings
The literature pertaining to risks of asbestos is extensive; of particular interest is the set of
studies examining environmental exposures to constituents of LAA (e.g., tremolite) or other
amphiboles. This literature provides findings consistent with those identified for LAA.
Several communities have been exposed in environmental or residential settings to
tremolite or tremolite-chrysotile mixtures from natural soils and outcroppings as well as
construction materials found in the home (see Table 4-18). Studies on these affected populations
(published as early as 1979) reported an increased risk of pleural and peritoneal malignant
mesothelioma (Sichletidis et al.. 1992b: Barisetal.. 1987: Langeretal.. 1987: Barisetal.. 1979).
Clinical observations include a bilateral increase in pleural calcification accompanied by
restrictive lung function decrements as the disease progresses, a condition known as "Metsovo
lung," named after a town in Greece where this abnormality was observed among
tremolite-exposed residents (Constantopoulos et al., 1985). These health effects are consistent
with the health effects documented for workers exposed to commercial forms of asbestos.
4-51
-------
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
(Eshisehir district)
-2,000
-5,000
-4,000
New Caledonia
-40,000
Fiber: tremolite,
tremolite/chrysotile mixtures
Exposure: indoor
0.089 fiber/mL; outdoor
0.013 fiber/mL
Size: not available
Fiber: tremolite
Exposure: Variable (1 to
>200 fibers/mL)
Size: length <10 um, diameter
0.2 um
Fiber: tremolite, chrysotile
Exposure: indoors
0.01-17.9 fibers/cc
Size: not available
Fiber: tremolite
Exposure: not available
Size: not available
Mesothelioma
Men
Women
SIR
SIR
53
144
Pleural plaques prevalence -14%
Diffuse pleura! thickening prevalence
-10%
Mesothelioma SIR -280
Pleural plaques prevalence -45%
Mesothelioma four incident cases among
198 people with pleural plaques over
15-yr follow-up period.
Pleural plaques prevalence -24% among
people over age 40 yrs, with increasing
prevalence with increasing age.
Longitudinal study (1988-2003) also
observed increase prevalence and extent
(surface area) of plaques.
Mesothelioma SMR 41
Lung cancer
Men
Women
SMR
SMR
-1.0
2.4
Metintas et al.
(2005) Metintas et
al. (2002)
Yazicioslu et al.
(1980) Yazicioslu
(1976)
Barisetal. (1979)
Constantopoulos et
al. (1987)
Constantopoulos et
al. (1985)
Bazas etal. (1985)
Sichletidis et al.
(2006) Sichletidis
etal. (1992a)
Sichletidis et al.
(1992b)
Luce et al. (2000)
4-52
-------
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 fibers/mL-yr, 5.5%
>20 fibers/mL-yr
Size: length >5 um
2,500 women follow-up through 2004:
(»)
Mesothelioma (30)
Lung cancer (30)
Ovarian cancer (9)
Pneumoconiosis (2)
SMR
Not reported
-1.9
-1.4
-11
2,460 people initially exposed age <15 yr:
Cancer type
Mesothelioma
Lung
Brain
Leukemia
Ovary
(«)
Women
(13)-
90
(5) -2.0
(4) -3.6
(4) -3.0
(6) -3.3
SIR
Men
(29) - 60
(3) -1.0
(5) -3. 4
(7) -4.2
-
(SMRs and SIRs based on means of two
different censoring methods)
Reid etal. (2013)
Reid et al. (2007)
Hansenetal. (1998)
Hansenetal. (1997)
Hansenetal. (1993)
SIR = standardized incidence ratio; SMR = standardized mortality ratio.
Although it is not a constituent of LAA, crocidolite is another type of amphibole asbestos
that has been studied with respect to health effects arising from environmental exposures.
Several studies have examined cancer risk and pneumoconiosis risk among nonworker residents
of Wittenoom, Australia, an area surrounding a crocidolite asbestos mine and mill [(Reid et al.,
2007; Hansen et al., 1998) see Table 4-18]. Increased risk of mesothelioma and pneumoconiosis
and more modestly increased risk of lung cancer were reported in these studies.
4.2. SUBCHRONIC- AND CHONIC-DURATION STUDIES AND CANCER
BIOASSAYS IN ANIMALS—ORAL, INHALATION, AND OTHER ROUTES OF
EXPOSURE
Laboratory animal studies of exposure to Libby Amphibole or tremolite asbestos show
effects similar to those observed in occupationally exposed human populations, including pleural
pathology, mesothelioma, and lung cancer. Tremolite is an amphibole asbestos fiber that is a
component of LAA (-6%). Also, in early studies, LAA was defined as tremolite. Therefore,
laboratory animal studies examining the effect of tremolite exposure have been reviewed and are
4-53
-------
summarized below to potentially increase understanding of the effects and mechanisms of LAA.
Detailed study summaries can be found in Appendix D and summarized in Tables 4-19 and 4-20.
As noted in Section 3, the primary route of human exposure is inhalation. Thus, studies that
expose animals through a pulmonary route are the most relevant for hazard identification. No
inhalation studies have been performed for LAA, but chronic intrapleural injection studies in
hamsters demonstrate carcinogenicity following exposure. The chronic inhalation and
intrapleural injection laboratory animal studies with tremolite asbestos demonstrated pleural
pathology and carcinogenicity in rats. These studies support the epidemiology studies of LAA
exposure (see Section 4.1) and aid in informing the mechanisms of LAA-induced disease.
4-54
-------
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
Effects3
Pleural 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)
Smarttet
al. (2010)
Shannahan
etal.
(201 la)
4-55
-------
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 ugFeCls; 0.5 mg
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. (20 11)
4-56
-------
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)
4-57
-------
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 ugFeCls; 0.5 mg
LA, 0.5 mg FeLA;
0.5 mgLA+ 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)
4-58
-------
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)
4-59
-------
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
um
Mean fiber
diameter
0.29 ±0.19
um
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 fluid; 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.
4-60
-------
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 um
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
pleural sarcomas
Sample 2: 22/28
pleural sarcomas
Reference
McConnell
etal.
(1983a)
Davis et al.
(19851
Davis et al.
(1991)
Smith et al.
(19791
Wasner et
al. (1982)
Stanton et
al. (1981)
4-61
-------
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
1x15 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)
Roller et al.
(19961
Bernstein et
al. (2005b)
Bernstein et
al. (2003)
Pfau et al.
(2008)
BW = body weight; PBS = phosphate buffer saline.
aWhen 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.
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., 2005b; Bernstein et al., 2003; Davis etal., 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. (2005b) 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
4-62
-------
observed in the controls. These results show that Wistar rats exposed to tremolite exhibited
increased numbers of pulmonary lesions and possibly tumors.
4.2.2. Intratracheal Instillation Studies
Intratracheal instillation has been used to examine the effect of exposure to Libby
Amphibole (Cvphert et al.. 2012b: Cyphert etal.. 2012a: Shannahanet al.. 2012a: Shannahan et
al.. 2012c: Shannahan etal.. 2012b: Shannahan etal.. 2012d: Padilla-Carlin etal.. 2011:
Shannahan etal.. 2011 a: Shannahan et al.. 20 lib: Smarttet al.. 2010: Putnam et al.. 2008) and
tremolite asbestos (Blake et al.. 2008: Pfau et al.. 2008: Sahuetal.. 1975). These studies
exposed C57B1/6 mice (100 ug/mouse), Wistar-Kyoto (WKY) rats (0.25 or 1 mg/rat) or Fischer
344 rats (0.25 or 6.5 mg/rat) once to LAA and analyzed the results up to 2 years postexposure.
Putnam et al. (2008) observed statistically nonsignificant increases in collagen following
exposure to LAA, as well as gene expression alterations related to membrane transport, signal
transduction, epidermal growth factor signaling, and calcium regulation. Smarttet al. (2010)
followed up this study by analyzing specific genes by quantitative reverse transcription
polymerase chain reaction for genes involved in collagen accumulation and scar formation
(CollAl, CollA2, ColSAl). LAA exposure led to increased gene expression of CollA2 at
1 week postinstillation and Col3 Al at 1 month postexposure. Both studies observed increased
inflammation; however, LAA exposure demonstrated minimal inflammation that did not
progress in the time points examined. These studies demonstrate that exposure to LAA may lead
to inflammation and fibrosis.
Shannahan et al. (201 la) exposed two rat models of human cardiovascular disease to
LAA to determine whether the preexisting CVD in these models would impact the lung injury
and inflammation following exposure. Healthy WKY rats were compared to spontaneously
hypertensive (SH) and spontaneously hypertensive-heart failure (SHHF) rats following exposure.
All rats (male only) were exposed to 0, 0.25, or 1.0 mg/rat via intratracheal instillation and were
examined at 1 day, 1 week, and 1 month postexposure. No changes were observed
histopathologically; however, changes were observed in markers of homeostasis, inflammation,
and oxidative stress. While inflammation and cell injury were observed in all strains, no
strain-related differences were observed following exposure to LAA (Shannahan et al., 201 la).
A series of studies further examined SH, SHHF, and WKY rats over several durations of
exposure to identify potential biomarkers of LAA exposure and determine if asbestos exposure
shifts biomarker expression in healthy rats to resemble CVD (Shannahan et al., 2012a:
Shannahan et al., 2012d). Acute-phase response molecules involved in inflammatory responses
such as a2-macroglobulin and al-acid glycoprotein, as well as the metabolic molecule
lipocalin-2 were generally increased 1 day after exposure regardless of duration (Shannahan et
al., 2012a). In addition, LAA generally did not change biomarker expression similarly to the
CVD rat strains (Shannahan et al., 2012b). However, the expression of two vasoconstriction
4-63
-------
genes, eNOS and ETR-A, were altered in Libby Amphibole-exposed WKY rats to resemble
untreated SH and SHHF rats (Shannahan et al., 2012b). Biomarkers for cancer were largely
unaffected in all three strains following LAA exposure (Shannahan et al., 2012a).
In a follow-up study to further examine the role of iron in the inflammatory response to
LAA exposure, Shannahan et al. (201 Ib) exposed SH rats to LAA alone and with bound Fe as
well as with an iron chelator (deferoxamine [DEF]). Exposure to LAA led to significant
increases in inflammatory markers (e.g., neutrophils, interleukin [ILJ-8) with the greatest
increase occurring in the presence of DEF. Iron bound to LAA was not released following
instillation except in the presence of DEF as supported by the lack of increase of iron in
bronchoalveolar lavage fluid (BALF). These results suggest that chelation of iron bound to LAA
as well as endogenous proteins increases the toxicity of LAA in vivo.
A pair of studies further examined the effect of iron in the context of Libby
Amphibole-induced lung injury and inflammasome activation. DEF treatment in addition to
LAA significantly affected cyclooxygenase-2 (COX-2), IL-6, and CCL-11 in lung tissue
compared to LAA treatment alone (Shannahan et al., 2012c). Addition of iron to LAA
significantly altered NF-kp and IL-lp compared to LAA alone (Shannahan et al., 2012c).
However, iron overload and DEF treatment generally were not significantly changed from each
other, suggesting that iron has little impact on the inflammasome cascade. Histological
examination and gene array analysis of inflammatory genes in WKY, SH, and SHHF rats did not
identify significant differences in the progression of pulmonary fibrosis between the three strains
(Shannahan et al., 2012d). These data do not indicate that the iron overload conditions that are
characteristic of the cardiovascular disease-rat strains amplify the pulmonary effects of LAA.
Padilla-Carlin et al. (2011) exposed Fischer 344 rats (male only) to LAA (0.65 or 6.5 mg/rat) or
amosite (0.65 mg/rat; positive control) by intratracheal instillation to examine inflammatory
response for 3 months postexposure. LAA exposure led to statistically significant increases of
neutrophils in BALF as early as 1 day postexposure, with other inflammatory markers (e.g.,
protein, lactate dehydrogenase [LDH], gamma-glutamyl transpeptidase [GOT]) increased
statistically significantly at different time points during the 3-month-period postexposure.
However, on a mass basis, amosite produced a greater inflammatory response as measured by
inflammatory markers (e.g., neutrophil influx, gene expression changes) and histopathological
analysis demonstrating interstitial fibrosis. Examination of male Fischer 344 rats from this study
at 2 years postexposure demonstrated that LAA induced a significant fibrogenic (but not
carcinogenic) effect (Cyphert et al., 2012b). Response to LAA exposure in this study was less
potent than amosite on a mass basis. Further comparison of LAA to other fiber types (Sumas
Mountain chrysotile [SM], El Dorado tremolite [ED], and Ontario ferroactinolite [ON])
demonstrated that LAA exposure increased inflammatory markers at 1 day postexposure which
returned to control levels by 3 months (Cyphert et al., 2012a). LAA exposure also led to an
increased fibrogenic response at 3 months postexposure. As compared to other fibers tested,
4-64
-------
fibrogenic response was correlated with the fiber length and width of each fiber, with
SM-exposed rats demonstrating the greatest fibrogenic response (SM > LAA > ON > ED).
These studies demonstrate a statistically significant increase in inflammatory response to LAA in
mice and rats as measured in BALF by cytology, histopathology, and gene expression analysis.
Follow-up studies are needed to inform the chronic effects of exposure to LAA.
Laboratory animal studies of tremolite intratracheal instillation exposure have been
performed in mice in doses ranging from 60 ug to 5 mg. Male Swiss albino mice exposed to
tremolite (5 mg) via intratracheal instillation demonstrated histological changes (Sahu et al.,
1975). Microscopic results following exposure to tremolite showed acute inflammation of the
lungs at 7 days postexposure, including macrophage proliferation and phagocytosis similar to
that observed with amosite and anthophyllite. Limited progression of fibrotic response was
observed at 60 and 90 days postexposure, with no further progression of fibrotic response.
4.2.3. Injection/Implantation Studies
There are no laboratory animal studies examining intraperitoneal injection or
implantation of LAA. Biological effects following exposure to tremolite have been examined in
five intraperitoneal injection studies (Roller et al., 1997, 1996; Davis et al., 1991; Wagner et al.,
1982: Smith etal., 1979: Smith, 1978) and one implantation study (Stanton et al., 1981).
Studies by Smith and colleagues (Smith etal., 1979: Smith, 1978), Wagner etal. (1982),
Davis etal. (1991), Roller etal. (1997), and Roller etal. (1996) demonstrated that intrapleural
injections of tremolite asbestos16 is associated with an increase in pleural fibrosis and
mesothelioma in hamsters and rats compared to controls or animals injected with less fibrous
materials. Doses ranged from 10-25 mg/animal for each study, and although carcinogenesis was
observed in these studies, there was a variable level of response to the different tremolite forms
examined. Although these studies clearly show the carcinogenic potential of Libby Amphibole
or tremolite asbestos fibers, intrapleural injections bypass the clearance and dissolution of fibers
from the lung after inhalation exposures. Further, limited information was provided confirming
the presence or absence of particles or fibers less than 5 um in length in these studies, limiting
the interpretation of results.
One laboratory animal study examined the effect of tremolite exposure following
implantation of fibers in the pleural cavity. Stanton et al. (1981) examined tremolite and
described a series of studies on various forms of asbestos. Fibers embedded in hardened gelatin
were placed against the lung pleura. As an intrapleural exposure, results might not be
comparable to inhalation exposures, as the dynamics of fiber deposition and pulmonary
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).
4-65
-------
asbestos samples from the same lot were described as being in the optimal size range for
carcinogenesis; the fibers were distinctly smaller in diameter than the tremolite fibers Smith et al.
(1979) used. These samples both had a high number of fibers in the size range (>8-um long and
<0.25 um in diameter; i.e., "Stanton fibers"). Exposure to both tremolite samples led to
mesotheliomas in 21 and 22 of 28 rats exposed. The Stanton et al. (1981) study also used talc,
which did not lead to mesothelioma production.
There are no studies currently available in laboratory animals exposed to LAA by
inhalation. However, the chronic intraperitoneal injection study in hamsters (Smith et al., 1979;
Smith, 1978) demonstrated tumor formation following exposure to tremolite obtained from the
Libby, MT mine. No other chronic inhalation studies of LAA are available. A recent study in
rats examining the impact of preexisting cardiovascular disease on pulmonary inflammation
demonstrated an increase in inflammatory markers following exposure to LAA via intratracheal
instillation in SH rats as compared to normal healthy controls exposed to the same dose
(Shannahan et al., 201 Ib). More recent studies examined gene expression changes (Hillegass et
al., 2010; Putnam et al., 2008) and early protein markers of fibrosis (Smartt et al., 2010) in mice
exposed to LAA via intraperitoneal injection. These studies demonstrated an increase in gene
and protein expression related to fibrosis following exposure to LAA. Tremolite fibers, although
obtained from different locations throughout the world, consistently led to pulmonary lesions
and/or tumor formation with various routes of exposure (inhalation, injection, instillation) and in
multiple species [rats, hamsters, and mice (Bernstein et al., 2005b; Bernstein et al., 2003; Roller
et al., 1997, 1996; Davis et al., 1985; Wagner et al., 1982; Stanton et al., 1981)1. Although
comparing potency of the various forms of tremolite is difficult given the limited information on
fiber characteristics and study limitations (e.g., length of follow-up postexposure), these results
show potential increased risk for cancer (lung and mesothelioma) following exposure to
tremolite asbestos.
The results of the studies described above show the fibrogenic and carcinogenic potential
of Libby Amphibole and tremolite asbestos. Further, the more recent studies by Salazar et al.
(2013). Salazar etal. (2012). Blake et al. (2008). and Pfau et al. (2008) support human studies
demonstrating potential autoimmune effects of asbestos exposure (see Section 4.3.1).
4.2.4. Oral
No studies in laboratory animals with oral exposure to LAA were found in the literature.
However, one chronic cancer bioassay was performed following oral exposure to tremolite.
McConnell et al. (1983a) describe part of a National Toxicology Program study (NTP, 1990b)
performed to evaluate the toxicity and carcinogenicity of ingestion of several minerals, including
tremolite. The tremolite (Governeur Talc Co, Governeur, NY) used was not fibrous. No
significant tumor induction was observed in the animals with oral exposure to tremolite animals.
Although nonneoplastic lesions were observed in many of the aging rats, these were mostly in
4-66
-------
the stomach and occurred in both controls and exposed animals. The observed lesions included
chronic inflammation, ulceration, and necrosis of the stomach (McConnell etal., 1983a).
McConnell et al. (1983 a) suggested that nonfibrous nature of this tremolite sample could account
for the lack of toxicity following exposure in this group of animals. Also, oral studies of
asbestos, in general, show decreased toxicity and carcinogenicity as compared to inhalation and
implantation/injection studies (Condie, 1983).
4.2.5. Summary of Animal Studies for Libby Amphibole and Tremolite Asbestos
Tables 4-19 and 4-20 summarize the studies described in this section, with full study
details available in Appendix D. Limited in vivo studies have been performed exposing
laboratory animals to LAA. One intrapleural injection study using tremolite from the Libby, MT
area is included in this section under LAA because earlier terminology for LAA was often
tremolite (Smith, 1978). Hamsters in this study exposed to LAA developed fibrosis and
mesothelioma following exposure. Intratracheal instillation studies of LAA in rats showed
increased collagen gene expression at 2-years postexposure (Cyphert et al., 2012a).
Sub chronic-duration studies in mice (Smartt et al., 2010; Putnam et al., 2008) demonstrated gene
and protein expression changes related to fibrosis production following exposure to LAA.
Finally, short-term-duration studies in rats demonstrated an increase in inflammatory and
cardiovascular disease markers following exposure to LAA (Padilla-Carlin et al., 2011;
Shannahan et al., 201 la: Shannahan et al., 201 Ib).
Because tremolite is a component of LAA, results from tremolite studies were also
described. In general, fibrous tremolite has been shown to cause pulmonary inflammation,
fibrosis, and/or mesothelioma or lung cancer in rats (Bernstein et al., 2005b: Bernstein et al.,
2003: Davis etal., 1991: Davis etal., 1985: Wagner etal., 1982) and hamsters (Smith et al.,
1979). The single short-term-duration study on mice showed limited response to tremolite (Sahu
et al., 1975). The one chronic-duration oral study (McConnell et al., 1983a) did not show
increased toxicity or carcinogenicity; this study, however, used only nonfibrous tremolite, which
later studies showed to be less toxic and carcinogenic than fibrous tremolite (Davis etal., 1991).
Chronic inflammation is hypothesized to lead to a carcinogenic response through the
production of reactive oxygen species and increased cellular proliferation (Hanahan and
Weinberg, 2011). Although limited, the data described in Section 4.2 suggest an increase in
inflammatory response following exposure to LAA and tremolite asbestos similar to that
observed for other durable mineral fibers [reviewed in (Mossman et al., 2007)]. Whether this
inflammatory response then leads to cancer is unknown. Studies examining other types of
asbestos (e.g., crocidolite, chrysotile, and amosite) have demonstrated an increase in chronic
inflammation as well as respiratory cancer related to exposure [reviewed in (Kamp and
Weitzman, 1999)]. Chronic inflammation has also been linked to genotoxicity and mutagenicity
following exposure to some particles and fibers (Driscoll etal., 1997: Driscoll et al., 1996:
4-67
-------
Driscoll etal., 1995). The evidence described above suggests chronic inflammation is observed
following LAA and tremolite asbestos exposure; however, the role of inflammation and whether
it leads to lung cancer or mesothelioma following exposure to LAA is unknown.
ROS production has been measured in response to both LAA and tremolite asbestos
exposure. Blake et al. (2007) demonstrated an increase in the production of superoxide anions
following exposure to LAA. Blake et al. (2007) also demonstrated that total superoxide
dismutase (SOD) was inhibited, along with a decrease in intracellular glutathione (GSH), both of
which are associated with increased levels of ROS. These results are supported by a recent study
in human mesothelial cells (Hillegass et al., 2010) (described in Section 4.4 and Appendix D).
Increased ROS production was also observed in human airway epithelial cells (HAECs)
following exposure to LAA (Duncan et al., 2010) (described in Section 4.4 and Appendix D).
This increase in ROS and decrease in glutathione are common effects following exposure to
asbestos fibers and particulate matter. Pfau et al. (2012) examined the role of the amino acid
transport system x~c, which is one of the pathways murine macrophages use to detect and
respond to stressful conditions. This study demonstrated that ROS production increase system
x~c activity. Although ROS production is relevant to humans, based on similar human responses
as compared to animals, information on the specifics of ROS production following exposure to
LAA is limited to the available data described here. Therefore, the role of ROS production in
lung cancer and mesothelioma following exposure to LAA is unknown.
Recent studies have also examined the role of the inflammasome and iron in the
development of fibrosis in male SH rats. The role of inflammasome activation and iron in the
development of LAA-induced fibrosis was studied in Shannahan et al. (2012c). Lung tissue
expression of inflammatory cytokines CCL-7, Cox-2, CCL-2, and CXCL-3 was increased
4 hours following LAA exposure. Conversely, LAA exposure reduced IL-4 and CXC1-1 in the
BALF. Finally, the ratio of phosphorylated ERK (pERK)/extracellular signal-related kinases
(ERK), which is an upstream activator of the inflammasome cascade, was increased in the lung
of LAA exposed rats 1 day post exposure. Rats treated with LAA + DEF or LAA + Fe had
significantly different levels of Cox-2 in the BALF and IL-6 in lung tissue but all other endpoints
were not significantly different. These data suggest that the concentration of iron does not
impact the activation of the inflammasome cascade and cytokines downstream of the pathway in
LAA-exposed animals.
In another study examining the role of iron in lung disease, Shannahan et al. (2012d)
valuated the effect of Fe overload on LAA-induced lung injury in rats with cardiovascular
disease. Gene array analysis demonstrated that LAA exposure upregulated inflammatory-related
genes such as NF-kp and cell cycle regulating genes such as matrix metalloproteinase-9 in WKY
rats but inhibited these same cluster of genes in SH and SHHF animals 3 months after
instillation. Histological examination of lung sections observed greater Fe staining of
macrophages in SHHF rats compared to WKY and SH rats 1 and 3 months post exposure;
4-68
-------
however, no differences in the progression of pulmonary fibrosis were noted between the three
strains. Altogether, these data do not suggest that the iron overload conditions that are
characteristic of the cardiovascular disease strains amplify the pulmonary effects of LAA.
4.3. OTHER DURATION- OR ENDPOINT-SPECIFIC STUDIES
4.3.1. Immunological
Salazaretal. (20131 Salazaretal. (2012). Rasmussen and Pfau (2012). Blake et al.
(20081 Pfau et al. (2008). Serve etal. (2013) and Hamilton et al. (2004) examined the role of
asbestos in autoimmunity in laboratory animal or in vitro studies. Blake et al. (2008) performed
in vitro assays with LAA (see Section 4.4), and both studies performed the in vivo assays with
tremolite. C57BL/6 mice were instilled intratracheally for a total of two doses each of 60 ug
saline and wollastonite or Korean tremolite sonicated in sterile phosphate buffered saline (PBS),
given 1 week apart in the first 2 weeks of a 7-month experiment. Sera from mice exposed to
tremolite showed antibody binding colocalized with SSA/Ro52 on the surface of apoptotic blebs
(Blake et al., 2008). In Pfau et al. (2008), by 26 weeks, the tremolite-exposed animals had a
significantly higher frequency of positive antinuclear antibody (ANA) tests compared to
wollastonite and saline. Most of the tests were positive for double-stranded DNA (dsDNA) and
SSA/Ro52. Serum isotyping showed no major changes in immunoglobulin (Ig) subclasses (IgG,
IgA, IgM), but serum IgG in tremolite-exposed mice decreased overall. Further, IgG immune
complex deposition in the kidneys increased, with abnormalities suggestive of
glomerulonephritis. No increased proteinuria was observed during the course of the study.
Local immunologic response was further studied on the cervical lymph nodes. Although total
cell numbers and lymph-node sizes were significantly increased following exposure to tremolite,
percentages of T- and B-cells did not significantly change.
Hamilton et al. (2004) investigated the ability of LAA, crocidolite, and paniculate matter
2.5 um in diameter or less (PIVh.s, collected over a 6-month period in Houston, TX, from EPA
site 48-201-1035) to alter the antigen-presenting cell (APC) function in cultured human alveolar
macrophages. Asbestos exposure (regardless of type) and PIVh.s up-regulated a Thl
lymphocyte-derived cytokine, interferon gamma (IFNy), and the Th2 lymphocyte-derived
cytokines IL-4 and IL-13. However, extreme variation among subjects was noted in the amount
of response. In addition, no correlation was present between the response of an individual's cells
to asbestos versus particulate matter, suggesting that more than one possible mechanism exists
for a particle-induced APC effect and individual differential sensitivities to inhaled bioactive
particles. Rasmussen and Pfau (2012) examined the role of macrophages in the development of
autoantibody production following exposure to LAA. LAA exposure alone did not affect cell
proliferation or antibody production; however, culturing lymphocytes with macrophage medium
following exposure to LAA did lead to increased cellular proliferation and antibody production.
Serve et al. (2013) examined a possible role of autoimmunity in fibrosis by an in vitro
4-69
-------
examination of potential mechanisms of MCAA leading to collagen deposition, a precursor to
fibrosis. This study demonstrated MCAA binding leads to increased collagen deposition through
altering matrix metalloproteinase (MMP) expression.
In two studies examining potential autoimmune effects of LAA exposure, Salazar et al.
(2013) and Salazar et al. (2012) examined the potential impact of LAA exposure on rheumatoid
arthritis and on ANA increases associated with systemic autoimmune disease. Salazar et al.
(2013) and Salazar et al. (2012) conducted a series of studies to establish the effects of LAA
exposure on autoimmune disease. The first set of studies utilized the collagen-induced arthritis
and peptidoglycan-polysaccharide (PG-PS) models of rheumatoid arthritis to determine whether
LAA exposure increased the onset, prolonged, or intensified the joint inflammation characteristic
of the disease (Salazar et al., 2012). Female Lewis rats were instilled biweekly for 13 weeks
with a total dose of 0, 0.15, 0.5, 1.5, and 5.0 mg LAA followed by induction with either model of
arthritis. LAA 5.0 mg reduced the magnitude of the swelling response in the cell-mediated
PG-PS model; however, neither the onset nor the duration of swelling was affected by LAA
exposure. LAA 1.5 and 5.0 mg and amosite 0.5 and 1.5 mg reduced total serum IgM. LAA
5.0 mg and amosite 1.5 mg reduced anti-PG-PS IgG in the serum 17 weeks after the final
instillation. Finally, the number of rats positive for ANA was increased only at the low exposure
concentrations of LAA in PG-PS-treated and nonarthritic rats. These results suggest that LAA
may have a modest inhibitory effect on the PG-PS rat model but may enhance responses to other
systemic autoimmune diseases.
In a follow-up study, Salazar et al. (2013) explored in greater detail the effect of LAA
exposure on ANA over time and the antigen specificity of the ANA. Female Lewis rats were
intratracheally instilled under the conditions in the previous study (Salazar et al., 2012). Serum
samples were analyzed every 4 weeks from the beginning of the instillations up to termination at
Week 28. Because elevated ANA are commonly associated with kidney disease, proteinuria was
assessed every 3 weeks beginning at Week 6 until termination of the experiment.
Histopathological analysis was also performed on the kidneys. ANA was increased 8 weeks
postexposure to LAA 5.0 mg. By Week 28, all doses of LAA except 1.5 mg increased ANA in
the serum. Analysis of the antigen specificity found that only the LAA at 1.5 mg significantly
increased antibodies specific for extractable nuclear antigens and the Jo-1 antigen. Urinalysis
found that all doses of LAA exposure induce moderate levels of proteinuria, but this effect was
not dose responsive. No dose-related histopathological effects were observed. Altogether, these
data suggest that LAA exposure increases autoimmune antibodies in the serum, but no evidence
of autoimmune disease was identified. However, the lack of SAID in the Lewis rat may be due
to strain-specific factors and suggests that other animal models may be more appropriate for
studying autoimmune effects of LAA.
4-70
-------
Although the number of studies is limited, the results suggest a possible effect on
autoimmunity following exposure to LAA. Further studies are needed to increase understanding
of this potential effect.
4.4. MECHANISTIC DATA AND OTHER STUDIES IN SUPPORT OF THE MODE OF
ACTION
For asbestos in general, International Agency for Research on Cancer (IARC) has
proposed a mechanism for the carcinogenicity of asbestos fibers [see Figure 4-4; (IARC, 2012)].
Asbestos fibers lead to oxidant production through interactions with macrophages and through
hydroxyl radical generation from surface iron. Inhaled fibers that are phagocytosed by
macrophages may be cleared or lead to frustrated phagocytosis, which results in macrophage
activation, release of oxidants, and increased inflammatory response, in part due to
inflammasome activation. Free radicals may also be released by interaction with the iron on the
surface of fibers. Increased oxidant production may result in epithelial cell injury, including
DNA damage. Frustrated phagocytosis may also lead to impaired clearance of fibers, with fibers
being available for translocation to other sites (e.g., pleura). Mineral fibers may also lead to
direct genotoxicity by interfering with the mitotic spindle and leading to chromosomal
aberrations. Asbestos exposure also leads to the activation of intracellular signaling pathways,
which in turn may result in increased cellular proliferation, decreased DNA damage repair, and
activation of oncogenes. Research on various types of mineral fibers supports a complex
mechanism involving multiple biologic responses following exposure to asbestos (i.e.,
genotoxicity, chronic inflammation/cytotoxicity leading to oxidant release, and cellular
proliferation) in the carcinogenic response to mineral fibers [see Figure 4-4, reviewed in (IARC,
2012)]. These complexities of fiber toxicity need to be considered when analyzing MO A for
asbestos [as reviewed in (Aust et al., 2011; Broaddus et al., 2011; Bunderson-Schelvan et al.,
2011: Huang etal.. 2011: Mossman et al.. 2011)1.
4-71
-------
-------
Important considerations in evaluating the available mechanism and MOA data are fiber
characteristics, route of exposure, dose metric, as well as study design and interpretation.
Specific fiber characteristics impact the fiber toxicokinetics (reviewed in Section 3), and in turn
the biologic response to fibers. For example, fiber length is an important determinant of fiber
clearance, with shorter fibers generally being cleared more efficiently as longer fibers result in
frustrated phagocytosis. Mechanisms of carcinogenesis may accordingly vary based on fiber
characteristics. The biologic response to respirable fibers is also influenced by the route of
exposure. Inhalation exposure studies are the most informative. Intratracheal instillation and
aspiration exposures bypass normal clearance mechanisms, and therefore affect fiber dosimetry.
Concerning dose metric, some studies suggest that the dose should be determined based on fiber
length, width, number, or surface area (Case etal., 2011; Mossman et al., 2011). However, the
majority of studies of fibers have been performed using mass as a dose. Finally, an important
consideration in analysis of in vitro studies is the cell types used, particularly related to the
ability to internalize fibers and produce an oxidative stress response. The discussions below
highlight these considerations in presenting the available mechanistic evidence for LAA.
Limited in vitro studies have been conducted with LAA from the Zonolite Mountain
mine. These studies demonstrated an effect of LAA on inflammation and immune function
(Duncan etal.. 2014: Duncan et al.. 2010: Blake et al.. 2008: Blake et al.. 2007: Hamilton et al..
2004), oxidative stress (Hillegass et al., 2010), and genotoxicity (Pietruska et al., 2010). Similar
endpoints have been examined in vitro following exposure to tremolite asbestos (Okayasu et al.,
1999: Wylieetal., 1997: Suzuki and Hei, 1996: Athanasiou et al., 1992: Wagner etal., 1982).
Results from in vitro studies have demonstrated potential biological mechanisms of oxidative
stress and inflammation in response to exposure to Libby Amphibole and tremolite asbestos. As
discussed in Section 4.2, laboratory animal studies examining the effects of tremolite exposure
have been reviewed and are summarized to potentially increase understanding of the effects and
mechanisms of LAA, because tremolite is a component of LAA (-6%). These studies are
summarized below and in Tables 4-21 and 4-22, with detailed study descriptions available in
Appendix D.
4-73
-------
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
XRCCl-deficient)
Human mesothelial cell
lines (LP9/TERT-1 and
HKNM-2)
Primary HAECs
Fiber type
LAA or crocidolite
LAA or 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
8 or 24 h
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)
4-74
-------
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.
4-75
-------
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, FD14, S157,
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)
4-76
-------
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.
(19991
A[L] cells = hamster hybrid cells containing human chromosome 11; BGL = (3-glucuronidase; HTE = hamster tracheal epithelial; NIEHS = National Institute
of Environmental Health Sciences; RCF-1 = refractory ceramic fibers ; RPM = rat pleural mesothelial; SHE = Syrian hamster embryo.
4-77
-------
4.4.1. Inflammation and Immune Function
Chronic inflammation following inhalation exposure to asbestos has been studied for
decades in both humans and animals [reviewed in (Mossman et al., 2011; Rom et al., 1991)1.
This inflammatory response has been attributed to the role of alveolar macrophages, which
attempt to engulf asbestos fibers to clear them from the respiratory tract [reviewed in Mossman
et al. (2011) and Takemura et al. (1989)]. Mechanistic studies have focused on the
chemokine/cytokine response of these macrophages and the subsequent signaling pathways that
are activated. More recently, studies have also examined the role of inflammasome activation
following exposure to asbestos in immune activation and fiber clearance (Hillegass et al., 2013;
Biswas et al.. 2011: Dostert et al.. 2008).
Increased cytokine and chemokine production has been observed following exposure to
LAA as well as other amphibole asbestos. Hamilton et al. (2004) showed an increase in Thl and
Th2 cytokines following exposure to LAA, crocidolite, and particulate matter, suggesting a
similar effect of exposure to these materials on immune function. Analysis of these results is
limited, as the use of primary cells in culture led to an extremely variable response. Two studies
by Blake et al. (2008) and Blake et al. (2007) further examined the effect of LAA on immune
response in vitro in mouse macrophages. These studies demonstrated that the size of the LAA
material is such that it was able to be internalized by macrophages (<10 um), and this
internalization resulted in an increase in ROS production. These studies also showed a variable
cytotoxic response, because LAA exposure did not result in a statistically significant increase in
cytotoxicity, while crocidolite did. DNA damage also was increased in crocidolite-exposed cells
but not in LAA-exposed cells. An increase (relative to controls) in autoantibody formation
following exposure to LAA was also observed. Studies that examined cellular response to
tremolite also found that fiber characteristics (length and width) play a role in determining ROS
production, toxicity, and mutagenicity (Okayasu et al., 1999; Wagner et al., 1982).
Gene expression alterations ofIL-8, COX-2, heme oxygenase (HO)-l, as well as other
stress-responsive genes as compared to amosite was observed in primary HAEC following
exposure to LAA. Comparisons were made with both fractionated (aerodynamic diameter
<2.5 um) and unfractionated fiber samples (Duncan et al., 2010). Crocidolite fibers (UICC)
were also included in some portions of this study for comparison. Primary HAECs were exposed
to 0, 2.64, 13.2, and 26.4 ug/cm2 of crocidolite, amosite, amosite 2.5 (fractionated), LAA, or
LAA 2.5 (fractionated) for 2 or 24 hours in cell culture. Minimal increases in gene expression of
IL-8, COX-2, or HO-1 were observed at 2 hours postexposure to all five fiber types; at 24 hour
postexposure, however, a dose-response was observed following exposure to all fiber types with
the results showing a proinflammatory gene expression response (Duncan et al., 2010).
Cytotoxicity was determined by measuring LDH from the maximum dose (26.4 ug/cm2) of both
amosite and LAA samples, with less than 10% LDH present following exposure to all
four samples. A follow-up study with the same design by Duncan et al. (2014)was performed to
4-78
-------
examine the in vitro determinants of asbestos fiber toxicity, comparing two samples each of
LAA (LA2000, LA2007) and amosite (UICC, RTI) asbestos. Primary human airway epithelial
cells (HAEC) were exposed for 24 hours to 2.64, 13.2 or 26.4 ug/cm2 LAA and amosite asbestos
that had been analyzed for fiber size distribution, surface area, and surface-conjugated iron (see
Table D-l 1). Most characteristics were similar, except RTI amosite consisted of longer fibers.
Fiber toxicity was measured by cytotoxicity (LDH assay), levels of ROS production, as well as
IL-8 mRNA levels as a measure of relative proinflammatory responses. Cytotoxicity levels were
similar for all four samples at the highest dose, but statistically significant compared to the
no-treatment control. Results on an equal mass basis demonstrated a statistically significant
increase in IL-8, IL-6, COX-2, and TNF mRNA levels for all four amphiboles at the highest two
doses. The greatest increase in IL-8 mRNA levels was following exposure to the RTI amosite
sample, while both LAA samples and the UICC amosite resulted in a similar level of response to
each other. Therefore, IL-8 was used to further analyze the dose metrics for this response.
Surface iron concentrations and surface reactivity was quantified with respect to hydroxyl radical
production to assess the effect of these properties on IL-8 mRNA expression. Surface iron
concentrations were similar for the two LAA samples and for the two amosite samples, but the
amosite samples had much greater surface iron as compared to the LAA samples. UICC amosite
had slightly greater iron as compared to RTI. A strong correlation was observed between fiber
dose metrics of length and external surface area. When these metrics were used in place of equal
mass dose, the differential IL-8 mRNA expression following exposure to these four samples was
eliminated. These results support a limited cytotoxicity and increased inflammatory cytokine
response of both amosite and LAA under these concentrations and time frames.
The role of macrophages in the development of autoantibody production following
exposure to asbestos was examined in a study performed by Rasmussen and Pfau (2012)
culturing CH12.LX B-lymphocytes, a murine Bl lymphocyte cell line, with LAA alone did not
affect proliferation or antibody production. However, culturing RAW264.7 macrophages with
LAA, collecting the macrophage medium, and culturing CH12.LX lymphocytes in the
conditioned medium reduced CH12.LX proliferation and increased IgGl, IgG3, and IgA
production. Further analysis found that both IL-6 and tumor necrosis factor (TNF)-a were
elevated in the medium of Libby Amphibole-treated macrophages, but only IL-6 increased IgG
and IgA production. However, these data also indicate that activated macrophages may regulate
CH12.LX antibody production. Altogether, these data suggest a potential mechanism for
macrophages to regulate asbestos-induced autoantibody production in Libby-exposed residents.
Chronic inflammation is also associated with oxidative stress, mechanisms of which
following exposure to LAA were also studied in human mesothelial cells (Hillegass et al., 2010).
Gene expression changes related to oxidative stress following exposure to 15 x io6 um2/cm2
4-79
-------
LAA17 as compared to the nonpathogenic control (75 x 106 um2/cm2 glass beads) in the human
mesothelial cell line LP9/TERT-1 for 8 and 24 hours. Gene ontology of these results
demonstrated alterations in genes related to signal transduction, immune response, apoptosis,
cellular proliferation, extracellular matrix, cell adhesion, and motility, and only in one gene
related to reactive oxygen species processing. Oxidative stress was observed to be both dose-
and time-dependent in cells exposed to LAA. GSH levels were transiently depleted following
2-8 hours exposure to the higher dose of LAA, with a gradual recovery up to 48 hours in
LP9/TERT-1 cells (HKNM-2 not analyzed). These studies demonstrate that LAA exposure
leads to increases in oxidative stress as measured by ROS production, gene expression, protein
and functional changes in oxidative stress proteins (SOD), and GSH level alterations in human
mesothelial cells.
The role of inflammasome activation and iron in the development of LAA-induced
fibrosis was studied in Shannahan et al. (2012d). Male SH rats were instilled with a single
exposure to 0 or 0.5 mg LAA, DBF, 21 ug FeCb, 0.5 mg LAA + 21 ug FeCb, or 0.5 mg
LAA + 1 mg DEF. Tissues were collected 4 hours and 1 day postexposure. LAA instillation
increased lung expression of the inflammasome-related molecules cathepsin B, Nalp3, NF-kp,
apoptosis-associated speck-like protein containing a CARD (ASC), IL-lp, and IL-6 expression
4 hours postexposure. Lung tissue expression of inflammatory cytokines CCL-7, Cox-2, CCL-2,
and CXCL-3 was increased 4 hours following LAA exposure. These data suggest that LAA
exposures leads to the activation of the inflammasome cascade and cytokines downstream of the
pathway in LAA-exposed animals. Li et al. (2012) also studied LAA inflammasome activation
and demonstrated in vitro in THP-1 cells that LAA activated the NLRP3 inflammasome but to a
lesser degree than chrysotile. However, this study showed that LAA exposure generated more
ROS production as compared to chrysotile. Although not studied, the authors suggest that
differences in fiber length and surface area may play a role in this differential inflammatory
response.
Libby Amphibole asbestos samples were characterized for this study with analysis of chemical composition and
mean surface area (Meeker et al.. 2003X 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).
4-80
-------
4.4.2. Genotoxicity
Genotoxicity, and more specifically, mutagenicity, are associated with tumor formation
through alterations in genetic material.18 Mutagenicity refers to a permanent effect on the
structure and/or amount of genetic material that can lead to heritable changes in function, while
genotoxicity is a broader term including all adverse effects on the genetic information (Eastmond
et al., 2009). Results of standard mutation assays like the Ames test, which analyze for point
mutations, have found asbestos and other mineral fibers to be negative or only marginally
positive (Walker et al., 1992). Several other studies, however, have shown that asbestos
exposure can result in a variety of chromosomal alterations, which are briefly discussed below.
Genotoxicity following exposure to asbestos fibers has been described as the result of
two distinct mechanisms, either ROS production leading to DNA damage, or physical
interference of mitosis by the fibers. For both DNA damage and mitotic interference, the fibers
must first enter the cell. Some studies have shown that a direct interaction between fibers and
cellular receptors might also lead to increased ROS production. ROS production is also related
to surface iron on fibers, with increased surface iron leading to increased ROS production
(IARC, 2012). ROS production is possibly a key event in fiber-induced direct DNA damage, as
observed following exposure to other forms of asbestos, while the indirect DNA damage requires
fiber interaction with cellular components (e.g., mitotic spindle, chromosomes).
ROS production and genotoxicity (micronuclei induction) following exposure to LAA
has been demonstrated in x-ray repair cross-complementing protein 1 (XRCCl)-deficient human
lung epithelial H460 cells (Pietruska et al., 2010). XRCC1 is involved in the repair mechanisms
for oxidative DNA damage, particularly single strand breaks. Micronuclei induction was
measured following treatment of cells by controls (positive, hydrogen peroxide; negative,
paclitaxel) and by 5 ug/cm2 fibers or TiO2 particles for 24 hours. Spontaneous micronuclei
induction was increased in XRCC1-deficient cells in a dose-dependent manner following
exposure to crocidolite and LAA as compared to unexposed cells. These results support a
potential genotoxic effect of exposure to both crocidolite and LAA.
Athanasiou et al. (1992) performed a series of experiments to measure genotoxicity
following exposure to tremolite, including the Ames mutagenicity assay, micronuclei induction,
chromosomal aberrations, and gap-junction intercellular communication. Although a useful test
system for mutagenicity screening for many agents, the Ames assay is not the most effective test
^Genotoxicity: 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).
4-81
-------
to detect mutations induced by mineral fibers. Mineral fibers can cause mutation through
generation of ROS or direct disruption of the spindle apparatus during chromatid segregation.
Fibers do not induce ROS in the Ames system, however, and the Salmonella typhimurium strains
do not endocytose the fibers. Only one study was found in the published literature that used the
Ames assay to measure mutagenicity of tremolite. Metsovo tremolite asbestos has been shown
to be the causative agent of endemic pleural calcification and an increased level of malignant
pleural mesothelioma (see Section 4.1). To measure the mutagenicity of Metsovo tremolite,
S. typhimurium strains (TA98, TA100, and TA102) were exposed to 0-500 ug/plate of asbestos
(Athanasiou et al., 1992). Metsovo tremolite did not yield a statistically significant increase in
revertants in the Ames assay, including in the TA102 Salmonella strain, which is generally
sensitive to oxidative damage. This study demonstrated clastogenic effects of tremolite,
including chromosomal aberrations and micronuclei induction. Tremolite exposure in Syrian
hamster embryo (SHE) cells did lead to a dose-dependent increase in chromosome aberrations
that was statistically significant at the highest doses tested [1.0-3.0 ug/cm2;/? <0.01; (Athanasiou
et al., 1992)]. A statistically significant, dose-dependent increase in levels of micronuclei was
demonstrated following tremolite exposure at concentrations as low as 0.5 ug/cm2 (p <0.01) in
BPNi cells after 24-hour exposure. Literature searches did not find tremolite tested for
clastogenicity in other cell types, but the results of this study suggest interference with the
spindle apparatus by these fibers. No analysis was performed to determine whether fiber
interference of the spindle apparatus could be observed, which would have supported these
results. No effect on the gap-junctional intercellular communication following tremolite
exposure was observed in both Chinese hamster lung fibroblasts (V79) and Syrian hamster
embryo BPNi cells, which are sensitive to transformation (Athanasiou et al., 1992).
Okavasu et al. (1999) analyzed the mutagenicity of Metsovo tremolite, erionite, and the
man-made refractory ceramic fiber (RCF-1). Human-hamster hybrid AL cells contain a full set
of hamster chromosomes and a single copy of human chromosome 11. Mutagenesis of the CD59
locus on this chromosome is quantifiable by an antibody complement-mediated cytotoxicity
assay. The study authors state that this is a highly sensitive mutagenicity assay, and previous
studies have demonstrated mutagenicity of both crocidolite and chrysotile (Hei et al., 1992). The
cytotoxicity analysis for mutagenicity was performed by exposing 1 x 105 AL cells to a range of
concentrations of fibers as measured by weight (0-400 ug/mL or 0-80 ug/cm2) for 24 hours at
37°C. A dose-dependent increase in CD59 mutant induction was observed following exposure
to erionite and tremolite, but not RCF-1.
In summary, one in vitro study examined genotoxicity of LAA by measuring DNA
adduct formation following exposure via murine macrophages [primary and immortalized;
(Blake etal., 2007)1. The data showed no increase in adduct formation as compared to
unexposed controls. A second study observed increases in micronuclei induction in both normal
human lung epithelial cells and XRCC1-deficient cells for both LAA and crocidolite asbestos
4-82
-------
(Pietruska et al., 2010). Two studies of tremolite examined genotoxicity. The first found no
significant increase in revertants in the Ames assay (Athanasiou et al., 1992), which is similar to
results obtained for other forms of asbestos. This study did find, however, that tremolite
exposure led to a dose-dependent increase in chromosome number and micronuclei formation,
which has also been described for other asbestos fibers [as reviewed in Hei et al. (2006) and
Jaurand(1997)]. Hei and colleagues (Okayasu et al., 1999) performed mutation analysis with
tremolite and found a dose-dependent increase in mutations in CD59 in hamster hybrid cells.
Genotoxicity analysis in humans, following exposure to LAA or tremolite, has not been
measured, although other types of asbestos fibers have led to increases in genotoxicity in primary
cultures and lymphocytes (Dopp et al., 2005; Poser et al., 2004). In general, these studies have
examined genotoxicity with a focus on ROS production as a key event. Although LAA- and
tremolite-specific data are limited to in vitro studies, given the similarities in response to other
forms of asbestos, there is some evidence to suggest genotoxicity following exposure to Libby
Amphibole and tremolite asbestos. However, the potential role of this genotoxicity in lung
cancer or mesothelioma following exposure to LAA is unknown.
4.4.3. Cytotoxicity and Cellular Proliferation
The initial stages of turnorigenicity may be an increased cellular proliferation at the site
of fiber deposition, which can increase the chance of cancer by increasing the population of
spontaneous mutations, thereby affording genotoxic effects an opportunity to multiply.
Increased cell proliferative regeneration may be associated with tumor clonal expansion and can
occur in response to increased apoptosis. In macrophages, increased cytotoxicity leads to an
increased oxidant release, which in turn may lead to increased cell damage, signaling activation
and inflammatory cell recruitment.
Wagner et al. (1982) examined the in vitro cytotoxicity of three forms of tremolite used
in their in vivo studies. LDH and p-glucuronidase were measured in the medium following
incubation of unactivated primary murine macrophages to 50, 100, and 150 ug/mL of each
sample for 18 hours. The Korean tremolite (Sample C) produced results similar to the positive
control: increased toxicity of primary murine macrophages, increased cytotoxicity of Chinese
hamster ovary cells, and increased formation of cells >25 |im diameter from the A549 cell line.
The tremolite sample from Greenland (Sample B) did result in increased toxicity over controls;
although to a lesser degree (statistics are not given). Although differential toxicity of these
samples was noted on a mass basis, data were not normalized for fiber content or size. The
inference is that differential results may be due, at least in part, to differential fiber counts.
Wylie et al. (1997) examined the mineralogical features associated with cytotoxic and
proliferative effects of asbestos in hamster tracheal epithelial (HTE) and rat pleural mesothelial
(RPM) cells with a colony-forming efficiency assay. HTE cells are used because they give rise
to tracheobronchial carcinoma, while RPM cells give rise to mesotheliomas. The results of the
4-83
-------
analysis with fiber exposure by mass (ug/cm2) show elevated colonies in HTE cells following
exposures to both asbestos fibers (p <0.05) at the lowest concentrations, while significant
decreases were observed for both asbestos fibers at the higher concentrations [0.5 ug/cm2,
p <0.05; (Wvlie et al., 1997)1. No proliferation was observed for either chrysotile or crocidolite
asbestos fibers in RPM cells, but cytotoxicity was observed at concentrations greater than
0.05 ug/cm2 (p <0.05). All talc samples were less cytotoxic in both cell types. Analyzing the
data for cytotoxicity and proliferation based on the exposure measurement demonstrated
differences in response depending solely on how the fibers were measured: by mass, number, or
surface area. These results show variability in interpreting the results of the same assay based on
the defined unit of exposure. Most early studies used mass as the measurement for exposure,
which can impact how the results are interpreted. When possible, further analysis of fiber
number and surface area would help elucidate the role of these metrics, particularly for in vivo
studies.
Tremolite and LAA exposure led to increases in both fibrosis and tumorigenicity in all
but one animal study, supporting a possible role for proliferation in response to these fibers (see
Tables 4-19 and 4-20). However, there are limited data to demonstrate that increased
cytotoxicity and cellular proliferation following exposure to LAA leads to lung cancer or
mesothelioma.
Summary
The review of these studies clearly highlights the need for more controlled studies
examining LAA in comparison with other forms of asbestos and for examining multiple
endpoints—including ROS production, DNA damage, inflammasome activation, and
proinflammatory gene expression alterations—to improve understanding of mechanisms
involved in cancer and other health effects. Data gaps still remain to determine specific
mechanisms involved in LAA-induced disease. Studies that examined cellular response to
tremolite also found that tremolite exposure may lead to increased ROS production, toxicity, and
genotoxicity (Okayasu et al., 1999; Wagner etal., 1982). As with the in vivo studies, the
definition of fibers and how the exposures were measured varies among studies.
4.5. SYNTHESIS OF MAJOR NONCANCER EFFECTS
The predominant noncancer health effects observed following inhalation exposure to
LAA are effects on the lungs and pleural lining surrounding the lungs. These effects have been
observed primarily in studies of exposed workers and community members, and are supported by
laboratory animal studies. Recent studies have also examined other noncancer health effects
following exposure to Libby Amphibole, including autoimmune effects and cardiovascular
disease; this research base is currently not as well developed as that of respiratory noncancer
effects. Adequate data are not available to differentiate the health effects of the predominant
4-84
-------
mineralogical forms composing LAA. Although the adverse effects of asbestiform tremolite are
reported in the literature, the contribution of winchite and richterite to the aggregate effects of
LAA has not been determined.
4.5.1. Pulmonary Effects
4.5.1.1. Pulmonary Fibrosis (Asbestosis)
Asbestosis is the interstitial pneumonitis (inflammation of lung tissue) and fibrosis
caused by inhalation of asbestos fibers and is characterized by a diffuse increase of collagen in
the alveolar walls (fibrosis) and the presence of asbestos fibers, either free or coated with a
proteinaceous material and iron (asbestos bodies). Fibrosis results from a sequence of events
following lung injury, which includes inflammatory cell migration, edema, cellular proliferation,
and accumulation of collagen. Asbestosis is associated with dyspnea (shortness of breath),
bibasilar rales, and changes in pulmonary function: a restrictive pattern, mixed
restrictive-obstructive pattern, and/or decreased diffusing capacity (ATS, 2004). In clinical
practice, fibrotic scarring of lung tissue consistent with mineral dust and mineral fiber toxicity is
most commonly identified as small opacities in the lung on radiographic examination. The
scarring of the parenchymal tissue of the lung contributes to changes in pulmonary function,
including restrictive pulmonary deficits due to increased stiffness (reduced elasticity) of the lung,
impaired gas exchange due, in part, to alveolar wall thickening, and sometimes mild obstructive
deficits due to asbestos-induced airways disease.
Workers exposed to LAA from vermiculite mining and processing facilities in Libby,
MT, as well as plant workers in Marysville, OH, where vermiculite ore was exfoliated and
processed, have been found to have an increased prevalence of small opacities on chest x-rays,
which is indicative of fibrotic damage to the parenchymal tissue of the lung (Rohs et al., 2008;
Amandus et al., 1987a: McDonald et al., 1986b: Lockey et al., 1984). Significant increases in
asbestosis as a cause of death have been documented in studies of the Libby worker cohort report
[see Table 4-6 for details; (Larson et al.. 201 Ob: Sullivan. 2007: Amandus and Wheeler. 1987:
McDonald et al., 1986a)]. For both asbestosis mortality and radiographic signs of asbestos
(small opacities), positive exposure-response relationships are described where these effects are
greater with greater cumulative exposure to LAA.
Deficits in pulmonary function consistent with pulmonary fibrosis have been reported in
individuals exposed to LAA in community-based studies.19 Data from the ATSDR community
screening, which included workers, provide support for functional effects from parenchymal
changes. The original report of the health screening data indicated moderate to severe
19The initial study of the Marysville, OH cohort measured and 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 et al.. 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
4-85
-------
pulmonary restriction in 2.2% of men (Peipins et al., 2003; ATSDR, 200 Ib). A recent reanalysis
of these data show that for study participants with small opacities viewed on the radiographs
(category 1/0 or greater) and DPT, the mean FVC is reduced to 78.76 (±3.64), 82.16 (±3.34),
respectively, of the expected value (Weill etal., 2011). A mean FVC of 95.63 (±0.76) was
reported for those with other pleural abnormalities versus 103.15 (±0.25) in participants with no
radiographic abnormalities. The strongest effects of diffuse pleural thickening and/or
costophrenic angle obliteration on FVC were seen among men who had never smoked (-23.77,
p <0.05), with smaller effects seen among men who had smoked (—9.77,p <0.05) and women
who had smoked (-6.73, p <0.05). Laboratory animal and mechanistic studies of LAA are
consistent with the noncancer health effects observed in both Libby workers and community
members. Pleural fibrosis was increased in hamsters after intrapleural injections of LAA (Smith,
1978). More recent studies have demonstrated increased collagen deposition consistent with
fibrosis following intratracheal instillation of LAA fibers in mice and rats (Padilla-Carlin et al.,
2011: Shannahan et al., 201 la: Shannahan et al., 20 lib: Smartt etal., 2010: Putnam et al., 2008).
Pulmonary fibrosis, inflammation, and granulomas were observed after tremolite inhalation
exposure in Wistar rats (Bernstein et al., 2005b: Bernstein et al., 2003) and intratracheal
instillation in albino Swiss mice (Sahu et al., 1975). Davis etal. (1985) also reported pulmonary
effects after inhalation exposure in Wistar rats, including increases in peribronchiolar fibrosis,
alveolar wall thickening, and interstitial fibrosis.
4.5.1.2. Other Nonmalignant Respiratory Diseases
Mortality studies of the Libby workers indicate increased mortality not only from
asbestosis, but also from other respiratory diseases. Deaths attributed to chronic obstructive
respiratory disease and deaths attributed to "other" nonmalignant respiratory disease were
elevated more than twofold [see Table 4-6; (Larson etal., 201 Ob: Sullivan, 2007)]. These
outcomes are consistent with asbestos toxicity, and the evidence of a positive exposure-response
relationship for mortality from all nonmalignant respiratory diseases supports this association.
4.5.2. Pleural Effects
Pleural thickening caused by mineral fiber exposure mainly includes two distinct
biological lesions: localized pleural plaques in the parietal (outer) pleura and diffuse pleural
thickening of the visceral (inner) pleura. Both of these forms of pleural thickening can be
identified on standard radiographs; however, smaller/thinner plaques and thinner diffuse
thickening may not be detected, particularly if they are not calcified or are obscured by other
normal chest structures. High resolution computed tomography is a radiographic method that is
more sensitive and specific than standard chest x-rays (i.e., it can detect pleural abnormalities
that are not evident on standard chest x-rays and it can more reliably exclude fat tissue that can
sometimes be mistaken for pleural thickening on standard chest x-ray
4-86
-------
Data from the ATSDR community health screening study indicate that the prevalence of
pleural abnormalities, identified by radiographic examination, increases substantially with
increasing number of exposure pathways (Peipins et al., 2003). A reanalysis of these data also
considered age, smoking history, and types of exposures (Weill etal., 2011). Increased pleural
thickening is reported for Libby workers, those with other vermiculite work, and those who had
worked in other jobs with dust exposures (in locations other than Libby, MT). The prevalence of
pleural plaques increased with age; in the 61-90 age group the prevalence was 38.3% and
12.7%, respectively, among those exposed only through household contacts and those exposed
through environmental exposure pathways. The community-based study in Minneapolis, MN
also provides evidence of increased risk of pleural abnormalities among residents surrounding an
exfoliation plant, with positive associations seen with measures of background and of
intermittent (activity-based) exposures (Alexander et al., 2012).
Increased pleural thickening (including LPT) is reported for both of the studied worker
cohorts, with evidence of positive exposure-response relationships (Larson etal., 2012a: Larson
etal.. 2010a: Rohs et al.. 2008: Amandus et al.. 1987a: McDonald et al.. 1986b: Lockev et al..
1984). Both McDonald et al. (1986b) and Amandus et al. (1987a) indicate that age is also a
predictor of pleural thickening in exposed individuals, which may reflect the effects of time from
first exposure. Smoking data were limited on the Libby workers and analyses do not indicate
clear relationships between smoking and pleural thickening (Amandus et al., 1987a: McDonald
etal., 1986b). Pleural thickening in workers at the Scott Plant (Marysville, OH) was associated
with hire on or before 1973 and age at time of interview but was not associated with BMI or
smoking history [ever smoked (Rohs et al., 2008)].
4.5.3. Other Noncancer Health Effects (Cardiovascular Toxicity, Autoimmune Effects)
Limited research is available on noncancer health effects occurring outside the
respiratory system and pleura. Larson et al. (201 Ob) examined cardiovascular disease-related
mortality in the cohort of exposed workers from Libby (see Section 4.1). Mechanistic studies
have examined the potential role of iron and the associated inflammation for both respiratory and
cardiovascular disease (Shannahan et al., 2012a: Shannahan et al., 2012c: Shannahan et al.,
2012b: Shannahan et al., 2012d: Shannahan et al., 201 Ib). Other studies examined the
association between asbestos exposure and autoimmune disease (Noonan et al., 2006) or
autoantibodies and other immune markers [(Pfau et al., 2005): see Table 4-15]. However,
limitations in the number, scope, and design of these studies make it difficult to reach
conclusions as to the role of asbestos exposure in either cardiovascular disease or autoimmune
disease.
4-87
-------
4.5.4. Summary of Noncancer Health Effects of Exposure to Libby Amphibole Asbestos
The studies of humans summarized in Section 4.1 have documented an increase in
mortality from nonmalignant respiratory disease, including asbestosis, in workers exposed to
LAA (Larson et al., 201 Ob; Sullivan, 2007; McDonald et al., 2004; Amandus and Wheeler,
1987). Additional studies have documented an increase in radiographic changes in the pleura
(pleural thickening) and parenchyma among employees of a manufacturing facility in
Marysville, OH that processed LAA-containing vermiculite ore (Rohs et al., 2008; Lockey et al.,
1984). Radiographic evidence of pleural thickening and interstitial damage (small opacities) are
also well documented among employees of the Libby vermiculite mining operations (Larson et
al.. 2012a; Larson et al.. 2010a; Amandus et al.. 1987a; McDonald et al.. 1986b). Positive
exposure-response relationships for these health effects for both occupational cohorts studied, as
well as the observed latency, support an association between exposure to LAA and these pleural
and/or pulmonary effects. Studies of community members exposed to LAA have documented
similar pleural abnormalities and pulmonary function deficits consistent with effects of tissue
damage caused by LAA (Weill etal., 2011; Whitehouse, 2004; Peipins et al., 2003). Animal
studies also support the toxicity of LAA to pleural and pulmonary tissues. Developing research
supports a role of inflammatory processes in the toxic action of LAA, consistent with the
observed health effects (Cvphert etal., 2012b; Shannahan etal.. 2012c: Shannahan et al.. 20 lib;
Duncan et al.. 2010; Hamilton et al., 2004). Taken together, the strong evidence in human
studies, defined exposure-response relationships, and supportive animal studies provide
compelling evidence that exposure to LAA causes nonmalignant respiratory disease, including
asbestosis, pleural thickening, and deficits in pulmonary function associated with mineral fiber
exposures. Existing data regarding cardiovascular effects and the potential for autoimmune
disease are limited.
4.5.5. Mode-of-Action Information (Noncancer)
The precise mechanisms causing toxic injury from inhalation exposure to LAA have not
been established. However, nearly all durable mineral fibers with dimensional characteristics
that allow penetration to the terminal bronchioles and alveoli of the lung have the capacity to
induce pathologic response in the lung and pleural cavity (ATSDR, 200la; Witschi and Last,
1996). The physicochemical attributes of mineral fibers are important in determining the type of
toxicity observed. Fiber dimension (width and length), density, and other characteristics, such as
chemical composition, surface area, solubility in physiological fluids, and durability, all play
important roles in both the type of toxicity observed and the biologically significant dose. As
described in Section 3, these characteristics also play a role in fiber dosimetry. Fibrosis results
from a sequence of events following lung injury, which includes inflammatory cell migration,
edema, cellular proliferation, and accumulation of collagen. Fibers do migrate to the pleural
space, and it has been hypothesized that a similar cascade of inflammatory events may contribute
4-88
-------
to fibrotic lesions in the visceral pleura. Thickening of the visceral pleura is more often localized
to lobes of the lung with pronounced parenchymal changes, and it has also been hypothesized
that the inflammatory and fibrogenic processes within the lung parenchyma in response to
asbestos fibers may influence the fibrogenic process in the visceral pleura. The pathogenic
mechanism(s) through which mineral fiber exposure effects parietal plaques is largely unknown.
There is currently insufficient evidence to establish the noncancer MOA for LAA.
Limited in vitro studies have demonstrated oxidative stress following LAA exposures in various
cell types (Duncan et al., 2010; Hillegass et al., 2010; Pietruska et al., 2010; Blake et al., 2007).
LAA fibers increased intracellular ROS in both murine macrophages and human epithelial cells
(Duncan et al., 2010; Blake et al., 2007). Surface iron and inflammatory marker gene expression
was increased following exposure to LAA in human epithelial cells [(Shannahan et al., 2012a;
Shannahan et al., 2012c; Shannahan et al., 2012b; Shannahan et al., 2012d; Shannahan et al.,
201 Ib; Duncan et al., 2010; Pietruska et al., 2010); see Table 4-18]. Tremolite studies
demonstrate cytotoxicity in various cell culture systems (see Table 4-22).
The initial stages of any fibrotic response involve cellular proliferation, which may be
compensatory for cell death due to cytotoxicity. Analysis of cellular proliferation has
demonstrated both increases and decreases following exposure to asbestos fibers in vitro and in
vivo depending on the specific fiber or cell type (Mossman et al., 1985; Topping and Nettesheim,
1980). Other studies have focused on the activation of cell-signaling pathways that lead to
cellular proliferation following exposure to asbestos (Scapoli et al., 2004; Shuklaetal., 2003;
Dingetal., 1999; Zanella et al., 1996; Zhang et al., 1993).
Although slightly increased compared to controls, cytotoxicity in murine macrophage
cells exposed to LAA was decreased compared to other fiber types (Blake et al., 2008).
Cytotoxicity was slightly, but statistically significantly, increased compared to an unexposed
control at 24 hours postexposure to LAA, while crocidolite exposure resulted in even higher
levels of cytotoxicity. No other in vitro study examined cytotoxicity following exposure to
LAA, although an increase in apoptosis was demonstrated in this same cell system (Blake et al.,
2008). Recent studies in mice exposed to LAA demonstrated increased collagen deposition and
collagen gene expression, markers of fibrosis (Smartt et al., 2010; Putnam et al., 2008).
Short-term-duration studies in rats also demonstrated an increased inflammatory response
(Padilla-Carlin et al., 2011; Shannahan et al., 201 la; Shannahan et al., 201 Ib). Tremolite or
LAA exposure both led to increases in fibrosis in all but one animal study, supporting a role for
proliferation in response to these fibers. Taken together with studies on other asbestos fibers,
these data suggest that cytotoxicity and cell proliferation may play a role in the noncancer health
effects following exposure to LAA.
Although continued research demonstrates that the LAA has biologic activity consistent
with the inflammatory action and cytotoxic effects seen with other forms of asbestos, the data are
not sufficient to establish a MOA for the pleural and/or pulmonary effects of exposure to LAA.
4-89
-------
4.6. EVALUATION OF CARCINOGENICITY
4.6.1. Summary of Overall Weight of Evidence
Under the EPA Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), LAA is
carcinogenic to humans following inhalation exposure based on epidemiologic evidence that
shows a convincing association between exposure to LAA and increased lung cancer and
mesothelioma mortality (Larson et al., 201 Ob: Moolgavkar et al., 2010; Sullivan, 2007;
McDonald et al., 2004; Amandus and Wheeler, 1987; McDonald et al., 1986a). These results are
further supported by animal studies that demonstrate the carcinogenic potential of LAA and
tremolite asbestos in rodent bioassays (see Section 4.1, 4.2, Appendix D). As a durable mineral
fiber of respirable size, this conclusion is consistent with the extensive published literature that
documents the carcinogenicity of amphibole fibers [as reviewed in (Aust et al., 2011; Broaddus
etal.. 2011: Bunderson-Schelvan et al.. 2011: Huang et al.. 2011: Mossman etal.. 2011)1.
EPA's Guidelines for Carcinogenic Risk Assessment (U.S. EPA, 2005a) indicate that for
tumors occurring at a site other than the initial point of contact, the weight of evidence for
carcinogenic potential may apply to all routes of exposure that have not been adequately tested at
sufficient doses. An exception occurs when there is convincing information (e.g., toxicokinetic
data) that absorption does not occur by other routes. Information on the carcinogenic effects of
LAA via the oral and dermal routes in humans or animals is absent. The increased risk of lung
cancer and mesothelioma following inhalation exposure to LAA has been established by studies
in humans, but these studies do not provide a basis for determining the risk from other routes of
exposure. Mesothelioma occurs in the pleural and peritoneal cavities and, therefore, is not
considered a portal-of-entry effect. However, the role of indirect or direct interaction of asbestos
fibers in disease at these extrapulmonary sites is still unknown. No information exists on the
translocation of LAA to extrapulmonary tissues following either oral or dermal exposure, and
limited studies have examined the role of these routes of exposure in cancer. Therefore, LAA is
considered carcinogenic to humans by the inhalation route of exposure.
4.6.1.1. Synthesis of Human, Animal, and Other Supporting Evidence
Libby, MT workers have been the subject of multiple mortality studies demonstrating
increased cancer mortality in relation to estimated fiber exposure. Occupational studies
conducted in the 1980s (Amandus and Wheeler, 1987: McDonald et al., 1986a) as well as the
extended follow-up studies published in more recent years (Larson et al., 201 Ob: Sullivan, 2007:
McDonald et al., 2004) and additional analyses of the extended follow-up (Moolgavkar et al.,
2010) provide evidence of an increased risk of lung cancer mortality and of mesothelioma
mortality among the workers exposed to LAA in the Libby vermiculite mining and processing
operations. This pattern is seen in the lung cancer analyses using an internal referent group in
the larger follow-up studies (Larson etal.. 201 Ob: Sullivan. 2007: McDonald et al.. 2004). with
cumulative exposure analyzed using quartiles or as a continuous measure, and in the studies
4-90
-------
reporting analyses using an external referent group [i.e., standardized mortality ratios; (Sullivan,
2007: Amandus and Wheeler. 1987: McDonald et al.. 1986a)1. McDonald et al. (2004) also
reported increasing risk of mesothelioma across categories of exposure; the more limited number
of cases available in earlier studies precluded this type of exposure-response analysis. This
association is also supported by the case series of 11 mesothelioma patients among residents in
or around Libby, MT, and among family members of workers in the mining operations
(Whitehouse et al., 2008), and by the observation of three cases of mesothelioma (two of which
resulted in death) in the Marysville, OH worker cohort identified as of June 2011 (Dunning et al.,
2012).
Although experimental data in animals and data on toxicity mechanisms are limited for
LAA, tumors that were observed in animal tissues were similar to those in humans (e.g.,
mesotheliomas, lung cancer) indicating the existing data are consistent with the cancer effects
observed in humans exposed to LAA. Smith (1978) reported an increased incidence of
mesotheliomas in hamsters after intrapleural injections of LAA. Additionally, studies in
laboratory animals (rats and hamsters) exposed to tremolite via inhalation (Bernstein et al.,
2005b: Bernstein etal., 2003: Davis etal., 1985), intrapleural injection (Roller et al., 1997, 1996:
Davis etal., 1991: Wagner etal., 1982: Smith etal., 1979), or implantation (Stanton et al., 1981)
have shown increases in mesotheliomas and lung cancers. Tremolite from various sources was
used and varied in fiber content and potency (see Section 4.2, Appendix D). The most sensitive
model for mesothelioma induction is the Syrian golden hamster following asbestos inhalation,
with different susceptibility between species attributed to more rapid translocation to the pleural
space (Donaldson et al., 2010). Although McConnell et al. (1983a) observed no increase in
carcinogen!city following oral exposure to nonfibrous tremolite, the ability of this study to
inform the carcinogenic potential of fibrous tremolite through inhalation is unclear, and these
study results contribute little weight to the evaluation of the carcinogen!city of fibrous LAA.
The available mechanistic information suggests LAA induces effects that may play a role
in carcinogenicity (see Section 4.2 Appendix D). Several in vitro studies have demonstrated
oxidative stress and genotoxicity following LAA exposures in various cell types (Duncan et al.,
2010: Hillegass et al., 2010: Pietruska et al., 2010: Blake et al., 2007). LAA increased
intracellular ROS in both murine macrophages and human epithelial cells (Duncan et al., 2010;
Blake et al., 2007). Additionally, surface iron, inflammatory marker gene expression, and
aneugenic micronuclei were increased following exposure to LAA in human epithelial cells
(Duncan etal., 2010; Pietruska et al., 2010). Tremolite studies demonstrate cytotoxic and
clastogenic effects (e.g., micronucleus induction and chromosomal aberrations) of the fibers in
various cell culture systems.
In summary, the epidemiologic data demonstrate an association between exposure to
LAA and increased cancer risk. Supporting evidence of carcinogenic potential was observed in
the limited number of laboratory animal studies exposed to LAA or tremolite (see Tables 4-19
4-91
-------
and 4-20 summarizing in vivo studies). Overall, the available evidence supports the conclusion
that LAA is carcinogenic to humans.
4.6.2. Mode-of-Action Information (Cancer)
4.6.2.1. Description of the Mode-of-Action Information
EPA guidance provides a framework for analyzing the potential mode(s) of action by
which physical, chemical, and biological information is evaluated to identify key events in an
agent's carcinogen! city (U.S. EPA, 2005a). Agents can work through more than one MO A, and
MOA can differ for various endpoints (e.g., lung cancer vs. mesothelioma). Reasonably, the
analysis of a MOA would start with some knowledge of an agent's biological activity that leads
to cellular transformation resulting in cancer. Although early steps in the process often can be
identified, carcinogenicity is a complex process resulting from multiple changes in cell function.
Due to the limited data available specific to LAA, the MOA of LAA for lung cancer and
mesothelioma following inhalation exposure cannot be established.
Occupational studies demonstrate human health effects (e.g., lung cancer, mesothelioma)
following exposure to LAA. Although the limited mechanistic data demonstrate biological
effects similar to those of other mineral fibers following exposure to LAA, the existing literature
is insufficient to establish a MOA for LAA for lung cancer or mesothelioma. These biological
effects following exposure to LAA and/or tremolite are demonstrated in a limited number of
laboratory animal and in vitro studies. Multiple key events for one particular MOA have not
been identified; therefore, the MOA for LAA carcinogenicity cannot be established. However,
multiple mechanisms of action (e.g., mutagenicity, chronic inflammation, cytotoxicity, and
regenerative proliferation) can be hypothesized based on the available asbestos literature. These
are described in Section 4.4, and discussed below.
4.6.2.2. Evidence Supporting a Mutagenic Mode of Action
Strength, consistency and specificity of the association
Only limited genotoxicity analysis following exposure to LAA or tremolite has been
reported, although studies of other types of asbestos fibers have shown increases in genotoxicity
both in vitro and in vivo [reviewed in (Huang et al., 2011)1. One in vitro study examined
genotoxicity of LAA by measuring DNA adduct formation following exposure via murine
macrophages [primary and immortalized; (Blake et al., 2007)]. The data showed no increase in
adduct formation as compared to unexposed controls. A second study observed increases in
micronuclei induction in both normal human lung epithelial cells and XRCC1-deficient cells for
both Libby Amphibole and crocidolite asbestos (Pietruska et al., 2010). Two studies of tremolite
examined genotoxicity. The first found no significant increase in revertants in the Ames assay
(Athanasiou et al., 1992), which is similar to results obtained for other forms of asbestos. This
study did find, however, that tremolite exposure led to a dose-dependent increase in chromosome
4-92
-------
number and micronuclei formation, which has also been described for other asbestos fibers [as
reviewed in (Hei et al., 2006; Jaurand, 1997)]. Hei and colleagues (Okayasu et al., 1999)
performed mutation analysis with tremolite and found a dose-dependent increase in mutations in
CD59 in hamster hybrid cells. Although LAA- and tremolite-specific data are limited to in vitro
studies, given the similarities in response to other forms of asbestos, there is some evidence to
suggest genotoxicity following exposure to Libby Amphibole and tremolite asbestos.
Dose-response concordance and temporal relationship
A dose-response concordance has not been established between the development of
genotoxicity and exposure to LAA or other amphibole asbestos. Genotoxicity studies of other
amphibole asbestos have examined gene mutations and chromosomal mutations, as well as DNA
damage resulting from ROS production following exposure. As recently reviewed by Huang et
al. (2011), there are a large number of in vitro studies that support possible genotoxic
mechanisms following exposure to fibers. There are fewer in vivo studies of the genotoxicity of
amphibole asbestos, and a very limited number of these were following inhalation exposure.
Some of these studies were performed in nonrelevant cell types for inhalation endpoints, and
some also were performed at doses higher than observed in environmental or occupational
asbestos exposures. Temporal relationship would be impacted by direct or indirect genotoxic
mechanism playing a role in asbestos-induced turnorigenesis. There is insufficient data to
conclude whether the observed genotoxic effects following exposure to amphibole asbestos
result from direct (e.g., spindle interference) or indirect (e.g., ROS production) mechanisms. The
available evidence suggests a role for both direct and indirect genotoxicity, but requires further
research (Huang et al., 2011). Therefore, although these results suggest a possible role for
genotoxicity in the MO A of LAA, dose-response concordance and a temporal relationship are
difficult to determine.
Biological plausibility and coherence
Although only limited genotoxicity studies of LAA and tremolite have been published,
these studies are supported by similar results for other amphibole asbestos studies that
demonstrate genotoxicity in both in vitro and in vivo studies [as reviewed in (Huang et al.,
2011)]. Although studies of asbestos genotoxicity need to be carefully reviewed to determine
relevance of routes of exposure, target cell, and dose, taking these parameters into account, this
review of these studies supports the biological plausibility of the genotoxicity of LAA.
4.6.2.3. Evidence Supporting Mechanisms of Action of Chronic Inflammation, Cytotoxicity,
and Cellular Proliferation
Chronic inflammation has been observed following fiber exposure, which is often
followed by fibrosis at the site of inflammation if the fibers persist [reviewed in (Mossman et al..
4-93
-------
2011)1. Macrophages phagocytose fibers and paniculate matter and are activated to trigger the
release of inflammatory cytokines, ROS, and growth factors. These responses lead to a sustained
inflammatory response that can result in fibrosis at the site of fiber deposition. Chronic
inflammation is hypothesized to contribute to a carcinogenic response through the production of
ROS and increased cellular proliferation (Hanahan and Weinberg, 2011).
The initial stages of any fibrotic response involve cellular proliferation, which may be
compensatory for cell death due to cytotoxicity. The same may be true for turnorigenicity, as
increased cell proliferation can increase the chance of cancer by increasing the population of
spontaneous mutations affording genotoxic effects an opportunity to multiply. Analysis of
cellular proliferation of epithelial cells has demonstrated both increases and decreases following
exposure to asbestos fibers in vitro and in vivo (Mossman et al., 1985; Topping and Nettesheim,
1980). Other studies have focused on the activation of cell-signaling pathways that lead to
cellular proliferation following exposure to asbestos (Scapoli et al., 2004; Shukla et al., 2003;
Ding et al.. 1999: Zanella et al.. 1996).
The inflammatory response to fibers in vivo has been studied following inhalation
exposure to many types of fibers but not for LAA [reviewed in (Broaddus et al., 2011; Mossman
et al., 2011; Mossman et al., 2007)1. Results following inhalation exposure to tremolite have
demonstrated increased inflammatory response as early as 1 day postexposure (Bernstein et al.,
2005b: Bernstein et al., 2003). Earlier data from Davis et al. (1985) following inhalation
exposure to other forms of tremolite showed increased fibrosis and carcinogenesis; however,
inflammatory response was not described. In vivo studies of LAA and tremolite through other
routes of exposure have demonstrated increased inflammation following exposure (Padilla-
Carlin et al., 2011; Shannahan et al., 201 la: Shannahan et al., 201 Ib). Inhalation studies
examining other types of asbestos (crocidolite, chrysotile, and amosite) have clearly
demonstrated an increase in chronic inflammation and respiratory cancer related to exposure
[reviewed in (Mossman et al., 2011)]. This effect is observed in animal studies for LAA and
tremolite and is relevant to humans based on similar responses in cohorts analyzed (Musk et al.,
2008: Hein et al., 2007: Levin etal., 1998).
Although limited, the data described here for LAA, and to a greater extent for tremolite,
suggest a similar response as to other amphibole asbestos. In vivo exposure to tremolite led to an
increase in inflammation for all studies where it was measured. This increase appeared in some
cases to depend on fiber size and morphology (Davis et al., 1991: Smith et al., 1979). In vitro
analysis of LAA showed increases in inflammatory cytokines (Hamilton et al., 2004) and in
proinflammatory gene expression (Duncan et al., 2010). Bernstein et al. (2005b) and Bernstein
et al. (2003) observed that exposure to tremolite led to pronounced inflammation as soon as
1 day after inhalation exposure in male Wistar rats. Inflammation also occurred in male albino
Swiss mice in an acute-duration study that did not lead to fibrosis or carcinogenesis, possibly due
to the short study duration [150 days (Sahu etal., 1975)].
4-94
-------
Chronic inflammation has also been associated with increased ROS production [reviewed
in (Aust et al., 2011; Kamp et al., 1992)]. Fibers can directly lead to the production of ROS by
iron-catalyzed generation through the Fenton reaction. ROS are also produced following
phagocytosis of fibers. ROS production following exposure to asbestos has been shown to be
associated with DNA damage (described below), chronic inflammation, and lipid peroxidation.
As described in the previous section, chronic inflammation may lead to increased cell
proliferation and DNA damage, which in turn may lead to tumor formation. The hydroxyl
radical produced has been shown to directly interact with DNA (Leanderson et al., 1988).
ROS production has been measured in response to both LAA and tremolite exposure.
The study of LAA (Blake et al., 2007) demonstrated an increase in superoxide anions, not
hydrogen peroxide, as has been demonstrated with crocidolite. Blake et al. (2007) also
demonstrated that total SOD was inhibited following exposure to LAA, along with a decrease in
intracellular glutathione. These results are supported by a recent study in human mesothelial
cells (Hillegass et al., 2010). Further, increased ROS production was also observed in human
airway epithelial cells following exposure to LAA (Duncan et al., 2010). This increase in ROS
and decrease in glutathione are common effects following exposure to asbestos fibers and
particulate matter. Limited studies, however, have examined the specific type of ROS produced
following exposure to each type of asbestos.
A dose-response concordance has not been established between the development of
chronic inflammation and exposure to LAA. Dose-response information is limited to inhalation
studies of other amphibole asbestos, which were recently reviewed (Case et al., 2011; Mossman
et al., 2011). Many of the early studies of amphiboles described above were performed using
only one dose. Therefore, while these studies demonstrate an exposure-response relationship
between amphibole asbestos and chronic inflammation, dose-response concordance cannot be
determined.
A temporal relationship has not been established between the development of chronic
inflammation and inhalation exposure to LAA. Chronic inhalation studies of tremolite
demonstrate an increase in chronic inflammation over time (Bernstein et al., 2005b: Bernstein et
al., 2003) that may lead to fibrosis, or possibly tumor formation. A similar pattern has also been
observed in inhalation studies with other amphibole asbestos [as reviewed in Mossman et al.
(2011)1.
Chronic inflammation following exposure to fibers has been associated with the
development of both malignant and nonmalignant lung and pleural diseases (Bringardner et al.,
2008; Mossman and Churg, 1998). In vivo and in vitro studies have shown increases in
inflammation and inflammatory markers following exposure to LAA and tremolite up to 1 month
following single intratracheal instillation exposures in animal models (see Section 4.2,
Appendix D). Although inhalation studies are limited, the results of those studies demonstrate an
increase in chronic inflammation overtime, similar to studies of other amphibole asbestos fibers
4-95
-------
(Mossman et al., 2011). Overall, the evidence described above suggests chronic inflammation is
observed following Libby Amphibole and tremolite asbestos exposure.
Although slightly increased compared to controls, cytotoxicity in murine macrophage
cells exposed to LAA was decreased compared to other fiber types (Blake et al., 2008). No other
in vitro study examined cytotoxicity following exposure to LAA, although an increase in
apoptosis was demonstrated in this same cell system (Blake et al., 2008).
Compensatory proliferation in epithelial cells following cytotoxicity can lead to an
increase in mutations (both spontaneous and induced). This increase is generally offset by
increased levels of apoptosis, as in Blake et al. (2008). Recent studies in mice exposed to LAA
demonstrated increased collagen deposition and collagen gene expression, markers of fibrosis
(Smartt et al., 2010; Putnam et al., 2008). Tremolite and LAA exposure led to increases in both
fibrosis and turnorigenicity in all but one animal study, supporting a role for proliferation in
response to these fibers. Taken together with studies on other asbestos fibers, these data suggest
that cytotoxicity and cell proliferation may play a role in tumor formation.
Neither a dose-response concordance nor temporal relationship has been established
between the development of cytotoxicity and regenerative cellular proliferation and exposure to
LAA. However, cytotoxicity and regenerative cellular proliferation has been observed following
exposure to LAA as well as other amphibole asbestos through other routes of exposure in in vivo
assays (e.g., intratracheal instillation). Also, increases in markers of proliferative response have
been observed in in vitro studies of LAA and other amphibole asbestos in epithelial cells. These
results suggest exposure to LAA may lead to increases in cytotoxicity and regenerative
proliferation; however, the data are not sufficient to determine a dose-response relationship.
It is generally accepted that sustained cell proliferation in response to cytotoxicity can be
a significant risk factor for cancer (Correa, 1996). Sustained cytotoxicity and regenerative cell
proliferation may result in the perpetuation of mutations (spontaneous or directly or indirectly
induced by the chemical), resulting in uncontrolled growth. It is also possible that continuous
proliferation may increase the probability that damaged DNA will not be repaired. Reparative
proliferation alone is not assumed to cause cancer. Tissues with naturally high rates of turnover
do not necessarily have high rates of cancer, and tissue toxicity in animal studies does not
invariably lead to cancer. Nevertheless, regenerative proliferation associated with persistent
cytotoxicity appears to be a risk factor of consequence.
4.6.2.4. Conclusions About the Hypothesized Modes of Action
Is the hypothesized mode of action sufficiently supported in the test animals?
There are a limited number of studies on the genotoxicity of LAA and/or tremolite.
However, the studies described in Section 4 suggest a possible role for mutagenicity in
asbestos-induced carcinogenicity. These studies showed chromosomal aberrations, increases in
micronuclei induction, and increased ROS production which has been shown to lead to
4-96
-------
mutagenicity (see Section 4, Table 4-21). One study of DNA adduct formation did not show any
DNA damage or adducts following exposure to LAA. Laboratory animal studies of other
amphibole asbestos have demonstrated similar results (Huang et al., 2011). Further research in
this area is needed in order to inform the possibility of a mutagenic MOA for LAA.
Chronic inflammation is observed following exposure to most fibers studied (Mossman et
al., 2011). Laboratory animal studies of LAA and tremolite demonstrated increases in
inflammation, inflammatory markers, and increases in inflammatory cells. Further, in vitro
studies have shown that exposure to LAA and tremolite lead to increases in expression of
inflammatory cytokines. Available data are limited but consistent with the hypothesis that a
MOA involving chronic inflammation contributes to asbestos-induced pulmonary and pleural
tumors, either independently or in combination with a mutagenic MOA. However, it has not
been determined whether chronic inflammation is a necessary precursor of carcinogenesis, and
experimental support for causal links, such as compensatory cellular proliferation or clonal
expansion of initiated cells, is lacking between toxicity and pulmonary or pleural tumor
formation. However, further research is needed to determine if this MOA could be established
for LAA and/or tremolite.
As reviewed in Section 4.2, in vivo and in vitro studies have shown a consistent cytotoxic
and proliferative response to LAA and/or tremolite. Therefore, it has been proposed that
cytotoxicity following pulmonary exposure to LAA and/or tremolite is a precursor to
carcinogenicity. A more biologically plausible MOA may involve a combination of chronic
inflammation, genotoxicity, and cytotoxicity, with genotoxicity increasing the rate of mutation
and regenerative proliferation enhancing the survival or clonal expansion of mutated cells.
However, this hypothesis has yet to be tested experimentally.
Is the hypothesized mode of action relevant to humans
Although limited for LAA, the evidence discussed above demonstrates that LAA
exposure results in genotoxicity in in vitro and in vivo studies in test animal species. Therefore,
the presumption is LAA would be genotoxic in humans. The few available data from human and
in vivo laboratory animal studies concerning the genotoxicity of amphibole asbestos suggest
consistency with this mechanism, but the studies are not sufficiently conclusive to provide direct
supporting evidence for a mutagenic MOA. This MOA is considered relevant to humans.
The evidence discussed above demonstrates that exposure to LAA and/or tremolite
asbestos lead to chronic inflammation. The available human data following exposure to LAA
and other amphibole asbestos suggest consistency with this mechanism being relevant to
humans. Data are inadequate to determine that a cytotoxic mechanism is operative following
exposure to LAA in exposed populations; however, none of the available data suggest that this
mechanism is biologically precluded in humans. Furthermore, both animal and in vitro studies
4-97
-------
suggest that LAA causes cytotoxicity at exposures that may induce pulmonary cancers,
constituting positive evidence of the human relevance of this hypothesized MOA.
Which populations or life stages can be particularly susceptible to the hypothesized mode of
action
A mutagenic MOA is considered relevant to all populations and life stages. According to
EPA's Cancer Guidelines (U.S. EPA, 2005a) and Supplemental Guidance (U.S. EPA, 2005bX
there may be increased susceptibility to early-life exposures for carcinogens with a mutagenic
MOA. The weight of evidence is insufficient to support a mutagenic MOA for LAA
carcinogenicity and in the absence of chemical-specific data to evaluate differences in
susceptibility, 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.
Populations that may be more susceptible include those with varied fiber toxicokinetics
related to potential anatomical, physiological, and biochemical differences which may impact
fiber dosimetry (see Section 4.7). No data are available as to whether other factors may lead to
different populations or life stages being more susceptible to the hypothesized MOA for
LAA-induced tumors (e.g., chronic inflammation, cytotoxicity or mutagenicity). For instance, it
is not known how the hypothesized key events in chronic inflammatory response (e.g., increased
oxidative stress) to fibers interact with known risk factors for human pulmonary or pleural
carcinomas.
As with chronic inflammation, populations that may be more susceptible to increased
cytotoxicity following exposure to LAA include those that may have varied fiber toxicokinetics
related to potential anatomical, physiological, and biochemical differences which may impact
fiber dosimetry (see Section 4.7). No data are available as to whether other factors may lead to
different populations or life stages being more susceptible to a cytotoxic MOA for LAA-induced
tumors. For instance, it is not known how the hypothesized key events (e.g., interference with
the spindle apparatus) in this MOA interact with known risk factors for human pulmonary or
pleural carcinomas.
Summary
Research on multiple types of elongate mineral fibers supports the role of multiple modes
of action following exposure to LAA. Of the MO As described above, the evidence that chronic
inflammation, genotoxicity and cytotoxicity, and cellular proliferation may all play a role in the
carcinogenic response to LAA is only suggestive (see Table 4-23). In vitro studies provide
evidence that amphibole asbestos is capable of eliciting genotoxic and mutagenic effects in
mammalian respiratory cells; however, direct evidence of mutagenicity in respiratory cells
following inhalation exposure is lacking. Results of the in vivo studies described here are
4-98
-------
consistent with the hypothesis that some forms of amphibole asbestos act through a MOA
dependent on cellular toxicity. This conclusion is largely based on the observations that
cytotoxicity and reparative proliferation occur following subchronic exposure and that
bronchiolar tumors are produced at exposure levels that produce cytotoxicity and reparative
proliferation. However, dose-response data in laboratory animal studies for damage/repair and
tumor development are limited because of the lack of inhalation studies using multiple doses of
fibers that exist. Although evidence is generally supportive of a MOA involving chronic
inflammation or cellular toxicity and repair, there is insufficient evidence to support these
hypotheses; thus, a linear approach is used to calculate the inhalation cancer unit risk in
accordance with the default recommendation of the 2005 Guidelines for Carcinogen Risk
Assessment (U.S. EPA, 2005a). It is possible that multiple MO As discussed above, or an
alternative MOA, may be responsible for tumor induction.
4-99
-------
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.
4-100
-------
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.
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.
4-101
-------
4.7. SUSCEPTIBLE POPULATIONS
Certain populations may be more susceptible to adverse health effects from exposure to
LAA. Because the adverse health effects resulting from exposure to LAA have been primarily
studied in occupational cohorts of adult white men (see Sections 4.1.1 and 4.1.3), there is limited
information on the effects to a broader population. A few studies, however, have examined
health effects resulting from nonoccupational exposure in other age groups, genders (i.e.,
females), and races or ethnicity groups. The data from these studies could inform whether any
differential risk exists for these groups (see Sections 4.1.2 and 4.1.4). However, note that
distinguishing true differences from chance variation in effect estimates is related to the sample
size and statistical power, which is usually limited in these studies. In addition, genetic
polymorphisms, preexisting health conditions, and differences in nutritional status may alter an
individual's response to LAA. Finally, coexposures to other substances (e.g., tobacco smoke or
particulate matter) may increase an individual's risk of adverse health effects from exposure to
LAA. When data are available, each of these factors is discussed below with respect to increased
susceptibility to cancer and noncancer effects from exposure to LAA. When information
specific to LAA is not available, the general literature on the toxicity of mineral fibers is briefly
referenced.
There are also factors that may influence one's exposure potential to asbestos based on
life stage or other characteristics. For example, children spend more hours outside and may
engage in activities which impact exposure potential compared to adults (U.S. EPA, 2006b:
NRC, 1993). Because life stage and activity patterns can increase the potential for health effects
from exposure, these factors define who may be more susceptible to health effects due to greater
exposure. Section 2.5 discusses this exposure potential, including how children, workers,
household contacts, and residents may be exposed to LAA.
4.7.1. Influence of Different Life Stages on Susceptibility
Individuals at different life stages differ from one another physiologically, anatomically,
and biochemically. Individuals in early and later life stages differ markedly from adulthood in
terms of body composition, organ function, and many other physiological parameters, which can
influence the toxicokinetics and toxicodynamics of chemicals and their metabolites in the body
(Guzelian et al., 1992). This also holds true for mineral fibers, including asbestos fibers (see
Section 3). This section presents and evaluates the literature on how individuals in early or later
life stages might respond differently and thus potentially be more susceptible to adverse health
effects of LAA exposure.
4-102
-------
4.7.1.1. Life-Stage Susceptibility
Humans in early life stages (i.e., conception through adolescence) can have unique
susceptibilities compared to those in later life stages because they undergo rapid physiological
changes during critical periods of development (Selevan et al., 2000). Furthermore, young
people are often exposed to xenobiotics via unique exposure pathways [i.e., transplacental
transfer and breast milk ingestion; (U.S. EPA. 2006b: NRC, 1993)1. The nature of these
alternate exposure pathways, and the lack of studies that accurately document exposure levels
and outcomes in the very young, contribute to the difficulty in assessing the relative
susceptibility of early life stage exposure to amphibole asbestos.
No in utero exposure data exist for LAA but limited observations in stillborn infants
indicate transplacental transfer of tremolite (Hague et al., 1998; Hague et al., 1996) and other
asbestos and nonasbestos fibers does occur (Hague etal., 1998; Hague et al., 1996; Hague et al.,
1992; Hague et al., 1991). Transplacental transfer of asbestos was also demonstrated in animals
following maternal exposure by gavage (Hague etal., 2001) or injection [(Hague and Vrazel,
1998; Cunningham and Pontefract, 1974); see Section 3]. These studies did not evaluate the
sources or levels of exposure, and injection studies are a less relevant route of exposure
compared to inhalation. Based on these studies, LAA fibers may be transferred through the
placenta, resulting in prenatal exposure at any stage of fetal development.
A number of studies have attempted to determine the impact of in utero and early life
exposure on the developing child. Those analyses performed in the very young include reports
of stillbirth (Hague et al., 1998; Hague et al., 1996) and death among infants and young children
(age 1-27 months) due to sudden infant death syndrome and bronchopulmonary dysplasia
(Hague and Kanz, 1988). These studies found higher levels of asbestos in the lungs of those who
died compared to unexposed individuals. In an infant study, the authors speculate that there was
either a preexisting abnormal lung physiology in these children that contributed to a reduced
ability to clear fibers from the lung, or the children had an increased exposure to asbestos (Hague
and Kanz, 1988). Those studies conducted in older children include reports of pleural and
diaphragmatic calcifications (Epler etal., 1980) and altered immune and respiratory conditions
(Shtol' et al., 2000). Although the data are suggestive of increased sensitivity in infants, no
definitive conclusion can be reached.
In experimental animal studies, the effects of in utero and early life exposure to asbestos
are equivocal. Rats' offspring that were exposed to tremolite had decreased body-weight gain at
weaning and 8-weeks old compared to controls (NTP, 1990b; McConnell et al., 1983a). This
finding was observed in similar studies with other forms of asbestos (NTP, 1990a, 1988, 1985;
McConnell et al., 1983a) but not replicated in others (McConnell et al., 1983b; NTP, 1983).
Embryonic toxicity was noted in a few experimental animal studies. Crocidolite injected into
pregnant mice resulted in altered limb differentiation in cultured embryos [(Krowke et al., 1983)
abstract], and chrysotile suspended in drinking water and given to pregnant mice resulted in
4-103
-------
decreased postimplantation survival in cultured embryos (Schneider and Maurer, 1977).
However, chrysotile ingested via drinking water did not affect embryonic survival in vivo in
pregnant mice (Schneider and Maurer, 1977). Altogether, the data provide no clear evidence for
increased susceptibility following early life or in utero asbestos exposure.
Several studies have examined the susceptibility of asbestos exposure on young children,
including how fiber deposition is affected by the physiological differences in children's lungs.
Evidence suggests that fiber deposition is increased in the lungs of children compared with adults
(Bennett et al., 2008; Isaacs and Martonen, 2005; Asgharian et al., 2004; Phalen and Oldham,
2001; Oldham et al., 1997; Schiller-Scotland et al., 1994; Phalen etaL 1985). Nasal deposition
of particles was lower in children compared to adults—particularly during exercise (Becquemin
et al., 1991). The lung and nasal depositional differences are partially due to structural
differences across life stages that change the depositional pattern of different fiber sizes, possibly
altering the site of action, and resulting in differential clearance and subsequent health effects.
However, it is unclear whether the lung surface, body weight, inhalation volume, or exposure
patterns are most determinative of dose.
There are a few studies analyzing noncancer outcomes in children exposed to LAA. A
Libby medical screening program collected data on 7,307 participants, including 600 children
aged 10-17 years, which represents 8.2% of the cohort (Peipins et al., 2003). Pulmonary
function tests showed that none of these children had moderate or severely restricted lung
function (ATSDR, 2002, 2001b). This program also studied chest radiographs for those 18 years
or older (Noonan et al., 2006; Peipins et al., 2003; ATSDR, 200 Ib), but x-rays were not
conducted on children. Among 1,003 adolescents and young adults (ages 10 to 29) who were
-------
or cancers observed with early lifetime exposures not seen with adult exposures. There are no
published reports that can directly answer these questions for exposure to LAA. Few cancers
occurring in childhood have been documented in children exposed to any form of asbestos.
Examples of cases include a 17-year-old exposed to chrysotile and tremolite (Andrion et al.,
1994) and a 3-year-old exposed to chrysotile (Lieben and Pistawka, 1967), both of whom
developed mesothelioma. Notably, childhood mesothelioma may have an etiology that is
different from that of the disease seen in adults, further confounding interpretation of these data
(Cooper et al., 1989).
Studies involving populations exposed to other types of asbestos have yielded equivocal
results on the carcinogenic effects following exposures occurring earlier in life, and few evaluate
very early life exposures. One study in the United Kingdom described occupational exposure to
chrysotile, crocidolite, and amosite for a group of 900 women. First exposure from ages
15-24 years led to a higher relative mortality risk for lung and pleural cancer compared with
women who were first exposed at older ages (SMR 30 based on 12 observed and 0.4 expected,
SMR 8 based on 4 observed and 0.5 expected, and SMR 6.7 based on 6 observed and 0.9
expected in the first exposure at ages 15-24, 25-34, and >35 years, respectively; (Newhouse et
al., 1972). In a study in Wittenoom, Western Australia, 27 individuals were diagnosed with
mesothelioma who had been environmentally exposed to crocidolite (i.e., residents of the town
but not directly employed in the area's crocidolite mining and milling industry); 11 of these
subjects were <15 years old at the time of exposure (Hansen et al., 1998). One-third of all the
subjects were younger than 40 years old when diagnosed, but the authors found no increase in
mesothelioma mortality rates when analyzed by age at first exposure. However, risk was
significantly increased based on time from the first exposure, duration of exposure, and
cumulative exposure (Hansen et al., 1998). Additional studies of this cohort found that the
mesothelioma mortality rate was lower for those first exposed (based on age residence in the area
began) to crocidolite at ages <15 years [n = 24; mesothelioma mortality rate 47 per
100,000 person-year) compared with those first exposed at ages >15 years (« = 43; mesothelioma
mortality rate 112 per 100,000 person-year; (Reid et al., 2007)1. The hazard ratio for age at first
residential exposure of >15 years compared with <15 years was 3.83 (95% CI: 2.19, 6.71),
adjusting for cumulative exposure, gender, and an interaction term for gender and cumulative
exposure. Altogether, these studies do not clarify whether exposure during childhood yields
different adverse health effects compared with exposure during adulthood.
Relatively few studies have examined the effects of asbestos exposure in juvenile
animals. Oral exposure to nonfibrous tremolite did not increase tumors in the offspring of rats
compared to controls (NTP, 1990b; McConnell et al., 1983a). Similar studies of other forms of
asbestos reported an increase of various neoplasms in the offspring (NTP, 1990a, 1988, 1985;
McConnell etal., 1983b: McConnell etal., 1983a), but another study reported none (NTP, 1983).
No cancer bioassays have been performed in juvenile animals exposed to LAA. Based on these
4-105
-------
very limited and inconclusive studies on other forms of asbestos, no conclusions can be drawn
about differential risk of adverse health effects after early life stage exposure to LAA compared
to exposure during adulthood. It is unknown whether early life stage exposure compared to adult
exposure increases susceptibility for adult cancers, as measured by increased incidence, severity,
or disease progression, or by decreased latency.
Later life stage is generally defined as >65 years old. Because pulmonary function
(volume and rate of breathing) decreases with age (Weiss, 2010), increased deposition of fibers
in the lung from exposures in later life stages is unlikely. Older adults could be more susceptible
to the effects of LAA due to the gradual age-related decline in physiological processes. For
instance, clearance of fibers from the lung might be reduced because cough reflex and strength of
older adults is less effective and the cilia are less able to move mucus out of the airway (U.S.
EPA, 2006a). Additionally, decreased immune function, increased genetic damage, and
decreased DNA repair capacity can result in increased susceptibility with age (U.S. EPA, 2006a).
These age-associated alterations could decrease fiber-induced DNA damage repair but might
also reduce the incidence of fiber-induced DNA damage due to decreased phagocytosis or
inflammation. Specific data pertaining to age-varying effects of LAA on these processes are not
available.
Because the risk of many types of noncancer effects increases with age, an increasing rate
of specific diseases with increasing age can be expected among individuals exposed at some
point in their lives to LAA. Radiographic tests among those exposed to LAA show that older
age, which in some occupational settings may be highly correlated with time since first exposure
(TSFE), is one of the factors most associated with pleural or interstitial abnormalities (Rohs et
al.. 2008: Horton et al.. 2006: Muravov et al.. 2005: Peipinsetal.. 2003: AT SDR. 200 Ib:
Amandus et al., 1987a: McDonald et al., 1986b: Lockey et al., 1984). Abnormal radiographs
also increase with age in general population studies (Pinsky et al., 2006). In a community health
screening study, an increased risk of rheumatoid arthritis among individuals ages >65 years was
observed in relation to several measures reflecting exposure to LAA [e.g., worked for W.R.
Grace, used vermiculite for gardening; (Noonan, 2006)1. However, the available studies do not
provide a basis for evaluating the timing of the exposure in relation to these outcomes. No
conclusions can be drawn about differential risk of noncancer after later life stage exposure to
LAA compared to exposure earlier in life.
No studies assessing the carcinogenic effect of exposures occurring in older age groups
are available for LAA or other amphiboles. Note that health effects observed among individuals
exposed to LAA are likely to increase with age due to the long latency period for the exposure
response for asbestos and lung cancer and other chronic diseases. However, this type of
observation would not directly address the question of whether exposures at older ages have a
stronger or weaker effect compared with exposures at younger ages.
4-106
-------
4.7.2. Influence of Gender on Susceptibility
A discussion of gender-related differences in risk from asbestos exposure raises several
important issues, such as gender-related differences in exposure patterns, physiology, and
dose-response (Smith, 2002). For example, nasal breathing filters out particles, and men tend to
breathe less through their nose during exercise than women do (Bennett et al., 2003). Bennett et
al. (1996) showed a gender difference in fractional deposition (defined as the ratio of particles
not exhaled to total particles inhaled) of particles 2 um in mass median aerodynamic diameter.
This particle diameter is within the range of LAA particles reported in Table 2-2. This study
found that, in general, women had a greater retention of particles compared to men because men
had higher ventilation rates compared to women; however, the overall deposition rate was higher
in the men (Bennett et al., 1996).
Most occupational studies for LAA have examined the effects of exposure only in men
(Moolgavkar et al., 2010: Sullivan. 2007: McDonald et al.. 2004: Amandus et al.. 1988:
Amandus et al., 1987b: Amandus and Wheeler, 1987: McDonald et al., 1986a: McDonald et al.,
1986b). There is limited information specifically on women exposed to LAA. In the Libby, MT
community studies, no gender-related trends in mortality due to lung or digestive cancer were
observed (ATSDR, 2000). These limited data do not provide a basis for drawing conclusions
regarding gender-related differences in adverse health effects from LAA.
4.7.3. Influence of Race or Ethnicity on Susceptibility
Race and ethnicity often are used in medical and epidemiological studies to define
various groups of the population. These categories could be surrogates for differences in
exposure (e.g., occupation, socioeconomics, behavior) or biology (e.g., physiology, genetics), in
which case these factors may play a role in susceptibility as well. Nasal structure and lung
architecture can influence the depositional patterns for both particles and fibers. One study of
18 Caucasians (ages 8 to 30 years) and 14 African Americans (ages 8 to 25 years) reported
increased ventilation rates during exercise in the African Americans [matched on gender, age,
height, and weight; (Cerny, 1987)]. Another study (11 Caucasians and 11 African Americans,
ages 18 to 31 years) reported decreased nasal deposition efficiency (for particle sizes of 1-2 um,
which is in the range of those for LAA reported in Table 2-2) in African Americans compared to
Caucasians (Bennett and Zeman, 2005). Furthermore, nasal breathing during exercise occurred
less in Caucasians compared to African Americans in this study (Bennett et al., 2003).
Of the occupational and residential studies for LAA, the vast majority of subjects with
known race were white, precluding the ability to conduct an analysis of racial and
ethnicity-related differences in the mortality risks within the Libby worker cohort. In a study of
occupational exposure to chrysotile asbestos in a textile factor, lung cancer mortality risk in
relation to exposure was lower in nonwhite males (0.84, 95% CI: 0.52-1.27) compared to white
males (2.34, 95% CI: 1.94-2.79), although a statistically significant increase in SMR was
4-107
-------
observed for nonwhite males at high exposure levels [>120 fibers-yr/mL; (Hein et al., 2007)1.
This observed difference could be due to a lower prevalence of smoking among nonwhite
compared with white males (Hein et al., 2007).
4.7.4. Influence of Genetic Polymorphisms on Susceptibility
XRCC1 is a DNA damage repair gene. A recent study demonstrated that
XRCCJ-deficient cells exposed to LAA or crocidolite asbestos demonstrated increased levels of
micronuclei induction (Pietruska et al., 2010). Two other studies examined XRCC1
polymorphisms in relation to disease risk with other types of asbestos exposure. Zhao et al.
(2006) found no association between XRCC1 polymorphisms and asbestosis in asbestos-exposed
workers. A study by Dianzani et al. (2006), however, did find an association between XRCC1
and asbestos-induced lung disease in a population exposed to asbestos pollution. Further work is
necessary, with clear definitions of patient populations and their exposure levels, so that these
studies and others can be compared to determine if XRCCJ polymorphisms increase
susceptibility to adverse health effects following exposure to LAA.
Superoxide dismutases are free radical scavengers that dismutate superoxide anions to
oxygen and hydrogen peroxide. SODs are expressed in most cell types exposed to oxygen.
Several common forms of SODs occur and are named by the protein cofactor: copper/zinc,
manganese, iron, or nickel. A recent study observed no significant alterations in levels of
intracellular SOD following a 3-hour exposure to LAA in mice (Blake et al., 2007). Other
studies in humans and mice have examined SOD expression in relation to other types of asbestos
exposure. Manganese SOD activity was elevated in biopsies of human asbestos-associated
malignant mesothelioma, although no genotypic differences were found to be related to this
change in activity (Hirvonen et al., 2002). Other studies have focused on the role of extracellular
superoxide dismutase (EcSOD) and asbestos-induced pulmonary disease (Kliment et al., 2009;
Gao et al., 2008; Fattman et al., 2006; Tan et al., 2004). These studies have suggested a
protective effect of EcSOD, because mice that lack this form of SOD have increased sensitivity
to asbestos-induced lung injury (Fattman et al., 2006). Familial studies showing an unusually
high incidence of mesothelioma suggest that genetic factors might play a role in the etiology of
mesothelioma (Ugolini et al., 2008; Huncharek, 2002; Roushdy-Hammady et al., 2001), although
whether a genetic factor or a common environmental element leads to the similar responses in
these families is difficult to determine. Increased interest in the role of genetic factors in
asbestos-related health outcomes has led to several analytical studies on specific genetic
polymorphisms. A review of 24 published reports (19 studies) discusses the current state of
knowledge regarding genetic susceptibility associated with asbestos-related diseases (in
particular, malignant pleural mesothelioma). Results from several studies demonstrated an
association between asbestosis-related diseases and G<57M7-null polymorphism, whereas results
for other polymorphisms were conflicting (Neri et al., 2008). Some polymorphisms discussed in
4-108
-------
Neri et al. (2008) are in genes for N-acetyl-transferase 2; glutathione-S-transferases (GSTs);
SOD; CYP1A1, CYP2D6; neurofibromatosis 2 (Nf2); p53; andXRCCJ. Although occupational
asbestos exposure was assessed, the type of asbestos is generally unknown in these studies.
Limited animal studies have examined the role of genetic variations related to asbestos
exposure, including specific signaling pathways (Shukla et al., 2007), DNA damage repair (Lin
et al., 2000; Ni et al., 2000), and tumor suppressor genes (Vaslet et al., 2002; Kleymenova et al.,
1997; Marsella et al., 1997). Genetic alterations of particular interest for mesothelioma include
those involved in tumor suppression (p53, Nf2) and oxidative stress (SOD, GSTs). Nf2 andp53
are frequently altered in mesotheliomas, but no consistent mutations have been found (Cheng et
al., 1999; Mayall et al., 1999: Bianchi etal., 1995). Alterations in expression of antioxidant
enzymes like SOD and GST in mesothelioma can yield cells more resistant to oxidative stress as
compared to normal cells due to increased antioxidant activity (Ramos-Nino et al., 2002:
Rahman and MacNee, 1999). No studies that examine the role of cell-cycle control genes were
found following exposure to LAA. Additionally, no information on other genetic
polymorphisms in relation to disease risk among those exposed to LAA was identified in the
available literature.
4.7.5. Influence of Health Status on Susceptibility
Preexisting health conditions could potentially alter the biological response to asbestos
exposure. Mesothelioma risk has been hypothesized to be related to immune impairment
(Bianchi and Bianchi, 2008) and Simian virus 40 (SV40) exposure in humans (Carbone et al.,
2007: Kroczynska et al., 2006: Cristaudo et al., 2005: Foddis et al., 2002: Bocchetta et al., 2000:
Mavall etal., 1999). Coexposure to asbestos and SV40 has been associated with p53-related
effects in vitro (Foddis et al., 2002: Bocchetta et al., 2000: Mavall etal., 1999), and cell signaling
aberrations in vivo (Kroczynska et al., 2006: Cristaudo et al., 2005). However, the influence on
cancer risk is unknown, as these lines of research are not fully developed and have not been
applied specifically to LAA.
Obesity can compromise inhalation exposure, as increased particle deposition in the lungs
of overweight children (Bennett and Zeman, 2004) and adults (Graham et al., 1990) has been
observed. Individuals with respiratory diseases could have compromised lung function that
alters inhalation exposure to LAA. For example, individuals with chronic obstructive pulmonary
disease (COPD) have increased inhalation volume (Phalen et al., 2006) and increased fine
particle deposition (Phalen et al., 2006: Bennett et al., 1997: Kim and Kang, 1997) and retention
(Regnis et al., 2000). Similarly, studies have reported an increase in coarse particle
(aerodynamic diameter >5 um) deposition in individuals with cystic fibrosis (Brown and
Bennett, 2004: Brown etal., 2001). For people exposed to LAA, an increased risk for interstitial
lung abnormalities was observed for those with a history of pneumonia (Peipins et al., 2003). In
another study, bronchial asthma was examined as a potential confounding variable for
4-109
-------
asbestos-related effects on pulmonary function, although no confounding was observed
(Whitehouse. 2004).
4.7.6. Influence of Lifestyle Factors on Susceptibility
Smoking can impair clearance of particles from the lung (Camner, 1980; Cohen et al.,
1979) and increase deposition of asbestos fibers (Sekhon et al., 1995; McFadden et al., 1986).
These effects could lead to the retention of more inhaled asbestos fibers and for a longer period
of time in smokers compared to nonsmokers, even when controlling for initial exposure.
Evidence of smoking-related susceptibility to pulmonary effects of asbestos was reported by
Christensen and Kopylev (2012) using data from the O.M. Scott, Marysville, OH plant cohort
described by Rohs et al. (2008). The amount of LAA exposure required to elicit the same
increase in risk of localized pleural thickening was considerably lower (sixfold) for smokers
compared with nonsmokers.
No studies were identified that examined lifestyle factors specifically with respect to
LAA and cancer susceptibility. Lifestyle factors such as exercise, nutritional status, and smoking
habits could affect the biological effects of asbestos exposure through various mechanisms. For
example, those with more physically demanding jobs or those who regularly engage in vigorous
exercise might experience increased lung deposition from fine particles or fibers compared to
those with a more sedentary lifestyle (Phalen et al., 2006; Becquemin et al., 1991). Randomized
controlled trials of vitamin supplementation (beta-carotene and retinol) have been conducted for
asbestos-related lung cancer, but results do not support a protective effect (Cullen et al., 2005).
For lung cancer, a synergistic relationship between cigarette smoking and asbestos
exposure has been demonstrated (Wraith and Mengersen, 2007; Hammond et al., 1979; Selikoff
and Hammond, 1979). Research has suggested that asbestos fibers might also enhance the
delivery of multiple carcinogens in cigarette smoke, and that cigarette smoking decreases the
clearance mechanisms in the lungs and could, therefore, lead to an increase in fiber presence in
the lungs (Nelson and Kelsey, 2002). Smoking likely causes genetic alterations associated with
lung cancer (Landi et al., 2008) that might increase the carcinogenic risk from exposure to
asbestos. Benzo[a]pyrene, a component of tobacco, also has been observed to enhance the
carcinogenic effects of asbestos (Loli et al., 2004; Kimizuka et al., 1987; Mossman et al., 1984;
DiPaolo et al.. 1983: Mossman et al.. 1983: Reissetal., 1983).
4.7.7. Susceptible Populations Summary
A very limited amount of information is available on exposure to LAA early in life to
determine if early exposure could lead to increased risk of asbestos-induced disease later in life.
Due to the long latency period of some diseases in relation to asbestos exposure in general,
adverse effects are more likely to be observed with an increase in age or, more specifically, with
increased time since first exposure. Further, asbestos exposure during specific life stages may
4-110
-------
lead to alternate exposure pathways and health outcomes, but there are limited studies assessing
the effect of exposure to amphibole asbestos by specific life stage. In the absence of
chemical-specific data to evaluate differences in susceptibility by specific life stage, 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 (see Section 4.6.2.4 and 4.6.2.5). The number of women who have been
occupationally exposed to LAA is very small, and health risks have not been evaluated
specifically for this group. Differences between men and women in residential sources and types
of exposure (e.g., types of activities done in the household) also preclude the possibility of
drawing conclusions regarding the relative susceptibility of women compared with men to health
effects of exposure to LAA. Similarly, sufficient data are not available to draw conclusions
regarding racial or ethnic variation in susceptibility to diseases caused by exposure to LAA. In
addition, the potential modifying effects of genetic polymorphisms, preexisting health
conditions, nutritional status, and other lifestyle factors have not been studied, specifically as
related to exposure of LAA and health outcomes.
4-111
-------
5. EXPOSURE-RESPONSE ASSESSMENT
5.1. ORAL REFERENCE DOSE (RfD)
An oral RfD was not derived. Oral exposure was not assessed because inhalation is the
primary route of concern and oral data for Libby Amphibole asbestos20 (LAA) is lacking.
5.2. INHALATION REFERENCE CONCENTRATION (RfC)
An RfC is defined as "an estimate (with uncertainty spanning perhaps an order of
magnitude) of an exposure (including sensitive subgroups) that is likely to be without an
appreciable risk of adverse health effects over a lifetime" (U.S. EPA, 2002). Consequently,
studies that relate these adverse health effects to exposure levels are necessary for RfC
derivation. Preferred study characteristics for RfC derivation include adequate
exposure-response information, ideally with quantitative exposure estimates to distinguish
exposure levels in the study subjects, and adequate duration of follow-up to identify health
effects of interest.
Overview of the Methodological Approach
The noncancer effects which were evaluated in populations with exposure to LAA (see
Sections 4.1.2 and 4.1.3) are pulmonary effects (including asbestosis, pleural thickening
[localized or diffuse], and other nonmalignant respiratory disease), cardiovascular disease-related
mortality, and autoimmune effects. Localized pleural thickening (LPT) was deemed the most
sensitive and was thus selected as the critical effect to derive the RfC (see Section 5.2.2.3). A
benchmark response (BMR) of 10% extra risk was selected for exposure-response modeling (see
Section 5.2.2.5.
RfCs are based on human data when appropriate epidemiologic studies are available.
The general approach to developing an RfC from human epidemiologic data is to quantitatively
evaluate the exposure-response relationship for that agent to derive a specific estimate of its
effect on the risk of the selected outcome in the studied population. For the current assessment,
the first step was to identify the most appropriate data set available to quantitatively estimate the
effects of LAA exposure on pleural effects. Studies of three different cohorts provide such
quantitative exposure-response information. Two are of occupationally exposed cohorts. The
first one is Libby Workers (Larson et al., 2012a) and the second one is Marysville workers (Rohs
et al., 2008). The third consisted of community members with nonoccupational exposure, who
resided around the Western Minerals plant in Minneapolis (Alexander et al., 2012). Upon
20The 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.
5-1
-------
evaluating these three cohorts, the Marysville workers were selected as the most appropriate for
derivation of the RfC. The Libby workers had generally higher levels of occupational exposure
and additional, unquantified exposures outside of the workplace (i.e., residential exposures).
While data on a critical predictor of the risk of pleural thickening (the time since first exposure
[TSFE]) was available for the occupational exposures, information on TSFE for the residential
exposures from living in Libby, MT prior to working in the mining or related operations is not
known. Substantial uncertainty also exists in the exposure estimates for the Minneapolis study,
and although some information on residential history was used by the investigators, it is unclear
whether this information applied just to the residence or whether there was information on TSFE
for the different exposure routes [see (Kelly et al., 2006)].
Among the Marysville workers, there were differences in the availability of exposure and
health outcome data over time. No industrial hygiene measurements were available before 1972.
Health examinations were performed at two time points, 1980 and 2002-2005, using different
x-ray reading protocols and different film readers. Thus, the subgroup of workers with the
highest quality exposure and outcome evaluation information, was determined to be those
workers who were hired in 1972 or later, and who had health examinations performed in
2002-2005; this group was selected as the primary analytic data set for derivation of the RfC
(see Section 5.2.2.2). Once the relevant data describing a well-defined group of individuals
along with their exposures and health outcomes were selected, a suite of appropriate statistical
model forms was evaluated. Before performing any modeling, biological and epidemiological
features were considered to determine a priori which variables and which models would be most
suitable for the given exposure and health outcome (see Section 5.2.2.6.1). Based on these
considerations, the Dichotomous Hill model was considered to be the most flexible and
potentially most suitable model form; however, all model forms suitable for dichotomous
epidemiological data were examined. Each model was evaluated for adequate fit to the data,
with each person's individual-level exposures and outcomes modeled using a variety of exposure
metrics. Appropriate covariates, which may be important predictors of LPT risk, were evaluated
for potential confounding in the statistical model.
In the primary analytic data set (the subcohort of workers hired in 1972 or later), all
univariate models examined had adequate fit and for each model form, mean exposure was
shown to have the best relative fit (compared with either cumulative or residence time-weighed
exposure metrics). Among the model forms, the relative fits were comparable, and thus the
Dichotomous Hill model (with plateau fixed at 85%) using the mean exposure metric was
selected as the primary model for RfC derivation. When evaluating nonexposure-related
covariates, none were found to fit the criteria for a confounder (i.e., they were not associated
with both the outcome and the exposure) and were not significant predictors of LPT risk when
included in the final model. Time since first exposure, one of the key covariates evaluated, was
associated with the exposure in the primary analytic data set but not the outcome. This is likely
5-2
-------
because there was a relatively narrow range of TSFE values (i.e., low variability) in the primary
analytic data set. However, based on the epidemiological literature, TSFE is expected to be a
major predictor of LPT risk and thus an important variable in evaluating the exposure-response
relationship with LAA. Because inclusion of TSFE in the model did not improve model fit (and
it was not a significant predictor), alternative strategies were explored to incorporate the effect of
TSFE into the exposure-response model. EPA decided to use a hybrid modeling strategy. First,
the effect of TSFE was estimated in a larger subset of the Marysville workers (all those with
health evaluations in 2002-2005, regardless of hire date) with a broader range of TSFE values,
using the same model as for the primary analytic data set. Next, this estimated effect was carried
over to the model for the primary analytic data set (workers hired in 1972 or later with health
examinations in 2002-2005) as a fixed regression coefficient, and the benchmark concentration
and lower limit of the benchmark concentration (BMCL) estimated.
The BMCL from the "hybrid" modeling approach was used as the point of departure.
Uncertainty factors were then applied to derive an RfC (see Section 5.2.3). Alternative analyses
are presented in Section 5.2.4 and 5.2.5 with a summary in Section 5.2.6. Uncertainties in this
noncancer assessment are described in detail in Section 5.3.
5.2.1. Choice of Principal Study
5.2.1.1. Candidate Studies
While there are studies of health effects in humans, no studies in laboratory animals on
the inhalation route of exposure are suitable for derivation of an RfC because the available
animal studies lack adequate LAA exposure-response information and are of a short-term
duration.
Multiple studies have identified several noncancer health effects in humans that could be
considered as potential critical effects for the derivation of an RfC. The noncancer health effects
range in severity from mortality to pleural abnormalities. Five mortality studies of cohorts of
workers who mined, milled, and processed Libby vermiculite identified increased risk of
mortality from noncancer causes including nonmalignant respiratory disease—especially
asbestosis and chronic obstructive pulmonary disease (COPD) (Larson et al., 201 Ob: Sullivan,
2007: McDonald et al.. 2004: Amandus and Wheeler. 1987: McDonald et al.. 1986a)—as well as
cardiovascular disease (Larson et al., 201 Ob). Because an RfC is intended to be a level that is
likely to be without appreciable risk of deleterious effects, these mortality studies were not
considered as candidates for RfC derivation because other human studies exist that provide
evidence of an association between LAA and less severe outcomes generally occurring at lower
levels of exposure, such as parenchymal and pleural abnormalities. More detailed discussion of
the choice of the critical effect for the RfC is presented in Section 5.2.2.3 and Appendix I.
Studies conducted among two cohorts of occupationally exposed workers have shown
radiographic evidence of health effects on the lung and pleura (a thin tissue surrounding the lung
5-3
-------
and lining the chest cavity). These effects include pleural thickening and fibrosis of the lung
(Larson etal.. 2012a: Larson et al.. 2012b: Larson et al.. 201 Ob: Rohs et al.. 2008: Amandus et
al.. 1987a: McDonald et al., 1986b: Lockey et al., 1984). Studies of exposed community
members in Libby, MT and Minneapolis, MN have also reported evidence of health effects on
the lung and pleura [(Alexander et al., 2012: Weill et al., 2011: Muravov et al., 2005: Peipins et
al.. 2004b: Whitehouse. 2004: Peipins et al.. 2003): see Section 4.1.2].
Although data exist that define exposures from some activities in the Libby, MT
community studies (see Section 2.3), the available exposure data were insufficient to estimate
exposure at the individual level. Only studies that include exposure measurement data allowing
estimation of individual exposures and identify appropriate health effects are considered for RfC
derivation (Alexander et al., 2012: Larson et al., 2012a: Rohs et al., 2008: Amandus et al., 1987a:
McDonald et al., 1986b: Lockeyet al., 1984). Among these six candidate principal studies (see
Figure 5-1), one study was of the community surrounding a vermiculite processing facility in
Minneapolis, MN (Alexander et al., 2012), three were occupational studies of exposed workers
in Libby, MT (Larson et al.. 2012a: Amandus et al.. 1987a: McDonald et al.. 1986b). and two
were studies in workers from the Marysville, OH facility (Rohs et al.. 2008: Lockey et al.. 1984).
The studies by Larson et al. (2012a) and Rohs et al. (2008) represent the most recent evaluations
of the occupational studies of exposed workers in Libby, MT and Marysville, OH workers,
respectively, and were considered as candidate principal studies for the derivation of the RfC,
along with the study of the Minneapolis community by Alexander et al. (2012).
5-4
-------
Libby, MT
,
Minneapolis, MN
Occupation-based |
(vermiailite mining and milling workers; 1
WR Grace)
->
4
*
t
i
"t!
NIOSH
Amandus et aL, 1987a
[McGill University
McDonald etal.,1986b
x^-
ATSDR
Larson et aL, 2012a
t_ _J
\
1
/
/
Occupation-based
(fertilizer and other lawn products
production; OM Scott)
L. J
^
/'
/
( '
/* N
LockeyetaL, 1984
s, ;
Rohset al.,2008
'-• .-*'
\
I
1
1
/
Community -based
(area around Western Minerals
k insulation materials plant)
/'""
/ <
L Minnesota Dept of Health and ATSDR
« Alexander etal., 2012
i
X^
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.
Each study has adequate reporting of the studied populations, methods of assessment of
health outcome(s) of interest, and statistical analyses. Each study also demonstrated associations
between exposure to LAA and radiographic signs of nonmalignant respiratory effects,
specifically pleural thickening (circumscribed and/or localized and/or diffuse) and small
interstitial opacities (indicative of parenchymal damage) (ILO, 2002, 1980, 1971). Table 5-1
summarizes the candidate principal studies. See Section 4.1.1 for detailed study information and
results.
5-5
-------
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) Self-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) Parenchymal changes
(2) Pleural changes'3:
pleural plaques, DPT
(2) Pleura! changes: LPT
(any thickening, with or
without calcification,
excluding solitary
costophrenic angle
blunting); DPT (any
pleural thickening,
including costophrenic
angle blunting, with or
without calcification)
5-6
-------
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
serf-reported activities;
based on air dispersion
modeling based on stack
emissions and activity-
based sampling
Median: 2.42 fibers/cc-yr
(cases) and
0.59 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 pleura! 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.
5.2.1.2. Evaluation of Candidate Studies and Selection of Principal Study
The candidate studies were further evaluated in terms of quality attributes that would
support their use as a principal study in the derivation of an RfC. When selecting among
candidate principal studies, several factors, summarized in Table 5-2, are generally considered.
5-7
-------
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 environmental 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 overtime); 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 pleura! 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 et al.. 2008) and can appear long after the
initial exposure (LilisetaL 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 etal.. 2008; Jarvholm.
1992).
Stronger studies will have data on TSFE for the relevant exposures.
5-S
-------
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 parenchymal and pleura! 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.
Pleura! and parenchymal 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.
Two of the studies were conducted in occupationally exposed populations (Larson et al.,
2012a: Rohs et al., 2008), while the third was conducted in community residents without
occupational exposure (Alexander et al., 2012). Each of the studies provided estimates of
cumulative LAA exposure (in fibers/cc-yr). However, there were differences in exposure
sources and intensity. Of the two occupational studies, one (Larson et al., 2012a) occurred in a
setting where both occupational and nonoccupational exposures were relevant due to the close
proximity of the local vermiculite mining and milling operations to the Libby, MT community.
Nonoccupational exposures in the Libby, MT community were not quantified and thus were not
accounted for in the overall estimates of individual exposure. In the other study (Rohs et al.,
2008), exposures were generally lower and considered to be limited to the occupational setting
because most of the employees showered and changed into civilian clothes at the end of the work
shift. Therefore, nonoccupational exposure in the Marysville workers was assumed to be
minimal. However, in both cases, the exposure estimates for earlier years are subject to
uncertainty. For example, data on job and department were missing for the majority of the
workers in the Libby facility hired before 1960 (Larson et al., 2012a). In the Marysville facility,
no fiber measurements exist before 1972 (Rohs et al., 2008), although exposure estimates for this
period were constructed based on measurements taken in subsequent years (see Appendix F).
The third study, by Alexander et al. (2012), was conducted among Minneapolis community
residents (including a higher proportion of women compared to the other studies). The
researchers attempted to estimate individual community members' exposure based on facility
emissions and the individual's specific activities that were considered to be related to exposure
(e.g., installing or removing vermiculite insulation or playing in or around waste piles).
However, exposure estimates were constructed from modeled emissions based on very sparse
data from the facility's discharge stacks and activity-based exposure reconstruction, and as a
result are considered to have greater uncertainties.
5-9
-------
As noted in Table 5-2, relatively lower exposure levels are advantageous for developing
an RfC (given sufficient numbers of individuals with the health effect of interest) due to
uncertainties inherent in extrapolating from high-intensity (e.g., occupational) exposure levels to
low-intensity (e.g., environmental) exposure levels. A limitation of the studies conducted among
workers at the Libby facility is that the exposure levels experienced for some job codes are high
compared with those in the other two studies (see Table 5-1; e.g., based on the interquartile range
(IQR) of exposure from the 25th percentile value of 0.4 fiber/cc-yr to the 75th percentile value of
15.8 fibers/cc-yr, 25% of participants had cumulative exposures above 15.8 fibers/cc-yr).
Another limitation of these studies for conducting exposure-response analysis for LPT is that
many of the Libby workers were likely to have also been residents in Libby both before and
during their employment at the mining and related operations, so their actual TSFE to any LAA
exposure may be longer than their TSFE to occupational exposure to LAA. Therefore, data on
this important variable is uncertain. Thus, the Libby workers study (Larson et al., 2012a) is less
preferable for RfC derivation. The other two studies (Alexander et al., 2012; Rohs et al., 2008)
had generally lower exposure levels in comparison; however, greater uncertainty exists in the
exposure estimates for the Minneapolis cohort because few measurements of facility emissions
into the ambient air (Adgate et al., 2011). Indeed, the authors estimate that the numerical
uncertainty in exposure estimates is likely to be at least an order of magnitude, perhaps much
greater. Further, it is unclear whether TSFE is well characterized for the nonoccupational
exposures in the Minneapolis cohort. In contrast, the study of workers at the O.M. Scott plant in
Marysville, OH (Rohs et al., 2008) used exposure estimates based on extensive industrial
hygiene sampling data, individual worker histories, and employee focus interviews. Thus, Rohs
et al. (2008) is the preferred study for derivation of the RfC.
5.2.2. Methods of Analysis
5.2.2.1. Exposure Assessment
EPA collaborated with a research team at the University of Cincinnati to update the
exposure reconstruction for use in the job-exposure matrix (JEM) for all workers in the
Marysville, OH cohort, taking into account additional industrial hygiene data that were
unavailable for previous studies conducted in this cohort (Rohs et al., 2008; Lockey et al., 1984).
Exposure estimates for each worker in the O.M. Scott Marysville, OH plant were developed
based on the arithmetic mean of the available industrial hygiene data from the plant. The
exposure assessment procedure is described in Appendix F. In brief, occupational exposure was
estimated for each worker and adjusted to a cumulative human equivalent exposure for
continuous exposure, incorporating adjustments for different inhalation rates in working versus
nonworking time. These adjustments take into account the extensive seasonal changes in work
hours at the Marysville facility (see Appendix F).
5-10
-------
5.2.2.2. Data Sets for Modeling Analyses
Section 5.2.1.2 describes the selection of the Rohs et al. (2008) cohort as the principal
study. As explained below, EPA further considered the differential quality of exposure data for
different years within this data set and concluded that estimation of an RfC would be improved if
the primary data set for exposure-response modeling was restricted to the subset of workers hired
in 1972 and later, when higher quality exposure information was available.
As described in Section 5.2.1.2, the Marysville workers evaluated by Rohs et al. (2008)
formed the principal analytic group for derivation of the RfC, with exposure information updated
and augmented by the University of Cincinnati in collaboration with EPA. As noted in
Section 4.1.1.2.2 and Appendix F, the more reliable exposure estimates are considered to be
those from 1972 and later, as these data were based on analytical measurements. Therefore, the
primary modeling to derive a point of departure (POD) was conducted among the subgroup of
workers evaluated by Rohs et al. (2008) that began work in 1972 or later and had no previous
occupational exposure to asbestos (119 workers: 13 cases of localized pleural thickening and
106 unaffected individuals). However, information from workers who were hired before 1972,
as well as from workers who were evaluated only in the earlier study by Lockey et al. (1984),
were also considered in separate analyses (details and results of the analysis are in Appendix E).
In each case, to avoid any potential bias from previous unmeasured occupational exposure to
asbestos, only the data from those who did not report any previous occupational exposure to
asbestos were used.
Table 5-3 and Figure 5-2 present summary characteristics for the three analytic groups.
The first is the combined information for the 1980 (Lockev et al.. 1984) and 2002-2005 (Rohs et
al., 2008) evaluations, comprising all workers without previous exposure to asbestos; a detailed
description of how these data were combined is in Appendix E. The second group is all workers
evaluated in 2002-2005 without previous exposure to asbestos (as described by (Rohs et al.,
2008). The third group is a subset of the workers evaluated in 2002-2005, hired in 1972 or later,
without previous exposure to asbestos (primary analytic group). For the groups comprising only
individuals evaluated in 2002-2005, exposure estimates covered the period from start of work
through the date of job stop or at the time vermiculite ceased to be used in 2000, whichever
occurred earlier.
5-11
-------
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
n
252
%
100
Individuals evaluated in
2002-2005, hired in 1972 or
later
n
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 since first
exposure (yr)
Exposure duration
(yr) — duration of
exposed time (i.e.,
accounting for
gaps)
Body mass indexb
Cumulative
exposure
(fibers/cc-yr)
Mean exposure
(fibers/cc)
Residence
time-weighted
(RTW) exposure
(fibers/cc-yr2)c
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:
0.12-3,500.66
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)
5-12
-------
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 Lockev 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.
CRTW 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
Interstitial
changes
Individuals evaluated in 2002-2005, hired in 1972 or later
Total/?- 119
Localized / \ Diffuse
pleural / , pleural
Ihickeningl thickening
(n =12)
Figure 5-2. Radiographic outcomes among Marysville, OH workers.
Only the data from those who did not report any previous occupational exposure
to asbestos were used. 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).
5-13
-------
As stated previously, fiber measurements started in the Marysville plant in 1972, and
exposures before this time were estimated by University of Cincinnati scientists, based on focus
group interviews with 15 long-term former workers and the times when engineering changes
were made to control dust in the facility (see Appendix F). Exposure estimates for the period
before 1972 can be considered less certain compared with those estimates more directly based on
industrial hygiene data. The University of Cincinnati analysis assumed that early exposure levels
in the plant are twice those measured in 1972 (see Appendix F). The greater uncertainty of the
pre-1972 exposure estimates led to EPA's decision to focus the analysis on the group of workers
hired in 1972 or later. Although it is generally true that the use of more data is an advantage for
statistical analyses because it allows for the computation of more statistically precise effect
estimates, this increased precision can be offset by a negative impact on the accuracy of the
effect estimate if an increase in sample size is accompanied by greater exposure misclassification
or other biases.
In summary, the primary analytic group was the Marysville workers evaluated by Rohs et
al. (2008) who were hired in 1972 or later; however, additional information from workers hired
before that date was also used in modeling and sensitivity analyses.
5.2.2.3. Selection of Critical Effect
A critical effect is defined as "The first adverse effect, or its known precursor, that occurs
to the most sensitive species as the dose rate of an agent increases" (U.S. EPA, 2011). Three
endpoints are suitable for consideration as critical effects for the derivation of an RfC for LAA
where health effects data and exposure information are available in the principal study (Rohs et
al., 2008): (1) parenchymal changes viewed as small interstitial opacities in the lung,
(2) localized pleural thickening (LPT), or (3) diffuse pleural thickening (DPT) as defined in ILO
(2000). Each of these represents persistent changes to normal tissue structure.
Small interstitial opacities (asbestosis) are widely accepted as adverse; the American
Thoracic Society (ATS) states that "asbestosis is usually associated with dyspnea, bibasilar rales,
and changes in pulmonary function: a restrictive pattern, mixed restrictive-obstructive pattern,
and/or decreased diffusing capacity" (ATS, 2004). Similarly, DPT is also widely accepted as
adverse, with the ATS stating that "decrements associated with diffuse pleural thickening reflect
pulmonary restriction as a result of adhesions of the parietal with the visceral pleura. Restrictive
impairment is characteristic, with relative preservation of diffusing capacity (pattern of entrapped
lung)" (ATS. 2004).
Statements from the consensus groups vary as to whether pleural plaques impact lung
function. Regarding pleural plaques, the ATS notes that this endpoint is also associated with
decrements in lung function:
5-14
-------
Although pleural plaques have long been considered inconsequential markers of
asbestos exposure, studies of large cohorts have shown a significant reduction in
lung function attributable to the plaques, averaging about 5% of FVC, even when
interstitial fibrosis (asbestosis) is absent radiographically...The presence of
circumscribed plaques can be associated with restrictive impairment and
diminished diffusing capacity on pulmonary function testing, even in the absence
of radiograph!c evidence of interstitial fibrosis. (ATS, 2004).
However, the statement goes on to note that findings of significant pulmonary deficits are
not consistent, and that "most people with pleural plaques alone have well-preserved lung
function." In addition to the ATS document, the American College of Chest Physicians (ACCP)
(Banks et al., 2009) published a Delphi study conducted to gauge consensus among published
asbestos researchers, and found that these researchers statistically rejected the statement that
"Pleural plaques alter pulmonary function to a clinically significant degree" (although noting that
some researchers strongly agreed with the statement, and the response rate was relatively low at
<40%). Therefore, EPA undertook a systematic review to evaluate the magnitude and extent of
the pulmonary function deficits associated with pleural plaques and LPT, described in
Appendix I. The review demonstrates that these deficits can be considered adverse. LPT is a
pathological change associated with decreased pulmonary function, and thus is considered an
appropriate adverse effect for deriving the RfC (see Section 5.2.2.3 and Appendix I). Based on
the association of LPT with pulmonary function decrease, LPT is an appropriate health effect for
derivation of an RfC. Because interstitial opacities, DPT, and LPT are all appropriate candidate
endpoints, the critical effect was chosen as that which is the first to appear, or which occurs at
the lowest levels of exposure. A summary of the systematic review of the pleural plaque data is
discussed below.
Larson et al. (2012a) evaluated the timing of appearance and exposure levels at which
pleural and parenchymal abnormalities occur on chest radiographs of vermiculite workers at the
Libby facility relative to hire date (i.e., time since first occupational exposure). In this
retrospective analysis, the study authors reported that the health endpoint with the shortest
median time to appearance was circumscribed pleural plaques with a median latency of
8.6 years, compared to median latency times of 27.0 years for DPT and 18.9 years for
parenchymal changes (small interstitial opacity profusion of 1/0 or greater). Although all
workers experienced generally high exposure, cumulative fiber levels were lowest for those with
circumscribed pleural plaques (median of 44.1 fibers/cc-yr), compared to those with DPT
(median of 317.8 fibers/cc-yr), and highest for those with parenchymal changes (median of
235.7 fibers/cc-yr for those with major profusion category 1 abnormalities, 678.4 for
category >2/l, and 1,303.4 for category >3/2). Similarly, Rohs et al. (2008) found that for all
workers in that study, on average the cumulative fiber exposure for those workers with LPT only
(3.45 fibers/cc-yr) was lower compared to those with DPT only (8.99 fibers/cc-yr) or with any
5-15
-------
interstitial changes (alone or with either LPT or DPT; 11.86 fibers/cc-yr). These results indicate
that LPT may be the most sensitive of the effects examined, as the radiographic outcome most
likely to occur soonest after first occupational exposure, and the outcome most likely to appear at
relatively lower cumulative exposure levels. The clinical perspective suggests that LPT do not
clinically impair lung function for most people. This perspective was stated by the American
College of Chest Physicians (Banks et al., 2009) regarding the 2004 ATS statement: "Data were
cited showing that large studies of workers with pleural plaques had approximately a 5% mean
decline in FVC compared to asbestos workers without pleural plaques. In this report, the experts
concluded that the presence of pleural plaques did not decrease lung function to a significant
extent"—that is, they concluded that the observed decrements were not clinically significant to
an individual patient. EPA's systematic review of the literature and formal meta-analysis found
decrements in the same range—statistically significant decreases in mean FVC of 4.09% (4.08%
when studies without limitations are used) and in FEVi of 1.99% (3.87% when studies without
limitations are used). In addition, although few of the studies evaluated reported results for
diffusing capacity (as evaluated by DLco, the diffusing capacity of the lung for carbon
monoxide), these studies did observe statistically significant or nearly significant decreases
between those with no radiographic abnormalities and those with LPT or pleural plaques and
without any evidence of other radiographic abnormalities.
As stated by ATS (2004), the majority of individuals with pleural plaques may have well-
preserved lung function. However, for individuals who are already at the lower end of the
"normal" range of function, already have compromised function, or have increased vulnerability
or susceptibility due to other factors (such as chronic disease, other environmental exposures,
smoking, etc.), even a relatively small decrease in lung function can be important, but once
averaged into the whole study population (i.e., looking at only average changes in the whole
group) the sensitive individuals' contribution to the population-wide change in mean pulmonary
function measures is muted.
Accordingly, there is a difference in considering what may be significant from a clinical
perspective compared to an epidemiological perspective. The clinician's focus is the individual
patient, and decisions made in that context (i.e., benefits/risks of medical treatments or tests). In
contrast, the population-level (risk assessment) perspective considers any changes in the
population distribution of pulmonary function and the potentially increased risks of adversity to
subpopulations of the general population. When considering an entire population with a
distribution of lung function parameters, even small changes in the mean of that distribution
means that a much larger proportion of the exposed population is shifted down into the lower
"tail" of the lung function distribution. This line of thinking is well understood in the recent
examples of lead and IQ (U.S. EPA, 2013 a) and respiratory function and ozone (U.S. EPA,
2013b). Early childhood exposure to lead can lead to decrements in intelligence as measured by
IQ. Depending on the exposure level to lead, a mean deficit of 2 IQ points would not be
5-16
-------
measurable nor lead to a clinical finding of harm in individuals, but from a epidemiologic or
population-level perspective, a downward shift in a portion of the entire IQ distribution by 2 IQ
points would be expected to push many individuals already in deficit further into a more clearly
"adverse" state. Similarly, even small decrements in lung function on the population level could
push "borderline" individuals into a state of clinically significant decreased lung function.
In addition, ATS (2004) stated "The presence of plaques is associated with a greater risk
of mesothelioma and of lung cancer compared with subjects with comparable histories of
asbestos exposure who do not have plaques." While references provided in the ATS (2004)
statement (Hillerdal and Henderson, 1997; Hillerdal, 1994) do not directly support the ATS
(2004) statement, a recent large (5,287 retired workers, 17 mesothelioma cases) study (Pairon et
al., 2013) found a statistically elevated risk of mesothelioma in a group with plaques (parietal
and diaphragm) compared to a no-plaques group, using computer tomography (CT). The study
authors found that, after adjusting for cumulative exposure index and TSFE, the risk of
mesothelioma in the plaques (parietal or diaphragm) group was statistically elevated (hazard rate
(HR) = 6.8, 95% CI: 2.2-21.4) compared to the risk of mesothelioma in the exposed workers
without pleural plaques.
In the Marysville workers evaluated in 2002-2005, differences in exposure patterns are
also apparent among outcome groups (see Table 5-4). Exposure to LAA was lower among those
with no radiographic abnormalities compared to those with LPT and those with DPT and/or
interstitial opacities. Of the candidate critical effects, LPT has the shortest TSFE, and is more
likely to appear at lower levels of LAA exposure. LPT is associated with adverse decrements on
pulmonary function. Thus, LPT is selected as the critical effect from among the noncancer
radiographic endpoints evaluated in the principal study for RfC derivation.
5-17
-------
Table 5-4. Characteristics of workers at the O.M. Scott plant in Marysville,
OH, with health evaluations in 2002-2005 who did not report any previous
occupational exposure to asbestos
N
Time (years)
since first
exposure
(TSFE), range
TSFE, median
(interquartile
range, IQR)
Mean exposure
(fibers/cc), range
Median (IQR)
Cumulative
exposure
(fibers/cc-yr),
range
Median (IQR)
Residence
time-weighted
(RTW) exposure
(fibers/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)
0.5313-3,500.66
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
1,693.86
(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)
62.7209-3,011.68
120.5295
(74.3389-944.9676)
Table 5-4 and Figure 5-2 both highlight a complexity in that these radiographic changes
are not mutually exclusive—individuals may have one or more changes simultaneously, in any
combination. Among the 66 individuals with LPT, 10 also had DPT or interstitial opacities, and
these 10 individuals are noticeably different with regards to TSFE and exposure compared to
LPT cases without other radiographic changes, consistent with the results of Larson et al.
(2012a). When restricting to the subgroup of individuals hired in 1972 or later, there are
106 individuals with no radiographic abnormalities, 12 individuals with LPT only, and one
individual with both LPT and DPT. The individual with both LPT and DPT had a TSFE of
31.52years, similar to the median TSFE of 29.71 years among the 12 individuals with LPT only.
However, this individual had higher estimated mean exposure (0.46 fiber/cc, compared to a
median of 0.08 fiber/cc) and cumulative exposure (9.13 fibers/cc-yr, compared to a median of
1.82 fibers/cc-yr), compared to the other 12 LPT cases. The primary analysis considers as the
critical effect all LPT cases together, contrasted to those without radiographic abnormalities, but
5-18
-------
the effect of separating out those with multiple radiographic outcomes is examined in the
sensitivity analyses (see Section 5.3.5).
In addition, an alternative critical effect of "any pleural thickening" (APT) is considered.
Note that in the case of the subcohort of workers evaluated in 2002-2005 and hired in 1972 or
later, this definition is equivalent to a critical effect of LPT because no individuals had DPT
alone.
5.2.2.4. Selection of Explanatory Variables to Include in the Modeling
As with the mode of action (MOA) for carcinogen!city, the MOA for LPT and the results
of other asbestos epidemiology studies could potentially inform noncancer modeling decisions
and suggest exposure metrics to use in modeling. The following text discusses the plausibility of
exposure metrics proposed in the MO A/epidemiology literature.
As noted in Section 4.4, important considerations in evaluating the available mechanism
and MOA data are fiber characteristics, route of exposure, dose metric, as well as study design
and interpretation. Specific fiber characteristics impact the fiber toxicokinetics (reviewed in
Section 3), and in turn, the biologic response to fibers. Fiber dimensions play a role in
translocation, a clearance mechanism that may lead to inhaled fibers moving from the lung to the
pleura. Data gaps still remain to determine specific mechanisms involved in LAA-induced
pleural disease. The review of studies in Section 4.4 clearly highlights the need for more
controlled studies examining LAA in comparison with other forms of asbestos and for examining
multiple endpoints—including reactive oxygen species (ROS) production and proinflammatory
gene expression alterations—to improve understanding of mechanisms involved in noncancer
health effects. Although research demonstrates that the LAA has biologic activity consistent
with the inflammatory action and cytotoxic effects seen with other forms of asbestos, the
conclusion of Section 3 of this assessment is that the data are not sufficient to establish an MOA
for the pleural and/or pulmonary effects of exposure to LAA.
A general understanding of the biology and the epidemiology of LPT can still inform the
modeling as to which explanatory variables should be considered in the models, how the
variables should be considered or statistically parameterized, and whether they should be
retained in the model. From a general understanding of the respiratory effects of asbestos, the
intensity of exposure (i.e., concentration), the duration of exposure, and the timing of exposure in
relation to subsequent diagnosis of LPT (i.e., TSFE) have been shown to be univariate predictors
of pleural plaques in multiple epidemiologic studies as discussed below and, therefore, merit
specific consideration in this modeling effort.
Timing of exposure
Fibrosis in the pleural tissues needs time to develop and become visible by x-ray (Larson
et al., 2010a). It has been shown that the prevalence of fibrotic lesions progresses as a function
5-19
-------
of time (Rohs et al., 2008) and can appear long after the initial exposure (Lilis et al., 1991).
Many investigations of the exposure-response relationship for pleural plaques has found that
TSFE is a significant explanatory variable (Paris et al., 2009; Paris et al., 2008; Jarvholm, 1992).
This suggest that TSFE should be considered as a potential explanatory variable in the modeling.
It is important to understand that even when an individual explanatory variable may be an
important univariate predictor of the risk of LPT, in more complex modeling with two or more
explanatory variables, the relationship observed in univariate modeling may no longer hold. One
reason for this is that two variables can be highly correlated in many occupational cohorts, thus,
the regression modeling may indicate that, for the data at hand, there is more unique information
to explain the risk of LPT in one variable than in the other.
Intensity of exposure
A general understanding of toxicology suggests that, for a given duration of exposure,
exposure at higher intensities (concentrations) will likely result in higher burdens of fibers in the
alveolar region of the lung, and potentially in the pleural tissue as well. Therefore, for a given
duration of exposure, there is a reasonable expectation that people exposed at higher intensities
of LAA would experience greater risk of being diagnosed with LPT than people exposed at
lower exposure intensities. Similarly, from general principles, at a given intensity of exposure,
greater duration of exposure results in higher tissue concentrations of fibers. Epidemiologic
evidence from a large cohort of asbestos-exposed workers has reported that exposure intensity
(concentration) can be an important predictor of being diagnosed with pleural plaques — even in
a multivariate model along with TSFE (Paris etal., 2009; Paris et al., 2008). As noted in Section
4.1.2.1.1.1, LPT was introduced as a term in the 2000 ILO guidance. LPT includes plaques on
the chest wall and at other sites (e.g., diaphragm). Plaques on the chest wall can be viewed either
face-on or in profile. A minimum width of about 3 mm is required for an in-profile plaque to be
recorded as present according to the 2000 ILO guidance. Jarvholm (1992) fit a mathematical
model for the incidence of pleural plaques based on concentration and TSFE, which the author
considered to have a biological interpretation. This suggests that exposure intensity should be
considered as a potential explanatory variable in the modeling.
Duration of exposure
Important characteristics of amphibole fibers are their biodurability and biopersistence.
Due to the slow clearance of amphibole fibers from the lung, the fiber burden in the alveolar
region of the lung is expected to increase for a given exposure intensity as the duration of
exposure increases. This may be true of the pleural tissues as well—but little scientific
information is available on the time course of potential fiber accumulation in pleura. Amphibole
fibers may remain biologically active for many years while the fibers are in residence in the
tissues, although this biological activity may vary with time. For example, depending on the
5-20
-------
composition and structure of the fiber, certain fibers may cease to have surface activity in
biological media, or may have different biological activity in this media (Pezerat 2009).
Further, some asbestos fibers can become covered with an iron-protein coat in the lungs of
exposed individuals [i.e., forming ferruginous bodies; (Dodson et al., 1993; Churg and Warnock,
1981)1. The biological effect of this coating is unclear, but may alter the activity of the fibers.
Epidemiologic evidence from studies of asbestos-exposed workers [e.g., Clinet al. (2011)]
indicates that cumulative exposure, a metric of exposure that encompasses both exposure
intensity as well as duration, can be an important predictor of the probability of being diagnosed
with pleural plaques. This suggests that duration of exposure should be considered in modeling.
Cumulative exposure (as an expression of concentration and duration) should be considered as a
potential explanatory variable in the modeling. Therefore, modeling results using both exposure
intensity, C, and cumulative exposure, CE, as the exposure metric are considered. Another
exposure metric related to both exposure intensity and duration is called residence time-weighted
exposure, a metric of exposure that can be used to more heavily weigh earlier exposures.
Residence time-weighted exposure is also considered for modeling.
Other explanatory variables
Other explanatory variables of interest include those that may be confounders of the
explanatory variables' statistical relationships with the risk of LPT. These include body mass
index (BMI), age, and smoking (complete list from the table of potential confounders). Each of
these was assessed as a potential confounder prior to modeling the main explanatory variables of
interest.
5.2.2.5. Selection of the Benchmark Response
Selecting a benchmark response (BMR) involves making judgments about the statistical
and biological characteristics of the data set and about the applications for which the resulting
benchmark concentration (BMCs)/lower limit of the BMC (BMCLs) will be used. An extra risk
of 10% is recommended as a standard reporting level for quantal data. Biological considerations
may warrant the use of a BMR of 5% or lower for some types of effects (e.g., frank effects), or a
BMR greater than 10% (e.g., for early precursor effects) as the basis of the POD for a reference
value (U.S. EPA. 2012).
LPT is a persistent change to normal tissue structure and is associated with a decrement
in lung function on a population level (~5 and -2.5% decrements in percentage predicted FVC
and FEVi, respectively). Larson et al. (2012a) showed a statistically significant increased risk of
people with LPT having "restrictive spirometry" and concluded that this abnormality may result
in lung function impairment. However, the available data do not lead EPA to conclude LPT
should be considered a frank effect and thus EPA selects a BMR of 10% extra risk for this
endpoint.
5-21
-------
As noted in Section 5.2.3.1, an alternative critical effect of APT was also considered as
an alternative analysis. For this outcome, a BMR of 10% was also used, given that (as shown in
Figure 5-2) a significant majority of cases were LPT.
5.2.2.6. Exposure-Response Modeling
LPT was selected as the critical effect based on the adverse health effects associated with
pleural thickening specific to this diagnosis (TLO, 2002). Note that for the primary analytic data
set (workers evaluated in 2002-2005 and hired in 1972 or later), the number of individuals with
LPT is the same as the number with either "any pleural thickening" or "any radiographic
change" (i.e., there is no difference in the number of cases or the estimates of risk) because the
single case with DPT also had LPT (and thus is included as an LPT case) and no cases of
interstitial abnormalities occurred. However, in the larger cohort of workers (all workers, and
those evaluated in 2002-2005 regardless of hire date), there were individuals with these more
severe outcomes as well as LPT (see Figure 5-2).
The exposure-response relationship was modeled as described below, and PODs were
estimated using BMC methodology. For inhalation data, the BMC is defined as the exposure
level that results in a specified BMR. The RfC is derived from the lower 95% confidence limit
of the BMC, referred to as the BMCL, which accounts for statistical uncertainty in the model fit
to the data. All analyses were performed using SAS® statistical software v. 9.3. BMCLs were
obtained by the profile likelihood method as recommended by Crump and Howe (1985) using
the nonlinear mixed modeling procedure (PROC NLMIXED) in SAS (Wheeler. 2005).
5.2.2.6.1. Considerations of appropriate model forms and explanatory variables. The process
and considerations for exposure-response modeling of the Marysville data were guided by EPA's
2012 Benchmark Dose Technical Guidance (U.S. EPA, 2012). As outlined in that document,
there are several stages of exposure-response modeling. Once the appropriate data set(s),
endpoint(s), explanatory variables(s), and BMR are determined, the next step is to choose an
appropriate statistical model form or set of model forms to evaluate (e.g., logistic, probit,
Dichotomous Hill, etc.). Among this set of models, the overall model fit and the fit in the region
of the BMR are evaluated to determine which models adequately represent the data. Finally, one
or more models are selected from the group of adequately fitting models to derive a POD for the
reference value. Regarding the selection of models to evaluate, the Benchmark Dose Technical
Guidance (see p. 26) states: "The initial selection of a group of models to fit to the data is
governed by the nature of the measurement that represents the endpoint of interest and the
experimental design used to generate the data. In addition, certain constraints on the models or
their parameter values sometimes need to be observed and may influence model selection." In
the Marysville data, a number of factors must be considered to determine an appropriate
modeling strategy: the nature of the data set, ability to estimate the effects of exposure and of
5-22
-------
important covariate(s), the existence of a plateau or theoretical maximum response rate in a
population, and the ability to estimate a background rate of the outcome in a population. Each
factor is described below, and consideration of these factors in total resulted in a preference for
the Dichotomous Hill model, with a set of additional model forms suitable for evaluation of
sensitivity to model selection.
• Nature of the data set: For the Marysville workers data set, the outcome data are
dichotomous (presence or absence of an effect), and thus, appropriate models are
those suitable for dichotomous endpoints. The Marysville workers underwent
radiographic evaluation in 2002-2005 to ascertain the presence or absence of
radiographic abnormalities (i.e., prevalence data). Radiographic outcomes are coded
as present or absent, leading to a dichotomous response structure. Appropriate
models for this type of data include models such as logistic, probit, log-logistic,
log-probit, Dichotomous Hill, and Michaelis-Menten (see Table 5-5). Goodness of
model fit for these models may be evaluated using the Hosmer-Lemeshow goodness-
of-fit statistic (Hosmer and Lemeshow, 2000): a low/>-value (<0.05) indicates poor
fit, while a higher/>-value indicates adequate fit. Note that the computation of this
statistic involves dividing the data set into bins, based on the predicted probability of
the (dichotomous) outcome. The standard procedure is to use 10 bins (i.e., deciles),
and this approach was used for all analyses shown here.
• Effect of exposure: Because the data set include estimates of individual exposure and
the goal is to derive an RfC, appropriate models need to include an independent
exposure variable. All the models listed above can estimate the effect of changes in
exposure on risk of the outcome, although the parameter that reflects the magnitude
of that effect changes across models. In models where exposure is included without
logarithmic transformation, the b parameter corresponding to exposure is interpreted
as a "slope" and represents the change in outcome per unit change in exposure. In
models where exposure is natural log transformed, the interpretation is somewhat
different; both the a and b parameters determine the shape of the exposure-response
relationship (e.g., a + b x ln(x) = ln(exp(a) x xb). Thus, b is a power parameter and
behaves more like a shape parameter in this context, while exp(a) behaves like a
traditional slope. This is an important distinction in interpreting models including
natural log transformation of the exposure (i.e., log-logistic, log-probit,
Michaelis-Menten, Dichotomous Hill).
5-23
-------
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(V) -
l'^' 1 + exp[-a - b x x]
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
p(x) 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^ hJ-n 1
pCx.r; M5 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 = [fr1 (BMR x (l - (a)) + («)) - a]/b
Log-logistic: BMC = exp[((-ln ((1/BMR) - 1)) - a)lb]
Log-probit: BMC = exp[(<&-1 (BMR) 0- a)/b]
Michaelis-Menten: BMC = exp[(-ln((Plateau—bkg)/((l—bkg) x BMR) - l)-a]
Dichotomous Hill: BMC = exp[(-ln ((Plateau—bkg)l((\— bkg) x BMR) - 1) - a)lb]
Dichotomous Hill withtime(T) covariate: BMC = exp [(-In ((Plateau—bkg)l((\ - bkg) x BMR) - 1) -
a - c x T)lb]
Plateau in models: Some model forms have an explicit parameter representing a
plateau (e.g., Dichotomous Hill and Michaelis-Menten), which is an asymptotic
quantity interpretable as the maximum prevalence of the outcome that would ever be
observed at very high levels of the predictor variables in the model (e.g., high levels
of LAA exposure). In contrast, certain model forms (e.g., log-logistic and log-probit)
do not estimate a separate plateau parameter, but instead have maximum asymptotic
values of 100%. EPA wanted to understand whether the considered models implied
an asymptotic maximum incidence for the endpoint (i.e., a "plateau") and to evaluate
the sensitivity to alternative specified values for that plateau. Thus, EPA selected a
model with an explicit plateau term for the option of fitting that plateau term to the
5-24
-------
data and for the ability to do sensitivity analysis with alternative fixed plateau values
to evaluate the sensitivity of results.
Plateau—further considerations: For models with an explicit plateau parameter, EPA
considered whether to let the plateau term be fit to the data or to select a fixed value
prior to fitting the model. As described below, EPA chose to fix the value of the
plateau prior to fitting these models to better consider a broader set of data on pleural
thickening. However, EPA also conducted sensitivity analysis on the impact of this
assumption for model results. Importantly, this plateau parameter of an asymptotic
maximum prevalence cannot be directly observed in, and is not well estimated from
the Marysville data because none of the workers experienced high enough exposure
and follow-up. In the group of workers defined for primary exposure-response
analysis (i.e., those hired in or after 1972 with radiographs performed in 2002-2005),
the TSFE averaged 28.4 years and ranged from 23.14 to 32.63 years.
Exposure-response models that include TSFE or otherwise incorporate the timing
between exposure periods and observation, such as models using the residence
time-weighted (RTW) exposure metric, could allow for the estimation of a plateau,
but the limited data on the effect of elapsed time in the workers hired in or after 1972
does not support a reliable estimate of the asymptotic maximum prevalence. In
addition, standard radiographs may not have perfect sensitivity or specificity to
identify the outcomes of interest (thus "observed" prevalence may differ from
"actual" prevalence). Models that do not include time from exposure to the x-ray
observation would be estimating a plateau that might similarly be extrapolating on
dose and might not appropriately estimate the impact of a longer follow-up period.
For the RfC, the question is what happens when individuals are exposed over a
lifetime (assumed to be 70 years). This may be difficult to answer if a given model
results in a plateau significantly lower than what might result from sufficient duration
or follow-up time. One option is to fix the plateau parameter at a value informed by
the existing literature on observed prevalence in populations that had higher
exposures and longer TSFE values. In a cross-sectional study of Libby workers and
residents seen at a clinic in Libby, MT, Winters et al. (2012) observed a prevalence of
76% for pleural thickening, although the maximum TSFE was not known. Previous
studies in populations exposed to asbestos (potentially amphibole and/or
nonamphibole) have reported prevalences of pleural thickening of 82.4% among U.S.
insulators with >40 years since first exposure (Lilis et al., 1991) and prevalence of
pleural plaques of 85.7% among Swedish shipyard workers with 50-54 years since
first exposure (Jarvholm, 1992). Thus, a reasonable option would be to use the
Dichotomous Hill or Michaelis-Menten model and fix the plateau term at a value (i.e.,
85%) consistent with maximum observed prevalence rates in the asbestos literature.
EPA performed a sensitivity analysis (see Section 5.3.4) to evaluate the effect of
assumptions regarding the plateau on the POD.
Effect of covariates: As with the discussion for effect of exposure, a desirable model
attribute is the ability to estimate the effect of additional covariates. EPA evaluated a
variety of possible covariates and determined that in the primary analytic data set of
individuals evaluated in 2002-2005 and hired in 1972 or later, none showed evidence
of potential confounding of the LAA exposure-LPT relationship. Each of the models
listed above (logistic, probit, log-logistic, log-probit, Dichotomous Hill, and
Michaelis-Menten) allow for the inclusion of covariates. Specifically for modeling
5-25
-------
LAA exposure and risk of LPT, one of the most important covariates to consider is
TSFE. As described above, the prevalence of pleural plaques (LPT was introduced as
a diagnostic term in the ILO guidance in 2000) has been shown to increase as TSFE
increases, even in the absence of continued asbestos exposure. Although the
literature indicates TSFE is the most important time-related factor, other factors may
be important to consider, including age at examination, hire year, job tenure (time
elapsed from job start to job stop), and exposure duration (taking into account gaps in
exposure). There are also nontime-related factors which may influence the
association between LAA exposure and risk of LPT. These include gender, smoking
status, and BMI. Smoking is a particularly important variable to consider when
evaluating respiratory health outcomes. Each of these factors was investigated in the
primary data set. To be a potential confounder, the factor must be associated with
both LAA exposure and LPT, and must not be an intermediate in the causal pathway
between exposure and outcome. The association with natural log-transformed LAA
exposure in the subcohort was assessed using a linear regression model, and the
association with LPT was assessed using a logistic regression model (see Table 5-6).
While many of the time-related factors (with the exception of age at x-ray
examination) as well as male gender and former smoker status were associated with
each of the three exposure metrics, none were associated with risk of LPT. Thus,
none of the factors met the criteria of being associated with both LAA exposure and
LPT, and none were considered as potential confounders. Further consideration of
potential confounding and effect modification is addressed in the uncertainty analyses
described in Section 5.3.3.
• Background rate: There may be a nonzero background rate of LPT in the population,
and it may be desirable to estimate this rate explicitly rather than using a model that
implicitly assumes a background rate of zero. Certain model forms (e.g., log-logistic,
log-probit, Dichotomous Hill, Michaelis-Menten) include an explicit parameter
representing the background rate of the response while others (e.g., logistic, probit)
do not include this parameter. Establishing a background rate for LPT prevalence in
the population is challenging, as estimates from previous studies in a variety of
populations vary widely (Weill etal., 2011; Rogan et al., 2000; Zitting, 1995; Cordier
etal.. 1987; Rogan etal.. 1987; Castellan et al.. 1985; Anderson et al.. 1979);
however, these previous studies do indicate that the background rate is unlikely to be
zero. Because there is not a clear indication of what the background rate is in an
unexposed population, models that allow estimation of a background rate rather than
assuming it to be zero, were considered to have greater weight.
5-26
-------
Table 5-6. Evaluation of association between covariates and exposure, and
between covariates and LPT.a Cells display beta coefficient (standard
error),/j-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.
Based on the considerations outlined above, EPA developed the following list of desirable model
features (see Table 5-7):
a) Models suitable for a dichotomous outcome
b) Ability to estimate the effect of exposure via inclusion of slope (models using
untransformed exposure), or slope and shape parameters (models using natural log-
transformed exposure)
c) Ability to estimate effect of covariates
5-27
-------
d) Ability to estimate or specify the plateau
e) Ability to estimate the background rate of LPT in the study population
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
The epidemiological models described above (logistic, probit, log-logistic, log-probit,
Dichotomous Hill, and Michaelis-Menten) are suited for dichotomous outcomes, and all allow
for the inclusion of covariates. However, the Michaelis-Menten model does not allow for
estimation of a separate shape parameter (b parameter is implicitly fixed at 1). The ability to
estimate both a and b (rather than imposing a presumed shape) provides greater flexibility in
exposure-response modeling, and thus the Dichotomous Hill model is preferred over the
Michaelis-Menten model. The logistic and probit models do not include separate parameters for
either the background rate or a plateau. The three remaining models—log-logistic, log-probit,
Michaelis-Menten, and Dichotomous Hill—do allow for estimation of the effect of exposure and
the background rate, but only the Dichotomous Hill model includes a separate plateau parameter.
Therefore, the Dichotomous Hill model is considered to be the most flexible and potentially the
most suitable based on biological and epidemiologic properties in the absence of information on
actual model fit, which is also an important consideration in model selection. The
Michaelis-Menten, log-probit, and log-logistic models are also reasonable alternatives (note that
latter two models implicitly fix the plateau at 100%, above the maximum observed prevalence in
reported studies). These models are also evaluated for sensitivity to modeling properties and
assumptions. As described above, the plateau is an asymptotic parameter, and it may not be
possible to reliably estimate given the limitations of the data. The Marysville workers who
underwent health evaluations in 2002-2005 and whose job start date was on or after 1/1/1972
5-28
-------
had relatively low levels of exposure and a narrow range of TSFE, and it is likely that estimation
of the maximum prevalence of LPT by radiograph is not well supported in these data. One
option to address this difficulty is to fix the plateau parameter at a value consistent with the
asbestos literature reviewed above. This assessment will use a plateau value of 85%, based on
the literature regarding maximum observed prevalence in populations reported to have had long
follow-up periods and significant exposure to asbestos, although the sensitivity of the model to
this assumption is examined in Section 5.3.4.
5.2.2.6.2. Considerations of exposure metric, statistical model fit and selection of
exposure-response model. While the above description in Section 5.2.2.6.1 explains EPA's
preference to use the Dichotomous Hill model for exposure-response modeling in this data set
(along with a reasonable set of models to evaluate sensitivity to model form), the text below
explains EPA's evaluation of options for the specific exposure metrics to use, and the results of
the analysis of statistical fit among the considered models. As noted in Section 5.2.2.4, in the
absence of an MOA for pleural health effects, an understanding of the general biology and the
epidemiologic literature may help to inform the consideration of exposure metrics for
exposure-response modeling. For pleural effects, the timing of exposure, the intensity of
exposure, and the duration of exposure may all be important variables to predict the risk of LPT
and were considered in the regression modeling.
In the Marysville data, exposure information was collected for each individual based on
specific work location, season, and year. There are several ways in which these estimates of
exposure by person by year can be aggregated into a single measure. Exposure estimates can be
summed over each individual's work history to yield a cumulative exposure estimate for each
individual. The cumulative exposure may also be divided by duration of (occupational) exposure
to yield an average intensity of exposure (mean exposure). A third option for an exposure metric
is to weigh more heavily exposures occurring in the more distant past, using a RTW21 exposure
metric for which each year is weighted by the number of years it occurs prior to the year in
which prevalence is evaluated. These three expressions of exposure can be used to derive a POD
with appropriate adjustment of the units to arrive at the RfC.
Table 5-8 shows the univariate model results for each of the model forms evaluated,
using each of the three different parameterizations of exposure. All models had adequate
goodness of fit (GOF), as indicated by the Hosmer-Lemeshow ^-values (all had/?-values >0.7,
substantially higher than the standard cutoff value of 0.10, below which a model is considered to
fit the data poorly), and were carried forward for further consideration. Because each of the
models shown in Table 5-8 was evaluated on the same data set (same number of observations,
21The 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.
5-29
-------
and same response variable), it is appropriate to compare relative fit among them using the
Akaike Information Criterion (AIC). The AIC for the models ranged from a low value of 73.8
(probit and Michaelis-Menten models using mean exposure) to a high of 79.3 (logistic model
using RTW exposure). Note that the Dichotomous Hill model with estimated plateau yielded
unrealistic parameter estimates when using either cumulative or RTW exposure (e.g., slope
estimates >100) and were considered less reliable. Within each model form, mean exposure
consistently provided a superior fit (as evidenced by lower AIC) compared to either cumulative
or RTW exposure.
Each of the different candidate models shown in Table 5-8 is similar in general form, and
comparison of model fit informed by the biological and epidemiologic features of the models
does not strongly imply a preference for one model form over the others. For the mean exposure
model, the AIC values for the logistic, probit, log-logistic, log-probit, Dichotomous Hill model
with plateau fixed at 85%, and Michaelis-Menten model with plateau fixed at 85% ranged from
73.8 to 75.4, a difference of only 1.6 AIC units for the mean exposure model which indicates
essentially equivalent fits. Figure 5-3 shows a plot of the fit for the Dichotomous Hill model
with plateau fixed at 85% and the Michaelis-Menten model with plateau fixed at 85%.
5-30
-------
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
5-31
-------
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),^-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 (0.7 120)
Plateau = 0.7418 (0.5143)
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
5-32
-------
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 (fibers/cc-yr), Mean indicates mean exposure (fibers/cc), RTW indicates residence time weighted exposure (fibers/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.
5-33
-------
CL
_l
M—
O
ro
.a
2
CL
cq
o _
CN
O
p
O
3MCL
0.00
\
0.05
BMC10
\
0.10
DH
DH85
MM
MM85
A 4 bins
° 5 bins
0.15 0.20 0.25
Mean Exposure (f/cc)
0.30
Figure 5-3. Plot of exposure-response models for probability of localized
pleural thickening (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.
5-34
-------
As described above, the Dichotomous Hill model with plateau fixed at 85% possesses
desirable biological/epidemiological properties and is the most flexible of the evaluated models.
This model had an AIC at the lower range of the models evaluated (75.4 when using mean
exposure or 76.8 when using cumulative or RTW exposure); the AIC for the model using mean
exposure was within two AIC units of the best-fitting models, a difference that is generally
considered indistinguishable with respect to relative model fit (Burnham and Anderson, 2002).
Because it was considered the most flexible model, the Dichotomous Hill model with plateau
fixed at 85% was selected as the primary model for RfC derivation. This model was carried
forward through the extensive sensitivity analyses (see Section 5.3) because it was considered to
be the model most likely to be able to detect sensitivity to covariates and alternative model
parameterization.
The RfC is "an estimate (with uncertainty spanning perhaps an order of magnitude) of an
exposure (including sensitive subgroups) that is likely to be without an appreciable risk of
adverse health effects over a lifetime." (U.S. EPA, 1994b), where a lifetime is commonly
assumed to be 70 years. Thus, consideration of the effect of time (specifically time elapsed from
exposure to outcome evaluation) is an important aspect of deriving an RfC. The literature on
pleural abnormalities and asbestos generally supports a conclusion that the amount of time
elapsed between exposure and evaluation has a major impact on observed response. The model
form and/or the selection of an exposure metric should incorporate considerations of time
factors. As described above, in the primary data set (which had a very limited range of TSFE
values) neither TSFE nor any of the other covariates evaluated were significantly associated with
LPT. However, the epidemiologic literature is clear that the timing of exposure is an important
factor in evaluating risk over a lifetime, and it was considered critical to address the time-course
of exposure and LPT in deriving the RfC. Thus, one option would be to incorporate TSFE
through the choice of exposure metric. Neither mean nor cumulative exposure takes into account
the TSFE or the timing of subsequent exposures. The RTW exposure metric does incorporate
information for each year's exposure of the time between that exposure and evaluation of the
prevalence—exposure in a given time interval is weighted according to time elapsed, with
exposure occurring earlier given greater weight. This approach might be preferable to a model
including TSFE as a covariate along with mean or cumulative exposure because the RTW metric
weights each year of exposure according to the time elapsed until health evaluation. However,
while time prior to evaluation is "considered" when using the RTW exposure metric, the
subcohort with better exposure data is still limited in that the range of TSFE in this group of
workers is relatively narrow (from 23.14 to 32.63 years), which may explain the lack of
predictive value of TSFE in this group. Thus, utilizing this approach to estimate a concentration
yielding the same risk for a 70-year exposure is informed by a relatively small range of TSFE
values.
5-35
-------
Therefore, EPA considered explicitly including TSFE as a covariate in exposure-response
modeling, along with either mean or cumulative exposure. As noted above, less variability is
present in TSFE (and other time-related factors) for the subgroup of workers hired in 1972 or
later, compared to the larger study population of Marysville workers who underwent health
evaluations in 2002-2005 (regardless of hire date, n = 252). In this group, the median TSFE was
33.5 years (SD = 7.12 years) and the range was 23.1 to 47.3 years, considerably longer than the
subset of these workers whose job start date was on or after 1/1/1972. As noted in
Section 5.2.2.2, EPA selected the smaller subgroup for primary exposure analysis because the
exposure data after 1972 is of higher quality, even though the range in TSFE is more limited.
However, it is possible to use a larger subset of the Marysville workers with a wider range of
time-related factors (but more uncertain exposure information) to model the effect of TSFE, then
include this effect as a fixed parameter when modeling the exposure-response relationship
among the primary analytic group of workers who underwent health evaluations in 2002-2005
and were hired >1972. This hybrid model may be used to calculate a BMCL for a scenario of
lifetime exposure (70-years duration and 70-years TSFE). This procedure could use either mean
exposure or cumulative exposure. The use of RTW may not be appropriate because RTW
includes consideration of the timing of exposures and thus a model using RTW in the exposure
metric and TSFE as a covariate might have two variables both reflecting the timing of exposure.
Modeling procedure
1) Fit the model (Dichotomous Hill with plateau fixed at 85%) in all the workers
evaluated in 2002-2005 (regardless of hire date) with TSFE and LAA exposure
(either represented as cumulative exposure [CE] or as mean exposure [C]) as
predictor variables.
, rr,^ , , Plateau- bka
p(x, T) = bka -\ -- - - (5-1)
r^ > •> a i+exp[-a-fcxln(X)-cxr] v '
2) Use the regression coefficient for TSFE calculated in (1), represented by "c", as a
fixed parameter in the model for workers who underwent health evaluations in
2002-2005, using data only on those hired in 1972 or later; fit the model to the data
on this subcohort using the individual data on both TSFE and LAA exposure as
independent variables — note that as in the larger cohort of all workers evaluated in
2002-2005, LAA exposure may be modeled as either mean exposure or cumulative
exposure.
3) Use some fixed value of TSFE to estimate the benchmark value and the lower limit
conditional upon that TSFE.
Benchmark value = exp( — -* - ) (5.2)
5-36
-------
The bivariate modeling results for this hybrid model option are shown in Table 5-9. Both
mean and cumulative exposure metrics were evaluated on the basis of two measures of model fit.
The AIC allows the comparison of the fit among two or more models of the same dependent
variable, and AIC results within approximately two units can be considered to be of equivalent
fit. However, two studies in the epidemiologic literature also compared mean concentration and
cumulative exposure in relation to the risk of pleural plaques and found that when also including
TSFE as an explanatory variable (as in the results shown in Table 5-9), mean exposure provided
a significantly better model fit (Paris et al., 2008). In addition, another study (Jarvholm, 1992)
proposed model including TSFE and intensity of exposure and stated that this model is
biologically interpretable; statistical modeling of pleural thickening (18% diffuse) showed that
duration of exposure does not matter when TSFE is included in the model (Lilis etal., 1991).
5-37
-------
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 ln(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 fiber/cc (Ratio = 3.5)b
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 = 0.0711
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 fiber/cc-yr (Ratio = 3.2)
"Results in this table are from two different data sets with different number of individuals and thus, the AICs for
the models cannot be compared between the two data sets.
bFor the model of Marysville workers who underwent health evaluations in 2002-2005 (without regard to date of
hire), both the BMC and BMCL at 28 years TSFE were approximately 3-fold lower than the values in the primary
analysis
For the Marysville data, the model with cumulative exposure did not fit well on the other
measure of model fit. In the larger group of workers evaluated in 2002-2005 (regardless of hire
date), fit of cumulative exposure had a low/>-value (i.e., <0.10) for the Hosmer-Lemeshow
goodness-of-fit statistic. Mean exposure did provide an adequate fit and was, therefore, carried
forward in the analysis as the primary basis of the derivation of the lifetime RfC in Section 5.2.3.
Although the model using cumulative exposure did not show an adequate fit based on the
Hosmer-Lemeshow goodness-of-fit test, the model fit as measured by the AIC was nearly
equivalent to the AIC for the mean exposure model, and the beta estimated for the effect of
TSFE was similar to that estimated in the model using mean exposure. Therefore, the derivation
of a chronic RfC based on the CE model is also shown in Section 5.2.3 for comparison.
5-38
-------
EPA concluded in Section 5.2.2.2 that the subcohort with the better exposure data yields
the more reliable estimate of the relationship of LPT to exposure concentration for the derivation
of the RfC. The primary analysis (Table 5-9) applying the hybrid model to the subcohort had
fewer cases and lower exposure concentrations (which are more relevant to environmental
exposure levels) than the larger cohort and hence, the statistical power for the primary analysis is
diminished and the/>-value for the beta for exposure parameter in the subcohort is 0.078. The
EPA's Benchmark Dose Technical Guidance [U.S. EPA (2012), see Figure 2A, p. 16] addresses
the general issue of when there is a lack of statistical significance due to low power and states
that studies of low power remain feasible for BMD modeling. The Benchmark Dose Technical
Guidance [U.S. EPA (2012), p. 15-16] further notes that when other data support a conclusion
that there is a relationship with dose, it can be acceptable to use results from a database for which
the relationship is not statistically significant. There exists a clear exposure-response
relationship between LAA and the risk of pleural abnormalities in the original study report
\p < 0.001; Rohs et al. (2008)], and there exists a clear exposure-response relationship in the
workers who underwent health evaluations in 2002-2005 (p<0.0006; see Table 5-9). Therefore,
while the effect estimate for exposure is less precise in the subcohort than in the larger group of
workers, this estimate is considered to be a more reliable estimate of the true effect of LAA on
the risk of developing LPT for the purpose of deriving the RfC.
In the model using mean exposure intensity in the larger data set of individuals evaluated
in 2002-2005, the beta coefficient for TSFE was 0.1075. This coefficient value was transported
to the equivalent model in the subset of workers hired in 1972 or later; at a TSFE of 28 years (the
median in this group of workers), the BMC and BMCL were 0.092 fiber/cc and 0.026 fiber/cc,
respectively. The two-step hybrid modeling results in a higher BMCL: both the BMC and
BMCL from the modelling of the larger cohort were about 3-fold lower at TSFE=28 years than
the preferred values from the hybrid model.
Ideally, the objective of the RfC derivation is to compute the BMCL for 70 years (i.e., a
lifetime) of exposure. However, the maximum observed TSFE in the primary cohort was
32.6 years, which while clearly a chronic exposure, cannot be expected to approximate a lifetime
exposure in this particular circumstance when TSFE is the most important predictive factor for
the prevalence of LPT. The BMCL values calculated at longer TSFEs were very low (e.g., a
BMCL of 2.7 x 1Q-6 at TSFE of 70 years) and the ratio of BMC to BMCL grows exponentially
such that these values at 70 years are considered to be unreliable (e.g., the BMC:BMCL ratio is
1,000 at TSFE of 70 years). This is likely because longer TSFE values are extrapolated well
outside of the range of the data, and attempting to extrapolate beyond -30 years leads to greater
statistical uncertainty. Consequently, the BMC and BMCL values corresponding to 28 years
were selected as the primary modeling result, with the BMCL of 0.026 fiber/cc serving as the
POD for RfC derivation with some additional accommodation of the uncertainty in using less
than lifetime exposure data to estimate lifetime risks.
5-39
-------
At the BMCL (2.6 x icr2 fiber/cc), the final model leads to an estimated probability of
LPT of 0.06 at 28-years TSFE, and 0.61 at 70-years TSFE (see Figure 5-4). This 10-fold
increase in the probability of the critical effect going from the median TSFE in the cohort, to the
full lifetime, is used to derive a data-informed subchronic to chronic uncertainty factor (UF) for
RfC derivation in the following section.
P(LPT) at the BMC and at the BMCL, estimated at TSFE=28 years
DH model with plateau=85%, mean exposure (flee)
0.8
0.7
S 0.4
0.3
0.2
0.1
0.0
10
20
30 40
TSFE (years)
PLOT + + + P(LPT)forBMCof0.0923f/cc * * * P(LPT)for BMCL of 2.6E-2ffcc
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.
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)
Among the available studies that could provide exposure-response data for the
relationship between LAA exposure and risk of LPT, consideration of study attributes led to the
selection of a study of the Marysville, OH workers evaluated in 2002-2005 as the primary data
5-40
-------
set for RfC derivation [(Rons et al., 2008) see Section 5.2.1]. An updated job-exposure matrix
was available for this follow-up of the original study group, with a refined understanding of
exposure to LAA throughout plant operation (see Section 5.2.3.1 and Appendix F). However,
due to remaining uncertainties in exposures prior to 1972, EPA elected to focus on
exposure-response modeling for the subgroup of plant employees hired in 1972 or later (see
Section 5.2.3.2). The critical effect selected for derivation of the RfC is LPT, a persistent change
to normal tissue structure associated with decreased pulmonary function.
Using a 10% BMR for LPT, a BMC of 0.092, and a BMCLio of 0.026 fiber/cc were
calculated for the mean exposure model (see Table 5-9). Following EPA practices and guidance
(U.S. EPA, 2002, 1994b), application of the following default and data-informed UFs was
evaluated resulting in a composite UF of 300.
• An interspecies uncertainty factor, UFA, of 1 is applied for extrapolation from animals
to humans because the POD used as the basis for the RfC was based on human data.
• An intraspecies uncertainty factor, UFn, of 10 was applied to account for human
variability and potentially susceptible individuals. Only adults sufficiently healthy
for full-time employment were included in the principal study and the study
population was primarily male. Other population groups, such as the elderly, women,
children, and those with preexisting health conditions, were not evaluated in the
principal study but may have a different response to LAA exposure.
• An uncertainty factor for extrapolating from a lowest-observed-adverse-effect-level
(LOAEL) to a no-observed-adverse-effect-level (NOAEL), UFL, of 1 was applied
because this factor was addressed as one of the considerations in selecting a BMR for
BMC modeling.
• A database uncertainty factor, UFo, of 3 was applied to account for database
deficiencies in the available literature for the health effects of LAA.
Although a large database exists for asbestos in general, only four study populations
exist for LAA specifically: the Minneapolis, MN community study, the Marysville,
OH worker cohort, the Libby worker cohort, and the ATSDR community screening
(which includes some Libby worker cohort participants). Studies conducted in three
of these populations, the Libby worker cohort (Larson et al., 2012a), Minneapolis
community study (Alexander et al., 2012), and Marysville workers (Rohs et al.,
2008), have all demonstrated substantial numbers of LPT cases occurring at the
lowest exposure levels examined in each study (Christensen et al., 2013), lending
confidence to the use of LPT as a critical effect and (Rohs et al., 2008) as the
principal study for RfC derivation.
However, studies in the Libby population have also demonstrated an association
between exposure to LAA and autoimmune effects (Marchand et al., 2012; Noonan et
al., 2006; Pfau et al., 2005). Because these studies did not provide exposure-response
information, it is unknown whether a lower POD or RfC would be derived for these
effects. For other (non-Libby) forms of amphibole asbestos, there is evidence for
5-41
-------
autoimmune effects from a study of individuals in a community exposed to tremolite.
In the Metsovo population, there were changes in immune parameters in tremolite-
exposed individuals without pleural plaques, and additional immune markers
(including autoantibodies) were increased in individuals with pleural plaques (Zerva
et al., 1989). Also it has been hypothesized that shorter asbestos fibers reach the
pleura via passage through lymphatic channels (Peacock et al., 2000), although
experimental evidence is lacking for this or alternative potential mechanisms of fiber
migration. This uncertainty in the sequence of health effects (pleural or autoimmune)
is the basis for selecting a UFo of 3.
In addition, this assessment applied a data-informed UF. A data-informed
subchronic-to-chronic uncertainty factor, UFs, of 10 was applied because, for this
particular health endpoint, even -30 years of observation (Rohs et al., 2008) is
insufficient to describe lifetime risks. Although chronic exposure has been generally
defined as more than approximately 10% of lifetime, the EPA's RfC guidance (U.S.
EPA, 1994b) states that for human data "[t]he best data to use for calculating an RfC
would be a population study of humans that includes sensitive individuals exposed for
lifetime or chronic duration, and that evaluates the critical endpoint or an appropriate
early marker for the disease.... However, the amount of exposure in a human study
that constitutes subchronic is not defined, and could depend on the nature of the effect
and the likelihood of increased severity or greater percent response with duration."
Similarly, the 2014 Guidance for Applying Quantitative Data to Develop Data-
Derived Extrapolation Factors for Interspecies and Intraspecies Extrapolation (U.S.
EPA, 2014) advises that "[extrapolation is most scientifically robust when data are
first evaluated before using defaults." Consistent with those documents, EPA first
evaluated the data on follow-up time (in this case TSFE) both with respect to the data
on LAA and what is known from the literature about other amphiboles and asbestos
generally before considering extrapolation.
For LAA, the study by Rohs et al. (2008) is a follow-up on the Lockev et al. (1984)
investigation, and while the earlier study showed the prevalence of all pleural
abnormalities together as 2.0% in 1980 (see Table 4-8), the study by Rohs et al.
(2008) showed that in 2002-2005, the prevalence of LPT increased to 26.2% (in
workers without other asbestos exposure; see Table 5-3 and Figure 5-2), a 13.1-fold
increase in 22-23 years with very low additional exposure during the additional years
of follow-up.
There are no epidemiologic data on the relationship between TSFE and the
prevalence of pleural plaques for other amphiboles. There are data on other
epidemiologic cohorts exposed to general asbestos. While the type of fiber exposure
is not defined in any of the other studies of pleural plaques and TSFE, it could be
reasonably inferred from listed occupations to be either mostly chrysotile or mixed
chrysotile and amphibole exposures.
One study of shipyard workers likely exposed to mostly chrysotile that also used x-
ray to diagnose pleural plaques (Jarvholm, 1992) presented data by 5-year intervals of
TSFE from 0 to 54 years. Although people with higher TSFE were likely exposed to
higher concentrations (due to relative lack of historical industrial hygiene), the
prevalence of having pleural plaques was consistently greater with longer TSFE and
the prevalence increased with increasing time. The prevalence increased from 1.7%
5-42
-------
at 12 years of TSFE to 24% at 27 years of TSFE and then to 86% at 52 years of
TSFE—a rise of 50-fold in 40 years. A 3.6-fold increase was observed for the last 25
years. The prevalence at 27 years of TSFE is similar to prevalence as in the
Marysville, OH cohort which had a 26.2% prevalence at about 33 years of TSFE.
One study of multiple occupations with undetermined asbestos exposure, likely
mostly chrysotile or mixed chrysotile and amphibole exposures, that used HRCT to
diagnose pleural plaques (Paris et al., 2009) presented data by quartiles of TSFE.
Note that, since HRCT is known (ATS, 2004) to identify 1.5-2 times more plaques
than x-ray detects, the relationship with TSFE may be different than with x-ray data.
In this FIRCT study, the prevalence increased 3.2 fold in 42 years, from 11.7% at 18
years of TSFE to 37.6% at 60 years of TSFE. However, unlike Jarvholm (1992). the
intervals of TSFE in this study were of widely divergent length. For example, the
growth between two narrowly defined intervals of similar length was 1.7-fold in only
5.5 years from TSFE of 39.5 years to TSFE of 45 years. Note that in both of these
studies, the growth rate was consistent, but uneven over periods of observations.
It is clear from supporting studies on exposure to other types of asbestos (some of
which use HRCT rather than x-ray data and therefore are identifying a different
outcome) that the qualitative pattern holds true that prevalence of pleural plaques
continue to increase with TSFE as high as 50-60 years, but the range of rates varies
across these studies from a 3.2-fold increase after 42 years to a 50-fold increase after
40 years TSFE. Data specific to LAA (where x-rays were used to diagnose pleural
plaques or LPT) shows a 13.1-fold increase in the 22-23 years observed between the
original analysis of the LAA-exposed cohort by Lockey et al. (1984) and the follow-
up study by Rohs et al. (2008).
The risk of pleural plaques, and thus LPT, continues to increase throughout life (even
with less-than-lifetime exposures). Because the RfC is intended to apply to
noncancer effects due to lifetime exposure, the limitation in the observed TSFE in the
principal study and in the supporting epidemiologic data strongly suggests
extrapolation from known exposures to lifetime exposures.
As the first attempt at extrapolation, EPA considered basing the point of departure on
the modeled BMCL for lifetime exposure and follow-up (TSFE = 70 years) and the
target benchmark response rate of 10% extra risk. However, there is considerable
uncertainty in that extrapolation as indicated by a ratio of 1,000 when comparing
lifetime BMC to lifetime BMCL (BMC7oy = 0.0027: BMCL70y = 2.7 x 10'6). Next,
EPA examined the ratio of the central estimates from the model. EPA calculated the
ratio of the model-predicted BMC at TSFE of 28 to the modeled BMC at TSFE of 70
years; that ratio is approximately 34.
As noted at the end of Section 5.2.2.6.2, EPA also used the model to examine the
estimated increase in the central estimate of the response rate for increases in TSFE
from the 28 years used to derive the point of departure to a TSFE of 70 years. That
ratio depends on the concentration whose risk is being evaluated, so EPA calculated
that ratio at the point of departure concentration (0.026 fiber/cc) that is the primary
modeling result. For that concentration, the central estimate of the risk at TSFE = 70
years is -10-fold greater than the central estimate of the risk at TSFE = 28 years
(from 6% to 61%) as shown in Figure 5-4.
5-43
-------
There is some uncertainty as to whether the prevalence would increase in the same
manner for additional years of TSFE outside the range of observed data as it does
within the range of the observed data on LAA. There are limited non-Libby asbestos
data and extrapolations that point to an uncertainty value of about 3. However,
having considered the epidemiologic data and the potential extrapolation information,
EPA concluded that a data-informed UFs of 10 is appropriate in this situation where
most of the epidemiologic data, including the LAA data, predicts a 10-fold increase
or more in the prevalence of pleural plaques (or LPT). Some extrapolations project a
factor that might be as high as 34-fold, but only if the rate of increase remained the
same at larger values of TSFE. For these reasons, a data-informed UFs=10 was
applied.
Composite UF= UFA x UFH x UFL x UFD x UFS= 1 x 10 x 1 x 3 x 10 = 300
The derivation of the RfC from the morbidity studies of the Marysville, OH worker
cohort [i.e., Rohs et al. (2008)1 was calculated from a POD, BMCLio for LPT of 0.026 fiber/cc,
and dividing by a composite UF of 300. As derived below, the chronic RfC is 9 x 10~5 fiber/cc
for LAA:
Chronic RfC for LPT = BMCLio - UF (5-3)
= 0.026 fiber/cc - 300
= 8.67 x icr5 fiber/cc, rounded to 9 x icr5 fiber/cc
Note that for the primary RfC for LAA and all the alternative RfCs for LAA, the fiber
concentrations are presented here as continuous lifetime exposure in fibers/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 jim. Historical PCM analysis (1960s and early 1970s) generally
had less resolution, and fibers with minimum widths of 0.4 or 0.44 jim 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.
While this assessment is informed by studies of other types of asbestos, it is not a
complete toxicity review of other amphiboles or of chrysotile asbestos.
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
As shown in the uncertainty analyses in Section 5.3.5 (see Table 5-17), use of an
alternative critical effect of APT results in an almost identical BMCLio as derived for the
primary analysis using LPT as the critical effect. In the subcohort, the number of cases of APT
is identical to the number of cases of LPT, and in the larger group of workers evaluated in
5-44
-------
2002-2005 used to estimate the effect of TSFE, the number of cases of APT is very similar to
the number of LPT cases (n = 69 cases of APT and n = 66 cases of LPT). In this larger group,
the regression coefficient for the effect of TSFE was very close when using either of these two
endpoints—values of 0.1108 for APT and 0.1075 for LPT. Thus, using the same uncertainty
factors as described above, the chronic RfC is 9 x 10"5 fiber/cc for LAA, using an alternative
endpoint of APT:
Chronic RfC for APT = BMCLio-UF (5-4)
= 0.027 fiber/cc - 300
= 9.0 x io~5 fiber/cc, rounded to 9 x 10'5 fiber/cc
This value is identical to the primary RfC derived using a critical effect of LPT, when
rounded to one significant digit. These results provide additional support to further substantiate
the primary RfC.
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
As shown in the uncertainty analyses in Section 5.3.5 (see Table 5-17), use of an
alternative critical effect of any radiographic change (ARC) results in an almost identical
BMCLio as derived for the primary analysis using LPT as the critical effect. In the subcohort,
the number of cases of ARC is identical to the number of cases of LPT, and in the larger group
of workers evaluated in 2002-2005 used to estimate the effect of TSFE, the number of cases of
APT is very similar to the number of LPT cases (n = 71 cases of ARC, and n = 66 cases of LPT).
In this larger group, the regression coefficient for the effect of TSFE was very close when using
either of these two endpoints—values of 0.1115 for ARC and 0.1075 for LPT. Thus, using the
same uncertainty factors as described above, the chronic RfC is 9 x 10~5 fiber/cc for LAA, using
an alternative endpoint of ARC:
Chronic RfC for ARC = BMCLio-UF (5-5)
= 0.027 fiber/cc - 300
= 9.0 x io~5 fiber/cc, rounded to 9 x 10'5 fiber/cc
This value is identical to the primary RfC derived using a critical effect of LPT, when
rounded to one significant digit. These results provide additional support to further substantiate
the primary RfC.
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
As shown in the analyses in Section 5.2.2 (see Table 5-9), the model with TSFE and CE
yields a similar overall fit to the model based on TSFE and mean exposure as judged by AIC
values within two units (242.6 vs. 242.7). However, the CE model yielded an unacceptably low
value for the Hosmer-Lemeshow goodness-of-fit test (p =0.04 <0.1) when also including TSFE
in the model, and therefore, the mean exposure model was selected for the derivation of the RfC.
This same pattern was corroborated in the analysis of the combined cohort in Appendix E where
the same bivariate Dichotomous Hill model with plateau fixed at 85% passed the
Hosmer-Lemeshow goodness-of-fit test for the model with TSFE and mean exposure but yielded
an unacceptably low value for the model with TSFE and CE (p = 0.006 <0.1; see Table E-4).
However, as CE has been a traditional exposure metric of asbestos exposure, there may
be interest in an estimate of the RfC for such a model based on TSFE and CE. The use of an
alternative model based on the TSFE and CE data yields a BMCLio of 0.577 fiber/cc-yr using
LPT as the critical effect among individuals evaluated in 2002-2005 and hired in 1972 or later at
a TSFE of 28 years (the median in this group of workers). In order to adjust the POD to units of
concentration (fibers/cc), this POD was divided by the mean duration of occupational exposure
(18.23 years) to yield an adjusted BMCLio of 0.0317 fiber/cc. Using the same uncertainty
factors as described above, the value of an RfC for LAA based on a model fit to TSFE and CE
would be 1 x 10"4 fiber/cc:
Chronic RfC for LPT = BMCLio - UF (5-6)
= 0.0317 fiber/cc-300
= 1.06 x io~4 fiber/cc, rounded to 1 x 10'4 fiber/cc
This value is approximately 11% higher than the primary RfC derived using a critical
effect of LPT and a model fit to TSFE and C. As noted previously, this result using CE is
included for comparative purposes with the primary RfC, but was not selected as the primary
RfC as the model based on CE did not indicate an adequate fit to the Marysville subcohort of
workers evaluated in 2002-2005 and hired in 1972 or later when also including TSFE in the
model.
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
EPA also conducted modeling of the full Marysville cohort to substantiate the derivation
of the RfC in the primary analytic subset of workers evaluated in 2002-2005 and hired in 1972
or later and considered in the previous sections. The "combined cohort" was assembled using all
5-46
-------
individuals who participated in the health examination in 1980 (Lockey et al., 1984) and
2002-2005 (Rohs et al., 2008), and who were not exposed to asbestos from a source outside of
the Marysville facility (see Table 5-3 and Appendix E for details). Due to differences in the
1980 x-ray evaluations compared with the 2002-2005 x-ray evaluations, an alternative critical
effect of APT was used in the combined cohort modeling.
The modeling of the combined cohort is described in detail in Appendix E. A suite of
model forms (univariate and bivariate) and exposure metrics (mean exposure, cumulative
exposure, RTW exposure) were evaluated using the combined cohort of 434 individuals. Two
models (Cumulative Normal Dichotomous Hill and Cumulative Normal Michaelis-Menten
models), which incorporated TSFE into the plateau term rather than as an independent predictor
alongside the exposure metric, were also evaluated as a supplement to the standard suite of
models.
The results for the BMC and BMCL for the three endpoints (LPT—defined as LPT
diagnosed in 2002-2005 and PT in 1980, APT, and ARC) are presented in Tables E-3, E-4, and
E-5. Any pleural thickening is selected as the preferred endpoint for the alternative derivation of
the reference concentration (RfC) for the combined cohort because the endpoint of APT is more
inclusive (it includes those with LPT and those with DPT in the absence of LPT) and eliminates
the uncertainty regarding the type of pleural thickening observed in the 1980 study (Lockey et
al., 1984) using the 1971 ILO guidance. The two models selected for derivation of an RfC were
(1) the same bivariate Dichotomous Hill model with fixed plateau used in the primary analysis
based on TSFE and mean concentration and (2) the cumulative normal Dichotomous Hill model
with fixed background rate of APT based on TSFE and CE. Additionally, in keeping with the
primary analysis of the subcohort hired in 1972 or later, an analogous hybrid analysis of the
subcohort based on the combined cohort was also conducted based on the two selected models.
All of these results are presented at a value of TSFE = 70 years for those models that were able
to reasonably extrapolate the TSFE value outside the observed range (47 years maximum) and
for the median value of TSFE in the combine cohort (25 years) for models where the
extrapolation to 70 years was not reasonable.
Following EPA practices and guidance (U.S. EPA, 2002, 1994b) as discussed in
Section 5.2.3, a composite uncertainty factor (UF) of 300 is used when deriving the RfC from the
POD calculated at the median TSFE (25 years). This includes an uncertainty factor of 10 to
account for intraspecies variability (UFn = 10), a factor of three to account for database
uncertainty (UFo = 3) and an extra factor (UFs) of 10 to account for the lack of information on
people at risk for a complete lifetime (UFs = 10). When using the POD based on the BMCL
calculated at TSFE = 70 years, the additional adjustment factor of 10 is not necessary and a
composite UF of 30 is used (UFn =10 and UFo = 3). The calculations of the RfC for the
combined cohort and the Rohs subcohort using both options are shown in Table 5-10. The RfCs
are rounded to one significant digit.
5-47
-------
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)
CN DH (CE,TSFE) a
CN DH (CE,TSFE)
CN DH (CE,TSFE) a
Calculation
RfC = (3.4 x 10-2)/300 = 1 x 10-4fiber/cc
RfC = (6.3 x 10~2)/300 = 2 x 10~4 fiber/cc
RfC = (3.5 x 10~2)/300 = 1 x 10~4 fiber/cc
RfC = (7.5 x 10~4)/30 = 3 x 1Q-5 fiber/cc
RfC = (8.4 x 10-4)/30 = 3 x IQ-5 fiber/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).
^Hybrid model informed by the combined cohort; background rate of APT was fixed at 3% to facilitate model
convergence.
For comparison, the above values all fall within approximately threefold when compared
to the primary RfC for LPT of 9 x 1CT5 fiber/cc derived in Section 5.2.3 from the Marysville
workers who underwent health evaluations in 2002-2005 and were hired in 1972 or later. These
results provide additional support to further substantiate the primary RfC.
5.2.6. Summary of Reference Concentration Values (RfCs) for the Different Health
Endpoints and Different Sets of Workers in the Marysville Cohort
The primary derivation of the reference concentration is based on the critical effect of LPT
in the Marysville workers who underwent health evaluations in 2002-2005 and were hired in
1972 or later. Multiple alternative values were derived using other health endpoints, other
regression models, and other sets of workers from the Marysville cohort. The results of the
different derivations are shown in Table 5-11. The range of values is from 3 x 10~5 fiber/cc to
2 x 10~4fiber/cc.
5-48
-------
Table 5-11. Multiple derivations of a reference concentration from the Marysville, 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
Hired >1972
Hired > 1972
Hired > 1972
Hired > 1972
All hires
All hires
All hires
LPT
APT
ARC
LPT
APT
APT
APT
Measured
Measured
Measured
Measured
Measured > 1972
Measured > 1972
Measured > 1972
None
None
None
None
None
None
None
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
CNDH;TSFE = 25yr
Hybrid combined cohort
DH85; TSFE = 25 yr
TSFE and C
TSFE and C
TSFE and C
TSFE and CE
TSFE and CE
TSFE and C
TSFE and CE
9X1Q-5
9xl(T5
9xl(T5
lxl(T4
ixicr4
2xlO-4
lxl(T4
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
5-49
-------
Table 5-11. Multiple derivations of a reference concentration from the Marysville, OH cohort. Primary RfC value in bold.
(continued)
Marysville,
OH
Marysville,
OH
All hires
All hires
APT
APT
Measured > 1972
Measured > 1972
None
None
Measured
Measured
—
—
Combined cohort
CN DH; TSFE = 70 yr
Hybrid combined cohort
DH85; TSFE = 70 yr
TSFE and CE
TSFE and CE
3xlO~5
3xlO~5
5.2.5
App. E
5.2.5
App. E
Abbreviations: Health endpoint (LPT = localized pleura! 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,
DHgs = 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).
5-50
-------
5.3. UNCERTAINTIES IN THE INHALATION REFERENCE CONCENTRATION
(RFC)
Some sources of uncertainty remain in the derivation of the RfC. This section identifies
the major sources of uncertainty and, where possible, uses a sensitivity analysis to evaluate the
potential impact of these uncertainties on the point of departure.
5.3.1. Uncertainty in the Exposure Reconstruction
As in all epidemiologic studies, uncertainties are present in the exposure assessment. In
this case, some uncertainty lies in the employment history, and some individuals had extensive
overtime work. Employment history was self-reported during interviews with each individual
for the original study (Lockey et al., 1984), and any errors in this process could affect the LAA
exposure estimates. While the uncertainties related to a lack of quantitative measurements are
not relevant to the analysis of workers hired in 1972 or later, it is important to recognize that
exposure assessment post-1972 also has some limitations. The main source of uncertainty is
incomplete exposure measurements for some of the occupations/tasks before industrial hygiene
improvements that started about 1973 or 1974 and continued throughout the 1970s (see
Appendix F, Figure F-l).
Some uncertainty exists when the Libby ore was first used in the facility. Company
records indicated that the date was between 1957 and 1960, and the University of Cincinnati
used the best available information from focus group interviews to assign 1959 as the year of the
first usage of Libby vermiculite ore (see Appendix F). In 1957 and 1958, only vermiculite ore
from South Carolina was thought to be used. From 1959 to 1971, vermiculite ores from both
Libby, MT and South Carolina were used. From 1972 to 1980, vermiculite ores from Libby,
MT, South Carolina, South Africa, and Virginia were used, with Libby vermiculite ore being the
major source. Libby vermiculite ore was not used in the facility after 1980. However, industrial
hygiene measurements (based on PCM) collected after 1980 showed low levels of fibers in the
facility. PCM analysis does not determine the mineral/chemical make-up of the fiber and, thus,
cannot distinguish among different kinds of asbestos.
Uncertainty also exists in the data regarding the asbestos content in other vermiculite ore
sources before and after the Libby ore was used. As reported in Appendix C, EPA analysis of
bulk vermiculite ores from Virginia and South Africa showed the presence of only a few or no
amphibole asbestos fibers. The South Carolina vermiculite ore contained relatively more fibers
than the Virginia and South African vermiculite ores but still far fewer fibers than the Libby
vermiculite ore. Using the industrial hygiene data, the University of Cincinnati estimated that
the fiber content of the South Carolina ore was about 8.7% of that of the Libby ore (see
Appendix F). This result is consistent with data comparing South Carolina and Libby
vermiculite ores from samples tested in 1982 (U.S. EPA, 2000a). Based on the industrial
hygiene data, the concentration of fibers detected in the workplace was near background after
5-51
-------
1980. The exposure distribution in Marysville workers is summarized in Table 5-12, which
shows the mean, cumulative, and RTW exposure metrics for the full cohort of workers, the
subset of those evaluated in 2002-2005, and the primary analytic group of workers evaluated in
2002-2005 and hired in 1972 or later. To evaluate the potential impact of assumptions regarding
exposure after 1980 when the use of Libby vermiculite ores was considered to have ceased, EPA
conducted a sensitivity analysis using the hybrid Dichotomous Hill model with plateau fixed at
85%, but truncating exposures at 12/31/1980 (see Table 5-13). Note that except where stated
otherwise, this model (hybrid Dichotomous Hill model with plateau fixed at 85%) was used for
all sensitivity analyses. Using the truncated exposures, the POD increased from
2.6 x 10"2 fiber/cc, to 3.9 x 10"2 fiber/cc. However, note that for the model with truncated
exposures, the modeling in the larger group of workers evaluated in 2002-2005 showed a poor
fit (Hosmer-Lemeshow GOF/?-value <0.10). These results are included for comparative
purposes but should be interpreted with caution.
5-52
-------
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 (fibers/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 (fibers/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 (fibers/cc-yr), calculated using midpoint of season dates
Arithmetic
mean
Geometric
mean
193.3093
(519.3874)
Range:
0.0007-3,500.66
72.2260
(204.1052)
Range:
0.0003-1,373.22
19.4767
(4.2550-78.0944)
4.1835
(0.9229-25.2179)
294.38 (687.95)
Range:
0.12-3,500.66
110.14(270.86)
Range:
0.05-1,373.22
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 (fibers/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 (fibers/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)
5-53
-------
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 1980 (fibers/cc-yr), calculated using midpoint of season dates
Arithmetic
mean
Geometric
mean
189.5463
(517.2418)
Range:
0.0007-3,475.17
70.9061
(203.4978)
Range:
0.0003-1,365.69
16.0931
(2.7315-69.9986)
2.4223
(0.6805-23.5698)
287.78(685.61)
Range:
0.12-3,475.17
107.84 (270.24)
Range:
0.05-1,365.69
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.
5-54
-------
Table 5-13. Effect of truncating exposures after 1980 and of using
arithmetic or geometric mean to summarize multiple fiber measurements
Hosmer-Lemeshow
GOF/>-value
Model AIC
Alpha (intercept)
Bkg (background)
Beta for TSFE
Beta for In (mean
exposure)
BMCandBMCLat
28 yr (fibers/cc)
Hosmer-Lemeshow
GOF/>-value
Model AIC
Alpha (intercept)
Bkg (background)
Beta for TSFE
Beta for In (mean
exposure)
BMCandBMCLat
28 yr (fibers/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)
OH
0. 1075 (SE = 0.0281),
p = 0.0002
0.4819 (SE = 0.1390),
p = 0.0006
0.0923 and 2.6 x 1Q-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)
OH
0. 1234 (SE = 0.0276),
p< 0.0001
0.4012 (SE = 0.1213),
p = 0.0011
0.0298 and 9.1 x 10~3
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
0.276 land 3. 9 x 1Q-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),
p = 0.2659
Individuals with
radiographs in
2002-2005 (to get
beta for TSFE)
0.02812
245.6
-4.1735
(SE= 1.0727)
OH
0.1266
(SE = 0.0276),
p< 0.0001
0.3256
(SE = 0.1015),
;? = 0.0015
0.0796 and 9.9 x 1Q-3
5-55
-------
Another potential source of uncertainty in the exposure reconstruction is the method used
to average multiple fiber measurements for a given location (within the Marysville facility) and
time period. The arithmetic mean of multiple measurements was used here, but an alternative
approach would be to use the geometric mean, which has the effect of "dampening" outliers or
extreme values. EPA conducted a sensitivity analysis for exposures estimated using the
geometric mean of multiple fiber measurements and found that the PODs were lower than those
estimated for the primary analysis (9.1 x 10~3 fiber/cc considering all years of exposure, and
9.9 x 10~3 fiber/cc if exposures were truncated after 1980) (see Table 5-13). The lower PODs are
a result of the approximately threefold lower exposures estimated for each individual worker
when using the geometric mean; in the Marysville workers evaluated in 2002-2005 and hired in
1972 or later, the median of each individual's mean exposure intensity estimated using the
arithmetic versus the geometric mean of multiple fiber measurements in the plant were
0.0234 fiber/cc and 0.0085 fiber/cc, respectively. Note that for both models utilizing geometric
mean exposures, the modeling in the larger group of workers evaluated in 2002-2005 showed
poor fit (Hosmer-Lemeshow GOF/>-value <0.10). These results are included for comparative
purposes but should be interpreted with caution.
Potential coexposure to other chemicals was present in the Marysville facility (see
Section 4.1.2.2.2). These other chemicals were used after expansion of vermiculite ore in
another area of the facility. Industrial hygiene data showed very low levels of fibers in the areas
where the additional chemicals were added to the expanded vermiculite. In addition, none of
these chemicals are volatile. The most likely route of exposure to these chemicals is through
dermal contact. It is unlikely that any coexposure to these particular chemicals would alter the
exposure-response relationship of LAA in the respiratory system (see Section 4.1.2.2.2).
The University of Cincinnati research team assumed no exposure to LAA occurred
outside of the workplace for the Marysville workers. The interviews with the Marysville
workers revealed that about 10% of the workers reported bringing raw vermiculite home. These
interviews also revealed that changing clothes before leaving the workplace was standard
practice at the end of the shift, and approximately 64% of the workers showered before leaving
the workplace. For these workers, it is likely that additional exposure outside the workplace was
minimal. However, for the remainder of the workers, it is reasonable to assume that additional
exposure could have occurred at home. Additional data collected by the University of Cincinnati
research team document that no increased prevalence of pleural or parenchymal change
consistent with asbestos exposure was observed in household contacts of the workers from the
Marysville facility (Hilbert et al.. 2013).
5.3.2. Uncertainty in the Radiographic Assessment of Localized Pleural Thickening (LPT)
The use of conventional radiographs to diagnose pleural thickening has several
limitations. More severe and larger lesions are more reliably detected on radiographs. There are
5-56
-------
also potential interferences. Fat pads may be mistaken for pleural plaques or LPT because they
generally occur against the ribcage in a similar location (Gilmartin, 1979): this is one source of
potential disagreement among x-ray readers. Often signs of trauma (e.g., fractured ribs) and
radiographic signs of past tuberculosis infection can be seen and are noted by the reader. In
these cases, LPT would not be diagnosed. There is a certain amount of subjectivity when
viewing the x-rays in determining which features are representative of pleural thickening and
whether signs of alternative etiology can be noted; thus, multiple certified readers are generally
consulted, and a consensus of opinions determines the diagnosis. Regardless, the potential for
outcome misclassification still exists. However, uncertainty in the presence or absence of
localized pleural thickening in each individual is decreased by the use of three highly qualified
chest radiologists evaluating the radiographic films and the use of the majority vote of the
readers for the diagnosis.
BMI was investigated as a potential explanatory variable because fat pads can sometimes
be misdiagnosed as pleural thickening. The effect of such outcome misclassification would be to
attenuate the observed association between exposure and outcome. In the Marysville data, BMI
was not measured in the 1980 examination but was available for most participants of the 2000s
examination (available for most of those in the data sets used to derive the RfC). To address
whether fat deposits may affect outcome classification, EPA considered the effect of adding BMI
as a covariate in the model. However, BMI did not display an association with odds of localized
pleural thickening in this population (p = 0.6933).
5.3.3. Uncertainty Due to Potential Confounding
Along with the effect of BMI, other covariates were also evaluated for potential
confounding of the association between LAA exposure and LPT in the Marysville workers (see
Section 5.2.2.6.1). Covariates included both demographic characteristics (gender, smoking
status, BMI) as well as potentially exposure-related factors (hire year, job tenure, exposure
duration, and age at x-ray).
Smoking is a particularly important variable to consider when evaluating respiratory
health outcomes. Although data are mixed, a few studies suggest smoking may affect the risk of
developing asbestos-related pleural thickening or timing of such pleural thickening development.
However, no studies were identified that assessed the relationship between LPT specifically and
any measure of smoking status. Plaques as defined in earlier ILO classification systems have not
been associated with smoking in asbestos-exposed workers (Mastrangelo et al., 2009; Paris et al.,
2009: Koskinen et al.. 1998).
Some evidence indicates that small interstitial opacities (asbestosis) and asbestos-related
DPT may be associated with smoking. Studies among populations exposed to other general
types of asbestos have reported mixed effects on the impact of smoking on risk of radiographic
abnormalities; two studies reported a significant association between risk of all pleural
5-57
-------
thickening, including both pleural plaques and DPT (McMillan et al., 1980), or any pleural
abnormality (Welch et al., 2007) and smoking after controlling for some measure of asbestos
exposure. A larger number of studies reported borderline associations when examining risk of
pleural changes (Adgate et al., 2011: Paris et al.. 2008: Dement et al.. 2003: Zitting et al.. 1996:
Yano et al., 1993: Lilis et al., 1991: Baker et al., 1985) or no association with smoking (Soulat et
al.. 1999: Nerietal.. 1996: Ehrlich et al.. 1992: Delclos et al.. 1990: Rosenstock et al.. 1988).
Possible reasons for the different findings include varying quality of smoking information (some
used categories of ever/never or former/current/never, while others used pack-years) and
differences in the specific outcome studied.
Rohs et al. (2008) did not find a difference in smoking prevalence among those with and
without any radiographic changes but also did not report results controlling for LAA exposure,
or for LPT specifically.
None of the potential confounding factors examined were significantly associated with
LPT after controlling for LAA exposure in the primary data set of workers evaluated in
2002-2005 and hired in 1972 or later. However, the effect of each covariate was reexamined in
the primary (hybrid) model which utilized information from the larger set of workers evaluated
in 2002-2005, regardless of hire date (see Table 5-14). Each covariate was included (one at a
time) in the primary model, and the statistical significance and effect on the model parameters
evaluated. None were statistically significantly related to risk of LPT, with ^-values for the
corresponding beta coefficients ranging from 0.1533 (gender) to 0.9858 (hire year). Note that
because these main effects were not statistically significantly associated with LPT, they would
not be expected to modify the association between LAA exposure (controlling for TSFE) and
LPT. It is unlikely that there would be opposing effects for exposure and the covariates
examined that would "cancel each other out" and mask true effect measure modification. Thus,
effect modification was not considered to be an issue in these analyses. It is not surprising that
the time-related factors were not statistically significant (and had little effect on the estimated
beta coefficient for exposure) because these factors were highly correlated with TSFE (which
was already included in the primary model). The effect of exposure was higher when including
gender and smoking status (current, ex- or never smoker) and lower when including ever-smoker
status and BMI. Although the effect of gender was not significant, there were relatively few
women in the Marysville workers population (n= 16 in the workers evaluated in 2002-2005,
and n = 13 in the subgroup hired in 1972 or later). In the primary analytic group, the prevalence
of LPT was not very different by gender, but the comparison is limited by sample size—there
was 1 woman among the total of 13 with LPT (prevalence of 7.7%), while the other 12 cases
(including the individual with LPT and DPT) were among the 106 men (prevalence of 11.3%).
However, women also tended to have lower LAA exposure in this study population—for
example, median cumulative exposure was 0.17 (interquartile range: 0.05, 0.26) fiber/cc-yr
among women, compared with 0.56 (interquartile range: 0.23, 1.78) fibers/cc-yr among
5-58
-------
men—which could explain the lower prevalence. Larson et al. (2012a) found that among Libby
workers (93.2% of whom were male), the prevalence of LPT was 37% among men and 9%
among the 23 women included in the study, but exposure levels by gender were not provided in
the published report. In the Minneapolis community study, the prevalence of all pleural
abnormalities was 16.5% among men, compared to 4.6% among women [adjusted odds ratio and
95% confidence interval of 3.8 (1.6, 8.9); (Alexander et al., 2012)], but again, exposure levels by
gender were not reported. Thus, the potential for different effects by gender merits further
investigation.
In the Marysville workers, the variables representing smoking history (either current
versus ex- versus never smoker, or ever smoker versus never smoker) were not statistically
significant. However, the limited sample size (only three cases were never smokers) and limited
nature of the smoking information precluded further analysis of smoking; thus, further research
is needed on the effect of smoking in relation to LPT risk among asbestos-exposed populations.
5-59
-------
Table 5-14. Effect of including covariates into the final model
Hosmer-Lemeshow
GOF/>-value
Model AIC
Alpha (intercept)
Bkg (background)
Beta for TSFE
Beta for In (mean
exposure)
Beta for covariate
Hosmer-Lemeshow
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),
p = 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),
;? = 0.2548
-0.0422
(SE = 0.1391),
;? = 0.7621
5-60
-------
5.3.4. Uncertainty Due to Time Since First Exposure (TSFE)
Some uncertainty is associated with the length of follow-up of the Marysville cohort.
There was relatively little variation in TSFE among the workers evaluated in 2002-2005 and
hired in 1972 or later. It is anticipated that the prevalence of localized pleural thickening in the
study population will likely continue to increase with passage of time. However, EPA took that
into account both in modeling and in its application of uncertainty factors. EPA utilized
information from the broader group of workers evaluated in 2002-2005 (i.e., regardless of hire
date) and with a wider range of TSFE, to estimate the effect of time. However, because even this
larger group lacked information on full lifetime exposure (maximum TSFE of 47 years), the
modeling approach may not accurately reflect the exposure-response relationship that would be
seen with a longer follow-up time. As one approach to gauge the sensitivity of the model, the
plateau parameter—representing theoretical maximum prevalence of LPT when both exposure
and TSFE are very large—was investigated further (see Table 5-15). As described above, the
plateau parameter for the primary modeling was fixed at a literature-derived value of 85%
(Jarvholm, 1992; Lilis et al., 1991). The sensitivity of the POD to this assumption was
investigated by looking at two alternative fixed values, 70 and 100% (i.e., the log-logistic
model), as well as estimating the plateau parameter from the data. The value of 70% was
selected for sensitivity analysis because in a cross-sectional study of Libby workers and residents
seen at a clinic in Libby, Winters et al. (2012) observed a prevalence of 72% for pleural
thickening, although the maximum TSFE was not known. Note that for the model with
estimated rather than fixed plateau, the modeling in the larger group of workers evaluated in
2002-2005 showed poor fit (Hosmer-Lemeshow GOF/>-value <0.10); these results are included
for comparative purposes but should be interpreted with caution.
5-61
-------
Table 5-15. Effect of different assumptions for the plateau parameter
Hosmer-Lemeshow GOF
/>-value
Model AIC
Plateau
Alpha (intercept)
Bkg (background)
Beta for TSFE
Beta for In (mean
exposure)
BMC/BMCL at 28 yr
(fibers/cc)
Hosmer-Lemeshow GOF
/>-value
Model AIC
Plateau
Alpha (intercept)
Bkg (background)
Beta for TSFE
Beta for In (mean
exposure)
BMC/BMCL at 28 yr
(fibers/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 IQ-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. Ox IQ-2
5-62
-------
The effects on the POD resulting from different assumptions regarding the plateau were
small. When assuming different fixed values (70, 85, and 100%) the POD only ranged from
2.6 x io~2 to 2.7 x io~2 fiber/cc. Estimating the plateau from the data led to a slightly higher
POD of 3.0 x 10"2 fiber/cc. These results lend confidence that assumptions regarding maximum
prevalence of LPT in the population do not have a substantial impact on the estimated POD.
As described in Section 5.2.2.6.1, one option to incorporate TSFE would be to utilize the
RTW exposure metric, which incorporates aspects of both exposure duration and TSFE. This
option was not selected for RfC derivation due to the narrow range of TSFE among the primary
analytic group of Marysville workers who underwent health evaluations in 2002-2005 and
whose job start date was on or after 1/1/1972. However, this approach was used as a sensitivity
analysis, estimating the concentration that, if experienced over 70 years, would yield the BMR.
The model with RTW as the metric derives as its POD a "benchmark residence time-weighted"
quantity in units of fibers/cc-yr2 and its associated confidence interval. In order to convert the
benchmark quantity in units of fibers/cc-yr2 and its associated lower limit into a 70-year
exposure concentration (in units of fibers/cc), the constant 70-year concentration yielding that
RTW should be determined where 70 years is both the duration and the time elapsed between the
first year of exposure and the health evaluation. That concentration is equal to the benchmark
RTW (or its lower limit) divided by the residence time-weighted value for exposures across
70 years: 1+....+70 = [(70 x 71 years)/2], as sum of first TV natural numbers is equal to
[TV x (N+ l)]/2. The results of using RTW exposure with the preferred model (Dichotomous
Hill with plateau fixed at 85%) are shown in Table 5-16. For comparison, results also using the
RTW exposure metric but using alternate model forms are also shown; the model fits and results
are very similar for the Dichotomous Hill, log-logistic, and log-probit models, with the PODs all
-0.003 fiber/cc for a scenario of 70-years exposure duration and 70-years TSFE. The
Michaelis-Menten model provided a lower AIC (2 units lower than the Dichotomous Hill
model), and the POD was slightly higher (0.0057 versus 0.0034 fiber/cc).
5-63
-------
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^-value
AIC
Intercept (SE)
Background
rate (SE)
B (SE),/;-value
Benchmark
RTW
(fibers/cc-yr2)
Benchmark
RTW lower
limit
(fibers/cc-yr2)
BMC for
TSFE = 28 yr
(fibers/cc)a
BMCL for
TSFE = 28 yr
(fibers/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 fibers/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.
Advantages of this approach using the RTW exposure metric in the subcohort are that it
relies solely on the individuals with higher quality exposure information and consistent
radiograph evaluation, and uses an exposure metric that weights more heavily exposure
occurring in the more distant past. However, the modeling still relies solely on the subgroup of
workers with little variation in TSFE, and whose TSFE values are for less than a full lifetime.
This lack of variability in TSFE limits the ability to explore how risk of LPT varies across the
life span. However, it does provide an important comparison for the primary RfC; the BMCLs
estimated using this approach range from 0.018 to 0.035 fiber/cc, similar to the BMCL estimated
in the primary analysis (0.026 fiber/cc).
Another source of information regarding TSFE comes from the study by Larson et al.
(2010a) which examined serial radiographs conducted on a group of Libby vermiculite workers
5-64
-------
with pleural or parenchymal changes. The mean follow-up time was 21.6 years, with a
maximum of 44.9 years. They found that among those workers with localized pleural
thickening, all cases were identified within 30 years, and that the median time from hire to the
first detection of localized pleural thickening was 8.6 years. Although the retrospective
evaluation of radiographs is a different and more sensitive procedure, these findings indicate that
the range of follow-up time in the Marysville subcohort is likely sufficient to support the
exposure-response modeling developed in this current assessment. Note that the likelihood that
prevalence of localized pleural thickening is expected to increase over the life span is a principal
rationale cited for the selection of a subchronic-to-chronic UF of 10 in this current assessment.
5.3.5. Uncertainty in the Endpoint Definition
The critical effect selected for RfC derivation is localized pleural thickening. As a
sensitivity analysis, an alternative critical effect of any radiographic change was also investigated
and found to yield an essentially identical POD (i.e., a BMCL of 2.7 x io~2, compared with
2.6 x 10"2 in the primary analysis), as that using the same modeling approach in the primary
analysis. Almost no information existed on radiographic changes other than LPT in the primary
analytic group of workers (evaluated in 2002-2005, hired in 1972 or later) because only one case
of DPT was reported and that individual also had LPT. No individuals had interstitial changes.
However, some individuals had DPT and/or interstitial changes (with and without LPT) in the
larger group of workers evaluated in 2002-2005, which allowed investigation of the effect of
TSFE considering alternative endpoint definitions.
The primary analysis contrasted individuals with LPT (with or without other radiographic
endpoints) to those without any radiographic changes. In the group of workers evaluated in
2002-2005, this had the effect of excluding five individuals with DPT and/or interstitial changes,
but without LPT. In the subgroup of workers hired in 1972 or later, this distinction had no effect
because the single case of DPT also had LPT (and thus was included as a case), and no
individuals showed interstitial changes. As a sensitivity analysis, the modeling procedure was
repeated, using three alternative endpoint definitions (see Table 5-17). The first contrasts all
those with LPT (with or without other endpoints) to those without LPT. Thus, the comparison
group could include those with DPT and/or interstitial changes but without LPT. The second
model used an endpoint of "any radiographic change." The third sensitivity model contrasted
those with LPT only to those with no radiographic abnormalities. Note that for two of the
alternative models (those contrasting LPT with no LPT, and any radiographic change with no
radiographic change) the modeling in the larger group of workers evaluated in 2002-2005
showed poor fits (Hosmer-Lemeshow GOF ^-values <0.10); these results are included for
comparative purposes, but should be interpreted with caution.
5-65
-------
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
at28yr
(fibers/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)
0.1075 (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
(» = 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 >1 972,
« = 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. 09187
2.6 x 10-2
(Ratio = 3.5)
Any radiographic
abnormality vs. no
radiographic abnormalities
All individuals
with
radiographs in
2002-2005
(» = 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 >1 972,
w = 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 ID"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 >1972,
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. 09317
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 >1972,
w=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. 09287
2.7 x ID'2
(Ratio = 3.4)
5-66
-------
The effect of TSFE in these sensitivity analyses using different endpoint definitions was very
similar to that in the primary model and led to nearly identical PODs (2.5 to 2.7 x 10~2 fiber/cc).
These results lend confidence to the primary analysis.
Additionally, EPA conducted a sensitivity analysis using a multinomial model to incorporate
information from all outcome groups. The multinomial logistic model is similar to the logistic
model and compares each outcome group to the referent group (i.e., no radiographic change) and
estimates separate model parameters (intercepts and beta coefficients) for each comparison.
With only two outcome groups, the logistic and multinomial models are equivalent. As
described earlier, there were noticeable differences when contrasting individuals who had LPT
alone, with those who had LPT along with DPT and/or interstitial changes. Thus, two different
multinomial models were considered. In the first, the outcome groups were (1) no radiographic
change (referent), (2) any LPT (with or without other radiographic changes), and (3) DPT and/or
interstitial changes (without LPT) (see Table 5-18). In the second model, the outcome groups
were defined as (1) no radiographic change (referent), (2) LPT alone, (3) LPT along with other
radiographic changes, and (4) DPT and/or interstitial changes (without LPT). Each of these was
contrasted to a logistic model, which is equivalent except that those with DPT and/or interstitial
changes without LPT are excluded (i.e., as was done for the primary analysis).
5-67
-------
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
Beta2 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),
;? 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),
;? 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),
;? 0.0001
0.0958 (SE = 0.0651),
^ = 0.1413
0. 1 173 (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
Pi(x, t) = 11 + exp[-czj - bt x x - Q x t]
Where/), 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).
The effect of TSFE was very similar across all the models and outcome groups, with the
corresponding beta coefficent ranging from 0.0958 to 0.1221 (compared to 0.1075 in the primary
analysis). The effect of mean exposure was much more variable, and the corresponding beta
coefficients were notably higher for those with DPT and/or interstitial changes (either alone or
along with LPT) compared to those for LPT alone. These results are in accordance with the
descriptive statistics shown in Table 5-4, which highlighted that while exposure patterns were
different among outcome groups, there was relatively less variation in TSFE. These results lend
confidence to the effect of TSFE used in the primary analysis.
5-68
-------
5.3.6. Summary of Sensitivity Analyses
EPA conducted numerous sensitivity analyses for comparison with the primary analysis
used to derive the RfC.22 These included analyses to explore the effect of exposure assessment
decisions (e.g., use of the cumulative exposure metric, truncation of exposures post-1980, and
use of arithmetic versus the geometric mean for exposure reconstruction); potential confounding
factors (time-related and nontime-related); the effect of TSFE (e.g., assumptions regarding the
plateau and use of the RTW exposure metric); and the definition of cases and noncases (e.g.,
varying case/noncase groups, use of the multinomial model). The results of other sensitivity
analyses are summarized in Table 5-19. In each case, the estimated BMCL was within an order
of magnitude of the POD. The biggest impacts came from using cumulative exposure (rather
than mean exposure), truncating exposures after 1980, and using the geometric mean versus the
arithmetic mean for exposure reconstruction (differences of-68 to +50% from the POD).
Assumptions regarding the plateau parameter (or estimating the plateau rather than fixing its
value) had a very small effect on the BMCL (differences of 0 to +5% from the POD). Similarly,
small differences in case and noncase definition led to small changes in the BMCL (differences
of-3.9 to +3.9%). Finally, the use of RTW exposure alone (rather than mean exposure and
TSFE) as the predictor in the subset of workers evaluated in 2002-2005 and hired in 1972 or
later, led to a difference of-19.2% in the BMCL.
22The 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 fiber/cc.
5-69
-------
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
(« = 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 (fibers/cc)
0.0923/2.6 x 10-2
0.0266/8.2 x 10-3*
0.2761/3.9 x 10-2
0.0298/9.1 x 10-3
0.0796/9.9 x 1Q-3
0.0920/2.7 x 1Q-2
0.0924/2.6 x 1Q-2
0.1022/3.0 x 1Q-2
0.0918/2.6 x 1Q-2
0.0929/2.7 x 1Q-2
0.0931/2.5 x 1Q-2
0.0844/2.1 x 1Q-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
"The 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.
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
5-70
-------
observed in occupational cohorts and are given in fibers/cc of air as counted by PCM (OSHA,
1994; U.S. EPA, 1988a). Thus, when examining the available health effects data on cancer for
LAA, the best available studies at this time report exposure concentration in terms of fibers/cc
counted by PCM (see Section 4.1.4). The cancer effects identified in populations with exposure
to LAA (see Section 4.1.4) are cancer mortality from mesothelioma and lung cancer (see
Section 5.4.2.2 for other cancers identified in populations exposed to asbestos in general).
Therefore, the IUR represents the upper-bound excess lifetime risk of mortality from either
mesothelioma or lung cancer in the general U.S. population from chronic inhalation exposure to
LAA at a concentration of 1 fiber/cc of air.
lURs are based on human data when appropriate epidemiologic studies are available.
The general approach to developing an IUR from human epidemiologic data is to first
quantitatively evaluate the exposure-response relationship (slope) for that agent in the studied
population. For the current assessment, the first step was to identify the most appropriate data
set available to quantitatively estimate the effects of LAA exposure on cancer mortality. Once
the relevant data describing a well-defined group of individuals along with their exposures and
health outcomes were selected (see Section 5.4.2), an appropriate statistical model form (i.e.,
Poisson or Cox) was selected that adequately fit the specific nature of the data, and then each
person's individual-level exposures were modeled using a variety of possible exposure metrics
informed by the epidemiologic literature. Exposure-response modeling was conducted for each
cancer mortality endpoint individually (see Section 5.4.3). In some cases, the statistical model
forms and the specific metrics of exposure used for each cancer endpoint may have been
different. For example, the 1988 EPA general asbestos assessment found different model
forms/metrics for mesothelioma and lung cancer. Appropriate covariates, which may be
important predictors of cancer mortality, were included in the statistical models. These models
were then evaluated to assess how the different exposure metric representing estimated
occupational exposures fit the observed epidemiologic data. The empirical model fits were
compared against those models suggested by the epidemiologic literature before selecting one
model for mesothelioma mortality and one for lung cancer mortality.
The selected cancer exposure-response relationships (slopes) for mesothelioma (KM) and
lung cancer (KL), which were estimated from the epidemiologic data on the Libby workers
cohort, were then applied to the general U.S. population in a life-table analysis using age-specific
mortality statistics to determine the exposure level that would be expected to result in a specified
level of response over a lifetime of continuous exposure. EPA typically selects a response level
of 1% extra risk because this response level is generally near the low end of the observable range
for such data. Extra risk is defined as equaling (Rx -R0) + (1 -R0\ where Rx here is the lifetime
cancer mortality risk in the exposed population and R0 is the lifetime cancer mortality risk in an
unexposed population (i.e., the background risk). In the case of lung cancer, the expected
lifetime risk of lung cancer mortality in the unexposed general U.S. population is approximately
5-71
-------
5%; thus, this human health assessment seeks to estimate the level of exposure to LAA that
would be expected to result in a 1% extra lifetime risk of lung cancer mortality equivalent to a
lifetime risk of lung cancer mortality of 5.95%: [(0.0595 - 0.05) H- (1 - 0.05) = 0.01]. This
corresponds to a relative risk (Rx/Ro) of about 1.2, which is near the low end of the observable
range for most epidemiologic studies of cancer. For mesothelioma mortality, an absolute risk
was considered, rather than extra risk, for two reasons: (1) mesothelioma is very rare in the
general population and (2) mesothelioma is almost exclusively caused by exposure to asbestos
and other mineral fibers, including LAA. Because the background rate of mesothelioma is
negligible, absolute risk models of exposure-response were considered more appropriate than
relative risk models, thereby justifying the definition of the target response rate in absolute terms
rather than in relative terms.
A life-table analysis (see Appendix G for details) was used to compute the 95% lower
bound on the level of LAA at which a lifetime exposure corresponds to a 1% extra risk of lung
cancer mortality (1% absolute risk for mesothelioma) in the general U.S. population using
age-specific mortality statistics and the exposure-response relationships for each cancer endpoint
as estimated in the Libby worker cohort. This lower bound on the level of exposure serves as the
POD for extrapolation to lower exposures and for deriving the unit risk. Details of this analysis
are presented in Section 5.4.5. Cancer-specific unit risk estimates were obtained by dividing the
extra risk (1%) by the POD. The cancer-specific unit risk estimates for mortality from either
mesothelioma or lung cancer were then statistically combined to derive the final IUR (see
Section 5.4.5.3). Uncertainties in this cancer assessment are described in detail in Section 5.4.6.
5.4.2. Choice of Study/Data—with Rationale and Justification
This human health assessment is specific to LAA. The current assessment does not seek
to evaluate quantitative exposure-response data on cancer risks from studies of asbestos that did
not originate in Libby, MT. However, this assessment does draw upon the exposure-response
models developed for other kinds of amphibole asbestos, as described in the epidemiologic
literature, to address uncertainty in model selection.
The available sources of cancer data include the cohort of workers employed at the
vermiculite mining and milling operation in and around Libby, MT. This cohort has been the
subject of several epidemiologic analyses of cancer risks, described in detail in Section 4.1.4
(Larson et al.. 201 Ob: Moolgavkar et al.. 2010: Sullivan. 2007: Amandus and Wheeler. 1987:
McDonald et al., 1986a). There have also been published reports on cases of mesothelioma in
the Libby, MT area (Whitehouse et al.. 2008) and mortality data published by the AT SDR
(2000). However, published mortality data on Libby, MT residents (Whitehouse et al., 2008:
AT SDR, 2000) could not be used in exposure-response modeling due to lack of quantitative
exposure data.
5-72
-------
The most appropriate available data set with quantitative exposure data for deriving
quantitative cancer mortality risk estimates based on LAA exposure in humans is the cohort of
workers employed at the vermiculite mining and milling operation in and around Libby, MT
(hereafter referred to as the Libby worker cohort). These data are considered the most
appropriate to inform this human health assessment for several reasons: (1) these workers were
directly exposed to LAA, (2) detailed work histories and job-specific exposure estimates are
available to reconstruct estimates of each individual's occupational exposure experience, (3) the
cohort is sufficiently large and has been followed for a sufficiently long period of time for cancer
to develop (i.e., cancer incidence) and cause mortality, and (4) the broad range of exposure
experiences in this cohort provided an information-rich data set, which allowed evaluation of
several different metrics, or mathematical expressions, of exposure. Uncertainties in these data
are discussed in Section 5.4.6.
The only other available epidemiology study cohort exposed to LAA was the cohort of
workers from a Marysville, Ohio vermiculite processing plant [see Section 4.1.1.2; (Rohs et al.,
2008; Lockey et al., 1984)]. The study of pleural changes in this population was the basis of the
noncancer exposure-response analyses (see Section 5.3). Regarding mortality among the
Marysville workers, Dunning et al. (2012) reported two mesothelioma deaths and 16 lung cancer
deaths. The Libby worker cohort was a more suitable candidate for cancer exposure-response
modeling than the Marysville worker cohort due to the larger number of cases (see Table 5-3
compared to Tables 5-20 and 5-22).
5.4.2.1. Description of the Libby Worker Cohort
Cancer mortality in the Libby worker cohort has been extensively studied (see
Section 4.1.4). McDonald et al. (2004). McDonald et al. (2002). and McDonald et al. (1986a)
published three studies on a subset of the cohort. Scientists from NIOSH conducted two
epidemiologic investigations, resulting in several published reports on different subsets of the
cohort (Sullivan, 2007; Amandus et al., 1988; Amandus and Wheeler, 1987). Berman and
Crump (2008) published analysis of data lagged 10 years (provided by Sullivan). Moolgavkar et
al. (2010) reanalyzed the Sullivan (2007) data with mortality follow-up through 2001. Larson et
al. (201 Ob) analyzed an ATSDR reconstruction of the Libby worker cohort from company
records with exposure estimates obtained from NIOSH for each job title with mortality follow-up
through 2006.
According to Sullivan (2007), nearly all of these study subjects were workers on-site at
the Libby, MT vermiculite mine, mill, or processing plant. Although the mine and other
facilities were several miles from downtown Libby, MT, some of the study subjects worked at
vermiculite ore expansion plants, at the Export Plant, or at offices in the town (see
Section 4.1.1.2). Workers may have also been assigned jobs as truck drivers or jobs working in
the screening plant, railroad loading dock, expansion plants, or an office. Individuals'
5-73
-------
demographic and work history data were abstracted from company personnel and pay records. A
database created by NIOSH in the 1980s contained demographic data and work history starting
from September 1935 and vital status at the end of 1981 for 1,881 workers. NIOSH compared
these data with company records on microfilm, and work history data were reabstracted to ensure
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 (with follow-up to
2006).
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 (fibers/cc-yr)
Median cumulative exposure (fibers/cc-yr)
Range of cumulative exposures (no lag) (fibers/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.
5-74
-------
NIOSH updated the cohort vital status through 2006 using the National Death Index
[NDI (Bilgrad, 1999)] and these data were used for this analysis. Workers known to be alive on
or after January 1, 1979 (the date NDI began tracking deaths nationwide), but not found in the
NDI search, were assumed to have been alive on December 31, 2006 (Sullivan, 2007). Nearly
54% of workers in the cohort (n = 1,009) had died by December 31, 2006. NIOSH researchers
obtained death certificates from across the United States (while exposure occurred in and around
Libby, deaths could have occurred elsewhere) for deaths prior to 1979, and the causes of death
were coded to the ICD revision that was in effect at the time of death by a single National Center
for Health Statistics-trained nosologist. After 1979, ICD codes were obtained from the
NDI-Plus. For workers known to be deceased, the underlying cause of death was determined
from death certificates and coded to the ICD codes using the rubrics of the ICD revision in effect
at the time of death [ICD-5 (WHO. 1938). ICD-6 (WHO. 1948). ICD-7 (WHO. 1957). ICD-8
(WHO. 1967). ICD-9 (WHO. 1977). or ICD-10 (WHO. 1992)1.
Basic demographic information on the occupational cohort members was largely
complete. However, when data were missing, they were statistically imputed (i.e., estimated) by
NIOSH based on several reasonable assumptions regarding gender, race, and date of birth. For
example, seven workers with unknown gender were assumed to be male because 96% of the
workforce was male, and NIOSH's review of names did not challenge that assumption (Sullivan,
2007). Workers of unknown race (n = 935) were assumed to be white because workers at this
facility were known to be primarily white, and U.S. Census Bureau data from 2004 indicate that
95% of the local population identify themselves as white (Sullivan, 2007). Date of birth was
estimated for four workers with unknown birth dates by subtracting the cohort's mean age at hire
from the worker's hire date. The impact of this imputation procedure on the analytic results can
reasonably be expected to be minimal.
5.4.2.2. Description of 'Cancer Endpoints
The cancer exposure-response assessment focuses on two cancer endpoints,
mesothelioma and lung cancer, although there is evidence that other cancer endpoints may also
be associated with exposure to asbestos in general. The IARC has concluded that sufficient
evidence in humans is present that other types of asbestos (chrysotile, crocidolite, amosite,
tremolite, actinolite, and anthophyllite) are causally associated with mesothelioma and lung
cancer, as well as cancer of the larynx and the ovary (Straif et al., 2009). Among the entire
Libby worker cohort, only two deaths were found to be due to laryngeal cancer, and no deaths
from ovarian cancer occurred among the 84 female workers. Therefore, EPA did not evaluate
these other outcomes as part of this current assessment. The limited number of female workers
in this cohort is discussed later as a source of uncertainty in the derived estimates (see
Section 5.4.6).
5-75
-------
The endpoint for both mesothelioma and lung cancer was mortality, not incidence.
Incidence data are not available for the Libby worker cohort. Nevertheless, mortality rates
approximate incidence rates for cancers such as lung cancer and mesothelioma because the
survival time between cancer incidence and cancer mortality is short. According to the National
Cancer Institute's Surveillance Epidemiology and End Results (SEER) data on cancer incidence,
mortality, and survival (Howlader et al., 2013), the median length of survival with mesothelioma
is less than 1 year, with 2-year survival for males about 20%, and 5-year survival for males about
6%. For lung cancer, the median length of survival is less than 1 year, with 2-year survival for
males about 25% and 5-year survival for males about 17%. Therefore, while the absolute rates
of cancer mortality at follow-up may underestimate the rates of cancer incidence, it is considered
to be unlikely that such discrepancies would be of significant magnitude. The use of mortality
statistics instead of incidence statistics as a source of uncertainty in the derived estimates is
further discussed in Section 5.4.6.
It is well established in the literature that mortality rates calculated from death certificates
are lower than true mortality rates due to lung cancer and, to a larger degree, mesothelioma.
These discrepancies are due mainly to misdiagnoses and imperfect sensitivity of the coding
system. Lung cancer sensitivity23 ranged from 86% in an asbestos-exposed cohort (Selikoff and
Seidman, 1992), to 95% in the general population (Percy et al., 1981): mesothelioma sensitivity
ranges from 40% for ICD-9 (Selikoff and Seidman, 1992) to about 80% for ICD-10 (Camidge et
al., 2006; Pinheiro et al., 2004). This underestimation of the true mortality rate results in a lower
estimated risk compared with that which would be estimated based on the true rate. EPA
modeled the risk of mesothelioma mortality using an absolute risk model, while the risk of lung
cancer mortality is modeled using a relative risk model. The underestimation of risk is much
more pronounced for the absolute risk model (mesothelioma) than for the relative risk model
(lung cancer). For the relative risk model, misdiagnosis rates would need to be different with
respect to exposure levels, and this is unlikely among the Libby workers that were included in
the lung cancer analysis because nosologists are blinded to exposure levels when coding lung
cancer as a cause of death. Therefore, EPA considered use of a procedure to adjust risks for
mesothelioma underascertainment (see Section 5.4.5.1.1) —but not for lung cancer.
Mesothelioma did not have a distinct ICD code prior to introduction of the 10th revision
(ICD-10), which although released in 1992, was not implemented in United States until 1999.
Death certificates from 1940 to 1978 were reviewed by the NIOSH principal investigator
(Sullivan, 2007) to identify any mention of mesothelioma on the death certificate, as is the
standard procedure for assessing mesothelioma mortality and has been used in other analyses of
Libby worker cohort mesothelioma mortality (Larson et al., 201 Ob: McDonald et al., 2004). For
deaths in the Libby worker cohort occurring from 1979 to 1998, death certificates were obtained
3Sensitivity is measured by the percentage of actual lung cancer deaths that are detected.
5-76
-------
if the NDI identified the cause of death as one of the possible mesothelioma codes identified by
Marsh et al. (2001), as respiratory cancer, nonmalignant respiratory disease, digestive cancer, or
unspecified cancer. For deaths in the Libby worker cohort that occurred after 1998, the ICD-10
code for mesothelioma was used. In total, 18 mesothelioma deaths were identified by NIOSH
using the methods of Sullivan (2007), which serve as the basis for this current assessment;
19 mesothelioma deaths were identified by Larson et al. (201 Ob) for the same cohort from all
death certificates rather than from death certificates with one of the specifically targeted set of
causes of death identified above in Sullivan (2007).
Whitehouse et al. (2008) identified four mesothelioma cases among workers that, as the
authors suggested, were not included in the Sullivan (2007) study with mortality follow-up
through 2001; no other information was provided. Three mesothelioma cases from these four
were most likely accounted for during the update of the NIOSH cohort from 2001 to 2006, which
serves as the basis for this current assessment. Whitehouse et al. (2008) also provided detailed
information on 11 residential cases, but this information could not be used in exposure-response
analyses for this current assessment because there is no quantitative exposure information for
these cases and no information defining or enumerating the population from which these cases
arose.
Lung cancer mortality was based on the underlying cause of death identified by the ICD
code on death certificates according to the ICD version in use at the time of death. Based on
these different ICD codes, lung cancer mortality included malignant neoplasms of the trachea,
bronchus, and lung, and was identified by the following codes: ICD-5 code "047" (excluding
"47c, Cancer of unspecified respiratory organs"), ICD-6 codes "162" or "163," ICD-7 codes
"162.0" or "163" (excluding "162.2, Cancer of the pleura"), ICD-8 and ICD-9 code "162," and
ICD-10 codes "C33" or "C34." In all, there were 111 deaths with an underlying cause attributed
to lung cancer. All deaths after 1960 were coded as bronchus or lung because the ICD versions
in use at that time distinguished malignant neoplasms of the trachea as distinct from neoplasms
of the bronchus and lung. Other investigators of this cohort have used slightly different
definitions of lung cancer or used different follow-up periods, as described in Section 4.1.1.1
(Studies of Libby, MT Vermiculite Mining and Milling Operations Workers).
5.4.2.3. Description of Libby Worker Cohort Work Histories
NIOSH staff abstracted demographic data and work history data from company personnel
and payroll records. An individual's work history was determined from job change slips, which
recorded any new job assignment, date of change, and change in hourly pay rate (which differed
by the job assignment). Work history records span the time period from September 1935 to May
1982. Dates of termination were unknown for 58 of 640 workers (9%) who left employment
before September 1953. EPA adopted the assumption used by NIOSH (Sullivan, 2007) that
these people worked for 384 days, based on the mean duration of employment among all workers
5-77
-------
with known termination dates before September 1953. The majority of workers in this cohort as
a whole and those hired on or after January 1, 1960 worked at multiple jobs; many of the
workers switched jobs repeatedly, and the changes in exposures associated with changes in job is
accounted for through the use of the job- and time-specific JEM described in the following
section.
5.4.2.4. Description ofLibby Amphibole Asbestos Exposures
The operations at the mine and in and around Libby, the conditions of exposure, and the
job-specific estimates of exposure intensity have been thoroughly described in Section 4.1
(Sullivan. 2007: Amandus et al.. 1987b: McDonald et al.. 1986a). Briefly, miners extracted
vermiculite ore from an open-pit mine that operated on Zonolite Mountain outside the town of
Libby, MT. The ore was processed locally in a dry mill (1935-1974) and/or two wet mills
(1950-1974 and 1974-1990). The resulting concentrate was transported by railroad to
processing plants around the United States where the vermiculite was expanded for use in
loose-fill attic insulation, gardening, and other products (see Section 2.1). EPA adopted the JEM
developed and used by Sullivan (2007), which was in turn based on that used in the earlier
NIOSH study for jobs through 1982 (Amandus et al.. 1987b: Amandus and Wheeler. 1987). As
discussed in more detail in Section 4.1, Amandus et al. (1987b) defined 25 location operations in
the Libby facilities for which they estimated exposure intensity based on available information
(see Table 5-21). A job category may have involved more than one location operation, and the
8-hour time-weighted average (TWA) exposure (8-hour TWA) for each job category in the JEM
was calculated from the exposure intensity and time spent at each location operation (Amandus
etal.. 1987b).
5-78
-------
Table 5-21. Exposure intensity (fibers/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
-
5-79
-------
Table 5-21. Exposure intensity (fibers/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
5-80
-------
For the later data in Table 5-21 from 1967 through 1982, over 4,000 air samples analyzed
for fibers by PCM analysis were available to inform the exposure intensity estimates for the
25 location operations. Therefore, the JEM for 1967-1982 is based on direct analytic
measurements in air for each location operation (Amandus et al., 1987b). With the exception of
two location operations in the dry mill, no air samples were available for other location
operations at the mine and processing facilities before 1967. In order to estimate exposures that
occurred before that time, the NIOSH researchers interviewed plant employees and based
estimates of exposure intensities on known changes in operations over the years and professional
judgments regarding the relative intensity of exposure. Exposure intensity for 23 of the 25
pre-1967 location operations was extrapolated from post-1967 measurements based on reasoned
assumptions for each location operation (Amandus et al., 1987b).
In contrast to the exposure information available for 1967 through 1982, the amount and
quality of measurement data in the facility in earlier years were much more limited (Amandus et
al., 1987b). A total of 40 dust samples were taken, exclusively in the dry mill, over the years
1950-1964. Using these measurements, higher exposures were inferred to occur before 1964
than in later years.
Air samples collected by the State of Montana were available for the dry mill
from 1956-1969, but at that time these were analyzed for total dust, not asbestos fibers. Total
dust samples (collected by a midget impinger) were examined by light microscopy, but no
distinction was made among mineral dusts, debris, and asbestos fibers. All objects were counted
and reported in the units of million particles per cubic foot (mppcf). Amandus et al. Q987b)
developed a range of conversion ratios between total dust and asbestos fiber counts based on the
comparison of contemporaneous air sampling in the dry mill (see Section 4.1.1.2) and selected a
conversion ratio of 4.0 fibers/cc per mppcf to estimate exposure intensity for two location
operations in the dry mill for the years prior to 1967. Uncertainties in the selection of this
conversion ratio are described in detail in Section 5.4.6.1.2.1.
The exposure intensity (fibers/cc) for each of the location operations (see Table 5-21)
was used to calculate an estimate of daily occupational exposure for each job category in the
JEM. In developing a job exposure estimate for a given 8-hour workday, the estimated exposure
intensity at each location/operation was multiplied by the fraction of the day that someone in that
job spent at that location/operation, at that point in time. If a job involved working in more than
one location or operation, these estimates were summed over an 8-hour workday. The resulting
JEM available for this current assessment and previous epidemiologic studies of the Libby
worker cohort is based on the air concentration of fibers as enumerated by PCM, which measures
fibers longer than 5 um with an aspect ratio >3:1 [i.e., the fiber size regulated under the
Occupational Safety & Health Administration [OSHA] standard (OSHA. 2006)1. Additionally,
only fibers that are wide enough to be viewed on PCM can be detected with this method.
Amandus et al. (1987b) considered fibers >0.44 um in diameter to be visible by PCM in the
5-81
-------
historical filter analysis. More recent techniques have refined the PCM method, and fibers
greater than 0.25 um in diameter are now considered PCM fibers (IPCS, 1986). Uncertainties
related to difference in defining PCM fibers are discussed in Section 5.4.6.1.2.1.
Amandus et al. (1987b) recognized the uncertainty in the pre-1968 exposures estimates
for the cohort. Although there is some uncertainty in the dust-to-fiber conversion, this
conversion (4.0 fibers/cc per mppcf) was based on dust and fiber data contemporaneously
collected in the dry mill and only applied to the dry mill environment. Amandus et al. (1987b)
considered a range of possible conversion factors (1.2-11.5 fibers/cc per mppcf). Greater
uncertainty may lie with the reasoned assumptions used to extrapolate exposures to the early
decades for all location operations considered. For example, there were four location operations
for which Amandus et al. (1987b) estimated a range of possible exposure intensities—drilling,
ore loading, the river dock, and the bagging plant, where intensity of exposure may vary as much
as threefold between the low and high estimates (see Table 5-21). Finally, some workers were
employed after 1982 and up until 1993, when demolition of the facilities was completed (Larson
et al., 2010b). These exposures were not evaluated by Sullivan (2007) and were not included in
the NIOSH JEM. Because exposure concentrations in 1982 (see Table 5-21) were generally at or
below 1.2 fibers/cc, it is unlikely that the overall cumulative exposures of this limited set of
workers were significantly underestimated by not including exposures during this time.
Uncertainties in all aspects of the JEM and the associated exposure assessment are described in
Section 5.4.6.1.2.
There was one important limitation of the NIOSH work history data in assigning
exposure levels for each job. In the earlier study (Amandus and Wheeler, 1987), workers with
"common laborer" job assignments and some workers with unknown job assignments hired
between 1935 and 1959 all received the same, relatively low exposure levels estimated for the
mill yard (Sullivan, 2007). In addition, reabstracting work histories for the more recent study
(Sullivan, 2007) identified several job assignments not mentioned in the earlier publications.
Sullivan (2007) estimated exposure for the additional job and calendar time period-specific
combinations based on professional experience and review of exposure records from earlier
studies of the Libby worker cohort (Amandus et al., 1987b: Amandus and Wheeler, 1987;
McDonald et al.. 1986a). EPA found that of the 991 workers hired before 1960, 811 workers
(82%) had at least one job with an unknown exposure assignment, with 706 (71%) listing neither
job department nor job assignment. In the more recent study by Sullivan (2007), these workers'
exposures were all estimated using the same TWA exposure intensity estimated for all jobs
during that time period (66.5 fibers/cc). The lack of information on specific exposure
information for such a large portion of these early workers, during the time period when
exposures were higher, resulted in significant exposure misclassification and effectively yielded
exposure estimates that were differentiated only by the duration of each worker's employment.
Because of the lack of more specific measured fiber exposure data during this early period, EPA
5-82
-------
experienced difficulties in identifying an adequate exposure-response model fit for the complete
cohort including all hires. These difficulties are described in detail in Section 5.4.3.4. As a
result, the IUR analyses were based on the subset of workers hired after 1959 (i.e., on or after
January 1, 1960), totaling 880 workers (i.e., full cohort [n = 1,871] minus those hired before
1960 [n = 991]). Of these 880 workers hired after 1959, 28 (3%) workers had at least one job
with an unknown job assignment with nine having all job and department assignments between
1960-1963 listed as unknown. As described in Sullivan (2007). NIOSH estimated that these
workers had a TWA exposure intensity of 66.5 fibers/cc. Uncertainties in the exposure
assessment for this subcohort are described in Section 5.4.6.1.2.4. While the Sullivan (2007)
study was limited to the white male workers, EPA's analysis includes all workers regardless of
race or gender. Table 5-22 shows the 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 (fibers/cc-yr)
Median cumulative exposure (fibers/cc-yr)
Range of cumulative exposures (no lag) (fibers/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.
5-83
-------
Figure 5-5 shows a three-dimensional representation of the JEM used by Sullivan (2007)
and in this cancer exposure-response assessment (note that the figure does not include all jobs
and is meant to be illustrative rather than comprehensive). The three axes show the intensity of
fiber exposure as an 8-hour TWA (fibers/cc, vertical axis) for selected job categories over time
(horizontal axes). For several jobs, the estimated 8-hour TWA was greater than 100 fibers/cc for
the decades prior to 1963.
Exposure
Intensity
(fibers/cc)
Year
S'electedJobs
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
(fibers/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.
5-84
-------
5.4.2.5. Estimated Exposures Based on Job-Exposure Matrix (JEM) and Work Histories
Exposure-response modeling of epidemiologic data is based on several considerations as
summarized by Finkelstein (1985):
After identification of an occupational hazard one of the goals of occupational
epidemiology is to quantify the risks by determining the dose-response relations
for the toxic agent. In many circumstances little is known about the dose received
by target tissues; the data available usually pertain only to exposure to various
concentrations of the toxic material in the workplace. The calculation of dose
requires additional physiological and chemical information relating to absorption,
distribution, biochemical reactions, retention, and clearance.
In asbestos epidemiology the usual measure of exposure is the product of the
concentration of asbestos dust in the air (fibers or particles per mL) and the
duration of exposure to each concentration summed over the entire duration of
exposure (years).
Cumulative exposure (CE) has been the traditional method of measuring exposure in
epidemiologic analyses of many different occupational and environmental exposures and was the
exposure metric applied to the risk of lung cancer mortality in the IRIS assessment for general
asbestos (U.S. EPA, 1988a). That said, different health outcomes may be best described using
different exposure metrics. The risk of mesothelioma mortality in the IRIS assessment for
general asbestos (U.S. EPA, 1988a) used a different exposure metric based on a linear function
of concentration added to a function of TSFE and duration of exposure. Additional exposure
metrics were also assessed for both mesothelioma and lung cancer mortality risks.
Two alternative approaches to developing exposure metrics to describe the effects of
concentrations of asbestos dust in the air on the risks of mortality have also been proposed. The
first alternative was proposed by Jahr (1974), who studied silica-induced pneumoconiosis and
suggested that exposures to occupational dusts could be weighted by the time since exposure.
This yields an exposure metric that gives greater weight to earlier exposures. The second
alternative was proposed by Berry et al. (1979) who subsequently suggested the application of
exposure metrics that allowed for the clearance of dust or fibers by using a decay term on
exposures.
For the evaluation of mortality risk from mesothelioma for general asbestos, U.S. EPA
(1988a) used a different exposure metric than was used for lung cancer mortality, which factored
in the TSFE. As observed in U.S. EPA(1988a), it is important to note that different
characterizations of estimated occupational exposures may be reasonably expected to be
associated with different endpoints.
Many studies have been limited in the availability of detailed exposure data—especially
at the individual level. In the Libby worker cohort, detailed work histories were matched with
job-specific exposure estimates, allowing for the reconstruction of each individual's estimated
5-85
-------
occupational exposure over time. The individual-level data were developed based on a uniform
estimate of exposure in a given job during a given point in time multiplied by individual-level
data on duration. From this information-rich, individual-level data set from NIOSH, EPA
constructed a suite of the different metrics of occupational exposure which had been proposed in
the asbestos literature or used in the IRIS asbestos assessment (U.S. EPA, 1988a). This suite of
metrics was defined a priori to encompass a reasonable set of proposed exposure metrics to allow
sufficient flexibility in model fit to these data. The types of exposure metrics evaluated were
intended to allow for more or less weight to be placed on earlier or later exposures. These
simulated exposure metrics were derived mathematically to approximate underlying processes
that are not well understood (see Section 5.4.6). Thus, the empirical fit of various exposure
metrics to the observed epidemiologic data is evaluated statistically, and the exposure metrics
have epidemiological interpretation but do not necessarily have direct biological interpretations.
The first exposure metric—CE—is a simple addition of each day's exposure across time
(see eq 5-7). CE has been widely used in modeling risk of cancer in occupational epidemiology
and has been used for modeling lung cancer (Larson et al., 201 Ob: Moolgavkar et al., 2010;
Berman and Crump, 2008; Sullivan, 2007; McDonald et al., 2004) and mesothelioma (McDonald
et al., 2004) in the Libby worker cohort. When using this exposure metric in the risk model, all
exposures (other than for years removed from consideration based on a lag assumption) have
equal weight regardless of when they occurred and lead to the same estimated cancer risk
whether exposure happened early or later in life.
EPA calculated each individual's occupational CE to LAA over time from their date of
hire until the date they ceased to be employed in the Libby operations or until the date NIOSH
collected the work history data for those still employed in May 1982. Workers were assumed to
remain at their CE on the last day of work until death or the end of the follow-up period on
December 31, 2006. Each worker's CE at any time point (daily increment) since their date of
hire was computed as the sum of their exposure intensity (fibers/cc) on each specific
occupational day (xt) from Day 1 through Day k. Mathematically, this was defined as
k
CE at time tk = ^Xt (5-7)
7=1
Where
xt. = the estimated job-specific exposure intensity for the day tj, and
tk = the day on which the exposure is estimated.
A second exposure metric—RTW exposure—gives additional weight to early exposures.
By doing so, the RTW exposure metric allows the possibility that early exposures are more
influential on cancer mortality predictions in the model. Unlike many chemicals that are rapidly
5-86
-------
metabolized in the body and excreted, asbestos fibers are durable, and some remain in the body
for years. Fibers that remain in the lung may continue to damage lung cells and tissue unless
they are removed or cleared (see Section 3.2). Similarly, fibers that translocate to the pleura may
damage cells as long as they remain in this tissue. Therefore, a fiber exposure may not only
damage tissue during the initial exposure, but fibers may remain in these tissues, with tissue fiber
concentration as well as cellular and tissue damage accumulating over time. While this
represents a biological point of view, in an epidemiologic context in which the exposure is
ambient fiber concentration and the event of interest is simply a cause of death (rather than
survival time), it is uncertain what metrics of exposure might fit the observed data and could be
considered most appropriate, so EPA considered several.
The RTW exposure metric in this current assessment is sometimes called the cumulative
burden, or the area under the curve. A type of RTW metric was proposed for modeling of
mesothelioma mortality by Newhouse and Berry (1976) based on a general understanding of the
relationship between tumor incidence rate and time to cancer (Cook et al., 1969) as well as
animal models of mesothelioma (Berry and Wagner, 1969).
A similar type of RTW metric was proposed in Peto (1978) and was subsequently applied
by Peto etal. (1982), discussed by Finkelstein (1985), and applied in the derivation of the IUR in
the 1988 IRIS assessment for asbestos (U.S. EPA, 1988a). McDonald et al. (2004) and
Moolgavkar et al. (2010) used RTW-type metrics for modeling mesothelioma in the Libby
worker cohort, and McDonald et al. (2004) applied an RTW metric for modeling lung cancer
mortality in the Libby worker cohort.
In calculating RTW, each day's exposure is multiplied by the time since exposure
occurred up to the time tk when RTW is estimated (see eq 5-8). The intent of RTW CE is to
allow for earlier exposures to contribute greater weight.
RTW CE at time tk = ^ xt, x ('* ~ l}) (5'8)
Where
xt. = the estimated job-specific exposure intensity for the day tj, and
tk = the day k on which the exposure is estimated.
The CE and RTW exposure metrics result in sustained or increasing metrics of exposure
across time. However, some cellular and genetic damage can be repaired over time after
exposure, decreasing the cancer risk from exposure over time. Additionally, asbestos fibers are
cleared (removed) from the lung through natural processes and translocated to other tissues (see
5-87
-------
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= 2_, \ -V x exP
7=1
Tv
(5-9)
Where
xt. = the estimated job-specific exposure intensity for the day tj,
tk = the day k on which the exposure is estimated, and
Ti/2= half-life of 5, 10, 15 or 20 years.
x, x(/t-/.)xexp
RTW with half-life at time tk = ;=i
(5-10)
-------
In addition to the considerations described above for selecting metrics to represent
estimated ambient exposure to LAA for use in predicting the risk of mortality, there is the
important issue of potentially modifying the exposure metrics to account for cancer latency.
Without knowledge of the specific timing of etiologically relevant exposure that may initiate and
promote cancers that ultimately result in mortality, any exposure metric may include exposures
during some time period that do not have bearing on the risk of mortality. In the absence of such
information on the specific cancer latency associated with a specific exposure, Rothman (1981)
suggested that the most relevant exposure period could be identified by comparing the fit of
exposure metrics across multiple lag periods to allow for the identification of the optimal latency
period. This has since become a standard practice in occupational and environmental
epidemiology. Accordingly, exposure estimates for all exposure metrics were adjusted to
account for the time period between the onset of cancer and mortality. The lag period defines an
interval before death, or end of follow-up, during which any exposure is excluded from the
calculation of the exposure metric. Cohort members who died or were lost within the initial
years of follow-up were assigned lagged exposure values of zero if they had not been followed
for longer than the lag time. The various exposure metrics were lagged at 10, 15, and 20 years to
account for different potential cancer latencies within the limitations of the available data.
Metrics without a lag were fit for comparison purposes and to aid in identifying pattern with
increasing lag time but were not considered to be biologically reasonable, given that the outcome
under analysis is cancer mortality (specifically, mesothelioma and lung cancer), for which
latency periods of 10 years or more have been suggested for asbestos (U.S. EPA, 1988a).
Consequently, metrics that were not adjusted by lagging exposure in the final years before
mortality (or the end of follow-up) were not considered further in the development of an IUR for
LAA.
In addition to the exposure metrics used in the lung cancer mortality analysis, modeling
of mesothelioma mortality (see Section 5.4.3.1) included additional exposure models. The Peto
model (Peto etal., 1982; Peto, 1979) uses a cubic power function of TSFE and a linear function
of exposure concentration. The model developed by Peto was later adapted in the IRIS (U.S.
EPA, 1988a) asbestos assessment. The linear function of concentration was developed based on
estimated average workplace concentrations over several asbestos cohorts exposed to chrysotile,
amphiboles, and mixed fibers. The two amphibole-exposed cohorts were U.S workers exposed
to amosite and Australian workers exposed to crocidolite; for both cohorts there was little
exposure information at the time these reports were published. As Health Effects
Institute-Asbestos Research (HEI, 1991) noted "No extensive measurements of historical
exposure levels are available for the cohorts exposed predominantly to crocidolite or amosite."
The only other mesothelioma models proposed in the literature for amphibole asbestos were
developed by Berry (1991) for crocidolite. Berry et al. (2012) found that models that allow for
clearance (mathematically, multiplicative exp[-A x 7]) better match the actual mortality
5-89
-------
experience of the Wittenoom, Australia crocidolite cohort with more than 50 years follow-up,
compared with Peto-type models. In particular, they found that models with a higher power of
TSFE of 5.4 (compared to power = 3 for Peto) and a decay rate of 15%/year (half-life of
approximately 5 years) fits the observed data best, followed by models with a power of TSFE of
3.9 and a decay rate of 6.8%/year (half-life of approximately 10 years). However, the exposure
data calculated for the Berry analysis was based on just one study of airborne levels.
Nevertheless, cumulative exposures calculated from these have been shown to be internally valid
based on association with fiber lung burden (Berry et al., 2012).
Peto's model [also used in the 1988 IRIS assessment for asbestos (U.S. EPA, 1988a)1 is
(5-11)
where
Im = the observed deaths from mesothelioma/person-years (i.e., the mesothelioma
mortality rate),
C = the average concentration of asbestos in the air,
KM = an estimated slope describing the relationship between LAA exposure and
mesothelioma mortality, and
Qk = the function of the TSFE (i) and the duration of exposure (d):
?ortd+W,Qk = (t-lO)k-(t-lQ-d)k.
Alternatively, Im = C x Qk x KM x exp(-/( x f) defines the Peto model with clearance.
Possible values of X and k suggested in the literature (Berry etal., 2012) are X = 0.068 or 0. 15 and
k= 3.9 or 5.4. As the Peto model and the Peto model with clearance were both proposed in the
amphibole asbestos literature, these models were carried forward in the analysis below.
5.4.3. Exposure-Response Modeling
As discussed above, consideration of biology and previous epidemiologic studies
informed the range of models considered. There is not sufficient information to select models
for the epidemiology data on the basis of the biological mechanism of action for lung cancer or
mesothelioma (see Section 3). In this situation, EPA's practice is to investigate several modeling
options to determine how to best empirically model the exposure-response relationship in the
range of the observed data as well as consider exposure-response models suggested in the
5-90
-------
epidemiologic literature. For LAA, possible exposure metrics were explored for model fit to the
chosen model forms. The exposure metric options were selected to provide a range of shapes
that was sufficiently flexible to allow for a variety of ways that time and duration might relate to
cancer risk in the data being modeled.
The following sections provide information about modeling of the full cohort first, the
difficulties in identifying adequately fitting models to these data, and the decision to base the
analysis on a subcohort of workers that allowed for identifying adequately fitting models.
5.4.3.1. Modeling of Mesothelioma Exposure Response in the Libby Worker Cohort
In a population, not, or only sparingly, exposed to asbestos, the background incidence of
mesothelioma is extremely low, about 1 in a million (Hillerdal, 1983). The evaluated
exposure-response models examine the relationship of the absolute risk of mesothelioma
mortality attributable to LAA exposure, because it is not clear that a background risk of
mesothelioma mortality exists among people who were truly unexposed to asbestos (as opposed
to the relative risk model, which is used for lung cancer mortality; see Section 5.4.3.3). EPA
does not have a specific technical guidance for model selection based on human cancer data, but
as a general consideration, EPA's BMD Technical Guidance (U.S. EPA, 2012) states that "The
initial selection of a group of models to fit to the data is governed by the nature of the
measurement that represents the endpoint of interest and the experimental design used to
generate the data." Here, the most prominent feature of the data is the rarity of mesothelioma
deaths. Correspondingly, Poisson models are employed to estimate the absolute risk of
mesothelioma, as the Poisson distribution is an appropriate model for use with data that are
counts of a relatively rare outcome, such as observed mesothelioma deaths in the Libby worker
cohort. Other parametric survival models, such as the Weibull model have been used for
absolute risk calculation, but they are not generally used for data with rare outcomes.
Consequently, there are no examples in the literature of the Weibull or other parametric survival
model ever being used for modeling mesothelioma mortality. Previous analyses of
mesothelioma mortality in the Libby worker cohort also used the Poisson model (Moolgavkar et
al., 2010; McDonald et al., 2004). Mathematically, the Poisson distribution specifies the
probability of k events occurring as
k\ (5-12)
where X is parameterized with the exposure metric (defined in Section 5.4.2.5). Then, life-table
analysis is used to estimate risks in the general U.S. population for the derivation of the unit risk
of mesothelioma mortality (see Section 5.4.5.1). In the standard Poisson distribution, the
assumption is that the mean is equal to the variance. However, actual count data often exhibit
5-91
-------
overdispersion, a statistical consideration when the variance is larger than the mean; thus, EPA
evaluated potential for overdispersion.
Estimation of the exposure-response relationship for mesothelioma mortality was
performed using a Monte Carlo Markov Chain (MCMC) Bayesian approach with an
uninformative or diffuse (almost flat) prior [WinBUGS Version 1.4; (Spiegelhalter et al., 2003)1.
Use of diffuse priors is a standard procedure in Bayesian analysis, in situations like this one,
when there is no prior knowledge about the toxicity of LAA under a particular model. Because
this analysis focuses only on the Libby worker cohort and does not try to factor in data from
other sources in estimating potency, use of a diffuse prior is considered appropriate for this
analysis.
The benefit of using the WinBUGS software implementing Bayesian MCMC approach is
computational ease and that it provides a posterior distribution of the mesothelioma coefficient
(KM) rather than just a point estimate. A diffuse (high variance) Gaussian distribution, truncated
to exclude negative parameter values, is used as a diffuse prior. With such a prior, results of
MCMC analysis are expected to be similar to maximum likelihood estimation in a non-Bayesian
analysis. Standard practices of MCMC (Spiegelhalter et al., 2003) analysis were followed for
verifying convergence and sensitivity to the choice of initial values. The posterior distribution is
based on three chains with a burn-in of 10,000 (i.e., the first 10,000 simulations are dropped so
that remaining samples are drawn from a distribution close enough to the true stationary
distribution to be usable for estimation and inference) and thinning rate of 10 (i.e., only each 10th
simulation is used—thus reducing autocorrelation), such that 3,000 total simulations constitute
the posterior distribution of KM. The mean of the posterior distribution served as a central
estimate, and the 90% credible interval24 defined the 5th percentile and the 95th percentile of the
distribution, which served as bounds for the 95th lower and upper one-sided confidence intervals,
respectively.
The fit of multiple metrics of exposure, the Peto model and the Peto model with clearance
(see Section 5.4.2.5), as well as exposure intensity, duration of exposure, age at death or loss to
follow-up, and TSFE were compared using the Deviance Information Criterion (DIG). The DIG
(Spiegelhalter et al., 2002) is used in Bayesian analysis and is an analogue of the AIC, with
smaller values indicating a better statistical fit to the data. Use of the DIG and AIC is standard
practice in comparing the fit of nonnested models to the same data set with the same dependent
outcome variable but different independent covariates. According to Burnham and Anderson
(2002), "These methods allow the data-based selection of a 'best' fitting model and a ranking
and weighting of the remaining models in a predefined set." Because of the small number of
deaths from mesothelioma in absolute terms, only uni- and bivariate models (with age or TSFE
as the second covariate) were considered. Gender and race were not used as covariates because
4A credible interval is the Bayesian analogue of a confidence interval.
5-92
-------
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. Lags of more than
20 years were not considered as the first mesothelioma death in the workers happened between
15 and 20 years after starting employment.
5.4.3.2. Results of the Analysis of Mesothelioma Mortality in the Full Libby 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 as it
does not suffer from the lack of exposure information for many of the workers in earlier years
(see Section 5.4.3.4). 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 over dispersion 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
5-93
-------
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
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.
5-94
-------
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
aBecause 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.
It is likely that the poorer fit seen when using information on exposure concentration is
the result of the fact that duration of exposure is measured with comparatively little error, while
derivation of specific exposure concentrations may be subject to a sizable measurement error.
Moreover, as described in Section 5.4.2.3, for 706 of 991 (71%) workers hired from 1935 to
1959, only the duration of exposure was known, and the same exposure concentration was
estimated for them. Thus, the same average estimated exposure intensity for that time period had
been used for these workers (Sullivan, 2007). Particularly large exposure measurement error,
among more than two-thirds of the workers hired prior to 1960 who had the same estimated
exposure intensity, resulted in the duration of exposure being the best predictor of mesothelioma
mortality. Additionally, estimates of exposure intensity prior to 1968 have greater uncertainty
associated with them than more recent exposure measurements, which are based on fiber counts
in air samples analyzed by PCM. For the majority of job locations (23 of 25), no exposure
measurements were available before 1968, and exposures were estimated based on employee
interviews (in 1982) and what was known about major changes in operations between 1935 and
1967. For two exposure locations, the dust-to-fiber conversion ratio is based on measurements
taken in the late 1960s, so extrapolations from the mid 1960s to the early 1960s is likely to be
more certain than extrapolation further back in time. The metric using only duration of exposure
fit best and the additional incorporation of exposure intensity information, as expressed as the
CE, only worsened the fit. Therefore, it is unlikely that IUR estimates can be developed using
5-95
-------
the full cohort data because the early exposure values (which were predominantly inferred from
later data and based on missing exposure information) were not predictive of mesothelioma
mortality.
5.4.3.3. Modeling and Results of Lung Cancer Exposure Response in the Full Libby Worker
Cohort
As noted in the previous section, while the final analytic exposure-response modeling is
based in the subcohort of workers hires after 1959, it is important to understand how different
metrics of exposure that appear in the epidemiologic literature are related to risk in the full
cohort. A parallel set of tables is provided for the subcohort of workers hired after 1959 in
Section 5.4.3.6.
Tables 5-27 to 5-29 show the mortality rates of lung cancer mortality by duration of
exposure, age of first exposure, and TSFE for the full Libby worker cohort (n = 1,871). Note
that people in Montana may be different from those in the whole United States and therefore may
be a more appropriate comparison but the reference population is smaller and less stable. The
United States reference rates/risks are more stable. Looking at both SMRs allows the evaluation
of the rates/risks in Libby workers with greater context. Lung cancer rates in the Libby worker
cohort are substantially higher than mesothelioma rates (see Section 5.4.3.2). Basic stratified
models of lung cancer rates and standardized mortality ratios (SMRs) in this population show
increased rate with increases in duration of exposure greater than 5 years, age at first exposure
and TSFE.
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 * 10~4
White male SMRMontam
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).
5-96
-------
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 10-4
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 10-4
White male deaths/white male PY
White male rate x 10~4
White male SMRMontam
White male SMRu.s.
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).
EPA does not currently have specific technical guidance for model selection based on
human cancer data. However, the process and criteria used in the assessment are explained
below. Standard models from similar exposure-response analyses available in the epidemiologic
literature may be candidate models for exposure-response analyses when they are appropriate to
the epidemiologic data at hand.
As noted above for mesothelioma, models are selected and evaluated based on the nature
of the data set, which for lung cancer, warrants the ability to use the time-dependent data. The
mesothelioma mortality data were modeled using the Poisson model within a Bayesian
framework to estimate the absolute risk, because mesothelioma is very rare in the general
population not, or sparingly, exposed to asbestos (Hillerdal, 1983). While the Poisson model is
appropriate for modeling very rare events, the standard form does not allow for inclusion of the
time-varying nature of exposure. Lung cancer is more common than mesothelioma which does
not have a well-defined background risk in the absence of asbestos exposure. Thus, modeling of
lung cancer mortality is based on the relative risk rather than the absolute risk and was conducted
in a frequentist framework, which is the standard methodology for epidemiologic analyses. A
5-97
-------
frequentist framework is an alternative method of inference, drawing conclusions from sample
data with the emphasis on the observed frequencies of the data.
Standard epidemiologic models for relative risk include Poisson, logistic, conditional
logistic, and Cox models. Multistage clonal expansions models are also available. However,
only the Cox models and clonal expansion models can accommodate the analysis of
time-varying covariates as in the case of the Libby worker cohort. While different researchers
have used two-stage clonal expansion models to model asbestos-related health endpoints from an
occupational cohort of asbestos textile workers in South Carolina (Zeka et al., 2011; Richardson,
2009), divergent model results raise questions about the resilience of this method when applied
to epidemiologic cohorts. Specifically, the two-stage clonal expansion analysis by Richardson
(2009) fit the data well and was complementary and consistent with his accompanying Cox
regression analysis, while the two-stage clonal expansion analysis by Zekaetal. (2011) on the
same cohort population, but with a different length of mortality follow-up, did not completely
converge, indicating poor model fit. One issue is that epidemiologic cohorts may be less regular
in nature than toxicological studies in the sense that epidemiologic cohorts can be dynamic, with
people joining at different times, possibly leaving and then rejoining. By comparison, in animal
studies, it is more typical for all the subjects to undergo the identical exposure protocol.
Additionally, the degree to which the results of two-stage clonal expansion models depends upon
multiple additional assumptions is not yet well understood and EPA does not have reliable
information available on which to make the required assumptions for the Libby worker cohort
(e.g., the number of cells at risk, constraints on the spontaneous rates of first and second
mutations, allowing for a fixed lag between malignant transformation of a cell and death from
cancer, etc.). Therefore, in addition to the basic stratified models of lung cancer risk by duration
of exposure, by age at first exposure, and by TSFE, EPA selected the Cox model as the most
appropriate model for exposure-response modeling based on the suitability of this model to the
nature of the data set (i.e., time-dependent exposure information), the long history of usage in
analyses of occupational cohorts, and the commonality of usage in other epidemiologic analyses
of the Libby workers cohort.
No other standard epidemiological model formulations allow for the analysis of
time-varying exposures in the manner achieved by the Cox proportional hazards model. The
exposure-response relationship (proportional hazards ratio) determined in this model intrinsically
takes into account the effects of other causes of mortality that are unrelated to exposure (i.e.,
independent censoring). Further, all comparisons are made within the cohort by comparing the
mortality experience of people with different exposures within the same cohort population.
Nonetheless, the issue of competing risks that are dependent on exposure (e.g., asbestosis or
nonmalignant respiratory disease) is an acknowledged uncertainty for this and other types of
analyses (see Section 5.4.6).
5-98
-------
The Cox proportional hazards model (Cox, 1972) is one of the most commonly used
statistical models for the epidemiologic analysis of survival and mortality in cohort studies with
extensive follow-up, including studies of the Libby worker cohort (Larson etal., 201 Ob:
Moolgavkar et al., 2010). In the Cox proportional hazards model, the conditional hazard
function, given the covariate array Z, is assumed to have the form
where ft is the vector of regression coefficients, Ao(0 denotes the baseline hazard function, and T
denotes transposition of the vector. One of the strengths of this model is that knowledge of the
baseline risk function is not necessary, and no particular shape is assumed for the baseline
hazard; rather, it is estimated nonparametrically. The contributions of covariates to the hazard
are multiplicative. When Z represents exposure and f>TZ is small, the Cox proportional hazards
model is consistent with linearity of the dose-response relationship for low doses.
The Cox proportional hazards model assumes that a function of covariates (i.e.,
exposures) result in risks that are a constant multiple of the baseline hazard in unexposed
individuals over some timescale, typically calendar time or age. This proportionality is assumed
to be constant across the range of observed exposures, given the set of modeled covariates, and
can be evaluated across time. When the proportional hazards assumption holds, it is possible to
estimate the hazard ratio of exposure (relative risk) without estimating the hazard function in the
unexposed (or in the lowest exposures seen within the study group), because this baseline hazard
function drops out of the calculations.
Other methods common to occupational epidemiology, such as the use of standardized
mortality ratios (results shown above in Tables 5-27 through 5-29) typically rely upon
comparisons of the mortality experience in an exposed population group compared to that in the
general population. However, the comparison population may not always be appropriate due to
differences in general health status (e.g., the healthy worker effect) and differences in exposure
to other risk factors for a specific disease (e.g., smoking history). The lack of comparability
between the study population and the comparison population can lead to confounding by other
measured or unmeasured characteristics that may be statistically associated with both the
exposure of interest and the endpoint. The Cox proportional hazards model controls for such
potentially confounding characteristics by using a comparison group from within the study
population (i.e., internal controls). Internal controls are a statistically appropriate comparison
group because they are expected to be more similar in potentially confounding characteristics to
the remainder of the cohort, thereby controlling for both measured and unmeasured confounding
and helping ensure that comparisons are more statistically valid.
5-99
-------
5.4.3.3.1. Lung cancer mortality analysis in the Libby worker cohort. As described in the
previous section, quantitative exposure-response relationships for lung cancer mortality were
evaluated using the Cox proportional hazards model. Cox proportional hazards models of this
type require the specification of a timescale. Age is typically the time-related variable with the
strongest relationship to cancer mortality and was used as the timescale in these analyses. Use of
age as the timescale in a time-varying Cox proportional hazards model controls for age as a risk
factor by design rather than by parametric modeling and effectively rules out age as a potential
confounder. Individual covariates available to EPA in the complete analytic data set compiled
from the NIOSH data were evaluated for their ability to explain lung cancer mortality. These
included gender, race, birth year, age at hire, and various exposure-related variables including
TWA workplace intensity of exposure in fibers/cc, job type, and the start and stop date of each
different job. These data allowed for the computation of cumulative exposure, cumulative
exposure with application of a half-life, and RTW cumulative exposure, with and without
application of a half-life (see Section 5.4.2.5). Each exposure metric was also lagged by 0, 10,
15, or 20 years. The use of a lag period aims to account for the latency period between the onset
of lung cancer (which occurs some time before clinical diagnosis) and lung cancer mortality.
All lung cancer mortality analyses were conducted using SAS software version 9.1 (SAS,
Gary, NC). EPA fit the extended Cox proportional hazards model (Tableman and Kim, 2004;
Kleinbaum and Klein, 1996), which included both time-independent factors such as gender, race,
and date of birth, as well as time-dependent measures of LAA exposure over the entire time
course of each individual's lifetime from his or her date of hire until death or loss to follow-up.
The inclusion of date of birth in these analyses controls for potential birth cohort effect, which is
strongly related to smoking patterns as people of different generations develop different smoking
rates.
EPA's analyses of time-dependent exposure data included goodness-of-fit testing of the
proportionality assumption for the Libby worker cohort. Because Cox proportional hazard
models rely on the assumption that the hazard rate among the exposed is proportional to the
hazard rate among the unexposed, it is important to evaluate the model against this assumption.
Therefore, analyses of extended Cox proportional hazards models tested this assumption using a
Wald test on the model interaction term between the LAA exposure metric and the timescale
(i.e., age). As a general rule, a nonzero slope that is either increasing or decreasing indicates a
violation of the proportional hazards assumption. Wald tests for the complete cohort consistently
showed that the interaction term was a statistically significant predictor of lung cancer mortality
(p < 0.05) and was interpreted as evidence that the hazards did not remain proportional over
time. The cause of the lack of proportionality is unknown, but several likely explanations are
discussed in Section 5.4.3.4 below and in the discussion of uncertainties in Section 5.4.6.1.
5-100
-------
5.4.3.4. Rationale for Analyzing the Subcohort ofLibby Workers After 1959
Several possible explanations exist for the finding that duration of exposure was the best
fitting exposure metric for mesothelioma mortality, as well as the finding of the lack of
proportionality of hazards in the lung cancer mortality modeling.
• Duration of exposure, but not department code or job category, was known for 706 of
991 (71%) workers hired from 1935 to 1959. The same exposure concentration had been
estimated for almost all of these workers, likely resulting in a particularly large
measurement error for exposure in approximately one-third of the total cohort of
1,871 workers. Assigning the same exposure concentration to so many of the workers
hired before 1960, regardless of job, likely resulted in significant exposure
misclassification and may explain the superior fit for duration of exposure in modeling of
mesothelioma mortality relative to the other exposure metrics based on measured
exposures.
• Even where the job category was identified, few exposure data exist prior to 1968. For
the majority of job locations (23 of 25), no exposure measurements were available prior
to 1967, and so exposures were estimated based on employee interviews (conducted in
1982) to determine what was known about major changes in operations between 1935
and 1967. For two job locations, dust-to-PCM extrapolations are based on measurements
taken in the late 1960s; thus, extrapolating from the mid 1960s to the early 1960s is likely
to be more certain than extrapolating further back in time. Random error in these
exposure measurements would also generally attenuate the strength of association
between exposure and observed effect during the earlier years of mine operation, and
thus, a greater degree of measurement error in the earlier years could have resulted in the
lack in proportionality of the hazard ratios for lung cancer over time. A greater degree of
measurement error in the earlier years could also provide an explanation for the worse fit
of the mesothelioma models that incorporated these exposure measures.
• Another explanation for the lack of proportional hazards in modeling lung cancer
mortality may be that this cohort has an anomalous age structure due to the hiring of
much older individuals during the time of the Second World War. Among those workers
in the cohort hired prior to 1960, 9% were older than 50 years at the time of hire, and
22% were older than 40 years. Among those workers hired in 1960 or afterwards, only
4% were older than 50 years, and 14% were older than 40 years. Older workers differ
from younger workers in several potentially important ways that could alter their
response to exposures. Older workers were born in a different era, with different
nutritional and public health standards, which may influence mortality patterns.
• The lack of proportional hazards in modeling lung cancer mortality may also be a
reflection of confounding or effect modification, which can change in magnitude over
time. The most likely candidate for confounding or effect modification is smoking.
NIOSH records show that of the 1,871 workers in the full Libby workers cohort,
1,121 workers (60%) were missing smoking status data, while 750 (40%) had data with
values "S" (Smoker), "Q" (Former Smoker), or "TV" (Nonsmoker). Given this high
percentage of missing values, EPA did not consider these smoking data to be adequate
for use in the evaluation of confounding or effect modification. Effect modification by
age is another possibility and may also explain the lack of proportionality in the modeling
5-101
-------
of lung cancer mortality as has been noted by Richardson (2009) in a two-stage clonal
expansion model of lung cancer risk in a cohort of asbestos exposed workers; however,
similar two-stage clonal expansion modeling of lung cancer risk in the same cohort of
workers was unable to replicate that finding, which may be due to the reliance of this
methodology on additional assumptions which EPA does not have a basis for making
(see discussion of two-stage modeling in Section 5.4.3.3).
• Smoking rates and patterns among the subcohort of workers hired after 1959 are likely to
have been more similar because smoking rates change more slowly over shorter periods
of time than over longer ones. This restriction in time period of hiring would also result
in less variation by birth year cohort, which is strongly related to smoking patterns as
people of different generations develop different smoking rates. Thus, this restriction in
the time period of hiring may make the cohort members more similar to each other,
thereby possibly reducing the potential impact of any smoking-related confounding.
Further discussion of the relevance of smoking can be found in the section on
uncertainties (see Section 5.4.6).
When the assumption of proportionality is not met, the potential influence of
confounding factors in the full-cohort analysis of lung cancer mortality is of concern.
Additionally, the lack of job category information for 71% of the workers hired prior to 1960 and
greater measurement error in early exposures may result in significant random exposure
measurement error, which may bias the observed exposure-response relationships towards the
null.
Although duration of exposure was the best exposure metric for modeling mesothelioma
mortality in the full cohort, it does not allow quantitatively estimating an exposure-response
relationship to support IUR. In addition, violation of the underlying statistical assumptions
adversely affected modeling of lung cancer mortality in the full cohort. Therefore, EPA chose to
undertake a subcohort analysis of workers hires after 1959.
While it is generally true that the use of more data is an advantage in statistical analyses
because it allows for the computation of more statistically precise effect estimates, this advantage
could not be realized, because of the difficulty in deriving risks from the full cohort analysis (see
next section on uncertainties remaining in the subcohort). The reasons stated in Section 5.4.2 for
choice of Libby worker cohort data over other study populations are still valid for the subcohort.
In particular, (1) these workers were directly exposed to LAA, (2) detailed work histories and
job-specific exposure estimates are available to reconstruct estimates of each individual's
occupational exposure experience with only nine workers completely missing job and
department codes during the period of relatively high average time-weighted estimated exposure
intensity, (3) the subcohort is still sufficiently large and has been followed for a sufficiently long
period of time for cancer to develop (i.e., cancer incidence) resulting in mortality, and (4) the
broad range of exposure experiences in the subcohort provided an information-rich data set.
EPA initially examined the fit of these models using several exposure metrics to predict
mortality from mesothelioma and found that in this subcohort, the exposure metrics that included
5-102
-------
information on exposure concentration provided superior statistical fits to the exposure metrics
based only on employment duration. In this same subcohort, the assumptions of the Cox
proportional hazards model for analysis of lung cancer were also satisfied for the modeling of
time-varying exposure.
On the other hand, there are quantitative uncertainties related to the choice of the
subcohort. First of all, the numbers of cases of both lung cancer and mesothelioma are lower
than in the whole cohort. Second, the follow-up of subcohort, while in excess of 40 years, may
not be sufficiently long to encompass all potential lung cancer, especially, mesothelioma
mortality related to LAA exposures. Third, the subcohort is younger and overall mortality is
lower than in the full cohort. However, the choice of the subcohort is appropriate because of the
superior exposure information based on a higher percentage of estimated exposures from actual
measurements as opposed to inferred exposure values. The higher percentage of actual
measurements allows a more accurate dose-response evaluation [see the discussion in Lenters et
al. (2012) and Lenters et al. (2011)] on the impact the quality of the exposure information has on
estimates of dose-response relationships [see Bateson and Kopylev (2014)].
5.4.3.5. Results of the Analysis of Mesothelioma Mortality in the Subcohort
Of the 880 workers hired after 1959, 230 (26%) had died by December 31, 2006. The
number of mesothelioma deaths in the subcohort is seven (two deaths coded in ICD-10 and five
deaths coded in ICD-9). The mesothelioma death rate of 2.47 per 10,000 person-years for the
subcohort is similar to the mesothelioma death rate of 2.68 per 10,000 person-years for the full
cohort (18 mesothelioma deaths), with a difference of less than 10%.
Tables 5-30 to 5-32 show the mesothelioma mortality rate by duration of exposure, age of
first exposure, and TSFE. As in the full cohort, both duration of exposure and TSFE show a
relationship with mesothelioma mortality rate. However, unlike the full cohort, where there was
no relationship with age at first exposure, in the subcohort, ages greater than 25 may be
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 10"3 and 7.90 x 10"3, respectively. Therefore, as in the full cohort,
overdispersion is very unlikely.
5-103
-------
Table 5-30. Mesothelioma mortality rate in the subcohort of employees
hired after 1959 shown by duration of exposure (yr)
Deaths/PY
Rate x 10-4
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
<15yr
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
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.
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 are commonly used in Bayesian analyses. Information weights are
5-104
-------
computed by first assessing the differences between the best DIG and each of the others (A Did)
(see eq 5-14).
( i
DICwi. =exp —
2
(5-14)
where
,R is the number of models and
Z)/C Wi 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.
Metrics with higher DICs and lower information weights indicate a poorer model fit and
are not included in Table 5-33. The other exposure metrics that were evaluated included those
metrics used in the full cohort analysis (duration of exposure, TSFE, age at death or censoring,
RTW metrics, and CE with lag metrics), but none of these metrics fit as well as the metrics in
Table 5-33.
The two metrics with cumulative exposure lagged 15 and 10 years, both with 5-year
half-life, provided the two best fits as indicated by their lower DIC values and higher information
weights (see Table 5-33). Cumulative exposures lagged 10 or 15 years, both with 10-year
5-105
-------
half-life, provided the next two best fits according to DIG values, but models including each of
these metrics exhibited noticeably lower information weights than the best metric. All metrics in
Table 5-33 contain a decay term and have the same number of parameters in their corresponding
model, allowing for a direct comparison of the DIG values and information weights.
For models from the amphibole asbestos literature, in the subcohort hired after 1959, the
DIG value for mesothelioma using the Peto metric (see eq 5-11) is substantially higher
(DIG = 98.4) than for any of the metrics in Table 5-33. This indicates that the metric of exposure
used in the previous IRIS IUR (U.S. EPA, 1988a) does not provide as good a fit for the LAA
worker cohort as the other metrics of exposure in Table 5-33. Setting the power term on time
since first exposure (k in eq 5-11) in the IRIS IUR (U.S. EPA, 1988a) metric to the values of 2
and 4, as suggested by U.S. EPA(1986a), continues to yield substantially higher DIG values
compared to the fit values of the exposure metrics in Table 5-33 (DIG = 89.2 and 107.9,
respectively). For the Peto model with clearance, increasing the power term in the Peto model
with clearance to k = 3.9 with decay X = 0.068 and k = 5.4 with decay X = 0.015 decreased DIG
slightly from the standard Peto model itself (DIG = 95.4 and 95.3, respectively). Using CE
instead of C in decay models, as discussed in Berry et al. (2012), made the fit much worse, as
measured by DIG. The fit also degraded when using the Berry etal. (2012) models of the form
(C or CE) x (T- 5)k.
Next, EPA considered which covariates should be added to the model with the exposure
metric that provided the best fit. The addition of covariates "age at death or censoring" and
"TSFE" did not improve the fit, as measured by DIG (results not shown).
As discussed above, EPA tabulated the mesothelioma rates for the two best fitting metrics
in Table 5-33 and for alternative models proposed in the amphibole asbestos literature (i.e., Peto
model and Peto model with clearance) in Tables 5-34 to 5-38. These tables show information by
quintiles for each metric of exposure.
The first two tables (see Tables 5-34, 5-35) show a dose-response relationship between
mesothelioma deaths and values of each exposure metric—while the mesothelioma data are
sparse, higher values of metric correspond to higher rate of mesothelioma. Tables 5-36 to 5-38
show a somewhat less clear dose-response relationship for the Peto model and the Peto model
with clearance, as the relationship between metric and rate appears to be somewhat parabolic.
5-106
-------
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 10-4
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
5-107
-------
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 10-4
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
To further illustrate the fit of the Peto model to the Libby mesothelioma data, the
frequency of the values of exposure computed using the Peto method and two best-fitting
exposure metrics, which were CE with 5-year half-life and 10- or 15-year lag, were plotted and
the mesothelioma cases were noted (see Figures 5-6 to 5-8). These figures provide more
information than Tables 5-34 through 5-36. The histograms show in a relative scale the
frequency of occurrence of each value of the specific exposure metric, thereby revealing the
complete distribution and showing where exposure values of the cases were. In these figures, the
exposure metric has been transformed to the natural log scale which yields a more normalized
distribution. The figures show how the exposure values of the mesothelioma deaths relate to the
exposure values of the subcohort of workers hired after 1959. Better fitting models are expected
to show higher exposure values for the mesothelioma cases relative to those who did not die
from mesothelioma, while poorer fitting models are expected to show mesothelioma cases
scattered with equivalent density to the distribution as a whole and, thus, closer to the center of
the distribution of exposure metric.
5-108
-------
Peto metric and mesothelioma de
v v
ww
>,
o
^
CT
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.
From comparing Figures 5-6 (above) and Figures 5-7, and 5-8 (below), exposure values
of the mesothelioma deaths based on the Peto exposure metric are clearly closer to the center of
the distribution and, therefore, more like the values of those that did not die of mesothelioma.
Thus Peto exposure metric does not reveal a higher likelihood of death from mesothelioma
compared to the other exposure metrics. This is consistent with what was observed in
Tables 5-34 through 5-38—the Peto model does not fit the subcohort data as well as CE metrics
with decay fit. For the CE-based exposure metrics (see Figures 5-7 and 5-8), unlike the Peto
exposure metric (see Figure 5-6), mesothelioma deaths are concentrated at high values of the
exposure metrics and not as close to the center of the distribution.
5-109
-------
CE with lag 15 and half-life 5 year
o
£=
0)
=3
CT
0)
V W
vw
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.
5-110
-------
CE with lag 10 and half-life 5 year
>,
o
cr
o>
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:
1) While TSFE by itself, without personal exposure information, shows a clear
relationship with increases in mesothelioma rates with increasing TSFE (see
Table 5-32), adding information on concentration of fibers to the model (as the Peto
model does) appears to degrade the fit (compare Table 5-36 to Table 5-32). The
amphibole cohorts (amosite and crocidolite), which were used to derive the Peto
model and the Peto model with clearance had little, if any, personal exposure data.
Therefore, those models were predominantly based on the power of TSFE and
demonstrated a good fit to the exposure timing data from those cohorts. Lack of good
exposure data, and in particular personal exposure information, was a limitation of
these analyses. In the case of the Libby workers subcohort, personal exposure data is
available and when personal exposure is taken into account, the models do not appear
to fit as well.
5-111
-------
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-lev el 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
xlO5
(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)
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).
5-112
-------
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
xlQ-8
(0.04,0.15)
(0.34, 1.09)
(0.52, 1.72)
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).
5.4.3.6. Results of the Analysis of the Lung Cancer Mortality in the Subcohort
EPA based its final analyses for lung cancer mortality on the subset of workers hired after
1959. Thus, this analysis is based on 32 deaths from lung cancer25 (ICD-8: 2 deaths with the
code 162.1; ICD-9: 1 death with the code 162.2, 20 deaths with the code 162.9; ICD-10:
9 deaths with the code C349) out of 230 total deaths that occurred in the subcohort of
880 workers.
Tables 5-41 to 5-43 show lung cancer mortality rates by duration of exposure, age of first
exposure, and TSFE. As in the full cohort, duration of exposure, age at first exposure, and TSFE
all show relationships with lung cancer mortality rate.
25Note 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
trachea! 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.
5-113
-------
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 lO'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).
5-114
-------
As noted in Section 5.4.3.5, these marginal analyses do not specifically include the
effects of the exposure as well as both the duration and the TSFE. Therefore, EPA investigated
the overall fit of different exposure models and tabulated the results of several models that
include personal exposure information.
All multivariate Cox proportional hazards models with time-varying exposures were
initially fit, using one exposure metric at a time, to the subcohort hired after 1959 with covariates
for gender, race, and date of birth. Lung cancer mortality was modeled using CE and RTW
exposure, where each metric was potentially modified by four different half-lives (5, 10, 15, or
20 years). Each of these exposure metrics was also evaluated with four different lag periods to
allow for cancer latencies of 0, 10, 15, or 20 years. In all, 40 multivariate exposure-response
models were evaluated for the adequacy of the exposure metric to fit the epidemiologic data.
Each model and the comparative model fit statistics are presented in Table 5-44.
5-115
-------
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
CE20-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-valuea
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
5-116
-------
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-valuea
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.
aLikelihood ratio test (overall model fit where/) O.05 indicated an adequate fit).
The assumptions of the Cox proportional hazards model were reevaluated for the
subcohort. Restricting the cohort addressed each of the previously listed potential explanations
for the lack of hazard proportionality (see Section 5.4.3.3). First, measurement error for
exposures is likely to have been smaller after 1959 for several reasons. One reason is that
706 workers were removed from the analysis because job category and department code
information were missing during all of their employment prior to 1960. Also, beginning in 1968,
fiber concentrations by PCM analysis of site-specific air samples were available for all location
operations to inform the JEM.
Second, prior to 1968, the exposure intensity for 23 of 25 location operations was
estimated based on assumptions informed by employee interviews in the early 1980s. It is likely
the uncertainty of these assumptions increased the farther back in time that exposures were
estimated, making the earliest exposure estimates (1940s and 1950s) less certain than those only
a few years before fiber count data were available. Further, between 1956 and 1967,
5-117
-------
dust-to-PCM extrapolation data were used to estimate exposures in the dry mill based on
measurements taken in the late 1960s. Although there is some uncertainty in the conversion ratio
selected by Amandus et al. (1987b), dust-to-fiber conversions are likely to be less uncertain than
extrapolations further back in time to the 1950s and 1940s, where only one air sample for dust
was available in 1944. Thus, the potential attenuation effect of nondifferential measurement
error is likely to be reduced by examining the post-1959 cohort alone compared to the entire
cohort.
Third, smoking rates among this more narrowly defined subcohort are likely to have been
more homogeneous; thus, restricting analysis to this subcohort would help to limit any potential
confounding due to smoking.
Finally, EPA conducted goodness-of-fit testing of the extended Cox proportional hazards
model as applied to the subcohort hired post-1959. There was no evidence to reject the
hypothesis of proportionality, and the exposure models demonstrated adequate fits to the data,
with statistically significant effect estimates. In each of the Cox proportional hazards model
analyses with time-varying exposures—across all the exposure metrics and across all the lag
lengths—no violations of the assumption of proportionality of hazards were found.
As the exposure-response models cannot strictly be considered to be nested, a standard
measure of fit called the AIC (Burnham and Anderson, 2002) was used for comparison of
goodness of fit across models based on the same data set. In their text on model selection,
Claeskens and Hjort (2008) state that".. .for selecting a model among a list of candidates, AIC is
among the most popular and versatile strategies." Claeskens and Hjort (2008) also state that the
model yielding the smallest AIC is judged the best fitting and it is a common practice in
environmental epidemiology to simply select the single model with the best statistical fit (i.e., the
lowest AIC) among the models evaluated. While large differences in AIC values can reveal
important differences in model fit, small differences are less conclusive. For example, in a set of
models differing in AIC by two or fewer units, each can be considered to have a substantially
similar level of empirical support [(Burnham and Anderson, 2002): p. 70].
Table 5-44 shows the models and exposure metrics ordered by fit. Of interest is whether
there are models with distinct exposure metrics that adequately fit these data (as measured by
statistical significance of the models-value) and then whether a measure of relative fit exists
among these adequately fitting models. Of the 40 exposure-response metrics, 14 demonstrated
an adequate fit to the data as measured by the overall model fit, with the standard likelihood ratio
test being statistically significant (p <0.05), as well as having statistically significant exposure
metrics (p <0.05). However, note that only the nine models that demonstrated adequate model
and exposure metric fit and incorporated a lag period to account for lung cancer mortality latency
were advanced for potential use in developing a unit risk. While metrics that did not include an
adjustment for lag on the exposure metric to account for cancer mortality latency were fit to
these data for the sake of completeness, they were dropped from further consideration because
5-118
-------
they implicitly assume no passage of time between the initiation of cancer, subsequent promotion
of that cancer, and mortality.
Several general patterns were discernible with respect to which exposure metric(s) best
predicted lung cancer mortality when comparing AICs for relative model fit. The data show that
lagging exposure by 10 years best predicts lung cancer mortality compared to other lags. This
trend is seen across both the cumulative exposure without decay and the various half-life
cumulative exposure metrics where a 10-year lag of exposure best predicts lung cancer mortality
for all cumulative exposure metrics compared to other lags. Metrics with 15-year lags were
generally the next best in terms of fit. Another conclusion is that the models that included RTW
exposure metrics, regardless of half-life or lag, did not fit as well as the models that employed
cumulative exposure with different half-lives and lags.
Among the 40 exposure metric models that were evaluated, the exposure model with the
lowest AIC value was for cumulative exposure with a 10-year half-life for decay and a 10-year
lag for cancer mortality latency and had a models-value based on the likelihood ratio test of
0.0071 (see Table 5-44). This multivariate model controlled for age, gender, race, and date of
birth. This model estimated a KL (slope) of 1.26 x 10"2 per fiber/cc-yr based on a 365-day
calendar year,26 and the 95th percentile upper bound on this parameter was
1.88 x 10"2 per fiber/cc-yr. The/>-value for the LAA regression coefficient (slope) was <0.001,
indicating that this parameter was statistically significantly greater than zero. Table 5-45 shows
the slopes and confidence intervals for all retained metrics from Table 5-44. Figure 5-9 shows
the model residuals (in this case, the Schoenfeld residuals for Cox models) for the retained
models in Table 5-45. Patterns in such residuals such as an increasing or decreasing slope
overall can indicate lack of fit. Attention is directed at the pattern of residuals with respect to
age at death (the x-axis). None of the plots appears to show a meaningful deviation from a linear
function of age which indicates a lack of interaction between the model-predicted effect of
exposure on lung cancer mortality risk and age. That is, the risk of lung cancer mortality does
not appear to vary by age within the subcohort of workers at the Libby facility. There is no
indication of a systematic lack of fit across these models excepting the nonlinear departure in the
center of each residual plot which appears to be random and is minimized with the pattern in the
residuals if smoothed out to a greater extent. The model fit residuals are consistent with the
similarity of the AIC values in demonstrating similar model fit.
26The two-sided 90% confidence interval is (6.00 x 1Q-3, 1.88 x 1Q-2).
5-119
-------
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
^-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)
5-120
-------
Exposure Metric
Less smoothing
More smoothing
CE—no half-life; 10-yr lag
CE—5-yr half-life; 10-yr lag
CE—10-yr half-life; 10-yr lag
CE—15-yr half-life; 10-yr lag
CE—20-yr half-life; 10-yr lag
5-121
-------
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
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.
5-122
-------
According to the model results presented in Table 5-44, there were multiple exposure
metrics that predicted lung cancer mortality and exhibited statistically significant effect
estimates. Several other metrics were considered to fit nearly as well as the model with the
smallest AIC because their AIC values were within two units of the exposure model with the
lowest AIC, a proximity that can be considered to be a range that cannot clearly differentiate
among models (Burnham and Anderson, 2002). As each of the other exposure metrics was
based on a different configuration of the same exposure data, the different slopes (KLs) are not
directly comparable, but all adequately fitting lagged models also produce statistically significant
slopes for the exposure-response relationship (p <0.05). Of particular note are the results of the
cumulative exposure model with a 10-year lag for latency but without a decay function because
this model showed the lowest AIC among nondecay models.
The AIC values for models that included lag and/or half-life adjustments to the exposure
metrics were not penalized in the regression analyses for using these extra parameters because
these factors were not represented as covariates but rather were embedded in the computation of
the exposure metric. While these results were obtained using each instance of lag and/or half-life
terms in separate model fit, it may be appropriate to mathematically penalize the AICs for
inclusion of these additional parameters. AIC values, as typically computed by regression
software, include the addition of a penalty for model complexity as measured by the number of
parameters that are fit in the regression model (thereby increasing the AIC). In the AIC
calculations presented in Table 5-44, the models are treated as having the same number of
parameters because each model represents the same individual's time-varying exposures in a
different way but with a single exposure parameter in the regression models. For that reason, the
models are equally penalized in the software's AIC calculation. However, because an argument
can be made that exposure metrics that do not include a decay function, with an explicit half-life
term, are implicitly more parsimonious (simpler), a comparison of the AICs is not
straightforward. If the decay model fits were penalized for the inclusion of the decay function in
the computation of the exposure metric, then with such an adjustment, the relative fit of the CE
models would be somewhat improved in terms of their comparison with the values in Table 5-44
(AICs are generally penalized two units for each additional parameter).
Table 5-45 displays the lagged exposure-response models and metrics with adequate
model fit (p <0.05) to the epidemiologic data that were further considered. The units of the
slopes are fibers/cc-yr. These slopes and confidence intervals represent calendar year continuous
environmental exposure as described above and define the "Exposed Hazard Rate" in the
life-table procedure when multiplied by the exposure level (see Appendix G for details). The
plots in Figure 5-9 do not suggest a meaningful difference in model fits among the nine different
parameterizations of exposure with adequate fit.
As presented in Table 5-45, the CE model with 10-year half-life and lag provided an
adequate fit to the data based on the likelihood ratio test (p <0.05) and had the lowest AIC value.
5-123
-------
The cumulative exposure model with a 10-year lag also yielded a statistically adequate fit to
these data (p < 0.05), as did several decay models with a 15-year lag. These results demonstrate
reasonable uncertainty in the metric of exposure such that no single exposure model can be
definitively selected based on goodness of fit alone. Issues related to uncertainty in the choice of
exposure metric are described further in the section explaining the derivation of the combined
IUR of mesothelioma and lung cancer (see Section 5.4.5.3).
5.4.3.7. Sensitivity Analysis of the Influence of High Exposures in Early 1960s on the Model
Fit in the Subcohort
As discussed in Section 5.4.2.5, the comparison of model fit among various exposure
metrics is an empirical process and does not necessarily reflect a specific biological or other
factor as an underlying cause for model fit. Although data do not exist to evaluate biological
bases for model fit, other potential factors can be explored where data allow. For example,
because of concerns that very high (>100 fibers/cc) 8-hour TWA exposures during 1960-1963
(see Table 5-21) could have influenced the relative fit of the various exposure metrics, EPA
conducted a sensitivity analysis of the impact on the relative model fit of reducing all estimated
exposure intensities for 1960-1963 by 50%.
For modeling mesothelioma mortality on this revised data set, one change occurred in the
relative fit of 3rd and 4th best fit decay models, but the observation that exposure metrics
including decay fit better than exposure metrics without decay was unchanged (see Table 5-46).
However, the fit of all the metrics decreased slightly, with each DIG increased between 0.3 and
1.1. The metrics without decay and RTW metrics had DIG values higher than those in
Table 5-46. The revised data set DIG for the model used in IRIS IUR (U.S. EPA. 1988a)
was 97.9.
5-124
-------
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.
For modeling lung cancer mortality on this revised data set, no difference was present in
the order of the relative fit between the same exposure metrics that fit the subcohort of workers
hired after 1959 and those exposures estimated by Amandus et al. (1987b) for 1960-1963 (see
Table 5-47). The metrics based on the revised data set fit marginally better based on AIC.
5-125
-------
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 models-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.
This sensitivity analysis reduces some of the potential uncertainty in the results that may
have been attributed to exposure measurement error specific to the 1960-1963 time period when
some of the estimated exposures were particularly high.
5.4.3.8. Additional Analysis of the Potential for Confounding of Lung Cancer Results by
Smoking in the Subcohort
In the full cohort analysis, the proportional hazard assumption was not found to hold, and
one of the reasons for this failure was the possible presence of confounding by smoking, which
altered the proportionality of the hazard rate in the exposed workers compared to the baseline
hazard rate over time. Confounding, which can bias observed results when there is an
uncontrolled variable that is correlated with both the explanatory variable and the outcome
variable, is a distinct concept from effect-measure modification (e.g., synergy), which might
reflect different observed effects of exposure to LAA among smokers as compared to
nonsmokers. The extent of effect-measure modification cannot be assessed without adequate
data on smoking; however, the potential for effect-measure modification is discussed in
Section 5.4.6.
5-126
-------
As an additional check on the potential for confounding, a novel method was evaluated to
test for confounding by smoking in occupational cohorts that do not have data on smoking. A
method has been described by Richardson (2010) to determine if an identified exposure
relationship with lung cancer is confounded by unmeasured smoking in an occupational cohort
study. Richardson (2010) demonstrated that an exposure of interest (i.e., LAA) can be used to
predict an outcome other than lung cancer such as COPD, which is known to be caused by
smoking, but not thought to be related to the exposure of concern.27 If a positive relationship is
identified where no causal association is suspected, this would suggest that smoking and the
exposure metric (LAA) were positively correlated and that the identified exposure-response
relationship was, in fact, confounded by smoking. EPA implemented this methodology to model
the potential effects of LAA on the risk of COPD mortality (n= 18) on the subcohort of workers
hired after 1959. Using the exposure metric defined as cumulative exposure with a 10-year lag,
the extended Cox proportional hazards model with time-varying exposures estimated a slope
(beta) for COPD of-0.056 per fiber/cc-yr based on a 365-day calendar year. The/?-value for the
coefficient (slope) was 0.102, indicating that this parameter was not statistically significantly
different from zero. Using the exposure metric defined as cumulative exposure with a 10-year
half-life for decay and a 10-year lag for cancer latency, the extended Cox proportional hazards
model with time-varying exposures estimated a slope (beta) of-0.135 per fiber/cc-yr based on a
365-day calendar year. The/?-value for the coefficient (slope) was 0.116, indicating that this
parameter was not statistically significantly different from zero.
Summarizing these findings, EPA used the method described by Richardson (2010) to
evaluate whether exposures to LAA predicted mortality from COPD as an indication of potential
confounding by smoking and found a nonsignificant negative relationship, which was
inconsistent with confounding by smoking in the subcohort of workers hired after 1959.
5.4.4. Exposure Adjustments and Extrapolation Methods
The estimated exposures based on the JEM and work histories are discussed in
Section 5.4.2.5. Note that all potency estimates (i.e., KM or KL) presented with units of
fibers/cc-yr are for calendar year and not for occupational year, so no additional adjustment is
needed to address this difference as may have been found in other evaluations based on
occupational epidemiology cohort analyses. Adjustments for differences in breathing rates and
the number of hours of exposure in an occupational (8-hour) day as compared to a whole
(24-hour) day are not incorporated directly into the slope but rather applied in the derivation of
the central risk and unit risk estimates.
27Richardson (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.
5-127
-------
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 MOA 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
o
J
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
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).
5-128
-------
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
5.4.5.1.1. Adjustment for mesothelioma underascertainment. For mesothelioma, the
undercounting of cases (underascertainment) is a particular concern given the limitations of the
ICD classification systems used prior to 1999. In practical terms, this means that some true
occurrences of mortality due to mesothelioma are missed on death certificates and in almost all
administrative databases such as the National Death Index. Even after the introduction of a
special ICD code for mesothelioma with the introduction of ICD-10 in 1999, detection rates are
still imperfect (Camidge et al., 2006; Pinheiro et al., 2004), and the reported numbers of cases
typically reflect an undercount of the true number.
Kopvlev et al. (2011) reviewed the literature on this underascertainment and developed a
general methodology to account for the likely numbers of undocumented mesothelioma deaths
using the Libby worker cohort as an example. Because the analysis of mesothelioma mortality
was based on absolute risk, it was possible to compensate for mesothelioma underascertainment
in the Libby worker subcohort. Kopvlev et al. (2011) considered analyses when mesothelioma
type (i.e., pleural and peritoneal) is unknown and when it is known. As the number of peritoneal
mesotheliomas is partially known in the Libby worker subcohort, the method for known
proportion of pleural and peritoneal mesothelioma deaths is briefly described here.
Selikoff and Seidman (1992) provided information on the likelihood that individuals who
have been diagnosed as having mesothelioma will have that disease recorded (in some field) on
their death certificate. Their results are based on histopathological analysis (Ribak et al., 1991)
of a very large cohort of insulators, with more than 450 mesothelioma cases. Despite medical
advances, diagnosis of mesothelioma is still very challenging, and histopathology is still a
standard diagnostic tool today [e.g., Mossman et al. (2013)]. Using their results on the most
common misdiagnoses of mesothelioma (mesothelioma diagnosed as lung, colon, and pancreatic
cancers; and conversely, other diseases misdiagnosed as mesothelioma) and likelihoods of
corresponding misdiagnoses, Kopvlev et al. (2011) conducted a simulation study randomly
5-129
-------
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, Kopylev 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
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
5-130
-------
5.4.5.2. Unit Risk 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 MOA 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
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
5-131
-------
correspond to the measure of central tendency involves ECoi rather than LECoi. The ECoi
yielded a lifetime central estimate of 0.0260 per fiber/cc.
Using the results of the exposure model based on cumulative exposure with a 10-year lag
for cancer latency, the LECoi for the adult-only exposures was determined to be 0.191 fiber/cc.
This LECoi yielded a lifetime unit risk of 0.0679 per fiber/cc. The ECoi for the adult-only
exposures was determined to be 0.325 per fiber/cc. When divided into a POD of 1%, this ECoi
yielded a lifetime central estimate of 0.0399 per fiber/cc.
The resulting unit risks in Table 5-52 ranged from 0.0260 to 0.0679 per fiber/cc. This
shows that the unit risk (i.e., 0.0389 per fiber/cc) based on the exposure metric 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) is in the center of this range, and is thus statistically robust. However, because
this estimate is in the middle of the range, it does not capture the uncertainty across metrics with
similar goodness of fit. As noted (see Section 5.4.3.6., an argument can be made that the CE
metric with a 10-year lag and no half-life is implicitly more parsimonious (simpler) because it
was not explicitly adjusted to include decay, although this metric is mathematically equivalent to
CE metric with a 10-year lag and an infinitely long decay half-life. Conceptually, the AIC
values are penalized for increased model complexity (thereby increasing the AIC). The CE
metric with a 10-year lag does fit these data—both statistically and by examination of the
residuals. Further, the CE metric is a simpler and more straightforward metric, and has an
extensive tradition of use in the epidemiologic literature and in the practice of risk assessment.
5.4.5.3. Inhalation Unit Risk (IUR) Derivation for Combined Mesothelioma and Lung Cancer
Mortality
For mesothelioma, the exposure-response models developed by EPA using personal
exposure data on the subcohort (see Table 5-50) provided better fit to the subcohort data than the
Peto model and the Peto model with clearance that have been proposed in the asbestos literature
(see Table 5-51). These variations of the Peto model have been shown to predict mesothelioma
mortality more precisely than the original Peto model in a large crocidolite-exposed cohort of
6,908 workers with 329 mesothelioma deaths from Wittenoom, Australia (Berry et al.,
2012)—specifically, the best fitting models for that cohort was the Peto model with a power
k = 3.9 and decay rate of X = 0.068 (approximately 10-year half-life) and Peto model with a
power k = 5.4 and decay rate of X = 0.15 (approximately 5-year half-life).
The exposure-response models developed by EPA using personal exposure data on the
subcohort (see Table 5-50) show estimated lifetime unit risks of 0.074 and 0.122 per fiber/cc.
The two Peto models with different clearances and powers of & show lifetime unit risks of 0.086
and 0.035 per fiber/cc, respectively (see Table 5-51). The results of the two different approaches
to modeling, the first based only on the LAA data and the second based on a much larger
population of workers exposed to another but different amphibole asbestos, reveal a relatively
5-132
-------
small degree of uncertainty in the derivation of the mesothelioma lifetime unit risk. Therefore,
EPA selected the model derived directly from the Libby data with the strong support of the
model that best fit the mesothelioma risk in a much larger cohort of amphibole-exposed workers.
EPA's selected model based on cumulative exposure with a 10-year lag and 5-year half-life
yielded a lifetime unit risk for mesothelioma of 0.122 per fiber/cc which encompasses other three
risk estimates. Table 5-53 shows the combined IUR for LAA for mesothelioma and lung cancer
based on the selected mesothelioma model, the two best-fitting mesothelioma models from the
epidemiologic literature (the Peto model with clearance), as well as the combined IUR for LAA
based on 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
LAA) for different combination of mesothelioma and lung cancer models.a'b
Primary IUR value in bold.
Lung cancer
Mesothelioma
Combined central
estimate
(per fiber/cc)
Combined upper
bound
(per fiber/cc)
Selected IUR based directly on the Libby data
CE10
CE105-yr half-life
0.115
0.169
Best models from the epidemiologic literature (Peto model with clearance)
CE10
CE10
Peto with clearance
Decay rate of 6.8%/yr
Power of time = 3.9
Peto with clearance
Decay rate of 15%/yr
Power of time = 5.4
0.089
0.061
0.135
0.092
Alternative model from the epidemiologic literature (Peto model)
CE10
Peto
No decay
Power of time = 3
0.203
0.308
aNote that for the IUR values for LAA shown in this table, the fiber concentration are presented here as continuous
lifetime exposure in fibers/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
etal.. 1987b: Rendall and Skikne. 1980). Methods are available to translate exposure concentrations measured in
other units into PCM units for comparison.
bWhile this assessment is informed by studies of other types of asbestos, it is not a complete toxicity review of
other amphiboles or of chrysotile asbestos.
5-133
-------
For lung cancer, this assessment selected the upper bound among the lung cancer lifetime
unit risks from the plausible exposure metrics (regardless of the small residual differences in
quality of fit). Because there were few metrics with unit risks higher than the best fitting
metric's unit risk for lung cancer mortality endpoint, this method effectively selects the highest
lifetime unit risk among those considered for the lung cancer mortality endpoint. Based on the
selected model for lung cancer mortality using the cumulative exposure with a 10-year lag in the
Libby subcohort data yields a central unit risk estimate of 0.040 per fiber/cc and upper-bound
lifetime unit risk of 0.0680 per fiber/cc.
Once the cancer-specific lifetime unit risks are selected, the two are then combined. It is
important to note that this estimate of overall potency describes the risk of mortality from cancer
at either of the considered sites and is not just the risk of both cancers simultaneously. Because
each of the unit risks is itself an upper-bound estimate, summing such upper-bound estimates
across mesothelioma and lung cancer mortality is likely to overpredict the overall risk.
Therefore, following the recommendations of the Guidelines for Carcinogen Risk Assessment
(U.S. EPA, 2005a), a statistically appropriate upper bound on combined risk was derived in order
to gain an understanding of the overall risk of mortality resulting from mesothelioma and from
lung cancer.
Because the estimated risks for mesothelioma and lung cancer mortality were derived
using Poisson and Cox proportional hazards models, respectively, it follows from statistical
theory that each of these estimates of risk is approximately normally distributed. For
independent normal random variables, a standard deviation for a sum is easily derived from
individual standard deviations, which are estimated from confidence intervals: standard
deviation = (unit risk - central risk) + Zo.95, where Zo.95 is a standard normal quantile equal to
1.645. For normal random variables, the standard deviation of a sum is the square root of the
sum of the squares of individual standard deviations.
As shown in Table 5-50, the upper bound of the selected mesothelioma mortality unit
risks was 0.122 per fiber/cc (highest adjusted unit risk value). The associated central estimate of
risk was 0.075 per fiber/cc for mesothelioma mortality. Table 5-52 shows the upper bound of the
selected lung cancer mortality unit risk was 0.068 per fiber/cc (highest unit risk value based on
LECoi). The associated central estimate of risk was 0.040 per fiber/cc for lung cancer mortality.
It is important to mention here that the assumption of independence of the estimated risks
for mesothelioma and lung cancer mortality (note above) is a theoretical assumption, as there is
insufficient data on independence of mesothelioma and cancer risks for LAA. However, in a
somewhat similar context of different tumors in animals, NRC (1994) stated: ".. .a general
assumption of statistical independence of tumor-type occurrences within animals is not likely to
introduce substantial error in assessing carcinogenic potency." To provide numerical bounding
analysis of impact of this assumption, EPA used results of Chiu and Crump (2012) on the upper
and lower limits on the ratio of the true probability of a tumor of any type and the corresponding
5-134
-------
probability assuming independence of tumors. The lower limit is calculated by
[1 - min(pl,p2)]/(l -pi x P2) and the upper limit is min(l,2 -pi -p2)/(l -pi x p2).
Substituting the risk of lung cancer (pi) is 0.040 and the risk of mesothelioma (p2) is 0.075, the
lower limit is 0.963 and the upper limit is 1.003. A value of 1.0 indicates independence.
Because lower and upper values are both very close to the value of 1.0, this demonstrates that the
assumption of independence in this case does not introduce substantial error, consistent with
what NRC (1994) has stated.
In order to combine the unit risks, first obtain an estimate of the standard deviation of the
sum of the individual unit risks as:
V[ [[(0.122 - 0.075) H- 1.645]2 + (0.068 - 0.040) H- 1.645]2 ] = 0.033 per fiber/cc (5-15)
Then, the combined central estimate of risk of mortality from either mesothelioma or
lung cancer is 0.040 + 0.075 = 0.115 per fiber/cc, and the combined IUR is
0.115 + 0.033 x 1.645 = 0.169 per fiber/cc.
To illustrate the uncertainty in the selected IUR, Table 5-53 shows central risks and upper
bounds for the combined IUR for selected metrics for each cancer (CE10 for lung cancer and
CE10 with 5-year half-life for mesothelioma) and for selected lung cancer model (CE10) with
other mesothelioma models suggested in the literature from Tables 5-51. The selected IUR does
address issues of model uncertainty because a higher risk is only given by the Peto model, but
the Peto model tends to overestimate mortality from mesothelioma in asbestos cohorts with long
follow-up [e.g. Barone-Adesi et al. (2008) and Berry etal. (2012)1.
Age-dependent adjustment factor
As discussed in Section 4.7.1.1, there is no chemical-specific information for LAA, or
general asbestos, that would allow for the computation of a chemical-specific age-dependent
adjustment factor for assessing the risk of exposure that include early-life exposures.
The review of mode-of-action information in this assessment (see Section 4.6.2.2)
concluded that the available information on the mode of action by which LAA causes lung
cancer or mesothelioma is complex and a mode of action is not established at this time. Thus, in
accordance with 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 for substances that act through a mutagenic mode of action is not recommended.
5.4.5.3.1. Comparison with other published studies ofLibby workers cohort. For lung cancer,
two alternative analytic approaches to the use of EPA's extended Cox proportional hazards
models are considered here for the calculation of a unit risk of lung cancer mortality. All of the
choices are based on different analyses of the Libby worker cohort; however, inclusion criteria
5-135
-------
differ among the analyses as does the length of mortality follow-up. Each of the two alternative
approaches has two options to estimate the slope of the exposure-response relationship in place
of the regression slope estimated from the Cox proportional hazards model.
The first approach would be to use the published categorical results based on Sullivan
(2007), which offers two options: (1) estimate a slope to those categorical data or (2) use the
slope estimated in a published reanalysis of categorical data of the Sullivan (2007) cohort by
Berman and Crump (2008). The second approach would be to use the published regression
results of other researchers who modeled the underlying continuous data. There are two options
under this approach: (1) use the slope estimated by Larson et al. (201 Ob) or (2) use the slope
estimated by Moolgavkar et al. (2010).
For comparison purposes, the lung cancer unit risk from these alternatives is computed;
however, as all analyses are based upon different subsets of the Libby workers cohort and used
different analytic methods, the results are not necessarily interchangeable. Table 5-54
summarizes lung cancer risks derived from these studies.
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)"
Berman and
Crumrj (2008)b
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
fibers/cc-yr
xlO3
(calendar yr)
5.8
4.2
1.69
3.96
1.61
Risk based on
upper confidence
limit (UCL) on the
slope
(per fibers/cc)
0.068
0.037
0.011
0.079
0.010
"Reanalysis 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.
5-136
-------
The first alternative analytical approach to estimating the extra risk from a linear
regression of individual mortality data was to use a technique that is standard in EPA cancer risk
assessments (U.S. EPA, 2005a) when individual-level data are not available. This approach used
a weighted linear regression of standardized risk ratio (SRR) estimators for lung cancer mortality
in white males, as calculated in the NIOSH cohort analysis (Sullivan, 2007), with categorical
cumulative exposure and a 15-year lag. The Sullivan (2007) analysis was based only on those
workers who had not died or been lost to follow-up before January 1, 1960 (in contrast to
employment beginning after January 1, 1960), because the NIOSH software program (Life-Table
Analysis System) used for this analysis only has statistics on external comparison rates for
asbestosis (one of the primary outcomes of interest in the Sullivan (2007) analysis) beginning in
1960. The STAR analysis involves internal comparisons of lung cancer mortality rates in the
higher exposure categories to the lung cancer mortality rates in the lowest exposure category.
The weights used for the SRRs were the inverses of the variances. Midpoints of the exposure
intervals were used, and for the unbounded interval, the midpoint was assumed to be twice the
starting point of that interval.
Using this approach, a regression coefficient of 4.2 x 10"3 per fiber/cc-yr (standard error
[SE] = 7.7 x 10"4 per fiber/cc-yr, p = 0.03) was obtained from the weighted linear regression of
the categorical STAR results. Because the data from Sullivan (2007) were already adjusted for the
length of an occupational year (240 days) to the length of a calendar year (365 days), only the
standard adjustment for inhaled air volume was performed. The concentration estimate obtained
using this regression modeling and the life-table analysis procedure was LECoi = 0.272 fiber/cc,
resulting in the lung cancer unit risk of 0.0368 per fiber/cc.
The Berman and Crump (2008) reanalysis was based on the Sullivan (2007) summary
results except the authors used a lag of 10 years [personal communication with Sullivan in 2008
as cited by Berman and Crump (2008)1. They fit the IRIS IUR (U.S. EPA. 1988a) lung cancer
model to aggregate data using an extra multiplicative parameter a. In this model, the relative
risk at zero exposure is a rather than 1 (unity). With a = 1, their model did not fit, and with a
estimated, the fit was satisfactory. Berman and Crump (2008) chose the central estimate of the
slope from the fit with a estimated, but constructed an "informal" 90% confidence interval by the
union of two confidence intervals (this upper bound is shown in Table 5-54). This was done to
address uncertainty in the estimated parameter a, similar to what is done in this current
assessment with estimated lag and decay. Note also that Berman and Crump (2008) provided a
UF to adjust for several sources of uncertainty in exposures, resulting in an upper-bound risk of
0.3162.
The second alternative analytic approach to estimating the extra risk of lung cancer from
a Cox regression with time-dependent covariates of individual mortality data was to use the
results published by Larson et al. (201 Ob), with cumulative exposure and a 20-year lag. This
analysis of lung cancer mortality was based on the full cohort of 1,862 workers, updated until
5-137
-------
2006 and using the same model form as the current EPA analysis (the extended Cox proportional
hazards model). Larson et al. (201 Ob) reported a regression coefficient of
1.06 x lO'3 per fiber/cc-yr (SE = 3.1 x 10'4 per fiber/cc-yr,p = 0.0006).28 EPA assumed that the
cumulative exposures reported by Larson et al. (201 Ob) were based on years of occupational
exposure (240 days per year) during a 365-day calendar year. In order to account for exposure
on every day of the year for a calculation of unit risk, an adjustment for exposures during the
length of an occupational year (240 days) to the length of an calendar year (365 days) and an
adjustment for the volume of inhaled air were performed to match EPA's analyses. The
concentration estimate obtained using the Larson et al. (201 Ob) regression modeling and the
life-table analysis procedure was LECoi = 1.26 fibers/cc, resulting in a lung cancer unit risk of
0.0103 perfiber/cc.
Moolgavkar et al. (2010) also used the Cox proportional hazards model with
time-dependent covariates for analysis of the Sullivan (2007) cohort with a 15-year lag. The
parameter in this study estimates 1.11 x 10"3 per fiber/cc-yr (SE = 2.5 x 10'4 per fiber/cc-yr),
which is very close to the Larson et al. (201 Ob) value, and therefore, the lung cancer unit risk
based on their analysis would be very close to the Larson et al. (201 Ob) value. Comparison with
McDonald et al. (2004) is difficult because in that article, outcome is defined as respiratory
cancer (ICD-9 160-165), which is more expansive than other researchers' definitions of the
outcome as lung cancer, and their subcohort of 406 white men employed before 1963 was
exposed during a time period when exposure assessment was less reliable and more likely to
include significant exposure-measurement error. Nonetheless, the parameter estimate resulting
from the Poisson analysis by McDonald et al. (2004) was 3.6 x 10"3 per fiber/cc-yr.
The differences in the results in Table 5-54 appear to be mostly attributable to the time
periods of analysis and various degrees of exposure measurement error corresponding to these
time periods rather than the analytic approach. EPA based their analyses on the exposures that
occurred after 1959, while the Sullivan (2007), Larson et al. (201 Ob), and Moolgavkar et al.
(2010) analyses were based on the cohort including those hired before 1960, and McDonald et al.
(2004) included only workers hired before 1964. The small discrepancy between observed lung
cancer deaths between this current assessment and Larson et al. (201 Ob), described in
Section 4.1.1.1, is unlikely to play a role in the difference among risk estimates. Moreover, for
the subcohort hired after 1959, all deaths are included in the Larson et al. (201 Ob) lung cancer
counting rules.
As explained in detail in the discussion on uncertainty in the exposure assessment (see
Section 5.4.6), there were only several measurements from the 1950s and one from 1942, and
most of the exposure estimation for the early years of the cohort's experience was based on
28Note 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 * 10~3 per fiber/cc-yr
(SE = 2.48 x lO-3 per fiber/cc-yr,;? = 0.018).
5-138
-------
estimates of the ratio of dust to fibers estimated in the late 1960s and extrapolated backwards in
time for several decades. Moreover, 706 of the workers hired before 1960 (not necessarily short-
term workers) did not have estimated exposure measurements associated with their positions,
leading to a much larger measurement error. These limitations in the underlying exposure
assessment for the years before 1968 likely resulted in exposure measurement error that could
have attenuated the analytic regression results [see the discussion in Lenters etal. (2012) and
Lenters et al. (2011)] on the impact the quality of the exposure information has on estimates of
dose-response; also Bateson and Kopvlev (2014), thereby yielding a smaller effect estimate for
the whole cohort compared to the subcohort hired after 1959.
None of the approaches used by McDonald et al. (2004), Sullivan (2007), or Larson et al.
(201 Ob) could have been appropriately used for the unit risk of mesothelioma because these
approaches are not based on absolute risk metrics of association, which the current assessment
considers to be the relevant metric of association. Berman and Crump (2008) did not evaluate
the risk of mesothelioma. Moolgavkar et al. (2010) used an absolute risk model for
mesothelioma. These results are summarized in Table 5-55. The upper-bound results for the full
cohort presented by Moolgavkar et al. (2010) are about 80% of the U.S. EPAQ988a) estimate of
the mesothelioma slope factor, leading to an approximately 80% estimate of the mesothelioma
unit risk as dependence is linear in the mesothelioma slope factor (see eq 5-11). This is very
close to the current assessment's estimate based on the subcohort, which is also about 80% of the
U.S. EPA(1988a) estimate of mesothelioma risk. Duration of exposure, but neither department
code nor job category, was known for 706 of 991 (71%) workers hired from 1935 to 1959.
Because of that limitation, duration of employment is the best metric for the full cohort, and it
does not support exposure-response estimation.
5-139
-------
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.
(2010)"
Larson etal. (2010b)
Herman and Crump
(2008)"
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 = 0.122
Central = 0.075
No estimates of absolute
risk
Upper Bound -0.13
Central ~ 0.08
No estimates of absolute
risk
No estimates provided
aReanalysis 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,
5-140
-------
6) Uncertainty in control of potential confounding in modeling lung cancer mortality,
7) Uncertainty due to potential effect modification,
8) Uncertainty due to length of follow-up,
9) Uncertainty in use of life-tables to calculate cancer mortality inhalation unit risks,
10) Uncertainty in combining of risks to derive a composite cancer inhalation unit risk
(IUR), and
11) Uncertainty in extrapolation of findings in adults to children.
5.4.6.1.1. Uncertainty in low-dose extrapolation. A common source of uncertainty in
quantitative cancer risk assessments generally derives from extrapolating from high doses in
animals to low doses in humans. Compared to assessments based on animal data, the uncertainty
from low-dose extrapolation in this assessment, which uses occupational epidemiology data, is
considered to be lower for the following reasons. The NIOSH worker cohort developed by
Sullivan (2007) includes 410 workers employed less than 1 year among the 880 workers hired on
or after January 1, 1960. Although short-term workers on average experience a mean exposure
intensity per day worked greater than workers employed more than a year (Sullivan, 2007), the
cohort nevertheless includes many short-term workers with relatively low cumulative
occupational exposures. Further, inclusion of salaried workers in the NIOSH cohort (Sullivan,
2007) adds many workers with lower workplace exposure. Thus, while occupational exposure
concentrations may be generally higher than typical ongoing environmental concentrations, the
low-dose exposures in this occupational database may be more representative of nonoccupational
exposures.
While many occupational epidemiology studies are based on relatively high exposure
levels that are beyond the range of common environmental exposures, many in the Libby worker
cohort experienced exposures that were near or below the PODs derived from the life-table
analysis (i.e., the estimated PODs are in the range of the observed data). The POD for the
selected lung cancer mortality exposure metric was 0.191 fiber/cc. The POD for the selected
mesothelioma mortality exposure metric was 0.106 fiber/cc. Among the workers hired after
1959 who had at least 1 year of occupational exposure (n = 470; 20 lung cancer deaths), there
were 19 (4%) with average occupational exposure concentrations of less than 0.3 fiber/cc,
including one lung cancer death (5%).
Although data might have been modeled down to a very low cumulative exposure level,
the Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a) recommends defining a POD
for low-dose extrapolation to increase the stability of the IUR estimate at lower exposures where
fewer cancers might be expected. Thus, the uncertainty associated with low-dose extrapolation
is somewhat mitigated because the linear extrapolation from the dose associated with the POD
5-141
-------
from the life-table analyses of each cancer endpoint was encompassed within the observed data
range. Nonetheless, some uncertainty remains in the extrapolation from occupational exposures
to lower environmental exposures when using a POD.
5.4.6.1.2. Uncertainty in exposure assessment. Accurate exposure assessment is generally
considered to be a major challenge for occupational epidemiologic studies and is a challenge
well recognized by the NIOSH investigators (Amandus et al., 1987b). As stated previously in
Section 5.4.3.3, while it is generally true that the use of more data is an advantage in statistical
analyses because it allows for the computation of more statistically precise effect estimates, this
advantage in precision may be offset by a negative impact on the accuracy of the effect estimate
if an increase in sample size is accompanied by greater exposure misclassification or other
biases. In this case, EPA decided to base this LAA-specific human health risk assessment upon
the mortality experience of workers hired on or after January 1, 1960. EPA's use of the
subcohort analysis is based on the belief that it is important to accurately estimate the true
underlying exposure-response relationships by relying on the most accurate exposure data. The
use of this subcohort greatly reduces the uncertainty in exposure error compared to evaluations
based on the full cohort. More specifically:
a) Job category and department codes were completely unknown for 706 of the
991 workers' jobs from 1935 to 1959 (71% of the cohort for this time period). These
workers had the same estimated exposure concentration (66.5 fibers/cc) for all years
without this information. Examination of the post-1959 cohort removes this
significant source of exposure misclassification (only 9 of 880 subcohort workers did
not have department code and job category information).
b) Using the more recently hired cohort minimizes the uncertainty in estimated worker
exposures based on the JEM, which was informed by air sampling data available in
1956 and later years. Although uncertainties still exist in the task-specific exposure
estimates from 1960-1967, uncertainty in the assessment of earlier exposure levels is
considerably greater.
c) Exposure measurements were collected from the area samples and represented
exposures for all the workers with the same job code. Statistically, this causes the
Berkson measurement error effect, which is described later in this section.
As EPA exposure-response modeling for mesothelioma and lung cancer mortality is
based on the post-1959 subcohort, the remaining discussion of uncertainty in exposure
measurement will address these data.
5.4.6.1.2.1. Sources of uncertainty in job history information. Worker exposures for EPA
exposure-response modeling were calculated based on job histories and the JEM from 1960
through 1982 (see Figure 5-5). Overall, there is little uncertainty in the job history information.
5-142
-------
Regarding exposure estimation for the occupational cohort, the NIOSH investigators (Amandus
et al., 1987b) conducted a detailed retrospective exposure assessment to estimate the individual
worker exposures. NIOSH used extensive occupational exposure data to construct the
time-specific JEM, spanning decades (Amandus et al., 1987b). These data were reabstracted
from the workers' employment records for quality assurance (Sullivan, 2007). NIOSH records
on work histories and job-specific exposure extended from the 1930s through May 1982.
However, the vermiculite mining and milling operation continued on for several years, and some
workers were retained through 1993 for plant close-out activities. Only 148 members of the
post-1959 cohort (n = 880) were employed as of the May 1982 employment records when the
cohort was enumerated by NIOSH (Sullivan, 2007). Because exposure concentrations in 1982
(see Table 5-21) were generally below 1 fiber/cc with only two locations having concentrations
of 1.2 fibers/cc, it is unlikely that these workers' exposures were significantly underestimated.
Sources of uncertainty in exposure intensity for the identified location operations
The available exposure data that inform the JEM include over 4,000 air samples, the
majority of which were collected after 1967 (see Table 4-1). All of the job location exposure
estimates (see Table 5-21) from 1968-1982 were directly informed from air samples collected
on membrane filters and analyzed for fibers by PCM. The availability of site-and task-specific
air samples for these years provides a good basis for the exposure estimates. However, some
uncertainties exist in estimating asbestos exposures using air samples analyzed by PCM.
1) PCM analysis does not determine the mineral or chemical make-up of the fiber: The
PCM method defines and counts fibers based on the size (length, width, and aspect
ratio) of the particle without regard for the material that makes up the particle being
viewed. The PCM method was developed for use in occupational environments
where asbestos was present, and the nature of the fibers should be further evaluated to
confirm the fibers viewed under PCM are asbestos. McGill University researchers
evaluated the fibers collected on membrane filters in the early 1980s and confirmed
the presence of asbestos fibers in the tremolite-actinolite solution series consistent
with the LAA (McDonald et al., 1986a). NIOSH researchers confirmed the presence
of tremolite asbestos in bulk dust samples but not in air samples from the facility
(Amandus et al., 1987b). Although less specific to fibers, 60-80% of the airborne
dust in the mills in 1968 was tremolite, further supporting the presence of asbestos in
the air (based on State of Montana air sampling, and x-ray diffraction analysis by the
Public Health Service [correspondence, October 17, 1968]). However, although the
presence of mineral fibers in the actinolite-tremolite series was confirmed in the work
environment, it is possible that fibers were also counted by PCM from other materials
(such as textiles from clothes and packaging materials). Therefore, it is unknown
from these data what proportion of the counted PCM fibers were mineralogically
asbestos and what proportion were other materials present in the air workplace.
2) PCM defines fibers as particles with an aspect ratio of 3:1 or greater. There is an
ongoing debate in the literature on asbestos toxicity regarding the influence of aspect
5-143
-------
ratio on relative toxicity. Specifically, in mining environments, it has been speculated
that a larger proportion of low aspect ratio fibers from mineral dusts may significantly
impact the apparent cancer potency of the measured PCM fibers in those
environments (Berman, 2010; U.S. EPA, 1988a). Few data are available to
understand fiber morphology and fiber aspect ratios in the Libby cohort working
environment. Considering the post-1959 cohort, PCM fiber size distribution and
aspect ratio data only exist for a set of eight air samples (599 fibers) collected from
the wet mill and screening operations and analyzed by the NIOSH researchers
(Amandus et al., 1987b). For these air samples, over 96% of the fibers viewed by
PCM had an aspect ratio greater than 10:1 [see Table 4-2 (Amandus etal., 1987b)1.29-
However, because these samples were provided by the company in the early 1980s,
they do not represent conditions in the old wet mill or dry mill operations, which were
significantly dustier environments (Amandus et al., 1987b). It is possible that prior to
IH modifications in 1974, the dry and old wet mills generated proportionally more
mineral dusts than the screening plant and new wet mill operations after IH
modifications. No data are available for the mining environment, which would also
be expected to generate a range of mineral dusts. Therefore, there is a significant
uncertainty about the size and aspect ratio of fibers included in PCM fiber counts for
the majority of the post-1960 workers cohort, but it is not possible to judge the
direction or magnitude of such uncertainty.
3) The resolution of visible PCM fibers: 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 |im. Historical PCM analysis (1960s and early 1970s)
generally had less resolution, and fibers with minimum widths of 0.4 or 0.44 jim were
considered visible by PCM (Amandus et al., 1987b: Rendall and Skikne, 1980).
McDonald et al. (1986a) compared fibers viewed by PCM and TEM and estimated
that approximately one-third of the total fibers could be viewed by PCM. Because
38% of the fibers were <5 jim in length, this implies approximately 30% were not
viewable by optical microscopy for other reasons, such as width. However, it is
unknown what proportion of that 30% would be viewed with the minimum width
resolution of 0.25 |im for later optical microscopy. It is likely that early PCM counts
were underestimated relative to the later data for the cohort but by less than a factor
of 2.
Before 1968, no air sampling data were available for 23 of the 25 job location operations
(see Table 4-2), and the exposure estimates were extrapolated from later air sampling data.
Amandus et al. (1987b) recognized there was significant uncertainty in the extrapolation of
available air sampling data to previous time periods. The researchers considered major changes
in operations and interviewed employees in the early 1980s regarding previous years of
operation. The assumptions used to make these extrapolations are clearly stated for each of the
plant operations. For four operations, high and low estimates of pre-1968 exposures were
provided based on different sets of exposure assumptions (see Table 5-21). For ore loading,
29Although 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.
5-144
-------
there were negligible differences in the exposure estimates for the period from 1960-1967 (10.7
versus 9 fibers/cc). For drilling, the river dock, and the bagging plant, there were 3.4-, 2.6-, and
2.8-fold differences, respectively, between the high and low estimates of exposure between 1960
and 1968.
Dry mill exposures between 1960 and 1968 were informed by air sampling for total dust
collected in the dry mill facility from 1956-1969 (where total dust was collected by midget
impingers). Amandus et al. (1987b) derived a conversion factor of 4.0 fibers/cc per mppcf to
apply to the two location operations in the dry mill during these years. A range of conversion
factors was considered for the dry mill depending on how the dust and fiber air samples (PCM)
were grouped and averaged (1.2 to 11.5 fibers/cc per mppcf). A subset of dust and fiber samples
available over the same time period (1967-1968) resulted in a ratio of 8.0 fibers/cc per mppcf.
In contrast, a ratio of 1.9 fibers/cc resulted when total dust samples from 1969 were compared
with fiber samples from 1970. However, both of these subsets had limited numbers of samples
available. Therefore, the conversion factor of 4.0 fibers/cc per mppcf was selected based on
using the maximum samples available over a time period when the dry mill exposures were
considered similar: dust samples (1965-1969) and fiber samples (1967-1971).
5.4.6.1.2.2. Sources of uncertainty in the calculation of the job-exposure matrix (JEM). The
exposures in the JEM (see Figure 5-5) were calculated from the exposure intensities of the
various task-specific exposure intensities shown by job location operation (see Table 5-21). The
uncertainties in the exposure intensity for the job location operations will impact the JEM.
Additionally, for each of the job categories in the JEM, NIOSH researchers defined which tasks
(job location operations) were conducted and for what proportion of the work day. A TWA
exposure for each job category across time was calculated based upon these assumptions and the
task-specific exposure estimates. There is a measure of uncertainty in these assumptions for
each job category. Additionally, there is interindividual variation within the job categories.
These uncertainties are common to exposure reconstruction for epidemiological cohorts.
5.4.6.1.2.3. Uncertainty in the exposure metric. The PCM measurement is the available
exposure metric for analysis of the Libby worker cohort at this time. Currently, there is no
optimal choice of the best dose metric for asbestos, in general, and for LAA, in particular.
Uncertainties related to PCM analytical method are discussed in Section 2. Briefly, PCM cannot
distinguish between asbestos and nonasbestos material or differentiate among specific types of
asbestos. Further, due to limitations of this methodology, PCM does not take into account fibers
shorter than 5 jim in length.
5.4.6.1.2.4. Evaluation of the effects of uncertainties in exposure measurement. An
understanding of the effects of exposure measurement error on the risks estimated from
5-145
-------
epidemiologic analyses is important to place these possible exposure measurement errors in
context. The effect of exposure measurement error on estimates of the risk of mesothelioma or
lung cancer mortality attributable to exposure depends upon the degree to which that error may
be related to the likelihood of the outcome of interest (mesothelioma or lung cancer mortality).
Exposure measurement error that is similar among workers who died of lung cancer, and those
who did not die of lung cancer, is termed nondifferential exposure measurement error. Exposure
measurement error that is associated with the outcome (error that is differential with respect to
disease status) can cause bias in an effect estimate towards or away from the null, while
nondifferential exposure error typically results in bias towards the null (Rothman and Greenland,
1998). From the above evaluation of uncertainties, there is no indication that the uncertainties in
job history information, exposure estimates for specific tasks, or calculation of the JEM would be
differential based on the cancer health outcome data. Therefore, these uncertainties are
considered unrelated to disease status and the general result is likely to be an attenuation in risk
estimates towards the null (i.e., the addition of random noise to a clear signal tends to reduce the
clarity of the observed signal, and the avoidance of random noise results in a stronger observed
signal).
Generally speaking, if the exposure concentrations estimated by NIOSH were
systematically too high, then the associated risks of exposure estimated in the regression analysis
would be low because the same actual risk would be spread across a larger magnitude of
exposure. Similarly, if the exposure concentrations estimated by NIOSH were systematically too
low, then the associated risks of exposure estimated in the regression analysis would be too high.
From the above evaluation, the majority of the sources of uncertainty are not systematic. There
are a few areas of uncertainty that may be classified as biased:
1) High- and low-exposure estimates for four job location operations were provided
between 1960 and 1967. Amandus et al. (1987b) chose the high estimates of
exposure for these job location operations when calculating the JEM. Therefore,
there will be a bias towards the high end for the job categories informed by these
data. There was a 1.1- to 3.4-fold difference between the high and low estimates.
This difference will be less pronounced where these exposure concentrations are
averaged with other job location operations in the JEM, and across multiple jobs (as
was the case for the majority of the workers; see Figure 5-5).
2) Current PCM analysis would count more fibers relative to early PCM methods based
on minimum fiber width resolution. For example, Amandus et al. (1987b) used a
minimum width cutoff of 0.44 |im in their review of PCM fibers in the 1980s, which
may have resulted in as much as a twofold underestimate compared to current PCM
methods with a width resolution of 0.25 jim or less. Additionally, as PCM
methodology has developed over time, it is unknown when PCM results from
company records would be considered relatively standard to a minimum width
resolution between 0.2 and 0.25 jim. Also, prior to standardization of PCM to
0.25-|im minimum width, there was interlaboratory variability as well. Therefore, the
5-146
-------
size distribution of PCM fibers (e.g., minimum width) reported in the JEM may have
changed over time. Although theoretically a systematic bias, given the years for
which PCM data are available, this is likely an insignificant effect.
3) Asbestos was a contaminant of vermiculite that was the primary object of production.
Mine, old dry mill, and wet mill ambient air may have contained material other than
asbestos that could have contributed to PCM fiber count. The exposures in the old
dry and wet mills and mine location may have included a greater proportion of dust to
fibers than tasks using the ore and refined vermiculite after the new wet mill became
operational. It is possible there is a systematic overcount of fibers in the dusty
environment due to interference from nonmineral fragments. This likely affects the
exposure intensity for 23 of 25 job location operations within the mine and old dry
mill. Estimated exposures from job categories that include these operations may be
biased upwards.
Nondifferential measurement error in a continuous exposure can be of the classical or
Berkson type and typically arises in environmental and occupational settings as a mixture of the
two forms (Zeger et al., 2000). Classical measurement error occurs when true exposures are
measured with additive error (Carroll et al., 2006) and the average of many replicate
measurements, conditional on the true value, equals the true exposure (Armstrong, 1998). This
error is statistically independent of the true exposure being measured and attenuates true linear
effects of exposure, resulting in effect estimates in epidemiologic studies that are biased towards
the null (Heid et al., 2004; Zeger et al., 2000; Armstrong, 1998). Such errors occur, for example,
when the mean values of multiple local air samples are used.
Berkson measurement error is independent of the surrogate measure of exposure (Heid et
al., 2004; Berkson, 1950) and is present when the average of individuals' true exposures,
conditional on the assigned measurement, equals the assigned measurement. Berkson
measurement error can arise from the use of local area mean sampled exposures to represent the
individual exposures of people in that area—even when the estimated area mean is equal to the
true underlying mean (i.e., no classical measurement error). Examples of random variability in
personal behavior that may produce Berkson measurement error in personal exposure estimates
include the volume of air breathed per day among the workers and the effectiveness of an
individual's nasal filtration at removing contaminants. In general, Berkson measurement error is
not thought to bias effect estimates but rather increases their standard errors (Zeger et al., 2000).
However, some epidemiologic studies have suggested that Berkson measurement error can
produce a quantitatively small bias towards the null in some analyses (Bateson and Wright,
2010; Kim et al., 2006: Reeves etal., 1998: Burr, 1988). Uncertainties in the levels and time
course of asbestos exposure for the Libby workers also adds uncertainty in evaluating the relative
fit of different exposure metrics.
5-147
-------
5.4.6.1.2.5. Exposure to other kinds of asbestos and residential exposure. Another source of
uncertainty in the estimation of exposures in the Libby workers cohort is the potential
contribution of nonoccupational or residential exposures as well as exposures to other kinds of
asbestos in employment before or after working in Libby.
Many of the workers resided in Libby, MT before and/or after their employment at the
mining and milling facilities ended. The vermiculite from the mine had been used at numerous
sites around the town including baseball fields around the expansion plant, high- and
middle-school tracks, attic and wall insulation in homes, and as a soil amendment in gardens
(U.S. EPA, 2010a, 2001). Exposure to asbestos could have occurred among individuals outside
of the workplace, particularly through activities with the potential of stirring up soil or other
materials that had been mixed with the vermiculite (Weis, 2001). The results of community
sampling indicated that even 10 years after mill operations ceased, asbestos fiber concentrations
in the air could exceed OSHA standards established for the protection of workers during some
activities (Weis. 2001).
Therefore, the workers' actual personal exposures as the sum of occupational and
nonoccupational exposures are likely to have been underestimated by the use of estimated
Libby-related occupational exposure alone. The difficulty stems from the lack of data on
residential exposures and lack of information on pre- and postemployment residence of the
Libby workers. Nonoccupational exposures were likely to have been smaller in magnitude than
the occupational exposures, but workers may have lived in and around Libby, MT for many
more years than they were exposed occupationally. The effect of residential exposure could be
more prominent for workers with lower occupational exposure who resided in Libby for a long
time. Whitehouse et al. (2008) has reported several cases of mesothelioma among residents of
the Libby, MT region who were not occupationally exposed. However, because the report by
Whitehouse et al. (2008) details only the cases and does not define or enumerate the population
from which those cases were derived, computed relative risks from nonoccupational exposures
were not available. ATSDR (2000) reported higher relative risks of mesothelioma among the
population of Libby, MT, including former workers residing in Libby, but did not provide
relative risk for nonoccupational exposure. Instead, the ATSDR (2000) report on mortality
grouped cases among the former workers with nonoccupationally exposed cases. Therefore, it is
not clear what the magnitude of the contribution of workers' nonoccupational exposures was to
their overall risk.
Some of the occupational workers with lower exposures, such as short-term workers, may
have either been high school or college students working during the summer or may have been
transient workers who may not have stayed for a long time in Libby. It is interesting to note that
the lung cancer rates by age at first exposure show very low rates for those first exposed before
age 25 (see Table 5-42). Sullivan (2007) analyzed differences between short- and long-term
workers and reported little difference among the groups except for age at hire. As the short-term
5-148
-------
workers were younger on average, this supported the suggestions that some of the short-term
workers may have been college students working during the summer. This population of
short-term workers is not well defined; however, while it is possible that short-term transient
workers could potentially have been exposed to other kinds of asbestos or other lung carcinogens
in their non-Libby occupational career, which might have affected their pre- and post-Libby risk
profile for asbestos exposure, lung cancer rates for those with less than 1 year of exposure are in
line with those with less than 5 years of exposure (see Table 5-27). While their occupational
histories other than working in Libby are unknown, it is very unlikely that they include
exposures of the magnitude that were encountered in the Libby mine and mill. The impact of
these uncertainties on regression slopes is difficult to evaluate. However, the slope may be
somewhat underestimated, as an observed increase in risk would be attributed to a larger
exposure differential than might have been present due to the addition of nonoccupational
exposures. There will also be a downward bias from random exposure measurement error with
lower occupational exposure affected disproportionately; however, the magnitude of this bias
would be expected to be small.
5.4.6.1.2.6. Conclusion regarding uncertainty in exposure assessment. Overall, there are
likely to be multiple sources of uncertainty attributable to exposure measurement error. It is
possible that systematic error may have been introduced into the exposure intensities assigned to
several of the job location operations discussed above. In each case, these errors in estimating
exposures were overestimates, which in general, might lead to underestimations of risk for lung
cancer, but the results are unclear for the risk of mesothelioma. The magnitude of the potential
overestimates of drilling and dry and old wet mill exposures is uncertain. The dust-to-fiber
conversion ratio applied to the dry mill during 1960-1967 could be an over- or underestimate by
as much as twofold, as Amandus et al. (1987b) derived a conversion factor of 4.0 fibers/cc per
mppcf, but subsequent samples available during 1967-1968 resulted in a ratio of 8.0 fibers/cc
per mppcf, while samples from 1970 yielded a ratio of 1.9 fibers/cc. Random error in the
measurement of dust or fibers would likely have produced an underestimation of risk. There is
no known bias in the assumptions to extrapolate exposure to pre-1968 location operations
outside of the dry mill, and random bias would also likely have produced an underestimation of
risk.
5.4.6.1.3. Uncertainty in model form. For mesothelioma mortality, the Poisson model is
commonly used for rare outcomes and has been applied by McDonald et al. (2004) to model
mesothelioma risk in the Libby worker cohort. For lung cancer mortality, the Cox proportional
hazards model is a well-established method that is commonly used in cohort studies, including
by Larson et al. (201 Ob) and Moolgavkar et al. (2010) for the Libby worker cohort, because this
type of survival analysis takes into account differences in follow-up time among the cohort.
5-149
-------
Larson et al. (201 Ob) conducted Poisson analyses and reported that their lung cancer results
using this different model form were similar to those from their extended Cox proportional
hazards models, but those results were not shown.
Both of these model forms allow for the evaluation and control of important potential
confounding factors such as age, gender, and race, and for the modeling of exposure as a
continuous variable. Both model forms yielded exposure-response results with good fit to the
occupational exposure data. The default assumption of the extended Cox proportional hazards
model as well as the Poisson model is that all censoring (due to death or loss to follow-up) is
assumed to be independent of exposure to the LAA. However, exposure to LAA may cause
death from other causes such as asbestosis or nonmalignant respiratory disease (Larson et al.,
201 Ob), which is referred to as dependent censoring. The concern is that the observation of lung
cancer mortality may be precluded by mortality from other causes.
In the cohort of 880 workers hired after 1959, 32 died of lung cancer, while 10 died of
asbestosis, and 21 died of nonmalignant respiratory disease. The mean length of follow-up from
the date of hire until death for the workers who died of lung cancer was 24.9 years. However,
the mean length of follow-up for the workers who died of asbestosis or nonmalignant respiratory
disease was 30.4 years, so it does not appear that early deaths from other causes associated with
exposure to the LAA (Larson et al., 201 Ob) would have precluded many cases of lung cancer.
This implies that any potential bias in the lung cancer risk estimates due to dependent competing
risks is small.
With respect to mesothelioma mortality, note that the exposure-response modeling is
limited by the number of deaths. However, dependent censoring, as described above, is not
accounted for in the Poisson model and likely causes a downward bias in the estimation of risk.
The mean length of follow-up for the workers who died of mesothelioma was 30.1 years, and
there is some evidence that early deaths from other exposure-related causes precluded an
individual's risk of death from mesothelioma; only lung cancer exhibited a shorter average
follow-up time compared to mesothelioma, and in 419 cases of mesothelioma, mesothelioma and
lung cancer were never coidentified (Roggli and Vollmer, 2008).
5.4.6.1.4. Uncertainty in selection of exposure metric. There is uncertainty about what metric
should be used for modeling exposure to LAA. This current assessment evaluated models
proposed in the asbestos literature for modeling mesothelioma and models that include unlagged
and lagged cumulative exposure with and without a half-life of various lengths, and RTW
exposure with and without a half-life. In the analysis of comparative model fit based on the
empirical data, lagged cumulative exposure with a half-life provided the best fits for both
mesothelioma and lung cancer mortality associated with LAA. However, evaluation of 20-year
lag and longer lag times for mesothelioma was not possible, as the earliest mesothelioma death
happened less than 20 years from the start of the exposure; hence, exposure was zeroed out, and
5-150
-------
the fit of any model with 20-year lag was very poor. Latency time for mesothelioma may be as
long as 60-70 years [e.g., Bianchi and Bianchi (2009)], so the precise lag time is uncertain.
In evaluating the data on lung fiber burden, Berry et al. (2009) estimated the range of the
half-life for crocidolite to be between 5 and 10 years. That range is consistent with the finding of
a 5- to 10-year half-life with 10-15 years lag that provided the best fit to the Libby workers
cohort mesothelioma mortality data. Similarly, recent 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). The marginally best fit for the Libby
workers cohort lung cancer mortality data was for CE models with a 5- to 20-year half-life and
10-year lag. However, the precise lag and half-life times are somewhat uncertain. Sensitivity
analysis that excluded people with high exposure during 1960-1963 (see Section 5.4.3.6.4)
provides further evidence that distinguishing between various lags and decays may be difficult
with these data. A limitation of this sensitivity analysis is the decrease in the number of cases,
especially for mesothelioma. Resolving this uncertainty would require longer follow-up time,
which would allow for a subcohort analysis of workers hired in 1967 or afterwards (when
exposure estimates began to be based on PCM measurements) until a sufficient number of cases
would be available for additional analysis.
These simulated decay models were derived mathematically to approximate underlying
biological processes that are not well understood, and are based on maximizing the likelihood for
the workers cohort and may not necessarily apply to the environmental exposure patterns.
Nonetheless, while the mode of action for carcinogenicity is unknown, the models incorporating
a half-life in the exposure metric were clearly preferable for mesothelioma mortality, and the
goal of the regression modeling effort was to identify the best fitting exposure model for the
Libby worker cohort.
Table 5-53 illustrates uncertainty in the IUR due to exposure metric selection. The
quantitative uncertainty is about threefold.
5.4.6.1.5. Uncertainty in assessing of mortality corresponding to other cancer endpoints.
There is evidence that other cancer endpoints may also be associated with exposure to the
commercial forms of asbestos. IARC concluded that there was sufficient evidence in humans
that commercial asbestos (chrysotile, crocidolite, amosite, tremolite, actinolite, and
anthophyllite) was causally associated with lung cancer and mesothelioma, as well as cancer of
the larynx and the ovary (Straif et al., 2009). Among the entire Libby workers cohort, only
two deaths were found to be due to laryngeal cancer, and there were no deaths from ovarian
cancer among the 24 deaths of 84 female workers. The lack of sufficient number of workers to
estimate risk of ovarian cancer is an uncertainty in an overall cancer health assessment.
The remaining uncertainties attributed to assessing mortality corresponding to the cancer
endpoints are considered to be low.
5-151
-------
5.4.6.1.6. Uncertainty in control of potential confounding in modeling lung cancer mortality.
It is well known that smoking is a strong independent risk factor for lung cancer and may have a
synergistic effect with asbestos exposure (Wraith and Mengersen, 2007). In contrast, smoking is
not considered a risk factor for mesothelioma (Selikoff and Lee, 1978; Anderson et al., 1976).
As an important potential confounder of the lung cancer mortality analysis, the possible
effect of smoking on the estimated risk of lung cancer mortality associated with exposure to
LAA needs to be evaluated to the fullest extent possible. This consideration was discussed in
Amandus and Wheeler (1987) and in Section 4.1.1.3.
Additionally, W.R. Grace and Co. instituted a smoking ban on the property in 1979
(Peacock, 2003), which may have affected smoking habits, reporting of smoking habits, or both.
About 30% of the subcohort was still employed in 1979 and all of the post-1959 cohort had been
terminated by May 1982, so the effect of a workplace smoking ban on cohort smoking history
may explain the higher proportion of former smokers in the Amandus and Wheeler (1987) data.
Lung cancer risks in ex-smokers decrease over time compared to lung cancer risks in continued
smokers. A reduction of smoking in the Libby worker population may lead to fewer
observations of lung cancer deaths in later years of the cohort study than would have occurred in
the absence of the smoking restrictions. Changes in smoking behavior during the course of the
epidemiological observation period would lead to changes in the observed time course of lung
cancer death rates. This issue is related to potential effect modification of lung cancer mortality
described in Section 5.4.6.1.7.
Without high-quality individual-level data on smoking that could be used to control for
potential confounding, it is still possible to comment upon the likelihood and potential magnitude
of confounding and the impact any confounding would be expected to have on the lung cancer
mortality risk estimates. Confounding can be controlled for in a number of ways including by
modeling and by restriction. Restriction of the study population can reduce any potential
confounding by making the resulting population more similar. For instance, there can be no
confounding by gender when a study population is restricted to only men. This assessment
restricted the study population to those workers hired after 1959. Smoking habits have changed
over time, and it can reasonably be assumed that the range of smoking habits among those hired
after 1959 is less variable than that among the whole cohort, particularly because of the narrower
range of birth cohorts represented in this subcohort. This should have the effect of reducing
some of the potential for confounding. Analytic examinations of potential confounding are
discussed below.
Richardson (2010) describes a method to determine if an identified exposure relationship
with lung cancer is confounded by unmeasured smoking in an occupational cohort study. EPA
implemented this methodology to model the potential effects of LAA on the risk of COPD
mortality on the subcohort of workers hired after 1959 (see Section 5.4.3.8). Summarizing these
findings, EPA used the method described by Richardson (2010) to evaluate whether exposures to
5-152
-------
LAA predicted mortality from COPD as an indication of potential confounding by smoking and
found a statistically nonsignificant negative relationship, which was inconsistent with
confounding by smoking.
5.4.6.1.7. Uncertainty due to potential effect modification. Among the 32 deaths from lung
cancer in workers hired after 1959 that were used to estimate the unit risk of lung cancer
mortality (see Section 5.4.5.2), data on smoking listed 16 as smokers, 4 as former smokers, and
12 of the 32 had missing data. Thus, data to support an estimate of the risk of LAA among
known nonsmokers were not available.
It is theoretically possible that the risk of lung cancer mortality estimated in this current
assessment is a reflection of a positive synergy between smoking and asbestos, and that the
adverse effect of LAA among the potentially nonsmoking workers has been overestimated. The
unit risk of the lung cancer estimate herein and the combined mesothelioma and lung cancer
mortality IUR would then be health protective for any population that had a lower prevalence of
smoking than that of the Libby worker cohort. However, if the smoking ban did diminish the
effect of smoking, then any overestimation would be somewhat mitigated.
5.4.6.1.8. Uncertainty due to length of follow-up. There is some potential uncertainty regarding
the length of follow-up for cancer mortality, even more so with the restriction of the cohort to
those workers hired after 1959. The hire dates among this subset of the cohort ranged from
January 1960 to November 1981 (the mean date of hire was May 1971). Follow-up continued
until the date of death or December 31, 2006, whichever occurred first. Therefore, the range of
follow-up was from 25 to 46 years, with a mean of more than 35 years.
However, for mesothelioma mortality, the length of the latency period is considerably
longer. Suzuki and Yuen (2001) reviewed 1,517 mesothelioma cases from 1975 through 2000
and was able to estimate the latency for 800. Suzuki and Yuen (2001) reported 17% of cases had
a latency of less than 30 years with 52% of cases with a latency of less than 40 years. Bianchi
and Bianchi (2009) estimated the mesothelioma latency in 552 cases and reported mean latency
periods of 35 years among insulators, 46 years among various industries, and 49 years among
shipyard workers.
According to the results of Suzuki and Yuen (2001) and of Bianchi and Bianchi (2009) a
mean length of follow-up of 35 years may only have captured half of all eventual mesothelioma
mortality cases among the Libby workers hired after 1959. If this were so, then the unit risk of
mesothelioma mortality could be larger than was estimated from existing data, suggesting
continued examination.
5-153
-------
5.4.6.1.9. Uncertainty in use of life-tables to calculate cancer mortality inhalation unit risk
(IUR). The life-table procedure computes the extra risk of death from birth up to 85 years of
age, in part, because this is how national cancer incidence and mortality rate data that are one
basis of the life-tables are made available [see SEER (2010)1, Table 15.10, age-specific U.S.
death rates). Because the prevalence of cancer mortality is a function of increasing age, this
cut-off at age 85 ignores a small additional risk of lung cancer mortality among a small
percentage of people who have the higher background risk. This has the effect of slightly
underestimating the IUR that would be derived if the life-table were extended for an additional
period of time, accounting for longer life spans. Extension of the life-table analysis to people
over the age of 85 requires an additional assumption. Assuming that having attained the age of
85 years, the additional life expectancy is 5 years, then the lung cancer mortality unit risk based
on the LECoi would be somewhat larger—on the order of 5-10%—slightly more than the
additional mesothelioma mortality risk if the life-tables were extended.
5.4.6.1.10. Uncertainty in combining of risk for composite cancer inhalation unit risk (IUR).
For the purpose of combining risks, it is assumed that the unit risks of mesothelioma and lung
cancer mortality are normally distributed. Because risks were derived from a large
epidemiological cohort, this is a reasonable assumption supported by the statistical theory. EPA
conducted a bounding analysis and showed that the related uncertainty is very low.
5.4.6.1.11. Uncertainty in extrapolation of findings in adults to children. The analysis of lung
cancer mortality specifically tested the assumption that the relative risk of exposure is
independent of age within the observed age range of the occupational subcohort hired after 1959
and did not find evidence of age dependence, although such a dependence among a working-age
study population has been reported in another asbestos-exposed cohort (Richardson, 2009).
However, note that no comparable data are available to estimate the lifetime risk from early life
exposures. Note that default age-dependent adjustment factors (ADAFs) are not recommended
because a mutagenic MOA was not identified.
5.4.6.2. Summary
Section 5.4.6.1 details the several sources of uncertainty in the assessment of the cancer
exposure-response relationships and the use of those data to derive the inhalation unit risk. The
text that follows summarizes the primary sources of uncertainty and, where possible, the
expected direction of effect on the exposure-response risk estimates and the inhalation units risk.
1) Uncertainty in low-dose extrapolation (see Section 5.4.6.1.1)
• There remains some uncertainty in the extrapolation of risks based on
occupational exposure to environmental exposure levels but this uncertainty is
5-154
-------
considered to be low as the lower range of occupational exposure overlaps with
expected environmental exposure levels.
2) Uncertainty in exposure assessment, including analytical measurements uncertainty
(see Section 5.4.6.1.2)
• The JEM was based on the "high" exposure estimate for each job according to
Amandus et al. (1987b) and to some extent this could be an overestimate of
exposure. The associated cancer risk would be somewhat underestimated
resulting in a somewhat underestimated IUR.
• The JEM was largely based on estimated fiber concentration using PCM
measurement (with some extrapolations in time), and because PCM may count all
long and thin objects as fibers, these measurement could overestimate the true
LAA fiber concentrations leading to an overestimate of exposure and a somewhat
underestimated cancer risk resulting in a somewhat underestimated IUR.
• PCM measurements in the era of NIOSH measurements in Libby used a lower
resolution, and therefore, included only somewhat thicker fibers thereby counting
fewer fibers than would have been counted by later PCM standards. These earlier
measurement could underestimate the true LAA fiber concentrations leading to an
underestimate of exposure and an overestimate of cancer risk resulting in a
somewhat overestimated IUR.
• The PCM measurement is the available exposure metric for analysis of the Libby
worker cohort at the time of this assessment. Currently, there is no optimal choice
of the best dose metric for asbestos, in general and in particular, for LAA.
Uncertainties related to PCM analytical method are discussed in Section 2 and
such uncertainties cannot be related to the IUR at the time of this assessment.
• Random measurement error in the assignment of exposures could have the effect
of underestimating the risk of lung cancer mortality as that measure of risk is
based on a relative measure. The effect would be to somewhat underestimate the
risk of lung cancer resulting in a somewhat underestimated unit risk for lung
cancer. It is unclear what the impact of such measurement error would be on the
absolute risk of mesothelioma.
• Exposure to other kinds of asbestos and residential exposure to LAA may have
caused workers' actual personal exposures (as the sum of occupational and
nonoccupational exposures) to have been underestimated by the use of estimated
Libby occupational exposure information alone. This could underestimate the
true LAA fiber exposures leading to an overestimate of the associated cancer risk
resulting in a somewhat overestimated IUR.
3) Uncertainty in model form (see Section 5.6.4.1.3)
• For mesothelioma, the Poisson model is the standard epidemiologic form and it is
considered to be the most appropriate model form for rare health outcomes;
therefore, uncertainty is considered to be low. For lung cancer mortality, the Cox
proportional hazards model is the standard epidemiologic form. It is considered
5-155
-------
to be the most appropriate model form for health outcomes with time-varying
exposure data, and thus, uncertainty is considered to be low.
4) Uncertainty in selection of exposure metric (see Section 5.6.4.1.4)
• There is uncertainty about what metric should be used for modeling exposures to
LAA. Table 5-53 illustrates the uncertainty in the IUR due to exposure metric
selection. The quantitative uncertainty is about threefold.
5) Uncertainty in assessing mortality corresponding to other cancer endpoints (see
Section 5.6.4.1.5)
• The lack of sufficient numbers of workers to estimate the risk of other cancers
potentially related to LAA exposure is an uncertainty of unclear direction but is
considered to be low due to the rarity of those cancers.
6) Uncertainty in control of potential confounding in modeling lung cancer mortality
(see Section 5.4.6.1.6)
• The uncertainty in control of potential confounding by smoking is considered to
be low, as the described sensitivity analysis did not show evidence of potential
confounding.
7) Uncertainty due to potential effect modification (see Section 5.4.6.1.7)
• Smoking was not considered to be related to LAA exposure, and therefore,
smoking is not considered to be a likely effect modifier of cancer risk. Age has
been shown to be a potential effect modifier of lung cancer risk but there was no
evidence of this relationship in the subcohort.
8) Uncertainty due to length of follow-up (see Section 5.4.6.1.8)
• There is uncertainty related to the limited follow-up for cancer mortality, and it is
possible that with subsequent mortality follow-up the IUR could change in a
direction that is unknown.
9) Uncertainty in the use of life-tables to calculate cancer mortality IUR (see
Section 5.4.6.1.9)
• The life-table procedure computes the extra risk of death from birth to 85 years of
age. If the life-tables were extended from 85 to 90 years to account for longer life
spans, the selected lung cancer mortality unit risk (Table 5-52 shows this as
0.068) would be somewhat larger, about 5-10%, and the selected mesothelioma
unit risk (Table 5-50 shows this as 0.122) would be slightly less (about 3%).
Taking both effects into consideration, the uncertainty in the IUR is considered to
be low.
10) Uncertainty in combining of mortality risks to derive a composite cancer mortality
inhalation unit risk (IUR) (see Section 5.4.6.1.10)
5-156
-------
• EPA assumed that the cancer risks were independent, conducted a bounding
analysis and showed the related uncertainty to be very low.
11) Uncertainty due to extrapolation of findings in adults to children (see
Section 5.4.6.1.11)
• There is uncertainty in the assumption that risks are independent of age and that
children are at the same exposure-related risk as adults. The lack of published
information on cancer risks associated with exposures during childhood remains an
uncertainty of unclear magnitude.
5-157
-------
6. MAJOR CONCLUSIONS IN THE CHARACTERIZATION OF HAZARD AND
EXPOSURE RESPONSE
Libby Amphibole asbestos (LAA),30 present in vermiculite ore from the mine near Libby,
MT, is a complex mixture of amphibole fibers—both mineralogically and morphologically (see
Section 2.2). The mixture primarily includes winchite, richterite, tremolite, magnesio-riebeckite,
magnesio-arfvedsonite, and edenite (84:11:6:1:1:1) amphibole minerals that exhibit a range of
fiber morphologies [e.g., asbestiform, acicular, prismatic (Meeker et al., 2003)1. Given the
exposure potential to LAA—and its characteristic mineral composition—a hazard
characterization and cancer exposure-response assessment are presented.
As discussed in Section 1, no RfC for asbestos currently exists, and the EPA IRIS IUR
for asbestos is based on a synthesis of 14 epidemiologic studies that included occupational
exposure to chrysotile, amosite, or mixed mineral fibers (chrysotile, amosite, and crocidolite)
(U.S. EPA, 1988a). Some uncertainty exists in applying the resulting IUR to environments and
minerals not included in the studies considered for the asbestos IUR derivation (U.S. EPA,
1988a). Published mortality studies on the worker cohorts exposed to LAA have become
available since the derivation of the IRIS asbestos IUR [i.e., McDonald et al. (2004), McDonald
etal. (1986a). Amandus and Wheeler (1987), Sullivan (2007), Larson et al. (201 Ob), and
Dunning et al. (2012)1. This assessment documents noncancer and cancer health effects from
inhalation exposure to LAA. Neither an oral slope factor nor an oral reference dose for Libby
Amphibole asbestos were derived in this assessment. Oral exposure was not assessed because
inhalation is the primary route of concern and oral exposure data for Libby Amphibole asbestos
is lacking.
6.1. HUMAN HAZARD POTENTIAL
6.1.1. Exposure
Several different groups of humans have the potential for exposure to fibers from
vermiculite mined in Libby, MT, and hence the potential for exposure to the LAA associated
with this material. These groups include not only the former workers at the mine and mill site,
but also residents in the community of Libby, MT, as well as workers at other locations who
processed the vermiculite product. When the mine in Libby, MT, was active, miners, mill
workers, and those working in the processing plants were exposed to vermiculite ore, silica dust,
and amphibole structures released to air from the ore during the mining and processing
operations (Meeker et al., 2003; Amandus et al., 1987b; McDonald et al., 1986a). In some cases,
workers may have inadvertently transported contaminated materials from the workplace to
30The 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.
6-1
-------
vehicles, homes, and other establishments, typically on the clothing, shoes, and hair. This
transported material may have resulted in "take-home exposure" for the workers, their families,
and other coresidents. The magnitude of these historic take-home exposures was not measured,
so the levels at which individuals in the home might have been exposed are unknown.
While some vermiculite concentrate was exfoliated and used in Libby, MT, most of the
concentrate was transported to expansion plants at other locations across the country where it
was exfoliated and distributed. A review of company records from 1964-1990 indicates that
more than 6 million tons of vermiculite concentrate was shipped to over 200 facilities outside of
Libby, MT (ATSDR, 2008). Because expanded vermiculite from Libby was widely used in
numerous consumer and construction products in 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 from Libby, MT was most notably used as attic insulation
(vermiculite added insulation or VAI) under the brand name Zonolite (Versar, 2003), and as a
soil amendment for gardening, as a fireproofmg agent, and in the manufacturing of gypsum
wallboard. Residents living in communities near the expansion plants may also have been
subjected to some of the same LAA exposure pathways as was the Libby community. The 2008
ATSDR Summary Report observed that individuals in a community with a vermiculite
expansion and processing plant could have been exposed to LAA by breathing airborne
emissions from the facility or by inhalation exposure to contaminants brought into the home on
workers' clothing or from outdoor sources (ATSDR, 2008).
6.1.2. Fiber Toxicokinetics
Although oral and dermal exposure to fibers does occur, inhalation is considered the main
route of human exposure to mineral fibers, and therefore, has been the focus of more fiber
toxicokinetic analyses in the literature. As with other forms of asbestos, exposure to LAA is
presumed to be through all three routes of exposure; however, this assessment specifically
focuses on the inhalation pathway of exposure. Generally, fiber deposition in the respiratory
tract is fairly well defined based on fiber dimensions and density, although the same cannot be
said for fiber translocation to extrapulmonary sites (e.g., pleura). The deposition location within
the pulmonary and extrapulmonary tissues plays a role in the clearance of the fibers.
Fiber clearance from the respiratory tract can occur through physical and biological
mechanisms. Limited mechanistic information is available on fiber clearance mechanisms in
general, and no information specific to clearance of LAA fibers is available. Fibers have been
observed in various pulmonary and extrapulmonary tissues following exposure, suggesting
translocation occurs to a variety of tissues. Studies have also demonstrated that fibers may be
cleared through physical mechanisms (coughing, sneezing) or through dissolution of fibers.
Multiple fiber characteristics (e.g., dimensions, density, and durability) play a role in the
toxicokinetics and toxicity of fibers. The literature examining a variety of fiber determinants and
6-2
-------
their role in disease is extensive, with a focus on fiber length, width, and durability; however,
these studies are often contradictory, making conclusions difficult for fibers in general. This is
in part due to the variety of fibers analyzed, inadequate study design, and/or lack of information
on fiber dimensions in earlier studies. However, due to the importance in understanding the role
of these fiber determinants in the biological response, careful attention has been paid to these
fiber characteristics when analyzing research studies on LAA and asbestiform tremolite, an
amphibole fiber that comprises part of LAA (see Appendix D). No toxicokinetic data are
currently available specific to LAA or its components (e.g., winchite, richterite, tremolite,
magnesio-riebeckite, magnesio-arfvedsonite, and edenite). When available, fiber characteristic
data are presented in the discussion of each study in relation to the toxic endpoints described.
6.1.3. Noncancer Health Effects in Humans and Laboratory Animals
The predominant noncancer health effects observed following inhalation exposure to
LAA are on the lungs and pleural lining surrounding the lungs. These effects have been
observed primarily in studies of exposed workers and community members, and are supported by
laboratory animal studies. Recent studies have also examined other noncancer health effects
following exposure to LAA, including autoimmune effects and cardiovascular disease; this
research base is currently not as well developed as that of respiratory noncancer effects.
Adequate data are not available to differentiate the health effects of the predominant
mineralogical forms composing LAA. Although the adverse effects of asbestiform tremolite are
reported in the literature, the contribution of asbestiform winchite and asbestiform richterite to
the aggregate effects of LAA has not been determined.
Noncancer health effects identified in humans following inhalation exposure to LAA
include pleural abnormalities, asbestosis, and reduced lung function, as well as increased
mortality from noncancer causes. Two cohorts of workers exposed to LAA have been studied:
workers at the mine and related operations in Libby, MT and employees in the O.M. Scott plant
in Marysville, OH, where the vermiculite ore was exfoliated and used as an inert carrier in lawn
care products. Radiographic assessments of study participants in both cohorts identified
abnormalities consistent with asbestos-related disease, specifically pleural effects and small
interstitial opacities [indicative of interstitial fibrosis (Rohs et al., 2008; Amandus et al., 1987a;
McDonald et al., 1986b; Lockeyet al., 1984)1. These studies provided quantitative exposure
estimates and were considered suitable for exposure-response analysis to support an RfC
derivation. Additionally, five cohort mortality studies of Libby, MT workers identified an
increased risk of mortality from noncancer causes, including asbestosis and other nonmalignant
respiratory disease (Larson et al., 201 Ob; Sullivan, 2007; McDonald et al., 2004; Amandus and
Wheeler, 1987; McDonald et al., 1986a) and cardiovascular disease (Larson et al., 201 Ob).
ATSDR conducted health screening of community members in and around Libby, MT
(including past workers), and identified an increase in radiographic abnormalities with an
6-3
-------
increased number of exposure pathways (Peipins et al., 2004a: Peipins et al., 2003; AT SDR,
2001b). Other researchers have also used these data to identify the increased prevalence of
respiratory symptoms in children (Vinikoor et al., 2010) and to evaluate the prevalence of
radiographic abnormalities and reduced lung function in nonworker participants (Weill et al.,
2011). Radiographic abnormalities were more prevalent in mine/mill workers versus other
exposure categories (i.e., household contacts, 'dusty trades', and community-only exposures)
(Weill et al., 2011). Prevalence of pleural effects increased with age and within each exposure
group. Decreased pulmonary function (as percentage of the predicted forced vital capacity) was
reported for participants with radiographic abnormalities (Weill et al., 2011). A nested
case-control study based on ATSDR community health screening also identified a potential for
increased prevalence of autoimmune disease (Noonan et al., 2006), and other experimental work
has examined mechanistic steps relating to autoimmunity (Marchand et al., 2012; Pfau et al.,
2005). Further development of this area of research could provide additional insights into the
range of health effects possibly linked to LAA.
Laboratory animal and mechanistic studies of LAA are consistent with the noncancer
health effects observed in workers exposed to LAA in Libby, MT and Marysville, OH, as well as
exposed community members. Pleural fibrosis was increased in hamsters after intrapleural
injections of LAA (Smith, 1978). More recent studies have demonstrated increased collagen
deposition and inflammation consistent with fibrosis following intratracheal instillation of LAA
fibers in mice and rats (Cypher! et al., 2012b; Cyphertet al., 2012a; Padilla-Carlin et al., 2011;
Shannahan et al., 2011 a: Shannahan et al., 20lib: Smarttet al., 2010; Putnam et al., 2008).
Pulmonary fibrosis, inflammation, and granulomas were observed after tremolite inhalation
exposure in Wistar rats (Bernstein et al., 2005b; Bernstein et al., 2003) and intratracheal
instillation in albino Swiss mice (Sahu et al., 1975). Davis et al. (1985) also reported pulmonary
effects after inhalation exposure in Wistar rats, including increases in peribronchiolar fibrosis,
alveolar wall thickening, and interstitial fibrosis.
Limited research is available on noncancer health effects occurring outside the
respiratory system and pleura. Sullivan (2007) and Larson et al. (201 Ob) examined
cardiovascular disease-related mortality in the cohort of exposed workers from Libby (see
Section 4.1). Mechanistic studies have examined the potential role of iron and the associated
inflammation for both respiratory and cardiovascular disease (Shannahan et al., 2012a;
Shannahan etal., 2012c; Shannahan et al., 2012b; Shannahan et al., 2012d; Shannahan et al.,
201 Ib). Recent studies in the Libby, MT community examined the association between asbestos
exposure and autoimmune disease (Noonan et al., 2006) or autoantibodies and other immune
markers [(Pfau et al., 2005) see Table 4-15]. Mechanistic studies examining the role of LAA
exposure in autoimmune disease have shown limited effects but did observe an increase in
autoantibodies in the serum of exposed animals (Salazar et al., 2013; Salazar et al., 2012). These
results are supported by recent in vitro studies demonstrating increased autoantibodies to
6-4
-------
mesothelial cells, leading to collagen deposition (Serve etal., 2013). These recent studies have
examined the association between asbestos exposure and autoimmune disease; additional
research in this area could enhance understanding of this potential mode of action for noncancer
effects (Salazaretal.. 2013: Serve etal.. 2013: Rasmussen and Pfau. 2012: Salazar et al.. 2012:
Blake et al.. 2008: Pfau et al.. 2008: Hamilton et al.. 2004). Limitations in the number, scope,
and design of these studies make it difficult to reach conclusions about the role of asbestos
exposure in either cardiovascular disease or autoimmune disease.
Limited in vitro studies have demonstrated oxidative stress following LAA exposures in
various cell types (Duncan et al., 2014: Duncan etal., 2010: Hillegass etal., 2010: Pietruska et
al., 2010: Blake et al., 2007). LAA fibers increased intracellular ROS in both murine
macrophages and human epithelial cells (Duncan etal., 2010: Blake et al., 2007). The role of
surface iron on inflammatory marker gene expression and inflammasome activation was shown
to be increased following exposure to LAA in human epithelial cells [(Duncan et al., 2014:
Shannahan et al., 2012a: Shannahan et al., 2012c: Shannahan et al., 2012b: Shannahan et al.,
2012d: Shannahan etal., 20 lib: Duncan etal., 2010: Pietruska et al., 2010) see Table 4-18].
Tremolite studies also demonstrate cytotoxicity in various cell culture systems (see Table 4-22).
However, evidence is currently insufficient to establish the noncancer MOA for LAA.
6.1.4. Carcinogenicity in Humans and Laboratory Animals
There is convincing evidence of a causal association between exposure to LAA and
mesothelioma and lung cancer in workers from the Libby, MT vermiculite mining and milling
operations as well as workers from the Marysville, OH plant (Larson et al., 201 Ob: Sullivan,
2007: McDonald et al., 2004: Amandus et al., 1988: Amandus and Wheeler, 1987: McDonald et
al., 1986a). Whitehouse et al. (2008) documented 11 mesothelioma cases in nonworkers
exposed to LAA in Libby, MT. Increased lung cancer and mesothelioma deaths are also
reported for worker cohorts exposed to other forms of amphibole fibers (amosite and crocidolite)
(de Klerk et al., 1989: Seidman et al., 1986: Henderson and Enterline, 1979). These findings are
consistent with the increased cancers reported for communities exposed to various rocks and
soils containing tremolite fibers (Hasanoglu et al., 2006: Sichletidis et al., 1992b: Baris et al.,
1987: Langeretal.. 1987: Baris etal., 1979: Yazicioglu, 1976). Although potency, fiber
dimension, and mineralogy differ among amphiboles, these studies are supportive of the hazard
identification of LAA fibers described in this assessment.
Although experimental data in animals and data on toxicity mechanisms are limited for
LAA, tumors were observed in tissues similar to those seen in humans (e.g., mesotheliomas, lung
cancer) indicating that the existing data are consistent with the cancer effects observed in humans
exposed to LAA. Smith (1978) reported increased incidence of mesotheliomas in hamsters after
intrapleural injections of LAA. Additionally, studies in laboratory animals (rats and hamsters)
exposed to tremolite via inhalation (Bernstein et al., 2005b: Bernstein et al., 2003: Davis et al.,
6-5
-------
1985), intrapleural injection (Roller et al.. 1997. 1996: Davis etal.. 1991: Wagner et al.. 1982:
Smith et al., 1979), or implantation (Stanton et al., 1981) have shown increases in mesotheliomas
and lung cancers. The tremolite used in these studies was from various sources and varied in
fiber content and potency (see Section 4.2, Appendix D).
The available mechanistic information suggests LAA induces effects that may play a role
in carcinogenicity (see Section 4.2, Appendix D). Several in vitro studies have demonstrated
oxidative stress and genotoxicity following LAA exposures in various cell types (Duncan et al.,
2010: Hillegass et al., 2010: Pietruska et al., 2010: Blake et al., 2007). LAA increased
intracellular ROS in both murine macrophages and human epithelial cells (Duncan et al., 2010:
Blake et al., 2007). Additionally, surface iron, inflammatory marker gene expression,
inflammasome, and aneugenic micronuclei were increased following exposure to LAA in human
epithelial cells (Duncan et al., 2010: Pietruska et al., 2010). Tremolite studies demonstrate
cytotoxic and clastogenic effects (e.g., micronucleus induction and chromosomal aberrations) of
the fibers in various cell culture systems.
6.1.5. Susceptible Populations
Certain segments of the general population could be more susceptible to adverse health
effects from exposure to LAA. In general, factors that may contribute to increased susceptibility
from environmental exposures include lifestage, gender, race/ethnicity, genetic polymorphisms,
health status, and lifestyle (e.g., smoking). However, little data exist to address the potential of
increased susceptibility to cancer or noncancer effects from exposure to the LAA.
Most occupational studies of workers exposed to LAA have examined the effects only in
men because this group represents the vast majority of workers in these settings (Moolgavkar et
al., 2010: Sullivan, 2007: McDonald et al., 2004: Amandus et al., 1988: Amandus et al., 1987b:
Amandus and Wheeler, 1987: Amandus et al., 1987a: McDonald et al., 1986a: McDonald et al.,
1986b). The analysis presented here includes all workers; however, there were few women in
the cohort, and therefore, no determination can be made regarding increased susceptibility to
lung cancer or mesothelioma by gender. Gender-related differences in exposure patterns,
physiology, and dose-response are some of the factors that may contribute to gender-related
differences in risk from asbestos exposure (Smith, 2002). The limited data available from
community-based studies (ATSDR, 2000) do not provide a basis for drawing conclusions
regarding gender-related differences in carcinogenic effects from LAA. Racial diversity among
the workers in the studies examining LAA exposure is also limited, and data on ethnic groups are
absent, precluding the ability to examine racial and ethnicity-related differences in the mortality
risks within the Libby, MT worker cohort. Finally, the potential modifying effects of genetic
polymorphisms, preexisting health conditions, nutritional status, and other lifestyle factors have
not been studied sufficiently to determine their potential contribution to variation in risk in the
population.
6-6
-------
6.1.6. Mode-of-Action Information
Research on multiple types of elongate mineral fibers supports the role of multiple modes
of action following exposure to LAA. Of the MO As described in Section 4.4, the evidence that
chronic inflammation, genotoxicity and cytotoxicity, and cellular proliferation may all play a role
in the carcinogenic response to LAA is only suggestive (see Table 4-23). In vitro studies provide
evidence that amphibole asbestos is capable of eliciting genotoxic and mutagenic effects in
mammalian respiratory cells; however, direct evidence linking mutagenicity to respiratory cells
following inhalation exposure is lacking. Results of the in vivo studies described here are
consistent with the hypothesis that some forms of amphibole asbestos act through a MOA
dependent on cellular toxicity, based on the observations that cytotoxicity and reparative
proliferation occur following subchronic exposure and that bronchiolar tumors are produced at
exposure levels that produce cytotoxicity and reparative proliferation. However, dose-response
data in laboratory animal studies for damage/repair and tumor development are limited due to the
limited number of inhalation studies that used multiple doses of fibers. Although evidence is
generally supportive of a MOA involving chronic inflammation or cellular toxicity and repair,
there is insufficient evidence to establish a MOA; thus, a linear approach is used to calculate the
inhalation cancer unit risk in accordance with the default recommendation of the 2005
Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a). It is possible that multiple MOA
discussed above, or an alternative MOA, may be responsible for tumor induction.
6.1.7. Weight-of-Evidence Descriptor for Cancer Hazard
Under the EPA Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), LAA is
carcinogenic to humans following inhalation exposure based on epidemiologic evidence that
shows a convincing association between exposure to LAA fibers and increased lung cancer and
mesothelioma mortality (Larson et al., 201 Ob; Moolgavkar et al., 2010; Sullivan, 2007;
McDonald et al., 2004; Amandus and Wheeler, 1987; McDonald et al., 1986a). These results are
further supported by animal studies that demonstrate the carcinogenic potential of LAA fibers
and tremolite fibers in rodent bioassays (see Section 4.1, 4.2, Appendix D). As LAA is a durable
mineral fiber of respirable size, this weight-of-evidence descriptor is consistent with the
extensive published literature that documents the carcinogenicity of amphibole fibers [as
reviewed in (Aust etal., 2011; Broaddus etal., 2011; Bunderson-Schelvan et al., 2011; Huang et
al.. 2011: Mossman etal.. 2011)1.
EPA's Guidelines for Carcinogenic Risk Assessment (U.S. EPA, 2005a) indicate that for
tumors occurring at a site other than the initial point of contact, the weight of evidence for
carcinogenic potential may apply to all routes of exposure that have not been adequately tested at
sufficient doses. An exception occurs when there is convincing information (e.g., toxicokinetic
data) that absorption does not occur by other routes. Information on the carcinogenic effects of
LAA via the oral and dermal routes in humans or animals is absent. The increased risk of lung
6-7
-------
cancer and mesothelioma following inhalation exposure to LAA has been established by studies
in humans, but these studies do not provide a basis for determining the risk from other routes of
exposure. Mesothelioma occurs in the pleural and peritoneal cavities, and therefore, is not
considered a portal-of-entry effect. However, the role of indirect or direct interaction of asbestos
fibers in disease at these extrapulmonary sites is still unknown. There is no information on the
translocation of LAA to extrapulmonary tissues following either oral or dermal exposure, and
limited studies have examined the role of these routes of exposure in cancer. Therefore, LAA is
considered carcinogenic to humans by the inhalation route of exposure.
6.2. EXPOSURE-RESPONSE
6.2.1. Noncancer/Inhalation
There were three potential candidate studies for the derivation of the RfC—two of these
were occupationally exposed cohorts, that is, the Libby worker cohort (Larson et al., 2012a) and
the Marysville worker cohort (Rohs et al., 2008) and the third was of community residents in
Minneapolis, MN [Minneapolis cohort; (Alexander et al., 2012)1. Each of these studies provided
individual exposure estimates and documented increased hazard of pleural effects. As detailed in
Section 5.2.1, each of the available studies has strengths and weaknesses. The cohort of
Marysville, OH workers [Lockey et al. (1984) and the follow-up by Rohs et al. (2008)1 was
selected as the principal cohort over the Libby worker cohort for several reasons: (1) lack of
confounding by residential and community exposure; (2) availability of information on important
covariates (e.g., BMI); (3) an exposure-response relationship defined for lower cumulative
exposure levels (particularly the workers hired in 1972 or later and evaluated in 2002-2005);
(4) adequate length of follow-up; (5) use of more recent criteria for evaluating radiographs (ILO,
2002); (6) availability of high-quality exposure estimates based on numerous industrial hygiene
samples and work records (see Section 5.2.1 for details); and (7) availability of data on TSFE
matched to the exposure data. The study of Libby workers (Larson et al., 2012a) had many of
these same attributes (e.g., adequate follow-up and high-quality exposure estimates), but
exposure levels were generally higher in this group compared to the Marysville workers, and the
Libby workers may have experienced greater levels of undocumented "take home" and other
nonoccupational exposure for which TSFE data were more uncertain. The main limitation in the
study of Minneapolis community residents (Alexander et al., 2012) was relatively lower quality
exposure information; exposure estimates were based on a small number of total dust
measurements from stack emissions combined with air dispersion modeling, and the authors
estimate that the individual exposure estimates are likely to have an order of magnitude of
uncertainty. Thus, the study of Marysville workers (Rohs et al., 2008) was selected as the
principal study for RfC derivation.
The MOA for LPT and the results of other asbestos epidemiology studies could
potentially inform noncancer modeling decisions and suggest exposure metrics to use in
6-8
-------
modeling. However, the conclusion of Section 3 of this assessment is that the data are not
sufficient to establish a MOA for the pleural and/or pulmonary effects of exposure to LAA. A
general understanding of the biology and epidemiology of pleural health endpoints suggests that
the timing of exposure, the exposure intensity, and the duration of exposure may be important
explanatory variables and these variables were carried forward for modeling using three
exposure metrics called "mean exposure" or "mean concentration" (C), "cumulative exposure"
(CE), and "residence time-weighted exposure" (RTW).
LPT was selected as the critical effect from among the noncancer radiographic endpoints
evaluated in the principal study for derivation of the RfC, with a BMR of 10% extra risk. LPT
was selected because it is the study endpoint that generally appears soonest after exposure and at
the lowest levels of exposure (i.e., is deemed the most sensitive endpoint). LPT is a pathological
change associated with decreased pulmonary function, and thus is considered an appropriate
adverse effect for deriving the RfC (see Section 5.2.2.3 and Appendix I).
The RfC is derived based on data from the Marysville workers who were evaluated in
2002-2005 and hired in 1972 or later. These workers were selected due to the greater certainty
in their exposure assessment. BMC modeling was used to derive the POD. Statistical models
were evaluated based on biological and epidemiological considerations (see Section 5.2.2.6.1)
and EPA's Benchmark Dose Technical Guidance (U.S. EPA, 2012). Considerations included
(1) the nature of the data set (i.e., cross-sectional, dichotomous health outcome data), (2) ability
to estimate the effect of exposure and of covariates, (3) appropriate inclusion of a plateau term
representing theoretical maximal prevalence of the outcome, and (4) appropriate estimation of
the background rate of the outcome. A number of models were evaluated, and the Dichotomous
Hill model with the plateau parameter fixed at a literature-derived value of 85% was selected for
the derivation of a POD and sensitivity analyses. This model had very similar fit to others
evaluated and was thought to provide the greatest flexibility and ability to determine sensitivity
of model results to various assumptions. EPA considered several exposure metrics informed by
general biology and the epidemiologic literature, including mean exposure intensity, cumulative
exposure (which incorporates duration of exposure), and RTW exposure (which incorporates
TSFE by weighting more heavily exposures occurring in the more distant past).
Another important feature of the exposure-response analysis is the ability to include
effects of TSFE in the modeling. TSFE has been shown in the literature to be important in
evaluating risk of LPT, and studies have shown that prevalence of LPT can increase with
increasing TSFE, even after exposure has ceased. EPA evaluated TSFE as a predictor in the
primary analytic group of workers hired after 1972 and evaluated in 2002-2005, but found that
TSFE was not significantly associated with LPT in this group—likely due to the very low
variability in TSFE for this particular population. Thus, EPA used a hybrid modeling approach
to "borrow" information on the effect of TSFE from a larger subset of the Marysville workers
with greater variability in TSFE. The model was fit in the group of all workers evaluated in
6-9
-------
2002-2005 (regardless of hire date), including both LAA exposure and TSFE as predictors. The
regression coefficient corresponding to TSFE was then set as a fixed parameter in the model for
the primary analytic group of workers hired in 1972 or later. In this hybrid modeling, mean
exposure was used due to its superior model fit compared to cumulative exposure. RTW
exposure was not used since TSFE was included as a separate covariate (to avoid collinearity of
predictors). Using this modeling approach (details in Section 5.2.2.6.2), the resulting BMCio
under these modeling assumptions is 0.0.0923 fiber/cc; the corresponding lower 95% confidence
limit of the BMCio (BMCLio) is 0.026 fiber/cc.
The RfC is obtained by applying uncertainty factors as needed. Two default UFs and one
data-informed UF have been applied for a composite UF of 300 (intraspecies uncertainty factor,
UFn =10; database uncertainty factor, UFo = 3; data-informed subchronic-to-chronic uncertainty
factor, UFs =10) (see Section 5.2.5). As shown below, the chronic RfC is 9 x 10~5 fiber/cc for
LAA, calculated by dividing the POD by a composite UF of 300:
Chronic RfC = POD-UF (6-1)
= 0.026 fiber/cc - 300
= 8.67 x io~5 fiber/cc, rounded to 9 x io~5 fiber/cc
Note that for the primary RfC as well as for all the alternative RfCs, the fiber concentration are
presented here as continuous lifetime exposure in fibers/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 |im. Historical PCM analysis (1960s and early 1970s) generally had less resolution,
and fibers with minimum widths of 0.4 or 0.44 jim 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.
While this assessment is informed by studies of other types of asbestos, it is not a
complete toxicity review of other amphiboles or of chrysotile asbestos.
EPA conducted RfC derivation from the same subcohort with alternative definitions of
the health endpoint (see Section 5.2.3.1 and 5.2.3.2). The chronic RfC value for "any pleural
thickening" (APT) was also 9 x io~5 fiber/cc and the same value was derived for "any
radiographic change" (ARC). Although EPA based the primary value of the RfC on the model
based on mean exposure, Section 5.2.4 illustrates an alternative derivation of an RfC of
1 x icr4 fiber/cc from the same cohort with an alternative exposure metric of cumulative
exposure. EPA also conducted alternative modeling of the Marysville cohort, including all
individuals who participated in the health examination in 1980 (Lockey et al., 1984) and
2002-2005 (Rohs et al., 2008) and who were not exposed to asbestos from a source outside of
the Marysville facility (see Section 5.2.4 and Appendix E). The modeling of this full cohort
6-10
-------
(n = 434 individuals) was performed using an alternative critical effect of "any pleural
thickening" (APT), a slightly different definition of LPT (diagnostic criteria changed slightly
over time) and with "any radiographic change" (ARC). These analyses yielded five other RfC
values. A summary table of the primary and alternative derivation of the RfC is provided in
Table 5-11 in Section 5.2.5; all eight alternative derivations of the RfC were within threefold of
the primary RfC, ranging from 3 x 1CT5 fiber/cc up to 2 x 1CT4 fiber/cc. This series of
derivations further substantiates the primary RfC derived from the Marysville workers evaluated
in 2002-2005, and hired in 1972 or later.
Confidence in the principal study is considered medium. The data used are human
epidemiological data which are preferred to animal bioassays, and the principal study (Rohs et
al., 2008) is conducted in a population of occupationally exposed workers with long-term,
relatively low intensity, exposures. While deriving the primary analysis from the group of
workers evaluated in 2002-2005 and hired after 1972 resulted in a smaller data set with fewer
cases to model, alternative RfC derivations based on the larger group of workers without
restriction on the date of hire (and many more cases) yielded similar values of the RfC. The
exposure assessment in the principal study is based on measured data. The main source of
uncertainty in the exposure estimates is incomplete exposure measurements for some of the
occupations/tasks before industrial hygiene improvements that started about 1973 or 1974 and
continued throughout the 1970s (see Appendix F, Figure F-l). The principal study assessed the
health outcome cross sectionally and this may underrepresent the true health burden as
individuals with more severe disease could have left employment or may have died and not been
included in the follow up study, resulting in an underestimation of overall toxicity. However, for
health outcomes not considered to be frank effects, such as LPT, this underestimation should be
minimal. Further, Rohs et al. (2008) compared the study participants with the complete study
population and found no evidence of major differences in the two group's exposure distributions.
Thus, the potential for selection bias is considered to be low. In terms of the sensitivity of the
principal study to detect the critical effect (LPT) by radiograph, it is known that HRCT can
identify asbestos-related lesions in the respiratory tract that cannot be identified by standard
radiographs [e.g., Jankovic et al. (2002). Lebedova et al. (2003). and Simundic et al. (2002)1.
Thus, the technology employed for determining the prevalence of radiographic changes in the
Marysville cohort will likely underestimate the prevalence of pleural lesions that could be
detected using HRCT.
Confidence in the completeness of the overall database is medium. The database consists
of long-term mortality and morbidity studies in humans exposed via inhalation to LAA. The
mortality studies do not provide appropriate data for RfC derivation for pleural abnormalities,
although the two other morbidity studies (Alexander et al., 2012; Larson et al., 2012a) do support
the conclusion that low levels of exposure to LAA is associated with increased prevalence of
LPT. It is known that inhaled asbestos fibers migrate out of the lung and into other tissues (see
6-11
-------
Section 3.1), which leads to uncertainty regarding the assumption that other health effects would
not be expected. While a potential for autoimmune effects and cardiovascular disease is noted in
exposed individuals, data are insufficient to provide a quantitative exposure response relationship
for these endpoints. It is unknown whether an RfC based on these other health would result in a
higher or lower estimate for the RfC. Nor is there evidence as to whether any of these other
effects would occur earlier following exposure to LAA than LPT occurs. No data exist on
general systemic effects in laboratory animals or humans. Therefore, overall confidence in the
RfC is medium, reflecting medium confidence in the principal study and medium confidence in
the completeness of the overall database.
6.2.1.1. Uncertainty and Sensitivity Analyses for Reference Concentration (RfC) Derivation
It is important to consider the sources of uncertainties in the derivation of the RfC for
LAA. These include the following:
Measurement error in exposure assessment and assignment. The estimated exposure for
each individual relied on self-reported employment history, which may be subject to
recall error. Only data from 1972 and later were used for an RfC derivation based on
lack of fiber measurements prior to this date, but some uncertainty remains due to the
limited amount of industrial hygiene data collected in 1972-1973. There is also
uncertainty in the post-1972 data regarding asbestos content and potency of fibers
originating from other ore sources (Virginia, South Carolina, and South Africa).
Although LAA was not used in the facility after 1980, industrial hygiene measurements
collected after 1980 showed low levels of fibers. Regarding nonoccupational exposure,
any exposure to LAA outside of the workplace is not likely to contribute significantly to
cumulative exposure; -10% of workers reported bringing raw vermiculite home, and the
majority showered and changed clothes before leaving the workplace. As a sensitivity
analysis, EPA evaluated the change in the POD when setting all exposure measurements
after 1980 to zero, and when using the geometric mean rather than the arithmetic mean to
summarize multiple fiber measurements for a given task/location/time period. These
analyses showed a difference of-65 to +50% in the POD.
Radiographic assessment of localized pleural thickening. Conventional
radiographs—rather than more sensitive and specific high-resolution computed
tomography—were used to determine the health outcome. Localized pleural thickening
may be difficult to detect on these radiographs, leading to the potential for outcome
misclassification. However, uncertainty in the reading of x-rays in each individual is
considered minimal because determinations of pleural outcomes detectable on x-rays
were based on agreement among independent classifications of each radiograph done by
multiple qualified chest radiologists.
Use of an alternative critical effect. In addition to the primary analysis using a critical
effect of LPT, EPA also derived an RfC based on the alternative endpoint of any pleural
thickening (APT), and based on any radiographic change (ARC). In the primary analytic
group of workers evaluated in 2002-2005 and hired in 1972 or later, the two alternative
endpoints are identical to the critical effect of LPT, because the one individual with DPT
6-12
-------
also had LPT, and none had interstitial changes. In the larger group of workers used to
estimate the effect of TSFE (all evaluated in 2002-2005), there were 69 cases of APT
and 71 with ARC, of which the majority (n = 66) were LPT. Consequently, the RfC
derived using these two alternative endpoints of APT or ARC were identical (to one
significant digit) to that derived for LPT, 9 x 10'5 fiber/cc.
Length of follow-up. Time from first exposure to x-ray was 23.2-32.7 years in the
primary analytic group of workers evaluated in 2002-2005, and hired in 1972 or later
(mean of 28.2 years). The literature shows that the prevalence of LPT may increase with
time, beyond this observed range of time from first exposure. The lack of observed data
beyond -30 years after first exposure (on average) is a source of uncertainty when
characterizing the exposure-response relationship for a full lifetime of exposure (e.g.,
70 years).
Model Form. A number of model forms were explored in the initial stages of analysis,
and generally showed reasonably close fits as measured by the AIC. The Dichotomous
Hill model with a plateau fixed at 85% was selected for RfC derivation due to its greater
flexibility and ability to evaluate sensitivity to model assumptions. EPA also evaluated
the sensitivity of the fixed plateau parameter and found that the POD changed very little
(<16%) when fixing the plateau at different values (70, 100%) or when estimating the
plateau from the Marysville data.
Effect ofcovariates. Information on a number of covariates was available for the
Marysville workers, including demographic characteristics (gender, smoking status,
BMI) as well as potentially exposure-related factors (hire year, job tenure, exposure
duration, and age at x-ray). The potential for these factors to confound the association
between LAA exposure and LPT was investigated in two ways. First, each was evaluated
for association with the exposure, association with the outcome, and whether it was an
intermediate in the pathway between exposure and outcome (i.e., did they meet the
theoretical definition for a confounder). By these standards, none of the covariates was a
confounder. Second, each covariate was included in the final model to evaluate the
impact on the estimated effects of LAA exposure and TSFE; the differences were quite
small, and none of the covariates were significantly associated with risk of LPT in these
models.
6.2.2. Cancer/Inhalation
6.2.2.1. Background and Methods
The most appropriate data set for deriving quantitative cancer risk estimates based on
LAA exposure in humans is the cohort of workers employed at the vermiculite mining and
milling operation near Libby, MT (see Section 4.1.1.1). The Libby, MT worker cohort has been
the focus of two epidemiologic investigations by the NIOSH scientists. A database created by
NIOSH in the 1980s contains demographic data, work history, and vital status at the end of May
of 1982 for 1,881 workers at the vermiculite mine, mill, and processing plant in Libby, MT (see
Section 4.1.1.1). Vital status follow-up was completed by NIOSH through 2006 using the
National Death Index (Bilgrad, 1997). Nearly 54% of workers in the cohort (n = 1,009) had died
6-13
-------
by December 31, 2006. The data from this update (provided by NIOSH) is the basis of EPA
exposure-response modeling.
EPA does not have sufficient information to select models for the epidemiology data on
the basis of the biological mechanism of action for lung cancer or mesothelioma (see Section 3).
In this situation, EPA's practice is to investigate several modeling options to determine how to
best empirically model the exposure-response relationship in the range of the observed data, as
well as to consider exposure-response models suggested in the epidemiologic literature. For
LAA, possible exposure metrics were explored for model fit to the chosen models forms. The
exposure metric options were selected to provide a range of shapes that was sufficiently flexible
to allow for a variety of ways that time and duration might relate to cancer risk in the data being
modeled. EPA then evaluated how well the models and exposure metric combinations fit the
data being modeled. Then EPA calculated a reasonable upper bound on risk using selected
exposure metrics. This is explained in more detail below and in Section 5.4.5. However, there
are uncertainties in the modeling of the epidemiological data that may impact the IUR and these
are described in Section 6.2.8 below and in greater detail in Section 5.4.6.
In the Libby, MT worker cohort data developed by NIOSH and used by EPA in this
assessment, detailed work histories, together with job-specific exposure estimates, allowed for
the reconstruction of each individual's occupational exposure experience over time to define
multiple exposure metrics. From this information-rich, individual-level data set from NIOSH,
EPA constructed a suite of the different metrics of occupational exposure which had been
proposed in the asbestos literature or used in EPA health assessment on general asbestos
exposures (U.S. EPA, 1988a) as well as modifications proposed (Berry etal., 2012). This suite
of models was defined a priori to encompass a reasonable set of proposed exposure metrics to
allow sufficient flexibility in model fit to these data. These exposure metrics were evaluated in
analytic-regression models to test which exposure metrics were the best empirical predictors of
observed cancer mortality, and the better fitting models were advanced for consideration as the
basis of the exposure-response relationship for the IUR. The types of exposure metrics evaluated
were intended to allow for variations of the classic metric of cumulative exposure, allowing for
more or less weight to be placed on earlier or later exposures. These simulated exposure metrics
were derived mathematically to approximate underlying processes that are not well understood.
Thus, the empirical fit of various exposure metrics to the observed epidemiologic data is
evaluated statistically, and the exposure metrics have epidemiological interpretation but do not
necessarily have direct biological interpretations.
Exposure estimates for all exposure metrics were adjusted to account for the time period
between the onset of cancer and mortality. The lag period defines an interval before death, or
end of follow-up, during which any exposure is excluded from the calculation of the exposure
metric. There was one important limitation of the NIOSH JEM. Of the 991 workers hired
before 1960, 706 workers with unknown department code and unknown exposure assignments
6-14
-------
hired between 1935 and 1959 had the same average estimated exposure intensity. The lack of
information on specific job assignments for such a large portion of these early workers when
exposures were higher resulted in the misclassification of the exposure and effectively yielded
exposure metrics that were differentiated only by the duration of each worker's employment.
For this reason and because there was little measured fiber exposure data during the earlier
period, identifying an adequate exposure-response model fit was unsuccessful. The two biggest
problems were that the duration of employment was the best-fitting metric for modeling
mesothelioma and that the Cox model assumptions were violated in modeling lung cancer
mortality (see Section 5.4.3.4). As a result, this assessment developed a subcohort analysis by
dividing the whole cohort into two groups: those hired before 1960 and those hired after 1959.
This removed all but nine cohort members with missing department code and job category
information and lessened the effect of estimates of early exposures where no air sampling data
were available. For the subcohort of those hired after 1959, those two biggest problems were
resolved: the assumptions of the Cox model were satisfied, and a lagged cumulative exposure
with a decay (rather than duration of exposure, as for the full cohort) was the best-fitting metric
for mesothelioma.
Of the 880 workers hired after 1959, 230 (26%) had died by December 31, 2006. The
number of mesothelioma deaths in the subcohort is relatively small (n = 7, two deaths coded in
ICD-10 and five deaths coded in ICD-9), but the rate of mesothelioma mortality was very similar
in the subcohort (24.7 per 100,000 person-years versus 26.8 per 100,000 person-years for the full
cohort [18 mesothelioma deaths], a difference of less than 10%).
6.2.3. Modeling of Mesothelioma Exposure Response
A Poisson model is employed for estimating the absolute risk of mesothelioma following
exposure to LAA, as the Poisson distribution is an appropriate model to use with data that are
counts of a relatively rare outcome, such as observed mesothelioma deaths in the Libby, MT
worker cohort. Estimation of the exposure-response relationship for mesothelioma using the
Poisson model was performed in WinBUGS software by a MCMC Bayesian approach with an
uninformative or diffuse (almost flat) prior. The model was run to fit the mortality data to
exposure data for various exposure metrics described above. To comparatively evaluate how
much better one model fits than another, the DIG was used. DIG is used in Bayesian analysis
and is an analogue of AIC (Burnham and Anderson, 2002). Use of the DIG and AIC is standard
practice in comparing the fit of nonnested models to the same data set with the same dependent
outcome variable but different independent covariates.
Modeling of mesothelioma mortality included an exposure metric with a cubic function
of time (see eq 5-9), originally proposed in Peto et al. (1982) and employed in derivation of the
IUR for asbestos (U.S. EPA, 1988a, 1986a), as well as modifications of Peto model (Peto model
6-15
-------
with clearance) proposed in the asbestos literature (Berry et al., 2012). See Section 5.4.3 for
further details.
For the subcohort hired after 1959, two cumulative exposure metrics with decay provided
the best model fits. Both metrics had a common 5-year half-life, with lag times of either 10 or
15 years. The Peto model and Peto model with clearance did not fit as well. As it is less likely
that exposure during the last few years before death were contributory to the development of the
cancer and cancer mortality, the zero-lag metrics were dropped from further consideration. The
selected metric as well as the Peto model and Peto model with clearance were retained for
derivation of the IUR (see Section 6.2.7 below and for additional detail see Section 5.4.5). The
selected exposure metric was cumulative exposure with a 5-year half-life and a 10-year lag time
with a central estimate for the slope (KM) of 3.11 x 10'4 per fiber/cc-yr with a 95% upper
confidence limit (UCL) of 5.08 x 10'4 per fiber/cc-yr.
6.2.4. Unit Risk Estimates for Mesothelioma Mortality
The increased risk of mesothelioma mortality attributable to continuous fiber exposure
was estimated using a life-table procedure based on the general U.S. population. The life-table
procedure involved the application of the estimated LAA toxicity to a structured representation
of the general U.S. population in such a manner as to yield age-specific risk estimates for cancer
mortality in the presence or absence of exposure to LAA (see Section 5.4.5; Appendix G).
A default linear low-dose extrapolation was used because the mode of action by which
LAA causes mesothelioma was not established. The lower limit on the effective concentration
(LECoi) yielded a unit risk for mesothelioma mortality of 0.053 per fiber/cc (POD of 1% divided
by the LECoi).
The value of the effective concentration (EC) that would correspond to the measure of
central tendency is the ECoi. This value is used in the derivation of a combined risk of
mesothelioma and of lung cancer. The ECoi yielded a lifetime central estimate value of
0.032 per fiber/cc.
For mesothelioma, the undercounting of cases (underascertainment) is a particular
concern given the limitations of the ICD classification systems used before 1999. In practical
terms, this means that some true occurrences of mortality due to mesothelioma are missed on
death certificates and in almost all administrative databases such as the National Death Index.
Even after introduction of special ICD code for mesothelioma with introduction of ICD-10 in
1999, detection rates are still imperfect (Camidge et al., 2006; Pinheiro et al., 2004), and the
reported numbers of cases typically reflect an undercount of the true number. Kopylev et al.
(2011) reviewed the literature on this underascertainment and developed methods to account for
the likely numbers of undocumented mesothelioma deaths.
To compensate for mesothelioma underascertainment attributable to ICD coding, the
mesothelioma mortality unit risk was further adjusted following the analysis of Kopylev etal.
6-16
-------
(2011). The adjusted mesothelioma central (i.e., maximum likelihood estimate) risk,
corresponding to the best-fit metric, was 0.044 per fiber/cc, and the adjusted mesothelioma
mortality unit risk was 0.074 per fiber/cc.
The adjusted mesothelioma risks for the Peto model and Peto model with clearance
ranged from 2-fold lower (0.035 per fiber/cc) to 3.6-fold higher (0.265 per fiber/cc). Thus, there
is uncertainty in mesothelioma risks generated from similar-fitting models from different
exposure metrics (see details in Section 5.4.5.3).
6.2.5. Modeling of Lung Cancer Exposure Response
All multivariate extended Cox models were fit to the subcohort hired after 1959 with
covariates for gender, race, date of birth, and exposure. Exposure for each of the 40 exposure
parameterizations was calculated independently, and the fit of these exposure metrics was
evaluated one at a time. Of the 40 exposure-response metrics, 14 demonstrated an adequate fit to
the data as measured by the overall model fit with the likelihood ratio test (p < 0.05) as well as
having statistically significant exposure metrics (p < 0.05). However, only the nine models that
demonstrated adequate model and exposure metric fit and incorporated a lag period to account
for cancer latency were considered further in the development of the IUR (see Tables 5-43 and
5-50).
Lagging exposure by 10 years was a better predictor of lung cancer mortality compared
to other lags. As it is less likely that exposure during the last few years before death were
contributory to the development of the cancer and cancer mortality, the zero lag metrics were
dropped from further consideration. The residence time-weighted cumulative exposure, both
with and without decay of the exposure metric, did not fit these lung cancer mortality data well
compared to the other models (see Table 5-44); this form of exposure metric does not
demonstrate evidence of an empirical fit to these epidemiologic data.
The model with the smallest AIC was for cumulative exposure with a 10-year half-life for
decay and a 10-year lag for cancer latency. The extended Cox model estimated a beta (the lung
cancer slope factor: KL) of 1.26 x io~2 per fiber/cc-yr based on a 365-day year, and the
95th percentile upper bound was 1.88 x 10"2 per fiber/cc-yr. The/?-value for the LAA regression
coefficient beta (slope) was <0.001. The slopes and confidence interval for the other exposure
metrics, which had similar fits to these data are reported in Table 5-45. Uncertainty in the choice
of the exposure metric is considered in the derivation of the unit risk (see details in
Section 5.4.5.2), representing the range of unit risks that are derived from these similarly fitting
metrics. The model results that were ultimately selected to reflect the upper bound among the
range of results were based on the cumulative exposure with a 10-year lag exposure metric
(CE10). The extended Cox model estimated a beta (slope) of 5.28 x io~3 per fiber/cc-yr based on
a 365-day year, and the 95th percentile upper bound was 1.00 x io~2 per fiber/cc-yr.
6-17
-------
6.2.5.1. Analysis of Potential Confounding of Lung Cancer Results by Smoking in the
Subcohort
EPA used two approaches to address the confounding issue, including restriction of the
cohort and an analytic evaluation of the potential for confounding by smoking including the
method described by Richardson (2010). Richardson (2010) describes a method to determine
whether an identified exposure relationship with lung cancer is confounded by unmeasured
smoking in an occupational cohort study. EPA implemented this methodology to model the
potential effects of LAA on the risk of COPD mortality on the subcohort of workers hired after
1959 (see Section 5.4.3.8). Summarizing these findings, EPA used the method described by
Richardson (2010) to evaluate whether exposures to LAA predicted mortality from COPD as an
indication of potential confounding by smoking and found a nonsignificant negative relationship,
which was inconsistent with confounding by smoking.
6.2.6. Unit Risk Estimates for Lung Cancer Mortality
The increased risk of lung cancer mortality attributable to continuous fiber exposure was
estimated using a life-table procedure based on the general U.S. population. The life-table
procedure involved applying the estimated LAA-specific toxicity to a structured representation
of the general U.S. population in such a manner as to yield age-specific risk estimated for cancer
mortality in the presence or absence of exposure to LAA (see Section 5.4.5; Appendix G). A
default linear low-dose extrapolation was used because the mode of action by which LAA causes
lung cancer was not established.
The nine exposure-response models retained in Table 5-45 all had reasonably similar
goodness of fit. No single model stands out as clearly statistically superior; however, there is a
range of quality of fit within the set that could be considered to have adequate fit. The lung
cancer mortality unit risks are shown in Table 5-52.
Using the results of the exposure model based on cumulative exposure with a 10-year lag
for cancer latency, the LECoi yielded a lifetime unit risk of 0.0679 per fiber/cc. The value of the
risk that would correspond to the measure of central tendency involves the ECoi rather than the
LECoi. The ECoi yielded a lifetime central estimate of 0.0399 per fiber/cc.
The resulting unit risks in Table 5-52 ranged from 0.0260 to 0.0679 per fiber/cc, for a
lifetime continuous exposure. This shows that the unit risk based on the exposure metric 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) is in the center of this range (i.e., 0.0389 per fiber/cc). This estimate is in
the middle of the range of possible unit risks and does not capture the uncertainty across metrics
with similar goodness of fit (see details in Section 5.4.6).
The model results selected to represent the upper-bound risk among the range of
reasonable results are based on a CE10 metric with a 10-year lag. The model results selected to
6-18
-------
reflect the upper bound among the range of results are based on the CE10 exposure metric with a
10-year lag, providing a unit risk of 0.0679 per fiber/cc.
6.2.7. Inhalation Unit Risk (IUR) Derivation Based on Combined Mesothelioma and Lung
Cancer Mortality from Exposure to Libby Amphibole Asbestos
Once the cancer-specific lifetime unit risks are selected, the two are then combined. It is
important to note that this estimate of overall potency describes the risk of mortality from cancer
at either of the considered sites and is not just the risk of both cancers simultaneously. Because
each of the unit risks is itself an upper-bound estimate, summing such upper-bound estimates
across mesothelioma and lung cancer mortality is likely to overpredict the overall risk.
Therefore, following the recommendations of the Guidelines for Carcinogen Risk Assessment
(U.S. EPA, 2005a), a statistically appropriate upper bound on combined risk was derived to gain
an understanding of the overall risk of mortality resulting from mesothelioma and from lung
cancer. For mesothelioma, the exposure-response models developed by EPA using personal
exposure data on the subcohort (see Table 5-50) provided better fit to the subcohort data than the
Peto model and the Peto model with clearance that have been proposed in the asbestos literature.
For lung cancer, this assessment selected the upper bound among the lung cancer lifetime unit
risks from the plausible exposure metrics (regardless of the small residual differences in quality
of fit). Because there were few metrics with unit risks higher than the best fitting metric's unit
risk for lung cancer mortality endpoint, this method effectively selects the highest lifetime unit
risk among those considered for the lung cancer mortality endpoint.
Table 6-1 shows cancer-specific unit risks as well as combined risk of mesothelioma and
lung cancer. The IUR value of 0.17 per fiber/cc, continuous lifetime exposure, accounts for
important quantitative uncertainties in the selection of the specific exposure metric that may have
remained in an IUR that might have been based on the best-fitting exposure models alone.
Additional uncertainties are discussed in detail in Section 5.4.6.
6-19
-------
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 models.a'b Primary
IUR value in bold.
Lung cancer
Mesothelioma
Combined central
estimate per fiber/cc
Combined upper
bound per fiber/cc
Selected IUR based directly on the Libby data
CE10
CE105-yr half-life
0.115
0.169
Best models from the epidemiologic literature (Peto model with clearance)
CE10
CE10
Peto with clearance
Decay rate of 6.8%/yr
Power of time = 3.9
Peto with clearance
Decay rate of 15%/yr
Power of time = 5.4
0.089
0.061
0.135
0.092
Alternative model from the epidemiologic literature (Peto model)
CE10
Peto
No decay
Power of time = 3
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 fibers/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.
While this assessment is informed by studies of other types of asbestos, it is not a complete toxicity review of
other amphiboles or of chrysotile asbestos.
Age-dependent adjustment factor
As discussed in Section 4.7.1.1, there is no chemical-specific information for LAA, or
general asbestos that would allow for the computation of a chemical-specific age-dependent
adjustment factor for assessing the risk of exposure that includes early-life exposures.
The review of mode-of-action information in this assessment (see Section 4.6.2.4 and
4.6.2.5) concluded that the available information on the mode of action by which LAA causes
lung cancer or mesothelioma is complex and a mode of action is not established at this time.
Thus, in accordance with 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 for substances that act through a mutagenic mode of action is not
recommended.
6-20
-------
6.2.7.1. Comparison with Other Published Studies ofLibby, MT Workers Cohort
Several published studies have previously evaluated risk of mesothelioma and lung
cancer [i.e., Sullivan (2007), Berman and Crump (2008), Larson et al. (201 Ob), and Moolgavkar
et al. (2010)1 in the Libby, MT workers cohort. For mesothelioma, only Moolgavkar et al.
(2010) provided an exposure-response relationship for absolute risk of mesothelioma mortality
that would be comparable with this current assessment. Based on the full cohort, with mortality
data through 2001 and a modification of the Peto/Nicholson exposure metric, life-table analysis
would provide an upper-bound unit risk of approximately 0.13 per fiber/cc continuous lifetime
exposure. Therefore, use of the exposure response modeling of Moolgavkar et al. (2010), would
provide an IUR for excess mesothelioma mortality in close agreement with the IUR derived in
this assessment (see Section 5.4.5.3.1 for more details).
For lung cancer, all of the studies provide exposure-response relationships in terms of
relative risk of lung cancer mortality, and thus, may provide risk estimates comparable to this
assessment. However, inclusion criteria, length of mortality follow-up, and analytic methods
differ among the analyses—thus, the results are not necessarily interchangeable. For comparison
purposes, the lung cancer unit risks from these studies are computed from life-table analyses (see
Table 5-54). The lung cancer unit risks calculated based on the published literature, ranged from
0.010 to 0.079 per fiber/cc (based on the upper confidence limit). This is in close agreement
with this current assessment where an upper-bound estimate of 0.068 per fiber/cc, continuous
lifetime exposure is derived (see Section 5.4.5.3.1 for more details).
6.2.8. Uncertainty in the Cancer Risk Values
It is important to consider the uncertainties in the derivation of the mesothelioma and
lung cancer mortality risks in this assessment in the context of uncertainties in animal-based
health assessments. This assessment does not involve extrapolation from high doses in animals
to low doses in humans. The current assessment 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 in Section 5.4.6 explore
uncertainty in the derivation of the IUR in order to provide a comprehensive and transparent
context for the resulting cancer mortality risk estimates.
Section 5.4.6.1 details the several sources of uncertainty in the assessment of the cancer
exposure-response relationships and the use of those data to derive the inhalation unit risk. The
text that follows summarizes the primary sources of uncertainty and, where possible, the
expected direction of effect on the exposure-response risk estimates and the inhalation units risk.
6-21
-------
1) Uncertainty in low-dose extrapolation (see Section 5.4.6.1.1)
• Some uncertainty remains in the extrapolation of risks based on occupational
exposure to environmental exposure levels, but this uncertainty is considered to
be low as the lower range of occupational exposure overlaps with expected
environmental exposure levels.
2) Uncertainty in exposure assessment, including analytical measurements uncertainty
(see Section 5.4.6.1.2)
• The JEM was based on the "high" exposure estimate for each job according to
Amandus et al. (1987b), and to some extent this could be an overestimate of
exposure. The associated cancer risk would be somewhat underestimated
resulting in a somewhat underestimated IUR.
• The JEM was largely based on estimated fiber concentration using PCM
measurement (with some extrapolations in time), and because PCM may count all
long and thin objects as fibers, these measurement could overestimate the true
LAA fiber concentrations, leading to an overestimate of exposure and a somewhat
underestimated cancer risk resulting in a somewhat underestimated IUR.
• PCM measurements in the era of NIOSH measurements in Libby used a lower
resolution, and therefore, included only somewhat thicker fibers, thereby counting
fewer fibers than would have been counted by later PCM standards. These earlier
measurements could underestimate the true LAA fiber concentrations, leading to
an underestimate of exposure and an overestimate of cancer risk resulting in a
somewhat overestimated IUR.
• The PCM measurement is the available exposure metric for analysis of the Libby
worker cohort at the time of this assessment. Currently, there is no optimal choice
of the best dose metric for asbestos, in general and in particular, for LAA.
Uncertainties related to PCM analytical method are discussed in Section 2, and
such uncertainties cannot be related to the IUR at the time of this assessment.
• Random measurement error in the assignment of exposures could have the effect
of underestimating the risk of lung cancer mortality because that measure of risk
is based on a relative measure. The effect would be to somewhat underestimate
the risk of lung cancer, resulting in a somewhat underestimated unit risk for lung
cancer. It is unclear what the impact of such measurement error would be on the
absolute risk of mesothelioma.
• Exposure to other kinds of asbestos and residential exposure to LAA may have
caused workers' actual personal exposures (as the sum of occupational and
nonoccupational exposures) to have been underestimated by the use of estimated
Libby occupational exposure information alone. This could underestimate the
true LAA fiber exposures leading to an overestimate of the associated cancer risk
resulting in a somewhat overestimated IUR.
6-22
-------
3) Uncertainty in model form (see Section 5.6.4.1.3)
• For mesothelioma, the Poisson model is the standard epidemiologic form and
considered to be the most appropriate model form for rare health outcomes;
therefore, uncertainty is considered to be low. For lung cancer mortality, the Cox
proportional hazards model is the standard epidemiologic form. It is considered
to be the most appropriate model form for health outcomes with time-varying
exposure data and thus uncertainty is considered to be low.
4) Uncertainty in selection of exposure metric (see Section 5.6.4.1.4)
• There is uncertainty about what metric should be used for modeling exposures to
LAA. Table 5-53 illustrates the uncertainty in the IUR due to exposure metric
selection. The quantitative uncertainty is about threefold.
5) Uncertainty in assessing mortality corresponding to other cancer endpoints (see
Section 5.6.4.1.5)
• The lack of sufficient numbers of workers to estimate the risk of other cancers
potentially related to LAA exposure is an uncertainty of unclear direction but is
considered to be low due to the rarity of those cancers.
6) Uncertainty in control of potential confounding in modeling lung cancer mortality
(see Section 5.4.6.1.6)
• The uncertainty in control of potential confounding by smoking is considered to
be low as the described sensitivity analysis did not show evidence of potential
confounding.
7) Uncertainty due to potential effect modification (see Section 5.4.6.1.7)
• Smoking was not considered to be related to LAA exposure, and therefore,
smoking is not considered to be a likely effect modifier of cancer risk. Age has
been shown to be a potential effect modifier of lung cancer risk but there was no
evidence of this relationship in the subcohort.
8) Uncertainty due to length of follow-up (see Section 5.4.6.1.8)
• There is uncertainty related to the limited follow-up for cancer mortality, and it is
possible that with subsequent mortality follow-up, the IUR could change in a
direction that is unknown.
9) Uncertainty in the use of life-tables to calculate cancer mortality IUR (see
Section 5.4.6.1.9)
• The life-table procedure computes the extra risk of death from birth to 85 years of
age. If the life-tables were extended from 85 to 90 years to account for longer life
spans, the selected lung cancer mortality unit risk (Table 5-53 shows this as
0.068) would be somewhat larger, about 5-10%, and the selected mesothelioma
unit risk (Table 5-53 shows this as 0.122) would be slightly less (about 3%).
6-23
-------
Taking both effects into consideration, the uncertainty in the IUR is considered to
be low.
10) Uncertainty in combining mortality risks to derive a composite cancer mortality IUR
(see Section 5.4.6.1.10)
• EPA assumed that the cancer risks were independent, conducted a bounding
analysis, and showed the related uncertainty to be very low.
11) Uncertainty due to extrapolation of findings in adults to children (see
Section 5.4.6.1.11)
• There is uncertainty in the assumption that risks are independent of age and that
children are at the same exposure-related risk as adults. The lack of published
information on cancer risks associated with exposures during childhood remains
an uncertainty of unclear magnitude.
6-24
-------
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: 459-465.
http://dx.doi.org/10.1093/annhvg/32.inhaled_particles VI.459
Amandus. HE: Wheeler. R: Jankovic. J: Tucker. J. (1987b). The morbidity and mortality of vermiculite miners and
millers exposed to tremolite-actinolite: Part I. Exposure estimates. Am J Ind Med 11: 1-14.
Anderson. HA: Lilis. R: Damn. SM: Fischbein. AS: Selikoff. IJ. (1976). Household-contact asbestos neoplastic risk.
AnnN Y Acad 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.
HumPathol 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. OccupEnvironMed 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/annhyg/32.inhaled_particles 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.20Q310-1436ST
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.
7-1
-------
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]. JToxicol 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: Page. T: Greene. RE. (1985). Respiratory illness in the construction trades. I. The significance of
asbestos-associated pleura! 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.org/10.1111/j. 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. EurRespir 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/iapplphysiol.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.
7-2
-------
Berkson. J. (1950). Are there two regressions? J Am Stat Assoc 45: 164-180.
Berman. 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. Ill l/i.1539-
6924.2010.01581.x
Herman. 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: Castranova. V: Donaldson. K: Fubini. B: Hadlev. J: Hesterberg. T: Kane. A: Lai. D: McConnell EE:
Muhle. H: Oberdorster. G: Olin. S: Warheit DB: Group.. IRSIW. (2005a). Testing of fibrous particles:
short-term assays and strategies. Inhal Toxicol 17: 497-537. http://dx.doi.org/10.1080/08958370591001121
Bernstein. D: Rogers. R: Smith. P. (2005b). 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 pleura!
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.2012.23
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.
Bianchj 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.
Bianchj 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
7-3
-------
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.
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: Pi 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 Environ Med 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: Holiaa 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.co m/statistics/statistical+theorv+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
7-4
-------
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
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]. J Toxicol Environ Health B Crit Rev 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/(SICm098-
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 exposure to Libby amphibole asbestos and localized
pleural thickening. J Occup EnvironMed 55: 1350-1355.
http://dx.doi.org/10.1097/JOM.Ob013e3182a3bb25
Christensen. KY: Kopylev. L. (2012). Localized pleural 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/annhY g/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]. Am J Pathol 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
din. 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
7-5
-------
Cohen. D: Ami. SF: Brain. JD. (1979). Smoking impairs long-term dust clearance from the lung. Science 204: 514-
517.
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(05)80207-5
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 pleural 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/iic.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
Health 42: 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). Placenta! 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: Shannahan. 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 and brucite 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
7-6
-------
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.
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: 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
Mutagen599: 124-134. http://dx.doi.0rg/10.1016/i.mrfmmni.2006.02.005
Ding. M: Dong. Z: Chen. F: Pack. D: Ma. W: Ye. J: Shi. X: Castranova. V: Vallyathan. 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'Sullivan. 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'Sullivan. 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/ajim.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.
7-7
-------
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 andLibby 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/IPCS
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
Ehrhch. 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: Sirvestri. 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 pleural 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
7-8
-------
Hammond. EC: Selikoff IJ: Seidman. H. (1979). Asbestos exposure, cigarette smoking and death rates. Ann N Y
Acad Sci 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. http://dx.doi.0rg/10.1164/airccm.159.l.9712108
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
Acad Sci 643: 419-429.
Hague. AK: Ah. 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-Hohrfeld. 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
7-9
-------
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. (1994). Pleural plaques and risk for bronchial carcinoma and mesothelioma. A prospective study. Chest
105: 144-150.
Hillerdal. G: Henderson. DW. (1997). Asbestos, asbestosis, pleura! plaques and lung cancer [Review]. Scand J
Work Environ Health 23: 93-103.
Hirvonen. A: Tuimala. J: OHikainen. 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
(1971) (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 (1980). 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
IOM (Institute of Medicine). (2006). Asbestos: Selected cancers. Washington, DC: National Academies Press.
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/ehc5 3. htm
7-10
-------
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.
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/annhvg/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.org/10.1111/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: Rorunno. 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.
(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/10.1016/80140-6736(87)90305-9
7-11
-------
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/10.1111/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 Environ Med 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/siweh.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.
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.9Q88311
Lockev. JE. (1985) Pulmonary hazards associated with vermiculite exposure. (Master's Thesis). University of
Cincinnati, Cincinnati, OH.
7-12
-------
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: Dement J: Richardson. D: Wolf. S. (2010). Asbestos fibre dimensions and lung cancer mortality among
workers exposed to chrysotile. Occup Environ Med 67: 580-584.
http://dx.doi.org/10.1136/oem.2009.050120
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 JEpidemiol 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 X174 RFIDNA. Occup Environ Med 51: 200-204.
Magnanj 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: Lloyd. 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/annhvg/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
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.
7-13
-------
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. BrJ 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. In U 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 trachea! 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 pleural abnormalities in people who had
indeterminate chest radiographs: the Libby, MT, experience. Int J Hyg Environ Health 208: 87-99.
Musk. AW: de Klerk. NH: Reid. A: Ambrosini. GL: Fritschi. L: Olsea 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
7-14
-------
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 pleural mesothelioma and other asbestos-associated diseases
[Review]. MutatRes 659: 126-136. http://dx.doi.0rg/10.1016/i.mrrev.2008.02.002
Neri. S: Boraschj P: Antonelli. A: Falaschi. F: Baschieri. L. (1996). Pulmonary function, smoking habits, and high
resolution computed tomography (HRCT) early abnormalities of lung and pleural 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,
NC: U.S. Department of Health and Human Services, National Toxicology Program.
http://ntp.niehs.nih.gov/ntp/htdocs/LT rptsAr279.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.
7-15
-------
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-Carlin. 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: Anduiar. 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/djs513
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: Navajas. 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.tb 18720.x
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
7-16
-------
Pfau. JC: Seib. 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
autoirnmunity 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/S0034-5687(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. AnatRec 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
Reid. A: Berry. G: de Klerk. N: Hansen. J: Hevworth. J: Ambrosini. G: Fritschi. 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
7-17
-------
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: Olsea N: Sleith. J: Samuel L: Aboagye-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 VTJ.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 and NHANES 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? HumPathol 39: 307-
315. http://dx.doi.0rg/10.1016/j.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.
Rom WM: Travis. WD: Brodv. AR. (1991). Cellular and molecular basis of the asbestos-related diseases [Review].
Am J Respir Crit Care Med 143: 408-422. http://dx.doi.0rg/10.1164/airccm/143.2.408
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. Am J Epidemiol 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.
http://YOsemite.epa. gov/sab/sabproduct.nsf/0/4F2A245C7160456B85257B030072E5D3/$File/EPA-SAB-
13-001 -unsigned.pdf
7-18
-------
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. J Toxicol 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 activationis 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. Toxicol Lett 72: 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
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
7-19
-------
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/S0891-5849(03)00060-l
Shukla. A: Lounsbury. KM: Barrett. TF: Gell J: Rincon. M: Butnor. KJ: Taaties. 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
pleura! 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
pleura! mesothelioma with endemic pleura! calcifications in a rural area in Greece. Med Lav 83: 326-329.
Simundic. I: Jankovic. T: Tocilj. J: Gjakun. K: Juric. I: Roglic. J: Captain. V. (2002). Diagnostic value of clinical,
radiologic and functional examinations of patients with pulmonary and pleura! 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 on biologic 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 Stat Methodol 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. Comput Biol Med 41: 565-573. http://dx.doi.0rg/10.1016/i.compbiomed.2011.05.009
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.
7-20
-------
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. Int J Hyg 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.
Takemura. T: Rom. WN: Ferraris. VJ: Crystal. RG. (1989). Morphologic characterization of alveolar macrophages
from subjects with occupational exposure to inorganic particles. Am Rev Respir Dis 140: 1674-1685.
http://dx.doi.org/10.1164/airccm/140.6.1674
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). (199Ib). 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/ZyPURL.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
7-21
-------
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 Report]. (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/ZyPURL.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. (EPA/630/R-03/003F). Washington, DC: U.S. Environmental
Protection Agency, Risk Assessment Forum.
http://www.epa.gov/raf/publications/pdfs/childrens supplement_final.pdf
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 intemet/registiT/termreg/searchandretrieve/glossariesandkevwordlists/search.do
?details=&glossarvName=IRIS%20Glossary
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/publications/pdfs/benchmark dose guidance.pdf
7-22
-------
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: U.S. Environmental Protection Agency, National
Center for Environmental Assessment, http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm?deid=255721
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. EPA (U.S. Environmental Protection Agency). (2014). IRIS summary sheet for asbestos (CASRN 1332-21-4).
Washington, DC: U.S. Environmental Protection Agency, Integrated Risk Information System. Retrieved
from http://www.epa.gov/iris/subst/0371 .htm
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: Firiberti. 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.org/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://cfpub.epa.gov/ols/catalog/advanced displav.cfm?&FIELDl=SUBJECT&INPUTl=Vermiculite%20a
ttic%20insulation&TYPEl=EXACT&LOGICl=AND&COLL=&SORTFIELD=YR&SORTORDER=DES
C&item count=l
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.0rg/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, radiological changes and
exposure: Analysis of ATSDR data from Libby, MT, USA. 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/ngOl 10-14
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/sugi30/201-30.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
7-23
-------
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.
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. StatMed26: 1150-1169. http://dx.doi.org/10.1002/sim.26Q2
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; Hodkinson. D; Winters. R; Mossman. BT. (1997).
Mineralogical features associated with cytotoxic and proliferative effects of fibrous talc and asbestos on
rodent tracheal epithelial and pleural 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; Funaki. M; Maeda. K; Matsunaga. C; Yamaoka. K. (1993). Effect of smoking on pleural
thickening in asbestos workers. Br J Ind Med 50: 898-901. http://dx.doi.org/10.1136/oem.50.10.898
Yazicioglu. S.. (1976). Pleural calcification associated with exposure to chrysotile asbestos in southeast Turkey.
Chest 70: 43-47. http://dx.doi.org/10.1378/chest.7Q. 1.43
Yazicioglu. S; Ilcavto. R; Balci K; Sayli. BS; Yorulmaz. B. (1980). Pleural calcification, pleural 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; Yen. 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
Zhang. Y; Lee. TC; Guillemin. B; Yu. MC; Rom WN. (1993). Enhanced IL-l"beta" and tumor necrosis factor-
"alpha" release and messenger RNA expression in macrophages from idiopathic pulmonary fibrosis or after
asbestos exposure. J Immunol 150: 4188-4196.
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.
7-24
-------
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 pleural abnormalities in representative adult
population sample. Chest 107: 126-131.
Zitting. AJ: Karjalainen. A: Impivaara. O: Kuusela. T: Maki. J: Tossavainen. A: Jarvisalo. J. (1996). Radiographic
small lung opacities and pleural abnormalities in relation to smoking, urbanization status, and occupational
asbestos exposure in Finland. J Occup Environ Med 38: 602-609.
7-25
-------
APPENDIX A. EPA RESPONSE TO MAJOR EXTERNAL PEER-REVIEW AND
PUBLIC COMMENTS
The 2011 External Review Draft (ERD) of the U.S. Environmental Protection Agency's
(EPA's) Toxicological Review of Libby Amphibole Asbestos underwent a formal external peer
review in accordance with EPA guidance on peer review (U.S. EPA, 2006c). In August 2011,
EPA released the assessment for public review and comment, and held a public listening
session on October 6, 2011 in Arlington, VA. In December 2011, EPA's Science Advisory
Board announced a public peer-review meeting on the draft assessment that was held on
Feb 6-8, 2012 in Alexandria, VA. In March 2012, the SAB announced two public
teleconferences of the SAB Libby Amphibole Asbestos Panel to discuss the Panel's draft
review report on May 1 and May 8, 2012. In January 2013, EPA's SAB released the final
report from their review of EPA's draft assessment entitled Toxicological Review of Libby
Amphibole Asbestos (August 2011) (SAB. 2013).
The SAB was tasked with evaluating the following: the accuracy, objectivity, and
transparency of the EPA assessment and the data and methods used to synthesize the scientific
evidence for health hazards. In this Appendix, the specific peer-review recommendations from
the Letter to the Administrator are followed by recommendations from SAB's Response to
EPA's Charge Questions. Individual recommendations from SAB (2013) are quoted verbatim
wherever possible. Page numbers for each quotation are also noted. In some instances, sets of
comments were paraphrased by EPA and so noted.
There were public comments provided directly to EPA on the ERD, as well as public
comments provided to the SAB Libby Amphibole Asbestos panel and the Chartered SAB. This
appendix summarizes the main comments made by the public and responds to those comments.
A letter characterized by its authors as a "Request for Correction" on the draft IRIS assessment
was received by EPA on February 26, 2014
(http://www.epa.gov/OUALITY/informationguidelines/iqg-list.html), with supplemental
information provided on June 25, 2014; many of the previous public comments to the SAB
were included as attachments to this Request for Correction. The response to public comments
addresses the main issues raised in this letter and its supplemental materials.
Section A. 1 responds to the major SAB peer review recommendations to EPA
summarized in the SAB Letter to the Administrator.
Sections A.2 through A.7 respond to more detailed SAB recommendations, with each
section addressing a different general topic. Section A.8 responds to public comments on
specific topics, with each subsection addressing a different general topic. Section A.9 responds
to general public comments on the ERD.
A-l
-------
A.l. MAJOR SAB RECOMMENDATIONS IN SAB LETTER TO THE
ADMINISTRATOR WITH EPA RESPONSES
Major SAB Recommendation Letter #1: [Letter to the Administrator, p. 1] "Localized
pleural thickening is an appropriate health endpoint for the derivation of the inhalation reference
concentration (RfC). It is an irreversible structural, pathological alteration of the pleura and is
generally associated with reduced lung function. The SAB has identified additional references
and recommends that the agency include a more detailed review of the literature to further
support this conclusion."
EPA Response: In response to the SAB's identification of additional references and
recommendation that the Agency include a more detailed review of the literature, EPA
conducted a more detailed review of the literature examining the relationship between
lung function measures and localized pleural thickening (LPT) or pleural plaques. LPT
was introduced as a term in the 2000 ILO guidance. LPT includes plaques on the chest
wall and at other sites (e.g. diaphragm). Plaques on the chest wall can be viewed either
face-on or in profile. A minimum width of about 3 mm is required for an in-profile
plaque to be recorded as present according to the 2000 ILO guidance. The additional
systematic review not only included the additional references noted by the Science
Advisory Board, but comprises a systematic and well-documented literature search and
review of the published literature. This work is presented in Appendix I and discussed in
Section 5.2.2.3.
This additional literature review and analysis demonstrates that pleural plaques and LPT
are associated with a decrease in two key measures of lung function, and that these
decreases are unlikely to be due to other factors such as excess body fat or undetected
changes in lung tissue (other than the pleural plaques) that might have also been caused
by exposure to asbestos. Thus, these additional references and analysis support the
EPA's conclusions in its External Review Draft, and the SAB advice to EPA that LPT is
an appropriate health endpoint for the derivation of the inhalation reference
concentration.
EPA's literature search identified epidemiology studies examining lung function in
asbestos-exposed populations with and without pleural plaques. Twenty studies relating
changes in forced vital capacity (FVC) to the presence of pleural plaques and 15 studies
relating changes in forced expiratory volume in 1 second (FEVi) to the presence of
pleural plaques were included in a meta-analysis.
A meta-analysis of the identified studies conducted by EPA estimated a statistically
significant decrement of 4.09 (95% CI: -5.86, -2.31) and 1.99 (95% CI: -3.77, -0.22)
percentage points respectively in predicted forced FVC and FEVi attributable to the
presence of pleural plaques.
Additional analyses indicated that these decrements are not likely to be due to limitations
in the study designs or conduct, undetected subclinical fibrosis, or misidentification of
pleural plaques due to subpleural fat pads. Further, the extent of plaques was found to
correlate with the degree of lung function decrement, and longitudinal studies indicate
that decrements increase with longer follow-up.
A-2
-------
These findings support the conclusion that pleural plaques and LPT are an appropriate
health endpoint for the derivation of the RfC.
Major SAB Recommendation Letter #2: [Letter to the Administrator, p. 1] "The SAB
supports the derivation of an RfC for LAA based on radiographic evidence of localized pleural
thickening in an occupationally exposed Marysville, Ohio, cohort. However, the SAB
recommends that the EPA conduct additional analyses to substantiate the RfC (to the extent data
permit) of pleural abnormalities using the recently published studies on two other cohorts."
EPA Response: EPA notes that alternative phrasings of this recommendation were
included in the executive summary (p. 1) as well as in the SAB's response to EPA's first
charge question on the Selection of Critical Studies and Effects (see Section 3.2.3.1 of the
SAB Report—p. 14). For clarity, EPA quotes the detailed SAB response on page 14
here:
"Another suggestion for providing support and perspective to the Marysville
findings is to conduct analogous analyses (to the extent the data permit) of pleural
abnormalities among the Libby workers cohort (Larson et al., 2012) and among
the Minneapolis exfoliation community cohort (Alexander et al., 2012; Adgate et
al., 2011). The Libby workers have higher, well characterized occupational
exposures compared to the Marysville cohort. The Minneapolis cohort of
nonworkers generally had estimated exposures at the lower end of the Marysville
cohort but included women and children, thus providing a cohort more
representative of the general population. However, because the Minneapolis
cohort had estimated, not measured exposures, it would not be suitable for the
primary RfC analysis. Similarly, because the Libby workers have both
environmental and occupational exposures, this cohort should not be used for
primary RfC analysis."
As recommended by the SAB, EPA examined two recently published studies of pleural
changes in persons exposed to Libby Amphibole asbestos at their homes in Minneapolis,
MN, and of pleural changes in persons with occupational exposure in Libby, MT
(Alexander et al. (2012): Larson et al. (2012). These studies were evaluated along with
the critical study of pleural changes in persons with occupational exposure in Marysville,
OH (Rohs et al.. 2008).
The evaluation of these studies is summarized in the final assessment in Section 5.2.1 and
a review of the three studies was published in a peer-reviewed journal article in the
Journal of Occupational and Environmental Medicine (Christensen et al., 2013). The
evaluation of these studies (in both the publication and in the assessment) included
examination of various aspects including study population, study design, outcome
evaluation, and exposure characteristics.
All three studies demonstrated that inhalation exposure to LAA is associated with
increased risk of LPT even at the lowest levels of exposure in each study (Christensen et
al., 2013). The results of these three studies provide additional support to EPA's
conclusion that low levels of exposure to Libby Amphibole asbestos is associated with
increased prevalence of LPT.
A-3
-------
EPA evaluated whether the study of residential exposure in Minneapolis could provide
useful information as to whether children or women had a different response to exposure
than did adult men even if the Minneapolis study was not the strongest database for
estimating a benchmark response. However, the overall quality of the exposure
assessment in this investigation and the lack of detail on the various routes of exposure
for men compared to women complicates the evaluation of any effect modification by
gender at this time. Likewise, the data on risks in children were also limited.
The EPA analysis of the Marysville cohort remains EPA's preferred basis for deriving an
RfC; the Marysville cohort had exposure concentrations closer to residential
concentrations in Libby, relatively high-quality exposure estimates, and the ability to
identify the time of first exposure to Libby Amphibole asbestos. In contrast, the
Minneapolis study had more uncertain estimates of exposure than did the study of the
Marysville workers. While the Libby workers had reasonably good estimates of
occupational exposures for workers whose work history information was available, the
occupational exposure levels were higher in Libby than in Marysville. In addition, Libby
workers overall exposure levels included additional residential exposures and data were
not available as to when that residential exposure started. This is a drawback for
modeling the noncancer effects because time since first exposure (TSFE) was determined
to be a very important variable for modeling the pleural changes (see response for
Letter #3 comment, below) and that time of first exposure was unavailable for many of
the Libby workers who were also residents in Libby.
Major SAB Recommendation Letter #3: [Letter to the Administrator, p. 2] "The SAB
recommends that more justification be provided for the selection of the 'best' model for
noncancer exposure-response analysis. The SAB also recommends examining other exposure
metrics besides the simple cumulative exposure, such as time-weighting of exposures. In
addition, more justification is needed for the selection of 10% extra risk as the benchmark
response since it is not consistent with the guideline for epidemiological data in EPA's
Benchmark Dose Technical Guidance (U.S. EPA, 2012)."
EPA Response: In accordance with the SAB recommendation, EPA provides a more
thorough explanation of its selection of the best model for noncancer exposure-response
analysis. EPA examined exposure metrics other than cumulative exposure, such as mean
exposure concentration, and time-weighting of exposures. EPA also provides more
explanation of its selection of 10% extra risk as the benchmark response rate, explaining
how in this case the selection is consistent with EPA''s Benchmark Dose Technical
Guidance (U.S. EPA. 2012).
EPA provides a more thorough explanation of model selection and exposure metrics in
Section 5.2.2.6 and in Appendix E. Following the guidance in the final updated
Benchmark Dose Technical Guidance (U.S. EPA, 2012), EPA explained that there are
several stages of exposure-response modeling. Once the appropriate data set(s),
endpoint(s) and BMR are determined, an appropriate set of statistical model forms is
selected and evaluated for model fit to determine which models adequately represent the
data. Among those models with adequate fit, one or more models are selected to derive a
point of departure for the RfC. Regarding the selection of models to evaluate, the
Benchmark Dose Technical Guidance (U.S. EPA, 2012) notes that additional criteria may
be used, "governed by the nature of the measurement that represents the endpoint of
interest and the experimental design used to generate the data" (page 26). When
A-4
-------
modeling the Marysville data, certain biological and epidemiological features must be
considered, including the nature of the data set, ability to estimate the effects of exposure
and of important covariate(s), the existence of a plateau or theoretical maximum response
rate in a population, and the ability to estimate a background rate of the outcome in a
population.
For the primary modeling in Section 5.2.2.6., EPA selected the Dichotomous Hill model,
(a minor variation on the model proposed in its External Review Draft, the
Michaelis-Menten model) because it allowed fuller consideration of the biological and
epidemiological features described above.
Evaluation of the three exposure metrics considered for the primary analytic data set
(Marysville workers with health evaluations performed in 2002-2005 and hired in 1972
or later) showed that mean exposure consistently led to improved model fit across the
range of model forms evaluated, in comparison with either cumulative or residence
time-weighted exposure (see Section 5.2.2.6).
Time since first exposure (TSFE), which is known from the epidemiological literature to
be an important determinant of LPT risk, was not a significant predictor in this data set.
In order to incorporate TSFE, a "hybrid" modeling approach was taken, as recommended
by the SAB. Here, the effect of TSFE was estimated using a broader subset of the
Marysville workers, with a wider range of TSFE values. This estimated effect of TSFE
was carried over to the modeling performed in the primary analytic data set as a fixed
effect. In this "hybrid" modeling, mean exposure provided adequate goodness of fit,
while cumulative exposure did not. Thus, while the External Review Draft used
cumulative exposure, the primary analysis in the final draft uses mean (occupational)
exposure concentration to derive an RfC.
In an alternative analysis (see Appendix E) that combines data across two health
evaluations (1980 and 2002-2005), EPA selected both the Dichotomous Hill model using
mean occupational exposure concentration and a variant of the Dichotomous Hill model
where TSFE is incorporated into the plateau term (the "cumulative normal" Dichotomous
Hill model). For the cumulative normal Dichotomous Hill model, EPA utilized the
cumulative exposure metric (which was proposed in its External Review Draft) because
of some expectation that it might better reflect the accumulated impact of inhaled
asbestos and because it provided adequate goodness of fit for this particular model form
and data set. As explained in Section 5.2.5, this alternative analysis yielded potential
reference concentrations that ranged from threefold lower than the selected reference
concentration to twofold higher than the selected reference concentration.
EPA considered its choice of a benchmark response and includes a more thorough
explanation of this is Section 5.2.2.5. EPA concluded that a benchmark of 10% extra risk
remains appropriate because LPT represents a persistent, structural change to the pleura,
but is not severe enough to justify a lower BMR. While EPA has sometimes utilized
much lower BMRs when using epidemiology data, that usage is usually in connection
with very large epidemiology studies of cancer endpoints that often have power to detect
small changes in extra risk.
Note that with regards to exposure metrics, the cumulative exposure measure (done on an
annual basis) is the sum, in units of fibers/cc-years, of the work season-specific time-
A-5
-------
weighted concentrations. The mean exposure measure is the cumulative occupational
exposure divided by the duration of occupational exposure. EPA additionally considered
a time-weighted measure, the "residence time-weighted" exposure metric. Here, the
average exposure in each time interval is multiplied by the number of time intervals
elapsed between that exposure and the x-ray evaluation of pleural abnormalities; these
multiplied exposures are then summed across the individual's work history. The
calculation of these exposure metrics is described in Section 5.2.2.6.2 and in Appendix E.
The result of the above changes in model and exposure metric and some other similar
changes is that the RfC in the final assessment is about 4.5-fold higher than the RfC in
the External Review Draft.
Major SAB Recommendation Letter #4: [Letter to the Administrator, p. 2] "A composite
uncertainty factor of 100 was applied to the point of departure to obtain the RfC. EPA applied
an uncertainty factor of 10 to account for human variability and sensitive subpopulations, and a
database uncertainty factor of 10 to account for database deficiencies in the available literature
for the health effects of LAA. The SAB recommends that the EPA reevaluate the use of a
default database uncertainty factor of 10 as part of the consideration of additional studies;
additional data (e.g., Minnesota cohort and data on other amphiboles) might support a lower
value, such as 3, for the database uncertainty factor. In addition, the SAB recommends EPA
revisit its judgement of a subchronic-to-chronic uncertainty factor and a LOAEL-to-NOAEL
uncertainty factor of onefold."
EPA Response: EPA has reconsidered the choice of uncertainty factors (see
Section 5.2.3). In the External Review Draft, EPA did not apply an uncertainty factor (or,
equivalently, divided by an uncertainty factor of 1) to account for adjustment from a
LOAEL to a NOAEL, or to adjust for using subchronic exposure data to estimate a
chronic RfC. An uncertainty factor of 10 was applied to reflect database uncertainty (due
to a limited amount of information on pleural effects after exposure to LAA, and the
potential for autoimmune effects) and an uncertainty factor of 10 was applied to account
for human variability in response.
With respect to adjustment for LOAEL to NOAEL, EPA guidance does not call for such
an uncertainty factor when benchmark dose modeling is used (as it was here) to derive a
confidence interval around an estimate of the concentration associated with an
appropriate benchmark response rate. As explained in response to Recommendation #3,
EPA determined and more thoroughly explained why it concluded a benchmark response
rate of 10% was appropriate, and through exposure-response modeling, determined a
confidence interval on the concentration for that response rate. Hence, EPA did not
change the conclusion from its External Review Draft that an uncertainty factor other
than one is needed for a LOAEL to NOAEL adjustment.
With respect to the adjustment from subchronic data to chronic data, EPA reconsidered
and concluded that a data-informed increase to the uncertainty factor (UFs) value from 1
to 10 was appropriate. Although chronic exposure has been generally defined as more
than approximately 10% of lifetime, the EPA's RfC guidance (U.S. EPA. 1994) states
that for human data"[t]he best data to use for calculating an RfC would be a population
study of humans that includes sensitive individuals exposed for lifetime or chronic
duration, and that evaluates the critical endpoint or an appropriate early marker for the
disease... .However, the amount of exposure in a human study that constitutes subchronic
A-6
-------
is not defined, and could depend on the nature of the effect and the likelihood of
increased severity or greater percent response with duration." This was EPA's
conclusion, despite the fact that the average duration of worker exposure in the key study
was more than 7 years.
Also, the External Review Draft had not modeled the impact of time since first exposure
(TSFE) in the primary modeling analysis that was restricted to workers on whom there
was better exposure data (those hired in 1972 or later), in part because there was less
variation in TSFE in that subcohort. The SAB also recommended that EPA utilize the
full cohort to investigate the impact of TSFE while also using the subcohort hired in 1972
or later to relate exposure to effect. Thus, EPA followed a modeling approach that
provided information on the joint effect of TSFE and exposure concentration and
considered how to use those model results to evaluate the likely impact of additional
follow-up, or TSFE, beyond that observed in the principal study.
EPA concluded that an uncertainty factor of 10 is appropriate because the
exposure-response modeling demonstrated that the range of TSFE in the Marysville
workers may not be of sufficient length to appropriately describe the effects of a lifetime
(i.e., 70 years) of exposure to LAA. EPA performed an analysis on the impact of TSFE,
and found that longer TSFE led to a substantial increase in the risk of LPT (see
Section 5.2.2.6.2), with an approximately 10-fold increase in risk when comparing a
TSFE of 70 years (i.e., a lifetime of exposure) to a TSFE of 28 years (the median in the
primary analytic data set). Based on this analysis, EPA concluded a data-informed
uncertainty factor of 10 is appropriate to reflect lifetime exposure.
In the External Review Draft, EPA had noted the uncertainty in not having lifetime data
in the database uncertainty factor as one factor that contributed to the rationale for a
database uncertainty factor (UFo) of 10. However, the SAB recommended that EPA
consider increasing the uncertainty factor for the duration of study (i.e., the UFs) while
reducing the UFo. EPA concluded that while some database uncertainties remain, there is
a basis to reduce the database uncertainty from 10 to 3. Since the release of the External
Review Draft, two newly published studies provided further information on the pleural
and parenchymal health effects of exposure to Libby Amphibole asbestos (Alexander et
al., 2012; Larson et al., 2012). Both of these studies support the derivation of the RfC
based on pleural effects among Marysville workers. However, some uncertainty remains
regarding autoimmune effects, and consequently, the database UF has been reduced to 3.
With respect to human variability, neither the SAB nor EPA concluded there was a basis
for a change to the uncertainty factor of 10 in EPA's External Review Draft. The
Marysville data (and the Libby data) comprise occupational workers (primarily men)
sufficiently healthy for full-time employment, and thus are not likely to capture the full
range of human responses and potential sensitive subpopulations.
Major SAB Recommendation Letter #5: [Letter to the Administrator, p. 2] "The SAB agrees
that the weight of evidence for LAA supports the descriptor 'Carcinogenic to Humans by the
Inhalation Route' in accordance with EPA's Guidelines for Carcinogen Risk Assessment. The
SAB views the mode of carcinogenic action of LAA as complex, and recommends that the
agency conduct a formal mode of action analysis in accordance with EPA's Guidelines for
Carcinogen Risk Assessment. Based on this formal analysis, the agency may still conclude that
the default linear extrapolation at low doses is appropriate."
A-7
-------
EPA Response: EPA acknowledges that the mode of carcinogenic action of LAA is
complex and multifactorial, and EPA has conducted a formal mode-of-action (MOA)
analysis in accordance with EPA's Guidelines for Carcinogen Risk Assessment in
Section 4.6 of the Toxicological Review. As recommended by the SAB, the focus of this
analysis is LAA, with some discussion of other amphiboles for context when appropriate
literature was available. Further discussion of the mechanistic data in support of the
MOA for asbestos in general has been included in Section 4.4, with the formal
carcinogenic MOA focused on mutagenicity, chronic inflammation, and cytotoxicity for
LAA in Section 4.6. The formal mode of carcinogenic action framework analysis
demonstrated that although evidence is generally supportive of an MOA involving
chronic inflammation or cellular toxicity and repair, there is insufficient evidence to
determine an MOA for LAA. Thus, a linear approach is used to calculate the inhalation
cancer unit risk in accordance with EPA's Guidelines for Carcinogen Risk Assessment.
Section 4.6.2.2 has also been revised to reflect that there are insufficient data to
determine whether a mutagenic mode of action for LAA is supported.
Major SAB Recommendation Letter #6: [Letter to the Administrator, p. 2] "The SAB
supports the selection of the Libby worker cohort for the derivation of the inhalation unit risk
(IUR) and agrees that the use of the subcohort post-1959 for quantification may be reasonable
due to the lack of exposure information for many of the workers in earlier years. The SAB has
suggested sensitivity analyses that would explore the implications of the selection of the
subcohort. The SAB finds it appropriate to use lung cancer and mesothelioma as endpoints for
the derivation of the IUR. The SAB recommends a more detailed discussion and justification of
how the use of mortality data rather than incidence data may have resulted in an undercount of
cases of lung cancer and mesothelioma and what implications, if any, it may have for the
derivation of the IUR."
On Page 19 of the SAB Report (a related more detailed comment): "Use of the
subcohort post-1959 seems reasonable due to the lack of exposure information for many of
the workers in earlier years. Out of 991 workers hired before 1960, 811 had at least one job
with an unknown job assignment and of these 706 had all department and job assignments
listed as unknown. It would seem highly problematic to include workers with limited or no
job information in the model. However, at least some information existed for the remaining
285 workers. The EPA should strengthen the analysis to calculate an overall Standardized
Mortality Ratio (SMR) for the Libby worker full- and subcohorts for lung cancer, using both
Montana and U.S. data for comparison. The later cohort also had lower levels of exposure to
asbestos, which would be closer to the lower levels found in the environment."
EPA Response: Per the SAB recommendation, EPA has added analyses of the Libby
worker full- and subcohorts for lung cancer, using both Montana and U.S. data for
comparison as well as parallel analyses of mesothelioma rates in the Libby worker full-
and subcohorts. Sections 5.4.3.2 and 5.4.3.5 include new tables on the rates of
mesothelioma and related text. New tables on the rates of lung cancer as well as SMRs
and related text are included in Section 5.4.3.3 and 5.4.3.6. Because the rate of lung
cancer mortality in Montana is lower than in the United States as a whole the SMRs
based on Montana rates are somewhat higher. While such computations could not
control for exposure because job history information was largely missing for the early
hires, the rates and risks by categories of duration, age, and TSFE generally appeared to
show similar patterns with highest duration and TSFE having noticeably higher rates.
A-8
-------
Absent similar quality exposure data on the early hires, it is difficult to assess the
potential sensitivity of selecting the subcohort. In addition, EPA's revised
Section 5.4.5.3.1 which compares EPA analyses with other published analyses of the
Libby full cohort and concluded that the risk was not underestimated from the analysis of
the subcohort.
In response to the SAB recommendation, EPA has also provided more detailed
discussion of the use of mortality data rather than incidence data. Because mortality rates
approximate incidence rates when the survival time between cancer incidence and cancer
mortality is short, and median survival for both mesothelioma and lung cancer were less
than 1 year, it is considered to be unlikely that such discrepancies would be significant.
The revised text is shown in Section 5.4.2.2.
Major SAB Recommendation Letter #7: [Letter to the Administrator, p. 2] "The draft
assessment clearly described the methods selected to conduct the exposure-response modeling
for lung cancer and mesothelioma. However, the SAB recommends that the agency provide
more support for its choice of statistical models for the exposure-response analysis. The SAB
also recommends consideration of several models in addition to the Poisson and Cox models
used in the draft assessment."
EPA Response: In response to the SAB recommendation, EPA has provided more
support for its choice of models. EPA has strengthened the presentation of the relative
merits of alternative models, including standard epidemiologic models such as Poisson,
logistic, and Cox, as well as the Weibull model for mesothelioma and two-stage clonal
expansion model for lung cancer. EPA has also enhanced its justification of the selected
models with revised text on models for mesothelioma in Section 5.4.3.1 and for lung
cancer in Section 5.4.3.3. Poisson and Cox models are traditional models that are widely
used in occupational epidemiology cohort analyses. They are well suited to the Libby
subcohort data and have been used by many investigators of the Libby worker cohort in
particular. EPA carefully considered the relative merits of the various alternative models,
noting, for example, that the Weibull model is generally not used for data with rare
outcomes such as mesothelioma, and that EPA did not have available reliable data from
the Libby cohort on which to make assumptions required for use of the two-stage clonal
expansion model. Thus, EPA retained the Poisson and Cox models in the revised
analyses for mesothelioma and lung cancer, respectively.
Major SAB Recommendation Letter #8: [Letter to the Administrator, p. 2] "The agency has
been overly constrained by reliance on model fit statistics as the primary criterion for model
selection. The SAB recommends graphical display of the fit to the data for both the main models
and for a broader range of models in the draft document to provide a more complete and
transparent view of model fit. The SAB also recommends that the EPA consider literature on
epidemiological studies of other amphiboles for model selection for dose-response assessment,
since the size of the Libby subcohort used in the exposure-response modeling is small."
EPA Response: To supplement the evaluation criteria for exposure-response model
selection for the Libby cancer subcohort beyond the use of model-fit statistics alone, EPA
has added graphical displays for a range of models for both mesothelioma (see
Section 5.4.3.5) and for lung cancer (see Section 5.4.3.6) to provide a more complete and
transparent view of model fit. These graphics further support the reasonable nature of the
A-9
-------
selected model for mesothelioma and lung cancer. EPA has also added graphical
displays of model fit for the noncancer analyses.
EPA has considered the epidemiologic literature on other amphiboles and has now
included additional analytic models on amphibole-related mesothelioma (model proposed
by Peto etal. (1982) and its modifications proposed by Berry etal. (2012)). The results
of these models support the selected model in the External Review Draft.
Major SAB Recommendation Letter #9: [Letter to the Administrator, p. 3] "The EPA has
summarized many sources of uncertainty, sometimes quantitatively, as well as the direction and
magnitude of the likely impact of each source of uncertainty. The SAB recommends that model
uncertainty be evaluated by estimating risks using a more complete set of plausible models for
the exposure-response relationship. This sensitivity analysis, while not a full uncertainty
analysis, would make explicit the implications of these key model choices."
EPA Response: With respect to model uncertainty in the cancer exposure-response
analyses, EPA did identify additional uncertainty based on SAB's recommendation to
more fully investigate models suggested by the epidemiologic literature, and this is
discussed in Section 5.4.5.3. EPA estimated risks using literature-based models for
mesothelioma and presented LAA unit risks in Table 5-52 demonstrating twofold
uncertainty around the final IUR value.
Major SAB Recommendation Letter #10: [Letter to the Administrator, p. 3] "Finally, the
SAB has identified critical research needs for epidemiological studies, mode of action, and
measurement methods for LAA to strengthen future LAA assessment."
EPA Response: EPA has conducted the Toxicological Review of Libby Amphibole
Asbestos based on the best available data and literature available at the time of the
assessment. EPA does recognize that ongoing scientific research in the fields of
epidemiology, MO A, and exposure measurement methods will further inform future
assessments of the toxicity and dose response of LAA.
A.2. MINERALOGY - OTHER MAJOR SAB COMMENTS AND
RECOMMENDATIONS WITH EPA RESPONSES:
SAB Mineralogy #1: [Section 3.2.1 of the SAB Report, p. 10]: "In general, the SAB finds that
this section provides an important foundation for understanding the nature of Libby Amphibole
asbestos (LAA) as related to evaluation of potential exposures. There are places where the
clarity and accuracy of the section can be improved, and these are detailed below."
EPA Response: Section 2 of the LAA has been revised for accuracy and clarity.
Additional details concerning the amphibole mineral species have been added to the text
and table along with a discussion of the mining operations and temporal evaluation of the
amphibole content of the ore over the period of mine operation.
SAB Mineralogy #2: [Section 3.2.1 of the SAB Report, p. 10]: "There is a mismatch between
the mineralogical detail embodied in the definition of mineral species and the detail available
relative to specific exposures in Libby. Specifically, mineral species define a very specific
structure (e.g., amphibole) and a specific composition or range of compositions (e.g., winchite or
tremolite). Given that these factors affect a mineral's physical and chemical behavior, they may
in principle be factors to consider for potential hazard. The SAB recognizes that this level of
A-10
-------
detail is not typically available for toxicity studies to allow its application to the evaluation of
LAA per se. In general, however, the observed unique aspects of amphibole asbestos support the
evaluation of LAA through comparison with other amphiboles based on particle morphology and
amphibole designation. Nevertheless, the SAB encourages a rigorous and accurate description of
LAA in Section 2, perhaps while noting the potential ambiguities in the use of mineral-species
names in other studies."
EPA Response: EPA agrees that there is a mismatch between the mineral species
identified in the LAA mixture and the availability of mineral-specific physical and
chemical behavior. EPA has revised Section 2 to reflect the available information on
particle morphology and mineralogy of amphibole asbestos. Unfortunately, of the
mineral constituents identified in LAA and aside from studies of LAA as a mixture, only
tremolite has been investigated in laboratory in vitro and in vivo studies, and it is the only
regulated asbestiform in the LAA mixture. With the exception of magnesio-riebeckite,
which rarely exhibits an asbestiform habit, all of the other constituents (winchite,
richterite, tremolite, magnesio-arfvedsonite, and edenite) can occur in an asbestiform
habit and exhibit similar particle morphologies (diameter, length, and aspect ratios; see
Sections 2.2.3). As further explained in Section 2.4.1, the differences among the calcic,
soda-calcic, and the sodic amphiboles relates to cation ratios (based on the number of
cation atoms per formula unit) for sodium, sodium plus potassium, and aluminum plus
calcium on the [Nas] and [Ca + Nae] site as shown in Figure 2-6. Table 2-1 illustrates
further the similarities between the optical and crystallographic properties of the mineral
species contained in the LAA mixture (see Section 2.4.1). It is not possible with the
LAA mixture to assign a mineral-specific biologic activity to any one of the species or to
assign biologic significance among rather small differences in cation ratios for the
specific minerals. All of the mineral forms in LAA are respirable and all exhibit similar
particle morphologies and there is no published evidence to indicate that there is or is not
a difference in the biologic activity among the LAA mineral species.
SAB Mineralogy #3: [Section 3.2.1 of the SAB Report, p. 11]: "Discussions of mineralogy
and morphology in Sections 2.2.1.1 and 2.2.1.2 are good, with appropriate discrimination
between methods/definitions that are applied to mineral field samples collected from the site
versus terms/definitions that are applied to environmental samples collected via air monitoring
(line 16 of page 2-9 and lines 4 and 5 of page 2-10)."
EPA Response: Section 2.2 has been edited to clarify and correct some of the chemical
formulas and add information concerning particle morphology (see Section 2.2.3).
Additional references have been added and definitions corrected (see Text Box 2-1).
SAB Mineralogy #4: [Section 3.2.1 of the SAB Report, p. 11]: "Section 2.1 is generally
sufficient for providing a background on historical aspects of the mining operations in Libby,
Montana."
EPA Response: Section 2.5 (what was formerly Section 2.1) has been slightly expanded
to include a more complete description of the mining operations at Libby and a
discussion of historical content of amphiboles in the ore mined from Libby.
SAB Mineralogy #5: [Section 3.2.1 of the SAB Report, p. 11]: "Section 2.2 needs
modification. This section should lay a foundation for understanding the nature of Libby
Amphibole (e.g., mineralogical characteristics such as composition and morphology),
A-ll
-------
information on how the material may vary spatially and temporally (with respect to mining
operations), and other factors that may impact exposures. The section does contain much
relevant information. There are parts of the section that are incorrect and misleading;
recommendations to address these issues include:"
SAB comment p. 11: "Consistent use of terminology associated with particle morphology.
The section mixes a number of terms that address particle morphology, and these are
critically important in assessing potential exposures and subsequent impacts. As an example,
'fibers (e.g., acicular...)' implies fibrous and acicular are the same, when in conventional
usage they are different [e.g., see Veblen and Wylie (1993)]. A tight use of terms that are
defined up front should be followed in the EPA document even when a lax use of terms may
exist in the literature cited. A partial attempt is provided in Section 2.2.1.2, but it could be
expanded and carefully vetted with respect to accepted terminology. The four most
important terms to lay out clearly are fibrous, acicular, prismatic, and asbestiform. If the
report's intent is to note differences in these terms, they should be discussed; if the
conclusion is that there are poorly defined distinctions, that topic also should be discussed.
One specific example of inaccurate usage is the term 'prismatic,' which by definition is
'prism'-shaped (meaning parallel sides; it is incorrectly used in multiple places)."
EPA Response: Sections 2.2.3 and Text Box 2-1 have been edited to provide a more
consistent terminology and definitions of particle morphologies. Unfortunately, there are
several definitions for asbestiform, acicular, prismatic, or fibrous morphologies that are
often used in an incorrect context in the published literature. For the purposes of this
assessment, the mineralogical definition is used in the text (Lowers and Meeker, 2002).
According to their report and survey of the literature, there are definitions based on
industrial, interdisciplinary, medical, mineralogical, and regulatory usages and they all
differ. For consistency, throughout the revised document EPA has chosen to use the
mineralogical definitions for clarity and simplicity. A more complete listing of key
definitions can be found in Appendix H of this document (Lowers and Meeker, 2002).
SAB comment p. 11: "Double-check all mineral formulae. There are numerous incorrect
compositions in the report; although some of these may be typographic errors (which, of
course, should be fixed), some may be incorrectly reported. An example of one incorrect
formula is that attributed to vermiculite, which is listed incorrectly as:
[(Mg,Fe,A)3(Al,Si)20io(OH)2-4H20]."
EPA Response: EPA has reviewed and edited Section 2.2.2 to provide correct mineral
formulations in Figure 2-4.
SAB comment p. 11 "Double check that all mineral-species definitions used are accepted
mineralogical standards. Mineral species are fundamental terms that describe a material with
a specific structure and a specific composition or range of compositions; both factors are
primary determinants of a material's properties. Indeed, at the heart of this report is the
definition of likely exposures to (and risks from) inhaled particles and other fibers based on
the use of mineral-species names. The problems in this category are probably most
widespread in Section 2.2.1.1, which details amphibole mineralogy (which is central to the
report). For example, anthophyllite is not a Libby amphibole."
EPA Response: EPA has edited Section 2.2 to correct and use a single mineralogical
definition for particle morphologies in the text. Additions/edits to Sections 2.2.3 through
A-12
-------
2.4.1 and 2.4.2 have added information on atomic differences among the various mineral
species identified in LAA. Additional information on optical and crystallographic
properties of the amphiboles has been added to the text and Table 2-1.
The use of anthophyllite in Section 2.2.1 of the External Review Draft was intended to
illustrate that other amphiboles are referred to as asbestos; it was not intended to imply
that anthophyllite was a constituent in the LAA mineral mixture. It has been replaced
with actinolite in the revised document.
SAB Mineralogy #6: [Section 3.2.1 of the SAB Report, p. 11-12]: "The SAB appreciates the
discussions that highlighted the complexity and variability of LAA in the context of
compositional solid solutions, emphasizing that even the use of mineral-species names for LAA
may mislead readers to believe that LAA is represented by a few discrete materials as opposed to
a mixture of materials with varying compositions. Overall, the mineralogy section could benefit
from some technical editing. It presents some irrelevant material (e.g., Section 2.2.1, which is a
general description of silicate mineral hierarchy), omits some critical information (e.g.,
Section 2.2.1.1 does not provide the mineralogical definitions of key minerals like winchite or
richterite), and presents some erroneous and irrelevant characterizations (e.g., some of the
vermiculite-mineralogy descriptions in Section 2.2.2)."
EPA Response: Section 2.2 has been revised and edited considering the review
comments from the SAB. While the general description of silicate mineral hierarchy may
not be key to understanding LAA mineralogy, it provides a generalized scheme for
structurally related compounds that may occur concomitantly with amphibole asbestos.
The subsection and table describing vermiculite have been removed because the primary
concern of this section is LAA.
Table 2-1 was added to the text in Section 2.4.1 to provide structural formulas and
provide optical and crystallographic properties of the mineral species identified in LAA.
SAB Mineralogy #7: [Section 3.2.1 of the SAB Report, p. 12]: "The report provides a good
summary of available information on the LAA. One specific observation that could be added is
one reported by Sanchez et al. (2008), namely that they observed no correlation between
morphology (fibrous vs. prismatic) and major-/minor-element chemistry. Webber et al. (2008)
similarly concluded that there was no correlation between mineral species and fiber width for
respirable fibers. In other words, this is consistent with the implication that the large set of
compositional data from Meeker et al. (2003) shown in the report reflects the range of
compositions associated with inhaled-fiber exposures."
EPA Response: Section 2.4.2 has been edited to include the observations of Sanchez et
al. (2008) and Webber et al. (2008).
SAB Mineralogy #8: [Section 3.2.1 of the SAB Report, p. 12]: "Discussion on page 2-10
glosses over a serious shortcoming of phase contrast microscopy (PCM); namely, its inability to
detect fibers narrower than -0.25 um. These thin fibers are among the most biologically potent
according to the Stanton-Pott hypothesis. The fact that only a third of the Transmission Electron
Microscopy (TEM)-visible Libby fibers were PCM-visible is buried in (McDonald et al., 1986).
Furthermore, Text Box 2-2 does not adequately contrast the capability of EM versus PCM.
EM's capability to yield elemental composition via Energy Dispersive Spectroscopy (EDS) and
A-13
-------
Wavelength Dispersive x-ray Spectroscopy (WDS) provides information to identify different
asbestos types. PCM, in contrast, cannot even determine if the fiber is mineral. Furthermore, the
Selected Area Electron Diffraction (SAED) capability of TEM allows determination of
crystalline structure, e.g., amphibole versus serpentine. Finally, Box 2-2 incorrectly states that
scanning electron microscopy (SEM) 'produces three-dimensional (3-D) images'. Rather, SEM
produces 2-D images that reveal surface structure of particles."
EPA Response: The description of the analysis of asbestos fibers has been edited and
moved to its own section, Section 2.3. The revised section addresses analysis of bulk
materials (vermiculite and soil) and air filters. The bulk material analysis presents
general methods of polarized light microscopy (PLM) and x-ray diffraction as current
methods for analysis.
The description of the analysis of air samples by PCM and TEM has been edited to
clarify the limitations of current counting methods. PCM analysis of fibers is limited by
the resolution of the light microscope (cannot distinguish fibers <0.25 jim in diameter)
and not all fibers observed on the filter are actual asbestos fibers. The lack of fiber size
resolution may tend to underestimate actual fiber counts because fibers <0.25 jim are not
resolved. The counting rules used for reporting PCM fibers are not regulations—they
merely describe the size and shape of the fibers counted in an optical field. The Text
Box 2-1 has been revised appropriately.
The description of the analysis of air samples using TEM has been edited and expanded.
The discussion of EDS and SAED has been corrected and a discussion of how these
analytical tools are used to identify the mineralogy of specific fibers observed in a grid
field. TEM analysis of mineral fibers is used to confirm fiber analysis by PCM, and one
generally records the total fibers counted on a sample grid and the number of phase
contrast microscope equivalent (PCMe) fibers for assessing human exposure. Both
values are recorded along with fiber size dimensions to gauge fiber size dimension and
distribution. TEM analysis allows the microscopist to determine the mineralogy of a
fiber of interest and to compare the ionic spectrum of the fiber to a known standard,
thereby providing identification of the fiber. Asbestos fibers from the Rainy Creek
complex are unique in having elevated sodium and potassium content in their atomic
structure, which makes their analysis unlike similar amphiboles from other regions
nationally or internationally.
SAB Mineralogy #9: [Section 3.2.1 of the SAB Report, p. 12]: "The electron microscopy
section on page 2-11 could be clarified. SEM and TEM provide higher resolution to allow better
particle morphological analysis. Electron diffraction allows mineralogical assessment. Energy
dispersive x-ray analysis allows elemental composition determination, which can corroborate the
mineralogical determination. X-ray diffraction (XRD) mentioned in this section is useful for
bulk sample mineralogy measurements."
EPA Response: The electron microscopy section in Section 2.3.1 has been corrected
and revised.
A-14
-------
A.3. FIBER TOXICOKINETICS—OTHER MAJOR SAB COMMENTS AND
RECOMMENDATIONS WITH EPA RESPONSES:
SAB Fiber Toxicokinetics #1: Set of Related SAB Comments from p. 16, 20 and 21:
[Section 3.2.3.2 of the SAB Report, p. 16]: "In general, the listing of the laboratory animal
studies in Tables 4-15 and 4-16 and the underlying data summary in Appendix D are
appropriate and complete. However, Tables 4-15 and 4-16 and the summary data in
Appendix D do not include the distribution of fiber lengths, and Section 4.2.5 is therefore
deficient as a summary of animal studies for LAA and tremolite, in terms of not discussing
how the content of long fibers in the administered materials had an influence on the effects
observed."
[Section 3.2.3.2 of the SAB Report, p. 16]: "The report text in Section 4.2.5 also is deficient
in not discussing how the contents of long fibers in the administered materials had an
influence on the effects observed. Therefore, the issue of the influence of fiber dimensions,
and especially fiber length, needs to be strengthened. The LAA fiber dimensions, listed in
Table D-5 (page D-6) should be moved to the main text in Section 4.4, Mechanistic Data and
Other Studies in Support of the Mode of Action. A recent paper by Berman (2011), which
was not cited in the draft report, suggests that cancer risk coefficients for various amphiboles
are more consistent when fiber length was taken into consideration. Berman (2011) also
suggests that the health risks presented by amphibole are greater than those of chrysotile."
[Section 3.2.4.4 of the SAB Report, p. 20]: "It is generally accepted that the toxicity and
carcinogenicity of mineral and synthetic vitreous fibers are governed by fiber dimensions, in
vivo durability, and dose, and that all long amphibole fibers are very durable in vivo. Thus,
the differences in biological potency among the various amphibole fiber types are due
primarily to their differences in dimensions, especially in their fiber length distributions
Berman (2011). The SAB noted that the text in Sections 4.2 and 4.3, and the tables cited
therein, are deficient in not citing all that is known about the dimensions of the administered
fibers."
[Section 3.2.4.4 of the SAB Report, Recommendations, p. 21]: "Areas of needed
improvement in the report include: (1) a discussion on known determinants of fiber toxicity;
and (2) the differences in fiber size distributions between LAA and other known
amphiboles."
EPA Response: EPA revised the assessment to clarify the role of fiber determinants in
toxicity in general (see Section 3) and how the fiber determinants of LAA inform the
toxicity of LAA versus other amphiboles (see Section 4.2-4.4). EPA has moved the
requested text on fiber dimensions from the Appendix D to the main document and
included fiber characteristics for all studies in Tables 4-19 and 4-20 when
available. Further, the EPA has drafted a new section (see Section 3.3) on the
"Determinants of Toxicity" as part of the general description of the toxicokinetics of
fibers, which includes SAB recommended references, including Berman (2011). This
section addresses, in general, the role of fiber toxicity determinants, including length, in
the biological response to fiber exposure. For example, in early studies, fiber length has
been correlated with disease status, with shorter fibers (<2 um) being associated with
asbestosis while longer fibers (>5 um) associated with mesothelioma (Lippmann,
1990). However, more recent studies have also suggested a role for surface area or
A-15
-------
surface chemistry, particularly surface iron, in disease status [reviewed in (Aust et al.,
2011)1. Specific information on fiber characteristics was not available for all studies on
LAA and tremolite, but this information was included in Appendix D and Sections 4.2
and 4.3 in tables for each study when available. A more detailed discussion of the impact
of these determinants of LAA and tremolite in the biological response to these fibers is
included in Sections 4.4 through 4.6.
SAB Fiber Toxicokinetics #2: Set of Related SAB Comments from p. 8 and pp. 12-14:
[Section 3.1.1 of the SAB Report, p. 8]: "SAB has identified sections where extraneous
and repetitive materials could be deleted. For Section 3, since the focus of the draft
document is on Libby amphibole fibers, it would be better to limit the literature reviews and
discussions to those dealing with the family of amphibole fibers. Chrysotile asbestos fibers
are very different from amphibole fibers in terms of their airborne concentration
measurement errors and uncertainties, much lower biopersistence, faster clearance, and
different translocation pathways."
[Section 3.2.2 of the SAB Report, p. 12-13]: "The discussion of general fiber
toxicokinetics is not clear, nor concise, especially since it fails to distinguish between
chrysotile and amphibole fibers. Furthermore, it is inaccurate in many places, as noted
below."
"In view of the fact that the focus of the document is on Libby Amphibole fibers, it
would be better to limit most of the literature reviews and discussions to those dealing
with the various kinds of amphibole asbestos fibers. Chrysotile asbestos fibers, which are
not a significant complication in exposures to Libby vermiculite, are very different from
amphibole fibers in terms of their: (a) airborne concentration measurement errors and
uncertainties (HEI, 1991); (b) much lower biopersistence (Bernstein et al., 2005b;
Bernstein et al., 2005a: Bernstein et al., 2004): and (c) clearance and translocation
pathways and rates (Bernstein et al., 2005b: Bernstein et al., 2005a: Bernstein et al.,
2004)"
[Section 3.2.2 of the SAB Report, p. 13]: "There are some misstatements on fiber
deposition and dosimetry in the document."
"The authors should draw on more authoritative and comprehensive reviews in the
literature [e.g., (Mossman et al., 2011; Lippmann, 2009)]. One misstatement in the draft
is that impaction is affected by fiber length. Another is that interception is affected by
aspect ratio. The document should cite the work by Sussman et al. (199la) and Sussman
et al. (1991b) that demonstrates that interception of amphibole (crocidolite) fibers is only
demonstrably in excess when fiber lengths are >10 jim. Also, the report should cite the
work of Brody and colleagues (Warheit and Hartsky, 1990; Brody and Roe, 1983; Brody
et al., 1981) on chrysotile fiber deposition in the alveolar region in rodents. In terms of
deposition sites, there should be no significant difference between chrysotile and
amphibole fibers."
[Section 3.2.2 of the SAB Report, p. 13]: "Another misstatement is that mucociliary
clearance is complete within minutes or hours rather than the true time frame of hours to a
few days (Albert et al., 1969). The authors also need to acknowledge that particles
depositing in the alveolar region can reach the tracheobronchial tree in two ways: (a) on
A-16
-------
surface fluids drawn onto the mucociliary escalator by surface tension, and (b) by passing
through lymphatic channels that empty onto the mucociliary escalator at bronchial
bifurcations. The report also should acknowledge that macrophage-related clearance of
fibers is only applicable to short fibers that can be fully phagocytosed. Nearly all of the
references to chrysotile in the discussion of translocation should be deleted. The Libby
asbestos fibers are essentially all amphibole fibers, and there is very little commonality
among serpentine and amphibole fibers in terms of translocation or long-term retention."
[Section 3.2.2 of the SAB Report, p. 13]: "There are also toxicokinetic misstatements in
Section 4.2 describing cancer bioassays in animals. The section should cite the inhalation
study of Davis et al. (1985) with fibrous tremolite, which is very similar to Libby amphibole.
Also, this section should discuss the tremolite inhalation study of Bernstein et al. (2003) and
(Bernstein et al., 2005b) that is cited in Table 4-16, as well as the more recent study by
Bernstein et al. (2011) that demonstrated pleural translocation in rats using noninvasive
means following airborne amosite asbestos exposure. The study examined animals for up to
1 year following a short 1-week exposure to amphibole and characterized the size of fibers
that were present in parietal pleura. Noncancer inflammatory pleural changes were
demonstrated associated with fiber translocation. This paper shows rapid translocation of
fibers to the pleura (at least of rodents) and it should be referenced for completeness on
toxicokinetic issues. Furthermore, the results of the various studies cited in Section 4.2 are
almost all very difficult to interpret with respect to the toxic effects that were, or were not,
reported, since no information was provided in Tables 4-15 and 4-16 on the key dosimetric
factor of fiber dimensions. There were comprehensive summaries of available information
on fiber dimensions of materials administered in the bioassays in Appendix D, including
numbers of long fibers, but Section 4.2.5 is deficient as a summary of animal studies for
LAA and tremolite because it does not discuss how the content of long fibers in the
administered materials had an influence on the effects observed."
EPA Response: EPA agrees with this set of SAB comments and has made revisions to
address them. EPA has edited the Toxicokinetics section of the Toxicological Review to
reflect the SAB recommendation to limit discussions to amphibole asbestos in order to
more appropriately focus the discussion on fibers more relevant to LAA. Further, EPA
has corrected any misstatements and included the references requested, as appropriate.
A.4. NONCANCER HEALTH EFFECTS—OTHER MAJOR SAB COMMENTS AND
RECOMMENDATIONS WITH EPA RESPONSES:
SAB Noncancer Health Effects #1: [Section 3.2.3.2 of the SAB Report, p. 16]: "The EPA
draft document discusses the different types of minerals present in LAA and it is uncertain how
the various components relate to adverse health effects. LAA contains -6% tremolite and there
is clear evidence from human and animal studies that tremolite causes adverse health effects in
humans and experimental animals. However, since LAA also contains winchite (84%) and
richterite (-11%), it would be prudent to determine whether these mineral forms contribute to the
adverse health effects of LAA or whether there are interactive effects of winchite or richterite
that modify the toxicity of tremolite. The SAB recommends that this issue be highlighted since
it is well-known that tremolite is highly fibrogenic and causes malignant mesothelioma (MM).
However, the contribution of winchite or richterite to adverse health effects is apparently
unknown."
A-17
-------
EPA Response: The contribution of the individual mineral types present in LAA on
adverse health effects following exposure to LAA is currently unknown. There is limited
information on these components individually, with peer-reviewed publications
examining the role of these individual components on adverse health effects available
only for tremolite. EPA included these studies of tremolite to inform conclusions related
to the mechanisms of action for LAA. EPA has further clarified the purpose of including
tremolite studies in the Toxicological Review of Libby Amphibole Asbestos at the
beginning of Section 4.2. Further discussion of the mineralogy of LAA can also be found
in Section 2. As described by Meeker et al. (2003), LAA is made up of winchite,
richterite, and tremolite. Tremolite makes up less than 10% of the complex mixture that
is LAA. EPA included analysis of in vitro and in vivo studies on tremolite in Section 4.2
and Appendix D. SAB requested clarification as to the purpose of including these
studies, but not any studies of winchite or richterite. It is not known at this time whether
the biological effects of LAA are induced by individual fiber types in the LAA mixture
(i.e., tremolite, winchite, or richterite) or by the complex mixture itself. There is
currently limited peer-reviewed published literature on LAA and on the individual fiber
types in the LAA mixture, particularly in vivo inhalation studies. Because tremolite
makes up a small percentage of the LAA material of interest, information about the
toxicity and carcinogenicity of tremolite may support conclusions related to the
biological response to LAA. At this time, there are no comparable peer-reviewed
published literature on winchite and richterite.
SAB Noncancer Health Effects #2: [Executive Summary of the SAB Report, p. 3]: "The
SAB agrees that the database of laboratory animal and mechanistic studies pertaining to LAA is
appropriately presented in the report and its Appendices for support of its analysis of the human
effects observed. However, the SAB finds the body of the document deficient in not utilizing
what is known about the dimensions of the administered fibers from Appendix D. It is generally
accepted that differences in biological potency among the various amphibole fiber types are due
primarily to differences in dimensions, especially in fiber length distributions. The SAB also
recommends that Section 4.6.2.2 be modified to reflect that there are insufficient data to
determine the mode of action for LAA."
EPA Response: Multiple fiber characteristics, including length, width, and durability,
play a role in the toxicokinetics and toxicity of fibers. While there is extensive literature
on the role of fiber determinants of toxicity relative to adverse health effects for fibers in
general, the studies are often contradictory, making it difficult to draw conclusions for
specific fiber characteristics. However, in response to the SAB recommendations, an
increased discussion of the role of fiber characteristics, including fiber dimensions, in the
biological effects of asbestos has been included in Section 3.3 (Determinants of
Toxicity). In Section 4, discussion of fiber dimensions was included for each study when
available. In general, when information for each study was available, the role of fiber
dimensions individually or cumulatively in the biological response was discussed. This
is discussed in Section 3 for asbestos in general, with further discussion specific to LAA
available in Sections 4.5 and 4.6. Although this information helps to inform MOA
hypotheses for LAA, EPA has concluded, as the SAB notes, there is insufficient
information at this time to reasonably establish a most likely MOA for LAA.
SAB Noncancer Health Effects #3: [Section 3.2.4.4 of the SAB Report, p. 21]: "Section 4.2
should start with a discussion of the relevance of routes of exposure, and then should proceed to
A-18
-------
discuss inhalation data, followed by a discussion of data from other, less relevant routes of
exposure."
EPA Response: The EPA has revised Section 4.2 to include statements on the relevance
of the inhalation route of exposure for studying health effects of fibers, and to discuss the
inhalation data prior to the review of the data from studies that were performed with an
alternate route of exposure. As noted in Section 3, the primary route of human exposure
to asbestos is inhalation. Therefore, studies that expose animals through a pulmonary
route are the most relevant for hazard identification.
SAB Noncancer Health Effects #4: [Section 3.2.4.4 of the SAB Report, p. 20]: "It is generally
accepted that the toxicity and carcinogen!city of mineral and synthetic vitreous fibers are
governed by fiber dimensions, in vivo durability, and dose, and that all long amphibole fibers are
very durable in vivo. Thus, the differences in biological potency among the various amphibole
fiber types are due primarily to their differences in dimensions, especially their fiber length
distributions Berman (2011). The SAB noted that the text in Sections 4.2 and 4.3, and the tables
cited therein, are deficient in not citing all that is known about the dimensions of the
administered fibers."
EPA Response: Information on fiber dimensions has been included when available for
all laboratory animal studies of LAA and tremolite in Tables 4-19 and 4-20. Discussion
of the role of these dimensions in the biological response to fibers is further discussed in
Section 3. For example, in early studies, fiber length has been correlated with disease
status, with shorter fibers (<2 um) being associated with asbestosis while longer fibers
(>5 um) associated with mesothelioma (Lippmann, 1990). However, more recent studies
have also suggested a role for surface area or surface chemistry, particularly surface iron,
in disease status [reviewed in Aust etal. (2011)1. Multiple fiber characteristics, including
length, width, and durability, play a role in the toxicokinetics and toxicity of fibers. As
discussed in Section 3.3, while there is extensive literature on the role of fiber
determinants of toxicity relative to adverse health effects for fibers in general, the studies
are often contradictory, making it difficult to draw conclusions for specific fiber
characteristics.
A.5. CARCINOGENICITY—OTHER MAJOR SAB COMMENTS AND
RECOMMENDATIONS WITH EPA RESPONSES:
SAB Carcinogenicity #1: Set of Three Related SAB Comments From:
1) Section 3.2.4.2 of the SAB Report, p. 18: "A formal mode of action analysis in
accordance with EPA's Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a)
has not been conducted in the draft assessment. The mechanisms by which amphibole
fibers produce malignancy and fibrosis are complex and likely to be multifactorial in
nature. The induction of reactive radical species through persistent interaction of fibers
with target cells, the involvement of chronic inflammatory response, the activation of
certain oncogenes and inactivation of yet-to-be-identified suppressor gene(s), have been
proposed as possible mechanisms. In addition, various in vitro and in vivo studies have
shown that fiber dimensions, surface properties, shape and crystallinity, chemical
composition, physical durability, and exposure route, duration, and dose are important
determinants of the biological potency of fibers."
A-19
-------
"With the LAA, neither the fairly limited amount of research conducted using in vivo as
well as in vitro assays that are described in the review, nor the more extensive body of
published work on other asbestiform minerals, which is also summarized, lead to clear
conclusions as to a single mode of carcinogenic action. The SAB agrees with the EPA
conclusion that the laboratory-based weight of evidence for the mode of action of LAA is
weak. Given the limited database available in the literature and some limited support
from data on carcinogenesis by other amphiboles, the EPA's conclusion that there is
insufficient information to identify the mode of carcinogenic action of LAA may be
justified. However, there are extensive data suggesting multiple mechanisms of
carcinogenic action of other amphibole asbestos fibers (IARC, 2012). The SAB finds
that, given the available information, the default linear extrapolation at low doses may be
appropriate."
2) Section 3.2.4.2 of the SAB Report, Recommendation, p. 18: "A formal mode of action
analysis for LAA should be conducted in accordance with EPA's Guidelines for
Carcinogen Risk Assessment (U.S. EPA, 2005a)."
3) Section 3.2.4.4 of the SAB Report, Recommendation, p. 21: "Section 4.6.2.2 should
be modified to reflect that there are insufficient data to determine if a mutagenic mode of
action for LAA is supported."
EPA Response: Please see response above to Major SAB Recommendation Letter #5.
SAB Carcinogenicity #2: [Section 3.1.1 of the SAB Report, p. 8] "There is inconsistency in
the tone of the conclusions in Section 4.7.1.1 (Lifestage Susceptibility) and in Section 6.3.3
(Applications to Early Lifetime and Partial Lifetime Environmental Exposure Scenarios for IUR)
to either support or refute early life stage susceptibility. The SAB recognizes that no firm
conclusion can be drawn about differential risk of adverse health effects after early life stage
exposure to LAA compared to exposure during adulthood, due to the limited and inconclusive
studies on other forms of asbestos. However, the available limited evidence pointing to excess
risk for exposures during childhood needs to be considered when considering a margin of
safety."
EPA Response: The susceptibility section (see Section 4.7) has been revised to reflect
the current state of the science on susceptibility to fibers, with a focus on the consistency
of the tone and conclusions on the early-life susceptibility to fibers. The weight of
evidence (WOE) 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 are not recommended.
A.6. INHALATION REFERENCE CONCENTRATION (RFC)—OTHER MAJOR SAB
COMMENTS AND RECOMMENDATIONS WITH EPA RESPONSES:
SAB Inhalation Reference Concentration (RfC) #1: [p. 1 Executive Summary] "SAB
recommends additional analyses/cohorts to strengthen and support the RfC since the size of the
Marysville subcohort is small."
EPA Response: As noted above (see response to Major SAB Recommendation
Letter #2), EPA evaluated the two newly available studies of Libby workers and
Minneapolis community residents and have added these results to Section 5.2.1. These
A-20
-------
studies, although not suitable for quantitative analyses for the derivation of the RfC,
qualitatively inform the development of the RfC because they indicate that LAA is also
associated with pleural effects at low levels of exposure. In addition, EPA included
numerous sensitivity analyses to support the RfC (see Section 5.3 and Appendix E).
SAB Inhalation Reference Concentration (RfC) #2: [p. 1 Executive Summary] "In addition to
localized pleural thickening (LPT), the SAB suggests that the EPA consider any x-ray
abnormalities as the outcome: LPT, diffuse pleural thickening (DPT), or asbestosis."
EPA Response: EPA has derived values for chronic RfCs based on "any pleural
thickening" and "all radiographic changes" as a sensitivity analysis of alternative
endpoint definitions in Section 5.2.3. Section 5.2.3 and Appendix E also show PODs for
alternative endpoint definition. The results in Section 5.2.3 for the three endpoint
definitions show equivalent toxicity values. Additionally, EPA included as a sensitivity
analysis a multinomial modeling approach, which simultaneously models all of the
different outcomes (i.e., LPT, DPT, and interstitial changes) in the larger subset of
workers with more recent health evaluations (regardless of hire date; see Section 5.3.5).
SAB Inhalation Reference Concentration (RfC) #3: [p. 1 Executive Summary] "The SAB
also suggests that the EPA conduct analogous analyses (to the extent the data permit) of pleural
abnormalities among the Libby workers cohort and the Minneapolis Exfoliation Community
cohort."
EPA Response: Please see response above to Major SAB Recommendation Letter #2:
[Letter to the Administrator, p. 1].
SAB Inhalation Reference Concentration (RfC) #4: [p. 3 Executive Summary] "With regard
to the exposure metric, the SAB recommends that the EPA reevaluate the raw exposure data and
review pertinent sampling documentation to bolster its use of the geometric mean to represent
the job group exposures, rather than an estimate of the arithmetic mean. The agency should
consider whether a sensitivity analysis using the minimum variance unbiased estimator (MVUE)
of the mean is warranted in the development of the cumulative exposure metric."
EPA Response: In response to this comment, EPA conducted an extensive re-evaluation
of the data and the approach to estimation of job group exposures, as described in
Appendix F. Evaluation of an updated job exposure matrix resulted in a decision to use a
cumulative exposure metric based on the arithmetic mean since this is the method used
for sampling in the field, rather than using the MVUE or some other statistical procedure
to develop the cumulative exposure (CE) metric. The updated industrial hygiene data
were used in these calculations (15 duplicate data points were excluded); use of the
arithmetic rather than the geometric mean resulted in exposure estimates that were
approximately threefold higher. These updated exposure measurements are used to
support derivation of the RfC, and analogous results using the original geometric-mean
based estimates are presented in the uncertainty discussion (see Section 5.3.1).
SAB Inhalation Reference Concentration (RfC) #5: [p. 3 Executive Summary] "EPA's
approach to the primary exposure-response modeling was generally appropriate, but the SAB
recommends that the procedure be refined and the document should provide a clearer description
of how the 'best' model was chosen, in accordance with EPA's 2012 Benchmark Dose Technical
Guidance (U.S. EPA, 2012). Since the Marysville cohort does not support precise estimation of
A-21
-------
the plateau, the EPA should consider fixing the plateau level based on a study of highly exposed
asbestos insulation workers."
EPA Response: Please see response to Major SAB Recommendation Letter #3 for
more detail on how EPA addressed the comment regarding modeling approach and model
selection. With regards to the plateau, EPA reviewed the literature [e.g., see (Winters et
al., 2012; Jarvholm, 1992; Lilis et al., 1991c)1, and in the primary modeling, fixed the
plateau at 85% consistent with a study of highly exposed workers. EPA explored the
impact of this assumption in sensitivity analyses and found that the results were similar to
the primary analysis (see Section 5.3.4 and Appendix E).
SAB Inhalation Reference Concentration (RfC) #6: [p. 4 Executive Summary] "The SAB
suggests examining other exposure metrics besides the simple cumulative exposure, such as
time-weighting of exposures. In addition, the document uses a 10% Extra Risk (ER) as the
benchmark response level (BMR) which is not typically used for human quantal response data.
The SAB recommends that EPA explain what features of the data set or outcome variable led the
agency to choose a BMR that is considerably greater than the norm for epidemiological data."
EPA Response: Regarding exposure metrics, EPA evaluated mean and residence
time-weighted (RTW) exposure metrics, in addition to the CE metric included in the
ERD analyses. The mean exposure metric was found to provide adequate goodness of fit
and the best relative model fit, and was thus carried forward for RfC derivation.
Regarding the BMR selection, please see response to Major SAB Recommendation
Letter #3 and Section 5.2.2.5; EPA followed the EPA's Benchmark Dose Technical
Guidance when selecting a BMR. Briefly, EPA characterized LPT as having the lowest
severity among the available pleural outcomes and thus selects a BMR of 10% extra risk
for this endpoint.
SAB Inhalation Reference Concentration (RfC) #7: [p. 4 Executive Summary] "The SAB
recommends a revised strategy for evaluation of confounders and covariates. Since the quantity
of interest in the analyses of the Marysville cohort is the point of departure (POD), the evaluation
of the various covariates should be made with respect to this quantity. The SAB suggests that
the covariates fall into two classes: exposure-related covariates (various exposure metrics and
TSFE [time since first exposure]) and nonexposure-related covariates (age, body mass index
[BMI], gender, and smoking status). For nonexposure-related covariates, no additional primary
analyses are needed. For exposure-related covariates, the SAB recommends that additional work
be done to refine the models to consider alternative exposure metrics, as well as the inclusion of
TSFE or other time-related variables in the analyses of the full cohort."
EPA Response: The primary modeling to support derivation of the RfC is performed in
the subset of workers with more recent health evaluations and hired in 1972 or later (i.e.,
highest quality exposure information). In this primary data set, EPA evaluated
confounding using both a theory-based method (whether the potential confounder is
associated with both the exposure and with the outcome; see Section 5.2.2.6.1) as well as
a data-based method (including each potential confounder in the final model to assess its
statistical significance; see Section 5.3.3). No evidence of confounding was found in
either case. With regards to TSFE specifically, EPA utilized a larger subset of workers to
estimate the effect of TSFE and included this information in the primary exposure-
response modeling. Comparable modeling of the full cohort is described in Appendix E.
A-22
-------
SAB Inhalation Reference Concentration (RfC) #8: [p. 4 Executive Summary] "The modeled
POD is based on cumulative exposure estimates for the worker cohort examined. The SAB
recommends using the full 70-year lifetime when converting cumulative to continuous exposure
rather than 60 (70 minus the lag of 10 used for exposure in the POD derivation); i.e., do not
correct for the lag of 10 for a 10-year lagged exposure, since the time of disease onset is not
known in prevalence data."
EPA Response: EPA revised its analyses based on SAB comments (see Section 5.2.2),
and as a result, the primary model uses concentration; thus, it does not require the
division by 70 to extrapolate to the full lifetime of 70 years. In the complementary
analyses of the combined data from both health evaluations in Appendix E, analyses
based on CE is divided by 70 (rather than 60) years, as recommended.
SAB Inhalation Reference Concentration (RfC) #9: [p. 4 Executive Summary] "The
uncertainty factors deserve additional consideration and analysis. A composite uncertainty factor
of 100 (an intraspecies uncertainty factor of 10 to account for human variability and sensitive
subpopulations; and a database uncertainty factor of 10 to account for database deficiencies) was
applied to the POD for derivation of the RfC. Although it may be difficult to identify specific
data on LAA to support departure from the default value of 10 for human variability, concern for
the impact on susceptible subpopulations, especially women and children, remains an issue.
Consideration of additional data (Minnesota cohort and data on other amphiboles) might support
a lower value, such as 3, for UFo. In addition, a subchronic-to-chronic uncertainty factor higher
than 1 may be used, given that the mean and maximum exposure duration in the study are well
below the lifetime exposure of interest."
EPA Response: Please see response above to Major SAB Recommendation
Letter #4. In the revised analyses, the data set UF has been reduced to 3, while the
subchronic-to-chronic UF has been increased to 10 based on the evaluation of the role
of TSFE in determining LPT risk (see Section 5.2.2.6.2).
SAB Inhalation Reference Concentration (RfC) #10: [p. 5 Executive Summary] "There also
is concern that the BMR of 10% for a severe endpoint is not reflected by the choice of a
LOAEL-to-NOAEL uncertainty factor (UFL) of 1."
EPA Response: Please see response to Major SAB Recommendation Letter #4. The
LOAEL-to-NOAEL UF was retained at 1 because BMD modeling was used in derivation
of the POD, rather than a LOAEL or NOAEL.
SAB Inhalation Reference Concentration (RfC) #11: [3.1.1 of the SAB Report,
Recommendations, p. 9] "An overall summary set of tables or figures describing the major
cohorts (Libby workers, community, Marysville plant), the types/timelines of exposure, and the
studies associated with each would help orient the readers of the document."
EPA Response: EPA have included a figure (see Figure 4-1) and text discussion
summarizing the studies conducted in the three different locations (Montana, Ohio, and
Minnesota), depicting the type of study population and type of health effect(s) examined.
In addition, a table and text describing the three candidate principal studies (Alexander et
al.. 2012: Larson et al.. 2012: Rohs et al.. 2008) in Section 5.2.1, and more detailed tables
of the demographic characteristics of the Marysville study population are included in
Section 5.2.2.2.
A-23
-------
SAB Inhalation Reference Concentration (RfC) #12: [3.1.1 of the SAB Report,
Recommendations, p. 9] "The draft document could be enhanced with quantitative comparison
of the environmental exposures that have taken place in other geographic regions of the world
(i.e., the Anatolia region of Turkey and Greece) (Metintas et al., 2012; Carbone etal., 2011;
Metintasetal.. 2010; Gogou et al.. 2009; Constantopoulos. 2008; Metintas et al.. 2008;
Sichletidis et al., 2006) with the Libby, Montana, community with regard to airborne tremolite.
This comparison should include numbers of fibers and fiber size distribution in relation to health
effects."
EPA Response: A new Section has been added (see Section 4.1.5: Comparison with
Other Asbestos Studies—Environmental Exposure Settings) that responds to these
suggestions and includes a summary table describing exposure (fiber type, exposure
level, and fiber size, where available) and health effects information for communities
exposed in environmental or residential settings to tremolite or tremolite-chrysotile
mixtures and communities with environmental exposure to crocidolite, another type of
amphibole asbestos. The health effects reported in these studies are consistent with those
documented for workers exposed to commercial forms of asbestos.
SAB Inhalation Reference Concentration (RfC) #13: [3.2.3.1 of the SAB Report, p. 14] "The
rationale for the use of the Marysville, Ohio, cohort for development of the RfC was well
described and scientifically supported. However, there are clear drawbacks to this cohort due to
the lack of exposure sampling prior to 1972 when most of the cohort began work, the use of
self-reported work histories, the end of Libby vermiculite use in 1980, and the mixture of
vermiculite sources used throughout the life of the plant. These drawbacks are offset by the
solely occupational exposure of this cohort, the use of better quality radiographs taken for
research purposes, and the use of 2000 [International Labour Organization] ILO standards for
reading radiographs. The selection of the subcohort for the main analysis has a clear and strong
rationale. (There were 118 workers who began work in 1972 or later when exposure data were
available and who had x-rays from the 2002-2005 exam.) The full cohort of 434 workers was
used for analyses to substantiate the subcohort findings."
EPA Response: EPA acknowledges that the cohort of Marysville workers has both
strengths and limitations, as identified in the SAB's above recommendation. EPA's
primary analysis uses the subset of workers with more recent health evaluations and hired
in 1972 or later to address some of these limitations (e.g., this subset was selected due to
the availability of higher quality exposure information and more recent health
evaluations); this strategy is supported by the SAB recommendation in SAB Inhalation
Reference Concentration (RfC) #17, below. EPA recognizes that the range of TSFE is
limited in this subset; thus, EPA used the larger group of workers with more recent health
evaluations (regardless of hire date) to estimate the effect of TSFE and included this in
the primary modeling (see Section 5.2.2.6.2). In addition, modeling of the full cohort is
described in Appendix E. The potential uncertainty due to the end of Libby vermiculite
use in 1980 and the mixture of vermiculite sources used throughout the life of the plant is
discussed in Section 5.3.1 and Appendix F.
SAB Inhalation Reference Concentration (RfC) #14: [3.2.3.1 of the SAB Report, p. 14]
"Although the SAB agrees that the Marysville subcohort represents the best population upon
which to base the RfC, there was discussion about the need for additional analyses/cohorts to
strengthen and support the RfC since the size of the Marysville subcohort was small. One
suggestion is to use the Marysville cohort but include any x-ray abnormalities as the outcome
A-24
-------
(LPT, diffuse pleural thickening [DPT], or asbestosis). In addition, cause of death might be
assessed for those who died between the two exams. Another suggestion for providing support
and perspective to the Marysville findings is to conduct analogous analyses (to the extent the
data permit) of pleural abnormalities among the Libby workers cohort (Larson etal., 2012) and
among the Minneapolis exfoliation community cohort (Alexander et al., 2012; Adgate et al.,
2011)."
EPA Response: Please see response to Major SAB Recommendation Letter #2 and
SAB Inhalation Reference Concentration (RfC) #2.
SAB Inhalation Reference Concentration (RfC) #15: [3.2.3.1 of the SAB Report, p. 15] "In
addition to localized pleural thickening, the SAB also suggests that the EPA consider looking at
LPT, DPT, and small opacity profusion score together as an outcome. There is evidence that
LPT is not always the first adverse effect that is detected on chest radiographs, and some
individuals with LAA exposure can develop either DPT or increased profusion of small opacities
without developing evidence of LPT. Combining outcomes is appropriate, since DPT and small
opacity profusion also are effects of asbestos exposure and the goal is to define an exposure level
below which LAA is unlikely to have adverse health effects."
EPA Response: Please see response to SAB Inhalation Reference Concentration
(RfC) #2.
SAB Inhalation Reference Concentration (RfC) #16: [3.2.3.1 of the SAB Report,
Recommendation, p. 15] "The SAB suggests the EPA assessment clarify the range of endpoints
that generally can be used to derive an RfC."
EPA Response: EPA included in Section 4 a revised description of the radiographic
endpoints evaluated in the relevant epidemiological studies. In Section 5, the selection of
LPT as the critical endpoint is further explained (see Section 5.2.2.3); in brief, LPT is
most likely to appear sooner after exposure, and at lower levels of exposure, making it
the most sensitive of the available endpoints. In addition, EPA has conducted sensitivity
analyses that included any pleural thickening and any radiographic changes as the critical
effects (see Section 5.2.6).
SAB Inhalation Reference Concentration (RfC) #17: [3.2.3.1 of the SAB Report,
Recommendation, p. 15] "The agency should include a more detailed review of the literature to
support the selection of LPT through detailing the studies that show the relationship between
LPT and both pathologic and physiologic abnormalities, and also risk of other noncancer
asbestos-related diseases."
EPA Response: In response to the specific SAB recommendation, EPA has conducted a
more detailed and comprehensive review of the literature, and performed a meta-analysis
of studies examining the relation between pleural plaques or LPT and pulmonary function
measures. This work is presented in Appendix I as support for the selection of LPT as
the critical effect. This analysis concluded that pleural plaques—and subsequently
LPT—are associated with statistically significant decrements in both forced vital capacity
(FVC) and forced expiratory volume in 1 second (FEVi).
SAB Inhalation Reference Concentration (RfC) #18: [3.2.3.1 of the SAB Report,
Recommendation, p. 15] "In addition to LPT, the document should include an analysis that uses
A-25
-------
all radiographic outcomes (LPT, DPT, and small opacities), recognizing this change may have
little impact on the current analysis."
EPA Response: Please see response to Inhalation Reference Concentration (RfC) #2.
SAB Inhalation Reference Concentration (RfC) #19: [3.2.5.1 of the SAB Report,
Recommendation, p. 22] "Consider sensitivity analyses of additional exposure metrics,
particularly those weighting earlier life exposures more heavily."
EPA Response: EPA has evaluated different exposure metrics, including mean and
RTW exposure metrics (see Section 5.2.2.6.2 and Appendix E.) in addition to the CE
metric included in the ERD analyses. In the subcohort of Marysville workers hired in
1972 or later and evaluated in 2002-2005, the best fitting exposure metric was mean
exposure (C) (see Section 5.2.2.6.2), and this metric was carried forward for primary RfC
derivation. This use of C is a change from the CE metric used in the ERD dated August
2011.
SAB Inhalation Reference Concentration (RfC) #20: [3.2.5.2 of the SAB Report, p. 22] "This
response focuses on the primary analysis of the Marysville subcohort. Additional comments on
the analysis of this cohort can be found in response to Question 4 in Section 3.2.5.4. The SAB
found that the various exposure-response models that were examined were reasonably well
described. However, the SAB recommends a clearer description of how the 'best' model was
chosen. It appears that EPA fits a series of quantal response models, retained models with
adequate fit according to the Hosmer-Lemeshow test (presumably based onp > 0.1, but if so, this
should be stated). Then, among the retained models, the authors selected the model with the
lowest Akaike Information Criteria (AIC). From a statistical standpoint, this methodology can
be justified. However, it is not clear how well aligned it is with the guidance for selection of the
POD in the updated version of EPA's Benchmark Dose Technical Guidance (U.S. EPA, 2012).
Thus the SAB recommends the EPA revise the approach to be better aligned with the Benchmark
Dose Technical Guidance document."
EPA Response: Please see response to Major SAB Recommendation Letter #3.
SAB Inhalation Reference Concentration (RfC) #21: [3.2.5.2 of the SAB Report, p. 23]
"Consistent with the tone of the Benchmark Dose Technical Guidance (U.S. EPA, 2012), the
SAB recommends that a thoughtful approach to model selection be used, including consideration
of biological/epidemiological plausibility, and desirable model features, combined with careful
examination of the data, model fit, and application of the AIC. The SAB highlights the
following points:
1) Model fit (visual comparison of model predictions to data and/or local smoother estimates
from data) in the region of the benchmark response rate (BMR) should play a role in model
selection.
2) The fitted Michaelis-Menten model has an upper plateau of 60% LPT incidence, while a
study of highly exposed asbestos insulation workers reported a prevalence of 85% (Lilis et
al., 1991c). The Marysville cohort does not support precise estimation of the plateau.
Thus, EPA should consider fixing the plateau at a level justified by the literature.
A-26
-------
3) Other exposure metrics besides the simple cumulative exposure, such as time weighting of
exposures, should be considered. The Dichotomous Hill model is attractive because it
allows estimation of an exposure parameter, allowing the exposure effect to scale as
covariates are added, the exposure metric changed, or the plateau fixed."
EPA Response: Please see response to Major SAB Recommendation Letter #3. Based
on the revised model considerations and selection process, the Dichotomous Hill model
was chosen as the primary model for RfC derivation, largely for the reasons outlined by
the SAB (see Section 5.2.2.6.1).
SAB Inhalation Reference Concentration (RfC) #22: [3.2.5.2 of the SAB Report, p. 23] "The
authors explain that their choice of a 10% Extra Risk (ER) as the BMR is in line with the EPA's
Benchmark Dose Technical Guidance. However, that rate is generally considered to apply
specifically to the analysis of quantal data sets from animal studies, which is the context in which
it was developed. In the EPA's Benchmark Dose Technical Guidance, it is mentioned that a
BMR of 1% ER is typically used for human quantal response data because epidemiologic data
often have greater sensitivities than bioassay data. The authors should explain what features of
the data set or outcome variable led them to choose a BMR that is considerably greater than the
norm for epidemiologic data."
EPA Response: Please see response to Major SAB Recommendation Letter #3.
SAB Inhalation Reference Concentration (RfC) #23: [3.2.5.2 of the SAB Report,
Recommendation, p. 23] "Consider model features and balance plausibility, localized fit, and
EPA's 2012 Benchmark Dose Technical Guidance (U.S. EPA, 2012) when choosing the best
model and explain decisions in more detail."
EPA Response: Please see response to Major SAB Recommendation Letter #3.
SAB Inhalation Reference Concentration (RfC) #24: [3.2.5.2 of the SAB Report,
Recommendation, p. 23] "In conjunction with updating and better justifying the primary
analysis, evaluate the impact of different time weightings of the exposure metric."
EPA Response: Please see response to Inhalation Reference Concentration (RfC) #19.
SAB Inhalation Reference Concentration (RfC) #25: [3.2.5.2 of the SAB Report,
Recommendation, p. 23] "Either lower the BMR to be more consistent with common practice
for epidemiological data or provide more justification for the 10% BMR used to calculate the
POD."
EPA Response: Please see response to Major SAB Recommendation Letter #3.
SAB Inhalation Reference Concentration (RfC) #26: [3.2.5.3 of the SAB Report, p. 24] "It is
not clear that the scientific basis of using time since first exposure (TSFE) is well founded. EPA
should consider what TSFE is supposed to be measuring and how it is related to other variables
in the data set (specifically age and exposure). There is some suggestion in the draft document
that in this data set it is a surrogate measure of intensity since people with larger TSFEs would be
more likely to have been exposed to higher levels of LAA present during the early time periods.
This perspective should help identify modeling options."
A-27
-------
EPA Response: TSFE is the time between the first day of exposure and the day of the
most recent health examination, which includes the duration of exposure and any time
after exposure ceases until the day of the health examination. Results in the literature
show that TSFE is a key determinant of prevalence, with prevalence increasing as TSFE
increases [e.g., see (Paris et al., 2009; Paris et al., 2008; Jarvholm, 1992; Lilis et al.,
1991c)1. As discussed in the text in Appendix E, in the full cohort of all Marysville
workers, there is a correlation between TSFE and CE because exposure was not constant
over time but was highest in the early years when vermiculite ore was used. However,
there are individual workers with high CE and low TSFE as well as workers with low CE
and high TSFE, supporting the conclusion that TSFE is a key variable.
SAB Inhalation Reference Concentration (RfC) #27: [3.2.5.3 of the SAB Report, p. 24] "The
SAB also finds that the method for incorporating TSFE into the full cohort analysis is not well
justified. Currently, the EPA uses TSFE as a predictor for the plateau in the Cumulative Normal
Michaelis-Menten model. No biological justification is given for why this maximum proportion
would vary with TSFE."
EPA Response: Upon further analysis in response to SAB comments, EPA is no longer
relying on the Cumulative Normal Michaelis-Menten model in this assessment and
instead has selected the Dichotomous Hill model, a minor variation on the Michaelis-
Menten model, for the primary analysis. For alternative analysis (see Appendix E), EPA
selected both the Dichotomous Hill model using mean occupational exposure
concentration and a variant of the Dichotomous Hill model where TSFE is incorporated
into the plateau term (the "cumulative normal" Dichotomous Hill model). With regard to
use of a cumulative normal model form where TSFE is incorporated into the "plateau"
term, the text has been modified to make clear that this form was evaluated because this
is what plots of the raw data suggested.
SAB Inhalation Reference Concentration (RfC) #28: [3.2.5.3 of the SAB Report, p. 25]
"Improve the scientific justification for using TSFE in the full cohort analysis; this justification
will include an explanation of its meaning in the context of this data set."
EPA Response: As discussed above in response to SAB Inhalation Reference
Concentration (RfC) #26, Appendix E has been revised to incorporate the evidence from
the literature that shows TSFE is an important explanatory variable; many studies show
that prevalence increases as TSFE increases. With regard to use of a cumulative normal
model form where TSFE is incorporated into the "plateau" term, the text has been
modified to make clear that this form was evaluated because this is what plots of the raw
data suggested.
SAB Inhalation Reference Concentration (RfC) #29: [3.2.5.3 of the SAB Report, p. 25]
"Revise the full cohort analysis to change the approach to incorporating TSFE, removing it from
the model of the plateau. As part of the revision, the SAB suggests assessments be made to
determine whether it is appropriate to use (a) the Dichotomous Hill model, (b) TSFE in the linear
predictor alongside cumulative exposure and/or use an alternative exposure metric that explicitly
incorporates TSFE, and (c) the approaches recommended for the subcohort such as a fixed
plateau. As appropriate, such analyses should include assessment of the functional form of
TSFE."
A-28
-------
EPA Response: As described in the response to Major SAB Recommendation Letter
#3, the analysis of the full cohort in the revised assessment evaluates a range of univariate
and bivariate models and a range of exposure metrics including residence-time weighting
that incorporates TSFE (see Appendix E). The analysis has been expanded to include a
parallel detailed evaluation using the "cumulative normal" Dichotomous Hill model
utilizing the cumulative exposure metric and TSFE, as well as the Dichotomous Hill
model based on mean occupational exposure and TSFE, where TSFE is included
alongside the exposure metric in the exponential term.
SAB Inhalation Reference Concentration (RfC) #30: [3.2.5.3 of the SAB Report, p. 25] "The
SAB recommends that the EPA present the lower 95% confidence limit of the benchmark
concentration (BMCL) estimates from a set of reasonable and plausible models, and selections of
data, which will both inform selection of a preferred model and illustrate the range of model
uncertainty."
EPA Response: As discussed in response to SAB Inhalation Reference Concentration
(RfC) #7, EPA's revised analysis of the full cohort includes BMCL values for a wide
range of alternative models, with special emphasis on the cumulative normal
Dichotomous Hill model and the Dichotomous Hill model. Lower 95% confidence limits
on the BMC were included in the presentation of the modeling results, for example, in
Table 5-8.
SAB Inhalation Reference Concentration (RfC) #31: [3.2.5.4 of the SAB Report, p. 25] "The
SAB recommends a revised strategy for evaluation of covariates. The target of inference for the
analyses of the Marysville cohort is the POD, which in this case is the BMCL. The evaluation of
the various covariates should be made with respect to this target of inference. The SAB suggests
the covariates fall into two classes: exposure-related covariates (various exposure metrics and
TSFE) and nonexposure-related covariates (age, body mass index [BMI], gender, and smoking
status). We provide recommended revised strategies for considering these two classes of
covariates that follow directly from consideration of the target of inference."
EPA Response: Please see response to SAB Inhalation Reference Concentration
(RfC) #7.
SAB Inhalation Reference Concentration (RfC) #32: [3.2.5.4 of the SAB Report, p. 25]
"Nonexposure related covariates: A decision on whether to control for the nonexposure-related
covariates should account for how the EPA wishes to determine and apply the RfC. The SAB
suggests a BMCL most directly applicable to all members of the general population is most
appropriate. This implies that the BMCL should be estimated from a model that includes
exposure covariate(s), but that is otherwise unadjusted. This is the same approach used in the
current draft document; only the rationale for the approach is different. The SAB suggests it
would be informative to conduct sensitivity analyses to examine how the BMCL varies across
subgroups defined by covariate values (e.g., older males or smokers). Because the Marysville
subcohort is a small data set, it is difficult to conduct this evaluation exclusively in the subcohort.
Therefore the SAB suggests that the EPA use the full cohort for the model selection and
parameter estimation components of sensitivity analyses incorporating these covariates."
EPA Response: Please see response to SAB Inhalation Reference Concentration
(RfC) #7. There was no evidence that the potential confounders evaluated were
significant predictors of LPT risk after adjusting for exposure to LAA and TSFE. Thus,
A-29
-------
there would be no significant effect modification (i.e., variation in risk across strata) from
these factors.
SAB Inhalation Reference Concentration (RfC) #33: [3.2.5.4 of the SAB Report, p. 26] "For
this activity the EPA would use its selected final model after excluding all exposure variables
(e.g., the Dichotomous Hill model with fixed background, fixed plateau, and after dropping
exposure variables). After fitting a model with a specific set of nonexposure-related covariates
in the full cohort, one can estimate a 'risk score' (i.e., the linear predictor for the
nonexposure-related covariates). This risk score would be included as a single term (as either an
unsealed offset or scaled by its estimated coefficient) in the subcohort analysis. Similar to the
approach presented in Table E-5, these analyses can be used to produce a new table of
subgroup-specific conditional BMCLs; these values will give some evidence of how the target of
inference varies by subgroup. In addition, weighted averages of the conditional BMCLs can be
computed to reflect population average BMCLs for specific covariate distributions in target
populations. For instance, Gavlor et al. (1998) gives a formula for the upper tail of a 95%
confidence interval, and this formula can be extended to obtain BMCLs for weighted averages."
EPA Response: Please see responses to SAB Inhalation Reference Concentration
(RfC) #7and#32.
SAB Inhalation Reference Concentration (RfC) #34: [3.2.5.4 of the SAB Report, p. 26]
"Exposure-related covariates: The inclusion of exposure-related covariates in the model is
fundamental to the inference. The EPA has done excellent preliminary work, and the SAB has
provided recommendations in Sections 3.2.5.2 and 3.2.5.3 of this report about how to revise the
approach. In addition, the SAB recommends that the EPA consider taking several further steps.
First, alternative exposure metrics should be assessed directly in the subcohort data set to
determine whether they fit the data better. In particular, alternative metrics (such as residence
time-weighted exposure) that more heavily weight more distant exposure may be more
biologically plausible because individuals exposed at an earlier age might be more susceptible to
the damaging effects of asbestos. Second, TSFE should be considered for addition to the model.
Since TSFE is complete and equally well estimated across all members of the cohort, the full
cohort can be used to determine how to model this variable. Similar to the approach
recommended for the sensitivity analyses discussed above, this would be done using the model
intended for the subcohort, but omitting exposure variables other than TSFE. Then, the
functional form of TSFE selected using the full cohort can be added to the subcohort analysis,
either as an unsealed offset term or as a scaled covariate. Given biological understanding of the
disease process, for models with both estimated exposure and TSFE included, it would be
appropriate to report the BMCL conditional on a large TSFE."
EPA Response: As recommended by the SAB, EPA investigated alternative exposure
metrics (cumulative, mean, RTW). In addition, EPA used the "hybrid" approach
suggested by the SAB in which the effect of TSFE is estimated in a larger subset of
Marysville workers (those with more recent health evaluations, regardless of hire date)
and this effect is carried over into the primary modeling performed among those workers
with more recent health evaluations and who were hired in 1972 or later (see
Section 5.2.2.6.2). A parallel analysis based on the full cohort of Marysville workers is
provided in Appendix E.
SAB Inhalation Reference Concentration (RfC) #35: [3.2.5.4 of the SAB Report, p. 26]
"Additional comments on covariates: TSFE:
A-30
-------
(1) TSFE deserves careful consideration for both biological and data set-specific reasons. It is an
important determinant of LPT both because individuals' lung tissues exposed at an earlier age
might be more susceptible to the damaging effects of asbestos and because asbestos' effect over
time is increasingly damaging. It is correlated with exposure in this data set since subjects with
the longest TSFE were exposed in the early years of the cohort when exposures were higher. It
is also more accurately estimated than exposure. (2) The SAB does not agree with the use of the
Cumulative Normal Michaelis-Menten model to adjust for TSFE because it makes the
assumption that the TSFE only affects the plateau. This has not been justified biologically or in
the context of features of this particular data set. Instead, the SAB recommends that EPA
consider alternative approaches to account for TSFE."
EPA Response: Regarding: (1) Please see responses to SAB Inhalation Reference
Concentration (RfC) #34. Regarding (2) For the analysis of the full cohort in
Appendix E, EPA investigated a variety of model forms that incorporated TSFE. These
included bivariate log-logistic and bivariate Dichotomous Hill models in which TSFE
was included as an independent predictor of prevalence alongside the exposure metric.
EPA also investigated models in which TSFE was incorporated in the plateau term
(Cumulative Normal Dichotomous Hill and Cumulative Normal Michaelis-Menten).
EPA is no longer relying on the Cumulative Normal Michaelis-Menten model in this
assessment.
SAB Inhalation Reference Concentration (RfC) #36: [3.2.5.4 of the SAB Report, p. 27]
Additional comments on covariates: Smoking:
"(1) Smoking is included in the follow-up by Rohs et al. (2008). However, the ever/never
categorization of smoking is much less informative than the pack-year analysis of smoking used
in the earlier study by Lockey etal. (1984). (2) There is an important discussion of the evidence
linking pleural changes and smoking in Footnote 34 on page 5-6. This information could be
moved into the body of the report, and amplified somewhat. A table summarizing the relevant
studies (irrespective of type of amphibole asbestos) summarizing the evidence regarding the role
of smoking would be useful."
EPA Response: Please see response to SAB Inhalation Reference Concentration
(RfC) #7. Smoking was investigated along with other covariates, but in the revised
analyses, was not found to be a potential confounder and was not significant in the final
model. However, EPA has moved the information from the footnote to the main body of
the text in the sections discussing uncertainty due to potential confounding (see
Section 5.3.3).
SAB Inhalation Reference Concentration (RfC) #37: [3.2.5.4 of the SAB Report, p. 27]
Additional comments on covariates (Gender): "There is little discussion of gender, except in
places where the number of females is listed as too few to analyze in any detail. The SAB did
not regard this as a serious concern because it is reasonable to assume that females and males
have similar probabilities of developing LPT."
EPA Response: Gender was investigated along with other covariates but, in the revised
analyses, was not found to be a potential confounder and was not statistically significant
in the final model (see Section 5.2.2.5.1). EPA agrees with the SAB that risk of LPT is
unlikely to vary greatly according to gender.
A-31
-------
SAB Inhalation Reference Concentration (RfC) #38: [3.2.5.4 of the SAB Report, p 27] "The
SAB recommends that a table be included summarizing the results of the various sensitivity
analyses and how they change the POD."
EPA Response: A section (with a table as suggested) summarizing the sensitivity
analyses has been included at the end of Section 5 (see Section 5.3.6) and in Appendix E.
SAB Inhalation Reference Concentration (RfC) #39: [3.2.5.4 of the SAB Report, p. 27]
"Exposure-dependent censoring: The exposure-dependent censoring discussion is based on
results from Rohs et al. (2008) that inappropriately separated deceased nonparticipants from the
remaining nonparticipants. Once all nonparticipants are combined there is no evidence of
exposure-dependent censoring. Furthermore, exposure-dependent sampling by itself does not
lead to bias in risk estimates. The important issue for bias is whether two individuals with the
same exposure, one diseased and the other not, are equally likely to participate in screening.
There has been no strong rationale presented that would indicate that such differential selection
has occurred in this cohort."
EPA Response: EPA has rewritten the description of this study (see Section 4.1.2.2.2) to
clarify that no exposure-dependent censoring is apparent when combining deceased and
living nonparticipants.
SAB Inhalation Reference Concentration (RfC) #40: [3.2.5.4 of the SAB Report,
Recommendation, p. 27] "Revise consideration of covariates to focus on their impact on the
target of inference.
1) For nonexposure-related covariates, this only alters the presentation; no additional primary
analyses are needed. Sensitivity analyses conditional on subgroups defined by covariates can
be added.
2) For exposure-related covariates, additional work is needed to refine the models to consider
alternative exposure metrics, as well as the inclusion of TSFE or other time-related variables
in analyses of the full cohort. The SAB encourages the EPA to either fully justify analyses
based on the Cumulative Normal Michaelis-Menten model in the context of this particular
data set, or replace them."
EPA Response: Regarding: (1) Please see response to SAB Inhalation Reference
Concentration (RfC) #24. Regarding (2) Please see response to SAB Inhalation
Reference Concentration (RfC) #35. EPA is no longer relying on the Cumulative
Normal Michaelis-Menten model in this assessment.
SAB Inhalation Reference Concentration (RfC) #41: [3.2.5.4 of the SAB Report,
Recommendation, p. 27] "Revise this discussion of Rohs et al. (2008) to make note (perhaps in
a revised table) that the dose distribution in participants is similar to the overall dose distribution
of the original full cohort. Furthermore, revise the discussion of exposure dependent sampling to
distinguish this from bias differential sampling in the sense above."
EPA Response: Please see response to SAB Inhalation Reference Concentration
(RfC) #39.
A-32
-------
SAB Inhalation Reference Concentration (RfC) #42: [3.2.5.5 of the SAB Report,
Recommendation, p. 28] "The SAB recommends EPA indicate more clearly in Section 5.2.3.1.
that 'year' is in the numerator in the exposure metric 'fibers/cc-year,' and to describe more
clearly how cumulative exposure is derived."
EPA Response: The primary model for RfC derivation uses C (fibers/cc). As discussed
in Section 1.1: "For LAA, the RfC is expressed as a lifetime daily exposure in fibers/cc
(in units of the fibers as measured by PCM)."
Although the units of cumulative exposure are written as fibers/cc-year, in the epidemiologic
literature, it actually means fibers/cc times years of exposure and could alternatively be written
as (fibers/cc) x years. Details of how CE estimates were derived are in Appendix F, and
the approach is summarized in Section 5.2.2.1: "In brief, occupational exposure was
estimated for each worker and adjusted to a cumulative human equivalent exposure for
continuous exposure, incorporating adjustments for different inhalation rates in working
versus nonworking time. These adjustments take into account the extensive seasonal
changes in work hours at the Marysville facility (see Appendix F)."
SAB Inhalation Reference Concentration (RfC) #43: [3.2.5.6 of the SAB Report, p. 28] "The
use of a UFn of at least 10 is standard in considering health protective levels based on effects in
the workforce, which is generally healthier and less diverse than the general population. In fact,
publications are available that discuss whether a factor of 10 is sufficient to cover all sensitive
subpopulations, especially children (OEHHA. 2008: Dourson et al.. 2002: Miller. 2002:
Scheuplein et al., 2002: Hattis etal., 1999). Some treatment of the question of interindividual
variability is offered in the later summary of conclusions (see Section 6 of the EPA document).
There is no specific evidence on the relative sensitivity of children to the noncancer effects of
Libby asbestos, although some indications with other amphiboles suggest the possibility of
enhanced effects following exposure at younger ages (Bennett et al., 2008: Isaacs and Martonen,
2005: Hague et al., 1998: Hague et al., 1996). Overall, it seems unlikely that a departure from
the default guideline value of UFn =10 could be justified within the existing guidelines, but
concerns remain for the impact on susceptible subpopulations, especially women and children."
EPA Response: Please see response to Major SAB Recommendation Letter #4. A UF
for intraspecies variability of 10 is used in derivation of the RfC.
SAB Inhalation Reference Concentration (RfC) #44: [3.2.5.6 of the SAB Report, p. 29] "EPA
explains and justifies the selection of a UFo of 10 based on the limited number of studies of
exposure to Libby asbestos (Libby workers, [Agency for Toxic Substances and Disease Registry]
ATSDR community study and Marysville workers) and the lack of evaluation of potentially
more sensitive alternative endpoints. The SAB finds that this uncertainty factor would not be
reduced even if improved exposure estimates allowed consideration of the full cohorts (or a
larger fraction thereof)."
EPA Response: Please see response to Major SAB Recommendation Letter #4.
Briefly, in reevaluating uncertainty factors, EPA applied a UFo of 3, recognizing the
limited number of studies for LAA specifically, but also that LAA has been associated
with autoimmune effects (see Section 5.2.3).
SAB Inhalation Reference Concentration (RfC) #45: [3.2.5.6 of the SAB Report, p. 29]
"However, some additional data have recently been published for the community surrounding a
A-33
-------
Minnesota expansion plant (Alexander et al., 2012; Adgate et al., 2011). Although there appears
to be a rationale for at least an initial consideration of LAA as a unique material (to provide an
unbiased comparison with other amphiboles), this SAB review has identified very substantial
grounds for considering this material as having composition, physical properties, and biological
effects that are very similar to those seen for other amphiboles. The most relevant comparison
would be to tremolite, since Libby Amphibole is -6% tremolite, an amphibole that is known to
cause cancer and noncancer effects in human populations. However, it is uncertain how other
components of Libby Amphibole (richterite and winchite) interact as a mixture with tremolite to
modify toxicity. This consideration of data on other amphiboles is particularly pertinent to
discussions of the mode of action, as well as the exposure-response relationships, for Libby
Amphibole. In light of this similarity it appears reasonable, and indeed necessary, to at least
debate the question of whether the available data on noncancer health effects of amphiboles are
sufficient to mitigate the acknowledged data shortage for Libby Amphibole itself. Therefore, the
SAB considers that additional data (e.g., the Minnesota cohort and data on other amphiboles)
might support a lower value, such as 3, for UFo."
EPA Response: In EPA's revised assessment, a UFo of 3 was selected; please see
response to Major SAB Recommendation Letter #4 and Section 5.2.3 for more details.
SAB Inhalation Reference Concentration (RfC) #46: [3.2.5.6 of the SAB Report, p. 29] "On
the other hand, there are substantial remaining uncertainties that are not addressed by these
additional data, including those raised by consideration of the severity of the endpoint and the
selection of the BMR (see below). This uncertainty should also be revisited by EPA in its
judgement of an uncertainty factor of onefold for a LOAEL-to-NOAEL uncertainty factor
(UFL). It can also be argued that a subchronic-to-chronic uncertainty factor (UFS) higher than 1
should be used, given that the mean and maximum exposure duration in this study are both well
below the lifetime exposure of interest. This uncertainty should also be revisited for EPA in its
judgement of an uncertainty factor of onefold for UFs."
EPA Response: Please see response to Major SAB Recommendation Letter #4 for
more details. In the reevaluation of uncertainty factors, EPA retained a UF of 1 for
LOAEL-to-NOAEL uncertainty, but increased the subchronic-to-chronic uncertainty
factor to 10.
SAB Inhalation Reference Concentration (RfC) #47: [3.2.5.6 of the SAB Report, p. 29] "It
may be appropriate for EPA to select a value of 10 for UFo, or a similar uncertainty spread
across several factors, but EPA needs to reevaluate selection of this factor explicitly once all the
additional information has been incorporated in the discussion."
EPA Response: In reevaluating uncertainty factors, EPA selected a UFo of 3; for more
details, please see Section 5.2.3 and the response to Major SAB Recommendation
Letter #4.
SAB Inhalation Reference Concentration (RfC) #48: [3.2.5.6 of the SAB Report,
Recommendation, p. 30] "Review additional data, in particular the exposure-response
relationship for noncancer endpoints in the Minneapolis community cohort."
EPA Response: Please see response to Major SAB Recommendation Letter #2 for
more details; briefly, because of lack of TSFE data, the Minneapolis community cohort
could not be used for exposure-response.
A-34
-------
SAB Inhalation Reference Concentration (RfC) #49: [3.2.5.6 of the SAB Report,
Recommendation, p. 30] "Determine whether this new analysis supports the existing analysis
based on the Marysville data, and if so whether this warrants reduction of the value of UFo since
the limited data basis for the original analysis has been expanded."
EPA Response: Please see response to Major SAB Recommendation Letter #4.
SAB Inhalation Reference Concentration (RfC) #50: [3.2.5.6 of the SAB Report,
Recommendation, p. 30] "Reassess the selection of the BMR to reflect the severity of the
chosen endpoint in the Marysville cohort and the precision available in the data. Whether or not
the chosen BMR is changed, present this analysis in the document rather than simply asserting
that a 'default' value for the BMR was chosen. Similar consideration should be applied to the
Minneapolis cohort to provide a valid comparison. This consideration needs to be linked to
discussion of the selection of a value for UFL as noted below."
EPA Response: Please see response to SAB Inhalation Reference Concentration
(RfC) #6. In brief, EPA clarified the selection of the BMR in Section 5.2.2.5 and
selected the BMR of 10% extra risk based on the characterization of LPT as having the
lowest severity among available pleural outcomes.
SAB Inhalation Reference Concentration (RfC) #51: [3.2.5.6 of the SAB Report,
Recommendation, p. 30] "Review additional sources of uncertainty:
1) Timescale of cohort coverage, normally addressed by UFs if this is a significant concern
rather than including this as a component of UFo which already has several major issues to
account for.
2) Additional uncertainty resulting from target population diversity (including women and
children, specific subpopulations of concern not represented in the cohort), and endpoint
severity."
EPA Response: Please see response to Major SAB Recommendation Letter #4.
With respect to human variability, neither the SAB nor EPA concluded there was a basis
for changing the uncertainty factor (UFn) of 10 in EPA's External Review Draft. The
Marysville data (and the Libby data) comprise occupational workers (primarily men)
sufficiently healthy for full-time employment, and thus are not likely to capture the full
range of human responses and potential sensitive subpopulations.
Finally, with respect to database uncertainty (UFo), EPA concluded that, while some
database uncertainties remain, there is a basis to reduce the database uncertainty from 10
to 3. Since the release of the External Review Draft, two newly published studies
provide further information on the pleural and parenchymal health effects of exposure to
Libby Amphibole asbestos (Alexander et al., 2012; Larson et al., 2012). Both of these
studies support the derivation of the RfC based on pleural effects among Marysville
workers and thus support a reduction in the UFo. However, some database uncertainty
remains regarding autoimmune effects, and consequently, the database UF has been
reduced to 3.
A-35
-------
SAB Inhalation Reference Concentration (RfC) #52: [3.2.5.7 of the SAB Report, p. 30] "In
the report there are two sections on uncertainty for the RfC: an application of uncertainty factors
following standard EPA practice (see Section 5.2.4), and a discussion of the uncertainties in the
overall methodology and approach (see Section 5.3). This response focuses on the latter.
Overall the SAB found the discussion to be thorough, detailed, and logical. The document can
be improved by harmonizing the full set of uncertainty discussions, including both the discussion
of RfC uncertainty and the related discussion of the IUR uncertainty (see the SAB response to
question 5 under Section 3.2.6.5 below). In addition, the RfC uncertainty assessment can be
strengthened. A key consideration of any assessment is whether the estimated RfC is adequately
protective of public health. The SAB recommends that additional work be done to substantiate
the RfC estimate through additional sensitivity analyses and discussion of results and insights
from other data sets (e.g., cause of death for the deceased nonparticipants in Rohs et al. (2008)
and the Minneapolis exfoliation community cohort (Alexander et al., 2012))."
EPA Response: EPA included numerous sensitivity analyses to address issues regarding
the exposure metric, assumptions in exposure assignment, model form and assumptions,
and the effect of covariates. These are described in the sections on uncertainty in
Section 5.3 and in Appendix E.
SAB Inhalation Reference Concentration (RfC) #53: [3.2.5.7 of the SAB Report, p. 30] "In
considering other studies, the appropriate assumption is that LAA fibers have the same
mechanisms of toxicity and quantitative risk relations as that of other asbestos fibers. In
sensitivity analyses, consider alternative exposure metrics (prioritizing residence time-weighted
metrics and excluding exposures after 1980), methods to fine-tune the RfC estimate from the
subcohort (particularly fixing rather than estimating the plateau, allow the slope parameter to be
estimated, use a lifetime of 70 regardless of the exposure metric), and added sensitivity analyses
in the full cohort using suggestions from the SAB."
EPA Response: Please see response to SAB Inhalation Reference Concentration
(RfC) #21. The primary model for RfC derivation is the Dichotomous Hill model with
plateau fixed at 85%, as suggested by the SAB.
SAB Inhalation Reference Concentration (RfC) #54: [3.2.5.7 of the SAB Report,
Recommendation, p. 31] "Harmonize the uncertainty discussions across the document."
EPA Response: EPA has made revisions to provide greater harmonization in the
discussion of uncertainty for cancer and noncancer effects (see Sections 5.3 and 5.4.6).
The uncertainty analyses pertaining to the derivation of the RfC are summarized in
Table 5-17 and indicate that the uncertainty in the POD due to the factors examined
(uncertainty in the exposure reconstruction, in the radiographic assessment of the critical
effect, from potential confounding, in the effect of TSFE, in the endpoint definition, and
in the choice of critical effect) is less than an order of magnitude.
SAB Inhalation Reference Concentration (RfC) #55: [3.2.5.7 of the SAB Report,
Recommendation, p. 31] "Substantiate the RfC estimate through
1) Additional sensitivity analyses of the subcohort;
2) Discussion of results from other studies;
A-36
-------
3) Additional sensitivity analysis of the full cohort; and
4) Summarizing in tabular form the results of the various sensitivity analyses and model
alternatives, to show how they affect the POD."
EPA Response: Please see response to SAB Inhalation Reference Concentration
(RfC) #21.
A.7. INHALATION UNIT RISK (IUR)—OTHER MAJOR SAB COMMENTS AND
RECOMMENDATIONS WITH EPA RESPONSES:
SAB Inhalation Unit Risk (IUR) #1: [Overall Clarity, SAB Section 3.1.1, p. 8] "A table
comparing these results with the results from the earlier 1988 EPA analysis (U.S. EPA, 1988) on
asbestos would be helpful."
EPA Response: Section 1.1.1 describes the IRIS Assessment for Asbestos (U.S. EPA,
1988) with specific results in Table 1-1, which can be compared with the results of the
current assessment.
SAB Inhalation Unit Risk (IUR) #2: [Selection of Critical Study and Endpoint, SAB
Section 3.2.4.3, p. 20] "Tables 5-6 and 5-8 are mis-titled, since the tables include the number of
deaths from mesothelioma and lung cancer as well as demographic and exposure data. The titles
should either be changed and additional causes of death included in the tables or new tables
should be created that focus on the causes of death. Provision of data on other major categories
of mortality, including numbers of [chronic obstructive pulmonary disease] COPD,
cardiovascular, colorectal cancer, and other cancer deaths, could provide useful information on
the representativeness of the mortality experience of these cohorts."
EPA Response: The corresponding tables have been amended to include additional
information on mortality from other causes and the titles have been changed. The new
tables are titled:
Table 5-Cancer-l (ERD Table 5-6). Demographic, mortality, and exposure
characteristics of the Libby worker cohort
Table 5-Cancer-3 (ERD Table 5-8). Demographic, mortality, and exposure
characteristics of the subset of the Libby worker subcohort hired after 1959.
SAB Inhalation Unit Risk (IUR) #3: [IUR Exposure-response Modeling, SAB
Section 3.2.6.1, p. 33] "Poisson regression analyses: the mathematical form of the regression
function should be given, and discussion of whether the potential for over-dispersion was
assessed."
EPA Response: The mathematical form has been provided as Equation 5-8. A
discussion of the possibility of overdispersion (when the variance exceeds the mean in a
Poisson distribution) has been included in Section 5.4.3.1 with results shown in
Sections 5.4.3.2 and 5.4.3.5 indicating a lack of evidence for overdispersion in either the
full cohort of all workers or the subcohort.
A-37
-------
SAB Inhalation Unit Risk (IUR) #4: [IUR Exposure-response Modeling, SAB
Section 3.2.6.1, p. 33] "Cox proportional hazards modeling: the reasons should be given for not
conducting a Bayesian analysis as was done for the Poisson regression model for mesothelioma."
EPA Response: EPA has clarified the reasoning in Section 5.4.3.3. The revised
language is excerpted here: "While the Poisson model is appropriate for modeling very
rare events, the standard form does not allow for inclusion of the time-varying nature of
exposure. Lung cancer is more common than mesothelioma and does have a known
background risk. Thus, modeling of lung cancer mortality is based on the relative risk
rather than the absolute risk and was conducted in a frequentist framework, which is the
standard methodology for epidemiologic analyses. A frequentist framework is an
alternative method of inference drawing conclusions from sample data with the emphasis
on the observed frequencies of the data."
SAB Inhalation Unit Risk (IUR) #5: [IUR Exposure-response Modeling, SAB
Section 3.2.6.1, p. 33] "Life-table analysis: the method used to estimate the hazard function for
the exposed population should be clearly spelled out in the text. Was it based on a
nonparametric estimate of the baseline hazard from the subcohort? Given that the SEER data
were used to calculate the background incidence of lung cancer, it would seem more appropriate
to use those data to estimate the baseline hazard and then to use the regression coefficient
obtained from the Cox model applied to the subcohort data to obtain the hazard of the exposed
group. Thus, the reasons for not using the SEER data to estimate the baseline hazard should be
explained."
EPA Response: EPA has clarified that lung cancer hazard function is based on the
nonparametric estimate of the baseline hazard from the subcohort, which was then
applied to the background mortality rates for lung cancer from SEER. Given the
potential for historical differences in the Libby subcohort compared with the U.S.
population (i.e., the potential for cohort effects), EPA prefers to estimate the hazard on
internal comparisons. As for projecting the expected disease burden going forward, EPA
believes the observed hazard rates are best applied to more recent background rates. EPA
has revised the description of the life-table analysis generally in Appendix G and
Section 5.4.1 as well as specifically for mesothelioma in Section 5.4.5.1 and for lung
cancer in Section 5.4.5.2.
SAB Inhalation Unit Risk (IUR) #6: [IUR Exposure-response Modeling, SAB
Section 3.2.6.1, p. 33] "Expand the discussion of model selection to explain the reliance on
model fit criteria for model selection. In particular, why should the broader epidemiologic
evidence on the time course of disease not argue at least for the presentation of more than one
statistical model?"
EPA Response: As described in the previous response to Major SAB
Recommendation Letter #7, EPA has strengthened the presentation of the relative merits
of alternative models and enhanced its justification of the selected models with revised
text on models for mesothelioma in Section 5.4.3.1 and for lung cancer in Section 5.4.3.3.
SAB Inhalation Unit Risk (IUR) #7: [IUR Exposure-response Modeling, SAB
Section 3.2.6.1, p. 34] "In a tabular form, summarize the fit results, POD estimates, and IUR
estimates from the full range of models considered in order to show the dependence of the IUR
estimate on model selection."
A-38
-------
EPA Response: Section 5.4.3.5 includes several new tables and figures summarizing the
fit results along with the unit risk estimates for mesothelioma, lung cancer, and the
combined IUR (see Section 5.4.5.3) to show the dependence of the IUR estimate on
model selection.
SAB Inhalation Unit Risk (IUR) #8: [IUR Exposure-response Modeling, SAB
Section 3.2.6.1, p. 34] "Present the fit to data graphically for both the main models and for a
broader range of models, including the Peto model. This step would provide a more thorough
and transparent view of fit, particularly in the region of the BMR, than is allowed by examining
summary statistical values alone."
EPA Response: New graphical presentations of model fits for mesothelioma, including
the Peto model, are shown in Section 5.4.3.5, and model fits for lung cancer are shown in
Section 5.4.3.6.
SAB Inhalation Unit Risk (IUR) #9: [IUR Exposure-response Modeling, SAB
Section 3.2.6.1, p. 34] "Provide in an appendix the details of the Nicholson/Peto model fit for
which the text currently states 'data not shown'."
EPA Response: Details of the Peto model fit are included in Section 5.4.3.5, which
includes additional results and descriptions of model fit, including new tables and figures.
SAB Inhalation Unit Risk (IUR) #10: [IUR Exposure-response Modeling, SAB
Section 3.2.6.1, p. 34] "Allow evaluation of the time dependence of disease by providing
tabulations of mesothelioma mortality rates and lung cancer SMRs by time since first exposure,
duration of exposure, and period of first exposure (for both the full and subcohorts of Libby
workers)."
EPA Response: As noted in a previous response, EPA has added the recommended
analyses of Libby worker full- and subcohorts for lung cancer, using both Montana and
U.S. data for comparison, as well as parallel analyses of mesothelioma rates in the Libby
worker full- and subcohorts. New tables on the rates of mesothelioma are shown in
Sections 5.4.3.2 and 5.4.3.5. New tables on the rates of lung cancer and SMRs are shown
in Sections 5.4.3.3 and 5.4.3.6.
SAB Inhalation Unit Risk (IUR) #11: [IUR Exposure-response Modeling, SAB
Section 3.2.6.1, p. 34] "Evaluate the feasibility of conducting an ancillary analysis of the full
Libby data set, including hires before 1959, using interval statistics or other traditional censoring
methods (not simple midpoint substitution). At a minimum, discuss the possible quantitative
uncertainties associated with using the smaller subcohort."
EPA Response: New tables on the rates of mesothelioma and the rates and SMRs for
lung cancer included all workers regardless of hire data as well as for those workers hired
after 1959. The statistical tradeoff and possible quantitative uncertainties associated with
using the smaller subcohort were discussed in Sections 5.4.3.4 and 5.4.6. These
quantitative uncertainties included the lower number of cases of both cancers than in the
whole cohort, the shorter follow-up time period for the subcohort, and the overall lower
mortality rate due to the subcohort being younger. EPA carefully considered the SAB
recommendation to use interval statistics or other traditional censoring methods and
reviewed the references provided by SAB. EPA concluded that the use of the subcohort
A-39
-------
was most appropriate for quantitative analyses, particularly due to the availability of
specific work histories and the higher percentage of exposure assignments based on
actual measurements as opposed to missed values.
SAB Inhalation Unit Risk (IUR) #12: [IUR Exposure-response Modeling, SAB
Section 3.2.6.2, p. 34] "The numbers of COPD deaths (n) in the subcohort that were the basis for
the analysis should be presented in the text."
EPA Response: The number of COPD deaths used in the analysis is shown in
Section 5.4.2.4.
SAB Inhalation Unit Risk (IUR) #13: [IUR Exposure-response Modeling, SAB
Section 3.2.6.2, p. 35] "The statements about the evidence against confounding by smoking
given by restriction of the cohort should be qualified by the assumptions required to justify them,
or deleted."
EPA Response: The statements have been further qualified. The following text is
shown in Section 5.4.3.4. "Thus, this restriction in the time period of hiring may make
the cohort members more similar to each other, thereby possibly reducing the potential
impact of any smoking-related confounding."
SAB Inhalation Unit Risk (IUR) #14: [IUR Exposure-response Modeling: p. 35] "The SAB
had no recommendations for further analyses" [with respect to the potential for lung cancer to
confound risks of smoking in this cohort].
EPA Response: EPA accepts the SAB recommendations for no further analyses relevant
to the potential for confounding of lung cancer risks by smoking.
SAB Inhalation Unit Risk (IUR) #15: [IUR Exposure-response Modeling, SAB
Section 3.2.6.2, p. 35] "The reference to three methods is confusing. There are actually only
two, the restricted cohort and the Richardson analysis for which two exposure metrics are
explored."
EPA Response: The discussion in Section 5.4.3.8 now refers to two methods, as noted
in the recommendation.
SAB Inhalation Unit Risk (IUR) #16: [IUR Exposure-response Modeling, SAB
Section 3.2.6.3, p. 35] "The EPA should acknowledge that the assumption of independence is a
theoretical limitation of the analysis, and should provide a fuller justification for this assumption.
EPA has cited the NRC (1994) analysis as suggesting the impact of this issue is likely to be
relatively small. This view is also echoed in U.S. EPA (2005a) Guidelines for Carcinogen Risk
Assessment. These provide the basis for a default assumption. However, it would be preferable
if this assessment discussed the evidence base and rationale for lung cancer and mesothelioma
specifically."
EPA Response: EPA has acknowledged the assumption of independence in
Section 5.4.5.3, and the revised text follows:
"It is important to mention here that the assumption of independence above is a
theoretical assumption, as there is no data on independence of mesothelioma and cancer
risks for LAA. However, in a somewhat similar context of different tumors in animals,
A-40
-------
NRC (1994) stated: '... a general assumption of statistical independence of tumor-type
occurrences within animals is not likely to introduce substantial error in assessing
carcinogenic potency.' To provide numerical bounding analysis of impact of this
assumption, EPA used results of Chiu and Crump (2012) on upper and lower limits on
the ratio of the true probability of a tumor of any type and the corresponding probability
assuming independence of tumors. The lower limit is (1 - min\pl,p2]) + (1 —pi x p2)
and upper limit is min(l,2 —pi —p2) + (1 —pi x p2). Substituting/?/ = risk of lung
cancer = 0.040 andp2 = risk of mesothelioma = 0.075, the lower limit is 0.963 and the
upper limit is 1.003 (a value of 1.0 indicates independence). Because lower and upper
values are both very close to the value of 1.0, this demonstrates that the assumption of
independence in this case does not introduce substantial error consistent with what NRC
(1994) has stated."
SAB Inhalation Unit Risk (IUR) #17: [IUR Exposure-response Modeling, SAB
Section 3.2.6.3, p. 35] "As a sensitivity analysis, the EPA should consider quantitatively
accounting for dependence in the risks of mesothelioma and lung cancer mortality either using a
method that models the dependence explicitly, or a bounding study that evaluates the numerical
consequences of the assumption of independence."
EPA Response: As noted in the response above, EPA has provided a numerical
bounding analysis to estimate the consequences of the assumption of independence. As
explained in response to the preceding comment, in this analysis the assumption of
independence does not introduce substantial error.
SAB Inhalation Unit Risk (IUR) #18: [IUR Exposure-response Modeling, SAB
Section 3.2.6.5, p. 37] "The SAB recommends that a more straightforward and transparent
treatment of model uncertainty would be to estimate risks using a more complete set of plausible
models for the exposure-response relationship (discussed in response to Question 1 in
Section 3.2.6.1), including the Poisson models. This sensitivity analysis would make the
implications of these key model choices explicit."
EPA Response: EPA's standard practice is to investigate several modeling options to
determine how to best empirically model the exposure-response relationship in the range
of the observed data as well as to consider exposure-response models suggested in the
epidemiologic literature. For lung cancer, a new discussion of potential alternative
models has been included in Section 5.4.3.3, including Poisson, logistic, Cox, and
multistage clonal expansion models. EPA selected the Cox model as the most
appropriate model for exposure-response modeling based on the suitability of this model
to the nature of the data set (e.g., time-dependent exposure information), the long history
of this model usage in analyses of occupational cohorts, and the commonality of usage in
other epidemiologic analyses of the Libby workers cohort. EPA's evaluation of
alternative approaches found no other standard epidemiological model formulations that
allow for the analysis of time-varying exposures in the manner achieved by the Cox
proportional hazards model.
For mesothelioma, a new discussion of alternative models has been included in
Section 5.4.3.1, including consideration of approaches such as parametric survival
models. EPA concluded that the Peto model and variations of the Peto allowing for
potential clearance are well supported in the epidemiologic literature. The Poisson model
A-41
-------
is an appropriate model for rare data. There are no examples of using other models for
modeling mesothelioma in similar situations.
EPA presents results for sensitivity analyses that were conducted for both mesothelioma
and lung cancer mortality in deriving combined inhalation unit risk in Section 5.4.5.3.
SAB Inhalation Unit Risk (IUR) #19: [IUR Exposure-response Modeling, SAB
Section 3.2.6.5, p. 37] "The SAB recommends that, as an initial step in conducting an integrated
and comprehensive uncertainty analysis, the agency provide a tabular presentation and narrative
evaluation of the IUR estimates based on a reasonable range of data selections (e.g., all or part of
the earlier hires as well as the 'preferred' subcohort), model forms, and input assumptions (as
discussed, in the response to question 1 in Section 3.2.5). These input assumptions should
include inter alia exposure metrics and externally defined parameters, as discussed in the
response to Question 1 in Section 3.2.5. As noted in the current cancer risk assessment
guidelines rOJ.S. EPA. 2005a) pages 3-29]:
The full extent of model uncertainty usually cannot be quantified; a partial
characterization can be obtained by comparing the results of alternative models. Model
uncertainty is expressed through comparison of separate analyses from each model,
coupled with a subjective probability statement, where feasible and appropriate, of the
likelihood that each model might be correct (NRC, 1994).
The SAB notes that ideally, the agency would develop a quantitative characterization of the
overall uncertainty in its IUR estimates by incorporating the major sources of uncertainty the
agency has identified in its evaluation. However, the SAB recognizes the challenge of
conducting such an analysis, and is not recommending that it be undertaken at this time."
EPA Response: Section 5.4.3.4 describes the challenges EPA faced in analyses of the
full cohort, attributing the difficulties to the lack of accurate information on job code and
job department among 71% of workers hired prior to 1960. In contrast, among those
workers hired after 1959, only 1% of workers lacked specific work histories. EPA
evaluated the feasibility of conducting an ancillary analysis of the full Libby data set to
include hires before 1959. As described previously, EPA added a discussion of the
quantitative uncertainties connected to the use of a smaller subcohort in Sections 5.4.3.4
and 5.4.6, as recommended by SAB. EPA determined that the use of higher quality
personal exposure information outweighs the limitations caused by a smaller size of the
subcohort because the use of poor exposure data leads to large measurement error and
results in the underestimation of the regression coefficient of the dose response [cf.
(Bateson and Kopylev. 2014: Lenters et al.. 2012: Lenters etal.. 2011)1.
A.8. PUBLIC COMMENTS
Section A.8 responds to public comments with each subsection addressing a different general
topic.
A.8.1. Mineralogy—Summary of Major Public Comments with EPA Responses:
None.
A-42
-------
A.8.2. Fiber Toxicokinetics—Summary of Major Public Comments and EPA Response:
Toxicokinetics Public Comment #1 (Paraphrased): Jay Flynn requested inclusion of specific
peer-reviewed, published literature on LAA and further discussion of the comparative toxicity of
LAA and other amphiboles.
EPA Response: EPA has included summaries of the peer-reviewed published literature
on LAA through March 2014 in the appropriate section of the Toxicological Review (see
Section 4.2 for in vivo, see Section 4.3 for in vitro) and full study descriptions in
Appendix D. As this Toxicological Review is specific to LAA, studies on other
amphiboles that do not make up the LAA mixture are not included in these summaries or
in Appendix D. However, a discussion of the determinants of fiber toxicity has been
included in Section 3 to discuss what is known about the comparative toxicity of various
fiber characteristics for all amphiboles. Further, the revised section on MOA includes
discussion of hypothesized MOA for other amphiboles in comparison to LAA.
A.8.3. Noncancer Health Effects—Summary of Major Public Comments with EPA
Responses:
Noncancer Health Effects Public Comment #1 (Paraphrased): Several commenters (Elizabeth
Anderson, John DeSesso, David Hoel, Jay Flynn, and Lawrence Mohr) stated that EPA failed to
demonstrate an association between LPT and decreased lung function, so that any lung function
decrease that might be associated is "insignificant" and thus LPT is not adverse by EPA's own
definition of "adverse."
EPA Response: EPA has provided an expanded description of the selection of the
critical effect for the derivation of the RfC in Section 5.2.2.3. EPA also conducted a
systematic review and meta-analysis of studies examining the relation between LPT and
pulmonary function measures. This work is presented in Appendix I.
This additional literature review and analysis demonstrates that pleural plaques and LPT
are associated with a decrease in two key measures of lung function, and that these
decreases are unlikely to be due to other factors such as excess body fat or undetected
changes in lung tissue (other than the pleural plaques) that might have also been caused
by exposure to asbestos. Thus, these additional references and analysis support the
EPA's conclusions in its External Review Draft, and the SAB advice to EPA, that LPT is
an appropriate health endpoint for the derivation of the inhalation reference
concentration.
EPA's literature search identified epidemiology studies examining lung function in
asbestos-exposed populations with and without pleural plaques; 20 studies relating
changes in forced vital capacity (FVC) to presence of pleural plaques and 15 studies
relating changes in forced expiratory volume in 1 second (FEVi) to presence of pleural
plaques were included in a meta-analysis.
A meta-analysis of the identified studies conducted by EPA estimated a statistically
significant decrement of 4.09 (95% CI: -5.86, -2.31) and 1.99 (95% CI: -3.77, -0.22)
percentage points respectively in predicted FVC and FEVi attributable to the presence of
pleural plaques.
A-43
-------
The definition of "adverse" in EPA's IRIS Glossary states that an adverse effect
".. .affects the performance of the whole organism or reduces an organism's ability to
respond to an additional environmental challenge." EPA analysis shows that the LPT
causes a statistically significant lung function decrease; such lung function decreases
reduce an organism's ability to withstand those additional environmental challenges that
further reduce lung function.
Another EPA definition of adversity for epidemiologic data states that reductions in lung
function such as FEVi are considered adverse respiratory health effects (U.S. EPA, 1994).
Additional analyses indicated that the decrements associated with the presence of LPT
are not likely to be due to limitations in the study designs or conduct, undetected
subclinical fibrosis or misidentification of pleural plaques due to subpleural fat pads.
Only several studies controlled for exposure, but the largest best controlled HRCT study
that also controlled for exposure found decrease in lung function similar to the decreases
above.
Further, the extent of plaques was found to correlate with the degree of lung function
decrement, and longitudinal studies indicate that decrements increase with longer
follow-up.
These findings support the conclusion that pleural plaques, and LPT, is an appropriate
health endpoint for the derivation of the RfC.
Noncancer Health Effects Public Comment #2 (Paraphrased): Several commenters (including
Lawrence Mohr) stated that EPA did not consider all of the scientific literature on LPT and that it
confuses LPT with DPT.
EPA Response: In response to the SAB's identification of additional references and
recommendation that the Agency include a more detailed review of the literature, EPA
conducted a more detailed review of the literature examining the relationship between
lung function measures and localized pleural thickening (LPT) and pleural plaques. That
systematic review not only included the additional references noted by the Science
Advisory Board, but comprises a systematic and well-documented literature search and
review of the published literature through the date of December 2013. This work is
presented in Appendix I and discussed in Section 5.2.2.3.
In a meta-analysis presented in Appendix I, EPA considered only studies that considered
pleural plaques in groups that did not contain any DPT or parenchymal abnormalities, so
that there would not be confusion of LPT with DPT.
Noncancer Health Effects Public Comment #3: "A new peer reviewed study published in
Chest (Clark et al., 2014) (note: citation updated from original comment to include publication
information) analyzes historic health data from the Libby, Montana vermiculite miners and finds
that plaques alone did not cause lung function deficits among miners exposed to LAA. No
statistically significant difference in lung function was found between miners with pleural
plaques alone and those with no radiography findings (using High Resolution Computed
Tomography ("HRCT")). EPA should evaluate and account for this study because it analyzes
Libby-specific data, making it one of the most relevant studies for this LAA assessment to
consider. Moreover, this study thoughtfully addresses bias and seeks to eliminate confounders
A-44
-------
present in many other studies. This study uses the most reliable diagnostic methods: HRCT and
multiple pulmonary function test parameters. It is well accepted in the medical community that
x-ray radiography is prone to misdiagnosis of pleural plaques (e.g., extrapleural fat can be
mistakenly identified as plaques) and underdiagnosis of other lung abnormalities (e.g., fibrosis)
that affect lung function. The HRCT data used in this study provide superior contrast sensitivity
and cross-sectional imaging format, and thus minimize the potential for bias from relying upon
x-rays. The study quality also is enhanced because it evaluates multiple pulmonary function test
parameters to distinguish among different types of lung decrements (such as obstructive lung
disease that is unlikely to be related to asbestos). In contrast to this new study, many other
studies that EPA has relied on reflect bias from reliance upon less accurate x-rays and limited
lung function testing.)" [Comment received by EPA on June 25, 2014.]
EPA Response: In response to Major SAB Recommendation Letter #1, EPA
conducted a systematic review and meta-analysis of the influence of localized pleural
thickening on lung function, including a separate meta-analysis of HRCT studies.
Although the Clark et al. (2014) was published after the cut-off date of December 31,
2013 for the systematic review and meta-analysis, EPA evaluated the Clark etal. (2014)
study as it relates to the meta-analysis. EPA found that inclusion of Clark etal. (2014)
would not materially change EPA's conclusions and in fact, the new paper is supportive
of EPA's conclusion (i.e., the summary estimate in the meta-analysis of HRCT studies
shows even greater decreases in lung function associated with LPT and the uncertainty
associated with the decrease is diminished with the inclusion of additional data from
Clark et al. (2014) as noted by the decrease in the width of the confidence interval).
Appendix I Meta-Analysis if
Meta-Analysis Including Clark paper
FVC -3.30% (-5.25;-1.34) -3.59% (-5.08,-2.10)
FEVi -1.96% (-6.01; 2.09) -2.60% (-5.94; 0.74)
Noncancer Health Effects Public Comment #4: "A second peer reviewed study (Moolgavkar et
al., 2014) (note: citation updated from original comment to include publication information)
rigorously assesses the body of literature that the Draft Assessment relies upon, and concludes
that: ... in light of the serious methodological limitations and inconsistent findings of these
collective studies, the overall weight of evidence does not establish an independent adverse
effect of pleural plaques on pulmonary function. This study quotes and then applies EPA-
established criteria as follows: "by the Agency's own definitions, for an effect to be considered
adverse, the presence of biological or pathologic changes is not sufficient. Rather, these changes
must additionally affect the performance of the whole organism or compromise the organism's
ability to respond to environmental changes."
"EPA should evaluate and account for this study because it assesses sources of bias and
confounders present in the body of literature that the LAA Draft Assessment relies upon."
[Comment received by EPA on June 25, 2014.]
EPA Response: The publication by Moolgavkar et al. (2014) reviews the literature
quoted in the 2011 External Review Draft. Their review is nonquantitative in nature and
A-45
-------
not a comprehensive review of the literature. In response to Major SAB
Recommendation Letter #1, EPA conducted a systematic review and meta-analysis of
the influence of localized pleural thickening on lung function and concluded that
localized pleural thickening is associated with statistically significant decrease in lung
function measures (see Appendix I). In Appendix I, EPA formally evaluated the
limitations of each study and conducted sensitivity analyses that confirmed the overall
conclusions.
The remainder of the publication repeats a number of public comments submitted to the
SAB and to EPA by the authors of Moolgavkar et al. (2014). EPA has responded to these
public comments in revisions of the final assessment and/or elsewhere in Appendix A.
Noncancer Health Effects Public Comment #5: Beveridge & Diamond (on behalf of W.R.
Grace) commented (Section III.Al.a) that "The Draft Assessment Does Not Satisfy EPA's Own
Definition of Adverse Effect and, as a Result, Fails to Meet the Agency's IQA Commitment to
Apply Its IRIS Policies and Procedures to Ensure and Maximize Quality."
EPA Response: Please see response to the Major SAB Recommendation Letter #1 and
Noncancer Health Effects Public Comment #1 above.
Noncancer Health Effects Public Comment #6: Beveridge & Diamond (on behalf of W.R.
Grace) commented (Section III.Al.b) that EPA "failed to demonstrate a 'causal' association
between LPT and functional impairment."
The comment goes on to say that the "Draft Assessment fails to explain how LPT could
plausibly cause, or be on a pathway to, impairment. Indeed, no mode of action has been
established. If the exposure to asbestos causes lung decrements, and if the exposure to asbestos
causes LPT, hypothetically one might find a correlation between LPT and lung decrements. But
this is insufficient to show that LPT itself causes lung decrements. The SAB Report does not
cure this deficiency."
EPA Response: EPA defines an "adverse effect" as a "biochemical change, functional
impairment, or pathologic lesion that affects the performance of the whole organism, or
reduces an organism's ability to respond to an additional environmental challenge"
(italics added) (U.S. EPA. 2002b).
In this assessment, the critical noncancer effect, localized pleural thickening, is a
persistent structural alteration of the pleura, which the EPA Science Advisory Board
noted they would consider a pathological change.
The revised LAA Toxicological Review has an extensive discussion of how LPT affects
lung function in Appendix I. While Appendix I does not use a summary descriptor of the
extent to which the evidence from EPA's systematic review of the literature on this
subject supports a decision on causality, it follows the same approach used in frameworks
for assessing causality and reaches the conclusion that there is an observed association
between the presence of LPT and decreased lung function, and that the association is
"unlikely to be explained by other causes of pulmonary function loss." It evaluates the
available studies (positive and null or negative) regarding the association between LPT
and lung function. It assesses the evidence for an association, consistency, and the
strength of association using a meta-analysis and its statistical significance. It examines
A-46
-------
whether confounders could explain the observed association and concludes that they are
unlikely to explain the observed association.
Appendix I does address several hypotheses that LPT might not affect lung function but
simply be a marker of exposure where the exposure affects lung function. Appendix I
also notes that there is otherwise not much information as to how LPT might affect lung
function. It is, however, recognized in EPA frameworks for evaluating causality that
often the mode of action for an effect is unknown, and that knowing a mode of action is
not a requirement for judging that there is, or not, a causal relationship.
Noncancer Health Effects Public Comment #7: Beveridge & Diamond (on behalf of W.R.
Grace) commented (Section III.A1 .c) that "The Draft Assessment's LPT findings relied on
irrelevant and non-probative DPT and visceral pleura data, and the Draft Assessment was
inaccurate in how it applied an ILO classification. Because of the merging of the analysis of DPT
(with known adverse effects) with LPT (which has not been shown to cause adverse effects), the
Draft Assessment's analysis is unreliable, inaccurate, and biased."
EPA Response: EPA carefully reviewed its understanding of terminology in the ILO
guidelines that define "LPT" and "DPT." The discussions of LPT and the scientific
literature in this final assessment uses consistent and accurate terminology. EPA
carefully applied ILO classification and avoided blurring the distinctions between
different radiological findings on the visceral pleura, such as DPT or other general or
unclassified pleural thickening.
In the meta-analysis presented in Appendix I where EPA further reviewed additional
literature on the association of LPT with decrements in lung function, EPA only
considered studies of workers with pleural plaques who did not have DPT or
parenchymal abnormalities in order to avoid confusion between LPT and DPT.
Please see response to the Noncancer Health Effects Public Comment #2 above.
Noncancer Health Effects Public Comment #8: Beveridge & Diamond (on behalf of W.R.
Grace) commented (Section III.Al.d) that "the Draft Assessment fails to identify and consider all
of the relevant literature, including authoritative papers and studies that contradict the Draft
Assessment's position, and studies that use the most sensitive radiographic diagnostic tool."
EPA Response: EPA performed a systematic identification and analysis of the relevant
literature and transparently reported the analysis and findings (see Appendix I). In
Appendix I, EPA specifically identified relevant HRCT studies and conducted meta-
analyses of x-ray and HRCT studies and of HRCT studies separately. EPA also
thoroughly evaluated consensus statements from ATS and ACCP.
Please see response to the Major SAB Recommendation Letter #1 and Noncancer
Health Effects Public Comment #1 above.
Noncancer Health Effects Public Comment #9 (Paraphrased): Beveridge & Diamond (on
behalf of W.R. Grace) commented (Section III.Al.e) that there are concerns arising from
potential confounding. One concern stated was that the draft assessment fails to consider and
account for important confounders and specifically that subpleural fat deposits can be mistaken
for pleural plaques (LPT) on lung x-rays. A second concern was that studies could be
complicated by risk factors for pulmonary decrements such as smoking and BMI.
A-47
-------
EPA Response: Each of these potential limitations has been addressed by EPA.
In its systematic literature review, EPA evaluated whether each study compared
participants with reference populations and in its meta-analyses of FVC and FEV1, EPA
only used studies with appropriate internal controls (see Appendix I).
EPA identified and accounted for potential confounders, effect modifiers and study
limitations such as: smoking and BMI. Additional analyses in Appendix I indicated that
the decrements associated with the presence of LPT are not likely to be due to limitations
in the study designs, undetected subclinical fibrosis or misidentification of pleural
plaques due to subpleural fat pads.
Please see response to the Major SAB Recommendation Letter #1 and Noncancer
Health Effects Public Comment #1 above.
Noncancer Health Effects Public Comment #10 (Paraphrased): Beveridge & Diamond (on
behalf of W.R. Grace) commented (Section III.Al.f) that there is no support for conjecture that
LPT is associated with pain and/or dyspnea.
EPA Response: A direct association between localized pleural thickening and chest pain
or dyspnea (defined as "difficult or labored breathing; shortness of breath") was
discussed in the draft assessment, but is not cited in the final assessment as the basis for
concluding that LPT is an adverse or deleterious effect. Instead, this final assessment
systematically reviewed the literature as to whether LPT affects lung function and
discusses that even a 5% change in mean lung function would be expected to functionally
impair some individuals with otherwise low lung function and would be expected to
reduce their ability to respond to additional environmental challenge. That analysis is
described in Appendix I.
Noncancer Health Effects Public Comment #11: Beveridge & Diamond (on behalf of W.R.
Grace) commented (Section III.B1) that "the Draft Assessment has not demonstrated that LPT is
an adverse effect as defined by EPA, because LPT has not been demonstrated to cause, or itself
to present, a functional impairment."
EPA Response: Please see response to the Major SAB Recommendation Letter #1 and
Noncancer Health Effects Public Comment #1 above.
Noncancer Health Effects Public Comment #12 (Paraphrased): Beveridge & Diamond (on
behalf of W.R. Grace) commented (Section IV. A) that EPA fails to identify relevant studies for
selection of critical effect.
EPA Response: EPA performed systematic identification and analysis of the relevant
literature and transparently reported the analysis and findings (see Appendix I). In
Appendix I, EPA identified relevant HRCT studies and conducted meta-analysis of x-ray
and HRCT studies and HRCT studies separately. EPA also thoroughly evaluated
consensus statements from ATS and ACCP.
Please see response to the Major SAB Recommendation Letter #1 and Noncancer
Health Effects Public Comment #1 above.
A-48
-------
Noncancer Health Effects Public Comment #13 (Paraphrased): Beveridge & Diamond (on
behalf of W.R. Grace) commented (Section IV.B) that EPA does not conduct a rigorous
weight-of-evidence evaluation.
EPA Response: EPA did conduct a rigorous "weight-of-evidence" evaluation of the
evidence regarding the relationship between LPT and lung function.
The revised LAA Toxicological Review has an extensive discussion of how LPT affects
lung function in Appendix I. While Appendix I does not use a summary descriptor of the
extent to which the evidence from EPA's systematic review of the literature on this
subject supports a decision on causality, it follows the same approach used in frameworks
for assessing causality and reaches the conclusion that there is an observed association
between the presence of LPT and decreased lung function, and that the association is
"unlikely to be explained by other causes of pulmonary function loss." It evaluates the
available studies (positive and null or negative) regarding the association between LPT
and lung function. It assesses the evidence for an association, consistency, and the
strength of association using a meta-analysis and its statistical significance. It examines
whether confounders could explain the observed association and concludes that they are
unlikely to explain the observed association.
Appendix I does address several hypotheses that LPT might not affect lung function but
simply be a marker of exposure where the exposure affects lung function. Appendix I
also notes that there is otherwise not much information as to how LPT might affect lung
function. It is, however, recognized in EPA frameworks for evaluating causality that
often the mode of action for an effect is unknown, and that knowing a mode of action is
not a requirement for judging that there is, or may be, a causal relationship.
Inhalation Reference Concentration (RfC) Public Comment #14 (Paraphrased): Beveridge &
Diamond (on behalf of W.R. Grace) commented (Section IV.C.2) that EPA does not follow its
own guidance, particularly on RfC critical effect.
EPA Response: Please see response to Noncancer Health Effects Public Comment #1
above.
Noncancer Health Effects Public Comment #15 (Paraphrased): Beveridge & Diamond (on
behalf of W.R. Grace) commented (Section IV.D) that certain methods and data (i.e., additional
follow-up by Dr. James Lockey and data collected by ATSDR) underlying the assessment are
not publicly available for public review, analysis and comment and stated that such data is
needed for a sound assessment. (The comment recommended EPA postpone completing its
assessment until those studies become publicly available and EPA receives public comment on
those data and their implications.)
EPA Response: During the peer review of the draft IRIS assessment of LAA, Dr.
Lockey publicly informed the EPA's Science Advisory Board (SAB) of his on-going
research (as noted by the public commenter). These data mostly consist of additional
follow-up regarding some of the workers in Marysville, Ohio and include studies of lung
volume and diffusion in about 154 of the 231 workers studied in 2004.
These data remain unpublished and therefore are not included in the IRIS assessment of
LAA. Additionally, at this time, after the drafting of the assessment, its peer review, and
A-49
-------
further review of supporting literature as recommended by the SAB, EPA does not
consider these studies necessary to evaluate the relationship between exposure to LAA
and the prevalence of localized pleural thickening in workers evaluated in prior studies of
Marysville workers. Nor does EPA now consider these studies needed to further support
the conclusion that LPT is adverse. Appendix I summarizes a range of studies that relate
the presence of LPT with decrements in lung function.
With respect to the data collected by the ATSDR in Libby, Montana, in 2000 and 2001,
this information is available in the published, peer-reviewed, literature. These publicly
available studies conducted by ATSDR were summarized in the external review draft
assessment and remain in the final IRIS assessment. The assessment provides a rationale
to explain why the ATSDR data were not relied upon and why EPA considered the
Marysville, Ohio data to provide a better dataset for derivation of the RfC for noncancer
effects and the NIOSH Libby worker dataset an appropriate dataset for derivation of the
cancer inhalation unit risk.
Although on-going analysis of those exposed to LAA will increase understanding of the
toxicity of this material, the data referenced by the public commenter do not serve as the
basis of the assessment and postponing completion of the IRIS LAA assessment is not
warranted.
Further discussion of EPA's selection of critical studies and endpoints is above in the
response to the Major SAB Recommendation Letter #1 and Noncancer Health Effects
Public Comment #1 above.
A.8.4. Carcinogenicity—Summary of Major Public Comments and EPA Response
Carcinogenicity Public Comment #1 (Paraphrased): Elizabeth Anderson raised an issue with
the consideration of the cancer mode of action (MO A) and the possibility of nonlinearity in
exposure-response.
EPA Response: The MOA section of the Toxicological Review (see Section 4.6) has
been revised to include a formal carcinogenic MOA analysis. Further discussion of the
mechanistic data in support of the MOA for amphibole asbestos in general has been
included in Section 4.4. Data gaps still remain to characterize specific mechanisms
involved in LAA-induced disease. The formal mode-of-carcinogenic-action analysis
demonstrated that there are insufficient data to determine an MOA for LAA given
available data. Therefore, EPA determined that a linear low-dose extrapolation was
appropriate. In the absence of a well-defined MOA, linearity of exposure-response below
the POD is assumed in the derivation of the IUR (EPA's Guidelines for Carcinogen Risk
Assessment (U.S. EPA, 2005aV). Please also see the response to Major SAB
Recommendation Letter #5 comment.
A.8.5. Inhalation Reference Concentration (RfC)—Summary of Major Public Comments
and EPA Response
Inhalation Reference Concentration (RfC) Public Comment #1: Suresh Moolgavkar stated
"The noncancer risk assessment is based on a cohort of workers at a Marysville, Ohio plant in
which Libby vermiculite was processed. The endpoint of interest was pleural abnormalities
(pleural thickening) on chest radiographs. The original data set considered by Rohs et al. (2008)
consisted of 280 workers with 80 cases of abnormalities on chest radiography. The Agency
A-50
-------
assessment was based on 119 workers with 12 cases of abnormalities. Thus, the Agency
discards 85% of cases for this assessment. The reasons given for this drastic reduction in the
cohort size are not tenable."
There were several related comments on the selection of the critical study for the derivation of
the RfC. Several commenters thought that the critical study population was too small and that
the full Marysville, OH cohort should be used.
EPA Response: EPA focused on the subset of workers who had the better quality
exposure data and more recent health evaluations for the derivation of the RfC.
EPA has used the modeling approach recommended by the SAB, which relies on the
larger subset of workers with more recent health evaluations (regardless of hire date), to
estimate the effect of TSFE on the risk of LPT and has combined this information with
the better quality exposure data in the primary modeling performed in the subcohort to
derive the RfC.
EPA has also performed modeling based on the fuller data set of all health evaluations
performed in 1980 and 2002-2005. That analysis is presented in Appendix E. The
primary and this complementary modeling of the full cohort yield a comparable RfC.
Inhalation Reference Concentration (RfC) Public Comment #2: Suresh Moolgavkar stated
"There is no evidence of a monotonic increasing exposure-response relationship for pleural
thickening in either the full cohort or the subcohort chosen for analysis by the Agency."
EPA Response: Monotonicity in the observed exposure-response data is not a
requirement for RfC derivation and may be sensitive to the number of strata into which
the data are divided.
In the analysis of the primary subcohort of workers from Marysville, OH, hired in 1972
or afterwards, the exposure-response relationship between mean intensity of exposure
and the risk of LPT is plotted in two different ways (see Figure 5-3). The two ways
divide the data into quartiles and quintiles and plot the exposure-response relationship.
These show increasing risk with increasing exposure. Plots of the exposure-response
relationship for the full cohort are shown in Appendix E and also show increasing risk
with increasing exposure. Figure 5-3 shows that the prevalence of LPT increases with
increasing exposure.
The monotonic models used to derive the exposure-response relationship adequately fit
the data (see Tables 5-4 and 5-9).
Inhalation Reference Concentration (RfC) Public Comment #3: Public comments were raised
regarding the source of data on Marysville workers exposed between 1971 and 1973. For
example, Suresh Moolgavkar stated: "The Agency says, '.. .more accurate exposure data are
considered to be those from 1972 and later, as these data were based on analytical
measurements.' Based on these considerations, the Agency chose from the Rohs cohort the
subcohort consisting of workers who began work in 1972 or later. The radiographic examination
of these workers was conducted over the period 2002-2005. However, in their paper, Rohs et al.
(2008) identified 1973, not 1971, as the year after which '.. .more comprehensive environmental
exposures were available...' The subcohort of workers hired after 1973 consists of
A-51
-------
94 individuals with 10 cases of pleural abnormalities. I have the Rohs database and it includes
an identifier for workers hired after 1973 but not for those hired after 1971. The report does not
explain this discrepancy."
EPA Response: Additional work (i.e., after publication of the Rohs et al. (2008) paper)
was done by the University of Cincinnati to refine and update the exposure estimates.
Please see Appendix F for details.
Inhalation Reference Concentration (RfC) Public Comment #4: Public commenters raised the
issue of potential confounders in the epidemiologic analyses. For example, Suresh Moolgavkar
stated: "I analyzed the data in the subcohort of individuals in the Rohs cohort who were first
employed after 1973.... With the usual assumption of a logit-linear relationship between
exposure and response in the logistic model, the coefficient for cumulative exposure is
statistically significant at the 0.05 level of significance. If, however, either age or body mass
index (BMI) are considered as confounders in a joint analysis, the coefficient for cumulative
exposure becomes insignificant. One of the important criteria enunciated by the Agency for
study selection for noncancer risk assessment is that the exposure-response relationship be robust
to adjustment for potential confounders. Thus, on page 5-11, the report states 'Amandus et al.
(1987) report that although cumulative exposure and age are both significant predictors of small
opacities, cumulative exposure was not significantly related to pleural abnormalities when age is
included in the model, thus limiting the usefulness of these data for RfC derivation based on
pleural abnormalities.' In listing the advantages of the Rohs subcohort the Agency used, the
report on page 5-14 (number 6) clearly states that it considers the absence of any evidence of
confounding in this data set a distinct advantage. I do not have access to the exact data used by
the Agency, but I have analyzed a closely related data set as described above and there is strong
evidence of confounding by both age and BMI. By its own criteria, the Agency should not be
using this data set for derivation of an RfC."
EPA Response: The commenter's concern is related to the potential for confounding of
the relationship between exposure to LAA and the risk of LPT. EPA evaluated
confounding using both a theory-based method (to ascertain whether the potential
confounder is associated with both the exposure and with the outcome; see
Section 5.2.2.6.1) as well as a data-based method (by including each potential confounder
in the final model to assess its statistical significance; see Section 5.3.3). No evidence of
confounding was found in either case. Comparable modeling of the full cohort is
described in Appendix E.
It is possible that the differences in interpretations of potential confounding are related to
difference in the exact set of data used by EPA and by the commenter.
EPA did assess the potential for confounding by age and by BMI using two different
approaches and did not identify such confounding of the exposure-response relationship
used to derive the RfC.
Inhalation Reference Concentration (RfC) Public Comment #5: Suresh Moolgavkar
commented ".. .the Agency uses various lags in the analyses of the subcohort. The use of lags
for the analyses of pleural abnormalities makes no sense. Lags, although I do not generally favor
them, can be used in analyses of hazard or incidence functions when the diagnosis of an
end-point, such as cancer, is made at a well-defined point in time. It makes absolutely no sense
to use lags in the analyses of prevalent conditions, which could have occurred many years before
A-52
-------
the condition was noted. In the Rohs database all radiography was performed between 2002 and
2005 when pleural abnormalities were noted. These could have occurred many years before the
radiography was done. What is the interpretation of a lag in this situation?"
EPA Response: EPA agrees with the commenter that lack of information on the timing
of the initial occurrence of the pleural changes makes it difficult to interpret lagged
exposures given the cross-sectional nature of the x-ray data. In the revised IRIS draft
assessment, EPA did not include lagged exposure metrics for this reason when modeling
the noncancer outcomes. Please also see response to SAB Inhalation Reference
Concentration (RfC) #8 comment. Lags were evaluated for the lung cancer and
mesothelioma risk modeling because for these cancers there is data available on the date
of death which is expected to be closely related to the date of cancer incidence due to the
short survival time for these cancers (see Section 5.4.2.2 for more details).
Inhalation Reference Concentration (RfC) Public Comment #6 (Paraphrased): Beveridge &
Diamond (on behalf of W.R. Grace) noted that the exposure metric for the derivation of the RfC
should be mean concentration rather than cumulative exposure.
EPA Response: EPA reconsidered the justification for model selection and the selected
exposure metric. EPA evaluated different exposure metrics, including mean and RTW
(see Section 5.2.2.6 and Appendix E) in addition to the CE metric included in the ERD
analyses. The recommended examination of alternative metrics of exposure has resulted
in a change from the use of CE in the draft to the use of C in the revision. Please also see
the response to Major SAB Recommendation Letter #3 comment.
Inhalation Reference Concentration (RfC) Public Comment #7 (Paraphrased): There were
several comments (including Beveridge & Diamond and David Hoel) on the selection of the
model for the derivation of the RfC. Comments stated that the model selection criteria were
unclear, that the Michaelis-Menten model should not be used to derive the RfC, and that the
merits of some models could not be appropriately distinguished based on model fit. Other
specific comments regarding the modeling included mention of background prevalence of
localized pleural thickening and the plateau parameters.
EPA Response: The SAB also commented that EPA should include biological and
epidemiological characteristics of the different models in the model selection. As noted
in the EPA Response to Major SAB Recommendation Letter #3, EPA provides a more
thorough explanation of its selection of the best model for noncancer exposure-response
analysis in Section 5.2.2.6 and in Appendix E.
Following the guidance in the updated Benchmark Dose Technical Guidance (U.S. EPA,
2012), EPA explained that there are several stages of exposure-response modeling. Once
the appropriate data set(s), endpoint(s) and BMR are determined, an appropriate set of
statistical model forms is selected and evaluated for model fit to determine which models
adequately represent the data. Among those models with adequate fit, one or more
models are selected to derive a point of departure for the RfC. Regarding the selection of
models to evaluate, the Benchmark Dose Technical Guidance notes that additional
criteria may be used, "governed by the nature of the measurement that represents the
endpoint of interest and the experimental design used to generate the data" (page 26).
When modeling the Marysville data, certain biological and epidemiological features must
be considered, including the nature of the data set, ability to estimate the effects of
A-53
-------
exposure and of important covariate(s), the existence of a plateau or theoretical maximum
response rate in a population, and the ability to estimate a background rate of the outcome
in a population.
For the primary modeling in Section 5.2.2.6., EPA selected the Dichotomous Hill model,
(a minor variation on the Michaelis-Menten model proposed in its External Review
Draft) because it allowed fuller consideration of the biological and epidemiological
features described above.
While EPA presents the results from the Michaelis-Menten model for purposes of
comparison, the final assessment does not rely upon the Michaelis-Menten model. EPA
explains in the final assessment how, based on advice from the SAB, it selected preferred
model forms and evaluated those forms, with evaluations of different exposure metrics,
with statistical comparisons, goodness-of-fit criteria, and graphical comparisons with
aggregated data. This is described in a revised Section 5.2.2.6 concerning model
considerations (including background prevalence, plateau, and ability to control for
potential confounders), model selection, and selection of the BMR, taking into account
EPA's newly available updated BenchmarkDose Technical Guidance (U.S. EPA, 2012).
Please also see response to Major SAB Recommendation Letter #3 comment.
Inhalation Reference Concentration (RfC) Public Comment #8 (Paraphrased): The American
Chemistry Council noted that the justification of the uncertainty factors was inadequate.
EPA Response: EPA has reconsidered the choice of UFs in light of the SAB
recommendations, revised analyses and newly available published studies.
Consequently, the database UF has been reduced to 3, while the subchronic-to-chronic
UF has been increased to 10 based on the evaluation of the role TSFE on LPT risk in the
Marysville data. Increasing the LOAEL-to-NOAEL UF is unnecessary because the POD
is based on BMD modeling. The basis for those decisions is explained in a revised
Section 5.2.3. Please also see response to the Major SAB Recommendation Letter #4
comment.
Inhalation Reference Concentration (RfC) Public Comment #9 (Paraphrased): There were
several comments (including the American Chemistry Council, Elizabeth Anderson, Terry Trent
and Karen Ethier) that the RfC would be below background concentrations, that the RfC would
be used for other amphiboles, and that it would be unfeasible to measure fiber concentration at
the RfC level.
EPA Response: This assessment is quantifying the toxicity of LAA. Background or
naturally-occurring levels of many material vary considerably across the United States.
A discussion of varying geogenic levels of materials such as asbestos has little relevance
in a summary of toxicological data.
There are instances in which exposure to a substance either poses health risks, or at least
cannot be determined to be unlikely to pose health risks, at commonly found
"background" levels. Thus, there is nothing inherently contradictory if an RfC is below
common environmental or other "background" exposures. EPA is unaware of a basis for
bounding this assessment based on background exposures at any Superfund site. In
addition, the RfC of 9 x 10~5 fibers/cc is above average ambient air concentrations
currently measured in Libby, MT.
A-54
-------
When there are practical implementation concerns about reference values below detection
limits or below background, those concerns are best addressed in the context of risk
management decisions.
Inhalation Reference Concentration (RfC) Public Comment #10 (Paraphrased): Beveridge &
Diamond (on behalf of W.R. Grace) commented (Section III.A2.a) that noncancer toxicity value
is based on unduly restricted and confounded data set.
EPA Response: Please see response to SAB Inhalation Reference Concentration
(RfC) #13.
Inhalation Reference Concentration (RfC) Public Comment #10 (Paraphrased): Beveridge &
Diamond (on behalf of W.R. Grace) commented (Section III.B.2) that the noncancer Draft
Assessment selected a statistical model that is not justified and not scientific.
EPA Response: Please see response to the Major SAB Recommendation Letter #3.Inhalation
Reference Concentration (RfC) Public Comment #11: Beveridge & Diamond (on behalf of
W.R. Grace) commented (Section IV.F and IV.G) that "The Draft Assessment Does Not Address
the Expected 'Central Tendency' Risks to Affected Populations and Fails to Set Forth Upper and
Lower Bound Estimates of Expected LAA Hazards to Affected Populations."
EPA Response: The assessment does provide central estimates of the point of
departures for noncancer effects, along with lower bounds on the point-of-departure
concentrations. These values provide information on the degree of uncertainty with
respect to the points of departure.
EPA also does not at this time have a methodology for reflecting central or upper-bound
estimates of the noncancer point of departure in statements about concentration unlikely
to be without significant deleterious effects such as the RfC.
A.8.6. Inhalation Unit Risk (IUR)—Major Public Comments with EPA Responses:
Inhalation Unit Risk (IUR) Public Comment #1: Public comments were received on the
importance of evaluating the quality of exposure assessments made in epidemiology studies and
as a consideration in selected studies to use for asbestos toxicity assessment. For example, Terry
Spear stated: "Asbestos risk assessments are sensitive to small changes in decisions about which
data to include or exclude. The following abstract from Burdorf and Heederik (2011) illustrates
this point: 'Mesothelioma deaths due to environmental exposure to asbestos in The Netherlands
led to parliamentary concern that exposure guidelines were not strict enough. The Health
Council of the Netherlands was asked for advice. Its report has recently been published. The
question of quality of the exposure estimates was studied more systematically than in previous
asbestos meta-analyses. Five criteria of quality of exposure information were applied, and
cohort studies that failed to meet these were excluded. For lung cancer, this decreased the
number of cohorts included from 19 to 3 and increased the risk estimate three- to sixfold, with
the requirements for good historical data on exposure and job history having the largest effects.
It also suggested that the apparent differences in lung cancer potency between amphiboles and
chrysotile may be produced by lower quality studies. A similar pattern was seen for
mesothelioma. As a result, the Health Council has proposed that the occupational exposure limit
be reduced from 10,000 fibers nT3 (all types) to 250 fibers nT3 (amphiboles), 1,300 fibers nT3
(mixed fibres), and 2,000 fibers nT3 (chrysotile). The process illustrates the importance of
A-55
-------
evaluating quality of exposure in epidemiology, since poor quality of exposure data will lead to
underestimated risk'."
EPA Response: EPA agrees on the importance of evaluating the quality of exposure
data as part of evaluating epidemiology studies. EPA cites the work cited by the
commenter by Burdorf and Heederik (2011) and follow-up work by Lenters et al. (2011)
and Lenters et al. (2012) when EPA discusses the importance of evaluating the quality of
the exposure data. EPA has cited these works as part of the justification for EPA's
decision to base its cancer risk estimates on the selected subcohort for which there is the
best exposure information (see Section 5.4.3.4).
Inhalation Unit Risk (IUR) Public Comment #2: Public comments were received on the use of
the smaller subcohort from the Libby worker study for quantitative risk assessment. For
example, Suresh Moolgavkar stated "The data set chosen for the cancer risk assessment is a
small subcohort of the full cohort of Libby miners. This subcohort discards the vast majority of
lung cancers and mesotheliomas in the Libby cohort, particularly in individuals over the age of
65. Thus, Agency risk assessments are based largely on younger individuals in the cohort and
ignore the ages at which cancer is most common."
EPA Response: EPA evaluated the potential uncertainties in basing the quantitative
analyses on a subcohort, and concluded that the availability of higher quality exposure
information outweighed the limitations caused by the smaller size of the cohort (see also
the response to SAB Inhalation Unit Risk (IUR) #19).
The concern of the commenter that the subcohort analysis does not include individuals of
all ages can be evaluated by reviewing EPA's presentation of the primary results in
comparison to those in the published literature of the full Libby worker cohort, which
includes individuals of all ages. These analyses are presented in Tables 5-52 and 5-53 in
Section 5.4.5.3.1. EPA believes that the estimates of cancer exposure-response based on
the subcohort provide better estimates due to the higher quality of exposure data.
In addition, the SAB stated that the ".. .use of the subcohort post-1959 for quantification
may be reasonable due to the lack of exposure information for many of the workers in
earlier years; out of 991 workers hired before 1960, 706 had all department and job
assignments listed as unknown."
Inhalation Unit Risk (IUR) Public Comment #3: Public comments were received regarding the
statistical methods used for inhalation unit risk derivation. For example, Suresh Moolgavkar
stated: "The Agency uses inappropriate statistical methods for analyses of the data on lung
cancer and mesothelioma. In particular the importance of duration of exposure in determining
risk is ignored. In the lung cancer analysis effect modification by age, which is strongly evident
in the Libby cohort, is not addressed."
EPA Response: The commenter states that the EPA used inappropriate statistical
methods for the analysis of lung cancer and mesothelioma; however, in later comments,
the same commenter states that "the proportional hazards model used by the Agency for
analysis of lung cancer in the Libby miners' cohort is standard."
The commenter states that the importance of duration is ignored; however, for lung
cancer, the time-varying proportional hazards model does account for duration of
A-56
-------
exposure and is the same model form used by the commenter in other asbestos analyses
of lung cancer risk (Moolgavkar et al., 2010).
The commenter makes the point that the observed lack of proportionality in the full
cohort analysis of lung cancer may be due to effect modification by age and cites an
analysis by Richardson (2009). Effect modification by age is a possible explanation of
the lack of proportionality in the modeling of lung cancer mortality as has been noted by
Richardson (2009) in a two-stage clonal expansion model of a cohort of asbestos-exposed
workers. However, similar modeling of lung cancer risk in the same cohort of workers
by other investigators (Zekaetal., 2011) was unable to replicate that finding. EPA did
evaluate the possibility of effect modification of the lung cancer mortality risk by age in
the Libby workers subcohort and did not identify such a phenomenon as summarized in
Section 5.4.3.5.
Inhalation unit risk (IUR) Public Comment #4: Public comments were made regarding the
potential impact of exposure error on risk estimates. For example, Suresh Moolgavkar stated:
"The Agency repeats the old canard (page 5-78 of the report) about nondifferential covariate
measurement errors leading to risk estimates biased towards the null. This statement, although
widely repeated by epidemiologists, is incorrect. First, not only must the misclassification be
nondifferential, it must satisfy other conditions [e.g., (Jurek et al., 2005)1 for the result to hold.
Second, the statement applies to the expectation of the risk estimate, not to the value of the
estimate from any single study. Thus, it is possible to have nondifferential misclassification that
satisfies all the required conditions but the result of a single study may actually overestimate the
risk. As Jurek et al. (2005) state, '... exposure misclassification can spuriously increase the
observed strength of an association even when the misclassification process is nondifferential
and the bias it produced is towards the null.' Similar discussion is provided by Thomas (1995)
and Weinberg et al. (1995)."
A related comment by Terry Spear stated that it is difficult or impossible to find true associations
between exposure and effect when exposure misclassification exists in epidemiological studies.
Systematic misclassifications will create falsely high- or low-risk estimates while random
misclassification may mask true associations altogether.
EPA Response: The commenter (Moolgavkar) is correct that, under certain conditions,
nondifferential measurement error can yield results away from the null in a single study.
However, under general conditions of nondifferential exposure measurement error, the
expectation of the risk estimate is biased towards the null. According to a highly
regarded textbook, nondifferential exposure error typically results in bias towards the null
(Rothman and Greenland, 1998).
The commenter has not provided any information to suggest that, in this case, one would
expect no bias or a bias in the other direction due to the inclusion of the early hires in the
Libby workers cohort for whom the majority had no data on work histories and thus no
specific data on their exposures.
As described in the discussion of uncertainties in the cancer exposure-response (see
Section 5.4.6.1.2.4), uncertainties related to exposure measurement error are considered
unrelated to disease status and the general result is likely to be an attenuation in risk
estimates towards the null (i.e., the addition of random noise to a clear signal tends to
A-57
-------
reduce the clarity of the observed signal, and the avoidance of random noise results in a
stronger observed signal).
Issues of the misclassification of exposure in general may also be considered for the
noncancer exposure-response analyses. In the Marysville data used to support the
derivation of the primary RfC, there is no evidence of systematic misclassification of
exposure. In addition, EPA focused on the subset of workers with the highest quality
exposure data to derive the RfC, reducing the probability of exposure misclassification.
Similarly, for the IUR, selection of the subcohort minimizes exposure misclassification as
described in Section 5.4.5.3.1.
While EPA agrees that significant systematic exposure misclassification can make it
more difficult to derive accurate risk estimates, EPA did not find evidence that systematic
misclassification is an issue in the derivation of the RfC or IUR.
Inhalation Unit Risk (IUR) Public Comment #5 (Paraphrased): Suresh Moolgavkar stated
that estimated half-lives for lung cancer and mesothelioma appear too short—especially for
mesothelioma.
EPA Response: EPA reviewed the epidemiologic literature and has noted that half-lives
have been used to predict cancer risks associated with asbestos. EPA evaluated the fit of
models with and without half-lives and found that for mesothelioma, the models based on
a half-life applied to cumulative exposure fit better than models without a half-life.
Half-lives also have been used for modeling the Wittenoom, Australia amphibole
asbestos cohort (Berry et al., 2012). Berry and colleagues found similar half-lives for
amphibole asbestos-related mesothelioma as EPA found for LAA and mesothelioma.
For lung cancer unit risk, EPA selected the cumulative exposure metric which does not
involve half-lives.
Inhalation Unit Risk (IUR) Public Comment #6 (Paraphrased): There were several comments
on the selection of the model for the derivation of the IUR. Commenters questioned the use of
the Poisson model for mesothelioma instead of the traditional use of the Peto model and
suggested the use of two-stage clonal expansion models for lung cancer instead of the traditional
Cox proportional hazards model.
EPA Response: As responded to SAB comment SAB Inhalation Unit Risk (IUR) #18,
EPA's standard practice is to investigate several modeling options to determine how best
to empirically model the exposure-response relationship in the range of the observed data
as well as consider exposure-response models suggested in the epidemiologic literature.
For lung cancer, a new discussion of potential alternative models has been included in
Section 5.4.3.3, including Poisson, logistic, Cox, and multistage clonal expansion models.
EPA selected the Cox model as the most appropriate model for exposure-response
modeling based on the suitability of this model to the nature of the data set (e.g.,
time-dependent exposure information), the long history of this model usage in analyses of
occupational cohorts, and the commonality of usage in other epidemiologic analyses of
the Libby workers cohort. EPA's evaluation of alternative approaches found no other
standard epidemiological model formulations that allow for the analysis of time-varying
exposures in the manner achieved by the Cox proportional hazards model.
A-58
-------
For mesothelioma, a new discussion of alternative models has been included in
Section 5.4.3.1, including consideration of approaches such as parametric survival
models. EPA concluded that the Peto model and variations of the Peto allowing for
potential clearance are well supported in the epidemiologic literature. The Poisson model
is an appropriate model for rare data. There are no examples of using other models for
modeling mesothelioma in similar situations.
EPA presents results for sensitivity analyses that were conducted for both mesothelioma
and lung cancer mortality in deriving combined inhalation unit risk in Section 5.4.5.3.
Inhalation Unit Risk (IUR) Public Comment #10 (Paraphrased): Beveridge & Diamond (on
behalf of W.R. Grace) commented (Section III.A2.b) that cancer toxicity value is based on
unduly restricted and confounded data set.
EPA Response: Please see response to the Major SAB Recommendation Letter #6 and
SAB Inhalation Unit Risk (IUR) #19.
Inhalation Unit Risk (IUR) Public Comment #11 (Paraphrased): Beveridge & Diamond (on
behalf of W.R. Grace) commented (Section III.B3) that the cancer Draft Assessment selected a
statistical models that are scientifically unsound.
EPA Response: Please see response to Inhalation Unit Risk (IUR) Public
Comment #6.
Inhalation Unit Risk (IUR) Public Comment #12: Beveridge & Diamond (on behalf of W.R.
Grace) commented (Section IV.F and IV.G) that "The Draft Assessment Does Not Address the
Expected 'Central Tendency' Risks to Affected Populations and Fails to Set Forth Upper and
Lower Bound Estimates of Expected LAA Hazards to Affected Populations."
EPA Response: The assessment does provide central estimates of the point of
departures for cancer effects, along with lower bounds on the point-of-departure
concentrations. These values provide information on the degree of uncertainty with
respect to the points of departure. In addition Table 5-53 provides estimate of central risk
of mesothelioma or lung cancer along with the upper bound on risk.
EPA does not, however, at the current time have a methodology for low-dose
extrapolation of cancer risk that yields a lower bound inhalation unit risk at lower doses.
A linear extrapolation from the confidence limits on the cancer points of departure might
not reflect the uncertainty at those lower concentrations.
A.9. OTHER GENERAL PUBLIC COMMENTS WITH EPA RESPONSES:
General Public Comment #1: Beveridge & Diamond (on behalf of W.R. Grace) commented
(Section III.A.3) that "The Assessment Fails to Address Information Presented by Commenters
Identifying Fundamental Flaws in the Draft Assessment's Analysis."
EPA Response: EPA has considered all the public and peer-reviewed comments in
developing this final revised assessment. This appendix describes how EPA responded to
the most significant peer review and public comments.
A-59
-------
General Public Comment #2 (Paraphrased): Beveridge & Diamond (on behalf of W.R. Grace)
commented (Section III.C) that draft assessment is incomplete and lacks transparency.
EPA Response: EPA considers the assessment complete and transparent in presentation
of selection of critical effect and in presentation of models and modeling results. Specific
points raised in that section are responded elsewhere as they repeat previous comments
made on RfC or IUR.
General Public Comment #3: Beveridge & Diamond (on behalf of W.R. Grace) commented
(Section IV.C.l) that EPA should "Implement the April 2011 Formaldehyde Peer Review Report
'Chapter 7' IRIS recommendations as 'best available science' and 'sound and objective scientific
practices'." The comment continued that EPA should "For those IRIS reforms that EPA has
instituted for other ongoing draft IRIS assessments, either implement these reforms for this IRIS
assessment or explain why the reforms do not represent 'best available science' or 'sound and
objective scientific practices'."
EPA Response: In April 2011, the National Research Council (NRC) released its
Review of the Environmental Protection Agency's Draft IRIS Assessment of
Formaldehyde. In addition to offering comments specifically about EPA's draft
formaldehyde assessment, the NRC made several recommendations to EPA for
improving the development of IRIS assessments. EPA agreed with the recommendations
and is implementing them following a phased approach that is consistent with the NRC's
"Roadmap for Revision," which viewed the full implementation of their
recommendations by the IRIS Program as a multi-year process.
The IRIS LAA assessment is in Phase 1 and has focused on a subset of the short-term
recommendations, such as editing and streamlining, increasing transparency and clarity,
and using more tables, figures, and appendices to present information and data. For
example, the assessment includes clear explanations of the methods used to develop the
LAA assessment, and descriptions of the decisions made in developing the hazard
identification and dose-response analyses. As recommended, supplementary information
and analyses are presented in appendices and standardized tables were incorporated for
clarity in evidence presentation. Additionally, detailed discussions of mechanistic studies
on the biological response to LAA were included and the weight-of-evidence discussion
was expanded to include a formal mode-of-action analysis. All critical studies and the
candidate studies for the derivation of the noncancer critical effect were thoroughly
evaluated using standard protocols for evidence evaluation that are provided in existing
EPA guidance with the evaluation criteria and study findings presented in tables in
Section 5. Furthermore, text has been expanded to include more description of the
considerations made in selecting the study that formed the basis for the quantitative
reference concentration and cancer risk estimates and selection considerations are also
summarized in a table. A detailed discussion of model selection for the epidemiological
data sets is included.
General Public Comment #4 (Paraphrased): Beveridge & Diamond (on behalf of W.R. Grace)
commented (Section IV.E) that the assessment lacks a meaningful discussion of the population
likely to be affected by the assessment.
EPA Response: The commenter states that the assessment should discuss whether this
assessment, especially its RfC for noncancer effects, will be applied to populations
A-60
-------
exposed to forms of asbestos not originating in the Rainy Creek complex near Libby, MT
and that the assessment should discuss population exposure to such other forms of
asbestos whose toxicity is likely to be compared to that of LAA.
This assessment is clearly identified as an assessment of the risks of exposure to asbestos
originating in the Rainy Creek complex near Libby, Montana. Populations exposed to
that material include populations in Libby and populations in locations at which people
have been exposed to asbestos originating near Libby.
While this assessment included data on the toxicity of other types of asbestos to inform
the mechanistic relationship between LAA and health effects, it did not include a full
literature review of other amphiboles and therefore cannot reach conclusions as to other
types of asbestos. That was beyond the scope of this assessment and does not affect to
quality, transparency, or utility of this assessment for characterizing the risks of exposure
to LAA.
General Public Comment #5 (Paraphrased): Beveridge & Diamond (on behalf of W.R. Grace)
commented (Section IV.H) that the assessment fails to identify significant uncertainties.
EPA Response: Please see response to the Major SAB Recommendation Letter #4.
General Public Comment #6 (Paraphrased): Beveridge & Diamond (on behalf of W.R. Grace)
commented (Section V) that the RfC may be below background.
EPA Response: Please see response to Inhalation Reference Concentration (RfC)
Public Comment #9.
A.10. OTHER GENERAL PUBLIC COMMENTS TO THE SAB WITH EPA
RESPONSES:
Other Public Comment #1 (Paraphrased): Some commenters (including the American
Chemistry Council) recommended EPA have an additional round of public comment and peer
review.
EPA Response: EPA considered whether the revisions to the draft assessment warranted
additional peer review and concluded that the changes made were in response to peer
review advice and public comments and did not of themselves need a new additional
round of public comment or peer review.
Other Public Comment #2 (Paraphrased): There were several comments (including from the
American Chemistry Council) on the SAB process such as that more time should be allowed for
public speakers to make comments; that there was no opportunity for meaningful interaction
between the public speakers and the SAB panel; that SAB should avoid policy
recommendations; that the panel should include all panelists' opinions; that there was not enough
statistical experience on the panel; that the panel process was too rushed; and that the panel was
unaware of EPA guidance.
EPA Response: The review of this assessment went through the standard SAB
peer-review process and was consistent with the EPA Peer-Review Guidance. There
were numerous opportunities for external parties to submit comments. External parties
were invited to submit comments to the docket during a 60-day public comment period
A-61
-------
from August 25 to October 24, 2011 as noted in the Federal Register. All public
comments to the docket were provided to the SAB for review. External parties were also
invited to present analyses and viewpoints to the EPA assessment staff and managers at a
"Public Listening Session" held on October 6, 2011. External parties had further
opportunities to make presentations and provide written input to the SAB review panel
and the full SAB during the initial SAB Panel meeting February 6-8, 2012 and at
subsequent teleconference meetings on May 1, May 8, July 25, and September 25, 2013.
The SAB Panel was constituted according to the process established by the SAB with
public comment on the expertise of the panel members and oversight by the full SAB.
Other Public Comment #3: The Sections 5.2.3.3 through 5.4.6.2 deal with statistical modeling
(pages 5-28 to 5-122). In these sections statistical models and complex equations are used to
analyze data from Libby Amphibole asbestos studies. If the reader of this section doesn't have at
least a degree in statistics then the contents are very unclear and difficult to understand or
analyze for accuracy of conclusions. Since releasing the initial draft at a town meeting in Libby
on May 3, has anyone been able understand this section of it? In this section a large quantity of
information on asbestos illness is derived from statistics. Sections 5.2.3.3 through 5.4.6.2 should
be deleted.
EPA Response: EPA has added overview text intended to provide a simpler explanation
of the basis for the assessment. EPA has rewritten the sections on model considerations
and selection to provide more clarity. EPA has also provided graphics that were not in
the External Review Draft. For purposes of transparency so that statistically-trained
readers can understand how EPA addressed methodological issues, the detailed statistical
information and explanations in the assessment are needed.
A-62
-------
A.11. 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. (1987). The morbidity and mortality of
vermiculite miners and millers exposed to tremolite-actinolite: Part III. Radiographic findings. Am J Ind
Medl 1:27-37.
Aust AE: Cook. PM: Dodson. RF. (2011). Morphological and chemical mechanisms of elongated mineral particle
toxicities [Review]. JToxicol Environ Health B CritRev 14: 40-75.
http://dx.doi.org/10.1080/10937404.2011.556046
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
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
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. Ill l/i.1539-
6924.2010.01581.x
Bernstein. D: Castranova. V: Donaldson. K: Fubini. B: Hadlev. J: Hesterberg. T: Kane. A: Lai. D: McConnell EE:
Muhle. H: Oberdorster. G: Olin. S: Warheit DB: Group.. IRSIW. (2005a). Testing of fibrous particles:
short-term assays and strategies. Inhal Toxicol 17: 497-537. http://dx.doi.org/10.1080/08958370591001121
Bernstein. D: Rogers. R: Smith. P. (2005b). 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: 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: Reid. A: Aboagye-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.2012.23
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.
A-63
-------
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.
Burdorf. A: Heederik. D. (2011). Applying quality criteria to exposure in asbestos epidemiology increases the
estimated risk. Ann Occup Hyg 55: 565-568. http://dx.doi.org/10.1093/annhyg/mer042
Carbone. M: Baris. YI: Bertino. P: Brass. B: Comertpav. S: Dogan. AU: Gaudino. G: Jube. S: Kanodia. S: Partridge.
CR: Pass. HI: Rivera. ZS: Steele. I: Tuncer. M: Way. S: Yang. H: Miller. A. (2011). Erionite exposure in
North Dakota and Turkish villages with mesothelioma. PNAS 108: 13618-13623.
http://dx.doi.org/10.1073/pnas.1105887108
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 exposure to Libby amphibole asbestos and localized
pleural thickening. J Occup Environ Med 55: 1350-1355.
http://dx.doi.org/10.1097/JOM.Ob013e3182a3bb25
Clark. KA: Flynn. JJ: Goodman. JE: Zu. K: Karmaus. WJ: Mohr. LC. (2014). Pleural plaques and their effect on
lung function in Libby vermiculite miners. Chest 146: 786-794. http://dx.doi.org/10.1378/chest. 14-0043
Constantopoulos. SH. (2008). Environmental mesothelioma associated with tremolite asbestos: Lessons from the
experiences of Turkey, Greece, Corsica, New Caledonia and Cyprus. Regul Toxicol Pharmacol 52: S110-
S115. http://dx.doi.0rg/10.1016/i.vrtph.2007.ll.001
Davis. JMG: Addison. J: Bolton. RE: Donaldson. K: Jones. AD: Miller. BG. (1985). Inhalation studies on the effects
of tremolite and brucite dust in rats. Carcinogenesis 6: 667-674. http://dx.doi.Org/10.1093/carcin/6.5.667
Dourson. M: Charnlev. G: Scheuplein. R. (2002). Differential sensitivity of children and adults to chemical toxicity -
II. Risk and regulation. Regul Toxicol Pharmacol 35: 448-467. http://dx.doi.org/10.1006/rtph.2002.1559
Gaylor. D: Ryan. L: Krewski. D: Zhu. Y. (1998). Procedures for calculating benchmark doses for health risk
assessment. Regul Toxicol Pharmacol 28: 150-164. http://dx.doi.org/10.1006/rtph.1998.1247
Gogou. E: Kerenidi. T: Chamos. V: Zintzaras. E: Gourgoulianis. KI. (2009). Mesothelioma mortality in Greece from
1983 to 2003. Int J ClinPract 63: 944-948. http://dx.doi.0rg/10.llll/i.1742-1241.2007.01334.x
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
Hattis. D: Banati P: Goble. R. (1999). Distributions of individual susceptibility among humans for toxic effects:
How much protection does the traditional tenfold factor provide for what fraction of which kinds of
chemicals and effects. AnnN Y Acad Sci 895: 286-316.
HEI (Health Effects Institute). (1991). Asbestos in public and commercial buildings: a literature review and
synthesis of current knowledge.
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/
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
Jarvholm. B. (1992). Pleural plaques and exposure to asbestos: a mathematical model. Int J Epidemiol 21: 1180-
1184.
A-64
-------
Jurek. AM: Greenland. S: Maldonado. G: Church. TR. (2005). Proper interpretation of non-differential
misclassification effects: expectations vs observations. Int J Epidemiol 34: 680-687.
Larson. TC: Antao. VC: Bove. FJ: Cusack. C. (2012). 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
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
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
pleura! fibrosis. Am J Ind Med 20: 1-15. http://dx.doi.org/10.1002/aiim.4700200102
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.9Q88311
Lippmann. M. (2009). Asbestos and other mineral and vitreous fibers. In M Lippmann (Ed.), Environmental
Toxicants: Human Exposures and Their Health Effects (3rd ed., pp. 395-458). New York, NY: John Wiley.
http://www.wilev.com/WilevCDA/WilevTitle/productCd-0471793353.html
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.
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
McDonald. JC: McDonald. AD: Armstrong. B: Sebastien. P. (1986). 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
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.
Metintas. M: Hillerdal G: Metintas. S: Dumortier. P. (2010). Endemic malignant mesothelioma: exposure to
erionite is more important than genetic factors. Arch Environ Occup Health 65: 86-93.
http://dx.doi.org/10.1080/19338240903390305
Metintas. M: Metintas. S: Ak. G: Erginel S: Alatas. F: Kurt. E: Ucgun. I: Yildirim. H. (2008). Epidemiology of
pleural mesothelioma in a population with non-occupational asbestos exposure. Respirology 13: 117-121.
http://dx.doi.org/10.1111/i. 1440-1843.2007.01187.x
Metintas. S: Metintas. M: Ak. G: Kalyoncu. C. (2012). Environmental asbestos exposure in rural Turkey and risk of
lung cancer. Int J Environ Health Res 22: 468-479. http://dx.doi.org/10.1080/09603123.2011.654330
Miller. A. (2002). Pleural plaques and lung function [Letter]. Am J Respir Crit Care Med 165: 305-306.
http://dx.doi.org/10.1164/airccm. 165.2.correspondence c
Moolgavkar. SH: Anderson. EL: Chang. ET: Lau. EC: Turnham. P: Hoel DG. (2014). A review and critique of U.S.
EPA's risk assessments for asbestos. Crit Rev Toxicol 44: 499-522.
http://dx.doi.org/10.3109/10408444.2014.902423
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
A-65
-------
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 CritRev 14: 76-121. http://dx.doi.org/10.1080/10937404.2011.556047
NRC (National Research Council). (1994). Science and judgement in risk assessment (pp. 908-909). Washington,
DC: National Academy Press.
OEHHA (California Office of Environmental Health Hazard Assessment). (2008). Air toxics hot spots risk
assessment guidelines technical support document for the derivation of noncancer reference exposure
levels. Oakland, CA: California Environmental Protection Agency.
http://www.oehha.ca.gov/air/hot spots/2008/NoncancerTSD final.pdf
Paris. C: Martin. A: Letourneux. M: Wild. P. (2008). Modelling prevalence and incidence of fibrosis and pleura!
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
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
Richardson. DB. (2009). Lung cancer in chrysotile asbestos workers: analyses based on the two-stage clonal
expansion model. Cancer Causes Control 20: 917-923.
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
Rothman. KJ: Greenland. S. (1998). Modern epidemiology (2nd ed.). Philadelphia, PA: Lippincott, Williams, &
Wilkins.
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.
http://vosemite.epa. gov/sab/sabproduct.nsf/0/4F2A245C7160456B85257B030072E5D3/$File/EPA-SAB-
13-001 -unsigned.pdf
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
Scheuplein. R: Charnlev. G: Dourson. M. (2002). Differential sensitivity of children and adults to chemical toxicity
I. Biological basis. Regul Toxicol Pharmacol 35: 429-447. http://dx.doi.org/10.1006/trph.20Q2.1558
Sichletidis. L: Chloros. D: Chatzidimitriou. N: Tsiotsios. I: Spyratos. D: Patakas. D. (2006). Diachronic study of
pleura! plaques in rural population with environmental exposure to asbestos. Am J Ind Med 49: 634-641.
http://dx.doi.org/10.1002/aiim.20334
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
Thomas. DC. (1995). Re: "When will nondifferential misclassification of an exposure preserve the direction of a
trend?" [Letter]. AmJEpidemiol 142: 782-784.
A-66
-------
U.S. EPA (U.S. Environmental Protection Agency). (1988). 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). (1994). 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). (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. (EPA/630/R-03/003F). Washington, DC: U.S. Environmental
Protection Agency, Risk Assessment Forum.
http://www.epa.gov/raf/publications/pdfs/childrens supplement_final.pdf
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/pubrications/pdfs/benchmarkdose guidance.pdf
Veblen. PR: Wylie. AG. (1993). Mineralogy of amphiboles and 1:1 layer silicates. Rev Mineral Geochem 28: 61-
137.
Warheit. DB: Hartsky. MA. (1990). Species comparisons of proximal alveolar deposition patterns of inhaled
particulates. Exp Lung Res 16: 83-99.
Webber. JS: Blake. DJ: Ward. TJ: Pfau. JC. (2008). Separation and characterization of respirable amphibole fibers
fromLibby, Montana. Inhal Toxicol 20: 733-740. http://dx.doi.org/10.1080/08958370801932544
Weinberg. CR: Umbach. DM: Greenland. S. (1995). Weinberg et al reply [Letter]. Am J Epidemiol 142: 784.
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
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
A-67
-------
PARTICLE SIZE DISTRIBUTION DATA FOR
LIBBY AMPHIBOLE STRUCTURES OBSERVED IN AIR
AT THE LIBBY ASBESTOS SUPERFUND SITE
July 14, 2010
Prepared by:
U.S. Environmental Protection Agency
Region 8
Denver, CO
With Technical Assistance from:
SRC, Inc.
Denver, CO
B-l
-------
APPROVAL PAGE
This report, Particle Size Distribution Data for Libby Amphibole Structures Observed in Air at
the Libby Asbestos Superfund Site, is approved for distribution.
•77/9
Bonita LaveH
U.S. EPArtfegion 8
Date /
B-2
-------
PARTICLE SIZE DISTRIBUTION DATA FOR
LIBBY AMPHIBOLE STRUCTURES OBSERVED IN AIR
AT THE LIBBY ASBESTOS SUPERFUND SITE
1.0 INTRODUCTION
Libby is a community in northwestern Montana that is located near a large open-pit vermiculite
mine. Vermiculite from this mine contains varying levels of a form of asbestos referred to as
Libby Amphibole (LA). In 1999, EPA Region 8 initiated environmental investigations in the
town of Libby and in February, 2002, EPA listed the Libby Asbestos Site (the Site) on the
National Priorities List. The Site includes the former vermiculite mine and residential homes,
commercial businesses, schools and parks that may have become contaminated with asbestos
fibers as a result of vermiculite mining and processing conducted in and around Libby as well as
other areas in the vicinity that may have been impacted by mining-related releases of asbestos.
Historic mining, milling, and processing operations at the Site, as well as bulk transfer of
mining-related materials, tailings, and waste to locations throughout Libby Valley, are known to
have resulted in releases of vermiculite and LA to the environment.
As part of the response actions taken pursuant to the Comprehensive Environmental Response,
Compensation and Liability Act, EPA has performed a number of investigations to characterize
the nature and extent of LA contamination of air, soil, dust and other media in and around the
community of Libby. Because available information suggests that the toxicity of asbestos is at
least partially influenced by the size of the inhaled asbestos particles, these investigations have
included the measurement of the dimensions (length and width) of LA particles observed in
samples collected from the Libby site.
The purpose of this report is to summarize size distribution data for LA particles that have been
observed in air samples collected at the site, and to utilize these data to make comparisons
between various subsets of the data to determine if any important differences in particles size
distributions can be recognized.
2.0 METHODS
2.1 Data Overview
EPA has been collecting samples of air since 2001 at the Libby site. Table 1 provides an
overview of the sampling programs that have generated these data. The raw data for the air
samples included in this assessment are provided in Appendix A (of this report).
Most of the samples that have been collected have been analyzed for asbestos by transmission
electron microscopy (TEM) using either ISO 10312 (1995) or AHERA (1986) counting rules, as
modified by site-specific modifications as described in modifications forms LB-000016 and
LB-000031 (provided in Appendix B). In all cases, the data that are recorded during the analysis
B-3
-------
of a sample include the length, width, and aspect ratio (length/width) of all particles that meet the
counting rules specified for the analysis.
2.2 Data Presentation
One convenient method for comparing the size distributions of two different sets of LA particles
is through a graph that plots the cumulative distribution function (CDF) for each particle set.
This graphical format shows the fraction of all particles that have a dimension less than some
specified value. This format is used in this document to present the distributions of length,
width, and aspect ratio.
There are a number of statistical tests that can be used to compare two distributions in order to
support a statistical statement about whether the distributions are "same" or "different." Such
comparisons are complicated by the fact that the distributions may be similar over some intervals
and dissimilar over other intervals. However, at present, data are not sufficient to know which
parts of the distribution are most important from a toxicological perspective. Therefore, this
document relies upon simple visual inspection to assess the degree of difference between various
regions of differing distributions.
B-4
-------
3.0 RESULTS
3.1 Data Validation
The Libby2 database and Libby OU3 database have a number of built-in quality control checks
to identify unexpected or unallowable data values during upload into the database. Any issues
identified by these automatic upload checks were resolved by consultation with the analytical
laboratory before entry of the data into the database. After entry of the data into the database,
several additional data verification steps were taken to ensure the data were recorded and entered
correctly. A total of 29,504 LA structures are included in Table 1. Of these structures, 25%
have undergone data validation in accord with standard site-wide operating procedures (SRC,
2008) to ensure that data for length, width, particle type, and mineral class are correct. Of the
structures that have undergone validation, only 39 of 7,464 (0.5%) structures had errors in
length, width, or mineral class. These errors were corrected and the database updated as
appropriate.
3.2 Consolidated Data Set
Originally, most samples of air at Libby were analyzed using a counting rule based on a fiber
aspect ratio of 5:1. More recently, most air samples are counted using an aspect ratio rule of 3:1.
Because this rule has varied over time, Libby-specific laboratory modifications LB-000016 and
LB-000031 (see Attachment 1) were created to document the historic modifications and
instructions that laboratories have followed throughout the Libby program.
Figure 3-1 presents the particle size distributions for 29,504 LA particles observed to date1 in air
samples collected at the Libby Asbestos Superfund site that have an aspect ratio of 5:1 or more,
along with the distributions for 11,451 particles that were counted using an aspect ratio rule of
3:1. As seen, the distributions are very similar. This is because the number LA particles that
have an aspect ratio > 3:1 and < 5:1 is a relatively small fraction of the total (7%).
For simplicity, all remaining analyses focus on the set of particles with an aspect ratio of 5:1 or
more.
3.3 Frequency of Complex Structures
Asbestos particles occur not only as fibers but also in more complex structures including
bundles, clusters, and matrix complexes. The frequency of these structure types in air samples
from Libby are summarized below:
'Based on a query of the Libby2 database on 12/08/09 and the Libby OU3 database on 2/9/10.
B-5
-------
Type2
Fiber
Bundle
Matrix
Cluster
Total
Number
23,933
2,366
3,150
54
29,504
Frequency
81%
8%
11%
0.2%
100%
As shown, most (81%) of the enumerated structures are fibers, with less than 20% complex
structures.
3.4 Comparisons of Stratified Data Sets
The data sets shown in Figure 3-1 are based on air samples that were collected at a number of
different locations around the site, and which were analyzed by several different methods. In
order to investigate whether there are any important differences in size distributions between
operable units, sampling locations (indoor, outdoor), activity (e.g., active or passive), and/or
analytical method, the consolidated data set was partitioned into a number of subsets, as follows:
Figure
Comparison
3-2
LA particles observed in air stratified by structure type
3-3
LA particles observed in air stratified by Operable Unit
3-4
LA particles observed in air stratified by sample type (ambient, indoor, outdoor ABS)
3-5
LA particles observed in air stratified by preparation method (direct vs indirect)
3-6
LA particles observed in air stratified by analysis method (ISO vs AHERA)
Figure 3-2 is a comparison of different structure types (fiber, bundles, and matrices). Clusters
were not included because there were too few for a distribution to be meaningful. As seen, the
length distribution for matrix particles is somewhat left-shifted compared to fibers. This is
perhaps expected because some portion of the fiber length in matrix fibers is obscured by the
matrix particle. In contrast, the length and thickness distributions for bundles are right-shifted
compared to fibers. This is expected because a bundle is several fibers lying in parallel.
Figure 3-3 compares the size distributions of LA at different operable units (OUs) at the site. As
seen, there appears to be little difference in structures from the different OUs.
2In some cases, the structure type assignment provided by the laboratory was not a valid choice according to the
recording rules for the specified analysis method. Table A-l in Appendix A presents the types of invalid structure
types and the structure class assumption that was made in order to include the structure in this report.
-------
Figure 3-4 shows the distribution of structure sizes for different types of air samples. Samples
have been placed into three groups: ambient air, indoor ABS, and outdoor ABS. As shown, the
length and width distributions for indoor and outdoor ABS samples are relatively similar, while
the length and width distribution for ambient air samples appear to be right shifted. However,
this observation should be considered to be relatively uncertain because of the small number
(136) of particles that constitute the ambient air data set.
Figure 3-5 compares the size distributions for samples using direct and indirect preparation
methods. As shown, there is little difference in the distributions or either length of width,
suggesting that preparation method does not have a significant impact on particle size.
Figure 3-6 compares the particle size distributions as a function of analytical counting rules. As
shown, the length and width distributions for particles analyzed using AHERA rules tend to be
somewhat right-shifted relative to the distributions for particles analyzed using ISO 10312 rules.
This apparent difference might be related either to differences in counting rules between
methods, or possibly to differences in the nature of samples analyzed by each method. In either
event, the difference between methods appears to be relatively small.
4.0 SUMMARY
Particle size data are available for nearly 30,000 LA structures that have been observed in air
samples collected at the Libby Asbestos Superfund site. Most (about 80%) LA particles are
fibers, with less than 20% complex structures (bundles, clusters, or matrices). LA particle
lengths typically range from a little less than 1 um up to 20-30 um, and occasionally higher. The
average length is about 7 um. Thicknesses typically range from about 0.1 um up to about 2 um,
with an average of about 0.5 um. Although some variations occur, particle size distributions are
generally similar between different locations and between different types of samples.
B-7
-------
APPENDIX A
RAW DATA: LA STRUCTURE DATA FROM THE LIBBY 2 DATABASE AND THE
LIBBY OU3 DATABASE
Libby2DB based on a download date of 12/8/09
Libby OU3 DB based on a download date of 2/9/10
See attached compact disc.
-------
APPENDIX B.
LIBBY-SPECIFIC LABORATORY MODIFICATION FORMS
LB-000016
LB-000031
Table 1. Air Sample Collection Programs
Program
Phase 1
Phase 1R
Phase 2
Phase 2R
CSS
SQAPP
Ambient Air
(AA)
OU4 Indoor/
Outdoor ABS
Indoor
Schools
Outdoor
Schools
Phase 2
(OU3)
Phase 3
(OU3)
Clean-up
Evaluation
Other
Program Description
Initial investigation sampling to assess nature and extent of potential
contamination. Includes source areas (e.g., screening plant, export plant),
commercial buildings, and residential properties.
Monitoring and confirmation sampling as part of clean-up activities.
Activity-based sampling (ABS) included four scenarios: 1) routine indoor
activities, 2) active cleaning, 3) simulated remodeling disturbances, 4) garden
rototilling.
Monitoring and confirmation sampling as part of Phase 2
Contaminant Screening Study of Libby properties to determine need for
remediation.
Sampling to address risk assessment data gaps. Included indoor ABS (routine
activities) and outdoor ABS (raking, mowing, playing), as well as clean-up
evaluation samples.
Ambient air monitoring program for 14 stations in OU4, 2 stations in OU2, 2
stations in OU6. Samples represent long-term (continuous 5-day) collection
periods.
Sampling to assess exposures during indoor ABS (passive & active activities)
and outdoor ABS (raking, mowing, playing) in OU4.
Stationary air sample collection from within Libby public schools
Outdoor ABS sampling from Libby public schools simulating exposures to
students and maintenance staff.
Ambient air sampling. Samples represent long-term (continuous 5-day)
collection periods.
ABS air sampling of ATV riding, hiking, camp fire construction
Sampling to monitor air and dust levels after completion of clean-up activities at
31 properties.
Includes various site-specific sampling investigations (e.g., Stimson Lumber,
Flyway, BNSF) and smaller-scale sampling programs.
Program Date
Range
Dec 1999-present
Jun 2000-present
Mar-Nov2001
Apr 2008-Nov 2009
Apr 2003-Oct 2006
Jun 2005-Oct 2006
Oct 2006- Jun 2008
Jul 2007- Jun 2008
Dec 2008
Jul-Sept 2009
July-Oct 2008
Aug-Nov 2009
Nov 2003-Feb 2004
Aug 2001 -present
Sampling and
Analysis Plan (s)
U.S. EPA (2000)
U.S. EPA (2000)
U.S. EPA (2001)
U.S. EPAf2002a)
U.S. EPA(2005c)
U.S. EPA (2006);
QOQTc)
U.S. EPAQQOTb);
(20Q7a)
U.S. EPA(20Q8a)
U.S. EPA(2009a)
U.S. EPA(2008c)
U.S. EPA(20Q9b)
U.S. EPA (2003)
various
Number of
LA
Structures (a)
328
18,525
867
1,717
3
1,456
136
5,603
9
5
67
59
5
731
(a) Restricted to LA structures recorded in accordance with a 5:1 aspect ratio rule.
LA structure counts are based on a download of Libby 2DB performed on 12-8-09 and the Libby OU3 DB on
2-9-10.
B-9
-------
Other
Program
1A
BN
CR
DM
El
EP
FC
FL
SL
LA Structures
9
17
3
1
1
104
184
146
266
Description
AIRS Site (418 Mineral Ave)
BNSF
Cumulative Risk Study
Demolition Sampling from 2006 only
BNSF Rail Yard Exclusion Zones
Export Plant
Flower Creek
WR Grace (Flyway site)
Stimson Lumber
B-10
-------
Figure 3-1. Particle Size Distributions of LA Particles in Libby Air Samples
Length
0.9 •
0.8 •
L_
O
0.1
^x7"
j
S
/
y
1
i
j
/
x-
X
'
1 10
Length (um)
^^
^^^
MIArSamples
100
Width
0.7 •
L_
o °'5
0.4 -
n ^ -
S
}
^
.r
j
•»
^
t
^^
J
0.01 0.1 1
Width (um)
^— AIIAir Samples
10
Aspect Ratio
n R -
u_
0
0.3 •
0.2 •
0.1 •
/
/
/
f
/
/
/
/
0.1 1 10
Aspect Ratio
/
,/
^
f
^
^
^^•™
All Air Sam pies
100
Data are based on a download of Libby 2DB performed on 12-8-09 and the Libby OU3 DB on 2-9-10.
All Air Samples
Number of Structures (29,504)
Type
F
B
M
C
Number Frequency
23,933
2,366
3,150
54
81%
8%
11%
0.2%
B-ll
-------
Figure 3-2. Particle Size Distributions of LA Particles in Libby Air Samples by Structure Type
0.9 •
0.8 •
0.7 •
U-
Q 0.5 •
O
0.4 •
0.3 •
Fibers
Matrices
Bundles
0.1
^^
\
xV
'S
x'
/
/
s
J
/
/
f
/
/
^
/
/
>
/
/
f
^
1 10 100
Length (um)
0.7 -
0.6 -
u_
O
0.1 -
Fibers
Matrice
Bundle
0.01
rjs
^
r
p
y
t
^
-r
)
/
-*•
^
^
s
—
iX
-^
-,
r ^
r
/
r
s*
/
s
.-^
^^
0.1 1 10
Width (um)
1.0
0.9 -
0.8 •
0.7
0.6
LJ_
Q 0.5
O
0.4 •
0.3
0.2
0.1
0.0
•Fibers
-Matrices
•Bundles
0.1
Aspect Ratio
10
100
Data are based on a download of Libby 2DB performed on 12-8-09 and the Libby OU3 DB on 2-9-10.
Structure
Type
F
B
M
N Structures
23,933
2,366
3,150
Clusters have not been included in this figure because N = 54 and this in not believed to be a sufficient number of structures.
B-12
-------
Figure 3-3. Particle Size Distributions of LA Particles in Libby Air Samples by Operable Unit (OU)
1.0
0.9 -
0.8 -
0.7 -
0.6 -H
Q 0.5 -
O
0.4
0.3
0.2
0.1 -
0.0
•OU1
•OU2
•OU3
OU4
•OU5
7
0.1
1 10
Length (um)
100
0.01
1.0
0.9
0.8
0.7
0.6
°-5
0.4
0.3
0.2
0.1
0.0
•OU1
•OU2
•OU3
OU4
•OU5
0.1
Aspect Ratio
100
Data are based on a download of Libby 2DB performed on 12-8-09 and the Libby OU3 DB on 2-9-10.
OU
1
2
3
4
5
N Structures
447
7,421
4,382
13,005
335
B-13
-------
Figure 3-4. Particle Size Distributions of LA Particles in Libby Air Samplesby Air Type
0.7 •
L_
O
n ^ •
0.1 •
^^IndoorABS
^^"Outdoor AmbientAir
Outdoor ABS
0.1
1
/
J{
* ^
S~
/
/
/
/
/
}
(
'
j
f
/
/
/*
f
r
/
>
f
S^JT~
y y
I
\
?=•
^^
^M
10 100
Length (urn)
0.9 •
0.8 •
n R .
L_
r\ n R .
0
0.1 •
n n •
^^IndoorABS
^^"Outdoor AmbientAir
Outdoor ABS
0.01
r
^
I
f <
7
J
}
I
J
)
f
f
J
/
J
?
/
s
sT
/
0.1 1 10
Width (urn)
0.9 •
0.8 •
0.7 •
n R •
L_
0
n 9 •
O-i .
n n •
Indoor ABS
^^"Outdoor Ambient A r
^^OutdoorABS
0.1
,,
r
/
J
j
/
f
j
/
f /
} /
J
j
x*
/
'
-?
y
z
J
f
:=
M
1 Aspect Ratio 10 1 00
Data are based on a download of Libby 2DB performed on 12-8-09 and the Libby OU3 DB on 2-9-10.
Samples Source
Ambient Air
Indoor ABS
Outdoor ABS
N Structures
136
891
5,953
B-14
-------
Figure 3-5. Particle Size Distributions of LA Particles in Libby Air Samples by Preparation Method
0.9 •
0.8 •
0.7 •
0.6 •
Q 0.5 •
O
0.4 •
0.3 •
0.2 •
0.1 •
n n •
^^D rect
Indirect
0.1
^X^
t
?
A
y
/>
s
/.
!/
/
/
/
fj
t
^•~-
JT
'/
^^
1 10 100
Length (urn)
1.0
0.9
0.8
0.7
0.6
LL
s °-5
0.4
0.3
0.2
0.1
0.0
D rect
Indirect
0.01
0.1
Width (urn)
10
1.0
0.9
0.8
0.7
0.6
LJ-
Q 0.5
O
0.4
0.3
0.2
0.1
0.0
•Direct
•Indirect
0.1
Aspect Ratio
10
100
Data are based on a download of Libby 2DB performed on 12-8-09 and the Libby OU3 DB on 2-9-10.
Preparation
Direct
Indirect
N Structures
17,578
11,926
B-15
-------
Figure 3-6. Particle Size Distributions of LA Particles in Libby Air Samples by Analysis Method
0.9 •
U.o •
Q 0.5 •
O
0.4 •
n n •
ISO
^— AHERA
0.1
^
r-i<
f
1
>
^
/
2
,
/
/
/
{
J
f
'
^
/
t
r
/
^^
Y
r
i
^
^—
10 100
Length (um)
0.8 •
0.6 •
LJ_
Q 05-
O
n A •
n ^ .
01 .
n n .
ISC
^— Ah
ERA
0.01
^
r
.
A
j\
f
j
s=f-
^
^
J
^
j
s
t
»
^
^^*
0.1 1 10
Width (um)
1 n -
0.9 •
n R .
u_
Q 0.5 •
O
0.4 •
n ^ .
n o .
01 .
n n •
^— ISO
AH
0.1
=KA
]
^
^
>
//
^
/
//
//
//
M
//
7
^^
f/
^Tl
1 Aspect Ratio 10 100
Data are based on a download of Libby 2DB performed on 12-8-09 and the Libby OU3 DB on 2-9-10.
Analysis Method
ISO
AHERA
N Structures
12,657
16,847
B-16
-------
5.0 REFERENCES
AHERA. (1986). Asbestos hazard emergency response act of 1986. (Pub. Law No. 99-519). Washington, DC: U.S.
House of Representatives, 99th Congress. http://www.gpo.gov/fdsvs/pkg/STATUTE-100/pdf/STATUTE-
100-Pg2970.pdf
ISO (International Organization for Standardization). (1995). Ambient air ~ Determination of asbestos fibres -
Direct transfer transmission electron microscopy method [Standard]. (ISO 10312:1995). Geneva,
Switzerland: International Organization for Standardization (ISO).
http://www.iso.org/iso/iso catalogue/catalogue to/catalogue detail.htm?csnumber=18358
SRC (Syracuse Research Corporation). (2008). Standard operating procedure for TEM data review and data entry
verification. (SOP No. EPA-LIBBY-09 (rev 1)). Denver, CO: U.S. Environmental Protection Agency.
U.S. EPA (U.S. Environmental Protection Agency). (2000). Sampling and Quality Assurance Project Plan Revision
1 forLibby, Montana [EPA Report]. Denver, CO.
U.S. EPA (U.S. Environmental Protection Agency). (2001). Phase 2 Sampling and Quality Assurance Project Plan
(Revision 0) for Libby, Montana [EPA Report]. Denver, CO.
http://www.epa.gov/libbv/QAPP Addendum A.pdf
U.S. EPA (U.S. Environmental Protection Agency). (2002). Final sampling and analysis plan, remedial
investigation, contaminant screening study, Libby Asbestos Site, Operable Unit 4 [EPA Report]. Denver,
CO.
U.S. EPA (U.S. Environmental Protection Agency). (2003). Final sampling and analysis plan addendum, post clean-
up evaluation sampling, contaminant screening study, Libby Asbestos Site, operable unit 4 [EPA Report].
Denver, CO.
U.S. EPA (U.S. Environmental Protection Agency). (2005). Supplemental remedial investigation quality assurance
project plan for Libby, Montana. Revision 1. Denver, CO: U.S. Environmental Protection Agency, Region
8.
U.S. EPA (U.S. Environmental Protection Agency). (2006). Sampling and analysis plan for outdoor ambient air
monitoring at the Libby asbestos site. Revision 1 [EPA Report]. Denver, CO.
U.S. EPA (U.S. Environmental Protection Agency). (2007a). Sampling and analysis plan for activity-based indoor
air exposures, operable unit 4, Libby, Montana, Superfund site. Final [EPA Report]. Denver, CO.
U.S. EPA (U.S. Environmental Protection Agency). (2007b). Sampling and analysis plan for activity-based outdoor
air exposures, operable unit 4, Libby, Montana, Superfund site. Final [EPA Report]. Denver, CO.
U.S. EPA (U.S. Environmental Protection Agency). (2007c). Sampling and analysis plan for outdoor ambient air
monitoring - Operable units 1, 2, 5, and 6. Final addendum [EPA Report]. Denver, CO.
U.S. EPA (U.S. Environmental Protection Agency). (2008a). Final sampling and analysis plan Libby public schools
~ Stationary air sample collection Libby asbestos site, Libby, Montana [EPA Report]. Denver, CO.
U.S. EPA (U.S. Environmental Protection Agency). (2008b). Phase II sampling and analysis plan for operable unit 3
Libby asbestos Superfund site. Part B: Ambient air and groundwater [EPA Report]. Denver, CO.
U.S. EPA (U.S. Environmental Protection Agency). (2009a). Final sampling and analysis plan for activity-based
outdoor air exposures at Libby public schools Libby asbestos site, Libby, Montana [EPA Report]. Denver,
CO.
U.S. EPA (U.S. Environmental Protection Agency). (2009b). Remedial investigation for operable unit 3 Libby
asbestos superfund site. Phase III sampling and analysis plan [EPA Report]. Denver, CO.
B-17
-------
APPENDIX C. CHARACTERIZATION OF AMPHIBOLE FIBERS FROM ORE
ORIGINATING FROM LIBBY, MONTANA, LOUISA COUNTY, VIRGINIA, ENOREE,
SOUTH CAROLINA, AND PALABORA, REPUBLIC OF SOUTH AFRICA
The O.M. Scott plant in Marysville, Ohio manufactured a number of products, including
fertilizers, dyes, and pesticides that were bound to a vermiculite carrier as a delivery vehicle.
The plant received ore from Enoree, South Carolina, Louisa County, Virginia, Libby, Montana,
and Phalaborwa, Republic of South Africa which was processed in an exfoliation furnace to
produce vermiculite used in the manufacture of their commercial products. Only ore from South
Carolina was used in 1957 and 1958. From 1959 to 1971, ores from South Carolina and Libby
were used. From 1972 to 1980, ores from Libby, South Africa, and Virginia were used. No ore
from Libby was used after 1980. Only ore from South Africa and Virginia were used after 1980
(see Appendix F).
EPA Region 8 obtained samples of ore from Libby, South Africa, and Virginia from Dr.
James Lockey, University of Cincinnati, and analyzed the samples to determine the particle size
distribution (length, width, and aspect ratio) using transmission electron microscopy and energy
dispersive spectroscopy to identify the mineral composition of the amphibole fibers. Dr. Lockey
obtained the South African and Virginia ore samples from the Marysville facility in 1980 and the
Libby ore (Libby #3 ore) from an expansion plant in Salt Lake City, Utah, in 1981. EPA
received a sample of ore from Enoree, South Carolina from the USGS historical collection,
Denver, CO (Vermiculite Ore [BO-4], approximately 10% gangue, Zonolite Co. Mine, Travelers
Rest, South Carolina 8/27/58).
The ore from the Rainey Creek complex (Vermiculite Mountain Mine, Libby, Montana)
resides in large ultramafic intrusive bodies that are rich in biotite and pyroxenite, and biotitite, a
rock composed almost completely of biotite Meeker et al. (2003). The ultramafic intrusions are
cut by deposits of syenite and carbonatite and much of the biotite has been hydrothermally
altered to hydrobiotite and vermiculite (Meeker et al., 2003; Frank and Edmund, 2001). The
pyroxenite has been altered to fibrous soda-rich amphiboles and contacts with pyroxenite
surrounding the biotitite contain the vermiculite ore zone containing diopside, hydrobiotite, and
apatite. Fibrous and nonfibrous amphiboles are located in both veins and disseminated
throughout the intrusive rock along cleavage planes of pyroxene. Amphiboles from Vermiculite
Mountain had been referred to as soda tremolite, richterite, soda-rich tremolite, tremolite
asbestos, and richterite asbestos by a number of investigators. In 2000, Wylie and Verkouteren
(2000) identified winchite as the principle amphibole in the Vermiculite Mountain deposit based
on chemical investigation referencing the classification system of Leake et al. (1997) and optical
properties. Meeker et al. (2003) investigated amphibole types from the mine complex using
electron probe microanalysis and x-ray diffraction analysis and reported the presence of
winchite, richterite, tremolite, and magnesioriebeckite. Magnesio-arfvedsonite and edenite were
C-l
-------
detected in low abundance. The amphibole composition of the Libby Amphiboles is roughly
winchite, richterite, tremolite, magnesio-riebeckite, magnesio-arfvedsonite, and edenite
(84:11:6:<1 :<1 :<1). The O.M. Scott facility received ore from the Vermiculite Mountain Mine
complex, Libby, Montana from 1959 through 1980.
The Palabora Igneous Complex located near Phalaborwa, Republic of South Africa is the
location of the Palabora mine. The Palabora ore deposit shares many features with the
Vermiculite Mountain mine complex including zoned deposits with ultramafic rocks
(pyroxenite) and intrusion by alkalic rock, primarily syenite. The primary mica at Palabora is
phlogopite rather than biotite, and the primary alteration product that forms vermiculite ore is
hydrophlogopite rather than hydrobiotite (Schoeman, 1989).
The Palabora ore is reported to contain little or no asbestiform fibers based on polarized
light microscopy by the Institute of Occupational Medicine in Edinburgh (IOM Consulting,
2008). Crude vermiculite from the Palabora complex was also reported to be free of asbestiform
fiber by polarized light microscopy (IOM, 2006). In both reports, the analysis by polarized light
microscopy were conducted with a detection limit of 1 ppm and since no chrysotile or amphibole
structures were detected, no further analysis by electron microscopy and x-ray diffraction were
conducted.
The ore from the Virginia Vermiculite mine in Louisa County, Virginia is described as
mafic rock intruded by a series of small pegmatites (Gooch, 1957). Meisinger (1979) classified
the deposits as Type 3, similar to the ores from Enoree, South Carolina. The formations consist
of potassic ultramafic bodies primarily biotite. The vermiculite ores are found primarily in
hydrobiotite portions of the biotite intrusions. The hydrobiotite deposits are preferentially mined
because of better commercial properties compared to vermiculite.
There is limited information on the asbestos content of the ores from the Louisa deposit.
Rohl and Langer (1977) reported both chrysotile and amphibole fibers in six ore samples from
the Louisa deposit. The chrysotile was reported and fibers and bundles while the amphibole was
described as widely composed with most of the fibers classified as actinolite. Moatamed et al.
(1986) analyzed the Virginia, Palabora, and Libby ore samples and reported traces of fibrous
amphibole asbestos identified as actinolite and actinolite in the form of cleavage fragments
having low aspect ratios. Amphibole content for both unexfoliated and exfoliated ores ranged up
to 1.3% amphibole asbestos.
Ores from the Enoree, South Carolina deposits are primarily hydrobiotite and biotite in
origin. Fluroapatite is a common mineral collocated with the hydrobiotite. Zircon is also widely
dispersed throughout the plutons along with minor accessory minerals including talc, chlorite,
chromite, rutile, titanite, corundum, anatase, and amphibole asbestos (Hunter, 1950). The
amphibole asbestos identified in the vermiculite deposit at Enoree has been classified as
tremolite (Libbv. 1975).
C-2
-------
Briefly, samples of ore and vermiculite were prepared following the procedure outlined
by Bern et al. (2002). Samples were dried, ground with a Wylie mill and mortar and pestle and
sieved through a 230-|im (60 mesh) sieve. Samples (exactly 2.0 g were mixed with 18 g of
analytical silica sand and placed in a fluidized bed asbestos segregator vessel to load 25 mm
MCE air sampling filters (0.8-jim pore size) (Januch et al., 2013). The fluidized bed asbestos
segregator was run for 3 minutes to load the filter cassettes with sufficient fibers for analysis by
transmission electron microscopy (TEM). The fluidized bed asbestos segregator preparation
method allows for analytical sensitivity for fiber detection in the range of 0.002% by mass
(Januch et al., 2013). Three filters were loaded for each of the ore and vermiculite samples.
After loading, the filters were prepared for TEM analysis by mounting on copper girds, carbon
coating, and subjected to TEM analysis (TEM-ISO 10312 method).
The laboratories followed fiber counting rules detailed in the Sampling and Analysis Plan
for the specific study. Total amphibole fibers and Phase Contrast Microscopy equivalent
(PCMe) fibers were counted for each of the ore/vermiculite samples. A total of 1.0-mm2 area or
a total of approximately 100 grid openings were counted for each filter to achieve the desired
analytical sensitivity. Energy dispersive spectroscopy (EDS) was performed on selected samples
from each of the vermiculite/ore samples to provide mineral characterization of individual fibers.
Fiber counts were recorded on NADES data sheets for further analysis. Only the Libby
vermiculite and Libby ore samples had sufficient fibers detected to perform a fiber size
distribution.
Fiber counts were determined by counting fiber numbers for a specific area of the filter
grid or a specific number of grid openings (whichever was achieved first) to determine total
fibers present. As shown in Table C-l, the number of fibers for the test materials varied greatly
depending on the source.
-------
Table C-l. Fiber detected in ore and expanded product
Sample type
Enoree (BO-4) ore
Virginia ore
Virginia expanded
South Africa ore
South Africa expanded
Libby # 3 ore
Libby expanded
Grid openings
285
146
146
146
146
148
153
Structures counted
LA
6
0
1
1
0
320
108
OA
1
0
0
0
0
0
0
c
0
0
0
2
0
0
0
Concentration (s/g)
LA
14,300
0
4,336
4,401
0
1,393,873
468,213
OA
3,400
0
0
0
0
0
0
C
0
0
0
8,801
0
0
0
LA = Libby amphibole; OA = Other amphibole; C = Chrysotile.
Note: the designation of fibers as LA in this instance reflects only a qualitative morphological comparison to
amphiboles of the Libby series.
The Libby #3 ore and the Libby #3 expanded material contained the greatest number of
fibers both in fiber counts on the filters and in calculated structures per gram of material.
Virginia expanded and South African ore contain amphibole structures represented by low fiber
counts. South African ore also contained chrysotile fibers as determined by morphology and
EDS analysis. The estimation of structures per gram of material indicated that there were
4,000 amphibole fibers per gram of material which was lower than the Libby ore samples.
Enoree ore contained amphibole fibers determined to be actinolite and anthophyllite based on
morphology and EDS analysis. Based on fluidized bed preparation, the ore contained
approximately 18,000 structures per gram of material which was lower than the Libby ore
samples. Numerous nonasbestiform minerals were also detected including biotite, micas, and
pyroxenes.
Amphiboles are a complex group of minerals characterized by double chains of silicate
tetrahedra and the generic chemical formula of: Ao-iEhCsTgC^OH^ where A, B, C, and T
represent the various cations. The modern classification system of amphiboles is described in
Leake etal. (1997). To classify the mineral species of the amphibole, it is not sufficient to
determine its composition; the various cations must be assigned to the specific A, B, C, and T
sites. The cutoffs of the compositional ranges allowed for each amphibole mineral species are
based on the number of the cations in the various sites. The methodology to classify an
amphibole is to first determine its elemental compositions (e.g., as expressed as weight
percentage oxide for each element or as atomic percentage for each element). Then a normalized
routine is applied to the raw elemental measurements to calculate the number of each of the
cations contained in one formula unit. (This is a simple arithmetic calculation since the cation
percentage have been measured and the stoichiometry must balance the charges of the cations
C-4
-------
and anions.) Generally, one formula unit is assumed to contain 23 oxygens. Next the sites are
filled up by assigning cations to them subsequently, specifically:
T: Si4+, A13+, and Ti4+
C: A13+ and Ti4+ (only after the T sites are filled first) and then Mg2+, Fe2+, Fe3+, and then
Mn2+.
B: Any remaining Mg2+, Fe2+, and Mn2+ (after the C sites are filled), all Ca2+, then Na+ if
there is any room left.
A: Na+and K+only
Once the cations are assigned to their sites, it is a simple matter to classify the minerals
based on the cutoffs of the compositions field allowed for each mineral.
The Libby amphibole group of minerals is a complex group of amphiboles consisting of
six minerals:
• Winchite, CaNa[Mg, Fe2+]4[Al, Fe3+]Si8O22[OH]2
• Richterite, NaCaNa[Mg, Fe2+, Mn, Fe3+]5Si8O22[OH]2
• Tremolite, Ca2Mg5Si8O22[OH]2
• Magnesio-riebeckite Na2[Mg3, Fe3+2]SigO22[OH]2
• Magnesio-arfvedsonite NaNa2[Mg4,Fe3+]SigO22[OH]2
• EdeniteNaCa2Mg5Si7AlO22[OH]2
Although this looks complex, the matter is simplified by the fortunate fact that all Libby
amphibole is characterized by a low amount of Al in the T site (and a correspondingly high Si
content). So, according to Leake's classification, if the Si (expressed as atoms per formula unit
[apfu]) is at least 7.5 and Al content in the T site is <0.5, all six Libby amphibole types can be
plotted on a graph of Na content of the B site versus the (Na + K) content in the A site. This
approach was described by Meeker et al. (2003) for the Rainy Creek complex.
Quantitative EDS spectra (TEM/EDS) were collected from all amphibole fibers found in
the South Africa, South Carolina, and Virginia samples, and six randomly-selected LA fibers in
each of the Libby ore and Libby expanded samples. Two bundles of asbestiform serpentine
(chrysotile) were found in the South Africa ore sample. EDS spectra were collected for one of
the bundles. The chemical formula of serpentine is Mg3Si2Os[OH]4. The EDS software package
collected and summarized each spectrum to determine the atomic percentage of each element of
interest.
Several assumptions were made in the treatment of the EDS data:
C-5
-------
1) Numbers of cations per formula unit are calculated on the basis of 23 oxygens. This
may or may not be correct, since an (OH) site in the amphibole crystal can be
occupied by either OH", F", Cl", or O2". The calculated cation numbers will be affected
if a significant quantity of O2" is in the OH site.
2) A persistent problem with amphiboles is that they can contain both ferric (3+) and
ferrous (2+) iron in the same crystal. For the purposes of this report all Fe was
assumed to be Fe2+. A routine for calculating the ratio of Fe2+to Fe3+ is described in
Leake etal. (1997) but it is very complex, applies to polished sections, and was not
attempted for this report.
3) For the purposes of this report, the T sites were filled completely full to 8 apfu with
all Si and then Al and Ti. The C sites were then filled to 5 apfu with any remaining
Al and/or Ti and then with Mg and Fe2+. All Ca and any Mg and, Fe remaining after
the C site was full were then assigned to the B site. Next, Na was assigned to the B
site until it was full (apfu), then any remaining Na and all K was assigned to the A
site. Quantitative EDS measurements were calibrated with the USGS's BIR-1G
basalt glass standard and the feldspar minerals albite and orthoclase.
Application of these assumptions to the TEM/EDS data produces a useable graph of the
Na and K content of the amphibole fibers. As shown in Figure C-l, Libby #3 ore and Libby #3
Expanded amphiboles were characteristic of winchite, richterite, edenite, and
tremolite-actinolite. Virginia Expanded and South African ore both contained amphibole fibers
characteristic of non-Libby (Na and K negative) in the tremolite series. Compositions of
amphibole fibers from the Libby Starting Material, which is a mixture of LA minerals traceable
from the mine at Libby and used as a reference material by environmental laboratories, is shown
on Figure C-l for comparison.
C-6
-------
Sodium and Potassium Content of Individual Amphibole
Fibers as Measured by TEM/EDS
HM
1 °-9 '
•2 0.8 -
lo.7-
O
"£ 0.6
01
Q. „ r-
Kin A Site (atoms
3 O O O O C
3 i-> k> Lu 4* U
X
Edenite Richterite Magnesio-
Arfvedsonite
• * *
, •
• * •
X
>* Magnesio-
k ^ * Winchite Riebeckite
, Tremolite-
>• X
{ Actinolite
* Libby #3 Ore
• Virginia Expanded
A South Africa Ore
X Enoree, SCOre
X Libby #3 Expanded
Libby Starting Material
£ 0.0 0.5 1.0 1.5 2.0
Na in B Site (atoms per formula unit)
Figure C-l. Cation values for Na in the B site and the Na + K in the A site
from individual amphibole fibers.
Following all assumptions described above and the approach of plotting Na in the B-site
versus Na + K in the A site as described by Meeker et al. (2003), the mineral species of the
Marysville fibers can be described as:
• The single Virginia amphibole asbestos fiber is an actinolite
• The single South African amphibole fiber is a tremolite
• Six of the Enoree amphibole fibers are actinolite and one is anthophyllite (OA)
• Five of the LA fibers from Libby are winchite
• One of the LA fibers is a richterite
• Two of the LA fibers are edenite
• Four of the LA fibers from Libby are actinolite
Actinolite, which has the chemical formula of Ca2(Mg,Fe2+)5SigO22[OH]2, is part of a
solid solution series with tremolite and occurs when some Mg is substituted by Fe2+. Actinolite
was not found in Meeker et al. (2003) analyses of samples from the mine at Libby, however,
some of those tremolite analyses would be classified as actinolite if all Fe was treated as Fe2+
(Meeker et al., 2003), which is how the analyses described above were treated.
Fiber size distributions for amphibole fibers from the Libby #3 ore and Libby #3
Expanded sources were conducted on the fibers counted during the TEM analysis of the filter
C-7
-------
grids (see Figure C-2). Due to the low number of fibers detected in the Virginia and South
Africa sources, it was not possible to develop a fiber size distribution for these fibers. The LA
fiber size data were plotted as a cumulative distribution frequency for fiber length, fiber width,
and aspect ratio. These data were compared to LA fibers collected in Libby as part of EPA's
ongoing ambient air monitoring program. The Libby ore and expanded material showed an
increased frequency of longer and wider fibers than the fibers from the Libby ambient air
sampling program. Aspect ratios were nearly identical. The differences between the length and
width frequency were not outside of the expected range for LA fibers and were consistent with
fiber size distributions for soil activity-based-sampling data from Libby.
Based on the TEM morphological analysis of filter grids, TEM/EDS analysis for the fiber
mineralogy, and the fiber size distribution data, it can be concluded that the amphibole fibers
detected in the Libby #3 ore samples from the Salt Lake Expansion facility are consistent with
data from authentic Libby amphibole fibers (Meeker et al., 2003) found in Libby, Montana.
Further, ore samples from Virginia and South Africa contained amphibole and chrysotile fibers
but at a much lower frequency of detection than the Libby amphibole ore.
-------
1.0
0.9
0.8
0.7
0.6
L
Q 0.5
O
0.4
0.3
0.2
0.1
0.0
0.1
Length
Libby #3 Ore
Libby#3 Ore Expanded
LibbyAir
Length (urn)
10
100
Q
O
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0,01
1.0
0.9
0.8
0.7
0.6
Q 0.5
O
0.4
0.3
0.2
0.1
0.0
Width
Libby #3 Ore
Libby #3 Ore Expanded
LibbyAir
0.1 1
Width (urn)
10
Aspect Ratio
•Libby#3Ore
-Libby #3 Ore Expanded
•LibbyAir
10
Aspect Ratio
100
1000
Figure C-2. Particle size distribution of LA amphiboles.
C-9
-------
C.I. REFERENCES
Bern. AG: Meeker. GP: Brownfield. I. (2002). Guide to analysis of soil samples from Libby, Montana for asbestos
content by scanning electron microscopy and energy dispersive spectroscopy. In Libby Superfund site
standard operating procedure, Attachment 3. (SRC-LIBBY-02 (Revl)). Washington, DC: U.S. Geological
Survey. http://www.iatl.com/content/file/methods/asbestos/Attachment%203%20-%20SOP%20SRC-
LIBBY-02%20Rev%201%20v2.pdf
Frank. D: Edmund. L. (2001). Feasibility for identifying mineralogical and geochemical tracers for vermiculite ore
deposits. (EPA 910-R-01-002). Seattle, WA: U.S. Environmental Protection Agency, Office of
Environmental Assessment, Region 10. http://www.epa.gov/rlOearth/OFFICES/OEA/risk/vermtracers.pdf
Gooch. EO. (1957). Vermiculite. Va Miner 3: 1-6.
Hunter. CE. (1950). Vermiculite of the southeastern states. In FG Snyder (Ed.), Symposium on mineral resources of
the southeastern United States (pp. 120127). Knoxville, TN: University of Tennessee Press.
IOM (Institute of Medicine). (2006). Asbestos: Selected cancers. Washington, DC: National Academies Press.
IOM Consulting. (2008). Sampling and analysis of crude vermiculite samples for possible asbestiform fibre and
quartz content (pp. 28). (609-02386). Surrey, England: Palabora Mining Co, Palabora Europe Ltd.
http://www.schundler.com/FIBER-MAIN-9-06.pdf
Januch. J: Brattin. W: Woodbury. L: Berry. D. (2013). Evaluation of a fluidized bed asbestos segregator preparation
method for the analysis of low-levels of asbestos in soil and other solid media. Analytical Methods 5: 1658-
1668. http://dx.doi.org/10.1039/C3AY26254E
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: Youzhi. 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.
Libbv. SC. (1975) The origin of potassic ultramafic rocks in the Enoree Vermiculite District, South Carolina.
(Master's Thesis). Pennsylvania State University, University Park, PA.
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.
Meisinger. AC. (1979). Vermiculite. In Minerals yearbook 1978 -1979, metals and minerals. Washington, DC: U.S.
Bureau of Mines.
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/S0013-9351(86)80183-9
Rohl. AN: Langer. AM. (1977). Mineral analysis of core samples from the Green Springs area. Virginia vermiculite
deposit: Unpublished letter report from Mt. Sinai School of Medicine (pp. 10). New York, NY: Mt. Sinai
School of Medicine.
Schoeman. JJ. (1989). Mica and vermiculite in South Africa. J S Afr Inst Min Metall 89: 1-12.
Wylie. AG: Verkouteren. JR. (2000). Amphibole asbestos from Libby, Montana: Aspects of nomenclature. Am
Mineral 85: 1540-1542.
C-10
-------
APPENDIX D. ANALYSIS OF SUBCHRONIC- AND CHRONIC-DURATION STUDIES
AND CANCER BIOASSAYS IN ANIMALS AND MECHANISTIC STUDIES
D.I. SUBCHRONIC- AND CHRONIC-DURATION STUDIES AND CANCER
BIOASSAYS
D.2. INHALATION
Davis et al. (1985) performed a chronic-duration inhalation study examining response to
tremolite asbestos. Specific-pathogen-free (SPF) male Wistar rats (n = 48) were exposed in a
chamber to 10 mg/m3 (-1,600 fibers/mL, >5 um) of commercially mined tremolite (South
Korea) for a total of 224 days (7 hours per day, 5 days per week) over a 12-month period. The
tremolite sample contained approximately 50% fibers 10 to 100 um long, using a fiber definition
of length >5 um, diameter <3 um, and aspect ratio >3:1. The results of the inhalation study
produced very high levels of pulmonary fibrosis, as well as 16 carcinomas and 2 mesotheliomas,
among the 39 tremolite-exposed animals (see Tables D-l and D-2). No pulmonary tumors were
observed in the controls.
Although Davis etal. (1985) did not describe the chrysotile data, the difference between
tremolite and chrysotile was stated to be statistically significant, with tremolite exposure
inducing more fibrotic and carcinogenic lesions (see Table D-l). These results show that rats
exposed to tremolite exhibited increased numbers of pulmonary lesions and tumors. Tumors
observed in other organ systems are also listed in Table D-2 and appear to be unrelated to
exposure. Although a method for an injection study is described in Davis etal. (1985), only the
inhalation results are presented. The injection study referenced in Davis et al. (1985) may be the
intraperitoneal injection experiments (Davis etal., 1991) using the same tremolite material.
Table D-l. Pulmonary fibrosis and irregular alveolar wall thickening
produced by tremolite exposure
Time after start of exposure
(number of rats examined)
Peribronchiolar fibrosis (SD)a
Irregular alveolar wall thickening (SD)b
Interstitial fibrosis (SD)b
12 mo
(» = 3)
23.0 (21.4-24.2)
35.2 (27.7-41.0)
0
18 mo
(» = 4)
13.4 (9.7-18.9)
27.7 (20.8-35.4)
3.0 (0-5.6)
27-29 mo
(» = 12)
-
-
14.5 (3.8-26.9)
SD = standard deviation.
Percentage of 100 squares counted in lung tissue area.
Percentage of total lung tissue area.
Source: Adapted from Davis et al. (1985).
D-l
-------
Table D-2. Tumors (benign and malignant) produced by tremolite exposure
Tumor site
Control (n = 36)
Tremolite (n = 39)
Pulmonary
Adenomas
Adenocarcinomas
Squamous carcinomas
Mesotheliomas
0
0
0
0
2
8
8
2
Other organ systems
Digestive/peritoneal
Urogenital
Endocrine
Musculoskeletal, integumentary
Reticuloendothelial/vascular
5
3
3
5
20
3
1
5
5
15
Source: Adapted from Davis et al. (1985).
Wistar rats were exposed for 13 consecutive weeks (6 hours per day, 5 days per week) to
either Calidria chrysotile asbestos or tremolite asbestos in a flow-past, nose-only inhalation study
(Bernstein et al., 2003) (see Table D-3). The tremolite samples had fiber counts of
100 fibers/mL of fibers longer than 20 um present in the exposure aerosol. Fibers were defined
as any object with an aspect ratio >3:1, length >5 um, and diameter <3 um, and all other objects
were considered nonfibrous particles. Counting was stopped when nonfibrous particle counts
reached 30, and fiber counting was stopped at 500 with length >5 um, diameter <3 um; a total of
1,000 fibers and nonfibrous particles were recorded (Bernstein et al., 2003). Lung tissue and
associated lymph nodes were examined by histopathology following tissue digestion. Associated
lymph nodes showed erythrophagocytosis (minimal severity) in one animal at all-time points,
compared to chrysotile and the control, which showed erythrophagocytosis (minimal severity)
only at 180 days.
D-2
-------
Table D-3. Chrysotile and tremolite fiber characteristics of fibers used in
inhalation exposure studies in rats
Fiber type
Chrysotile
Tremolite
Mean no.
fibers
evaluated
2,016
1,627
Mean no.
total
fibers/mL
48,343.2
3,128.1
Mean % total
fibers,
>20 um length
0.4
3.4
Mean diameter
um±SD
0.08 ±0.07
0.32 ±3.52
Mean length
um±SD
3.61 ±7.37
5.49 ±13.97
Diameter
range (um)
0.02-0.7
0.1-3.7
Length
range (um)
0.07-37.6
0.9-75
Source: Bernstein et al. (2003).
Table D-3 shows the comparison of number, concentration, and mean size distribution of
fibers used in this study. Note that the mean tremolite fiber diameter and length are much greater
than those of chrysotile, but the size ranges do overlap somewhat (Bernstein et al., 2003).
The long-term effects from the same exposure and counting methods discussed above
were described in Bernstein et al. (2005b), who present the full results through 1 year after
cessation of tremolite exposure in Wistar rats (n = 56). The long tremolite fibers, once deposited
in the lung, remain throughout the rat's lifetime. Even the shorter fibers, following early
clearance, remain with no dissolution or additional removal. At 365 days postexposure, the
mean lung burden was 0.5 million tremolite fibers >20 jim long and 7 million fibers 5-20 jim
long with a total mean lung burden of 19.6 million tremolite fibers. The tremolite-exposed rats
showed a pronounced inflammatory response in the lung as early as 1 day postexposure, with the
rapid development of granulomas (1 day postexposure) followed by the development of
pulmonary fibrosis characterized by collagen deposition within the granulomas. Increases in
alveolar macrophages and granulomas were observed at all-time points (1, 2, 14, 90, and
180 days) measured except 365 days. Pulmonary fibrosis increased starting at 14 days and
continued to be observed for up to 365 days. Slight interstitial fibrosis also was observed, but
only at 90 and 180 days postexposure. This study demonstrates that tremolite exposure leads to
pronounced inflammation and fibrosis (Bernstein et al., 2006). Tumors were not observed in this
study, and is a consistent observation with the time frame observed in other studies [i.e., 1-year
postexposure (Smith, 1978)1.
D.2.1. Intratracheal Instillation
A study by Putnam et al. (2008) was designed to explore gene-environment interactions
in the development of asbestos-related diseases. C57B1/6 mice were exposed once via
intratracheal instillation to Libby Amphibole asbestos (LAA)3 (Six Mix; 100 ug), crocidolite
3The 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.
D-3
-------
(100 ug), or saline (30 uL). Characteristics of fibers are described in Table D-4. Animals were
sacrificed, and the lungs were harvested 6 months postinstillation. The left lung was used for
ribonucleic acid (RNA) isolation, and the right lung was used for histology (email from
E. Putnam [University of Montana] to M. Gwinn [U.S. EPA] dated 02/26/09). Histology on
mouse lungs from each treatment group demonstrated an increase in fibrosis, as viewed by
Gomori's trichrome staining, following exposure to crocidolite and, to a lesser extent, LAA.
Histologic tissue was also exposed to Lucifer Yellow stain to further analyze variability in
collagen following exposure. Lucifer Yellow staining revealed an increase in collagen following
exposure to both crocidolite and LAA, but only crocidolite exposure led to a statistically
significant increase (p < 0.05). RNA was isolated from homogenized lungs and purified for use
in microarray analysis. Pooled RNA samples from mice in each exposure group were analyzed
on a lOK-element mouse oligonucleotide array (MWG Biotech), and expression was compared
to a mouse reference standard RNA. Gene-expression results were analyzed by GO Miner, and
genes exhibiting at least 1.25-fold upregulation or downregulation in treated lungs were
described. These included genes involved in membrane transport, signal transduction, epidermal
growth factor signaling, and calcium regulation for both crocidolite and LAA exposures, which
support the increase in collagen observed above. Some limitations to this study are the use of a
standard reference for gene-expression comparisons (as opposed to the saline controls), the
practice of describing genes only if a greater than twofold difference in expression is observed
and the use of pooled samples of homogenized whole lung that, in some cases, could dilute
variability among different areas of exposed lung (different lobes, fibrotic versus nonfibrotic).
Table D-4. Fiber characteristics for intratracheal instillation studies in mice
Material
LAA (Six Mix)
Crocidolite
Diameter (um)
0.61 ± 1.22
0.16 ±0.09
Length (um)
7.21 ±7.01
4.59 ±4.22
Aspect ratio
22.52 ± 22.87
34.05 ±43.29
Source: Smartt et al. (2010): Blake et al. (2007): Blake et al. (2008): Putnam et al. (2008).
A follow-up paper to Putnam et al. (2008) prepared by Smartt et al. (2010) examined the
increase of collagen in C57B1/6 mouse lung following exposure to crocidolite or LAA. The
paper also examined a few specific gene alterations by quantitative reverse transcription
polymerase chain reaction (RT-PCR). Animals (n = 3 to 6 mice per group) were dosed with the
same samples (see fiber characteristics in Table D-4 ) as described above (Putnam et al., 2008)
but were euthanized at 1 week, 1 month, and 3 months postinstillation. Treated mice were then
divided into two groups, with the left lung from the first group used for RNA isolation and the
right lung used for histology. The lungs from the second group were used for protein isolation
D-4
-------
and hydroxyproline assay (email from E. Putnam [University of Montana] to M. Gwinn
[U.S. EPA] dated 02/26/09). Similar to results from Putnam et al. (2008). Gomori's staining
demonstrated increased collagen and inflammation at the airways in lungs of mice exposed to
either LAA or crocidolite. These results were similar following exposure to both amphiboles,
with crocidolite effects appearing more severe at all-time points examined. No changes in the
pleura of the lungs that were indicative of potential mesothelioma were observed; such changes,
however, would not be expected in such a short time frame. This study also examined severity
of inflammation and found that, on average, crocidolite-exposed animals demonstrated minimal
inflammation at 1 week postinstillation, which then progressively worsened at 1 and 3 months
postinstillation. Although both asbestos exposures led to increased inflammation, LAA exposure
demonstrated minimal inflammation, which did not progress in the time points examined.
Gene-expression alterations were measured by quantitative RT-PCR for genes involved in
collagen accumulation and scar formation (CollAl, CollA2, and Col3Al). Although exposure to
both forms of asbestos at 1 week and 1 month postinstillation led to increased Col gene
expression, the levels and subtypes varied. LAA exposure led to increased gene expression of
CollA2 at 1 week postinstillation and Col3Al at 1 month postexposure, while crocidolite led to
no significant alterations in the expression of these genes. Both crocidolite and LAA exposure
led to increased CollAl gene expression as compared to the saline control at 1 week and
1 month postexposure. Due to these differences in expression, the authors also examined the
collagen protein levels in the lungs to compare with the gene-expression changes. Total collagen
content was determined by measuring the hydroxyproline content in the caudal aspect of the left
lung. As compared to saline-exposed mice, a significant increase in hydroxyproline was
observed at 1 week and 1 month following exposure to both crocidolite and LAA; however, only
lungs from crocidolite-exposed animals demonstrated a significant increase at 3 months
postexposure. These studies demonstrate that exposure to LAA lead to inflammation and
fibrosis, although with differences in the time and level of response from those of crocidolite.
Shannahan et al. (201 la) exposed two rat models of human cardiovascular disease (CVD)
to LAA4 to determine if the preexisting CVD in these models would impact lung injury and
inflammation following exposure. Healthy Wistar Kyoto (WKY) rats were compared to
spontaneously hypertensive (SH) and spontaneously hypertensive heart failure (SHHF) rats
following exposure. These rat models demonstrate pulmonary iron homeostasis dysregulation
(Shannahan et al., 2010). All rats (male only) were exposed to 0, 0.25, or 1.0 mg/rat via
intratracheal instillation and were examined at 1 day, 1 week, and 1 month postexposure. No
changes were observed histopathologically; however, changes were observed in markers of
homeostasis, inflammation, and oxidative stress. Bronchoalveolar lavage fluid (BALF) protein
was significantly increased in both the SH and SHHF rat models as compared to controls as early
4Median fiber dimensions as determined by TEM: length = 3.59 um; width = 0.23 um; aspect ratio >5:1.
D-5
-------
as 1 week postexposure. y-Glutamyl transferase (GOT) activity was increased in a
concentration-dependent manner with exposure to LAA at the earliest time point measured
(1 day), and was more pronounced in WKY rats as compared to SH and SHHF rats. Lactate
dehydrogenase (LDH) activity was also elevated in all strains but was more pronounced in the
SHHF rat model. Neutrophil increases were observed following exposure in all strains, peaking
at 1 day postexposure in all strains and persisting in the SH and SHHF rats until 1 month
postexposure. Macrophages showed similar results but persisted only in the SH rat model until
1 month postexposure. In order to determine any impact of exposure on iron homeostasis, BALF
ferritin and transferrin levels were measured in the lung. Increases in ferritin and transferrin
were observed in both SH and SHHF rats as compared to WKY controls. Nonheme iron was
also observed to be increased in only the SH rats at 1 day and 1 week postexposure. Markers of
inflammation (macrophage inflammatory protein [MIPJ-2) and oxidative stress (heme
oxygenase-1 [HO-1]) were elevated in both SH and SHHF as compared to WKY rats at baseline,
but limited exposure-related differences were observed. Limited changes were also observed in
ascorbate and glutathione (GSH) levels in BALF and lung tissue. Inflammation and cell injury
were observed in all strains (Shannahan et al., 201 la). In conclusion, this study showed the
potential for population variability related to CVD in response to exposure to LAA, including
markers of cellular injury, iron homeostasis, and inflammation.
Shannahan et al. (201 Ib) tested the hypothesis that LAA5 will bind iron and increase the
inflammogenic activity of fibers in vitro and acute lung injury and inflammation in vivo. The
authors examined the ability of LAA to bind exogenous iron in an acellular system and evaluated
iron-related alterations in the production of reactive oxygen species (ROS). The authors also
investigated the role of iron in the acute inflammogenic response in vitro, using human
bronchiolar epithelial cells, and in vivo using SH rats by modulating fiber-associated iron
concentrations. In a cell-free medium, LAA bound about 16 jig of iron/mg of fiber and
increased ROS generation about threefold. Generation of ROS was reduced by treatment with
deferoxamine (DEF), an iron chelator. To determine the role of iron in LAA ROS generation
and inflammation, BEAS2B cells (bronchiolar epithelial cell line) were exposed to LAA (50 jig),
iron-loaded LAA, or LAA treated with DEF. No conditions altered HO-1 or ferritin mRNA
expression. LAA by itself markedly increased IL-8 gene expression, which was significantly
reduced by iron-loaded LAA, but increased with LAA treated with DEF. To determine the role
of iron in LAA-induced lung injury in vivo, spontaneously hypertensive rats were exposed
intratracheally to either saline (300 jiL), DEF (1 mg), ferric chloride (21 jig), LAA (0.5 mg),
iron-loaded LAA (0.5 mg), or LAA plus DEF (0.5 mg). Neither ferric chloride nor DEF
increased BALF neutrophils compared to saline at 24 hours after treatment. LAA exposure led
to a statistically significant increase in BALF neutrophils (p <0.05). Loading of iron on LAA,
5Median fiber dimensions as determined by TEM: length = 3.59 um; width = 0.23 um; aspect ratio >5:1.
D-6
-------
but not chelation, slightly decreased inflammation (LAA + DEF > LAA > iron-loaded LAA). At
4 hours after exposure, LAA-exposed lung mRNA expression of MIP-2 was significantly
reduced in rats exposed to iron-loaded LAA, but increased by DEF
(LAA + DEF > LAA > iron-loaded LAA). Ferritin mRNA expression was elevated in rats
exposed to iron-loaded LAA compared to the LAA control. HO-1 expression was unchanged
following treatment with LAA. The study authors concluded that the acute inflammatory
response following exposure to LAA might be modified by the fiber's ability to complex iron,
rather than redox cycling of fiber-associated iron. The authors further concluded that iron
overload conditions may influence susceptibility to LAA-induced pulmonary disease.
Shannahan et al. (2012a) identified a number of serum biomarkers in healthy and CVD
rats following varying durations of exposure to LAA. These studies were conducted to
determine if asbestos-exposed healthy rats presented with biomarkers upregulated to CVD rats.
Rats were intratracheally instilled with 0, 0.25, 0.5, 1.0, or 5.0 mg in 300 uL. Four separate
study designs were employed. In the first study, WKY (healthy), SH (CVD), and SHHF (CVD)
rats were exposed to a single intratracheal instillation, and biomarkers were assessed 1 day and
3 months postexposure. In the second study, F344 rats were instilled once and samples were
collected 3 months and 1 year postexposure. In the third study, F344 rats were instilled biweekly
for 13 weeks and samples were collected 1 day and 2 weeks following the final instillation. In
the fourth study, WKY rats were instilled weekly for 4 weeks and serum samples were analyzed
1 day and 1 month following the final instillation. Acute-phase response (APR) molecules that
are involved in inflammatory responses such as a2-macroglobulin were upregulated 1 day after a
single instillation of 1 mg LAA in WKY and SH rats. In addition, 5 mg LAA increased
a2-macroglobulin 1 day and 2 weeks after the 13-week exposure. All other doses and exposure
endpoints did not affect a2-macroglobulin. Another APR molecule, al-acid glycoprotein, was
increased in WKY, SH, and SHHF rats 1 day following a single instillation and 3 months
postexposure in SH rats. In addition, al-acid glycoprotein was also increased 1 day and 2 weeks
after a 13-week exposure to 5.0 mg LAA in F344 rats. WKY rats also had non-dose-responsive
increases in al-acid glycoprotein 1 day after a 4-week exposure to 0.25 and 0.5 mg LAA. The
metabolic molecule lipocalin-2 was increased 1 day after a single instillation in WKY, SH, and
SHHF rats and 1 day and 1 month after a 4-week exposure. Biomarkers for cancer were largely
unaffected by LAA exposure. An exception to this was at 1 day after a single instillation in
WKY and SH rats, mesothelin was reduced in the serum. Altogether, the data suggest that the
modification of biomarker expression generally occurs rapidly and returns to homeostatic levels
1 day after instillation, regardless of duration.
In another study, Shannahan et al. (2012c) conducted a series of experiments to
determine the effect of LAA-induced pulmonary damage on the development of CVD, and to
identify early markers of lung and CVD in asbestos-exposed individuals. Three separate study
designs were utilized. In the first study, WKY, SH, and SHHF rats were instilled once with 0,
D-7
-------
0.25, or 1 mg LAA and examined 1 day, 1 week, 1 month, and 3 months postexposure. In the
second study, F344 rats were instilled once with 0, 0.15, 0.5, 1.5, or 5 mg LAA and examined
1 day, 3 days, 1 week, 2 weeks, 3 months, 1 year, and 2 years postexposure. In the third study,
F344 rats were instilled biweekly for 13 weeks with 0, 0.15, 0.5, 1.5, or 5 mg LAA and
examined 1 day, 2 weeks, and 2 years postexposure. WKY rats instilled with 1 mg LAA showed
a decreased rate of ADP-induced aggregation after 1 week, 1 month, and 3 months of exposure.
LAA at 1.5 and 5.0 mg increased platelet disaggregation 1 year postexposure in F344 rats. The
matrix metalloproteinase TIMP-2 showed a dose-dependent increase at 3 months postexposure in
F344 rats exposed to 0.25 and 1.0 mg LAA, but TIMP-2 was decreased in SH rats following
exposure to 1.0 mg LAA. Endothelial nitric oxide synthase and endothelin receptor-A (both
markers of vasoconstriction) were decreased and increased, respectively, in WKY rats at 1.0 mg
LAA. No other dose-responsive effects were noted for other inflammatory or vasoconstriction
markers. Altogether, these data suggest that LAA exposure may change the expression of some
biomarkers in healthy rats to resemble expression levels of cardiovascular compromised rats.
The role of inflammasome activation and iron in the development of LAA-induced
fibrosis was studied in Shannahan et al. (2012d). Male SH rats were instilled with a single
exposure to 0 or 0.5 mg LAA, DBF, 21 ug FeCb, 0.5 mg LAA + 21 ug FeCb, or 0.5 mg
LAA + 1 mg DEF. Tissues were collected 4 hours and 1 day postexposure. LAA instillation
increased gene expression in the lung of the inflammasome-related molecules cathepsin B,
Nalp3, NF-kfi, apoptosis-associated speck-like protein containing a CARD (ASC), IL-lfi, and
IL-6 expression 4 hours postexposure. Lung tissue expression of inflammatory cytokines CCL-7,
Cox-2, CCL-2, and CXCL-3 was increased 4 hours following LAA exposure. Conversely, LAA
exposure reduced IL-4 and CXCl-1 in the BALF. Finally, the ratio ofpERK/ERK, which is an
upstream activator of the inflammasome cascade, was increased in the lung of LAA-exposed rats
1 day postexposure. Rats treated with LAA + DEF or LAA + FeCb had significantly different
levels of Cox-2 in the BALF and IL-6 in lung tissue, but all other endpoints were not
significantly different. These data suggest that the concentration of iron does not impact the
activation of the inflammasome cascade and cytokines downstream of the pathway in
LAA-exposed animals.
In another study examining the role of iron in lung disease, Shannahan et al. (2012b)
evaluated the effect of Fe overload on LAA-induced lung injury in rats with CVD. WKY, SH,
and SHHF male rats were instilled once with 0, 0.25, or 1.0 mg of LAA. Blood, BALF, and lung
tissue were collected 1 week, 1 month, and 3 months postexposure. Gene array analysis
demonstrated that LAA exposure upregulated inflammatory-related genes such as NF-kfi and cell
cycle regulating genes such as matrix metalloproteinase-9 in WKY rats but inhibited these same
clusters of genes in SH and SHHF animals 3 months after instillation. Histological examination
of lung sections observed greater Fe staining of macrophages in SHHF rats compared to WKY
and SH rats at 1 and 3 months postexposure; however, no differences in the progression of
D-8
-------
pulmonary fibrosis were noted among the three strains. Altogether, these data do not suggest
that the iron overload conditions that are characteristic of the CVD strains amplify the pulmonary
effects of LAA.
Padilla-Carlin et al. (2011) investigated pulmonary and histopathological changes in male
F344 rats following exposure to LAA.6 The rats were administered a single dose of saline,
amosite (AM), (0.65 mg/rat), or LAA (0.65 or 6.5 mg/rat) by intratracheal instillation. At time
from 1 day to 3 months after exposure, bronchoalveolar lavage (BAL) was performed and the
right and left lung was removed for Rt-PCR and histopathological analysis, respectively. The
results showed that amosite exposure (0.65 mg/rat) resulted in a higher degree of pulmonary
injury, inflammation, and fibrotic events than the same mass dose of LAA. Both amosite and
LAA resulted in higher levels of cellular permeability and injury, inflammatory enzymes, and
iron-binding protein in both BALF and lung tissue compared to saline controls. In addition,
histopathological examination showed notable thickening of interstitial areas surrounding the
alveolar and terminal bronchioles in response to amosite and LAA. However, mRNA levels for
some growth factors (e.g., PDGF-A and TGF-1P), which contribute to fibrosis, were
downregulated at several time points. The authors concluded from this study that on a mass
basis, amosite produced greater acute and persistent lung injury.
In a continuation of the previous study, Cyphert et al. (2012b) compared the long-term
lung effects of LAA with amosite asbestos in the F344 rat. Male F344 rats were intratracheally
instilled with 0.65 or 6.5 mg LAA or 0.65 mg amosite in a single dose and monitored for 2 years.
At 2 years postexposure, there was a trend of increased collagen gene expression, a marker for
fibrosis, in all asbestos-exposed animals, but only the 0.65 mg dose of LAA reached statistical
significance. Mesothelioma markers, mesothelin (Msln) and Wilms' tumor gene (WT1), were
similarly increased in the lung at 1 year and 1 and 2 years postexposure to the low dose of LAA,
respectively. Epidermal growth factor receptor (EGFR) was increased in the lung at both doses
of LAA 2 years after instillation. Histological analysis noted a time-dependent and
dose-responsive increase in fibrosis scarring in LAA-exposed rats, but inflammation scoring did
not consistently induce dose-responsive or time-dependent increases in LAA-treated animals.
Fibrosis was significantly greater in the amosite-exposed animals at both 1 and 2 years
postinstillation. The data do not suggest that LAA induces significantly different types of effects
on carcinogenic, inflammatory, or fibrotic markers compared to amosite.
In another study establishing the pulmonary effects of different asbestos fibers, Cyphert
et al. (2012a) compared the effects of LAA with chrysotile and tremolite fibers on pulmonary
function in male F344 rats. Animals (eight/group) were treated with a single intratracheal
instillation of LAA (0.5 mg or 1.5 mg/rat), tremolite (0.5 mg or 1.5 mg/rat), and chrysotile
(0.5 mg or 1.5 mg/rat), and several markers of lung inflammation and injury were examined
6Median fiber dimensions as determined by TEM: length = 3.59 um; width = 0.23 um; aspect ratio >5.
D-9
-------
1 day and 3 months postexposure. After both, 1 day and 3 month exposures, both doses of LAA
exposure significantly increased the number of neutrophils in the BALF and biomarkers of lung
injury such as total protein, albumin, and LDH, relative to the control; however, the lung
alterations after 3-month exposures were greatly reduced relative to the 1-day data. Minimal and
mild levels of fibrosis were observed in the lung histopathology after 3 months in the low- and
high-dose levels of LAA. Relative to other fibers tested in these series of experiments, the LAA
fibers induced less fibrosis than the chrysotile fibers but were more pathogenic than the tremolite
sample. The study concluded that the severity of fibrosis is correlated to the length and aspect
ratio of the fibers.
In an early study, Sahu et al. (1975) described histological changes in the lungs of mice
exposed individually to amosite, anthophyllite, and tremolite. Fibers were described only as
<30-um long. Groups of 20 male albino Swiss mice were exposed to amosite, anthophyllite, and
tremolite at a single dose of 5 mg, and two animals from each group were sacrificed at 1,2, 7,
15, 30, 60, 90, 120, and 150 days postexposure. Microscopic results following exposure to
tremolite showed acute inflammation of the lungs at 7 days postexposure, including macrophage
proliferation and phagocytosis similar to that observed with amosite and anthophyllite. Limited
progression of fibrotic response was observed at 60 and 90 days postexposure, with no further
progression of fibrotic response.
Blake et al. (2008) and Pfau et al. (2008) examined the role of amphibole asbestos in
autoimmunity with both in vitro and in vivo assays. Blake et al. (2008) performed in vitro assays
with LAA, and both studies performed the in vivo assays with tremolite. C57BL/6 mice were
instilled intratracheally for a total of two doses each of 60 ug saline and wollastonite or Korean
tremolite sonicated in sterile phosphate buffered saline (PBS), given 1 week apart in the first
2 weeks of a 7-month experiment. Detailed fiber characteristics were described in Blake et al.
(2007) for wollastonite and LAA, but not for Korean tremolite (see Table D-4; wollastonite and
Korean tremolite not shown).
Blake et al. (2008) described autoantibody production following exposure to wollastonite
or tremolite, monitored biweekly with blood samples from saphenous vein bleeds and then by
cardiac puncture following euthanization. Specific autoantibodies were identified by
immunoblotting with known nuclear antigens. These autoantibodies were then incubated with
murine macrophage cells previously exposed to LAA, wollastonite, or vehicle control (binding
buffer containing 0.01 M HEPES, 0.14 M NaCl and 2.5 mM CaCh). Only sera from mice
exposed to tremolite showed antibody binding colocalized with SSA/Ro52 on the surface of
apoptotic blebs (Blake et al.. 2008).
In Pfau et al. (2008), serum and urine samples were collected and checked for protein
biweekly for 7 months following exposure to wollastonite or tremolite. By 26 weeks, the
tremolite-exposed animals had a significantly higher frequency of positive antinuclear antibody
tests compared to wollastonite and saline. Most of the tests were positive for dsDNA and
D-10
-------
SSA/Ro52. Serum isotyping showed no major changes in immunoglobulin subclasses (IgG,
IgA, IgM), but serum IgG in tremolite-exposed mice decreased overall. Furthermore, IgG
immune complex deposition in the kidneys increased, with abnormalities suggestive of
glomerulonephritis. No increased proteinuria was observed during the course of the study.
Local immunologic response was further studied on the cervical lymph nodes. Although total
cell numbers and lymph-node size were significantly increased following exposure to tremolite,
percentages of T- and B-cells did not significantly change. Because tremolite is part of the
makeup of LAA (6%), using tremolite-exposed mice might yield a similar response to
LAA-exposed mice. This same effect has been demonstrated following exposure to ultraviolet
radiation in skin cells, suggesting a similar mechanism (Saegusa et al., 2002).
Salazar et al. (2013; 2012) conducted a series of studies to establish the effects of LAA
exposure on autoimmune disease. The first set of studies utilized the collagen-induced arthritis
(CIA) and peptidoglycan-polysaccharide (PG-PS) models of rheumatoid arthritis to determine
whether LAA exposure increased the onset, or prolonged or intensified, the joint inflammation
characteristic of the disease (Salazar et al., 2012). Female Lewis rats were instilled biweekly for
13 weeks with a total dose of 0, 0.15, 0.5, 1.5, and 5.0 mg LAA followed by induction with
either model of arthritis. LAA at 5.0 mg reduced the magnitude of the swelling response in the
cell-mediated PG-PS model; however, neither the onset nor the duration of swelling was affected
by LAA exposure. LAA at 1.5 and 5.0 mg and amosite at 0.5 and 1.5 mg reduced total serum
IgM. LAA at 5.0 mg and amosite at 1.5 mg reduced anti-PG-PS IgG in the serum 17 weeks after
the final instillation. Finally, the number of rats positive for antinuclear antibodies (ANA) was
increased only at the low exposure concentrations of LAA in PG-PS-treated and nonarthritic rats.
These results suggest that LAA may have a modest inhibitory effect on the PG-PS rat model but
may enhance responses to other systemic autoimmune diseases (SAID).
In a follow-up study, Salazar etal. (2013) explored in greater detail the effect of LAA
exposure on ANA over time and the antigen specificity of the ANA. Female Lewis rats were
intratracheally instilled under the conditions in the previous study (described above). Serum
samples were analyzed every 4 weeks from the beginning of the instillations up to termination at
Week 28. Because elevated ANA are commonly associated with kidney disease, proteinuria was
assessed every 3 weeks beginning at Week 6 until termination of the experiment.
Histopathological analysis was also performed on the kidneys. ANA were increased 8 weeks
postexposure to LAA at 5.0 mg. By Week 28, all doses of LAA except 1.5 mg increased ANA
in the serum. Analysis of the antigen specificity found that only the LAA at 1.5 mg significantly
increased antibodies specific for extractable nuclear antigens (ENA) and the Jo-1 antigen.
Urinalysis found that all doses of LAA exposure induced moderate levels of proteinuria, but this
effect was not dose responsive. No dose-related histopathological effects were observed.
Altogether, these data suggest that LAA exposure increases autoimmune antibodies in the serum
but that no evidence of autoimmune disease is identifiable. However, the lack of SAID in the
D-ll
-------
Lewis rat may be due to strain-specific factors, suggesting that other animal models may be more
appropriate for studying the autoimmune effects of LAA.
D.2.2. Injection/Implantation
LVG:LAK hamsters were intrapleurally injected with tremolite obtained from the Libby,
MT, mine in an unpublished study by Smith (1978) prepared for W.R. Grace and Company.
These samples were identified as tremolite (22260p5; Sample 60) and 50% tremolite + 50%
vermiculite (22263p2, Sample 63). Both fiber samples were measured by optical phase
microscopy, and fibers were described as amorphous, irregularly shaped particles of about
5-15 um diameter, with Sample 60 (tremolite) also containing the occasional fiber up to 30 um
long. Fiber size for Sample 60 (tremolite) was also measured by scanning electron microscopy
(SEM) and determined to have a geometric mean length of 2.07 um, a geometric mean diameter
of 0.2 um, and an average aspect ratio of 10.36:1. Twenty-five milligrams of each of the two
samples were individually injected intrapleurally in LVG:LAK hamsters. Pathology was
examined at approximately 3 months postexposure in 10 animals from each group, with the
remaining animals observed until death, or 600 days postexposure, depending on the health of
the animal. Average survivorships were 410, 445, and 421 days in groups exposed to Sample 60,
Sample 63, and saline, respectively (see Table D-5). Pleural fibrosis was observed 3 months
postexposure, and mesothelioma was observed in both treatment groups between 350 and
600 days postexposure, with no mesotheliomas in control groups.
Table D-5. Pleural adhesions and tumors following intrapleural injection
exposure in LVGrLAK hamsters (25 mg)
Endpoint
Average adhesion rating3'13
Total tumors/animals0
Benign
Malignant
Mesothelioma
Control
0 (n = 10)
8/59
3/59
5/59
0/59
Sample 60 (tremolite)
3.3(w=10)
8/58
2/58
6/58
5/58
Sample 63 (tremolite and vermiculite)
3.6 (n = 10)
16/61
5/61
9/61
5/61
aAs analyzed in first group sacrificed (between 41 and 92 days postexposure).
bRating for pleura! adhesions: 0 = no adhesions; 1 = minimal adhesions; 4 = extensive adhesions.
These include adrenal adenoma, adrenal adenocarcinoma, lymphoma, pulmonary adenocarcinoma, adrenal and
salivary carcinoma, mesothelioma, rhabdomyosarcoma, hepatoma, thyroid carcinoma, subcutaneous carcinoma, and
malignant melanoma.
Source: Smith (1978).
D-12
-------
A subsequent study (Smith et al. (1979) was designed to determine whether
mesothelioma is a nonspecific result of mesothelial cells trapped in fibrous pleural adhesions,
occurring regardless of fiber type. Earlier studies by this group suggested that fibrosis and
tumors resulting from fiber exposure (chrysotile or glass) were related to fiber dimensions
(>20-um long, >0.75 urn in diameter) (Smith and Hubert. 1974). Injected fibrous talc (FD-14)
was used as a negative control in earlier studies and led to limited fibrosis and no tumor
formation. The characteristics of the FD-14 sample are described in the proceedings of Smith
and Hubert (1974). No further information could be found on the characteristics of the samples
used in this study.7 Because the talc contained 50% tremolite, 35% talc, 10% antigorite, and
5% chlorite, it was considered a tremolite sample by Smith (1978). When the sample was later
analyzed independently by Wylie et al. (1993), only 64 (12.8%) of 500 tremolite particles
measured met the National Institute for Occupational Safety and Health definition of a fiber
(>3:1 aspect ratio). Wylie et al. (1993) note, however, the sample consisted of very long fibers
of the mineral talc, with narrow widths and a fibrillar structure. A second tremolite sample
(Sample 275) used by Smith et al. (1979) was described as similar to FD-14, although no details
were given. The last two samples were prepared from a deposit of tremolitic talc from the
western United States (Sample 31) and from a specimen of asbestiform tremolite (Sample 72).8
Each of the four samples was examined microscopically, although the data were not
reported in the paper by Smith etal. (1979). The average fibers in Sample 72 were long, thin,
crystalline fibers (>20 um long, 0.4 um in diameter). Sample 31 appeared to have fewer long,
thin fibers than Sample 72, and many of the fibers in this sample were acicular. The
characteristics of the FD-14 sample were determined by phase microscopy (Smith and Hubert,
1974), but no characterization method was reported for the other three samples in this study.
Other samples used by this group have been analyzed by both optical and electron microscopy
(Smith, 1978; Smith and Hubert, 1974). The limited information on the fiber characteristics of
the samples used in these studies is provided in Table D-6. Note that no information was
provided confirming the presence or absence of particles or fibers less than 5 um in length in any
of the three papers by Smith and Hubert (1974), Smith etal. (1979), or Smith (1978). These data
deficiencies limit the interpretation of results from this study.
7This fiber is also analyzed in Wylieetal. (1993) and Stantonetal. (1981).
8 Although the source of this material is not reported, these studies parallel those in the unpublished study performed
by Smith etal. (19791 for W.R. Grace that used material from Libby, MT. Whether Sample 72 is material from
Libby, MT, or another location is unknown.
D-13
-------
Table D-6. Fiber characteristics and numbers of resulting tumors following
intrapleural injection of 10 or 25 mg fiber samples into LVGrLAK hamsters
Sample
FD-14
275
31
72
Average length"
Jim
5.7
N/D
>20
>20
Average
diameter3 (urn)
1.6
N/D
<0.4
<0.4
Tumors/survivors at 10 mgb
350 d
N/D
0/34
1/41
0/13
500 d
N/D
0/14
1/19
1/6
600 d
N/D
0/6
1/11
3/2
Tumors/survivors at 25 mgb
350 d
0/35
0/31
2/28
3/20
500 d
0/26
0/15
4/9
5/6
600 d
0/20
0/3
6/5
5/1
N/D = not described.
""Although average length and diameter are reported, what range of fibers was counted is unclear. Smith (1978)
(unpublished) states that only fibers greater than 5 um long are included. No other information is provided for these
samples.
bNumerator = cumulative number of animals with tumors; denominator = number of survivors.
Source: Smith et al. (1979): Smith (1978): Smith and Hubert (1974).
Following analysis of LVG:LAK intrapleurally injected with 10 or 25 mg of each of the
four samples of tremolite, Smith (1978) reported tumors at 350 days postexposure (25 mg) and
600 days postexposure (10 mg) for Samples 31 and 72 (see Table D-6). Although the number of
animals was not provided by Smith etal. (1979), previous studies by these authors reported using
50 animals per exposure group (Smith, 1978; Smith and Hubert, 1974). The results in Table D-6
present the cumulative number of tumors (numerator) at each time point analyzed over the
remaining survivors (denominator). The survival rate without tumor presentation was decreased
for animals exposed to Samples 72, 31, and 275. Smith et al. (1979) concluded that the FD-14
and 275 samples were noncarcinogenic, and Sample 31 was less carcinogenic than Sample 72.
Hamsters exposed to Sample 72 had extensive pleural fibrosis, which was observed to a lesser
degree in hamsters exposed to the other samples (Sample 72 > Sample 31 > Sample
275 = FD-14). No statistical information was reported for these results, and because the number
of background tumors in control animals was not provided, no statistical analysis can be
performed.
Both studies demonstrate that intrapleural injections of amphibole asbestos (tremolite or
LAA9) lead to an increase in pleural fibrosis and mesothelioma in hamsters compared to controls
or animals injected with less fibrous materials. The use of doses of equal mass for both studies
makes it difficult to compare potency among samples, as each sample could have vastly different
fiber number and total surface area. Although these studies clearly show the carcinogenic
potential of LAA fibers, intrapleural injections bypass the clearance and dissolution of fibers
from the lung after inhalation exposures.
'Assuming Smith etal. (1979) used LAA.
D-14
-------
Stanton etal. (1981) also examined tremolite and describe a series of studies on various
forms of asbestos. Fibers embedded in hardened gelatin were placed against the lung pleura. As
an intrapleural exposure, results might not be comparable to inhalation exposures because the
dynamics of fiber deposition and pulmonary clearance mechanisms are not accounted for in the
study design. Studies using two tremolite asbestos samples from the same lot were described as
being in the optimal size range (>8 um long and <0.25 um in diameter) for carcinogenesis; the
fibers were distinctly smaller in diameter than the tremolite fibers used by Smith et al. (1979).
Exposure to each of the two tremolite samples led to mesotheliomas in 21 and 22 of 28 rats
exposed. The Stanton etal. (1981) study also used talc, which did not lead to mesothelioma
production. This talc was found to be the same as that used by Smith et al. (1979) and later by
Wylie etal. (1993). Wylieet al. (1993) stated that, although the two trem elites were consistent
by size with commercial amphibole asbestos, the talc used contained fibers that were much
thinner and shorter, which is not typical of prismatic tremolite fibers.
Wagner etal. (1982) examined three types of tremolite (California talc, Greenland, and
Korea) using SPF Sprague-Dawley (n = 48) and Wistar (n = 32) rats, then followed up with a
range of in vitro tests using the same fiber samples. Rats were injected intrapleurally
(20 mg tremolite) at 8-10 weeks of age and allowed to live out their lives. Median survival
times after injections were 644 days (California talc), 549 days (Greenland tremolite), and
557 days (Korean tremolite). Positive controls had a decreased survival time due to an infection,
which limits the interpretation of these data. Also, this study was performed separately using
different rat strains for the three tremolite samples. The authors state that, although the
decreased control survival time and use of different rat strains limit the usefulness of the study
for quantitative analysis, the results can be described qualitatively. Of the three tremolites, only
the Korean tremolite showed carcinogenic activity producing mesothelioma (14/47 rats, 30%).
Analysis of the fiber characteristics showed the Korean sample had fibers that were longer than
8 um and a diameter of less than 1.5 um. The California talc and Greenland tremolite had little
to no fibers in this size range (see Table D-7). Follow-up in vitro assays in the sample
publication (Wagner et al., 1982) confirmed the in vivo results, with the exposure to Korean
tremolite resulting in increased LDH and p-glucuronidase (BGL) release, cytotoxicity, and
giant-cell stimulation.
D-15
-------
Table D-7. Fiber characteristics of three tremolite samples analyzed by in
vivo and in vitro methods (TEM measurements)
Sample
A
B
C
Location
California
Greenland
Korea
Fiber type
Flake-like
material
Medium-sized
fibrous material
Fine-fiber
material
Length
um
<6
<3
>8
Diameter
um
<0.8
<1.2
<1.5
No. of nonfibrous
particles (xlO4)
6.9
20.7
3.3
Total no. of
fibers
(xlO4)
5.1
4.8
15.5
No. of fibers >8 um
long (xlO3)
<1.5 fim diameter
1.7
0
56.1
TEM = transmission electron microscopy.
Source: Wagner etal. (1982).
Davis et al. (1991) examined six tremolites with differing morphologies through
intraperitoneal injections with male SPF Wistar rats. Four of the tremolites were from
Jamestown, California; Korea; Wales; and Italy; and two were from Scotland (Carr Brae and
Shinness). Of these, the three from California, Korea, and Wales were asbestiform, and the other
three were fiber bundles or prismatic (see Table D-8). Rats were exposed (n = 33 or 36) with
one intraperitoneal injection with samples that were 10 mg/2 mL-sterile PBS. Animals were
allowed to live out their full life spans or until signs of debility or tumor formation developed.
Although exposure was performed based on sample weight, each sample was analyzed to
determine the number of expected fibers per milligram and, therefore, per exposure. These
samples also were characterized further by counting fibers versus particles. Data were collected
for all fibers (aspect ratio >3:1) and particles (aspect ratio <3:1) of total fibers. A fiber was
defined as any component >8-um long and <0.25 um in diameter as measured by SEM (i.e.,
Stanton fibers).
D-16
-------
Table D-8. Fiber characteristics in a 10-mg dose (as numbers of fibers)
Sample
California
Wales
Korea
Italy
Carr Brae
Shinness
No. of
animals
36
36
33
36
33
36
No. of
mesotheliomas
36
35
32
24
4
2
No. of fibers in
1 mg of injected
dust (xlO5)
13,430
2,104
7,791
1,293
899
383
No. of fibers
>8 um long,
<0.25 um
diameter3 (xlO5)
121
8
48
1
0
0
No. of particles
in 1 mg injected
dust (xlO5)
18,375
4,292
13,435
20,137
9,490
5,901
Morphology
Asbestiform
Asbestiform
Asbestiform
Fiber bundles
Fiber bundles
Prismatic
"Stanton fibers.
Source: Davis etal. (1991).
The authors' overall conclusions were that all materials studied could cause
mesothelioma by this method of exposure, and the number of Stanton fibers was not sufficient to
explain the differences in response. Mesothelioma incidence was not correlated to Stanton
fibers, total particles, or mass of dust. The best predictor of mesothelioma incidence was total
fibers (see Table D-8). Although three samples were considered asbestiform (California, Wales
[Swansea], Korea), all samples had <1% of counted fibers defined as Stanton fibers. The highest
mesothelioma incidence was observed for the California sample, which contained the most
Stanton fibers (121 fibers per mg dust). The tremolite from Wales, resulted in
97% mesothelioma incidence yet contained only eight Stanton fibers per milligram (more than
90% less than in the California sample). In contrast, the Italy tremolite, although containing only
0.08% Stanton fibers, resulted in 67% mesothelioma incidence. Little is known, however, about
the characteristics of particles or fibers <5 um long. This study highlights two issues associated
with all fiber studies: the limits of analytical techniques and the variability in response based on
the metric used to measure exposure. This study also supports the premise that asbestos samples
containing fibers that are not long and thin can be carcinogenic.
The Roller etal. (1996) study was designed to provide data on the dose-response of
various fiber types in relation to their fiber dimensions (as measured by SEM). Fibers were
defined in this study as having an aspect ratio of greater than 5:1 for all lengths and widths.
Female Wistar rats (n = 40) were given either one intraperitoneal injection of 3.3 mg or 15 mg of
tremolite. Rats were examined for tumors in the abdominal cavity following a lifetime (up to
30 months) of observation. This paper described the fiber dimensions in depth (see Table D-9),
while limited discussion focused on the exposure results. This table shows the characteristics of
the fibers sorted first by aspect ratio and diameter, and the fiber size distribution binned by the
D-17
-------
length and diameter for those fibers with a length >5 jim. Results were described in this study in
a table as "positive rats" being those with histologically confirmed mesothelioma or
macroscopically supposed mesothelioma. No information was provided on how these
determinations were made. Exposure to 3.3 mg and 15 mg tremolite resulted in 9 mesotheliomas
in 29 animals (64 weeks postexposure) and 30 mesotheliomas in 37 animals (42 weeks
postexposure), respectively. This study demonstrates that intraperitoneal injection of tremolite
led to mesothelioma in Wistar rats. Analysis of other tissues was not described.
Table D-9. Characteristics of tremolite fibers intraperitoneally injected into
Wistar rats
Fiber number per
ng dust and mass fraction (%)
Aspect ratio (L/D) >5/l; D <2 urn
[(Roller et al., 1996) studvl
Length:
>5
No.
17.4
um
%
Mass
32
>10 um
No.
6.9
%
Mass
27
>20 um
No.
1.9
%
Mass
18
Aspect ratio (L/D) <3/l; D <3 urn
[WHO. 1985 as reoorted in (Roller et al.. 1996)1
Diameter:
>5
No.
18.4
um
%
Mass
43
>10
No.
7.0
um
%
Mass
35
>20 um
No.
2.0
% Mass
26
Fiber-size distribution for aspect ratio (L/D) >3/l (all lengths, all diameters; SEM)
% Total
fibers
L >5 um
10% <
22%
0.8
Length (um)
50% <
2
.4
90% <
9.2
99% <
29.4
Diameter (um)
10% <
0.14
50% <
0.27
90% <
0.67
99% <
1.49
SEM = scanning transmission microscopy.
Source: Roller etal. (1996).
D.2.3. Oral
McConnell et al. (1983) describe part of a National Toxicology Program study (NTP,
1990a, b, 1988, 1985) that was conducted to evaluate the toxicity and carcinogenicity of
ingestion of several minerals. This study examined chrysotile and amosite in both hamsters and
rats, and crocidolite and tremolite only in rats. This chronic bioassay was designed to encompass
the lifetime of the animal, including exposure of the dams from which the test animals were
derived. Although the study examined chrysotile, amosite, crocidolite, and tremolite, for the
purposes of this assessment, the focus is on the results from exposure to tremolite. The tremolite
(Gouverneur Talc Co., Gouverneur, NY) used was not fibrous. Instead, the material was
crystalline, as this form was a common contaminant in talc at the time of these studies
(McConnell et al.. 1983) (see Table D-10). Citing the Stantonetal. (1981) paper, McConnell et
al. (1983) stated that crystalline tremolite can become fibrous upon grinding. Tremolite was
D-18
-------
incorporated by 1% weight into NIH-31 feed and given to 250 male and female F344 rats from
birth until death (118 male and female controls).
Table D-10. Fiber characteristics and distribution of tremolite fibers analyzed
in feed studies in F344 rats
Characteristic
Mean width
Tremolite particles
% of Tremolite particles
Length interval"
<3 um
0.77
120
19.4
>3 um, <5 fim
1.78
61
9.85
>5 um, <10 fim
2.87
17
o
J
>10um
5.22
49
8
"Average groups, more detailed in primary paper.
Source: McConnelletal. (1983).
No significant tumor induction was observed in the animals with oral exposure to
tremolite. Although nonneoplastic lesions were observed in many of the aging rats, these were
mostly in the stomach and occurred in both controls and exposed animals. The lesions included
chronic inflammation, ulceration, and necrosis of the stomach (McConnell etal., 1983).
McConnell et al. (1983) suggested that nonfibrous tremolite could account for the lack of
toxicity following exposure in this group of animals. Also, oral studies of asbestos generally
show decreased toxicity and carcinogenicity as compared to inhalation and
implantation/injection studies.
D.3. MECHANISTIC DATA AND OTHER STUDIES IN SUPPORT OF THE MODE OF
ACTION
D.3.1. In Vitro Studies—LAA
Hamilton et al. (2004) examined the potential for fibers, including LAA, to modify the
function of antigen-presenting cells (APC). Analysis was performed at 24 hours with two forms
of asbestos (crocidolite [25 or 50 jig/mL] and LAA obtained from Site No. 30, Libby, MT [25 or
50 |ig/mL]) and ultrafine particulate matter (PIVh.s [particulate matter 2.5 microns in diameter or
less] [50 or 100 |ig/mL]). Limited information is provided by Hamilton et al. (2004) on fiber
characteristics. Samples from Site No. 30, however, are described as predominantly richterite
and winchite by Meeker et al. (2003). Primary human alveolar macrophages were incubated for
24 hours with LAA (25 or 50 |ig/mL), crocidolite (25 or 50 |ig/mL), or ultrafine particulate
matter (50 or 100 |ig/mL). Following incubation, cells were isolated from remaining particles
and nonviable cells, after which 0.25 x 106 macrophages were cocultured with autologous
lymphocytes (1 x 106 cells) in an 11-day APC assay. This assay analyzes the antigen-presenting
D-19
-------
function of the pretreated macrophages by stimulating the lymphocytes using tetanus toxoid as
the antigen. The supernatant was assayed for cytokines on Day 11, and Hamilton et al. (2004)
found that pretreatment with either asbestos or PIVh.s significantly upregulated both Thl and Th2
cytokines (interferon gamma [LFNy]; interleukin-4 [IL-4]; and interleukin-13 [IL-13]) (p <0.05).
Therefore, preexposure to either fibers or particles increased APC function, as reflected in
increased cytokine release after tetanus challenge. No significant differences, however, were
discernable between asbestos and PIVb.s pretreatment. The authors speculated that the variability
in response among samples assayed—presumably due to the use of primary cells—obscures
statistical significance. This study supports a role for fibers and PIVb.s in potentiating immune
response, although the specific role may be unclear as many agents can activate macrophages
prior to antigen challenge.
Recent studies (Blake et al., 2008; Blake et al., 2007) compared the response of murine
macrophages (primary and cell line RAW264.7) to LAA fibers and crocidolite asbestos fibers.
The LAA fibers (7.21 ± 7.01 um long, 0.61 ± 1.22 um in diameter) used in these studies were
obtained from the U.S. Geological Survey and were chemically representative of the Libby, MT,
mine (Meeker et al.. 2003). The crocidolite fibers (4.59 ± 4.22 um long, 0.16 ± 0.09 um in
diameter) used in these studies were provided by Research Triangle Institute, NC, and the
noncytotoxic control fiber (wollastonite, 4.46 ± 5.53 um long, 0.75 ± 1.02 um in diameter) was
provided by NYCO Minerals, NY. Cells were exposed for 24 hours to fiber samples measured
by relative mass (5 ug/cm2), after which the cells were analyzed by transmission electron
microscopy (TEM) to measure internalization. The results of the first study (Blake et al., 2007)
indicate that LAA fibers can both attach to the plasma membrane and be internalized by
macrophages, similar to the crocidolite fibers. These internalized fibers were primarily less than
2 um long and were found localized in the cytoplasm, in cytoplasmic vacuoles, and near the
nucleus following 3-hour exposure at a concentration of 62.5 ug/cm2. This same concentration
was selected for the remaining studies because it did not decrease cell viability for the LAA
(92%). Cell viability was decreased for crocidolite (62%), however, at this concentration. As a
result, the remaining assays would be expected to have decreased viability following exposure to
crocidolite, which may impact the levels of various responses. For example, the ROS
measurement would increase with increased cell number; therefore, some of the quantitative
results would be difficult to compare among fiber types unless normalized to cell number.
Oxidative stress was measured by the induction of ROS and the reduction in GSH levels.
These two measurements generally complement each other, as GSH is used to maintain
intracellular redox balance in cells in response to increased ROS levels. Both LAA and
crocidolite fiber internalization generated a significant increase (p < 0.05) in intracellular ROS as
quantified by the oxidation of 2,7-dichlorodihydrofluorescein to dichlorofluorescein with hourly
readings on a fluorescent plate reader. LAA exposure significantly increased ROS in a
dose-dependent manner (6.25, 32.5, and 62.5 ug/cm2) as early as 1 hour postexposure at the
D-20
-------
highest dose (p < 0.05) as compared to a no-treatment group. Only the highest concentration of
crocidolite was tested. The lower concentrations of LAA were not compared to crocidolite and
wollastonite, but a comparison of the highest exposure concentrations (62.5 jig/cm2) of LAA,
crocidolite, and wollastonite revealed greater ROS production following LAA exposure (1 hour,
p < 0.05). Blake et al. (2007) stated that similar results were seen in the primary cell line but did
not report the data. To differentiate the type of ROS produced, dehydroergosterol fluorescence
intensity levels were used, revealing that superoxide anion was significantly increased following
exposure to LAA as compared to controls. This observation was further confirmed with the use
of a free radical scavenger (PEG-SOD [polyethylene glycol-superoxide dismutase]) specific to
superoxide anion. This coexposure of LAA and PEG-SOD led to a significant decrease in ROS
as compared to cells exposed only to LAA (p < 0.05). Total intracellular superoxide dismutase
(SOD) activity was also measured following exposure to LAA and showed a decrease in activity
at 3 hours postexposure as compared to controls (p < 0.05). Crocidolite appears to increase
intracellular SOD activity at 24 hours postexposure. These three assays demonstrate that LAA
exposure leads to increased superoxide anion in macrophages, most likely by suppressing
activity of intracellular SOD.
GSH levels were found to be decreased in response to LAA and crocidolite exposure in
the macrophage cell line as compared to unexposed cells (p < 0.05). The decreased GSH levels
were more prominent following crocidolite exposure as compared to LAA. Crocidolite exposure
has been shown in other studies to lead to increased hydrogen peroxide but not superoxide anion
(Kamp and Weitzman, 1999; Kamp et al., 1992). The increased hydrogen peroxide from
crocidolite exposure can then lead to increased hydroxyl radical production (through interactions
with endogenous iron), and potentially, deoxyribonucleic acid (DNA) adduct formation. DNA
adduct formation (8-hydroxy-2'deoxyguanosine [8-OHdG]), 8-oxoguanine-DNA-glycosylase 1
(Oggl) levels, and DNA damage (comet assay) also were measured. A significant increase in
DNA damage in exposed macrophages, as measured by increases in both 8-OHdG formation and
expression of Oggl, a DNA repair enzyme that excises 8-OHdG from DNA following oxidative
stress, was observed following exposure to crocidolite but not LAA. Increased superoxide anion
following LAA exposure does not appear to yield oxidative damage similar to crocidolite. These
results suggest a chemical-specific response to each type of amphibole that yields varied cellular
responses. Therefore, the mechanism of action following response to LAA might be different
than that of crocidolite, also an amphibole fiber.
To determine if the ROS production was related to fiber number for both LAA and
crocidolite, cell-fiber interactions and fiber internalization were measured following exposure to
equal concentrations of crocidolite, LAA, and wollastonite (62.5 ug/cm2, 3 hours). With phase
contrast light microscopy, the number of cells interacting with one or more fibers was counted
(100 cells counted for each treatment). All murine macrophages bound or internalized at least
one fiber from the LAA sample (mean ± SD, 4.38 ± 1.06 internalized) or the crocidolite sample
D-21
-------
(3.28 ± 1.58 internalized) but not the wollastonite sample (Blake et al., 2007). No significant
differences were observed in the responses to LAA or crocidolite samples, suggesting that the
differences in measured ROS were not related to cell number. Fiber sizes varied between the
two samples, with the crocidolite sample containing a more homogeneous mixture of long fibers
(exact size not given), while the LAA sample contained a mixture of sizes and widths. These
characteristics were not analyzed to determine what, if any, role they might play in the varied
response.
The second study by Blake et al. (2008) reports the effects of in vitro exposure to LAA
on apoptosis by exploring autoimmune response following asbestos exposure. Although LAA
was not directly used in the autoimmune studies, the autoantibody (SSA/Ro52) is a known
marker of apoptosis, and the in vitro studies included treatment with LAA. RAW264.7 cells
exposed to LAA induced apoptosis over 72 hours, as measured by induction of poly
(ADP-ribose) polymerase cleavage and increased Annexin V staining. Redistribution of
SSA/Ro52 in apoptotic blebs was demonstrated in LAA-exposed RAW264.7 cells but not in the
unexposed controls and wollastonite-exposed RAW264.7 murine macrophages, further
confirming apoptosis following LAA exposure.
Rasmussen and Pfau (2012) studied the role of Bla B-lymphocytes in the development of
autoantibody production following asbestos exposure. CH12.LX B-lymphocytes, a murine Bl
lymphocyte cell line, were cultured with 35 ug/cm2 of LAA or 1 ug/mL of lipopolysaccharide
(LPS; positive control) for 48 hours. Asbestos exposure did not affect proliferation or antibody
production. CH12.LX B-lymphocytes cultured for 24 hours in RAW medium treated with
35 ug/cm2 LAA reduced CH12.LX proliferation and increased IgGl, IgG3, and IgA production
when normalized to cell number. The authors identified that IL-6 and TNF-a were both elevated
in the medium of asbestos-treated RAW macrophages. Treating CH12.LX B-lymphocytes with
recombinant IL-6 or TNF-a at similar concentrations as in the asbestos-treated macrophage
medium resulted in reduced CH12.LX proliferation. Interestingly, only the IL-6-treated
CH12.LX cells had increased IgG and IgA production. However, both high and low
concentrations of IL-6 increased IgG and IgA secretion, indicating that some other mechanism is
present in the asbestos-treated RAW medium that regulates CH12.LX antibody production.
These data suggest a potential mechanism for asbestos-induced autoantibody production in
LAA-exposed residents.
Li et al. (2012) exposed THP-1 cells (macrophage cell line) to LAA10 and chrysotile (0,
20, 40 ug/mL for 24 hours) to measure inflammatory response. This study measured cell death,
caspase activation and release of IL-lp to determine if each fiber type activated the Nod-like
receptor protein 3 (NLRP3) inflammasome. Results demonstrated that while both fiber types
appeared to activate NLRP3, chrysotile led to a greater effect as measured by cell death,
10 LAA, or Libby "Six-Mix" was used for this study. Fiber characteristics are described from previous studies.
LAA mean fiber length = 7.21um; mean surface area = 5 m2/g.
D-22
-------
activation of caspase-1, and release of IL-lp. However, results demonstrated that both fibers
also led to increased ROS production compared to the same mass dose of chrysotile as measured
by increases in expression of antioxidant enzymes, protein oxidation, and nitration and lipid
peroxides. In order to further study these differences in biological response to these two fibers,
BEAS-2B cells (bronchial epithelium cells) were exposed to supernatant from the THP-1 cells.
Both activated the MAPK cascade, increased ERK and MAP3K8 phosphorylation, and increased
AP-1 binding and IL-6 release. These results were attenuated with the addition of an IL-lp
antagonist (IL-1 Ra). This study demonstrated that although exposure to both fibers led to the
same biological responses, the level of response was variable. Although not studied, the authors
suggest that differences in fiber length and surface area may play a role in this differential
inflammatory response.
Serve et al. (2013) examined a possible role of autoimmunity in fibrosis by an in vitro
examination of potential mechanisms of mesothelial cell autoantibodies (MCAA) leading to
collagen deposition, a precursor to fibrosis. Nonmalignant, transformed human mesothelial cells
(MeT-5A) were exposed to serum samples from LAA-exposed populations. These samples were
identified as MCAA-positive or MCAA-negative and were pooled prior to exposure. MCAA
was found to be present and induced collagen deposition but not mesothelial cell differentiation.
The increase in collagen deposition observed was not through increased collagen synthesis but
SPARC-related collagen processing and associated with specific matrix metalloproteinases
(MMPs). This study demonstrated that MCAA binding leads to increased collagen deposition by
altering MMP expression.
Duncan et al. (2014) examined the in vitro determinants of asbestos fiber toxicity,
comparing two samples each of LAA (LA2000, LA2007) and amosite asbestos (UICC, RTI).
Primary human airway epithelial cells (HAEC) were exposed for 24 hours to 2.64, 13.2, or
26.4 ug/cm2 LAA and amosite asbestos, with each asbestos sample having been analyzed for
fiber size distribution, surface area, and surface-conjugated iron (see Table D-l 1). The asbestos
samples had similar characteristics, except RTI amosite, which consisted of longer fibers. Fiber
toxicity was measured by cytotoxicity (LDH assay), levels of ROS production, as well as IL-8
mRNA levels as a measure of relative proinflammatory responses. Cytotoxicity levels were
similar among all four samples at the highest dose, but statistically significant compared to
no-treatment control. Results on an equal-mass basis demonstrated a statistically significant
increase in IL-8, IL-6, cyclooxygenase-2 (COX-2), and TNF mRNA levels for all four
amphiboles at the two highest doses. The greatest increase in IL-8 mRNA levels followed
exposure to the RTI amosite sample, while response levels observed among the UICC amosite
and both LAA samples were not statistically significant. Therefore, IL-8 was used to further
analyze dose metric for this response. Surface iron concentrations and surface reactivity were
quantified with respect to hydroxyl radical production to assess the effect of these properties on
IL-8 mRNA expression. Surface iron concentrations were similar for the two LAA samples and
D-23
-------
for the two amosite samples, but the amosite samples had significantly greater surface iron as
compared to the LAA samples. UICC amosite had slightly greater iron compared with RTI
amosite. A strong correlation was observed between fiber dose metrics of length and external
surface area. When these metrics were used in place of equal-mass dose, the differential IL-8
mRNA expression following exposure to these four samples was eliminated.
Table D-ll. Characterization of amphibole samples (Duncan etal., 2014)
LA2000
LA2007
RTI amosite
UICC amosite
Particle count
n (total particles)
n (total EMP)a
561
450
510
250
588
292
525
178
Particle number/mg
Total particles x 107/mg
BMP x 107/mg
98.2
78.7
103
50.5
9.15
4.5
94.2
31.9
Particle size distribution
Total particle mean length (um)
Total particle mean width (um)
Total particle mean aspect ratio
EMP mean length (urn)
EMP mean width (urn)
EMP mean aspect ratio
3.7 ±0.2
0.36 ±0.02
12.8 ±0.6
4.4 ±0.2
0.30 ±0.01
15.5 ±0.6
2.3 ±0.2
0.36 ±0.01
8.4 ±0.7
3. 8 ±0.3
0.29 ±0.02
15.1 ± 1.2
6.4 ±0.6
0.44 ±0.01
16.9 ± 1.6
12.1 ± 1.2
0.37 ±0.01
32.4 ±3.0
2.1 ±0.3
0.43 ±0.01
5.6 ±0.6
4.3 ±0.5
0.27 ±0.01
13. 0± 1.0
Surface area
Total surface area by GA (m2/g)b
EMP surface area by TEM (m2/g)c
5.3
1.1
7.4
2.6
3.1
2.8
4.8
1.5
"Elongated mineral particle (EMP) defined as having an aspect ratio >3 : 1 .
bMeasured by Kr gas adsorption (GA) and BET analysis.
"Measured by TEM and calculated by using the equation SA = (L>
-------
negative controls. Fiber-size distribution for crocidolite and LAA is shown in Table D-12.
Micronuclei induction was measured following treatment of cells by controls as described above,
and by 5 jig/cm2 fibers or titanium dioxide (TiCh) particles for 24 hours. Following treatment,
cells were fixed, permeabilized, and blocked before being exposed to anticentromere antibodies,
and micronuclei were counted and scored as centromere negative arising from DNA breaks
(clastogenic) or centromere positive arising from chromosomal loss (aneugenic). Spontaneous
micronuclei induction was increased in XRCC1-deficient cells as compared to controls.
Wollastonite and titanium dioxide did not induce micronuclei in either cell type. Crocidolite and
LAA-induced dose-dependent increases in micronuclei formation in both cell types, including an
increase in the proportion of micronuclei in XRCC1-deficient cells (see Table D-13). LAA
exposure led to a decreased amount of micronuclei as compared to crocidolite. Specifically in
relation to clastogenic versus aneugenic micronuclei, crocidolite exposure led to mainly
clastogenic micronuclei, while LAA exposure led to a mixture of aneugenic and clastogenic
micronuclei. Nuclear bud formation was also observed but only with exposure to crocidolite and
bleomycin. Western blot analysis was performed to analyze protein expression related to DNA
damage repair (XRCC1) and cell cycle progression (p53, p21) (data not shown in publication).
The differences observed between crocidolite and LAA are most likely related to their
physicochemical differences. However, these results support a genotoxic effect of exposure to
both crocidolite and LAA.
Table D-12. Size distribution of UICC crocidolite and LAA used in Pietruska
et al. (2010V
Length (jim)
0.1-1.0
1.1-5.0
5.1-8.0
8.1-10.0
10.1-20.0
>20.1
% fibers in size range
Crocidolite
46.4
44.8
3.8
0.9
2.4
1.7
LAA
12.6
38.5
23.1
10.4
11.6
3.6
""Distribution by diameter also given in original manuscript.
Source: Adapted from Supplemental Material of Pietruska et al. (2010).
D-25
-------
Table D-13. Percent clastogenic micronuclei following exposure to LAA or
crocidolite
LAA (5 ug/cm2)
Crocidolite (5 ug/cm2)
H460 cells
7 1.5 ±3. 4%
57.2 ± 2.2%
XRCCl-deficient
86.0±1.2%a
65.1±2.2%a
ap <0.05 as compared to control cells.
Source: Pietruska et al. (2010).
Mechanisms of oxidative stress following exposure to LAA were also studied in human
mesothelial cells (Hillegass et al., 2010). Gene-expression changes were measured with
Affymetrix U133A microarrays (analysis with GeneSifter) following exposure to
15 x 106 |im2/cm2 LAA11 as compared to the nonpathogenic control (75 x 106 |im2/cm2 glass
beads) in the human mesothelial cell line LP9/TERT-1 for 8 and 24 hours. Gene expression of
only one gene (manganese superoxide dismutase [MnSOD; SOD2]) was altered following
exposure to LAA for 8 hours, while 111 genes had an altered gene expression following
exposure to LAA for 24 hours (altered by at least twofold as compared to controls).
The gene forMnSOD; SOD2 was observed to be significantly upregulated at both time
points (p <0.05) as compared to the nonpathogenic control. This gene was confirmed in normal
human pleural mesothelial cells (HKNM-2) by quantitative RT-PCR at 24 hours following
exposure to the nontoxic dose of LAA. Upregulation of three genes from this and previous
studies by these authors was confirmed by quantitative RT-PCR (SOD2, ATF, and IL-8) in
HKNM-2 cells exposed to both LAA and crocidolite asbestos. Gene ontology of these results
demonstrated alterations related to signal transduction, immune response, apoptosis, cellular
proliferation, extracellular matrix, cell adhesion and motility, and ROS processing. Follow-up
studies at both the nontoxic dose (15 x 106 |im2/cm2) and the toxic dose (75 x 106 |im2/cm2)
exposure levels in LP9/TERT-1 cells examined SOD protein and activity, ROS production, and
GSH levels. At 24 hours, SOD2 protein levels were increased following exposure to the toxic
dose of LAA (p <0.05) but not at 8 hours. Cells exposed to all doses of LAA and crocidolite
asbestos had increased copper-zinc superoxide dismutase (Cu/ZnSOD; SOD1) protein at
24 hours (p <0.05) but not at 8 hours. Although total SOD activity remained unchanged, a
dose-related SOD2 activity was observed following exposure to both doses of LAA for 24 hours,
but this appeared to be minimal and was not statistically significant (activities at 8 hours were
not examined). Oxidative stress was measured by dichlorodihydrofluorescein diacetate
11LAA samples for this study were characterized by analysis of chemical composition and mean surface area
(Meeker et al.. 2003X Doses were measured in surface area and described based on viability assays with fiber
samples as either nontoxic (15 x 106um2/cm2) or toxic (75 x 106um2/cm2).
D-26
-------
fluorescence staining detected by flow cytometry and was observed to be both dose- and
time-dependent in cells exposed to LAA and was increased following exposure to the toxic dose
of LAA (statistical analysis not possible). Oxidative stress was further supported by analysis of
gene expression of heme oxygenase 1 (HO-1) following exposure to LAA in both LP9/TERT-1
and HKNM-2 cells for 8 and 24 hours. HO-1 was significantly increased following exposure to
the toxic dose of LAA in both cell lines (p-value not given). GSH levels were transiently
depleted following 2-8 hours exposure to 75 x 106 |im2/cm2 levels of LAA, with a gradual
recovery up to 48 hours in LP9/TERT-1 cells (HKNM-2 not analyzed). Exposure to crocidolite
asbestos at the toxic dose led to a significant GSH decrease at all-time points up to 24 hours
(p <0.05). These studies demonstrate that LAA exposure leads to increases in oxidative stress as
measured by ROS production, gene expression, protein and functional changes in oxidative
stress proteins (SOD), and GSH-level alterations in human mesothelial cells.
Pfau et al. (2012) conducted a study to determine the effect of LAA exposure on the
amino acid transport system x~c which is one of the pathways murine macrophages detect and
respond during stressful conditions. RAW 264.7 murine macrophages were cultured in the
presence of LAA for 24 hours and then compared to the control substances silica, LPS, and
wollastonite. System x~c was increased in LAA-treated cells but not in silica or wollastonite
controls. ROS production increased system x~c activity. Furthermore, inhibition of system x~c
increased ROS production and reduced viability in LAA-treated cells but not silica-treated cells.
Altogether, these data suggest that system x~c may play a role in macrophage survival and
inflammation following LAA exposure.
The relative toxicity of LAA was measured by gene-expression changes of IL-8, COX-2,
and heme oxygenase (HO)-1, as well as other stress-responsive genes, as compared to amosite
(Research Triangle Institute, NC) in primary HAEC in vitro. Comparisons were made with both
fractionated (aerodynamic diameter <2.5 jim) and unfractionated fiber samples (Duncan et al.,
2010). Crocidolite fibers (UICC) were also included in some portions of this study for
comparison. Fractionation was performed using the water elutriation method (Webber et al.,
2008) and characterized as described in Lowers and Bern (2009). Primary HAECs were exposed
to 0, 2.64, 13.2, and 26.4 jig/cm2 of crocidolite, amosite, AM2.5 (fractionated), LAA, or LA2.5
(fractionated) for 2 and 24 hours in cell culture. Confocal microscopy was used to determine
fiber content in cells exposed for 4 and 24 hours to 26.4 jig/cm2 AM2.5 or LA2.5 only. At
4 hours postexposure, fibers were mainly localized on the periphery of the cell with some fibers
internalized. By 24 hours postexposure, most fibers appeared to be internalized and localized by
the nucleus. Cytotoxicity was determined by measurement of LDH from the maximum dose
(26.4 jig/cm2) of both, fractionated and unfractionated, amosite and LAA samples, with less than
10% LDH present following exposure to all four samples. Cytotoxicity was also determined for
just the fractionated samples of amosite and LAA by measuring intracellular calcein fluorescence
emitted by live cells and showed 95% and 99% viability for AM2.5 and LA2.5, respectively.
D-27
-------
These results support a limited cytotoxicity of both amosite and LAA under these concentrations
and time frames.
Gene-expression changes in specific inflammatory markers (IL-8, COX-2, HO-1) were
analyzed by quantitative RT-PCR for amosite, AM2.5, LAA, LA2.5, and CRO at both 2 and
24 hours postexposure (all doses). Minimal increases in gene expression of IL-8, COX-2, or
HO-1 were observed at 2 hours postexposure to all five fiber types; at 24 hours postexposure,
however, a dose response was observed following exposure to all fiber types. The smaller size
fractions resulted in differences in magnitude of gene-expression changes between AM2.5 and
LA2.5, with AM2.5 leading to greater induction of IL-8 and COX-2 as compared to LA2.5.
HO-1 levels were comparable between the two samples (see Table D-14). Gene expression of
transforming growth factor (TGF)-Bl was also quantified but only following exposure to AM2.5
and LA2.5 (all doses; data not shown in publication). Levels of IL-8 protein were also measured
following 24 hours exposure to AM2.5 and LA2.5 (all doses) and were statistically significant at
the two highest exposures (13.2 and 26.4 jig/cm2). Gene-expression changes were also
examined for 84 genes involved in cellular stress and toxicity using a 96-well RT-PCR array
format following 24 hours exposure to 13.2 jig/cm2 amosite, LAA, AM2.5, or LA2.5, or to
26.4 jig/cm2 LA2.5 only. The results show a proinflammatory gene-expression response.
Gene-expression profiles were similar between amosite and LAA, but differences were observed
between AM2.5 and LA2.5.
Table D-14. Gene-expression changes following exposure to 26.4 ug/cm2
amphibole asbestos for 24 hours3
Genes for specific
inflammatory markers
IL-8
COX-2
HO-1
Amosite (AM)
50 ±7.5
5.4 ±0.5
2.9 ±0.2
Amosite,
fractionated
(AM2.5)
120 ± 25
16 ±2.8
4.5 ±0.3
LAA
46 ± 8.3
9.0 ±1.7
2.5 ±0.2
LAA, fractionated
(LA2.5)
37 ±7.8
1.6 ±0.3
5.1 ±0.6
aAll results in fold change ± standard deviation as compared to untreated control cells.
Source: Duncan etal. (2010).
To determine if surface iron on the fibers played a role in the inflammatory response,
Duncan etal. (2010) also examined surface iron concentrations by two methodologies:
inductively coupled plasma optical emission spectroscopy and citrate-bicarbonate-dithionite.
Both assays determined AM2.5 appeared to have surface iron as measured by thiobarbituric
acid-reactive product formation following exposure to amosite, AM2.5, LAA, and LA2.5. Both
amosite samples were found to generate the greatest amount of hydroxyl radicals compared to
D-28
-------
the two LAA samples, with the fractionated AM2.5 and LA2.5 exhibiting small increases in
ROS produced compared to the unfractionated samples.
D.3.2. In Vitro Studies—Tremolite
In general, all fibrous tremolite samples were shown to be carcinogenic, with those
containing more of the longer, thinner fibers (>10 um length, <1 um diameter) being more potent
carcinogens. Most studies described here used weight as the measurement of fibers for exposure,
with the doses ranging from 0 to 40 mg/animal. One set of studies did expose animals with
fibers measured by number (100 fibers/cm3) (Bernstein et al., 2006; Bernstein et al., 2005b).
D.3.2.1. Cytotoxicity
Wagner et al. (1982) examined the in vitro cytotoxicity of three forms of tremolite (see
Table D-7) used in their in vivo studies. LDH and BGL were measured in the medium following
incubation of unactivated primary murine macrophages to 50, 100, and 150 ug/mL of each
sample for 18 hours. Cytotoxicity of Chinese hamster lung fibroblasts V79-4 was measured by
methylene blue staining (fiber concentrations not given). Giant-cell formation in A549 human
basal alveolar epithelial cell cultures was measured, using 100 and 200 ug/mL of each sample for
5 days. Crocidolite fibers were used as the positive control.
In all three assay systems, the Korean tremolite produced results similar to the positive
control: increased toxicity of primary murine macrophages, increased cytoxicity of Chinese
hamster ovary (CHO) cells, and increased formation of giant cells from the A549 cell line. The
tremolite sample from Greenland (Sample B) did result in increased toxicity over controls,
although to a lesser degree (statistics are not given). The authors speculated that the iron content
in Sample B might have contributed to these results. Although differential toxicity of these
samples was noted on a mass basis, data were not normalized for fiber content or size. The
inference is that differential results are due, at least in part, to differential fiber counts.
In a study to further elucidate the role of ROS following exposure to asbestos, Suzuki and
Hei (1996) examined the role of heme oxygenase (HO) in response to asbestos. HO is induced
in response to oxidative stress and functions to degrade heme; it might, therefore, prevent
iron-mediated hydroxyl radical production. All fibers tested led to an increase in HO, although
chrysotile (UICC) and crocidolite (UICC) led to a greater increase than tremolite (Metsovo,
Greece) and erionite (Rome, Oregon). No statistics, however, are described for these results.
This study focused on responses to 20 and 40 ug/mL of chrysotile and then used doses that
yielded 0.5 and 0.3 relative survival fractions for all other fibers (crocidolite, 20 and 40 ug/mL;
tremolite, 150 and 300 ug/mL; erionite, 200 and 400 ug/mL). Fibers were not characterized in
this paper. When normalized by survival fraction, the inductions of HO above the control were
3.89-, 3.86-, 2.75-, and 2.78-fold above background levels for chrysotile, crocidolite, tremolite,
and erionite, respectively. Limited information is provided on the results of tremolite exposures
D-29
-------
beyond an increase in HO following an 8-hour exposure. This increased HO following exposure
to tremolite demonstrates a response similar to that observed for crocidolite and chrysotile in this
study. Crocidolite is further analyzed with exposures to the antioxidants superoxide dismutase
and catalase, leading to a dose-dependent decrease in HO induction, which supports the role of
HO in oxidative stress.
Wylie etal. (1997) examined the mineralogical features associated with cytotoxic and
proliferative effects of asbestos in hamster tracheal epithelial (HTE) cells and rat pleural
mesothelial (RPM) cells with a colony-forming efficiency assay. HTE cells are used because
they give rise to tracheobronchial carcinoma, while RPM cells give rise to mesothelioma. Cells
were exposed to fibers by weight, number, and surface area (see Table D-15).
Table D-15. Fiber characteristics of five fibers examined in vitro for
cytotoxic (HTE cells) and proliferative effects (RPM cells)
Sample
FD14
SI57
CPS183
NIEHS crocidolite
NIEHS chrysotile
Description (% of sample)
Talc (37), tremolite (35), serpentine
(15), other (<2), unknown (12)
Talc (60), tremolite (12), unknown
(21), other (4), anthophyllite (3),
quartz (1)
Talc (50), quartz (12), unknown
(28), tremolite (4), other (4),
anthophyllite (3)
Riebeckite (100)
Chrysotile (100)
Surface area (mm2/g)
6.2 ±0.2
4.9 ±0.2
4.9 ±0.4
10.3 ±1.3
25.4 ±0.5
Fibers/ng
2.5 x 103
1.1 x 104
1.1 x 104
5.3 x 105
5.3 x 104
Fibers >5 urn/jig
0.8 x 103
4.8 x 103
9.2 x 103
3.8 x 105
3.4 x 104
NIEHS = National Institute of Environmental Health Sciences.
Source: Wylie etal. (1997).
Colony-forming efficiency assay results are expressed as the number of colonies in
exposed cultures divided by the control colonies multiplied by 100. Increases in colony numbers
indicate increased cell proliferation or survival in response to the exposure. Decreases in colony
numbers indicate toxicity or growth inhibition in response to the exposure. The results of the
analysis with fiber exposure by mass (ug/cm2) show elevated colonies in HTE cells following
exposures to both asbestos fibers (p <0.05) at the lowest concentrations, while significant
decreases are observed for both asbestos fibers at the higher concentrations (0.5 ug/cm2,/? <0.05)
(WvlieetaL 1997).
No proliferation was observed for either chrysotile or crocidolite asbestos fibers in RPM
cells, but cytotoxicity was observed at concentrations >0.05 ug/cm2 (p <0.05). All talc samples
were less cytotoxic in both cell types. Comparing results of these samples when exposure is
D-30
-------
measured by fiber number, the same number of crocidolite asbestos fibers >5-um long leads to
proliferation in HTE cells, but proliferation did not occur for FD14 fibers. The other two talc
samples showed both insignificant cytotoxicity (SIS7) and significant cytotoxicity (CPS183,
p <0.05). Therefore, when measured by fiber number, the results show differential responses for
the fibers analyzed, suggesting the mineralogy of the fibers is more important in determining the
biological response to fibers. In the RPM cells, however, similar responses were seen for all
fibers analyzed, except for the slight cytotoxicity of FD14 at 2.6 fibers/cm2. This suggests that
fiber number does play a role in biological response in this cell type.
The results of these samples in both cell lines demonstrated that the cellular responses
seemed unrelated to the surface area, which demonstrates the impact of the dose metric on data.
Analyzing the data for cytotoxicity and proliferation based on the exposure measurement
demonstrated differences in response depending solely on how the fibers were measured (e.g., by
mass, number, or surface area). These results show variability in interpreting the same assay
based on the defined unit of exposure. Most early studies used mass as the measurement for
exposure, which can impact how the results are interpreted. When possible, further analysis of
fiber number and surface area might help elucidate the role of these metrics, particularly for in
vivo studies.
D.3.2.2. Genotoxicity
Athanasiou et al. (1992) performed a series of experiments to measure genotoxicity
following exposure to tremolite, including the Ames mutagenicity assay, micronuclei induction,
chromosomal aberrations, and gap-junction intercellular communication. Although a useful test
system for mutagenicity screening for many agents, the Ames assay is not the most effective test
to detect mutations induced by mineral fibers. Mineral fibers can cause mutation through
generation of ROS or direct disruption of the spindle apparatus during chromatid segregation.
Fibers do not induce ROS in the Ames system, however, and the Salmonella typhimurium strains
do not endocytose the fibers. Only one study was found in the published literature that used the
Ames assay to measure mutagenicity of tremolite. Metsovo tremolite asbestos has been shown
to be the causative agent of endemic pleural calcification and an increased level of malignant
pleural mesothelioma (see Section 4.1). To measure the mutagenicity of Metsovo tremolite,
S. typhimurium strains (TA98, TA100, and TA102) were exposed to 0-500 ug/plate of asbestos
(Athanasiou et al., 1992). This assay demonstrated that, like most asbestos fiber types tested in
earlier studies, Metsovo tremolite did not yield a significant increase in revertants in the Ames
assay, including in the TA102 Salmonella strain, which is generally sensitive to oxidative
damage. Although these strains can detect ROS mutations, they would not be able to produce
ROS from fibers alone or through necessary signaling pathways, and they do not endocytose
fibers. Thus, negative results in the Ames assay do not inform the genotoxicity of Metsovo
tremolite.
D-31
-------
Furthermore, this study demonstrated the clastogenic effects of tremolite, including
chromosomal aberrations and micronuclei induction. Tremolite exposure (0-3.0 ug/cm2) in
Syrian golden hamster embryo (SHE) cells resulted in a statistically significant increase in
chromosomal aberrations (p <0.02) when all treatment groups were combined and then
compared to controls; however, no clear dose-response relationship was evident (Athanasiou et
al., 1992). Tremolite exposure in SHE cells did lead to a dose-dependent increase in
chromosome aberrations that was statistically significant at the highest doses tested
(1.0-3.0 ug/cm2) (p<0.01) (see Table D-16).
Table D-16. Micronuclei induction (BPNi cells) and chromosomal aberrations
(SHE cells) following exposure to tremolite for 24 hours
Asbestos dose (ug/cm2)
0
0.5
1.0
2.0
3.0
Micronuclei
incidence/1,000 cells
17
31a
70b
205b
Not tested
Chromosomal aberrations (including chromatid gaps,
breaks, isochromatid breaks, and chromosome type)
o
J
4
12C
9a
13C
"Significantly different from control (p O.05).
bSignificantly different from control (p <0.01).
'Significantly different from control (p <0.02).
Source: Athanasiou et al. (1992).
Micronuclei induction was measured in BPNi cells after 24-hour exposure to
0-2.0 ug/cm2 tremolite. A statistically significant dose-dependent increase in levels of
micronuclei was demonstrated following tremolite exposure at concentrations as low as
0.5 ug/cm2 (p <0.01). Literatures searches did not find tremolite tested for clastogenicity in other
cell types, but the results of this study suggest interference with the spindle apparatus by these
fibers. No analysis was performed to determine whether fiber interference of the spindle
apparatus could be observed, which would have supported these results.
To determine whether tremolite has some tumor promoter characteristics, Athanasiou et
al. (1992) further examined intercellular communication following exposure to 0-4.0 ug/cm2
tremolite in both Chinese hamster lung fibroblasts (V79) and SHE BPNi cells, which are
sensitive to transformation. Inhibition of gap-junctional intercellular communication has been
proposed to detect tumor-promoting activity of carcinogens (Trosko et al., 1982). No effect on
gap-junction intercellular communication following tremolite exposure was observed.
D-32
-------
Okayasu et al. (1999) analyzed the mutagenicity of Metsovo tremolite, erionite, and the
man-made ceramic (RCF-1) fiber. Whether this tremolite is the same as that used in previous
studies from this group is unclear. Tremolite from Metsovo, Greece, used in this study was
characterized as 2.4 ±3.1 jim long and 0.175 ± 0.13 jim in diameter (arithmetic mean) with the
number of fibers per microgram of sample equal to 1.05 x 105. Human-hamster hybrid A(L)
cells contain a full set of hamster chromosomes and a single copy of human chromosome 11.
Mutagenesis of the CD59 locus on chromosome 11 is quantifiable by antibody
complement-mediated cytotoxicity assay. The authors state that this is a highly sensitive
mutagenicity assay, and previous studies have demonstrated mutagenicity of both crocidolite and
chrysotile (Hei et al., 1992). The cytotoxicity analysis for mutagenicity was performed by
exposing 1 x io5 A(L) cells to a range of concentrations of fibers as measured by weight
(0-400 ug/mL or 0-80 ug/cm2) for 24 hours at 37°C. CD59 mutant induction showed a
dose-dependent increase in mutation induction for erionite and tremolite, but RCF-1 did not.
D.4. SUMMARY
In vitro studies have been conducted with LAA from the Zonolite Mountain mine. These
studies demonstrated an effect of LAA on inflammation and immune function (Duncan et al.,
2010: Blake et al.. 2008: Blake et al.. 2007: Hamilton et al.. 2004). oxidative stress (Hillegass et
al., 2010), and genotoxicity (Pietruska et al., 2010). These results suggest that LAA may act
through similar mechanisms as other forms of asbestos, but data gaps still remain to determine
specific mechanisms involved in LAA-induced disease.
Studies that examined cellular response to tremolite also found that fiber characteristics
(length and width) play a role in determining ROS production, toxicity, and mutagenicity
(Okayasu et al., 1999: Wagner etal., 1982). As with the in vivo studies, the definition of fibers
and the methods of fiber measurement vary among studies.
D-33
-------
D.5. REFERENCES
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
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. D: Rogers. R: Smith. P: Chevalier. J. (2006). The lexicological response of Brazilian chrysotile asbestos:
a multidose subchronic 90-day inhalation toxicology study with 92-day recovery to assess cellular and
pathological response. Inhal Toxicol 18: 313-332. http://dx.doi.org/10.1080/08958370500515871
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
Blake. D: Bonn. C: Cox. D: Cardozo-Pelaez. F: Pfau. J. (2007). Internalization of Libby amphibole asbestos and
induction of oxidative stress in murine 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
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-Carhn. DJ: Schladweiler. MC: Shannahan. 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
Davis. JMG: Addison. J: Bolton. RE: Donaldson. K: Jones. AD: Miller. BG. (1985). Inhalation studies on the effects
of tremolite and brucite 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
Duncan. K: Ohio. A: Dailev. L: Bern. A: Gibbs-Flournoy. E: Padilla-Carhn. 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
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
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.
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
D-34
-------
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
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.0rg/10.1016/i.cvto.2012.08.025
Lowers. HA: Bern. AM. (2009). Particle size characterization of water-elutriated Libby amphibole 2000 and RTI
international amosite. Reston, VA: U.S. Geological Survey, http://pubs.usgs.gov/of/2009/1242/
McConnell EE: Rutter. HA: Ulland. BM: Moore. JA. (1983). Chronic effects of dietary exposure to amosite
asbestos andtremolite inF344 rats. Environ Health Perspect 53: 27-44.
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.
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,
NC: U.S. Department of Health and Human Services, National Toxicology Program.
http://ntp.niehs.nih.gov/ntp/htdocs/LT rptsAr279.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 rptsAr277.pdf
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
Padilla-Carlin. 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
Pfau. JC: Seib. 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
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
D-35
-------
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
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
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.
Saegusa. J: Kawano. S: Koshiba. M: Havashi. N: Kosaka. H: Funasaka. Y: Kumagai. S. (2002). Oxidative stress
mediates cell surface expression of SS-A/Ro antigen on keratinocytes. Free Radic Biol Med 32: 1006-1016.
http://dx.doi.org/10.1016/S0891-5849(02)00797-9
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. J Toxicol 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
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
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
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
D-36
-------
Smith. WE. (1978). Final report on biologic tests of samples 22260p5 and 22263p2. Madison, NJ: Health Research
Institute, Fairleigh Dickinson University.
Smith. WE: Hubert. DP. (1974). The intrapleural route as a means for estimating carcinogenicity. In E Karbe; JF
Park (Eds.), Experimental lung cancer: Carcinogenesis and bioassays: International symposium held at the
Battelle Seattle Research Center, Seattle, WA, USA, June 23-26, 1974 (pp. 92-101). New York, NY:
Springer-Verlag.
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.
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.
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
Trosko. JE: Yotti LP: Warren. ST: Tsushimoto. G: Chang. CC. (1982). Inhibition of cell-cell communication by
tumor promoters [Review]. Carcinog Compr Surv 7: 565-585.
Wagner. J: Chamberlain. M: Brown. R: Berry. G: Poolev. F: Davies. R: Griffiths. D. (1982). Biological effects of
tremolite. BrJ Cancer 45: 352-360.
Webber. JS: Blake. DJ: Ward. TJ: Pfau. JC. (2008). Separation and characterization of respirable amphibole fibers
fromLibby, Montana. Inhal Toxicol 20: 733-740. http://dx.doi.org/10.1080/08958370801932544
Wylie. AG: Bailey. KF: Kelse. JW: Lee. RJ. (1993). The importance of width in asbestos fiber carcinogenicity and
its implications for public policy [Review]. Am Ind Hyg Assoc J 54: 239-252.
http://dx.doi.org/10.1080/15298669391354621
Wylie. AG: Skinner. HCW: Marsh. J: Snyder. H: Garzione. C: Hodkinsoa D: Winters. R: Mossman. BT. (1997).
Mineralogical features associated with cytotoxic and proliferative effects of fibrous talc and asbestos on
rodent tracheal epithelial and pleural mesothelial cells. Toxicol Appl Pharmacol 147: 143-150.
http://dx.doi.org/10.1006/taap.1997.8276
D-37
-------
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.I. EXPOSURE DATA
The U.S. Environmental Protection Agency (EPA) collaborated with a research team at
the University of Cincinnati (UC) to update the exposure reconstruction for use in the
job-exposure matrix (JEM) for all workers in the Marysville, OH cohort, taking into account
additional industrial hygiene data that were not available for previous studies conducted in this
cohort. As discussed in detail in Appendix F, exposure estimates for each worker in the O.M.
Scott Marysville, OH plant were developed based on available industrial hygiene phase contrast
microscopy (PCM) data from the plant. Figure E-l shows the average exposure concentrations
of fibers in air (PCM fibers/cc) of each department from 1957 to 2000, indicating the time
periods when industrial hygiene data for fiber concentration in air were not available
(1957-1971) and were available (1972 and after).
E-l
-------
No Industrial Hygiene Data
o
CL
Industrial Hygiene Data
-Trionizinga
-Plant Maintenance
Central Maintenance
-Background
4 •
1955
1960
1995
2000
Figure E-l. Exposure concentrations in Marysville, OH facility.
aTrionizing is a term used in the Marysville, OH facility and includes unloading of railcars
containing vermiculite ore (track), using conveyers to move the vermiculite ore into the expander
furnaces, separation of the expanded vermiculite from sand, blending lawn-care chemicals, and
drying and packaging of the final product. As no unexpanded ore was used in the pilot plant,
research, polyform, office, packaging, or warehouse, jobs in these categories were assigned as the
background exposure. Workers assigned to plant maintenance activities spent 50% of their time
in trionizing and 50% of their time in areas assigned as plant background. Workers assigned to
central maintenance spent 10% of their time in trionizing areas and 90% of their time in areas
assigned as plant background. Central maintenance jobs were eliminated in 1982 and contracted
out (see Appendix F).
In brief, the starting point for the JEM was the estimated concentration of fibers in air
(fibers/cc) of each department from 1957-2000. The details are presented in Appendix F. Using
available data on the date of hire and the departments in which each person worked and taking
into account extensive overtime for some workers in some seasons, the cumulative exposure
(CE; fibers/cc-yr)12 for each worker for each season for each year since the date of hire was
estimated. The final CE metric (fibers/cc-yr) was obtained by adding the seasonal exposure
value for each worker for the total duration of employment for that worker. Each worker's CE
was then adjusted to a cumulative human equivalent exposure concentration (CHEEC;
fibers/cc-yr) to represent exposure 24 hours/day and 365 days/year (assuming that any exposure
off site was zero) for the full duration of employment. Note: Although Appendix F uses the
12Although the units of cumulative exposure are generally written as fibers/cc-year in the epidemiologic literature, it actually
means fibers/cc times years of exposure.
E-2
-------
term CHEEC, the more conventional term, CE, is used in this appendix to refer to cumulative
exposure adjusted to an equivalent human exposure adjusted to 24 hours/day and 365 days/year.
Mean exposure concentration (C, fibers/cc) was calculated by dividing the CE value
(fibers/cc-yrs) by the duration of exposure (years), where exposure duration was calculated as the
sum of the days worked by each worker (accounting for time away from work) divided by
365.25 days/year. Residence-time-weighted exposure (RTW, fibers/cc-yrs2) was calculated as
follows:
RTW = X [CE(s)-t(s)]/365.25 (E-l)
where:
CE(s) = cumulative exposure (fibers/cc-yrs) occurring in season "s"
t(s) = number of days between the midpoint of season "s" and the date of x-ray
This RTW exposure metric includes consideration of time since first exposure (TSFE) in that it
more heavily weights exposures in the past.
E.2. DATA SETS FOR MODELING
The primary analysis in Section 5.2.3 of this assessment models data for Marysville
workers evaluated in 2002-2005 and hired in 1972 or later without previous exposure to asbestos
(Rohs et al., 2008). The cohort is defined as the Rohs subcohort for this appendix. This data set
was chosen for the primary analysis because it was considered to have the highest quality
information as the job exposure matrix is directly supported by the industrial hygiene data and
the radiographs were evaluated by the same readers using the same evaluation guidelines. The
primary analysis estimates the effect of TSFE (years) using the larger subset of workers
evaluated in 2002-2005, regardless of hire date and without previous exposure to asbestos (Rohs
et al., 2008). This cohort is defined as the Rohs cohort for this appendix.
The complementary analysis in this appendix combines the radiographic evaluations for
all workers who participated in the Lockey et al. (1984) study and the follow-up study by Rohs et
al. (2008) and without previous exposure to asbestos. This cohort is defined as the combined
cohort for this appendix. This strategy was adopted as it provided the maximum range in TSFE
to inform the dependence of adverse health outcomes on TSFE. Outcome assessments (i.e.,
chest x-rays) were performed at two different time points, 1980 and 2002-2005, by different
readers.
The summary statistics for the three cohorts are presented in Table 5-3 of the main
document.
Radiographs were evaluated by two B Readers with a consensus evaluation by a third
reader in the case of disagreement in the original study by Lockey et al. (1984). In the follow-up
by Rohs et al. (2008), a radiographic reading was considered positive "when the median
E-3
-------
classification from the three independent B Readings was consistent with pleural and/or
interstitial changes" (see p. 631). Lockey et al. (1984) used modified International Labour
Organization (ILO) 1971 standards: Rohs et al. (2008) used ILO 2000 standards. The ILO 1971
standards did not provide separate diagnostic categories for localized pleural thickening (LPT)
and diffuse pleural thickening (DPT). The ILO 1971 standards included diagnostic categories
for pleural plaques and for other pleural thickening (PT). See Table 3 in Lockey et al. (1984) for
the summary of the original x-ray results.
The full data set used to model the exposure-response relationship for the adverse health
outcome obtained was as follows. The radiographic data from Lockey et al. (1984) (n = 513)
and Rohs et al. (2008) (n = 280), were combined for a total of 793 x-ray evaluations (this
includes repeated x-rays on the same individual). X-rays obtained from workers who reported
exposure to asbestos at other locations were excluded from consideration (n = 793 - 105 = 688
x-ray evaluations).
For workers who were x-rayed in both Lockey et al. (1984) and Rohs et al. (2008), one of
the observations was excluded (as described below) so that no repeat x-ray observation for any
individual worker in the data set was used for modeling. For workers who were negative for
radiographic changes in Lockey et al. (1984) and did not participate in Rohs et al. (2008), the
Lockey et al. (1984) data were retained. For workers who were negative for radiographic
changes in Lockey et al. (1984) and participated in Rohs et al. (2008), the Rohs et al. (2008) data
were retained. For workers who were positive for radiographic changes in Lockey et al. (1984)
and also in Rohs et al. (2008), the 1984 study data were retained. Two workers were positive in
1984 and negative in 2008. In accord with recommendations from the UC research group
(Lockey, 2013), the 2008 study data were retained for these two workers. The different results in
these two readings could be the result of a temporary cause (localized acute inflammation, fat
tissue, or pleural effusion that resolved), reader variability, or changed ILO criteria for pleural
abnormalities and do not imply that the pleural abnormality is reversible. This procedure for
assembling the data set for the full cohort resulted in
n = 688 x-rays - 252 duplicates = 436 x-rays, representing 436 individual workers. Two
workers from Lockey et al. (1984) were excluded because their hire date and the x-ray date were
the same (n = 436 - 2 = 434). For each worker, the estimated CE corresponded to the date of the
x-ray retained for analysis. That is, if the 1980 x-ray was used, the individual's CE estimate
covered the period from start of work through the x-ray date in 1980. If the 2002-2005 x-ray
was used, CE covered the period from start of work through the date of job stop or 2000,
whichever occurred earlier. The facility stopped using any vermiculite in its products in 2000.
All of the data used for modeling (x-ray diagnosis and exposure reconstruction) are
available in Health and Environmental Research Online. All of the modeling was done with the
individual exposure and health outcome data.
E-4
-------
E.3. STATISTICAL ANALYSIS OF THE COMBINED COHORT
E.3.1. Endpoints Modeled
The x-ray changes identified in the 1980 study (Lockev et al.. 1984) and the 2002-2005
study (Rohs et al., 2008) included the following:
• LPT (2002-2005 and as pleural plaques in 1980)
• PT (1980 only)
• DPT (2002-2005 only)
• Small interstitial opacities (1980 and 2002-2005)
Lockey et al. (1984) used modified 1971 ILO classification and reported costophrenic
angle obliteration (CAO) only, pleural plaques, and pleural thickening. There were no workers
with CAO in the latter two categories. Workers with CAO only were not included as someone
with an adverse health outcome attributed to exposure to fibers. A total of 10 workers had
pleural abnormalities attributed to exposure to Libby Amphibole asbestos (LAA) fibers. In this
assessment, EPA excluded one worker with pleural abnormalities because of previous exposure
to asbestos. One worker with plaques and one worker with pleural thickening, but no plaques,
had no abnormalities in their 2002-2005 radiographs. The more recent results are used for these
two workers. Among the remaining seven workers with pleural abnormalities in 1980, five
workers were diagnosed with pleural thickening and pleural plaque and two workers were
diagnosed with pleural thickening but no plaque. The two workers with pleural thickening and
no plaque were included in the LPT category. For the modeling of the combined cohort, these
endpoints can be grouped into three categories, as follows:
• LPT (includes LPT and PT, but not DPT; 70 cases)
• Any pleural thickening (APT) (includes LPT, PT, and 3 cases of DPT without LPT or
PT; 73 cases)
• Any radiographic change (ARC) (includes APT and 3 cases of small interstitial
opacities without any pleural thickening; 76 cases)
Of these three alternative endpoints, APT is identified as the preferred metric of outcome
because it is more inclusive and eliminates the uncertainty regarding the type of pleural
thickening observed in the 1980 study (Lockev et al., 1984) using the modified 1971 ILO
guidance. However, for completeness, modeling was also performed for LPT and ARC and
these results are also presented.
E-5
-------
E.3.2. Investigation of Explanatory Variables and Potential Confounders
The explanatory variables (other than CE) investigated included those related to time
(hire year, TSFE, exposure duration, and age at x-ray) and those not related to time (other
covariates including gender, smoking status, and body mass index [BMI]).
Regression models were used to determine whether each covariate (time-related or other
covariates) would meet the definition of a confounder—that is, whether it is associated with the
exposure in the study population, is associated with the outcome, and is not intermediate between
exposure and outcome (i.e., does not lie on the causal pathway). The association with
time-related variables was assessed using a univariate linear regression model. For that model,
the outcome was the natural log-transformed exposure metric (CE, mean exposure, or RTW
exposure) and the predictor was the covariate of interest. The association with outcome was
assessed using a univariate logistic model, where the outcome was APT and the predictor was
the covariate of interest. The results are summarized in Table E-l.
E-6
-------
Table E-l. Evaluation of association between covariates and exposure, and
between covariates and occurrence of any pleural thickening (APT).a Cells
display beta coefficient (standard error), />-value for predictor.15
Association with
cumulative exposure
Association with
mean exposure
Association with
RTW exposure
Association with
APT
Time-related
Hire yr
TSFE
Exposure duration
Age at x-ray
-0.1873 (0.0109),
O.OOOl
0.0719 (0.0070),
O.OOOl
0.1072 (0.0073),
O.OOOl
0.0713 (0.0060),
O.OOOl
-0.1040 (0.0095),
O.OOOl
0.0156 (0.0057),
0.0060
0.0309 (0.0066),
O.OOOl
0.0266 (0.0051),
O.OOOl
-0.2556 (0.0149),
O.OOOl
0.1456 (0.0075),
O.OOOl
0.1784 (0.0087),
O.OOOl
0.1294 (0.0070),
O.OOOl
-0.0970 (0.0184),
O.OOOl
0.1119 (0.0153),
O.OOOl
0.0988 (0.0145),
O.OOOl
0.0737 (0.0113),
O.OOOl
Other covariates
Male gender
Ever smoker0
Current
Former
Smoking pack-yrs
BMI (evaluated in
2002-2005 only)c
2.0119 (0.3849),
O.OOOl
0.5500 (0.2109),
0.0094
-0.0212 (0.2548),
0.9336
0.9622 (0.2334),
O.OOOl
0.01703 (0.00532)
0.0015
-0.0289 (0.0204),
0.1570
1.2180 (0.2949),
O.OOOl
0.3232 (0.1592),
0.0430
0.1610(0.1952),
0.4101
0.4402 (0.1787),
0.0142
0.00739 (0.00406)
0.0695
-0.0196 (0.0172),
0.2564
2.6587 (0.5255),
O.OOOl
0.6811 (0.2893),
0.0190
-0.3559 (0.3448),
0.3026
1.4293 (0.3158),
O.OOOl
0.02696 (0.00722)
0.0002
-0.0306 (0.0219),
0.1644
1.8754 (1.0247),
0.0672
0.2219 (0.2761),
0.4216
-1.1280 (0.4763),
0.0179
0.7464 (0.2887),
0.0097
0.00624 (0.00635)
0.3259
-0.0256 (0.0262),
0.3288
""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 APT status and the predictor is the covariate of interest. Based on n = 434 individuals
(73 cases of any PT and 361 without PT).
bBold entries indicate statistically significant associations.
°Data on smoking status were missing for five individuals in the full cohort. Data on BMI were unavailable for
216 individuals in the combined cohort.
Based on the statistical significance of the beta coefficients, each of the four time-related
variables (hire year, TSFE, exposure duration, and age at x-ray) was, as expected, strongly
associated with measures of fiber exposure (mean, cumulative, and RTW exposure). These four
time-related variables were also highly correlated with each other, with correlation coefficients
ranging from absolute magnitudes of 0.51 (between hire year and TSFE) to 0.85 (between
duration and TSFE); this high correlation raises concerns about collinearity which can cause
instability in regression models if highly correlated variables are included together. There is no
indication from the general literature on asbestos or durable mineral fibers that age is a risk
E-7
-------
factor for pleural thickening in the absence of exposure. There is considerable support from the
general asbestos literature that TSFE is often the most influential explanatory variable when
analyzing the exposure-response relationship for asbestos fibers (Paris et al., 2009; Paris et al.,
2008; Jakobsson et al., 1995; Ehrlich et al., 1992; Jarvholm, 1992). Consequently, among the
time-related variables, only TSFE was considered further as a separate predictor variable in
exposure-response modeling, noting that both CE and RTW CE include exposure duration within
the exposure metric. The correlation between TSFE and exposure was also high for certain
metrics, with correlation coefficients of 0.23 and 0.36 for TSFE and CE, and TSFE and RTW
exposure, respectively (p <0.0001 for both). However, the correlation between TSFE and mean
exposure was not significant (correlation coefficient of 0.07 \p = 0.1334]). In evaluating results
of models containing TSFE and either cumulative or RTW exposure, the potential for
collinearity and resultant model instability was considered.
The other covariates investigated included gender, smoking status, and BMI. Although
gender was associated with each of the LAA exposure variables (see Table E-l), it was not
associated with the outcome and thus not considered to be a potential confounder (or effect
modifier). The analysis based on gender is limited by the small number of females included in
the full cohort (n = 31), but there is no indication from the general literature on asbestos that
males and females have a different probability of developing pleural thickening following
exposure to asbestos.
The analysis of smoking status (current, former, or never smoker) appears contradictory
in that current smoker status appears to have an inverse relationship with the risk of APT,
relative to never smokers. However, on further investigation, it was evident that current smokers
had much lower TSFE compared to former smokers (medians of 13.7 and 31.6 years,
respectively), and were also on average younger than former smokers (median age at x-ray of
46 years compared to 56 years). The apparent discrepancy of smokers seeming to have lower
risk of APT than nonsmokers could be due, therefore, to the increased risk from longer TSFE
among former smokers, rather than a protective effect among current smokers. In addition, the
analyses based on ever-smoker status or on pack-years did not indicate potential confounding
(see Table E-l). This is consistent with the conclusion of the ATS (2004) that smoking does not
affect the presentation of asbestos-related pleural fibrosis. Consequently, smoking status was not
included in further analyses.
BMI was investigated as a potential confounder because fat pads along the chest wall can
sometimes be misdiagnosed as pleural thickening in conventional x-rays. Thus, there might be a
positive relation between BMI and pleural thickening. The analysis of BMI as a confounder is
limited because data on BMI were not collected in the 1980 study (Lockey etal., 1984). Using
the available data, the analysis showed that BMI was not a potential confounder as it was not
associated with either exposure or outcome (see Table E-l). Accordingly, BMI was not included
in further analyses.
E-8
-------
CE is commonly used in modeling of some asbestos outcomes. What is known about the
distribution and retention of inhaled fibers is summarized in Section 3. For example, lung cancer
exposure-response for asbestos is usually modeled with CE. However, for pleural effects from
exposure to chrysotile asbestos, a mean exposure metric in a model that included TSFE was also
proposed (Paris et al., 2008). Therefore, for this assessment several exposure metrics were
investigated in the modeling, including mean exposure (fibers/cc), CE (fibers/cc-yr), and RTW
CE (fibers/cc-yrs2).
The importance of TSFE to date of x-ray is clearly illustrated by comparing the results of
Lockeyet al. (1984) with the results ofRohs etal. (2008). These two studies were conducted in
the same worker population (with some loss to follow-up) 24 years apart. In the initial study
(Lockey etal., 1984), only 2% of the individuals showed pleural changes; in the follow-up study
(Rohs et al., 2008), 28% of the individuals showed pleural changes. There was very little
additional exposure to fibers after 1980. This result is consistent with findings in other
occupational cohorts exposed to various forms of asbestos fibers that TSFE is a significant
explanatory variable for pleural thickening, even in the absence of continued exposure (Paris et
al., 2009: Paris et al., 2008: Jakobsson et al., 1995: Ehrlich et al., 1992: Jarvholm, 1992). An
important point of clarification is that TSFE is not the same as time to the first appearance of the
adverse health outcome. The pleural thickening or small interstitial opacities could have formed
at any time between the start of exposure and the time the endpoint was observed in the x-rays
(e.g., for a worker who was negative for APT in 1980 but positive in 2004, the APT could have
occurred anytime between 1980 and 2004, as only two x-ray events are available).
It has been suggested that pleural abnormalities increase in extent (Sichletidis et al.,
2006) and that the prevalence increases as a function of TSFE (Lilis et al., 1991c). The fibers
persist in the respiratory tract and pleural tissue for a long time and can continue to damage the
tissue even in the absence of continued exposure. An alternative explanation is that the fibrosis
is already initiated by the exposure, but additional time is needed for the lesion to progress in
size to be visible on the x-ray. There are no data available to distinguish these possibilities in the
Marysville cohorts. Therefore, models that include TSFE as an explanatory variable and allow
for the prevalence of APT to increase with longer follow-up even in the absence of continued
exposure were given some preference in this analysis.
E.3.3. Model Forms and Exposure Metrics
A range of model forms were investigated to determine which was most appropriate for
use in characterizing the exposure-response relationship to derive the point of departure (POD).
These models forms are summarized in Table E-2.
E-9
-------
Table E-2. Model forms considered in developing the point of departure
(POD)
Category
Univariate
X-C, CE,
RTW
Bivariate
X - L, CE
T = time from
first
exposure
Name
Log-Logistic
Dichotomous Hill
a) Estimated
plateau
b) Fixed plateau
Bivariate
log-logistic
Bivariate
Dichotomous Hill
a) Estimated
plateau
b) Fixed plateau
Cumulative normal
Dichotomous Hill
Cumulative normal
Michaelis-Menten
Code
UVLL
UVDH
UV DH FP
BVLL
BVDH
BV DH FP
CNDH
CNMM
Equation
1 - bkg
p(x) bkg 1 1 + exp[_a _ b x lnM]
Plateau — bkg
p(x) bkg \ l + exp[_a _ b x lnM]
1 - bkg
,„/•„ TA !,;,„! u
pb.T) bkg 1 1 + exp[_a _ b xln(;0_ c x T]
Plateau — bkq
•n(v T^ hl-n 1
pb.T) bkg \ 1 + exp[_a _ b xlnM _ c x T]
nfr n bj-3 i Plateau^ ~ bka
p(i,T) bkg \ 1 + exp[_a _ b x taM]
Plateau(T) = bkg + (1 - &fc0) x <£(r|m,s)
vh n _ fi/., , ^CeauOO - ifctf
PCx,r; 6/tfl i 1 + exp[_a _ lnM]
Plateau(T) = bkg + (1 - bkg) x *(r|m,s)
The background prevalence (bkg) of the health effect of interest was treated as an
estimated parameter for all models. The exposure metrics tested included mean exposure
concentration, CE, and RTW exposure. Univariate models that were tested included log-logistic
(UV LL) and Dichotomous Hill with estimated (UV DH) or fixed plateau (UV DH FP).
Bivariate models that were tested included log-logistic (BV LL) and Dichotomous Hill
with estimated (BV DH) or fixed plateau (BV DH FP) with the same exposure metrics as noted
above and with TSFE incorporated as an additional explanatory variable in the exponential term.
This is the more conventional way of incorporating an additional explanatory variable in a
logistic model.
As an additional approach, modified versions of the Dichotomous Hill and
Michaelis-Menten models were evaluated with the exposure metrics of mean and CE and with
the plateau term modeled as a function dependent on TSFE. These model forms were tested
because a plot of the shape of the prevalence curve (based on APT) as a function of TSFE at
fixed CE shows that that the curve begins low and then rises in a nonlinear fashion (see
Figure E-2 panel A), and that the "plateau" at high CE tends to increase as TSFE increases (see
Figure E-2 panel B). This behavior suggests that expressing the plateau as a function with an
S-shape could be suitable. Several S-shaped curves were tested, including the cumulative
E-10
-------
normal, cumulative gamma, and cumulative Weibull. Based on Akaike Information Criterion
(AIC), there was no significant difference in performance for any of these functions; therefore,
the cumulative normal function was chosen because of its familiarity and ease of use. The
resulting models are referred to as the cumulative normal Dichotomous Hill (CN DH) or the
cumulative normal Michaelis-Menten (CN MM). The effect of increasing TSFE in these model
forms is to increase the plateau (maximum prevalence at high exposure) and also to increase the
slope of the response (the increase in prevalence per unit increase in exposure). However, these
model forms do not allow TSFE to function as a separate predictor of prevalence alongside with
the exposure metric as in the BV LL and BV DH models. It is acknowledged that this is a less
conventional way of incorporating an additional explanatory variable in a logistic model.
However, these model forms based on the cumulative normal function are included so that the
results with this data set can be judged along with the results of other models.
E-ll
-------
Panel A: Prevalence vs. time since first exposure stratified by cumulative exposure.
100%
CEBins
10 20 30
Time Since First Exposure (yrs)
40
50
Bin
Number
CE1
CE2
CE3
CE4
CE5
Bin Lower
Bound
0.00
0.23
0.91
1.73
7.20
Bin Upper
Bound
0.23
0.91
1.73
7.20
100.00
Mean
Value
0.09
0.53
1.20
3.22
34.50
No. of
Workers
87
87
86
87
87
No. of
Cases
2
5
16
19
31
Prev
2.3%
5.7%
18.6%
21.8%
35.6%
Panel B: Prevalence vs. cumulative exposure stratified by time since first exposure
100% T
20
CE (f/cc-yrs)
TSFE Bins
Bin
Number
Tl
T2
T3
T4
T5
Bin Lower
Bound
0.0
9.8
23.4
29.8
36.3
Bin Upper
Bound
9.8
23.4
29.8
36.3
100.0
Mean
Value
4.4
15.7
26.0
32.7
42.4
No. of
Workers
87
80
93
87
87
No. of
Cases
1
5
7
20
40
Prev
1.1%
6.3%
7.5%
23.0%
46.0%
Figure E-2. Observed dependence of any pleural thickening (APT)
prevalence on cumulative exposure (CE) and time since first exposure
(TSFE).
E-12
-------
E.3.4. Benchmark Response
As discussed in Section 5.2.2.4, a benchmark response (BMR) of 10% extra risk is used
in this assessment. For the modeling of the full cohort a BMR of 10% extra risk is also used for
all endpoints (LPT, APT, or ARC). The vast majority of cases in the combined cohort are
classified as LPT. Using the same BMR across all endpoints also permits easier comparison of
the results.
E.3.5. Modeling Results
The modeling results are summarized in Table E-3 through E-5, below. For models that
include TSFE as an independent explanatory variable, the results are shown for two alternative
values: TSFE = 70 years and TSFE = 25 years. These two values were selected because
25 years is the median value of TSFE for the combined cohort, and consequently using the
values for TSFE = 25 years does not require extrapolation outside the observed range of the data.
Results were derived for TSFE = 70 years because the ultimate objective of this effort is to
derive a reference concentration (RfC) that is applicable to an individual exposed for 70 years.
E-13
-------
Table E-3. Modeling results for localized pleural thickening (LPT) in the combined cohort of Marysville workers evaluated
in 1980 or in 2002-2005
Model
UV
LL
UV
DHFP
UV
DH
BV
LL
BV
DHFP
BV
DH
CN
MM
Exp.
metric
C
CE
rtwCE
C
CE
rtwCE
C
CE
rtwCE
C,
TSFE
CE,
TSFE
C,
TSFE
CE,
TSFE
C,
TSFE
CE,
TSFE
C,
TSFE
CE,
TSFE
bkg a b c m s Plateau3
0.000 -0.921 0.338 - - - 1.000
0.014 -2.127 0.466 - - - 1.000
0.014 -3.780 0.534 - - - 1.000
0.000 -0.675 0.357 - - - 0.850
0.016 -1.960 0.502 - - - 0.850
0.014 -3.731 0.578 - - - 0.850
0.007 2.388 0.903 - - - 0.309
0.025 -0.767 1.992 - - - 0.325
0.020 -7.107 2.033 - - - 0.371
0.008 -4.377 0.338 0.112 - - 1.000
0.010 -5.190 0.348 0.103 - - 1.000
0.009 -4.434 0.403 0.125 - - 0.850
0.012 -5.382 0.409 0.114 - - 0.850
0.015 -5.151 0.667 0.190 - - 0.559
0.016 -6.387 0.615 0.160 - - 0.586
0.011 3.151 1.000 - 38.47 13.24 0.991
0.015 -0.273 1.000 - 38.27 14.00 0.988
H-Lp AIC
0.030 370.49
0.175 350.35
0.383 328.39
0.031 370.35
0.189 350.04
0.445 327.92
0.048 370.64
0.526 346.33
0.895 324.73
0.169 301.80
0.012 300.97
0.220 300.71
0.002 299.98
0.643 301.05
0.062 300.78
0.200 299.42
0.540 298.22
BMD BMDL
(70) (70)
2.3 x ID-2 5.9 x 1Q-3
8.6 x 10-' 2.6 x 10-'
1.9 xlQ1 8.7x10°
2.3 x ID-2 6.3 x 1Q-3
9.0 x 10-' 2.8 x 10-'
2.0 x 1Q1 9.1 x 10°
3.2 x ID-2 1.2 x ID-2
1.0x10° 6.7x10-'
2.1 xlQ1 1.4 xlQ1
5.4 x 1Q-8 <1 x 1Q-12
5.2 x \Q-6 1.8 x 1Q-11
1.5 x 1Q-7 1.2 x ID"12
1.2 x ID-5 1.4 x 10-'°
5.1 x 1Q-7 3.9 x 1Q-11
3.2 x 1Q-5 1.4 x 1Q-9
4.8 x ID"3 2.1 x ID"3
1.5 x 1Q-1 6.3 x 1Q-2
BMC BMCL
(70) (70)
2.3 x 1Q-2 5.9 x 1Q-3
1.2xlO-2 3.7x10-3
7.9 x ID-3 35 x 10-s
2.3 x 1Q-2 6.3 x 1Q-3
1.3 x IQ-2 3.9 x ID-3
8.0 x ID-3 37 x 10-s
3.2 x 1Q-2 1.2 x 1Q-2
1.5 x IQ-2 9.6 x ID-3
8.4 x ID-3 5 6 x iQ-3
5.4 x 1Q-8 <1 x 1Q-12
7.5 x ID-8 2.5 x 1Q-13
1.5 x 1Q-7 1.2 x 1Q-12
1.7xlO-7 2.0 xlQ-12
5.1 x 1Q-7 3.9 x 1Q-11
4.6 x 1Q-7 2.0 x 1Q-11
4.8 x ID-3 2.1 x ID-3
2.1 x ID-3 90 x iQ-4
BMD BMDL
(25) (25)
1.6 x 1Q-1 5.0 x 1Q-2
3.3 x 10° 1.0 x 10°
1.7X1Q-1 6.1 xlQ-2
3.5 x 10° 1.2 x 10°
1.9 x 1Q-1 8.4 x 1Q-2
3.8 x 10° 1.6 x 10°
7.8 x 1Q-2 3.2 x 1Q-2
1.8 x 10° 8.2 x 1Q-1
BMC BMCL
(25) (25)
1.6 x 1Q-1 5.0 x 1Q-2
1.3 x 1Q-1 4.2 x 1Q-2
1.7X1Q-1 6.1 xlQ-2
1.4 x IQ-1 4.9 x 1Q-2
1.9 x 1Q-1 8.4 x 1Q-2
1.5 x IQ-1 6.3 x 1Q-2
7.8 x 1Q-2 3.2 x 1Q-2
7.3 x 1Q-2 3.3 x 1Q-2
E-14
-------
Table E-3. Modeling results for localized pleural thickening (LPT) in the combined cohort of Marysville workers evaluated
in 1980 or in 2002-2005 (continued)
Model
CN
DH
Exp.
metric
C,
TSFE
CE,
TSFE
bkg a b c m s Plateau3
0.017 11.05 3.26 - 41.02 13.88 0.982
0.018 -0.233 1.592 - 39.59 14.58 0.981
H-Lp AIC
0.435 299.83
0.158 299.49
BMD BMDL
(70) (70)
3.0 x 1Q-1 2.8 x ID-2
BMC BMCL
(70) (70)
1.7xlO-2 2.0x10-3
4.2 x 1Q-3 4.0 x 1Q-4
BMD BMDL
(25) (25)
3.7 x ID"2 2.4 x 1Q-2
1.6 x 10° 8.8 x 10-'
BMC BMCL
(25) (25)
3.7 x ID"2 2.4 x 1Q-2
6.5 x 1Q-2 3.5 x 1Q-2
Grey Cells = Although a maximum likelihood solution was obtained at TSFE = 25, a value for lower limit of the benchmark concentration (BMCL) could not be derived.
Consequently, the model was derived for TSFE = 28, where a value for BMCL could be estimated. The median value for TSFE in the Rohs cohort is 28 yrs.
Tor CN models, the plateau term shown is for TSFE = 70 yrs.
Exp = exposure; R-Lp = Hosmer-Lemeshow/>-value; BMC = benchmark concentration; BMD = benchmark dose; BMDL = lower limit of the benchmark dose.
a, b, c, m, and s are model fitting parameters.
E-15
-------
Table E-4. Modeling results for any pleural thickening (APT) in the combined cohort of Marysville workers evaluated in 1980
or in 2002-2005
Model
UVLL
UVDH
FP
UVDH
BVLL
BVDH
FP
BVDH
CN
MM
CNDH
Exp.
metric
C
CE
rtwCE
C
CE
rtwCE
C
CE
rtwCE
C, TSFE
CE, TSFE
C, TSFE
CE, TSFE
C, TSFE
CE, TSFE
C, TSFE
CE, TSFE
C, TSFE
CE, TSFE
bkg a b c m s Plateau3
0.000 -0.846 0.350 - - - 1.000
0.013 -2.062 0.469 - - - 1.000
0.013 -3.760 0.545 - - - 1.000
0.000 -0.592 0.372 - - - 0.850
0.015 -1.896 0.509 - - - 0.850
0.013 -3.732 0.596 - - - 0.850
0.011 2.673 1.000 - - - 0.318
0.025 -0.760 2.198 - - - 0.335
0.020 -7.567 2.161 - - - 0.389
0.008 -4.422 0.360 0.116 - - 1.000
0.010 -5.263 0.356 0.107 - - 1.000
0.010 -4.546 0.443 0.133 - - 0.850
0.012 -5.549 0.431 0.121 - - 0.850
0.015 -5.393 0.707 0.199 - - 0.588
0.018 -7.165 0.709 0.186 - - 0.577
0.011 3.134 1.000 - 37.53 12.64 0.995
0.015 -0.243 1.000 - 37.48 13.42 0.992
0.017 11.172 3.326 - 39.89 13.20 0.989
0.018 -0.222 1.778 - 38.89 14.05 0.987
H-Lp AIC
0.061 378.25
0.193 357.63
0.355 333.67
0.063 378.06
0.210 357.23
0.425 333.03
0.091 377.95
0.611 352.14
0.936 328.53
0.238 303.30
0.022 303.38
0.455 301.68
0.006 301.94
0.724 301.62
0.068 302.08
0.232 300.86
0.665 300.27
0.534 300.53
0.243 301.04
BMD BMDL
(70) (70)
2.1 x ID"2 5.7 x ID"3
7.5 x iQ-i 2.3 x iQ-i
1.8 xlQ1 8.1x10°
2.2 x ID"2 6.2 x lO-3
8.0 x iQ-i 2.5 x iQ-i
1.8 x 1Q1 8.6 x 10°
3.3 x ID"2 1.2 x ID"2
9.9 x iQ-i 6.67 x IQ-i
2.1x10' 1.38x10'
7.1 x 1Q-8 <1 x 10-12
4.0 x IQ-6 2.16 x iQ-ii
2.2 x IQ-7 1.0 x 10-11
1.1 x 1Q-5 2.7 x 10-1°
6.0 x 1Q-7 1.4 x 10-1°
3.0 x 1Q-5 3.3 x 1Q-9
4.9 x 1Q-3 2.2 x 1Q-3
1.4 x ID"1 6.4 x ID"2
1.8 x 1Q-2 36 x 10-3
3.3 x iQ-i 53 x 10-2
BMC BMCL
(70) (70)
2.1 x ID"2 57 x iQ-3
1.1 x ID"2 3.3 x ID"3
7.2 x 1Q-3 33 x iQ-3
2.2 x ID-2 6.2 x iQ-3
1.1 x ID"2 3.6 x ID"3
7.3 x ID-3 35 x iQ-3
3.3 x 1Q-2 1.2 x 1Q-2
1.4 x ID"2 9.5 x 1Q-3
8.4 x ID-3 56 x iQ-3
7.1 x ID-8 <1 x 10-12
5.7 x 1Q-8 <1 x 10-12
2.2 x IQ-7 1.0 x 10-11
1.6 x 1Q-7 3.8 x 10-12
6.0 x 1Q-7 1.4 x 10-1°
4.3 x 1Q-7 4.7 x iQ-ii
4.9 x 1Q-3 2.2 x 1Q-3
2.0 x ID"3 9.1 x 1Q-4
1.8 x 1Q-2 36 x ID-3
4.7 x ID"3 7.5 x 1Q-4
BMD BMDL
(25) (25)
1.5 x ID-1 5.0 x 1Q-2
3.0 x 10° 9.6 x ID"1
1.7X10-1 6.3x10-2
3.3 x 10° 1.1 x 10°
1.9X10-1 8.7x10-2
4.0x10° 1.7x10°
7.2 x 1Q-2 3.1 x 1Q-2
1.7x10° 7.6x10-1
5.0 x 1Q-2 32 x 1Q-2
1.5 x 10° 8.6 x ID-1
BMC BMCL
(25) (25)
1.5 x ID-1 5.0 x 1Q-2
1.2 x ID-1 3.8 x 1Q-2
1.7X10-1 6.3x10-2
1.3 x iQ-i 4.2 x 1Q-2
1.9x10-1 8.7x10-2
1.6X10-1 6.7x10-2
7.2 x 1Q-2 3.1 x 1Q-2
6.7 x ID"2 3.1 x ID"2
5.0 x 1Q-2 32 x 1Q-2
6.0 x 1Q-2 3.4 x 1Q-2
Tor CN models, the plateau term shown is for TSFE = 70 yrs.
E-16
-------
Table E-5. Modeling results for any radiographic change (ARC) in the combined cohort of Marysville workers evaluated in
1980 or in 2002-2005
Model
UVLL
UVDH
FP
UVDH
BVLL
BVDH
FP
BVDH
CN
MM
CNDH
Exp.
metric
C
CE
rtwCE
C
CE
rtwCE
C
CE
rtwCE
C, TSFE
CE, TSFE
C, TSFE
CE, TSFE
C, TSFE
CE, TSFE
C, TSFE
CE, TSFE
C, TSFE
CE, TSFE
bkg a b c m s Plateau3
0.000 -0.732 0.383 - - - 1.000
0.015 -2.075 0.510 - - - 1.000
0.013 -3.859 0.580 - - - 1.000
0.000 -0.469 0.408 - - - 0.850
0.016 -1.904 0.553 - - - 0.850
0.000 -0.469 0.408 - - - 0.850
0.000 1.754 0.776 - - - 0.384
0.025 -0.849 2.139 - - - 0.358
0.020 -7.244 2.022 - - - 0.420
0.008 -4.303 0.416 0.118 - - 1.000
0.011 -5.263 0.410 0.107 - - 1.000
0.010 -4.443 0.507 0.136 - - 0.850
0.012 -5.593 0.495 0.122 - - 0.850
0.016 -5.300 0.774 0.202 - - 0.610
0.018 -7.299 0.787 0.189 - - 0.594
0.011 3.012 1.000 - 36.23 12.29 0.997
0.015 -0.356 1.000 - 36.13 13.10 0.995
0.016 9.507 2.883 - 38.64 12.89 0.993
0.018 -0.326 1.751 - 37.71 13.82 0.990
H-Lp AIC
0.055 382.856
0.238 360.193
0.374 334.455
0.057 382.701
0.262 359.875
0.057 382.701
0.073 383.619
0.570 356.772
0.882 331.410
0.237 304.545
0.006 304.776
0.379 302.957
0.092 303.329
0.869 303.397
0.439 303.896
0.289 303.611
0.346 303.006
0.202 303.621
0.443 303.906
BMD BMDL
(70) (70)
2.2 x 1Q-2 7.1 x ID"3
7.9 x 1Q-1 2.6 x 1Q-1
1.8 x 1Q1 8.4 x 10°
2.3 x 1Q-2 7.6 x lO-3
8.2 x 1Q-1 2.8 x 1Q-1
2.3 x 1Q-2 76 x 10-3
2.7x10-2 1 .OxlQ-2
9.9 x 1Q-1 6.6 x 1Q-1
2.1x10' 1.4x10'
3.6 x 1Q-7 1.0 x 1Q-10
2.1 x 1Q-5 4.1 x 1Q-9
8.2 x ID-7 6.1 x IQ-10
4.4 x io~5 2.1 x 1Q-8
1.4 x IQ-6 2.6 x 1Q-9
7.3 x 1Q-5 8.3 x 1Q-8
5.5 x 1Q-3 2.6 x 1Q-3
1.6 x 1Q-1 7.6 x ID"2
1.7x10-2 3.4x10-3
3.5 x IQ-1 5.4 x 1Q-2
BMC BMCL
(70) (70)
2.2 x ID-2 7 j x 10-s
1.1 x ID"2 3.7 x ID"3
7.1 x ID-3 34 x iQ-3
2.3 x ID-2 76 x iQ-3
1.2 x ID"2 4.1 x 1Q-3
9.2 x 1Q-6 3.1 x 1Q-6
2.7x10-2 l.OxlQ-2
1.4 x ID"2 9.4 x 1Q-3
8.4 x ID-3 56 x ID-3
3.6 x 1Q-7 1.0 x 1Q-10
3.0 x 1Q-7 5.9 x 1Q-11
8.2 x IQ-7 6.1 x IQ-10
6.3 x 1Q-7 3.0 x 1Q-10
1.4 x \Q-6 2.6 x 1Q-9
1.0 x \Q-6 1.2 x 1Q-9
5.5 x 1Q-3 2.6 x ID"3
2.3 x ID"3 1.1 x ID"3
1.7x10-2 3.4x10-3
4.9 x 1Q-3 7.7 x 1Q-4
BMD BMDL
(25) (25)
1.3 x IQ-1 5.0 x 1Q-2
2.6 x 10° 9.6 x 1Q-1
1.5 x 10-' 6.3 x 1Q-2
2.9 x 10° 1.2 x 10°
1.7x10-' 8.3x10-2
3.6 x 10° 1.6 x 10°
6.1 x 1Q-2 3.0 x 1Q-2
1.5 x 10° 7.3 x 1Q-1
4.9 x 1Q-2 33 x 10-2
1.4 x 10° 8.3 x IQ-1
BMC BMCL
(25) (25)
1.3 x IQ-1 5.0 x 1Q-2
1.0 x 1Q-1 3.9 x 1Q-2
1.5 x 10-' 6.3 x 1Q-2
1.2x10-' 4.7x10-2
1.7x10-' 8.3x10-2
1.4 x 1Q-1 6.4 x ID-2
6.1 x 1Q-2 3.0 x 1Q-2
5.8 x ID"2 2.9 x ID"2
4.9 x 1Q-2 33 x 10-2
5.5 x 1Q-2 3.3 x 1Q-2
Tor CN models, the plateau term shown is for TSFE = 70 yrs.
E-17
-------
The units in the benchmark dose (BMD) and lower limit of the BMD (BMDL) columns
are those of the exposure metric used in the model. When the exposure metric was based on CE,
the benchmark concentration (BMC) and lower limit of the BMC (BMCL) values were
calculated by dividing the BMD and BMDL by the value of TSFE. When the exposure metric
was RTW, the BMC and BMCL were calculated by dividing the integral of TSFE from 0 to
70 years (=702/2). The units of the BMC and BMCL are fibers/cc.
E.3.6. Considerations for Identification of the Preferred Model(s) for the Combined
Cohort
The following factors were considered in evaluating the model results in order to identify
models that might provide a sound basis for selection of a POD and derivation of an RfC.
• All models with an unacceptable Hosmer-Lemeshow (H-L) goodness-of-fit statistic
(p <0.10) were eliminated. This is in accord with the approach usually followed in
BMD modeling (U.S. EPA. 2012).
• Models with lower AIC values were generally preferred over models with higher AIC
values. Fits of models with AIC values that were within 2 units of each other were
considered to be approximately equivalent (U.S. EPA, 2012; Burnham and Anderson,
2002).
• Models with a relatively low fitted plateau (the maximum prevalence at high
exposure and long TSFE) were given lower priority. This factor was considered
because the prevalence of an adverse health outcome in individuals exposed to
asbestos fibers is expected to approach some relatively high value (e.g., 80-100%) in
situations with high exposure and long follow-up time (Winters etal., 2012;
Jarvholm. 1992: Lilis et al.. 1991c).
• Model results with a wide interval between the BMC and BMCL were given low
priority because the wide interval indicates an uncertain value for BMCL.
• Models that had a good visual agreement between observed and predicted responses,
especially in the region of the BMR, were preferred over models with poor
agreement. This is implemented by inspection of graphs of the predicted response
from the model with the observed data, stratified into bins. This is a subjective
evaluation and depends in part on how the observed data are binned.
E.3.7. Selection of the Preferred Models for the Combined Cohort
The model results presented in Tables E-3 to E-5 were reviewed with respect to the
factors described above. In general, findings were similar for all three endpoints, with the
following main conclusions:
• All univariate models (UV LL, UV DH, and UV DH FP) based on the three exposure
metrics (mean, cumulative, and RTW exposure) performed relatively poorly, as
E-18
-------
indicated by high AIC values greater than 25 units larger than the best fitting models
within each endpoint (see Tables E-3 to E-5) and/or H-L^-values below 0.1.
Consequently, this class of models was not retained for further consideration in the
derivation of the POD.
• Bivariate Dichotomous Hill models based on C or CE and TSFE were not considered
as the fitted plateau term was considerably lower than 85%.
• Bivariate models based on TSFE and C or CE where TSFE acts on the slope term
directly (BV LL and BV DH FP) generally yielded results with favorable AIC values.
However, models based on TSFE and CE had H-L^-values below 0.1 and were not
considered further. Models based on TSFE and C had adequate H-L^-values
(p >0.1) and had relatively narrow intervals between the BMC and the BMCL when
TSFE = 25 years (the median value for the combined cohort). In contrast, these same
models yielded results with extremely wide intervals between the BMC and the
BMCL when extrapolated to TSFE = 70 years. These results indicate the potential
for considerable model uncertainty when extrapolating beyond the range of the
observed data. Consequently, this class of models was retained for further
consideration in the derivation of the POD only for TSFE = 25 years. Because the
results at TSFE = 25 years are quite similar for the BV LL and BV DH FP models,
only the BV DH FP models were assessed further. This is consistent with the
modeling presented in Section 5 of the main document.
• Bivariate models, where the TSFE term acts on the plateau term (CN MM and CN
DH) with either mean or CE, demonstrated adequate goodness of fit with H-L
^-values >0.1, and had low AIC values, a plateau term that approaches 1.0 at high
TSFE, and relatively narrow intervals between the BMC and the BMCL
(BMC/BMCL ratios of approximately 2 at TSFE = 25 years and approximately 6 at
TSFE = 70 years). Consequently, this class of models where the TSFE variable acts
on the plateau term (CN DH and CN MM) was also retained for further
consideration in the derivation of the POD.
Based upon these considerations, five models were identified for further consideration
including the BV DH FP (C, TSFE),13 CN DH (C or CE, TSFE) and the CN MM (C or CE,
TSFE). Each of these models demonstrated adequate goodness of fit, and incorporates both
exposure and TSFE as predictors of the prevalence of pleural thickening, and have comparable
AIC values (usually within 2 units of each other).
Between the CN models (CN DH and CN MM) where the plateau term is a function of
TSFE, there was no clear and consistent statistical basis (i.e., goodness of fit or relative fit) for
distinguishing between C and CE as the preferred exposure metric. However, it should be noted
that there is not a large difference in the BMCL regardless of whether C or CE is used as the
exposure metric. The model using CE was used in this analysis of the combined cohort because
CE is commonly used in exposure-response modeling for asbestos. In addition, there was a
statistically significant association between duration of exposure and prevalence of APT in the
13This notation indicates the model form and the explanatory variable in parentheses.
E-19
-------
univariate analysis (see Table E-l). Finally, duration of exposure was not included as a separate
predictor because CE includes duration of exposure. A relationship between CE and the adverse
health effects could reflect a cumulative increase in internal dose for the fiber or could reflect
accumulating tissue damage.
Likewise, between the two CN models, there was little statistical basis for distinguishing
between the MM models and the DH models, and both yielded similar BMCL values. The CN
DH model was selected as being more flexible compared to the CN MM model because it treats
the shape term for the exposure metric as a fitting parameter as opposed to assigning the shape
term to 1 as in the CN MM model. Thus, the two models given highest priority for deriving a
POD for the combined cohort were the CN DH model with CE (for TSFE = 25 or
TSFE = 70 years) and the BV DH FP model with C (for TSFE = 25 years). Results from these
two model forms are presented in the remainder of this appendix.
In order to compare observed APT prevalence in the combined cohort (73 cases in
434 workers) to that predicted by the CN DH model using CE and TSFE as explanatory
variables, it is necessary to group the workers into bins according to TSFE and CE. The CE bins
were formed by dividing the cohort into four groups of approximately equal size, as shown in
Table E-6.
Table E-6. Cumulative exposure (CE) bins
CEbin
1
2
3
4
Bin lower bound
0.00
0.34
1.13
3.74
Bin upper bound
0.34
1.13
3.74
96.91
Number of workers in bin
109
108
108
109
The TSFE bins were formed by dividing the cohort into three groups of similar size, as
shown in Table E-7.
Table E-7. Time since first exposure (TSFE) bins
TSFE bin
1
2
3
Bin lower bound
0
20.0
30.0
Bin upper bound
20.0
30.0
50.0
Number of workers in bin
141
125
168
Using these binning rules yielded the prevalence of APT for each bin is shown in
Table E-8.
E-20
-------
Table E-8. Prevalence of any pleural thickening (APT) stratified into bins
TSFE bin midpoint
10
25
40
Cumulative exposure bin midpoint (fibers/cc-yr)
0.17
1/58= 0.02
2/44= 0.05
0/7 = 0
0.73
1/37 = 0.03
0/25 = 0
8/46 = 0.17
2.43
0/19 = 0
6/37 = 0.16
18/52 = 0.35
50.33
1/27 = 0.04
3/19 = 0.16
33/63 = 0.52
Figure E-3 shows the agreement between observed APT prevalence (shown as data
points) and predicted prevalence (shown as smooth lines) for the CN DH (CE, TSFE) model.
Panel A compares observed to predicted as a function of CE, stratified by average TSFE. The
red line represents the model predictions for TSFE = 70 years. Note that there are no workers
with this long a length of follow-up, so there are no observations to compare to the model
predictions for this curve. Panel B compares the observed and predicted prevalence as a function
of TSFE, stratified by CE. As illustrated, the agreement between observed and predicted
prevalence is relatively good in both dimensions. These graphs help illustrate the key feature of
the CN DH model, which is that the maximum prevalence at high CE is low for short TSFE, and
increases towards 1.0 only as TSFE increases towards 70 (see Panel A). Likewise, Panel B
shows that for TSFE of 70, prevalence is predicted to be low at low CE values, and prevalence
approaching the maximum does not occur until CE values reach relatively high levels (in the
range of 50 fibers/cc-yrs). Also note, even though TSFE does not act directly on the slope of the
exposure-response curve, because the plateau increases as TSFE increases (see Panel A), the
initial slope also increases as TSFE increases.
E-21
-------
Panel A: Observed vs. predicted as a function of cumulative exposure (CE)
30
CE (f/cc-yrs)
40
50
•Fit(T=10)
•Fit(T=25)
•Fit(T=40)
• Fit(T=70)
BMR
BMDL70
Data (T = 10)
Data (T=25)
Data (T=40)
60
Panel B: Observed vs. predicted as a function of time from first exposure (TSFE)
i.o
0.9 -
0.8 -
0.7 -
0.6 -
0.5 -
0.4 -
0.3 -
0.2 -
0.1 -
0.0
-Fit (CE = 0.32)
•Fit (CE = 1.31)
•Fit (CE = 50)
Data (CE = 0.32)
Data (CE = 1.31)
Data (CE = 50)
10
20
30 40
TSFE (years)
50
60
70
Figure E-3. Graphical display of predicted vs. observed any pleural
thickening (APT) prevalence for cumulative normal Dichotomous Hill (CN
DH) model fit to the combined cohort.
E-22
-------
Figure E-4 shows analogous graphs for the BV DH FP (C, TSFE) model. Binning was
performed as above, except that workers were stratified by C rather than CE. As illustrated in
Panel A, for this model the dependence of prevalence on C as a function of TSFE (see Panel A)
is generally similar in shape to that for the CN DH model (see Figure E-3 Panel A), although in
this case, the plateau value of 0.85 would ultimately be reached for high values of C for all
values of TSFE. As shown in Panel B, extrapolating from the model to TSFE = 70 predicts that
prevalence will approach or exceed 0.8 for any exposure concentration C of 0.02 fiber/cc or
higher.
E-23
-------
Panel A: Observed vs. predicted as a function of mean exposure
concentration (C)
o.o
3.5
4.0
Panel B: Observed vs. predicted as a function of time from first exposure
(TSFE)
o.o
30 40
TSFE (yrs)
Figure E-4. Graphical display of predicted vs. observed any pleural
thickening (APT) prevalence for the Bivariate Dichotomous Hill model with
fixed plateau (BV DH FP) with exposure parameters of mean exposure
concentration (C) and time since first exposure (TSFE) fit to the combined
cohort.
E-24
-------
Based on a visual inspection of the agreement between observed and predicted
prevalence (compare Figure E-3 with Figure E-4, for TSFE =10, 25, and 40 years), the CN DH
(CE, TSFE) and the BV DH FP (C, TSFE) show adequate fits between the observed and
predicted response in the region of the BMR. Consequently, results for both models are
presented in the remainder of this appendix.
E.4. MODELING OF THE ROHS SUBCOHORT INFORMED BY MODELING OF THE
COMBINED COHORT
In the primary analysis described in Section 5.2.2.5, it was determined that the data from
the Rohs subcohort were not sufficient to provide a reliable estimate of the dependence on TSFE
in the B V DH FP model, so the effect of TSFE was estimated in a two-step procedure where the
dependence on TSFE was first determined from a fit of the BV DH FP model to a larger cohort
of the Marysville workers without restriction based on hiring date (n = 252, the Rohs cohort),
and then carrying the estimated effect of TSFE (the c parameter) over to the group of
119 workers hired in 1972 or later as evaluated in 2002-2005. Table E-9 summarizes the results
of applying this same strategy based on the CN DH model.
• Row 1 shows the results of an attempt to fit the CN DH model to the Rohs subcohort
using the values for m and s (the parameters which characterize the dependence of the
plateau on TSFE) derived from a fit of the combined cohort to the CN DH model. As
shown, a solution was found for the BMD at a value of TSFE = 70 years; however,
the corresponding BMDL could not be estimated for TSFE = 70 years. At
TSFE = 25 years both the BMD and BMDL were estimated.
• Row 2 shows the same approach, except that the background term was assigned a
fixed value of 0.03 rather than being treated as a fitting parameter. As shown, this
reduction in parameter number allowed estimation of the BMDL and BMCL at both
TSFE = 25 and 70 years.
• Row 3 is very similar to the approach presented in Section 5.2.2.5.2, fitting the BV
DH FP model to the Rohs subcohort using a two-step procedure. The only difference
is that the value of the c parameter shown in Table E-9 is based on a fit of the BV DH
FP model to the combined cohort (n = 434) rather than the Rohs cohort (n = 252)
used in the primary analysis.
E-25
-------
Table E-9. Modeling results for any pleural thickening (APT), applying parameters derived from modeling in the
combined cohort of Marysville workers evaluated in 1980 or in 2002-2005, to the subcohort of Marysville workers
evaluated in 2002-2005 and hired in 1972 or later
Model
CNDH
CNDH
BV DH FP
Exposure
metrics
CE,
TSFE
CE,
TSFE
C,
TSFE
bkg a b c M s Plateau3
0.063 -92.662 78.27 - 38.89 14.05 0.987
0.030 -0.986 1.890 - 38.89 14.05 0.987
0.038 -2.760 1.272 0.133 - 0.850
H-L
p AIC
0.122 75.65
0.602 75.85
0.718 75.55
BMD BMDL
(70) (70)
3.2 x 10° -b
5.3 x 10"1 5.9 x 10~2
-
BMC BMCL
(70) (70)
4.5 x 10~2 -b
7.6 x 10~3 8.4 x 10~4
1.2 x 10~3 3.4 x 1Q-7
BMD BMDL
(25) (25)
3.3 x 10° 8.8 x 1Q-1
2.2 x 10° 8.7 x lO"1
-
BMC BMCL
(25) (25)
1.3 x 1Q-1 3.5 x ID"2
8.7 x 10~2 3.5 x 10~2
1.3 x 1Q-1 5.5 x 1Q-2
Grey cells indicate fixed parameter values.
a Value for Plateau in CN DH model is for TSFE = 70 yrs.
bFit is unstable; value could not be estimated.
E-26
-------
Figure E-5 compares the dependence of the BMC and BMCL values on TSFE for the CN
DH model fit to the combined cohort (n = 434, red lines) to that for the BV DH FP model fit to
the Rohs subcohort (n = 119, blue lines) using the two-step approach described above. As
shown, the two models yield generally similar values for BMC and BMCL values at 25 years
and for BMC values at 70 years. However, BMCL values are widely divergent at 70 years.
l.E+OO i
l.E-01
l.E-02
l.E-C
l.E-C
l.E-05
l.E-06
BMC based on CN DH (CEJSFE)
BMCL based on CN DH (CEJSFE)
BMC based on BV DH FP (CJSFE)
BMCL based on BV DH FP (CJSFE)
20
30
40 50 60
Time since first exposure (yrs)
70
80
Figure E-5. Benchmark concentration (BMC) and lower limit of benchmark
concentration (BMCL) values as a function of time since first exposure
(TSFE) for two models: the cumulative normal Dichotomous Hill (CN DH)
model using cumulative exposure and TSFE fit to the combined cohort, and
the Bivariate Dichotomous Hill Fixed Plateau (BVF DH FP) model using
mean exposure concentration (C) and TSFE fit to the Rohs subcohort using a
two-step procedure.
E.5. SELECTION OF A POINT OF DEPARTURE (POD) TO DERIVE AN RFC FROM
THE COMBINED COHORT AND THE ROHS SUBCOHORT
As discussed in Section E.3, EPA evaluated the combined cohort by fitting 19 different
combinations of models and exposure metrics to each of 3 different endpoints, and calculated
BMCL values for each of 2 different values of TSFE (25 and 70 years). Based on a
consideration of the H-L goodness-of-fit statistic, the AIC values, the magnitude of the
difference between BMC and BMCL values, and a consideration of visual agreement between
E-27
-------
observed and predicted prevalence values, three different combinations of model and exposure
metrics were identified as being preferred as candidates for selection of the POD:
• BV DH FP (C, TSFE = 25 years)
• CN DH (CE,TSFE = 25 and 70 years)
Recognizing that results were generally similar across all three endpoints (APT, LPT, and ARC),
the results based on APT were identified as being preferred. For the CN DH (CE, TSFE) model,
values from both TSFE = 25 and TSFE = 70 were judged to be potentially useful, and were
retained. For the BV DH FP (C, TSFE) model, results at TSFE = 70 were judged to be
unreliable due to the wide difference between BMC and BMCL, and only the results from
TSFE = 25 were retained.
As discussed in Section E.4, EPA also evaluated the Rohs subcohort by fitting the CN
DH model to the APT data, using a two-step fitting procedure where the coefficient of the TSFE
term was first determined by fitting the combined cohort, and then retaining that coefficient as a
constant when the model was fit to the subcohort. Similar to the combined cohort, BMCL values
at both TSFE = 25 and TSFE = 70 were judged to be credible, and were retained.
Based on this approach, the BMCL values listed in Table E-10 were identified as
plausible PODs for derivation of the RfC.
Table E-10. Benchmark concentration (BMC) and lower limit on
benchmark concentration (BMCL) values for several alternative strategies
TSFE
25yrs
70yrs
Cohort
Combined cohort
Combined cohort
Rohs subcohort
Combined cohort
Rohs subcohort3
Model (parameters)
CN DH (CE, TSFE)
BV DH FP (C, TSFE)
CN DH (CE, TSFE)
CN DH (CE, TSFE)
CN DH (CE, TSFE)
BMC (f/cc)
6.0 x 1Q-2
1.5 x KT1
8.7 x IQ-2
4.7 x 10-3
7.6 x 10-3
BMCL (f/cc)
3.4 x ID"2
6.3 x 1Q-2
3.5 x 1Q-2
7.5 x 1Q-4
8.4 x 1Q-4
Background fixed at 0.03, see Table E-9.
E.6. DERIVATION OF AN RFC FROM THE COMBINED COHORT AND THE ROHS
SUBCOHORT
Following EPA practices and guidance (U.S. EPA, 2002b, 1994) as discussed in
Section 5.2.3, a composite uncertainty factor (UF) of 300 is used when deriving the RfC from the
POD calculated at the median TSFE (25 years ). This includes an uncertainty factor of 10 to
account for intraspecies variability (UFn = 10), a factor of three to account for database
E-28
-------
uncertainty (UFo = 3) and an extra factor of 10 to account for the lack of information on people
at risk for a full lifetime. When using the POD based on the BMCL calculated at
TSFE = 70 years, the additional adjustment factor of 10 is not necessary and a composite UF of
30 is used (UFn = 10 and UFo = 3). The calculations of the RfC for the combined cohort and the
Rohs subcohort using both options are shown in Table E-l 1. The RfCs are rounded to one
significant digit.
Table E-ll. Alternative reference concentration (RfC) values
Cohort
Combined cohort
Combined cohort
Rohs subcohort
Combined cohort
Rohs subcohort
Starting from
TSFE = 25 yrs
TSFE = 25 yrs
TSFE = 25 yrs
TSFE = 70 yrs
TSFE = 70 yrs
Mode (parameters)
CN DH (CE,TSFE)
BV DH FP (C, TSFE)
CN DH (CE,TSFE)
CN DH (CE,TSFE)
CN DH (CE,TSFE)
Calculation
RfC = (3.4 x 10-2)/300 = 1 x 1Q-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 10~5 fibers/cc
RfC = (8.4 x 10-4)/30 = 3 x IQ-5 fibers/cc
For comparison, the above values all fall within approximately threefold when compared
to the primary RfC derived in Section 5 of 9 x 1CT5 fibers/cc.
E.7. SENSITIVITY ANALYSIS
EPA conducted a sensitivity analysis regarding choices of several alternative cohorts,
alternative endpoints, alternative exposure metrics, and alternative model fitting strategies. The
alternative cohorts included the combined cohort and the Rohs cohort.
The alternative endpoints included LPT, APT, or ARC including the total number of
individuals at risk for APT in the combined cohort (434 individuals) and the individuals with
APT and with exclusion of the 3 individuals with interstitial opacities only (431 individuals).
The alternative exposure metrics included the total CE for each worker and the CE with
lags of 5, 10, and 15 years. For the combined cohort using the CN DH model, there was no
variation in the POD as a function of lag time (these results are not presented). Another
alternative CE metric was constructed by setting all exposure to zero after 1980. This was done
because the Marysville facility discontinued use of Libby ore in 1980. Thus, exposure after 1980
included fibers from South Carolina ore, Virginia ore, Palabora ore, and perhaps residual fibers
from Libby ore remaining in the facility.
The results of the sensitivity analysis are summarized in Table E-12. For those results
with a narrow range in the interval between the BMC and the BMCL, this analysis shows a fairly
consistent POD (BMCLio at TSFE of both 25 and 70 years).
E-29
-------
Table E-12. Summary of sensitivity analysis using the cumulative normal
Dichotomous Hill (CN DH) model using cumulative exposure (CE) and time
since first exposure (TSFE) as explanatory variables
Cohort
Combined cohort (434)
Combined cohort, less those with interstitial opacity only (43 1)
Combined cohort, exposure after 1980 = 0 (434)
Rohs cohort (252)
Endpoint
LPT (70)
APT (73)
ARC (76)
APT (73)
APT (73)
LPT (66)
APT (69)
ARC (71)
BMCL (f/cc)
TSFE = 25
3.5 x 1Q-2
3.4 x 1Q-2
3.3 x 1Q-2
3.4 x 1Q-2
1.2 x 1Q-2
2.4 x 1Q-2
2.4 x 1Q-2
2.5 x 1Q-2
TSFE = 70
4.0 x !Q-4a
7.5 x IQ-4
7.7 x 1Q-4
7.1 x 1Q-4
3.8 x lQ-6a
6.1 x 1Q-4
1.0 X 1Q-3
1.1 X 1Q-3
aResult is considered less reliable because of wide interval between the BMC and the BMCL (BMC:BMCL ratio
>10 and >200, respectively).
To further evaluate the performance of the exposure-response modeling, the CN DH and
the BV DH FP models were used to calculate the number of APT cases that would be predicted
in the three cohorts using both the CN DH and BV DH FP models. The results are summarized
in Table E-13.
Table E-13. Observed and predicted numbers of any pleural thickening
(APT) when modeling in various subsets of the Marysville workers
Cohort
Combined cohort
Rohs subcohort
Rohs cohort
Hire
date
Any
>1972
Any
X-ray date
1980
x
2002-2005
x
x
X
N
434
119
252
APT cases
Observed
73
13
66
APT cases predicted
CNDH
(CE, TSFE)
One
step
Two
step3
72.6
12.9
68.8
10.8
68.7
BV DH FP
(C, TSFE)
One
step
Two
stepb
73.7
12.9
69.5
12.9
70.0
*m = 38.89, s = 14.055.
bc = 0.1333.
E-30
-------
E.8. REFERENCES
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.20Q310-1436ST
Burnham. KP: Anderson. DR. (2002). Model selection and multimodel inference: a practical information-theoretic
approach (2nd ed.). New York: Springer.
http://www.springer.co m/statistics/statistical+theorv+and+methods/book/978-0-387-95364-9
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.
Jakobsson. K: Stromberg. U: Albin. M: Welinder. H: Hagmar. L. (1995). Radiological changes in asbestos cement
workers. Occup Environ Med 52: 20-27.
Jarvhorm. B. (1992). Pleura! plaques and exposure to asbestos: a mathematical model. Int J Epidemiol 21: 1180-
1184.
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
pleura! fibrosis. Am J Ind Med 20: 1-15. http://dx.doi.org/10.1002/aiim.4700200102
Lockev. J. (2013). Personal communication to EPA from Dr. James Lockey, July 2, 2013. Cincinnati, OH: Lockey,
J.
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
Respir Dis 129: 952-958.
Paris. C: Martin. A: Letourneux. M: Wild. P. (2008). Modelling prevalence and incidence of fibrosis and pleura!
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
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
Sichletidis. L: Chloros. D: Chatzidimitriou. N: Tsiotsios. I: Spyratos. D: Patakas. D. (2006). Diachronic study of
pleura! plaques in rural population with environmental exposure to asbestos. Am J Ind Med 49: 634-641.
http://dx.doi.org/10.1002/aiim.20334
U.S. EPA (U.S. Environmental Protection Agency). (1994). 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). (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). (2012). Benchmark dose technical guidance. (EPA/100/R-
12/001). Washington, DC: Risk Assessment Forum.
http://www.epa.gov/raf/publications/pdfs/benchmark dose guidance.pdf
E-31
-------
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
E-32
-------
APPENDIX F.
WORKER OCCUPATIONAL EXPOSURE RECONSTRUCTION FOR THE
MARYSVILLE COHORT
Prepared by:
James E. Lockey, MD, MS
Carol Rice, PhD
Linda Levin, PhD
Eric Borton, MS
Timothy Hilbert, MS
Grace LeMasters, PhD
University of Cincinnati
Department of Environmental Health
Cincinnati, OH
Bob Benson, PhD
David Berry, PhD
U.S. EPA Region 8
Denver, CO
With technical support from:
William Brattin, PhD
SRC, Inc.
Denver, CO
-------
F.I. INTRODUCTION
This appendix presents the data and methods used to reconstruct fiber exposure levels for
workers at the O.M. Scott facility in Maysville, Ohio. It builds on the previous work of Dr.
James Lockey and coworkers who investigated possible effects of exposures to dust containing
Libby Amphibole asbestos (LAA) at the Marysville plant (Rohs et al., 2008; Lockey et al.,
1984).
The data used in the original exposure reconstruction, and as reported in the published
manuscripts, were based on the exposure measurements available at that time (Lockey et al.,
1984). The current exposure reconstruction is based on approximately three times as many
measurements as utilized in 1980 (899 vs. 325). These exposure measurements were obtained by
the U.S. Environmental Protection Agency (EPA), from O.M. Scott, and through trial documents
from the United States of America versus W.R. Grace et al., as well as the archived data used in
the 1980 exposure reconstruction.
F.2. DESCRIPTION OF THE EXPOSURE SETTING
Beginning in 1957 and continuing until 2000, the plant in Marysville manufactured a
number of lawn care products including fertilizers and pesticides that were bound to a
vermiculite carrier as a delivery vehicle. This is of potential concern because some types of
vermiculite ore contain asbestos fibers, and processing the vermiculite ore in the workplace
could have led to release of asbestos fibers to air and inhalation exposure of workers.
F.2.1. Vermiculite Ore Sources
Initially (1957-1958), vermiculite ore was obtained only from Enoree, South Carolina.
Beginning in 1959, vermiculite ores from both Libby, Montana and Enoree were used. At first,
Libby vermiculite ore was only about one-third of the total vermiculite used, but the fraction
from Libby increased from 1964 to 1972, such that by 1972 Libby was the predominant source
(>95%). Libby vermiculite ore continued as the predominant source until 1980, when its use
was discontinued (Borton etal., 2012). Other sources of vermiculite ore used at the plant
included Palabora, South Africa (first used in 1970) and Louisa County, Virginia (first used in
1979). In 2000, the company developed a new process and vermiculite usage ended.
This variation in vermiculite ore source is significant because different types of
vermiculite ores have varying amounts and types of asbestos content (see Appendix C). Of the
vermiculite ores used at the Marysville facility, the highest asbestos fiber content is observed for
LAA in Libby vermiculite ore, with lower levels of actinolite and anthophyllite in South
Carolina vermiculite ore, and very low levels of actinolite, tremolite, and chrysotile in South
African vermiculite ore and tremolite in Virginia vermiculite ores. Consequently, depending on
the time frame when workers were employed in the Marysville facility, workers may have been
exposed to a mixture of fiber types. Because fiber concentrations in air were measured using
F-2
-------
phase contrast microscopy, which does not distinguish fiber types, exposure metrics derived
from the measurements include all airborne fibers in the work area.
F.2.2. Qualitative Information Sources
Information on workplace activities and processes involving vermiculite was obtained
from multiple sources. First, O.M. Scott provided report that included information about the
plant, including maps of the plant layout prior to 1980. Second, archived files from Lockey et al.
(1984) were identified. Third, as a result of the recent W.R. Grace trial, additional material
relevant to the O.M. Scott plant was discovered. The Department of Justice (DOJ) was contacted
for the release of these data. Seven 4-inch binders were available for review and every page
(approximately 3,150 pages) was reviewed to identify information relevant to the current project.
Aspects of particular interest included the manufacturing process, usage and source of raw
materials, engineering and design changes in the plant, work practices, and exposure assessment
methodology. Approval was received from the DOJ to use the relevant data for this project.
Written reports, letters, memos, and notes contained background information on plant operations.
A total of 1,489 pages were read for potentially useful and pertinent information and abstracted
into a data file. From these records, the following information was obtained:
• Plant layout, including changes over time. This allowed the association of the
descriptions used on air sampling data forms/reports with jobs or departments within
the plant. A limited number of aerial images were available to identify major
structures.
• Process descriptions, including workers per shift, workers per department, sources of
raw materials, and raw material volume in number of railroad cars received, tonnage
of railroad cars from Libby and South Carolina, and tonnage of unexpanded
vermiculite received.
• A list of job titles and tasks for each department.
Lastly, two focus group discussions were conducted with workers who had been
employed at the plant in the 1957-1980 time frame (Borton et al., 2012). Gaps in understanding
were filled with information gathered from the focus groups, specifically regarding:
• Plant layout and changes over time, including engineering controls,
• Historical pattern of job rotations within department from 1957 to 1980,
• Time spent in work locations at the plant site,
• Overtime associated with departments and season, and
• Use/nonuse of respirators.
F-3
-------
F.2.3. Vermiculite Processing
Vermiculite was processed at the plant in the trionizing department. Trionizing is a term
used in the Marysville, OH facility and includes all operations where bulk vermiculite ore was
handled or processed. Raw vermiculite ore was delivered in railcars and unloaded outside into
hoppers for storage before being fed into an expander furnace. After expansion, a cyclone
separated the expanded vermiculite from other material before the vermiculite was dried,
crushed, and sized by screening. The expanded vermiculite was mixed with additives to form
the final product for lawn treatment (Lockev, 1985).
Because the potential for exposure to fibers released from vermiculite to air depended on
the type of activity being performed, exposure measurements in the trionizing department were
first assigned to each of the jobs, as follows:
• Track
• Blender
• Cleanup14
• Dryer
• Expander
• Feeder
• Mill
• Resin
The track job was further divided into track unload (exposures associated with the actual
unloading of vermiculite from railcars) and track other (exposures that occurred while working
in the railcar unloading area at times when unloading was not occurring).
F.2.4. Exposure Controls in the Trionizing Department
A number of exposure reduction efforts in the vermiculite expander operation have been
documented from archived files from the original Lockey study, focus groups, and material
released by the DOJ from the W.R. Grace trial. The first major engineering control was the
installation of a central vacuum system in 1961. Dust collectors were installed and improved
ventilation was initiated in 1968. Additional improvements, such as adding hoods and a bag
house to remove dust from the stoner deck exhaust and enclosing vibrating conveyers, were
implemented in 1970-1973. A more comprehensive and integrated approach to dust control
took place approximately in 1975/1976-1980. A number of engineering controls and work
14Since the initial 1980 study, cleanup has been recognized as one of the tasks through which the indoor trionizing
workers rotated. However, no industrial hygiene samples unique to cleanup were initially available and cleanup was
previously given the mean value of the other industrial hygiene measurements (Lockey. 1985). The newly available
measurements included samples specified as cleanup and these were assigned to the cleanup activity.
F-4
-------
practices were added during these years. In 1976, a major construction change isolated track
unloading activities from the production areas, reducing transfer of particulates into the plant
during raw material transfer (OSHA, 1979). Additional engineering controls included the
installation of more roof fans and dust collectors. Work practices emphasized vacuuming rather
than dry sweeping and improved sealing of leaks in the vermiculite expanders. During this time
period, routine weekly checks for leaks by maintenance personnel began. In 1980, wet scrubbers
were added to clean the air from areas not served by the bag house.
F.2.5. Respiratory Protective Equipment and Clothing Change Considerations
Respirator usage was inadequate (OSHA, 1979). Respirators were used only sporadically
due to heat in the production area and discomfort during use. Paper masks were preferred by
workers and were often reused from day to day. There was no documentation of fit testing of the
paper masks. Paper masks can provide some protection against the larger particles, but likely
provided little reduction in respirable particles, particularly when reused. Therefore, no
adjustment was made to lower the exposure estimates due to respirator use.
Per focus groups, workers were provided paid work time for required showers at the
facility after each production shift beginning in 1961-1962. Work coveralls were laundered
on-site after each work shift starting in approximately 1966. Street clothes were stored during
the work shift in locker rooms separated from the production area (Borton et al., 2012).
Consequently, off-site exposures to work-related fibers were not likely to have been significant.
F.2.6. Other Departments in the Facility
Workers in other departments in the plant where only expanded vermiculite or no
vermiculite was used were defined as having "plant background" exposure. These included the
following (Borton et al., 2012):
• Polyform.
• Office.
• Research lab.
• Pilot plant.
• Warehouse.
• Packaging.
The polyform process started in 1969 and was separate from any vermiculite operations
(Borton et al., 2012). Other departments included central maintenance and plant maintenance.
Workers in these departments spent part of their time in the trionizing area and part of their time
in jobs in areas categorized as plant background. The central maintenance department became a
contract service in 1983, and after this date most workers in central maintenance were not
F-5
-------
employees of O.M. Scott. However, some O.M. Scott employees continued to work in central
maintenance after 1983.
F.3. INDUSTRIAL HYGIENE DATA SOURCES
Three sources of industrial hygiene (IH) measurements of fiber concentrations in
workplace air were identified: sampling reports from O.M. Scott that included measurements at
the facility from 1972 to 1994, archived files from the Lockey et al. (1984) study, and the W.R.
Grace trial discovery material.
F.3.1. Document Evaluation, Data Entry, Cleaning, Editing, and Standardization
Air sampling reports included quantitative measurement of airborne dust and fiber
concentration associated with a department job. These records were computerized following an
approved data entry scheme. Records were double entered and verified.
Two identical Microsoft Access databases were created for initial and duplicate entry of
the quantitative data. Each individual performing data entry had a unique and separate database
to avoid possible data entry confusion. A random 10% check of entered data was conducted
throughout the data entry process to maintain quality of data, to address data entry questions and
to resolve potential database issues. Data entry differences were below 5% throughout the entry
process.
A final verification of data entry used SAS Version 9.2 PROC COMPARE to import the
initial and duplicate Access tables. All discrepancies were addressed by reviewing the original
document. The initial and duplicate Microsoft Access databases were archived. A copy of the
initial database was converted to Microsoft Excel format for standardization and ease of
analyses.
F.3.2. Process of Standardization
The standardization process included categorizing entered data into appropriate variable
fields, spell checking, identifying duplicate record entry from duplicate documents, merging
records for the same sample or measurement, evaluating data for completeness, and categorizing
groups of data based on type of sample or measurement.
Data were reviewed and edited to ensure the information was entered into the appropriate
data field. A frequency of the data fields using SAS 9.2 PROC FREQ identified spelling
differences and patterns to ensure correct labeling of the data. Additional data variables were
created depending on recognized need to distinguish important pieces of data.
A new variable called group ID was created to identify, track, and consolidate partial
and/or complete duplicate data into one unique sample. Partial data were identified on a
combination of sample date, sample record ID, sample result, volume, sampling time, and/or
document patterns. A document pattern would include instances where only a group of sample
F-6
-------
results were available in one document and another document(s) would match the exact sequence
of sample results.
Data were further categorized based on the type of sample. Categories include dust
samples, bulk samples, personal and area fiber samples, limit of detection (LOD) or
quantification (LOQ) samples, off-site locations, and time-weighted average (TWA) samples.
Some samples were collected with a direct-reading fibrous aerosol monitor, but these were not
used because no calibration information was included in the records. Thus, only the fiber count
data collected with a sampling pump were used. In addition, group IDs lacking a sample result,
sample year, or department were excluded.
The natural logs of personal and area samples were evaluated by year and department.
The ranges and means of the personal and area samples were approximately equal. When plotted
by year and department, the data were seen to be in the same range, with the values overlapping.
Therefore, personal and area sample data sets were merged and both were used for the
development of the Exposure Matrix. Group IDs with only LOD or LOQ values were grouped
by year and categorized as trionize or background. In order to assign an estimate for the LOD or
LOQ, the median value of each group was divided by two and assigned to all samples in that
group. Given the small number of LOD and LOQ samples (n = 35), it is unlikely any significant
bias was introduced using this method. TWA values were not used when the individual
measurements that comprised the TWA were already available.
Attempts in other studies to convert from total dust to fiber count have relied on
similarities in equipment or process where side-by-side samples were collected. However, no
side-by-side matched pairs of dust/fiber data were identified from this plant. Therefore, total
dust measurements were not converted to fiber counts and were not used as part of the fiber
exposure estimation.
F.4. OVERVIEW OF THE EXPOSURE DATA
F.4.1. Sampling and Analysis Methods
FALL Sampling
Collection of IH air samples to determine worker exposure to fibers started in 1972.
Samples were obtained by drawing air through a filter to capture airborne fibers. Initially,
samples were collected either the industrial hygienist carrying the sampler and "following the
worker" or by placing the sampler at a stationary location. Personal sampling began in 1976 by
using a pump and filter cassette worn by the worker.
No corporate plan for air sampling was found in the available documents. Air sampling
practices were discussed with the focus group participants who noted some instances of leaving
sampling pumps in control rooms during high dust activities such as the use of compressed air to
remove particulates from surface areas. This activity was not uniformly omitted from air
sampling results; however, there was no documentation that high-exposure work was excluded
F-7
-------
from the sampling efforts. In fact, in the early years, some activities recorded in the sampling
record included reference to compressed air "blow down," one of the activities associated with
potentially high exposures. Consequently, all sample results were considered representative of
conditions during collection and were included in the data set.
F.4.1.2. Analysis
Air filter samples were analyzed by a microscopist using phase contrast microscopy
(PCM) and the results were expressed as PCM fibers per cubic centimeters (fibers/cc) of air
(Bortonetal.. 2012). Fiber counting followed the NIOSH P&CAM 239 and 7400 counting
methods. In these methods, a countable fiber is defined as an elongated particle with a length
greater than 5 jim, a diameter less than 3 jim, and an aspect ratio (length:diameter) of 3:1 or
greater. This microscopic technique provides no information on the chemical or crystal structure
of elongated particles; therefore, the PCM fiber counts represent all elongated particles fitting the
counting criteria.
F.4.2. Summary Statistics
Table F-l shows a total of 899 IH samples were available for this analysis. Most (81%)
were collected in the trionizing departments where exposure to vermiculite and fibers tended to
be highest, and 19% of the measurements came from other (background) locations in the plant.
F-8
-------
Table F-l. Industrial hygiene fiber measurements by
document source
Document source
DOJ
EPA
UC
MULTIPLE a
Total (%)
Trionize
23
398
135
172
728 (81)
Background
0
122
45
4
171(19)
Total (%)
23 (2.6)
520 (57.8)
180 (20.0 )
176 (19.6)
899 (100)
a Results listed in two or more sources with duplicates removed
Table F-2 shows the number of samples stratified by year and by job. As shown, the first
fiber count measurements were available in 1972 and the last in 1994. The frequency of sample
collection was not uniform over time, with the highest numbers of samples being collected in
1976 and 1978.
F.4.3. Data Review and Assessment
Figure F-l provides a graphical display of the IH data from the trionizing department
plotted as a function of time. Note that the concentration scales are not the same in all panels.
Highest concentrations tended to occur during track unload, feeder, and expander jobs. Exposure
levels in most trionizing jobs showed a general tendency to decrease over time as engineering
controls improved and as Libby vermiculite use was discontinued.
Figure F-2 shows a graphical summary of data from nontrionizing (background)
departments and jobs. In this case, there are no clear distinctions among departments or jobs, so
the data are shown without stratification. One data point (a value of 4.03 fibers/cc that was
identified as having been collected in the lab) was identified as an outlier because it was
substantially higher than any other value in the background data set. This value is not considered
to be representative of exposures in background jobs, and was excluded from all further
evaluations. As indicated in the figure, although less dramatic than for the trionizing department,
there is also an apparent tendency for background exposure levels to decrease over time.
F-9
-------
Table F-2. Industrial hygiene fiber measurements by department and year
Category
Trionize
(indoor jobs)
Trionize
(outdoor jobs)
Background
All
Job
Blender
Cleanup
Dryer
Expander
Feeder
Mill
Resin
Total indoor
Track other
Track unload
Total outdoor
Cafeteria
Central maint.
Control
Research Lab
Office
Packaging
Plant maint.
Polyform maint.
Polyform
Poly packaging
Warehouse
Total
Grand total
1972
1
8
9
1
2
3
12
1973
1
38
39
1
1
40
1975
1
18
19
1
1
1
1
2
22
1976
15
83
10
1
109
6
6
115
1977
3
6
2
6
1
2
20
6
27
33
4
1
5
10
63
1978
21
26
2
51
12
22
11
145
23
15
38
15
28
1
9
1
54
237
1979
2
7
13
1
23
3
3
2
2
28
1980
5
10
1
1
17
4
2
6
23
1981
3
1
3
12
3
3
4
29
6
3
9
3
3
3
3
12
50
1982
1
6
3
2
3
4
19
2
3
5
3
3
1
7
31
1983
3
3
3
2
5
3
3
11
1984
1
2
3
5
4
15
6
6
12
3
2
6
11
38
1985
3
2
5
1
8
9
3
9
5
19
1986
10
11
7
28
18
6
24
1
3
2
2
6
6
3
23
75
1987
7
6
7
3
23
10
10
1
1
1
2
4
1
1
2
13
46
1988
10
6
1
4
21
9
1
10
1
5
6
4
16
47
1993
1
1
2
2
3
1994
7
3
7
17
12
12
1
9
10
39
Total
27
56
46
276
32
77
28
542
102
84
186
2
4
30
4
5
75
24
2
1
9
15
171
899
F-10
-------
I»
! -
i -
!:
*1 fi
. »i if r.
, V ,-.•.,,,, ... I -•..,.,,
—,—_*,»-*
i •> v -^ -. •.«. i <•-•._
! =
.
i
J •
•
TnrtH|CWW|
viva M»t aarn nm «BMi« HM. imM MM> v*w tauti
Figure F-l. Trionizing department data by year and job.
F-ll
-------
Jt C
Jin
I 30 -
£, 2 5 -
s
2 "> n -
c
o
i i-5 ~
0 5
6/13
(-?$
^~^ Outlier
I * d^* t fti ta
./68 3/8/71 12/2/73 8/28/76 5/25/79 2/1S/82 11/14/84 S/ll/87 5/7/90 1/3 1/33 10/28/95
Sampling Date
Figure F-2. Background data by year.
Note: Outlier is a data point collected from the research lab and is not considered representative of
background exposure; it was excluded from evaluation.
Figure F-3 plots observed concentrations as a function of sampling duration (the length of
time over which air was drawn through the filter). As seen, there is a clear tendency for samples
with the highest concentrations to have the shortest sampling durations, especially for track
unload and other trionizing jobs. This finding is expected because high concentrations of fibers
in this work process generally occur when overall particulate levels are high. The PCM
analytical method requires that the microscopist be able to visualize the fibers for counting, and
this cannot occur if the overall loading of the filter obscures elongated particles. Therefore,
sampling in high dust conditions must be for a short time interval (often 15 minutes or less) to
prevent overloading of the filter. If overloading occurs, the sample is void, and marked
"overloaded."
F-12
-------
2.0
1.8
1.6
1.4
1-2
1.0
0.8
0.6
Bkg
100
200 300 400
Sampling Duration (min)
6C3
Track (other)
100 200 300 400 500
Sarrpling Duration (min)
600
700
800
Trionize
Track Unload
300
250
50
D
100
200
300 400 500
Sampling Duration (ruin)
600
700
800
100 200 300
Sampling Duration (min)
400
500
Figure F-3. Relation between sampling duration and measured concentration.
F-13
-------
Short-duration samples may represent actual conditions in the workplace for a specific
job overall or for a short-term operation in a job. In the first instance, the sample result
represents the full duration of the job; in the second, the sample result would be time-weighted as
part of a job. No information was available to indicate worker exposure duration was related to
either sampling duration or exposure concentration. Consequently, all measurements were used
without any adjustments based on sampling duration.
F.5. DEVELOPMENT OF THE JOB-EXPOSURE MATRIX
F.5.1. General Strategy
A job-exposure matrix (JEM) is a table that provides estimated exposure levels in air
(fibers/cc) for workers in each job for each year. The exposure interval of interest for the
Marysville worker cohort begins in 1957 when vermiculite was first used in the plant and
extends to 2000 when vermiculite usage ended. Because measurements of fibers in the air are
available only for the central portion (1972-1994) of the exposure interval of interest
(1957-2000), the JEM was constructed in two steps:
Step 1: Industrial hygiene data collected between 1972 and 1994 were used to derive
estimates of yearly average concentrations by job during this interval. Exposure
levels in 1994 that were derived from industrial hygiene data were assumed to
remain constant until 2000.
Step 2: Information available from plant records and worker focus groups was used to
estimate concentrations from 1957 to 1971 by extrapolation from 1972 values.
Two alternative strategies were used to construct JEMs. The first strategy, implemented
by UC, was based on the log-transformed data, and the exposure metric provided in the JEM was
the geometric mean exposure concentration (Borton et al., 2012). This approach was used
because the probability of response is expected to be a nonlinear function of exposure, and use of
the log-transformed values helps minimize the effect of measurement error on the regression
model (Seixas et al., 1988). The second approach, implemented by EPA working in consultation
with UC, utilized the untransformed data, and the exposure metric provided in the JEM was the
arithmetic mean exposure concentration. This approach was used because toxicity values
derived by EPA are typically based on the long-term average exposure level rather than the
geometric mean exposure level (U.S. EPA, 1994). The details of these two approaches are
provided below.
F-14
-------
F.5.2. Derivation of a Job-Exposure Matrix (JEM) Based on Log-Transformed Data
F.5.2.1. Trionizing Department 1972-2000
The trionizing department included jobs from the entry of vermiculite into the plant
through final product. Jobs included track, screen/mill, feeder, dryer, expander, blender, resin,
and cleanup. Workers rotated through the various jobs within the department. Overall rotation
among jobs reported in the 1980 Lockey study (Lockev, 1985) was verified by focus groups.
As seen in Table F-2, the frequency of sample collection was sparse in many years,
limiting the calculation of a mean exposure level for each indoor trionizing job for each year.
This issue is particularly evident in the early years, as 147 of the 176 measurements in
1972-1976 are from the expander, with the remaining as follows: cleanup (16 measurements),
feeder (10), dryer (2), mill (1), blender (0), and resin (0).
Plots of the log-transformed IH measurements over time were made for individual
trionizing jobs. All samples that were below the level of detection (n = 35) were assigned
one-half the median of the limit of detection or limit of quantitation for the corresponding
department-year. Only the plot of expander data, representing 51% of all indoor trionizing
measurements, spanned the time frame of interest. Plots for the six nonexpander jobs, at the
dates available were generally consistent with the expander data plots. All of the indoor
trionizing jobs were in the same building where engineering controls in one area would likely
influence exposures both at the job where the control was implemented and also at nearby work
locations. Moreover, workers reported equal time spent in the various indoor jobs. Therefore, in
order to leverage the available data, it was determined that the exposure measurements for indoor
trionizing jobs should be combined. The outdoor track job included two very different work
activities: unloading railcars containing vermiculite (track unload) and general track work such
as bringing in the railcars and monitoring discharge (track other). The two track job activities
(unload and other) had a substantially larger range of sampling results and were treated
separately.
In accordance with this strategy, the following steps were implemented to derive the
geometric-mean-based JEM for the trionizing department from 1972 to 2000:
1. The data were log transformed.
2. A curve was drawn through the data set for all indoor trionizing jobs to estimate
annual log-mean values. Figure F-4 illustrates this curve. As values for 1980-1994
were similar and near the level of detection, the log-mean value for all the samples
was used and then extended until 2000. For all exposure values for the combined
indoor trionizing jobs from 1973-1978, a smooth-fitted curve was drawn using
Microsoft Excel to connect the log-mean values of "index years" (1973, 1976, and
1978) having a substantial number of exposure measurements (approximately 40 or
more). This approach was chosen to assure that stable log-mean values were used to
define the curve over this time period. The log-mean value for 1977 JJi
F-15
-------
measurements naturally fell on the curve between 1976 and 1978. Therefore, for the
1972-1979 time period, log-mean values for only 4 years (1972, 1974, 1975, 1979)
were lacking. The line connecting 1976 backward through 1973 provided values for
1974 and 1975 and the continuation of this line provided the value for 1972.
Connecting 1978 (index year) to 1980 provided the value for 1979. For each year,
the annual geometric mean exposure estimate was determined by exponentiation of
the log-mean value from the curve. The decline seen in exposures throughout the
1976-1978 time period is consistent with reports of implementing engineering
controls such as dust collection, enclosing vibrating conveyors, adding ventilators,
erecting a wall between the railroad track and the main building, and sealing leaks in
the system.
3. The log-transformed measurement results for track unload and track other were
plotted and a straight line produced to best fit the data points. The geometric mean
exposure for each year was determined by exponentiation of the value on the line for
that year.
4. For the trionizing department, it was estimated that 11% of work time was spent in
track and 89% in all other jobs. This is consistent with the previous weights used in
the 1980 Lockey study (Lockey, 1985) and confirmed by the focus groups.
5. The focus groups reported that when working track, track unload required about 25%
of the time and track other comprised about 75% of the track job time. Therefore, a
weighted average for exposure at track within the trionizing department was derived.
This 25% time estimate for track unload is higher than previously reported (Lockey,
1985).
F-16
-------
i Expander measurement
" result excluding Track
Fitted Curve
Annual mean of natural
log of measurements
Mean of natural log of
measurements with
40 or more values
(1973, 1976,1978)
Figure F-4. Natural logarithm of all usable industrial hygiene measurements across all indoor jobs within the
trionizing department, and the fitted line (red) used to represent the geometric mean.
F-17
-------
F.5.2.2. Trionizing Department 1957-1971
Estimation of exposure values in the trionizing department before 1972 (prior to exposure
measurements) required consideration of two factors: (1) changes in dust levels over time due to
the effects of dust control measures in the department and (2) changes in the vermiculite source
material used.
F.5.2.2.1. Adjustment for changing indoor dust levels. As noted above, a graphical display of
IH concentration values for indoor trionizing jobs indicated that all samples generally followed
the same pattern: higher in the early years of industrial hygiene sampling and declining
gradually over time. Further, the focus groups reported that no single engineering change
resulted in a dramatic reduction in the perception of dustiness in the plant. Thus, the workers'
recollections supported the findings from the industrial hygiene data demonstrating a smooth
decline in levels of exposure rather than a dramatic stepwise drop due to any one engineering
change.
Focus group participants who had worked in the trionizing department before 1972
reported that dust exposures in indoor trionizing jobs were at least two times higher in the 1960s
than in the 1970s. Therefore, the year 1972 was used as the start of the "gradual" retrospective
increase in exposure back to 1967 as 1972 was the first year when industrial hygiene
measurements were available, and the percentage of Libby vermiculite used was 93%. The year
1967 was selected because it was the year preceding engineering controls. Accordingly, a line
was drawn to connect these two points (see Figure F-4). Before 1967, estimates for fiber
exposure levels were extended backward in time, assuming no change in dust levels
retrospectively from 1967.
In contrast to the indoor trionizing jobs, the track unload and track other jobs were
outdoors and were likely unaffected by indoor plant engineering controls. Hence, estimates for
fiber exposure levels for track duties were not adjusted for a time-dependent change in dust
levels.
F.5.2.2.2. Adjustments for vermiculite raw material sources. Two primary sources of
information were located regarding vermiculite ore sources in the 1957-1972 time frame:
• An archived UC document from the original site investigation with estimates of
railroad car loads delivered to the plant per year. Documents indicate railroad cars
from Libby were 100-ton cars and from South Carolina 70-ton cars.
• The Chamberlain memo (internal O.M. Scott memo) provides information regarding
vermiculite ore sources for 1964-1972 in railroad car loads per year.
F-18
-------
Per the UC document, 100% of the vermiculite ore estimated to be used from 1957-1959
was from South Carolina. Per the Chamberlain memo, it was best estimated that Libby
vermiculite ore began arriving in 1960. Focus groups held by UC investigators with a
cross-sectional representation of former O.M. Scott employees placed the first use of Libby
vermiculite ore earlier, in 1958 or 1959. In the absence of definitive documentation, UC used its
best professional judgment to assign the start date for the use of Libby vermiculite ore as 1959.
Documentation was found from the original 1980 UC documents indicating an estimated
Libby tonnage contribution of 32% from 1959-1963. These percentages for 1959-1963 were
adopted for use in this project. After adjusting for the difference in railcar sizes, the
Chamberlain memo indicates that Libby tonnage usage increased from 57% in 1964 to 73% in
1965 to 92% in 1966. Table F-3 summarizes the distribution of unexpanded vermiculite sources
received at the plant between 1957 and 1971.
Table F-3. Vermiculite tonnage by year and source
Year
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
% Tonnage Libby
32
32
32
32
32
57
73
92
87
79
82
90
95
% Tonnage SC
100
100
68
68
68
68
68
43
27
8
13
21
18
10
5
Comment
No confirmation of Libby usage
No confirmation of Libby usage
Libby usage began per focus groups; Chamberlain memo3
says 1960
Chamberlain memo and 1980 UC document
Chamberlain memo and 1980 UC document
Chamberlain memo and 1980 UC document
Chamberlain memo and 1980 UC document
Chamberlain memo
Chamberlain memo
Chamberlain memo
Chamberlain memo
Chamberlain memo
Chamberlain memo
Chamberlain memo
Chamberlain memo
"Internal O.M. Scott memo.
To develop the relationship of fiber levels between South Carolina and Libby
vermiculite, IH samples associated with either 100% Libby or 100% South Carolina vermiculite
were identified. Two jobs with the highest number of samples from the same year from each
source were used to establish the relationship. The data are summarized in Table F-4, below.
F-19
-------
Table F-4. Relative concentrations of fibers in Libby and South Carolina
vermiculites
Data Set
1977 Track unload
1978 Expander
Count-weighted mean
Libby vermiculite
Sample count
13
8
Mean (f/cc)
7.85
0.55
5.07
South Carolina vermiculite
Sample count
11
7
Mean (f/cc)
0.82
0.20
0.58
The ratio of the count-weighted average of these samples is (5.07/0.58) is 8.7:1, and this
ratio was used for estimating the proportion of Libby versus South Carolina fiber exposure levels
from 1959 to 1971.
F.5.2.3. Exposure Estimates for Nontrionizing Departments
As noted above, departments using only expanded vermiculite or no vermiculite were
defined as having "plant background" exposure. These included the departments of polyform,
office, research, pilot plant, warehouse, and packaging. This decision was based on plots of
available sampling data showing similar levels and qualitative reports documenting that no fibers
were in the finished product.
Plant background exposure concentrations before 1972 were estimated using similar
methodology as for the trionizing department. It was assumed that background levels were not
affected by engineering control as in trionizing, but were influenced by the percentage of Libby
vermiculite used. Therefore, for the years prior to 1972, the measured plant background rate in
1972 was adjusted only for the yearly percentage of Libby vermiculite used. The 2 years before
Libby vermiculite usage, 1956 and 1957, were assigned concentration values equal to the level of
detection (0.01 fiber/cc). This is in line with industrial hygiene measurements post Libby
vermiculite usage through 1994.
Background exposure estimates derived as described above were applied to workers in
polyform, office, research, pilot plant, warehouse, and packaging.
Because maintenance workers spent some time in the trionizing department as well as in
background areas, the values for these workers were adjusted as follows:
• Plant Maintenance—although there were some differences of opinion in the focus
groups regarding where plant maintenance spent their time, the consensus reached
was to assign approximately 50% of time in trionizing and 50% in areas defined as
plant background for their work in shop and other departments.
• Central Maintenance—according to the focus groups, these employees worked
outside of trionizing for about 90% time (background) and within trionizing for about
10% time for installation of new equipment/parts. Around 1982-1983, the central
F-20
-------
maintenance department was contracted to outside personnel, although some O.M.
Scott workers continued to work in central maintenance.
F.5.2.4. Results: Job-Exposure Matrix (JEM) Based on Geometric Mean Exposure Levels
Table F-5 presents the JEM from 1957 to 2000 using the methodology detailed above.
Exposure concentrations represent the geometric mean exposure level, by job and year.
Table F-5. Geometric mean-based job-exposure matrix (JEM) for
Marysville workers
Year
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
Trionizing
(all jobs)
0.801
0.801
2.874
2.874
2.874
2.874
2.874
4.493
5.530
6.76
6.437
5.557
5.291
4.928
4.318
3.674
3.007
2.464
2.019
Plant
maintenance"
0.406
0.406
1.441
1.441
1.441
1.441
1.441
2.253
2.772
3.389
3.227
2.786
2.654
2.473
2.169
1.847
1.513
1.242
1.020
Central
maintenance1"
0.089
0.089
0.295
0.295
0.295
0.295
0.295
0.460
0.567
0.693
0.660
0.570
0.544
0.509
0.449
0.385
0.319
0.264
0.220
Background0
0.010
0.010
0.008
0.008
0.008
0.008
0.008
0.012
0.015
0.019
0.018
0.016
0.017
0.018
0.019
0.020
0.020
0.020
0.020
F-21
-------
Table F-5. Geometric Mean based job-exposure matrix (JEM) for
Marysville workers (continued)
Year
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995-2000
Trionizing
(all jobs)
1.391
0.150
0.086
0.077
0.063
0.063
0.060
0.060
0.055
0.055
0.052
0.052
0.052
0.052
0.052
0.052
0.052
0.052
0.052
0.052
Plant
maintenance"
0.705
0.090
0.053
0.044
0.036
0.036
0.035
0.035
0.032
0.032
0.031
0.031
0.031
0.031
0.031
0.031
0.031
0.031
0.031
0.031
Central maintenance1"
0.157
0.030
0.027
0.017
0.015
0.015
0.015
0.015
0.014d
0.014
0.014
0.014
0.014
0.014
0.014
0.014
0.014
0.014
0.014
0.014
Background0
0.020
0.020
0.020
0.010
0.010
0.010
0.010
0.010
0.010
0.010
0.010
0.010
0.010
0.010
0.010
0.010
0.010
0.010
0.010
0.010
"Assumes exposure occurs 50% in trionizing and 50% in background departments.
bAssumes exposure occurs 10% in trionizing and 90% in background departments.
background includes pilot plant, research, polyform, office, packaging, and warehouse.
dAfter 1983, central maintenance was outsourced, but some O.M. Scott workers continued in that position.
F.5.3. Derivation of a Job-Exposure Matrix (JEM) Based on Untransformed Data
The basic approach used by EPA for deriving a JEM based on the untransformed data
was generally similar to that used for the log-transformed data, with the following exceptions:
• Nondetects were assigned a value of zero rather than the detection limit (Cameron
and Trivedi. 2013: U.S. EPA. 2008b: Haasetal.. 1999: U.S. EPA. 1999).
• The IH data were fit to statistical models to characterize time trends, rather than using
interpolation among data-rich years.
F-22
-------
• Indoor trionizing jobs were modeled individually rather than combined into one data
set.
The details of this approach are described below.
F.5.3.1. Fitting Available Industrial Hygiene Data from 1972-1994
F.5.3.1.1. Trionizing department data. Industrial hygiene data collected in the trionizing
department between 1972 and 1994 were classified as being associated with nine different types
of jobs (blender, cleanup, dryer, expander, feeder, mill, resin, track other, and track unload).
Table F-6 provides summary statistics for these trionizing jobs. All values are shown to two
significant figures.
F-23
-------
Table F-6. Summary statistics for trionizing jobs
Job
Blender
Cleanup
Dryer
Expander
Feeder
Mill
Resin
Track other
Track
unload
1972-1975
n
0
1
2
64
0
0
0
0
2
Mean
-
5.3 x 10°
1.2 x 10°
5.7 x 10°
-
-
-
-
3.5 x 10°
Max
-
5.3 x 10°
2.1 x 10°
5.9 x 101
-
-
-
-
5.2 x 10°
1976-1980
n
24
52
6
157
23
39
13
33
53
Mean
1.8 x icr1
7.5 x KT1
6.1 x 1Q-2
1.6 x 10°
6.0 x 10°
6.2 x KT1
7.1 x IQ-2
1.2 x KT1
1.7 x 101
Max
1.2 x 10°
1.1 x 101
1.8 x KT1
4.8 x 101
5.0 x 101
6.1 x 10°
1.9 x 1Q-1
1.5 x 10°
2.5 x 102
1981-1984
n
o
J
o
J
11
24
5
13
12
18
22
Mean
1.4 x 1Q-2
2.0 x 1Q-2
5.0 x 1Q-2
6.3 x 1Q-2
2.8 x 1Q-2
4.9 x 1Q-2
5.4 x 1Q-2
3.2 x 1Q-2
9.0 x 10°
Max
1.9 x 1Q-2
5.0 x 1Q-2
1.1 x KT1
2.3 x 1Q-1
1.0 x 1Q-1
1.0 x 1Q-1
1.7 x 1Q-1
1.3 x 1Q-1
3.6 x 101
1985-1990
n
0
0
27
23
1
18
3
37
7
Mean
-
-
2.1 x 1Q-2
3.7 x 1Q-2
8.0 x 10-3
4.2 x 1Q-2
5.7 x 10-3
6.2 x 1Q-2
1.1 x 10°
Max
-
-
9.0 x 1Q-2
8.5 x 1Q-2
8.0 x 1Q-3
3.6 x 1Q-1
1.0 x 1Q-2
1.5 x 10°
2.1 x 10°
1991-1994
n
0
0
0
8
3
7
0
14
0
Mean
-
-
-
5.6 x 1Q-2
6.9 x 1Q-2
6.8 x 1Q-2
-
6.0 x 1Q-2
~
Max
-
-
-
1.7 x 1Q-1
1.0 x 1Q-1
2.0 x 1Q-1
-
2.2 x 1Q-1
-
All concentration values are PCM fibers/cc.
F-24
-------
As indicated, mean exposure levels vary among jobs, and also tend to decrease over time.
Because the data are insufficient to calculate a reliable estimate of the arithmetic mean exposure
level for each job for each year, the data for each job were fit to a statistical model to
characterize the rate of change over time. Several different modeling approaches were
evaluated, as described below.
F.5.3.1.1.1. Fitting method 1: local regression (LOESS). To investigate the form of the
regression curve relating sample concentrations to date of sample, a flexible nonparametric
fitting method was applied, using data for each job. Analyses were implemented by the SAS
procedure PROC LOESS (SAS for Windows, Version 9.3). Linear functions of time were
sequentially fit to "windows" of concentration values within a chosen radius (time span) of each
concentration value. A smooth LOESS curve was then drawn through the fitted values. Fitting
was performed by weighted least squares. The same radius was applied to each window of
job-specific data. A "smoothing parameter" determined the radius of the fitting windows. The
optimum smoothing parameter was determined by a grid search to identify the value that
minimized the Akaike Information Criterion with Correction, a criteria for determining model fit.
These nonparametric plots generally reflect a decrease in exposure over time with a
steeper decline in the mid-1970s followed by a shallower decline in later years. As shown in
Figure F-5, a smooth fit was obtained for indoor trionizing jobs, but the results were more erratic
and variable for the other jobs. This variability was judged to be related to variations in the
amount of data available over various time windows rather than to authentic variations in
concentration. On this basis, the LOESS approach was not pursued further. However, the
results did suggest that exponential models could be a reasonable parametric form.
F-25
-------
150
100
Track (Other)
10/25^95
4/19/01
Track Unload
6/ll/Sa 3/8/71 1Z/2/73 8/28/76 5/25/79 2/«/82 11/14/S4 S/11/S7 5/7/9O
Date
fiC
— 50
I 40
6/H/6S
0.40
0.35
t
'<» 0.20
0.05
O.OO
Indoor Trionizing Jobs
12/2/73
5/15/79
34
5/7/90
10/28/95 4/19/01
Background
•—.F'ttcd
• Measurwi
6/11/68 12/2/73 5/25/79 11/14/S4 5/7/90 10/28/95 4/19/01
Date
Figure F-5. Local regression (LOESS) fitting results.
F-26
-------
F.5.3.1.1.2. Fitting method 2: exponential models with job-specific slopes. The second fitting
method that was evaluated assumed a nonlinear regression model to describe the relationship
between fiber concentrations and time. At time t, it was assumed that
C(t) = fj(t) + et (F-l)
where p(f) = mean ofC(f) at time t, and e/is a normally distributed error term with mean
0 and variance structure as discussed below.
A two parameter exponential function was used to model mean fiber concentration at
time t:
XO = a x exp(-6 x t) (F-2)
The intercept parameter (a) and the slope parameter (b) were expressed in terms of
exponentiated functions [a = exp(ao), b = exp(#o)] to guarantee that a, b, and ju(i) could only take
on nonnegative values. Time t was coded as number of years from 1/1/1970 (an arbitrary frame
of reference) to the date of sampling to facilitate model convergence.
When the data were grouped by job and by year, a plot of the natural logarithm (In) of
variance versus the natural logarithm of mean concentration revealed that In-variance tended to
increase approximately as a linear function of the In-mean (see Figure F-6). Based on this, a
"power of the mean" variance function was chosen to describe the mean-variance relation, where
the dimension and value of the power parameter 6 were determined from the data. This broad
class of variance functions is commonly used in nonlinear regression analyses. Different models
for the variance function were tried, including the 1-parameter function, p(ff, and 2-parameter
function, 6\ + ju(ifi. Model convergence was consistently achieved with the 1-parameter power
function model and was not achieved with the 2-parameter function. Consequently, the variance
of the error term was modeled as a 1-parameter power function of the mean fiber concentration
at time t, multiplied by a scale parameter er2 reflecting the overall level of precision in C(f)
(similar to er2 in ordinary linear regression):
Var{C(0> = o2 x v(ff (F-3)
F-27
-------
15
10
OJ
c_>
C
ra
-5
-10
-15
Figure F-6. Variance in industrial hygiene (IH) data as a function of the
mean.
Regression parameters were estimated by iteratively reweighted least squares, in which
estimates of the mean and the variance were alternately updated until convergence. Initially, the
estimation of 6 was incorporated into the estimation of regression parameters. However, this
greatly increased data computations, and model convergence was usually not achieved.
Therefore, the estimate of the parameter 6 was obtained by including a grid search, which
identified values for which model convergence was obtained and provided the value that best fit
the data. A search of values from 0.1 to 2 was sufficient in each analysis to estimate 6. Post hoc
sensitivity analyses were performed in which other values of 9 were manually specified to
confirm that the chosen 6 was optimum. Results showed that a power of the mean model with
0—1 allowed model convergence for all areas. After model parameters were estimated, a2 was
estimated by calculating the mean-squared error (MSB), equal to the weighted sum of squared
deviations of observed minus mean concentrations, divided by the sample size minus number of
parameters (2 for this model). The weights were equal to the inverse of mean concentration to
the power 6 at each time. Analyses were implemented using the SAS procedure PROC NLIN
(SAS for Windows, Version 9.3).
When each job was fit individually, most yielded reasonable fits (see Figure F-7).
However, cleanup and blender yielded fits in which predicted concentrations for 1972-1973
were substantially higher than could be justified with known information about the
manufacturing process. The results for cleanup and blender were likely a result of the absence of
F-28
-------
data in the early time frame (1972-1973), and were considered to be unreliable. On this basis,
this approach (use of independent parameters for each job) was not pursued further.
F.5.3.1.1.3. Fitting method 3: exponential models with common slopes for grouped jobs. To
avoid the unrealistic results generated when each job was allowed to have a separate slope term,
a strategy of grouping jobs expected to show a similar rate of decline in airborne fiber levels was
employed to obtain more reliable and realistic fits. Based on the expectation that the rate of
decline in average exposure level was likely to be similar for trionizing jobs in the same general
area, the trionizing jobs were grouped into two categories: jobs located inside the trionizing
building (indoor trionizing jobs) and jobs located in the railroad yard (outdoor trionizing jobs).
Indoor jobs included blender, cleanup, dryer, expander, feeder, mill, and resin, while outdoor
jobs included track unload and track other. For each group, the data were fit to the model,
requiring the slope parameter (b) to be the same for all jobs within the same group. Results are
displayed in Figure F-8.
F-29
-------
Blender
i
2
06/li/e5 12/02/73 03/23/75 11/U/E4 O3/07/90 1O/13/93
390
300
C eanup
|
I
06/11/63 12/02/73 03/13/79
O3/H7/90 1B/15/93
Expander
11/14/B4 03/07/90
Reiin
06/ll/SS 11/02/71
11/14/St 03.11X7/90 ID/21/9}
Feeder
K/11,''«B U/D2/73
U/14/B4 D3.W90 ICv'iVSS
1-6
L4
: to
o.s
0.6
0.4
0.1
0.0
Track
Dryer
Mill
\
•D
230
.5 200
6
h
g ™
:.'
M
Track Un.oad
ne/n,'sa 12/02/73 03/23,^79
3/7,^0 1A>'2S/S3
Figure F-7. Trionizing department data stratified by job. Variance-weight fitting with independent 6 terms.
F-30
-------
Blender
\
Mjrli,''e^ IZ/OZ,^ CS/Z3/79 U,/14(B4 M/ttT/M
Cleanup
Expander
i
J «
I.
I »
Resin
Feeder
Tradt
L2
«-»
I-
Dryer
Mill
Track Unoad
i •
.£200
06/1US2 Ii'02.r73 03,13,179 1U14VS4 03/C7/90 Ul/21,'99
U/ttQ/73
G/ll,'e2 12/2/73
Figure F-8. Trionizing department data stratified by job. Variance-weight fitting with common b terms for indoor
and outdoor jobs.
F-31
-------
F.5.3.1.1.4. Fitting method 4: segmented exponential models. The fourth approach evaluated
was similar to the third approach, except the data were divided into two or three time segments,
with different exponential curves fit to each segment. This approach was based on the
expectation that the rate of decline in average exposure levels in the trionizing department was
related to the timing and effectiveness of various engineering controls. As discussed in
Section F.2, a number of different engineering controls were installed overtime, with the largest
decreases in dust level tending to occur in the 1976 to 1980 time frame. After 1980, Libby
vermiculite was no longer used, and exposure levels tended to be low and relatively constant.
Based on this, for indoor trionizing jobs, the data were fit using a three-segment approach, with
the time segments defined as follows:
Segment 1: Before 1/1/1976.
Segment!: 1/1/1976 to 12/31/1980.
Segment 3: 1/1/1981 and after.
Engineering controls installed to reduce indoor exposures in the trionizing department are
not expected to have had significant impact on the outdoor exposure levels, so outdoor trionizing
jobs (track other and track unload) were fit to a two-segment model, with the break point
between segments occurring at 1/1/1981, when Libby vermiculite was no longer used. Results
are shown in Figure F-9.
F-32
-------
Blender
I D
I U>
Expander
Resin
3 M
I"2
a "*
Ki .
Cleanup
•n
| u
i"
!:
c.
viva
t^»
Feeder
V
X
Track (Other)
| 18
!:
Dryer
Mill
Track Unload
i 1°
\
5
vat
lm
V
Figure F-9. Weighted exponential fits to indoor (3-segment) and outdoor (2-segment) trionizing jobs.
F-33
-------
F.5.3.1.1.5. Selection of the preferred fitting approach. In choosing between fitting Strategy 3
and fitting Strategy 4, two factors were considered: (1) statistical goodness of fit of the model
and (2) consistency with the general understanding of the impact of engineering controls at the
Marysville facility.
The goodness of fit of the estimation model was determined by calculating the MSB,
where MSB was calculated as the sum of the squared derivations between observed and
predicted values divided by n — p, where n is the number of data points and/? is the number of
model parameters. For both indoor and outdoor jobs, the segmented approach (see Strategy 4)
provided a lower MSB than the un-segmented approach (see Strategy 3), as shown in Table F-7.
Table F-7. Fitting statistics for trionizing
jobs
Data set
Indoor
Trionizing
Outdoor
Trionizing
No. of segments
1
3
1
2
MSE
5.80
5.08
33.6
31.5
In addition, a segmented approach is consistent with the approach used by the University
of Cincinnati for fitting the log-transformed data. This approach is also consistent with the
available information regarding the implementation and effectiveness of various dust control
techniques in the trionizing department. Hence, it is thought that the segmented approach better
represents changes over time, even though the model is somewhat more complex with more
regression parameters than the un-segmented models. The variance parameter 6 of the
segmented models was set at the value determined from the corresponding nonsegmented model,
and was altered slightly, if necessary, to assure convergence. Post hoc sensitivity analyses were
performed to validate that the optimum model fit was obtained. For these reasons, the
segmented fits were selected for use in calculation of the arithmetic-mean-based JEM for
trionizing jobs. Model parameters for the preferred models are shown in Table F-8.
F-34
-------
Table F-8. Parameter values for segmented exponential fits to trionizing jobs
Parameter
a (Segment 1)
a (Segment 2)
a (Segment 3)
b (Segment 1)
b (Segment 2)
b (Segment 3)
Blender
5.69 x 10°
4.34 x 102
1.66 x 1Q-2
2.02 x KT1
9.25 x KT1
6.14x 1(T6
Cleanup
8.81 x 10°
6.72 x 102
2.56 x 1Q-2
2.02 x 1Q-1
9.25 x 1Q-1
6.14x 1(T6
Drier
2.56 x 10°
1.95 x 102
7.45 x 1Q-3
2.02 x 1Q-1
9.25 x 1Q-1
6.14x 1(T6
Expander
1.24 x 101
9.44 x 102
3.60 x lO-2
2.02 x 1Q-1
9.25 x 1Q-1
6.14x 1Q-6
Feeder
5.36 x 101
4.09 x 103
1.56 x 1Q-1
2.02 x 1Q-1
9.25 x 1Q-1
6.14x 1Q-6
Mill
2.17 x 101
1.66 x 103
6.31 x 10-2
2.02 x 1Q-1
9.25 x 1Q-1
6.14x 1Q-6
Resin
5.78 x 10°
4.41 x 102
1.68 x 1(T2
2.02 x 1Q-1
9.25 x 1Q-1
6.14x 1(T6
Track
other
2.42 x 10°
5.46 x 1(T2
-
3.46 x KT1
9.12 x 1(T4
-
Track
unload
2.41 x 102
5.42 x 10°
-
3.46 x KT1
9.12x 1(T4
-
F-35
-------
F.5.3.1.1.6. Calculation of job-weighted average exposure within the trionizins department.
Workers in the trionizing department rotated among jobs, spending approximately equal amounts
of time in each job during each work cycle, including equal time at each of the two dryer
locations. When working at the outdoor track job, the employees reported that about 25% of the
time was spent at track unload and 75% was spent at track other. Based on this, the
job-weighting factors shown in Table F-9 were computed:
Table F-9. Job-weighting factors for trionizing department workers
Indoor
Blender
0.111
Cleanup
0.111
Dryer
0.222
Expander
0.111
Feeder
0.111
Mill
0.111
Resin
0.111
Outdoor
Track other
0.083
Track unload
0.028
The job-weighted average exposure across all jobs (/) for each year (f) in the trionizing
department was then calculated as:
Job-weighted average (t) = £ C(j, t) x JWF(f)
where C(j,f) = exposure concentration while working at job "/' in year "t."
(F-4)
F.5.3.1.1.7. Data for other departments ("background"). As discussed previously, industrial
hygiene measurements in locations where only expanded vermiculite or no vermiculite was used
were defined as having "plant background" exposure. These included measurements in
polyform, office, research, pilot plant, warehouse, and packaging. Measurements of fibers in the
air from these departments tended to be relatively low, with little distinction among departments.
Therefore, data for all background jobs were combined and fit as a single data set.
Both the nonsegmented and two-segment exponential fitting strategies were tested for the
background data set. Of these, the two-segment exponential approach was selected as being
optimum because it better reflects known changes in processes, and the mean square error was
slightly lower than for the nonsegmented model (see Table F-10).
Table F-10. Fitting statistics for background jobs
Data set
Background
No. of segments
1
2
Mean square error
0.020
0.018
Figure F-10 shows the model parameters and the two-segment exponential fit for the
background data set.
F-36
-------
Concentration (PCM f/cc)
Oc _
V
^^>^* A * i *
6/11/68 3/8/71 12/2/73 8/28/76 5/25/79 2/18/82 11/14/84 8/11/87 5/7/90 1/31/93 10/28/95
Sampling Date
Parameter
a (segment 1)
a (segment 2)
b (segment 1)
b (segment 2)
Value
0.491
0.022
0.294
0.013
Figure F-10. Two-segment exponential fit to background jobs.
F.5.3.2. Estimation of Exposure Levels from 1957 to 1971
Extrapolation of model-predicted exposure concentrations in 1972 backwards in time to
earlier years was performed as described in Section F.6.2. In brief, the extrapolation was based
on a consideration of relative dust levels, the relative amounts of vermiculite from Libby or
South Carolina, and the relative asbestos content of these types of vermiculite. The basic
equation used for extrapolation is as follows:
where:
Cj.y
C/,1972
Cj:y = (£,-,1972) ' Extrapolation fact or jy
Extrapolation factor jy = (Dust ratio)yy(FL + k x F5C)y
: Extrapolated fiber concentration for job "/' for year "/'
: Estimated concentration of fiber in job "/' for 1972
(F-5)
F-37
-------
Dust ratio/,^ = Estimated ratio of dust in air for job "/' in year "/' compared to dust
level in 1972
FL = Fraction of vermiculite derived from Libby in year y
Fsc = Fraction of vermiculite derived from South Carolina in year;;
k = Estimated relative concentration of fiber in South Carolina vermiculite
compared to Libby vermiculite
As discussed in Section F.6.2.2, for the indoor trionizing jobs, the dust ratio in 1967 was
assumed to be twice as high as in 1972, decreasing linearly over this time window. For all
background and track jobs, the dust ratio was assumed to be 1:1. Data on the relative amounts of
vermiculite from Libby and South Carolina were derived from company records (see Table F-3,
above), and the relative asbestos content of Libby vermiculite to South Carolina vermiculite was
estimated to be 8.7:1. Based on these values and estimates, extrapolation factors were calculated
as summarized in Table F-l 1.
F-3 8
-------
Table F-ll. Extrapolation factors for 1957-1972
Department
Trionize (all indoor
jobs)
Trionize (outdoor
jobs) and
background
Year
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
Dust ratio
2.00
2.00
2.00
2.00
2.00
2.00
2.00
2.00
2.00
2.00
2.00
1.80
1.60
1.40
1.20
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
FL
0.00
0.00
0.32
0.32
0.32
0.32
0.32
0.57
0.73
0.92
0.87
0.79
0.82
0.90
0.95
1.00
0.00
0.00
0.32
0.32
0.32
0.32
0.32
0.57
0.73
0.92
0.87
0.79
0.82
0.90
0.95
1.00
Fsc
1.00
1.00
0.68
0.68
0.68
0.68
0.68
0.43
0.27
0.08
0.13
0.21
0.18
0.10
0.05
0.00
1.00
1.00
0.68
0.68
0.68
0.68
0.68
0.43
0.27
0.08
0.13
0.21
0.18
0.10
0.05
0.00
k
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
0.115
Extrapolation factor
0.230
0.230
0.796
0.796
0.796
0.796
0.796
1.239
1.522
1.858
1.770
1.465
1.345
1.276
1.147
1.000
0.115
0.115
0.398
0.398
0.398
0.398
0.398
0.619
0.761
0.929
0.885
0.814
0.841
0.911
0.956
1.000
Extrapolation factor = Dust ratio * (FL+k * Fsc).
k = I/ratio; ratio = 8.7.
F-39
-------
F.5.3.3. Results: Job-Exposure Matrix (JEM) Based on Arithmetic Mean Exposure Levels
As described above, IH measurements from the plant were used to estimate yearly
arithmetic mean (AM) exposure levels in the trionizing department and in all other departments
(background) from 1957 to 2000. As described previously, plant maintenance workers were
assumed to be exposed 50% of the time in the trionizing department and 50% of the time in
background departments, and central maintenance workers were assumed to be exposed 10% of
the time in the trionizing department and 90% of the time in background departments.
Table F-12 provides the AM-based JEM developed using this methodology.
F-40
-------
Table F-12. Arithmetic Mean (AM)-based job-exposure matrix (JEM) for
Marysville workers
Year
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
Trionizing
(all jobs)
2.078
2.078
7.200
7.200
7.200
7.200
7.200
11.201
13.761
16.802
16.002
13.487
12.651
12.334
11.483
10.498
8.210
6.484
5.138
3.164
1.473
0.745
0.409
0.244
0.189
0.189
0.189
0.189
0.188
0.188
0.188
0.189
0.188
0.188
Plant
maintenance"
1.053
1.053
3.647
3.647
3.647
3.647
3.647
5.673
6.970
8.511
8.105
6.839
6.425
6.275
5.854
5.367
4.193
3.307
2.618
1.618
0.764
0.392
0.219
0.133
0.104
0.104
0.104
0.104
0.103
0.103
0.103
0.103
0.102
0.102
Central
maintenanceb
0.232
0.232
0.804
0.804
0.804
0.804
0.804
1.252
1.538
1.877
1.788
1.521
1.444
1.427
1.351
1.262
0.978
0.766
0.601
0.382
0.196
0.111
0.068
0.044
0.036
0.036
0.036
0.035d
0.035
0.035
0.035
0.035
0.034
0.034
Background0
0.027
0.027
0.094
0.094
0.094
0.094
0.094
0.146
0.179
0.219
0.209
0.192
0.198
0.215
0.225
0.236
0.175
0.130
0.097
0.073
0.054
0.040
0.030
0.022
0.019
0.019
0.019
0.018
0.018
0.018
0.018
0.017
0.017
0.017
F-41
-------
Table F-12. Arithmetic Mean (AM) based job-exposure matrix (JEM) for
Marysville workers (continued)
Year
1991
1992
1993
1994
1995-2000
Trionizing
(all jobs)
0.187
0.188
0.187
0.187
0.187
Plant maintenance"
0.102
0.102
0.102
0.102
0.102
Central
maintenance1"
0.034
0.034
0.033
0.033
0.033
Background0
0.017
0.017
0.016
0.016
0.016
"Assumed exposure 50% in trionizing and 50% in background departments.
bAssumed exposure 10% in trionizing and 90% in background departments.
background includes pilot plant, research, polyform, office, packaging, and warehouse.
dAfter 1983, central maintenance was outsourced, but some O.M. Scott workers continued in that job.
F.5.4. Selection of the Preferred Job-Exposure Matrix (JEM)
In occupational epidemiology and industrial health studies, evaluations of worker
exposure are often based on estimates of the geometric mean exposure concentration (Seixas et
al., 1988). However, EPA traditionally employs the arithmetic mean exposure level in
computing exposure and risk (U.S. EPA, 1994), and toxicity values employed by EPA in risk
quantification are based on arithmetic mean exposures. For this reason, EPA determined that the
JEM based on untransformed data (as described in Section F.6.3) is the most appropriate for use
in calculating cumulative worker exposure, as described in the following section, and for use in
deriving the reference concentration (RfC).
F.6. DEVELOPMENT OF A CUMULATIVE HUMAN EQUIVALENT EXPOSURE
CONCENTRATION
F.6.1. Basic Equation
In most occupational studies of worker exposure to asbestos, cumulative exposure (CE) is
expressed in units of fibers/cc-years, which is calculated as the product of average exposure
concentration at work (C, fibers/cc) and exposure duration (ED, the number of years at work):
C£(f/cc-yrs) = C (f/cc) x ED (yrs)
(F-6)
F.6.2. Extrapolation from Workplace Exposure to Continuous Exposure
When exposure-response data based on workers are used as the basis for evaluating
exposures and risks in people with continuous exposure (e.g., full-time residents), it is necessary
to convert the cumulative exposure value for each worker to a value that is appropriate for a
resident with continuous exposure:
F-42
-------
CE (continuous) = CE (workplace) x Adj. factor (F-7)
This adjustment accounts for the fact that workers are exposed only part of the day (while
at work), and also accounts for different breathing rates between the workplace and the
residence. In the absence of site-specific data, the adjustment factor for asbestos is calculated as
follows (U.S. EPA. 2014. 1994):
Adj factor = °cc"paona (F-8)
(BR-ET-ED)continuous
(l.25 m3 per ftr)-(8 hrsper day)-(5 work days per week) 50m3 nQC'Ti re c\\
(0.8333m3 per hr)-(24hrs per day)-(7 days per week) 140m3 ' ^ '
In the case of the Marysville cohort, a more complex adjustment is needed to convert
from workplace exposure to continuous exposure, because employees at the Marysville plant
often worked extended work schedules, both in terms of hours per day and days per week, and
these schedules depended on the time of year (season) due to seasonal variations in product
demand (OSHA, 1979). The focus groups were used to gain a more complete understanding of
these work schedules. The groups were comprised of long-term workers with pre- and post-1972
experience across all departments. Therefore, these groups were uniquely qualified to elucidate
the plant work schedules over the full time frame of interest, beginning in 1957.
Based on this understanding of plant operations, six departments were identified that had
a unique set of season-specific exposure parameters (hours/day, days per season):
1 . Trionizing (including track other and track unload).
2. Plant maintenance.
3. Central maintenance.
4. Polyform.
5. Background (office, research lab, pilot plant).
6. Background with extra time (warehouse, packaging).
For each of these departments, a seasonal adjustment factor was calculated using the
following general equation:
c 7 j- f *- ( 1.25m3 per ftr \ (ETd t \ (EDd {\
Seasonal adj. factor,, ,• = - f- - — — — — (F-10)
J ' a'1 \0.8333m3perhrJ \ 24 ) \ Nt ) ^ '
where:
ETd,i = Exposure time (hours/day) in department "d" during season "/'"
EDd,i = Number of days worked in department "d" during season "/'"
Ni = Number of days in season "/"
F-43
-------
For each worker, the date of any job change among these six departments was adjusted so
the change occurred at the starting month for the nearest season. Department-specific and
season-specific values of ET, ED, and TV are provided below, along with the corresponding
seasonal adjustment factors.
F.6.2.1. Trionizing, Plant Maintenance, Poly form, Warehouse, and Packaging
Each of these departments was characterized by a complex work schedule that included
substantial overtime, with the level of overtime work depending on season:
Spring
Season = January 1 to May 31
N= 151.25 days (includes 0.25 days to account for leap years)
Work schedule = 7 days/week, 12 hours/day, with New Years' Day off
ED= 151.25-1 = 150.25
ET= 12 hours/day
Seasonal adj. factor = (1.25/0.8333) x [12/24 x 150.25/151.25] = 0.745
Summer
Season = June 1 to August 31
TV =92 days
Work schedule = 5 days/week, 8 hours/day, with 2 week summer vacation
ED = (92 - 14) x 5/7 = 55.71 days
ET= 8 hours/day
Seasonal adj. factor = (1.25/0.8333) x [8/24 x 55.71/92] = 0.3028
Fall
Season = September 1 to December 31
N= 122 days
Work schedule = 5 days/week, 12 hours/day plus 2 days/week, 8 hours/day, with
Christmas Day off
EDI = 121 days x 5/7 = 86.43 days
En = 12 hours/day
ED2 = 121 x 2/7 = 34.57 days
ET2 = 8 hours/day
Seasonal adj. factor = (1.25/0.8333) x [(12/24 x 86.43) + (8/24 x 34.57)]/122 = 0.6730
F.6.2.2. Office, Pilot Plant, Research, and Central Maintenance
Each of these departments was characterized by a normal work schedule that did not
include overtime.
Spring
Season = January 1 to May 31
N= 151.25
Work schedule = 5 days/week, 8 hours/day, with New Years' Day off
ED = 150.25 days x 5/7 = 107.32 days
F-44
-------
ET= 8 hours/day
Seasonal adj. factor = (1.25/0.8333) x [8/24 x 107.32/151.25] = 0.3548
Summer
Season = June 1 to August 31
TV =92 days
Work schedule = 5 days/week, 8 hours/day, with 2 week summer vacation
ED = (92 - 14) x 5/7 = 55.71 days
ET= 8 hours/day
Seasonal adj. factor = (1.25/0.8333) x [8/24 x 55.71/92] = 0.3028
Fall
Season = September 1 to December 31
N= 122 days
Work schedule = 5 days/week, 8 hours/day, with Christmas Day off
£D = (122-l)x 5/7 = 86.43
ET= 8 hours/day
Season adj. factor = (1.25/0.8333) x [8/24 x 86.43/122] = 0.3542
In summary, the seasonal adjustment factors are as shown in Table F-13:
Table F-13. Seasonal adjustment factors
Departments
Trionizing, plant maintenance, polyform,
warehouse, packaging
Office, pilot plant, research, central maintenance
Spring
0.7450
0.3548
Summer
0.3028
0.3028
Fall
0.6730
0.3542
F.6.3. Calculation of Average Exposure Concentrations
Calculation of the average exposure concentration (C) for each worker is complicated by
the fact that some workers did not spend 100% of the time at work in a single location.
According to the focus group data, each worker was allowed approximately a 30-minute
break for lunch and two 15-minute breaks during the day. Therefore, regardless of job, every
worker was considered to have at least 1 hour of the total time at work spent at a background
exposure location. There was no documentation that a third 15-minute break was provided when
working longer than 8 hours in a day.
In addition, when overtime hours (more than 8 hours/day) were worked, workers in some
departments spent some of their extra hours in other departments. According to focus group
data, the only workers that worked extra hours outside of their own departments were those in
trionizing and polyform. Thus, a decision was needed on how to appropriate the amount of
overtime spent outside trionizing and polyform.
F-45
-------
1. Extra hours for polyform workers—According to the focus groups, polyform workers
first worked in their own department, and went to trionizing to work extra hours.
According to workers, about 75% of the daily overtime was in their own department.
Therefore, for each 4 hours worked beyond the normal 8 hour day, it is estimated that
polyform workers spent 3 hours in polyform and 1 hour in trionizing.
2. Extra hours for trionizing workers—As with polyform workers above, it is estimated
that for every 4 hours of overtime worked by trionizing workers, 3 hours were spent
in trionizing and 1 hour was spent in polyform.
In accord with these exposure parameters, the value of Cdi for each department "d" for
each season "/" was calculated as indicated in Table F-14.
F-46
-------
Table F-14 Equations for calculating Cdi values that account for breaks and interdepartment overtime
Department (d)
Trionizing
Polyform
Plant maintenance
Central maintenance
Warehouse, packaging,
office, pilot plant, research
Season (j)
Spring
10/12 x Ct + 2/12 xCb
1/12 x Ct +11/12 xCb
11/12 xCpm + 1/12 xCb
7/8 x Cam + 1/8 x Cb
Cb
Summer
7/8 xQ+1/8 xCb
Cb
7/8 x Cpm + 1/8 x Cb
7/8 x Ccm + 1/8 x Cb
Cb
Fall
5/7 x (10/12 x Q + 2/12 x Cb) + 2/7 x (7/8 x Q + 1/8 x Cb)
5/7 x (1/12 x Q+ H/12 x Cb) + 2/7 x Cb
5/7 x (H/12 x Cpm + 1/12 x Cb) + 2/7 x (7/8 x Cpm + 1/8 x Cb)
7/8 x Ccm + 1/8 x Cb
Cb
Ct = Concentration in trionizing department.
Cb = Concentration in background departments.
Cpm = Average exposure while performing plant maintenance activities.
Ccm = Average exposure while performing central maintenance activities.
F-47
-------
F.6.4. Calculation of Cumulative Human Equivalent Exposure Concentration (CHEEC)
Given the department-specific seasonal adjustment factors, the cumulative human
equivalent exposure concentration (CHEEC) for each worker is calculated as follows:
CHEEC (f/cc-year) = 2(Cdji x Seasonal Adj. factor^ x JVj/365.25) (F-ll)
where Cdi is the average concentration of fibers inhaled by a worker in department "d"
during season "/'," Ni is the number of days in season "/'," and the sum is calculated across
all seasons that the worker is exposed.
F.6.5. Verification of the Calculations
To verify the accuracy of the CHEEC calculations, several quality control checks were
conducted. The distribution was evaluated by reviewing the mean, median, standard deviation,
highest 10 values, and lowest 10 values. Several workers were also randomly selected and their
values were hand calculated to ensure all programming was appropriate.
F-48
-------
F.7. REFERENCES
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 Environ Med 54: 1350-1358. http://dx.doi.org/10.1097/JOM.Ob013e31824fel74
Cameron. AC: Trivedi. PK. (2013). Regression analysis of count data (2nd ed.). New York, NY: Cambridge
University Press.
Haas. CN: Rose. JB: Gerba. CP. (1999). Quantitative microbial risk assessment. New York: John Wiley and Sons,
Inc.
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.
OSHA (Occupational Safety & Health Administration). (1979). Industrial hygiene survey for OM Scott and Sons,
ScottslawnRd, Marysville, Ohio. November 29, 1978 to January 19, 1979. (Report number 06635-180).
Cincinnati, OH: Occupational Safety and Health Administration.
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
Seixas. NS: Robins. TG: Moulton. LH. (1988). The use of geometric and arithmetic mean exposures in occupational
epidemiology. Am J Ind Med 14: 465-477.
U.S. EPA (U.S. Environmental Protection Agency). (1994). 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). (1999). M/DBP Stakeholder meeting: statistics workshop
meeting summary. M/DBP Stakeholder Meeting, November 19, 1998, Governor's House, Washington DC.
U.S. EPA (U.S. Environmental Protection Agency). (2008). Framework for investigating asbestos-contaminated
superfund sites. (OSWER Directive #9200.0-68). Washington, DC: U.S. Environmental Protection Agency,
Office of Solid Waste and Emergency Response.
http://www.epa.gov/superfund/health/contaminants/asbestos/pdfs/framework asbestos guidance.pdf
U.S. EPA (U.S. Environmental Protection Agency). (2014). IRIS summary sheet for asbestos (CASRN 1332-21-4).
Washington, DC: U.S. Environmental Protection Agency, Integrated Risk Information System. Retrieved
from http://www.epa.gov/iris/subst/0371 .htm
F-49
-------
APPENDIX G. EXTRA RISK AND UNIT RISK CALCULATION
G.I. MESOTHELIOMA MORTALITY
The increased risk of mesothelioma mortality attributable to continuous fiber exposure
was estimated using a life-table procedure based on the general U.S. population. The life-table
procedure involved the application of the estimated Libby Amphibole asbestos 15-specific toxicity
to a structured representation of the general U.S. population in such a manner as to yield
age-specific risk estimates for mesothelioma mortality in the absence and presence of exposure
to Libby Amphibole asbestos. Baseline all-cause mortality rates were included in the life-table
in such a way as to enable computation of the specific absolute risk of mesothelioma mortality
while accounting for other competing causes of mortality. For each age-interval in the life-table,
the effect estimates of the Poisson regression model analysis (the absolute risk) were used to
estimate mesothelioma mortality at a particular exposure level. These age-specific absolute risks
can then be summed over a lifetime. Different exposure levels are evaluated to ascertain what
magnitude of exposure would be expected to produce 1% absolute risk of mesothelioma
mortality. By this method, the exposure-response relationship determined in the Libby worker
cohort is used to estimate mesothelioma mortality in the general U.S. population that would be
expected from continuous lifetime environmental exposure to various concentrations of Libby
Amphibole asbestos.
Assuming no background risk for mesothelioma, extra risk is the same as absolute risk.
Absolute risk estimates were calculated using the effect estimates derived from the modeling of
the mesothelioma mortality risk and a life-table analysis program that accounts for competing
causes of death.16 The unit risk of mesothelioma is computed using the 95% upper bound to
estimate an upper bound for extra risk of mesothelioma due to Libby Amphibole asbestos
exposure. The upper bound calculation is specific to the exposure metric parameters; the effect
of metric uncertainty in these values is discussed in Section 5.4.5.3. Because this human health
assessment derived a combined inhalation unit risk (IUR) for both mesothelioma and lung cancer
mortality, an interim value based on the central effect estimate (rather than the upper bound) is
also computed to avoid statistical concerns regarding the combination of upper bounds. Details
are shown in Section 5.4.5.3. In accordance with EPA's Supplemental Guidance for Assessing
Susceptibility from Early Life Exposure to Carcinogens (U.S. EPA, 2005b), the application of
15The 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.
16This program is an adaptation of the approach previously used by the Committee on the Biological Effects of
Ionizing Radiation (BEIR. 1988). A spreadsheet containing the extra risk calculation for the derivation of the
LEC01 for mesothelioma mortality is presented in Tables G-l.
G-l
-------
the age dependent adjustment factors for substances that act through a mutagenic mode of action
is not recommended (see Section 5.4.5.3).
U.S. age-specific all-cause mortality rates from the 2010 National Vital Statistics Report
(NVSR) for deaths in 2007 among all race and gender groups combined (Xu etal., 2010) were
used to specify the all-cause background mortality rates (R0) in the life-table analysis. The risk
with exposure (Rx) was computed up to age 85 years,17 assuming continuous environmental
exposure to Libby Amphibole asbestos. Conversions between occupational Libby Amphibole
asbestos exposures and continuous environmental asbestos exposures were made to account only
for differences in the amount of air inhaled per day during a higher effort occupational shift
(8 hours; 10 m3) compared to a standard 24-hour (20 m3) day (U.S. EPA, 1994) because results
were already based on a 365-day calendar year. The computation of the unit risk involved three
steps. The first step was to compute the unit risk for adults. This was achieved by initiating
exposure at age 16 years and maintaining continuous exposure throughout the remainder of life
while allowing for the incremental mathematical decay of previously accumulated exposure.18
An age of 16 years was used because it roughly matched the youngest age of a worker in
the subcohort and was consistent with the application of a similar life-table methodology when
the age-dependent adjustment factors (ADAFs) are applied; however, the application of
age-dependent adjustment factors was not recommended in this case (see Section 4.6.2.2). An
adjustment was also made in the life-table for the lag period, so that the age-specific risk
calculations began at 16+ (the length of the lag period) years of age. The standard assumption
used by the U.S. Environmental Protection Agency (EPA) is that the average lifetime spans
70 years. Because the adult-only-exposure unit risk excluded the first 16 years, the adult-only-
exposure unit risk based on 54 years was then rescaled for an entire lifetime of continuous
exposure by multiplying the interim value for adult-only-exposure by 70/54 to cover the
childhood years (<16 years) to compute the "adult-based" unit risk. After reseating, the resulting
"adult-based" lifetime unit risk estimate (in contrast to the unsealed "adult-only-exposure" unit
risk estimate obtained from the life-table calculations) may be prorated for less-than-lifetime
exposure scenarios in the same manner as would be used for an "adult-based" unit risk estimate
derived from a rodent bioassay.
Consistent with the Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), the
same data and methodology were also used to estimate the exposure level effective concentration
(ECx) and the associated 95% lower confidence limit of that exposure level effective
concentration (LEG*) corresponding to an absolute risk of 1% (x = 0.01). A 1%-risk level is
commonly used for the determination of the point of departure (POD) for low-dose extrapolation
17Note that 85 years is not employed here as an average lifespanbut, rather, as a cut-off point for the life-table
analysis, which uses actual age-specific mortality rates.
18Exposures in the life-tables were computed at the mid-point of each age interval and appropriately lagged.
G-2
-------
from epidemiological data, and the LEG value corresponding to that risk level was used as the
actual POD.
The following table illustrates the computational details of the unit risks for
mesothelioma mortality (see Table G-l). The result of Table G-l is shown in Table 5-49 and is
not adjusted for the underascertainment of mesothelioma described in Section 5.4.5.1.1. The unit
risks adjusted for underascertainment are shown in Table 5-49.
Column Definitions for Table G-l:
Column A: Age interval up to age 85.
Column B: All-cause mortality rate for interval /' (rate x 105) (Xu etal., 2010).
Column C: All-cause hazard rate for interval / (h*Xi) (= all-cause mortality
rate x number of years in age interval).
Column D: Probability of surviving interval / (qi) [= exp(-/z*jcz-)].
Column E: Probability of surviving up to interval /' (Si) (Si = 1; Si = Si-i x qi-\, for / > 1).
Column F: Lagged exposure at midinterval (x dose) assuming constant exposure was
initiated at age 16.
Column G: Mesothelioma mortality hazard rate in exposed people for interval (hi). To
estimate the LECoi, i.e., the 95% lower bound on the continuous exposure
giving an extra risk of 1%, the 95% upper bound on the regression
coefficient is used.
Column H: All-cause hazard rate in exposed people for interval /' (h*xi)
[= h*x, + (hx, - h,)].
Column I: Probability of surviving interval /' without dying from mesothelioma for
exposed people (qxi) [= exp(—h*Xi)].
Column J: Probability of surviving up to interval /' without dying from mesothelioma
for exposed people (Sxi) (Sx\ = 1; Sxi = Sxi-i x qxi-\, for / > 1).
Column K: Conditional probability of dying from mesothelioma in interval /' for
exposed people [= (hx + h*xi) x Sx,• x (1 - qxi)] (Rx, the lifetime probability
of dying from mesothelioma for exposed people = the sum of the
conditional probabilities across the intervals).
Note that the life-tables for mesothelioma mortality estimate the extra risk as the absolute
risk as there is no assumption of a background risk in the absence of exposure. In each of the
life-tables, inhalation exposure commences at age 16 years and continues at the same exposure
concentration for the duration of the life-table. This allows for the computation of an
"adult-only-exposure" occupational lifetime unit risk, which is then scaled by a ratio of 70:54 to
G-3
-------
account for risk over the standard 70-year lifetime. While exposure is initiated in the life-table at
age 16 years, this exposure is lagged to match the corresponding exposure-response models,
which provide the hazard rates per unit of exposure. For example, in Table G-l, Column F
shows exposure lagged by 10 years so that no lagged exposure appears in the table prior to age
26 years (16 + 10). Note that risks are initially shown in 1-year intervals because children's risk
intervals can be smaller, and there was a need to be able to begin exposures at 16 years.
G-4
-------
Table G-l. Mesothelioma extra risk calculation for environmental exposure to 0.1479 fiber/cc Libby
Amphibole asbestos using the metric of cumulative exposure with a 10-year exposure lag and a 5-year
half-life of exposure, as described in Section 5.4.5.3
A
Age
int.
<1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
B
All-cause
mortality
(xlO5)
684.5
28.6
28.6
28.6
28.6
13.7
13.7
13.7
13.7
13.7
18.7
18.7
18.7
18.7
18.7
61.9
61.9
61.9
C
All-cause
hazard
rate (/?*)
0.0068
0.0003
0.0003
0.0003
0.0003
0.0001
0.0001
0.0001
0.0001
0.0001
0.0002
0.0002
0.0002
0.0002
0.0002
0.0006
0.0006
0.0006
D
Prob. of
surviving
interval
(9)
0.9932
0.9997
0.9997
0.9997
0.9997
0.9999
0.9999
0.9999
0.9999
0.9999
0.9998
0.9998
0.9998
0.9998
0.9998
0.9994
0.9994
0.9994
E
Prob. of
surviving
up to
interval
(S)
1.0000
0.9932
0.9929
0.9926
0.9923
0.9920
0.9919
0.9918
0.9916
0.9915
0.9914
0.9912
0.9910
0.9908
0.9906
0.9904
0.9898
0.9892
F
Lagged exp.
mid. int.
(ATdose)
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
G
Exposed
meso.
hazard rate
(hx)
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
H
Exposed all-
cause haz.
rate
(h*x)
0.0068
0.0003
0.0003
0.0003
0.0003
0.0001
0.0001
0.0001
0.0001
0.0001
0.0002
0.0002
0.0002
0.0002
0.0002
0.0006
0.0006
0.0006
I
Exposed
prob. of
surviving
interval (qx)
0.9932
0.9997
0.9997
0.9997
0.9997
0.9999
0.9999
0.9999
0.9999
0.9999
0.9998
0.9998
0.9998
0.9998
0.9998
0.9994
0.9994
0.9994
J
Exposed
prob. of
surviving up
to int. (Sx)
1.0000
0.9932
0.9929
0.9926
0.9923
0.9920
0.9919
0.9918
0.9916
0.9915
0.9914
0.9912
0.9910
0.9908
0.9906
0.9904
0.9898
0.9892
K
Exposed cond.
prob. of meso. in
interval (Rx)
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
G-5
-------
Table G-l. Mesothelioma extra risk calculation for environmental exposure to 0.1479 fiber/cc Libby
Amphibole asbestos using the metric of cumulative exposure with a 10-year exposure lag and a 5-year
half-life of exposure, as described in Section 5.4.5.3 (continued)
A
Age
int.
18
19
20
21
22
23
24
25
26
27
28
29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
B
All-cause
mortality
(xlO5)
61.9
61.9
98.3
98.3
98.3
98.3
98.3
99.4
99.4
99.4
99.4
99.4
110.8
145.8
221.6
340.0
509.0
726.3
1,068.3
1,627.5
C
All-cause
hazard
rate (/?*)
0.0006
0.0006
0.0010
0.0010
0.0010
0.0010
0.0010
0.0010
0.0010
0.0010
0.0010
0.0010
0.0055
0.0073
0.0111
0.0170
0.0255
0.0363
0.0534
0.0814
D
Prob. of
surviving
interval
(?)
0.9994
0.9994
0.9990
0.9990
0.9990
0.9990
0.9990
0.9990
0.9990
0.9990
0.9990
0.9990
0.9945
0.9927
0.9890
0.9831
0.9749
0.9643
0.9480
0.9218
E
Prob. of
surviving
up to
interval
(S)
0.9886
0.9880
0.9874
0.9864
0.9854
0.9845
0.9835
0.9825
0.9815
0.9806
0.9796
0.9786
0.9777
0.9723
0.9652
0.9546
0.9385
0.9149
0.8823
0.8364
F
Lagged exp.
mid. int.
(ATdose)
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.144
0.401
0.626
0.821
1.268
1.701
1.918
2.026
2.080
2.107
2.121
2.127
G
Exposed
meso.
hazard rate
(hx)
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0001
0.0002
0.0003
0.0004
0.0006
0.0009
0.0010
0.0010
0.0011
0.0011
0.0011
0.0011
H
Exposed all-
cause haz.
rate
(h*x)
0.0006
0.0006
0.0010
0.0010
0.0010
0.0010
0.0010
0.0010
0.0011
0.0012
0.0013
0.0014
0.0062
0.0082
0.0121
0.0180
0.0265
0.0374
0.0545
0.0825
I
Exposed
prob. of
surviving
interval (qx)
0.9994
0.9994
0.9990
0.9990
0.9990
0.9990
0.9990
0.9990
0.9989
0.9988
0.9987
0.9986
0.9938
0.9919
0.9880
0.9821
0.9738
0.9633
0.9470
0.9209
J
Exposed
prob. of
surviving up
to int. (Sx)
0.9886
0.9880
0.9874
0.9864
0.9854
0.9845
0.9835
0.9825
0.9815
0.9805
0.9793
0.9780
0.9767
0.9706
0.9628
0.9512
0.9342
0.9098
0.8764
0.8299
K
Exposed cond.
prob. of meso. in
interval (Rx)
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0001
0.0002
0.0003
0.0004
0.0006
0.0008
0.0009
0.0010
0.0010
0.0010
0.0009
0.0009
G-6
-------
Table G-l. Mesothelioma extra risk calculation for environmental exposure to 0.1479 fiber/cc Libby
Amphibole asbestos using the metric of cumulative exposure with a 10-year exposure lag and a 5-year
half-life of exposure, as described in Section 5.4.5.3 (continued)
A
Age
int.
70-74
75-79
80-84
B
All-cause
mortality
(xlO5)
2,491.3
3,945.9
6,381.4
C
All-cause
hazard
rate (/?*)
0.1246
0.1973
0.3191
D
Prob. of
surviving
interval
(?)
0.8829
0.8209
0.7268
E
Prob. of
surviving
up to
interval
(S)
0.7710
0.6807
0.5588
F
Lagged exp.
mid. int.
(ATdose)
2.131
2.132
2.133
G
Exposed
meso.
hazard rate
(hx)
0.0011
0.0011
0.0011
H
Exposed all-
cause haz.
rate
(h*x)
0.1256
0.1984
0.3202
I
Exposed
prob. of
surviving
interval (qx)
0.8819
0.8201
0.7260
J
Exposed
prob. of
surviving up
to int. (Sx)
0.7642
0.6740
0.5527
K
Exposed cond.
prob. of meso. in
interval (Rx)
0.0008
0.0007
0.0005
Absolute Rx = 0.0 100
exp. = exposure, haz. = hazard, int. = interval, meso. = mesothelioma, mid. = mid-interval, Prob. = probability.
Absolute risk = 0.01000, exp. Level = 0.1479; occupational lifetime unit risk = 0.01/0.1479 = 0.0676 (based on occupational exposures beginning at age 16 yr);
scaled occupational lifetime unit risk = 0.0876 (scaled by ratio of 70:54 to account for risk over 70-yr lifetime).
G-7
-------
G.2. LUNG CANCER MORTALITY
Lung cancer mortality risk computations are very similar to mesothelioma mortality
computations above (see Section G.I), with one important difference that extra risk is used for
lung cancer. Extra risk is defined as equaling (Rx - Ro) + (1 - Ro), where R* is the lifetime lung
cancer mortality risk in the exposed population and Rois the lifetime lung cancer mortality risk
in an unexposed population (i.e., the background risk). U.S. age-specific all-cause mortality
rates from the 2010 National Vital Statistics Report (Xu et al., 2010) for deaths in 2007 among
all race and gender groups combined were used to specify the all-cause background mortality
rates (R0) in the life-table analysis. Cause-specific background mortality rates for cancers of the
lung, trachea, and bronchus were obtained from a Surveillance, Epidemiology, and End Results
(SEER) report on mortality during 2003-2007 (2003-2007 Surveillance Epidemiology and End
Results Table 15.10, age-specific U.S. death rates).
The following tables show details of the computations of the unit risks for lung cancer
mortality (see Tables G-2). The result of Table G-2 is shown in Table 5-52.
Column Definitions for Tables G-2:
Column A:
Column B:
Column C:
Column D:
Column E:
Column F:
Column G:
Column H:
Column I:
Column J:
Age interval up to age 85.
All-cause mortality rate for interval /' (x 105/year) (Xu etal., 2010).
Lung cancer mortality rate for interval / (x 105/year) (2003-2007
Surveillance, Epidemiology and End Results Table 15.10, age-specific U.S.
death rates).
All-cause hazard rate for interval /' (h*xi) (= all-cause mortality
rate x number of years in age interval).
Probability of surviving interval /' (qi) [= exp(-/z*x,)].
Probability of surviving up to interval /' (Si) (Si = 1; Si = Si-i x grMj for /' > 1).
Lung cancer mortality hazard rate for interval /' (hi) (= lung cancer mortality
rate x number of years in interval).
Conditional probability of dying from lung cancer in interval /'
[= (h H- h*i) x Si x (1 - qi)], i.e., conditional upon surviving up to interval /'
(R0, the background lifetime probability of dying from lung cancer = the
sum of the conditional probabilities across the intervals).
Lagged exposure at midinterval (x dose) assuming constant exposure was
initiated at age 16.
Lung cancer mortality hazard rate in exposed people for interval. To
estimate the LECoi, i.e., the 95% lower bound on the continuous exposure
G-8
-------
giving an extra risk of 1%, the 95% upper bound on the regression
coefficient is used, i.e., maximum likelihood estimate + 1.645 x standard
error.
Column K: All-cause hazard rate in exposed people for interval /' (h*Xi)
Column L: Probability of surviving interval /' without dying from lung cancer for
exposed people (qxi) [= exp(-/z*jcz-)].
Column M: Probability of surviving up to interval /' without dying from lung cancer for
exposed people (Sxi) (Sx\ = l;Sxi = Sxi-i x qxi-i, for / > 1).
Column N: Conditional probability of dying from lung cancer in interval /' for exposed
people [= (hxi + h*Xi) x Sx, x (1 - qxi)] (Rx, the lifetime probability of dying
from lung cancer for exposed people = the sum of the conditional
probabilities across the intervals).
In each of the life-tables, inhalation exposure commences at age 16 years and continues
at the same exposure concentration for the duration of the life-table. This allows for the
computation of an "adult-only-exposure" occupational lifetime unit risk, which is then scaled by
a ratio of 70:54 to account for risk over the standard 70-year lifetime. While exposure is initiated
at age 16 years, this exposure is lagged to match the corresponding exposure-response models,
which provide the hazard rates per unit of exposure. For example, in Tables G-2, Column I
shows exposure lagged by 10 years so that no lagged exposure appears prior to age 26 years.
G-9
-------
Table G-2. Lung cancer extra risk calculation for environmental exposure to 0.191 fiber/cc Libby Amphibole
asbestos using a linear exposure-response model based on the metric of cumulative exposure with a 10-year
exposure lag, as described in Section 5.4.5.3
A
Age
Int.
<1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
B
All-cause
mortality
(xlO5)
684.5
28.6
28.6
28.6
28.6
13.7
13.7
13.7
13.7
13.7
18.7
18.7
18.7
18.7
18.7
61.9
61.9
C
Lung CA
mortality
(xlO5)
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
D
All
cause
hazard
rate
(h*)
0.0068
0.0003
0.0003
0.0003
0.0003
0.0001
0.0001
0.0001
0.0001
0.0001
0.0002
0.0002
0.0002
0.0002
0.0002
0.0006
0.0006
E
Prob. of
surviving
interval
(?)
0.9932
0.9997
0.9997
0.9997
0.9997
0.9999
0.9999
0.9999
0.9999
0.9999
0.9998
0.9998
0.9998
0.9998
0.9998
0.9994
0.9994
F
Prob. of
surviving
up to
interval
(S)
1.0000
0.9932
0.9929
0.9926
0.9923
0.9920
0.9919
0.9918
0.9916
0.9915
0.9914
0.9912
0.9910
0.9908
0.9906
0.9904
0.9898
G
Lung
CA
hazard
rate
(h)
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
H
Cond. prob.
of lung CA
mortality
in interval
(«o)
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
I
Lagged
exp.
mid.
int.
(ATdose)
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
J
Exposed
lung CA
hazard
rate (hx)
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
K
Exposed
all-cause
haz. rate
(h*x)
0.0068
0.0003
0.0003
0.0003
0.0003
0.0001
0.0001
0.0001
0.0001
0.0001
0.0002
0.0002
0.0002
0.0002
0.0002
0.0006
0.0006
L
Exposed
prob. of
surviving
interval
(qx)
0.9932
0.9997
0.9997
0.9997
0.9997
0.9999
0.9999
0.9999
0.9999
0.9999
0.9998
0.9998
0.9998
0.9998
0.9998
0.9994
0.9994
M
Exposed
prob. of
surviving
up to int.
(Sx)
1.0000
0.9932
0.9929
0.9926
0.9923
0.9920
0.9919
0.9918
0.9916
0.9915
0.9914
0.9912
0.9910
0.9908
0.9906
0.9904
0.9898
N
Exposed
cond.
prob.
of lung CA
in interval
(**)
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
G-10
-------
Table G-2. Lung cancer extra risk calculation for environmental exposure to 0.191 fiber/cc Libby Amphibole
asbestos using a linear exposure-response model based on the metric of cumulative exposure with a 10-year
exposure lag, as described in Section 5.4.5.3 (continued)
A
Age
Int.
17
18
19
20
21
22
23
24
25
26
27
28
29
30-34
35-39
40-44
45-49
B
All-cause
mortality
(xlO5)
61.9
61.9
61.9
98.3
98.3
98.3
98.3
98.3
99.4
99.4
99.4
99.4
99.4
110.8
145.8
221.6
340.0
C
Lung CA
mortality
(xlO5)
0
0
0
0.1
0.1
0.1
0.1
0.1
0.2
0.2
0.2
0.2
0.2
0.5
2.1
7.9
20.2
D
All
cause
hazard
rate
(h*)
0.0006
0.0006
0.0006
0.0010
0.0010
0.0010
0.0010
0.0010
0.0010
0.0010
0.0010
0.0010
0.0010
0.0055
0.0073
0.0111
0.0170
E
Prob. of
surviving
interval
(?)
0.9994
0.9994
0.9994
0.9990
0.9990
0.9990
0.9990
0.9990
0.9990
0.9990
0.9990
0.9990
0.9990
0.9945
0.9927
0.9890
0.9831
F
Prob. of
surviving
up to
interval
(S)
0.9892
0.9886
0.9880
0.9874
0.9864
0.9854
0.9845
0.9835
0.9825
0.9815
0.9806
0.9796
0.9786
0.9777
0.9723
0.9652
0.9546
G
Lung
CA
hazard
rate
(h)
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0001
0.0004
0.0010
H
Cond. prob.
of lung CA
mortality
in interval
(«o)
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0001
0.0004
0.0010
I
Lagged
exp.
mid.
int.
(ATdose)
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.10
0.29
0.48
0.67
1.24
2.20
3.15
4.11
J
Exposed
lung CA
hazard
rate (hx)
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0001
0.0004
0.0011
K
Exposed
all-cause
haz. rate
(h*x)
0.0006
0.0006
0.0006
0.0010
0.0010
0.0010
0.0010
0.0010
0.0010
0.0010
0.0010
0.0010
0.0010
0.0055
0.0073
0.0111
0.0171
L
Exposed
prob. of
surviving
interval
(qx)
0.9994
0.9994
0.9994
0.9990
0.9990
0.9990
0.9990
0.9990
0.9990
0.9990
0.9990
0.9990
0.9990
0.9945
0.9927
0.9890
0.9831
M
Exposed
prob. of
surviving
up to int.
(Sx)
0.9892
0.9886
0.9880
0.9874
0.9864
0.9854
0.9845
0.9835
0.9825
0.9815
0.9806
0.9796
0.9786
0.9777
0.9722
0.9652
0.9545
N
Exposed
cond.
prob.
of lung CA
in interval
(**)
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0001
0.0004
0.0010
G-ll
-------
Table G-2. Lung cancer extra risk calculation for environmental exposure to 0.191 fiber/cc Libby Amphibole
asbestos using a linear exposure-response model based on the metric of cumulative exposure with a 10-year
exposure lag, as described in Section 5.4.5.3 (continued)
A
Age
Int.
50-54
55-59
60-64
65-69
70-74
75-79
80-84
B
All-cause
mortality
(xlO5)
509.0
726.3
1,068.3
1,627.5
2,491.3
3,945.9
6,381.4
C
Lung CA
mortality
(xlO5)
39.8
74.7
139.8
220.9
304.3
369.5
379.4
D
All
cause
hazard
rate
(h*)
0.0255
0.0363
0.0534
0.0814
0.1246
0.1973
0.3191
E
Prob. of
surviving
interval
(?)
0.9749
0.9643
0.9480
0.9218
0.8829
0.8209
0.7268
F
Prob. of
surviving
up to
interval
(S)
0.9385
0.9149
0.8823
0.8364
0.7710
0.6807
0.5588
G
Lung
CA
hazard
rate
(h)
0.0020
0.0037
0.0070
0.0110
0.0152
0.0185
0.0190
H
Cond. prob.
of lung CA
mortality
in interval
(«o)
0.0018
0.0034
0.0060
0.0089
0.0110
0.0114
0.0091
R0 = 0.0531
I
Lagged
exp.
mid.
int.
(ATdose)
5.06
6.02
6.97
7.93
8.88
9.84
10.79
J
Exposed
lung CA
hazard
rate (hx)
0.0022
0.0042
0.0080
0.0129
0.0181
0.0224
0.0235
K
Exposed
all-cause
haz. rate
(h*x)
0.0257
0.0368
0.0544
0.0832
0.1275
0.2013
0.3236
L
Exposed
prob. of
surviving
interval
(qx)
0.9747
0.9639
0.9470
0.9201
0.8803
0.8177
0.7236
M
Exposed
prob. of
surviving
up to int.
(Sx)
0.9384
0.9146
0.8815
0.8348
0.7682
0.6762
0.5529
N
Exposed
cond.
prob.
of lung CA
in interval
(**)
0.0020
0.0038
0.0069
0.0103
0.0131
0.0137
0.0111
Rx = 0.0625
CA = cancer, cond. = conditional, exp. = exposure, haz. = hazard, int. = interval, mid. = mid-interval, Prob. = probability.
Extra risk = 0.01001; exp. Level = 0.191; occupational lifetime unit = 0.01/0.191 = 0.0524 (based on occupational exposures beginning at age 16 yr); scaled
occupational lifetime unit = 0.0679 (scaled by ratio of 70:54 to account for risk over 70-yr lifetime).
G-12
-------
G.3. REFERENCES
BEIR (Committee on the Biological Effects of Ionizing Radiation). (1988). Health risks of radon and other
internally deposited alpha-emitters: BEIR IV. Washington, DC: National Academy Press.
http://www.nap.edu/openbook.php?isbn=0309037972
U.S. EPA (U.S. Environmental Protection Agency). (1994). 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). (2005). Guidelines for carcinogen risk assessment. (EPA/630/P-
03/001F). Washington, DC: U.S. Environmental Protection Agency, Risk Assessment Forum.
http://www.epa.gov/cancerguidelines/
Xu. JQ: Kochanek. KD: Murphy. SL: Tejada-Vera. B. (2010). Deaths: Final Data for 2007. Hyattsville, MD:
National Center for Health Statistics, http://www.cdc.gov/nchs/data/nvsr/nvsr58/nvsr5819.pdf
G-13
-------
APPENDIX H. GLOSSARY OF ASBESTOS TERMINOLOGY
The definitions associated with asbestos literature often vary depending on the source or
publication in which it is used. There are definitions applied to industrial, interdisciplinary,
medical, mineralogical, and regulatory usage of terms associated with the discipline involved
with mineral fiber reporting. The definitions are a source of ongoing debate within the asbestos
community centering on nomenclature. From the academic, industrial, and regulatory literature,
it is clear that there is disagreement and perhaps misunderstanding regarding some of the
terminology used by workers in various asbestos-related fields. For many of the definitions
contained herein and for perspectives on the evolution of these terms, the reader is referred to
Lowers and Meeker (2002). MOSH (201IX and NKC (1984). Risk assessment terminology for
the IRIS program can be found in the IRIS Glossary.
Acicular: The very long and very thin, often needle-like shape, that characterizes some
prismatic crystals. (Prismatic crystals have one elongated dimension and two other dimensions
that are approximately equal.) Acicular crystals or fragments do not have the strength,
flexibility, or other properties often associated with asbestiform fibers.
Actinolite: A calcic amphibole mineral in the tremolite-ferroactinolite solid solution series.
Actinolite can occur in both asbestiform and nonasbestiform mineral habits. The asbestiform
variety is often referred to as actinolite asbestos.
Amosite. A magnesium-iron-manganese-lithium amphibole mineral in the
cummingtonite-grunerite solid solution series that occurs in the asbestiform habit. The name
amosite is a commercial term derived from the acronym for "Asbestos Mines of South Africa."
Amosite is sometimes referred to as "brown asbestos."
Amphibole: A group of silicate minerals that may occur either in massive or fibrous
(asbestiform) habits.
Anthophyllite: A magnesium-iron-manganese-lithium amphibole mineral in the anthophyllite
gedrite solid solution series that can occur in both the asbestiform and nonasbestiform mineral
habits. The asbestiform variety is referred to as anthophyllite asbestos.
Asbestiform (minemlosical): A specific type of mineral fibrosity in which the fibers and fibrils
are long and 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 are therefore suitable for uses where incombustible, nonconducting, or
chemically resistant materials are required.
H-l
-------
Asbestos Structure. A term applied to any connected or overlapping grouping of asbestos fibers
or bundles, with or without other particles.
Aspect Ratio. The ratio of the length of a particle to its diameter.
Biopersistence: The ability to remain in the lung or other tissue. Biopersistence of mineral
fibers is a function of their fragility, solubility, and clearance.
Bundle: A group of fibers occurring side by side with parallel orientations.
Chrysotile. A mineral in the serpentine mineral group that occurs in the asbestiform habit.
Chrysotile generally occurs segregated as parallel fibers in veins or veinlets and can be easily
separated into individual fibers or bundles. Often referred to as "white asbestos," chrysotile is
used commercially in cement or friction products and for its good spinnability in the making of
textile products.
Cleavage Fragment. A fragment produced by breakage of a crystal in directions that are related
to the crystal structure and are always parallel to possible crystal faces. A mineral on an
approximately planar surface on a mineral that is controlled by its crystal structure.
Cluster: A group of overlapping fibers oriented at random.
Crocidolite. A sodic amphibole mineral in the glaucophane-riebeckite solid solution series.
Crocidolite, commonly referred to as "blue asbestos," is a varietal name for the asbestiform habit
of the mineral riebeckite.
Durability. The tendency of particles to resist degradation in body fluids.
Edenite. A calcic amphibole mineral in the hornblende solid solution series. Edenite occurs in a
blocky massive form or as fibrous asbestiform. It is present in trace levels in Libby Amphibole
asbestos.
Fiber (mineralosical): 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.
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
length greater than or equal to 5 jim and an aspect ratio greater than or equal to 3:1 (by PCM) or
5:1 (by transmission electron microscopy [TEM]).
Fibril: A single fiber which cannot be separated into smaller components without losing
its fibrous properties or appearance. 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, needle-like
grains or fibers.
Fragility. The tendency of particles to break into smaller particles.
H-2
-------
Libby Amphibole Asbestos (LAA). The term 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, as described in
Section 2.2.
Masnesio-arfvedsonite: A sodic amphibole mineral in the magnesio-arfvedsonite-arfvedsonite
solid solution series. It occurs in asbestiform and nonasbestiform habit. It occurs in trace levels
in Libby Amphibole asbestos.
Masnesio-riebeckite: A sodic amphibole mineral the magnesio-riebeckite-riebeckite solid
solution series. It occurs in nonasbestiform, blocky, massive and asbestiform habit. In Libby
Amphibole asbestos, it is infrequently identified in the asbestiform habit. It occurs in trace levels
in Libby Amphibole asbestos.
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.
Nonasbestiform: The term used to describe fibers not having an asbestiform habit. The massive
nonfibrous forms of the asbestos minerals have the same chemical formula and internal crystal
structure as the asbestiform variety but have crystal habits in which growth is more equivalent in
two or three dimensions instead of primarily one dimension. When milled or crushed,
nonasbestiform minerals generally do not break into fibers/fibrils but rather into fragments
resulting from cleavage along the two or three growth planes. Often, cleavage fragments can
appear fibrous.
Parting The tendency of a crystal or grain to break along crystallographic planes weakened by
inclusions or structural defects. Different specimens of the same mineral may or may not exhibit
parting. Twinned crystals often part along composition planes, which are lattice planes and,
therefore, potentially crystal faces. Parting is similar to cleavage.
Phase Contrast Microscopy (PCM): A form of light microscopy used to count fibers collected
on 25-mm or 37-mm cellulose ester air filters following NIOSH Method 7400 (commonly
referred to as PCM fibers). Fiber counting criteria include: fibers longer than 5 jim in length,
>0.25 |im in diameter with an aspect ratio of 3:1 or greater. Commonly used to assess
occupational exposures to mineral fibers.
Phased Contrast Microscope Equivalent (PCME). A subset of fibers counted by transmission
electron microscopy following ISO 10312 that were collected on cellulose filters. Fibers are
counted following the PCM counting rules. PCME fibers will be a subset of the total structures
counted under ISO 10312.
Prismatic: Having blocky, pencil-like elongated crystals that are thicker than needles.
Refractory Ceramic Fiber (RCF): An amorphous, synthetic fiber produced by melting and
blowing or spinning calcined kaolin clay or a combination of alumina (AbOs) and silicon
dioxide (SiO2). Oxides (such as zirconia, ferric oxide, titanium oxide, magnesium oxide, and
calcium oxide) and alkalis may be added.
Richterite. A sodic-calcic amphibole mineral in the richterite-ferro-richterite solid solution
series. It occurs in fibrous and nonfibrous habits.
H-3
-------
Solid Solution Series. A grouping of minerals that includes two or more minerals in which the
cations in secondary structural position are similar in chemical properties and size and can be
present in variable but frequently limited ratios.
Structure: A term used mainly in microscopy, usually including asbestos fibers, bundles,
clusters, and matrix particles that contain asbestos.
Thoracic-Size Particle. A particle with an aerodynamic equivalent diameter that enables it to be
deposited in the airways of the lung or the gas exchange region of the lung when inhaled.
Tremolite. A calcic amphibole mineral in the series tremolite-ferroactinolite. Tremolite can
occur in both fibrous and nonfibrous mineral habits. The asbestiform variety is often referred to
as tremolite asbestos. Due only to changes in the International Mineralogical Association's
amphibole nomenclature, subsets of what was formerly referred to as tremolite asbestos are now
mineralogically specified as asbestiform winchite and asbestiform richterite.
Winchite. A sodic-calcic amphibole mineral in the barroisite-ferro-barroisite solid solution
series. It occurs in fibrous and nonfibrous habits. It was formerly referred to as soda-tremolite
when first described in the Rainy Creek complex.
H.l. REFERENCES
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
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/
NRC (National Research Council). (1984). Asbestiform fibers: nonoccupational health risks. Washington (DC).
H-4
-------
APPENDIX I. EVALUATION OF LOCALIZED PLEURAL THICKENING IN
RELATION TO PULMONARY FUNCTION MEASURES
The outcome used to derive the reference concentration in this Toxicological Review is
localized pleural thickening (LPT) (in the absence of asbestosis, defined as small interstitial
opacities >1/0), as described by the International Labor Organization (ILO, 2002) and
implemented by Rohs et al. (2008). LPT is a persistent structural change to the pleura, and as
shown in this appendix, LPT is associated with decrements in pulmonary function. The
U.S. Environmental Protection Agency (EPA) sought information pertaining to the impact and
progression of LPT by conducting a systematic evaluation of cross-sectional and longitudinal
studies examining the relationship between LPT and pulmonary function, focusing on forced
vital capacity (FVC) and forced expiratory volume in 1 second (FEVi) as the primary measures
of pulmonary function.
LPT was not defined by the ILO until the 2000 guidelines were published (ILO. 2002).
Previously, the 1980 ILO guidelines defined circumscribed pleural thickening (plaques) and
diffuse pleural thickening (DPT), either with or without costophrenic angle obliteration. LPT
was introduced as a term in the 2000 ILO guidance. LPT includes plaques on the chest wall and
at other sites (e.g. diaphragm). Plaques on the chest wall can be viewed either face-on or in
profile. A minimum width of about 3 mm is required for an in-profile plaque to be recorded as
present according to the 2000 ILO guidance. Neither classification for pleural thickening (LPT
or DPT) in the 2000 ILO guidelines exactly corresponds with the previous ILO classification
systems for pleural thickening; LPT is defined differently than the previous category of pleural
plaques, and DPT is defined more narrowly due to the requirement for continuity with
costophrenic angle obliteration and a 3 mm minimum width for DPT extending up the lateral
chest wall.
Different researchers have used different terminology for circumscribed pleural
thickening or plaques when implementing the 1980 ILO guidelines, most often using the term
"pleural plaques." "Some studies clearly included in reported plaques sites other than the chest
wall, while other studies did not explicitly describe inclusion of plaques in other sites.
Because the "LPT" designation is fairly recent, few studies provide data for this specific
outcome. Therefore, EPA also considered studies examining the relationship between
circumscribed pleural thickening (plaques) as defined in the 1980 ILO guidelines and pulmonary
function.
The research question addressed by this review concerns the functional impact of LPT (or
pleural plaques): Is the presence of LPT (or pleural plaques) associated with decrements in
percent predicted pulmonary function? The search was conducted in September 2013 using the
PubMed and Web of Science databases; ToxNet, a toxicology database, was not used because
the focus of this review was on epidemiology studies. The search strings used in specific
1-1
-------
databases are shown in Table 1-1 and the search strategy is summarized in Figure 1-1, with
additional details of the process described below.
Table 1-1. Summary of search terms—asbestos, localized pleural
thickening, and pulmonary function
Database,
search date
PubMed
9/25/2013
No date restriction
Web of Science
9/25/2013
No date restriction
Merged
reference set
Terms
(("asbestos" [MeSH Terms] OR "asbestos" [All Fields] OR
"libby"[MeSH Terms] OR "libby"[All Fields]) AND ("pulmonary
function" [All Fields] OR "spirometry"[MeSH Terms] OR
"spirometry"[All Fields] OR FEV[A11 Fields] OR FVC[A11 Fields] OR
VC[A11 Fields] OR TLC[A11 Fields] OR "dyspnea" [All Fields]) AND
("pleural thickening" [All Fields] OR "pleural plaque" [All Fields] OR
"pleural plaques" [All Fields] OR "chest x-ray" [All Fields] OR
"radiographic"[All Fields] OR "computed tomography" [All Fields]
OR hrct[All Fields] OR profusion[All Fields])) AND
("humans" [MeSH Terms] AND English[lang])
Topic = ((asbestos AND ("pulmonary function" OR "spirometry" OR
FEV OR "forced expiratory volume" OR FVC OR "forced vital
capacity" OR VC OR "vital capacity" OR TLC OR "total lung
capacity" OR dyspnea) AND ("pleural thickening" OR "pleural
plaque" OR "pleural plaques" OR "chest x-ray" OR radiographic OR
"computed tomography" OR HRCT OR profusion)))
Duplicates eliminated through electronic screen (n = 47)
Additional duplicates eliminated through HERO (n = 58)
Hits
184
183
367
320
262
1-2
-------
Database Searches for Asbestos + LPT or Pleural Plaques + Pulmonary Function
{see Table 1-1 for keywords and limits)
Pubmed I I Web of Science
n = 184 (After duplicates removed)
.,/ n = 262 1.
Manual Screening For Pertinence
(Title/Abstract)
Excluded (n = 105)
105 Did not examine study question (pulmonary
function in relation to LPT or pleural plaques)
Manual Screening For Pertinence
(Title/Abstract/Full Text)
(n = 157)
Additional Search
Strategies
(review of references from
identified studies and
review articles)
Excluded (n = 127)
64 Did not examine study question
23 Pleural plaque group included DPT, "any" pleural
changes, or undefined pleural abnormalities
7 Includes asbestosis (without stratified analysis)
8 Additional methods details for an included study
10 Reviews or other background material
15 Miscellaneous
Primary Source Studies (/i = 30 + 8)
35 Cross-sectional
4 Longitudinal (includes 2 that also
have cross-sectional data}
1 Exercise protocol
M eta-An a lysis of FVC and FEV1 Percent Predicted, Cross-Sectional Studies,
Internal Comparison Group
20 Included
15 Excluded (10 external comparison only; 1 results presented as difference in absolute
values only; 1 sample sizes in relevant groups not reported; 3 quantitative results
not reported)
Figure 1-1. Summary of literature search for studies of relation between
localized pleural thickening (LPT) or pleural plaques and pulmonary
function.
1-3
-------
Based on the initial title and abstract screen, 58 additional duplicate citations were found
and 105 citations were excluded because they were not directly relevant to the study question
(e.g., no pulmonary measurements). The remaining 157 citations were selected for full-text
review by a group of three reviewers to determine whether any contained an analysis that
addressed the study question. Each paper was reviewed independently by two of the three
reviewers. In cases of disagreements or uncertainty (e.g., questions about the definition of
pleural abnormality used), the third reviewer also reviewed the paper and participated in the
consensus-building discussions. Studies were excluded at this step if the analysis group included
individuals with DPT or was based on undefined pleural abnormalities (n = 23), or if they
included individuals with parenchymal abnormalities (defined as x-ray profusion score greater
than 1/0, or high resolution computed tomography [HRCT] evidence of parenchymal
abnormality) without presenting a stratified analysis showing the results for the effect of pleural
plaques in the absence of asbestosis (n = 7). Thirty studies were selected for inclusion through
this process, and eight additional references were identified through (1) a review of references in
reviews and in the identified primary source studies and (2) by searching the Table of Contents
of relevant journals for newly released papers (September-December 2013) of selected journals
(American Journal of Industrial Medicine, Occupational and Environmental Medicine,
American Journal of Epidemiology, Epidemiology, Annals of Epidemiology) for a total of
38 primary source studies. In some instances, more than one publication presented data on the
same study participants or on a subset of the study participants, or provided additional
methodological details about a study. In these cases, these publications are treated as one related
set of studies (i.e., one entry in the summary tables and analysis). The references reviewed
through this process can be found on the Health and Environmental Research Online (HERO)
website (http://hero.epa.gov/Libby Amphibole Asbestos (Draft 2011)/).
In the next step of this review process, each of the selected studies was evaluated for
attributes related to study methods. Again, two of the three reviewers independently abstracted
information pertaining to selection of participants, protocols for x-ray or HRCT readings,
protocols for spirometry measurements, analytic approach, and consideration of smoking as a
potential confounder (see Table 1-2). This information was used to identify studies with
limitation(s) of sufficient magnitude to potentially affect the interpretation of the study results.
1-4
-------
Table 1-2. Information abstracted for initial study evaluation
Information abstracted
Notes regarding potential limitations
Study
participants
Geographic location
Source of exposure
Age
Duration of exposure
Time since first exposure (TSFE)
Smoking history
Current or retired workers
A short time since first exposure (i.e., <10 yr) or no information
on time since first exposure in a relatively young study
population (i.e., mean age <40 yr) considered a limitation, with
potential for "false negative" results (i.e., these studies would
miss an association that would be observed with longer
follow-up).
Imbalance in smoking prevalence between comparison groups
(i.e., pleural plaque vs. no pleural plaque groups) that was not
addressed in the analysis considered a limitation; impact on risk
estimate would depend on direction of the imbalance; similar
considerations for age, gender, and height if absolute values,
rather than predicted values, of pulmonary function parameters
were used.
Selection
process
Source, recruitment process
Exclusion/inclusion criteria
Comparison group: source,
recruitment, matching
Participation rates, final n
Clinic-based studies, studies based on recruitment for
medico-legal evaluations, or general screening studies with very
low participation rates (<20%) considered a limitation because
of concerns this process would result in differential selection
based on symptoms or other effects and exposure.
Measures:
x-ray or
HRCT
Type of x-ray views, number of
readers, training
Standards for classifying findings
[e.g., (ILO. 1980)1
Blinding to exposure and
medical history
Definition, size of pleural
abnormality group
Use of only one reader or of different readers in different
locations without discussion of training and reliability testing
considered a limitation because of concerns of outcome
misclassification resulting, in large studies, in attenuation of the
association of LPT with pulmonary function (direction of bias is
difficult to assess with small sample sizes); no information about
reading protocol also considered a limitation.
Lack of blinding to exposure history, medical history, and other
readings considered a limitation.
Measures:
spirometry
Protocol reference for
administration of pulmonary
function tests; number of
technicians, number of trials
Blinding to exposure and
medical history
Reproducibility (and use of
nonreproducible results)
Source of reference values or
equations
Use of absolute values, rather than predicted values, of
pulmonary function parameters considered a limitation (even if
adjustment for age, gender, and height was addressed in the
analysis) because it is difficult to compare to the majority of
studies reporting predicted values.
Lack of any details regarding procedures used in spirometry
considered a limitation, but no study provided all of the desired
details.
Analysis
Confirm that study includes
analysis of the association
between LPT and pulmonary
function measures with an
appropriate comparison group
Prevalence of smoking or mean
pack-yr by group; use of
smoking variable in the analysis
Analysis of "external comparison" only (i.e., comparison to an
unexposed referent group rather than an internal comparison to
an exposed referent group) or studies that provided pulmonary
function results (percentage predicted) for LPT or pleural plaque
group without a comparison group considered a limitation
because of issues of the comparability of the populations.
No adjustment for smoking when there is either no indication of
the degree of difference in smoking between groups or when
there was a large difference in smoking between groups (e.g.,
smoking prevalence >10% higher or mean pack-yr >10 pack-yr
higher in pleural plaque group) considered a limitation.
Other
Miscellaneous (e.g.,
discrepancies in sample size or
reported results)
1-5
-------
For the purpose of developing a summary effect estimate across studies, EPA considered
cross-sectional studies separately from longitudinal studies. Among the cross-sectional studies,
25 used an internal comparison group (i.e., comparison of pleural plaque versus no pleural
plaque groups among individuals with asbestos exposure), and ten included only an external
comparison group (i.e., the comparison was between asbestos-exposed individuals with pleural
plaques and people without asbestos exposure). Internal comparisons provide a better approach
to addressing issues of comparability and potential confounding (i.e., produce groups with
greater similarity with regards to exposure and other factors, such as smoking, socioeconomic
status, work status, and general health). Based on these considerations, the 10 studies with only
an external comparison group (Schneider et al., 2012; Chow et al., 2009; Sandrini et al., 2006;
Ameille et al.. 2004: Kilburn and Warshaw, 1991: Hillerdal 1990: Kilburn and Warshaw, 1990:
Hjortsberg et al., 1988: McLoud etal., 1985: Fridriksson et al., 1981) were not included in the
quantitative analysis.
The abstracted information relating to study methods are shown in a set of supplemental
tables included at the end of this appendix (see Supplemental Table I-A (cross-sectional studies,
internal comparison), Supplemental Table I-B (longitudinal studies), and Supplemental Table I-C
(cross-sectional studies, external comparison only).
1.1. ANALYSIS
After the initial evaluation of study attributes, the studies were again reviewed by sets of
two reviewers, focusing in more detail on the analysis and results. The reviewer assignments
allowed each of the three reviewers to have the responsibility for each of the papers either in the
initial abstraction of the methods details or of the results. The results were then displayed in
tabular form. Specific sets of studies were also displayed in graphical form, grouping results of
similar type (e.g., difference in percentage predicted [%predicted] FVC), as described in detail
below.
Each of the identified 20 cross-sectional, internal comparison studies that provided usable
data on (1) the number of individuals with and without pleural plaques and (2) mean values for
the respiratory measures of interest in each group were included in further analysis. Most, but
not all studies, also included either standard deviations (SDs) or standard errors (SEs) for these
estimates, as described below. Four studies reported vital capacity (VC) rather than FVC; these
four studies (Rui et al.. 2004: van Cleemput et al.. 2001: Singh etal.. 1999: Jarvholm and
Larsson, 1988) were included in the analysis together with the rest of the studies. In total,
15 x-ray studies and 5 HRCT studies were used for the analysis of mean difference in FVC;
10 x-ray studies and 5 HRCT studies were used for the analysis of mean difference in FEVi.
Summaries of the included studies are shown in Table 1-3; the five excluded studies are
summarized in Table 1-4, with reasons for exclusions noted.
1-6
-------
Table 1-3. Cross-sectional studies used in meta-analysis of mean difference in
percentage predicted forced vital capacity (FVC) or forced expiratory volume
(FEVi)
Reference, methods details
Results
X-ray studies
Bresnitz et al. (1993)
Philadelphia
Construction — elevator (union)
Selection bias: n total eligible not available
Information bias: x-rays — two B Readers,
blinded; spirometry — procedure reference,
no details
Confounding: internal comparison,
percentage predicted; excluded profusion
scores >1/0
Fiber type: unknown
Di Lorenzo et al. (1996)
Italy
Asbestos cement factory
Selection bias: 86% participation
Information bias: x-rays — two readers,
blinding not reported;
spirometry — procedure reference, some
details
Confounding: internal comparison,
percentage predicted; excluded profusion
scores >1/1
Fiber type: mixed
Duiic et al. (1993)
Croatia
Asbestos cement factory
Selection bias: 92% of current workers and
52% of retired workers participated
readers, blinded; spirometry — procedure
reference, some details
Confounding: internal comparison,
percentage predicted with additional
covariates; potentially inadequate
consideration of smoking; excluded
profusion scores < 1/1
Fiber type: mixed
From Table 2.
Mean (SD) percentage predicted, by group
FVC
FEVi
Bilateral and
unilateral
pleural
thickening
(» = 20)
85.8(10.6)
86.3(11.8)
No pleural
abnormalities
(» = 71)
89.4(16.2)
86.1 (19.7)
Mean
difference
-3.6
0.2
From Table 3.
Mean (SD) percentage predicted, by group
FVC
FEVi
Pleural plaques
(« = 10)
83.2(12.2)
76.5(14.3)
No bronchial,
parenchymal or
pleural disease on
x-ray (« = 9)
92.4(13.4)
86.9 (9.6)
Mean
difference
-9.2
-10.4
From Tables 2 and 4.
Mean (SD) percentage predicted, by group, unadjusted
FVC
FEVi
DLco
DLco (with
carboxyhemoglobin
correction)
Pleural plaques
(« = 55)
75.8 (12.7)a
86.8 (10.6)a
89.9(11.6)a
90.6 (12.6)a
No plaques
(« = 252)
92.2 (9.9)
89.0(12.0)
98.8(12.6)
96.8(12.7)
Mean
difference
-16.4
-2.2
-8.9
-6.2
aStatistically significant difference between groups with and without pleural
plaques; difference in FVC was also significant in model adjusting for exposure
and smoking.
N (%), by group
Restriction
Obstruction
Pleural plaques
(« = 55)
23(41.9)
4 (7.2)
No plaques
(« = 252)
41 (16.2)
23 (9.2)
RRa
(95% CI)
2.6(1.7,3.9)
0.80(0.23,2.2)
Restriction: FVC <80 %pred and FEV% >70%.
Obstruction: FEVi <80 %pred and FEV% <70%.
"Calculated by EPA.
RR = relative risk.
1-7
-------
Table 1-3. Cross-sectional studies used in meta-analysis of mean difference in
percentage predicted forced vital capacity (FVC) or forced expiratory volume
(FEVi) (continued)
Reference, methods details
Garcia-Closas and Christiani (1995)
Massachusetts
Construction — carpenters (union)
Selection bias: 1 6% of current workers and
3% of retired workers participated
Information bias: x-rays — two B Readers,
blinded; spirometry — procedure reference,
some details
Confounding: internal comparison,
percentage predicted with additional
covariates; excluded profusion scores >0/1
Fiber type: unknown
Hilt et al. (1987)
Norway
Asbestos-exposed workers
Selection bias: 96% of people with
abnormalities participated in repeat exam
Information bias: x-rays — departmental
radiologist followed by one B Reader,
blinding not reported;
spirometry — procedure reference some
details
Confounding: internal comparison,
percentage predicted with smoking variable;
did not discuss details of profusion scores
Other refs: Hilt et al. f!986b); Hilt et al.
(1986a)
Fiber type: unknown
Results
From Tables III
, IV, and V.
Mean (SD) percentage predicted, by group
FVC
FEVi
Pleural
plaques
(» = 64)
94.2 (14.7)
87.:
5(16.4)
No asbestosis
or DPT on
x-ray
(« = 457)
99.1 (12.0)
94.4(13.6)
p-value
(unadjusted, Mean
adjusted3) difference
(<0. 01, 0.11) -4.9
(0.01,0.13) -7.1
Prevalence, by group
Restriction (n =
4.2%)
Obstruction (n =
15.2%)
27,
= 96,
Mixed
(n = 24, 3.8%)
Pleural
plaques
»64)
5 (7.8)
10(15.6)
4 (6.5)
No asbestosis or
DPT on x-ray
(« = 457)
18(3.9)
42 (9.2)
6(1.3)
Adjusted OR
(95% CI)
1.27
(0.41,3.94)
1.03
(0.47, 2.22)
3.76
(1.45,12.33)
Restriction: FVC <80 %pred and FEV% >75%.
Obstruction: FEVi <80 %pred and FEV% <75%.
Mixed; FVC <80 %pred and FEVi <80 %pred and 60
-------
Table 1-3. Cross-sectional studies used in meta-analysis of mean difference in
percentage predicted forced vital capacity (FVC) or forced expiratory volume
(FEVi) (continued)
Reference, methods details
Jarvholm and Sanden (1986)
Sweden (Gothenburg)
Selection bias: n total eligible not available
(limited to nonsmokers)
Information bias: x-rays — one reader from
group of three chest physicians, blinding not
reported; spirometry procedure reference not
given, some details
Confounding: internal comparison,
percentage predicted; did not discuss details
of profusion scores
Fiber type: mostly chrysotile
Jarvholm and Larsson (1988)
Sweden (Gothenburg)
Asbestos-exposed workers
Selection bias: participation rate not
reported
Information bias: x-rays — one reader from
group of readers, blinding not reported;
spirometry — procedure reference not given,
some details
Confounding: internal comparison,
percentage predicted, stratified by smoking;
did not discuss details of profusion scores
Fiber type: unknown
Miller et al. (1992)
United States and Canada
Insulation workers
Selection bias: approximately 40%
participation; some information on mortality
by participation status
Information bias: x-rays — one B Reader,
some details
Confounding: internal comparison,
percentage predicted; potentially inadequate
consideration of smoking; stratified by
profusion score (O/- and 0/0)
Fiber type: mixed
Results
From Table 2 (no plaques) and Table 3 (Plaques).
Mean (SD) percentage predicted [subgroup «]
Pleural plaques
(« = 56)
Normal x-ray
(« = 88)
Mean
difference
FVC
Low
Heavy
Weighted average8
96.6 (10.8) [23]
90. 9 (12. 9) [33]
93.2(12.1)
100.0 (10.3) [54]
99.1 (14.0) [34}
99.7(11.9)
-3.4
-8.2
-6.5
FEVi
Low
Heavy
Weighted average*
108.0 (14.0) [23]
102.2 (17.5) [33]
104.6(16.2)
110.9 (13.1) [54]
110.7 (15.1) [34]
110.8(13.9)
-2.9
-8.5
-6.2
"Calculated by EPA.
From Table 5.
Mean (SD) percentage predicted
Current smokers3
VC
FEVi
Pleural plaques
(« = 53)
94.4(10.5)
103.1(13.4)
No pleural plaques
by x-ray (« = 425)
96.7(12.0)
102.6(14.6)
Mean
difference
-2.3
0.5
aData for former smokers and never smokers were not used because sample sizes
for these two groups were not reported.
From Table 3 (O/- and 0/0 groups).
Percentage predicted*
FVC
Circumscribed
pleural
thickening
(»=121)
86.8
No pleural
thickening
(« = 203)
89.8
Mean
difference
-3.0
aEPA assumed reported values are means; SD or SE not reported.
DPT definition required costophrenic angle blunting/obliteration.
1-9
-------
Table 1-3. Cross-sectional studies used in meta-analysis of mean difference in
percentage predicted forced vital capacity (FVC) or forced expiratory volume
(FEVi) (continued)
Reference, methods details
Miller et al. (2013)
United States (four states)
Selection bias: screening for medico-legal
evaluation
Information bias: x-rays — one B Reader,
no details
Confounding: internal comparison,
percentage predicted; potentially inadequate
consideration of smoking; stratified by
profusion score (0/0)
Fiber type: unknown
(Ohlson et al. (1985); Ohlson et al. (1984))
Sweden
Asbestos cement plant
Selection bias: 96% participation
Information bias: x-rays — one qualified
reader, blinding not reported;
some details
Confounding: internal comparison,
percentage predicted stratified by exposure
group; did not discuss details of profusion
scores
Fiber type: mostly chrysotile
Oliver et al. (1988)
United States (Pennsylvania)
Railroad workers
Selection bias:
Information bias: one B + one other reader,
blinding not reported;
details
Confounding: internal comparison,
percentage predicted stratified by exposure
duration and smoking status; excluded
profusion scores >0/1
Fiber type: unknown
Related reference: Oliver et al. (1985)
Results
From Table VI and Table VII.
Mean (SD) percentage predicted
FVC
DLco
LPT
group3
91.6(16.35)
89.5(21.68)
Normal x-ray
(« = 1,096)
96.6(15.87)
98.6(19.09)
Mean
difference
-5.0
-9.1
Calculated by EPA, based on sample-size weighted average of circumscribed only
(n = 290), and diaphragm (n = 83).
From Table 4 of Ohlson et al. (1985) (comb
constant proportion of pleural plaques acros
ine exposure categories, assuming
s exposure levels).
Mean percentage predicted (SD or SE not reported)3
FVC
FEVi
Pleural plaques
(« = 24)
97.8
97.0
No pleural plaques
(« = 51)
92.6
91.5
Mean
difference
5.2
5.5
Results support the statement in Ohlson etal. (1984) that pulmonary function
values among men with and without pleural plaques did not differ significantly
(quantitative results not reported).
"Calculated by EPA.
From text: smoking adjusted FVC = -4.3%
From Table II.
in — n n^n^v TTPVI — 9 i s in — n ^Q\
Mean (SD) percentage predicted
FVC
FEVi
DLco
FVC <80% [n]
Plaque
(« = 81)
86.0(0.17)a'b
80.3(21.3)a
97.0(21.3)
18.5[15]a
No plaque
(« = 278)
92.7(0.14)b
87.3(0.19)b
101.9(19.7)
9.0 [25]
Mean
difference
-6.7
-7.0
-4.9
RR (95% CI)b
2.1(1.1,3.7)
'"p <0.05 vs. no plaque.
bEPA noted that these SDs are considerably different from those reported in other
studies and so used imputed SD values for this study in the meta-analysis.
EPA used smoking adjusted results in the meta-analysis.
1-10
-------
Table 1-3. Cross-sectional studies used in meta-analysis of mean difference in
percentage predicted forced vital capacity (FVC) or forced expiratory volume
(FEVi) (continued)
Reference, methods details
Schwartz et al. (1990)
United States (Iowa)
Selection bias: 46% participation
Information bias: one experienced reader
(plus 10% validation study), blinded;
spirometry — procedure reference, some
details
Confounding: internal comparison,
percentage predicted; excluded profusion
scores >1/0
Fiber type: unknown
Related reference: Broderick et al. (1992)
Singh et al. (1999)
Australia
Selection bias: clinic-based recruitment
Information bias: one experienced reader,
blinding not reported;
no details
Confounding: small n\ internal comparison,
percentage predicted; potentially inadequate
consideration of smoking; did not discuss
details of profusion scores
Fiber type: unknown
Weill et al. (2011)
Montana (Libby)
Community-based
Selection bias: 79% participation
Information bias: x-rays — two out of three
B Readers consensus, blinding not reported;
spirometry — procedure reference, some
details
Confounding: internal comparison,
percentage predicted with additional
covariates; excluded profusion scores >1/0
Fiber type: LAA
Results
From Table 9 (excludes interstitial changes).
Percentage predicted8 (SD)
FVC
Circumscribed
pleural fibrosis No pleural fibrosis
(« = 178) (« = 797)
90.3(13.4) 94.7(16.8)
Mean
difference
-4.4
aEPA assumed reported values are means.
From Table 2.
Mean (SD) percentage predicted
VC
Pleural plaques No pleural disease
98.0(15.6) 101.2(10.6)
Mean
difference
-3.2
Calculated by EPA, based on reported SEs (4.5 and 4.0, respectively, for pleural
plaques and no pleural disease groups).
Pleural abnormality definition excludes costophrenic angle blunting/obliteration.
From Table 6, men.
Men
Never smokers
Ever smokers
Women
Never smokers
Ever smokers
From Table 4.
Beta (difference in percentage predicted) for
plaques compared with no abnormalities groups
-4.28
-4.43
(p<0.05)
(p<0.05)
Not reported
Not reported
(p>0.05)
(p>0.05)
Mean (SDa) percentage predicted
FVC
Circumscribed
pleural
thickening Normal
(« = 482) (« = 4,065)
95
.63(16.7) 103.15(15.9)
Mean
difference
-7.5
DPT definition requires costophrenic angle blunting/obliteration.
Calculated by EPA, based on reported SEs (0.76 and 0.25, respectively, for pleural
thickening and normal groups).
EPA also noted discrepancies between the text and Table 4 with respect to
definition of DPT and sample size in different groups.
EPA used Table 6 results for men in the meta-analysis.
1-11
-------
Table 1-3. Cross-sectional studies used in meta-analysis of mean difference in
percentage predicted forced vital capacity (FVC) or forced expiratory volume
(FEVi) (continued)
Reference, methods details
Zavalic and Bogadi-Sare (1993)
Croatia
Shipyard workers
Selection bias: participation rate not
reported
B Readers consensus, blinding not reported;
spirometry — procedure reference not
provided, some details
Confounding: internal comparison,
percentage predicted; excluded profusion
scores >1/0
Fiber type: unknown
Results
From Table 5 and Table 6.
FVC
FEVi
DLco
Mean (SD) percentage predicted, by group
Pleural plaques3
(« = 68)
88.4 (17.4)
85.7(13.6)
91.3(29.8)
No pleural plaques
(« = 101)
90.9(21.2)
86.0 (17.2)
90.1 (16.2)
Mean
difference
-2.5
-0.3
1.2
Calculated by EPA, based on sample-size weighted average within each table, and
then averaged across tables.
HRCT Studies
Clinetal. (2011)
France
Exposed workers (retired or inactive)
Selection bias: participation rate not
reported
blinded; Spirometry — procedure reference
not provided, multiple locations
Confounding: internal comparison,
percentage predicted with additional
covariates
Fiber type: unknown
Related ref: Paris et al. (2009)
Oldenburg et al. (2001)
Germany
Exposed workers
Selection bias: participation rate not
reported
not reported; spirometry — procedure
reference not provided
Confounding: internal comparison,
percentage predicted
Fiber type: unknown
From Table 3.
Mean (SD) percentage predicted, by group
FVC
FEVi
Isolated pleural
plaques (« = 403)
96.6(16.6)
97.9(19.4)
Normal CT scan
(« = 1,802)
100.4(16.6)
101.9(19.2)
Mean
difference
-3.8
-4.0
Adjusted for age, gender, body mass index (BMI), smoking, location of pulmonary
function testing, yr asbestos exposure, cumulative exposure index.
From Table 1 .
Mean (SD) percentage predicted, by group
FVC
FEVi
Plaques (« = 21)
88.8(13.89)
91.67(20.25)
No plaques
(« = 22)
89.89(11.86)
86.58 (28.09)
p-value for
difference
>0.05
>0.05
Mean percentage predicted, by group
Current and former
smokers
FVC
FEVi
Nonsmokers
FVC
FEVi
« = 16
86.5
86.28
n = 5
96.16
108.95
n= 15
86.97
78.76
n = l
96.13
103.36
Not reported
Not reported
Not reported
Not reported
Not reported
1-12
-------
Table 1-3. Cross-sectional studies used in meta-analysis of mean difference in
percentage predicted forced vital capacity (FVC) or forced expiratory volume
(FEVi) (continued)
Reference, methods details
Rui et al. (2004)
Italy
Referrals to an occupational medicine clinic
Selection bias: participation rate not
reported; participants had evidence of
pleural plaques on x-ray and subsequent
referral for HRCT
Information bias: HRCT — one reader,
blinding not reported.
Spirometry — procedure reference, some
details
Confounding: internal longitudinal
comparison, percentage predicted
Fiber type: unknown
Soulat et al. (1999)
France
Former nitrate fertilizer plant workers
Selection bias: 66.9% participation (48.6%
of all identified using company records)
Information bias: HRCT — one reader,
blinded; Spirometry — no procedure
reference
Confounding: internal comparison,
percentage predicted; potentially inadequate
consideration of smoking
Fiber type: mixed
van Cleemput et al. (2001)
Belgium
Asbestos cement factory
Selection bias: 83% participation
Information bias: Three readers, blinded;
used x-ray rather than HRCT to exclude
individuals with asbestosis from study
population
details
Confounding: internal comparison,
percentage predicted; potentially inadequate
consideration of smoking
Fiber type: mixed
Results
From Table 2 — Results from last follow-up visit.
Mean (SD) percentage predicted, by group
VC
FEVi
Plaques (« = 36)
90(10)
95 (14)
No plaques
(« = 67)
96(11)
102(13)
p-value for
difference
<0.05
0.05
From Table 4.
Mean (SE) percentage predicted, by group
FVC
FEVi
Plaques (« = 84)
110.2(2.03)
112.6(2.40)
No abnormalities
(« = 51)
108.9(2.60)
108.4(3.15)
p-value for
difference
Not reported
Not reported
From Table 3
Mean (SD) percentage predicted, by group
VC
FEVi
DLco
Plaques (« = 51)
110.5(13.4)
104.1 (12.9)
102.0(16.5)
No plaques
(« = 22)
109.8(14.9)
103.8(13.7)
97.2(15.5)
Mean
difference
0.7
0.3
4.8
CI = confidence interval; DLco = diffusing capacity of lung for carbon monoxide
1-13
-------
Table 1-4. Cross-sectional studies excluded from meta-analysis of mean
difference in percentage predicted forced vital capacity (FVC) or forced
expiratory volume (FEVi)
Reference, methods details
Results, reason for exclusion
X-ray studies
Bourbeau et al. (1990)
Canada (Quebec)
Construction — insulators (union)
Selection bias: 85% participation
Information bias: x-rays — two B Readers,
(1979) procedures with some details
Confounding: internal comparison, percentage
predicted with additional covariates
Rosenstock et al. (1988)
United States (Washington)
Plumbers and pipefitters
Selection bias: participation rate 20% in
Seattle, 7% in Tacoma
Information bias: x-rays — two readers,
blinded; spirometry — procedure reference not
reported, some details provided
Confounding: internal comparison, percentage
predicted; potentially inadequate consideration
of confounding
From Table 5 .
Mean difference (liters) in
absolute value, pleural plaques compared with no pleural plaques:
Difference (SE)
FVC -0.20
FEVi -0.35
(0.09)
(0.1)
(Excludes costophrenic angle obliteration and profusion >1/0; adjusted for age,
height, smoking, and parenchymal disease (based on Gallium-67 uptake
quantitation).
Excluded because results presented for absolute difference rather than
difference in percentage predicted; n for pleural plaques group after exclusions
not reported (approximately 50).
From Figure 4, profusion score O/- or 0/0:
Mean difference in percentage predicted FVC approximately 98 and 94%,
respectively in the no pleural disease and bilateral discrete groups.
Excluded because sample sizes in relevant groups not reported.
HRCT
Lebedova et al. (2003)
Czech Republic
Asbestos-processing plants
Selection bias: approximately 30% of random
selection from within groups defined on the
basis of x-rays taken in 2000
Information bias: HRCT — readers not
reported; blinding not reported
Confounding: internal comparison, adjusted
for smoking
From Table 5.
p- value
Pleural lesions Fibrosis
FVC 0.0019 0.0003
FEVi 0.0057 O.0001
Pleural — fibrosis interaction
0.0580
0.1498
Adjusted for smoking, chronic bronchitis, BMI, and ischemic heart disease.
Excluded because quantitative results not presented.
1-14
-------
Table 1-4. Cross sectional studies excluded from meta-analysis of mean
difference in percentage predicted forced vital capacity (FVC) or forced
expiratory volume (FEVi) (continued)
Reference, methods details
Results, reason for exclusion
Neri et al. (1996)
Italy
Exposed workers
Selection bias: 119/161 participated, reasons
for exclusion unlikely to be related to both
exposure and outcome;
Information bias: HRCT—two readers,
blinded to exposure; Spirometry—ATS
guidelines; Confounding: internal comparison
States that "No significant difference of pulmonary function tests was observed
between the subjects with pleural plaques detected on HRCT and workers with
normal pleura in absence of parenchymal involvement."
Excluded because quantitative results not presented.
Staples et al. (1989)
United States (California8)
Exposed workers
Selection bias: participation rate not reported
Information bias: two readers, blinded;
Spirometry—procedure reference not reported,
some details provided
Confounding: internal comparison
a Location not explicitly stated; EPA assumed
to be California based on affiliation of authors.
From text, page 1,507:
Analysis of "normal" group (n = 76) divided into with and without plaques;
VC and FEVi percentage predicted reported as "not significantly different" but
quantitative results not reported.
Excluded because quantitative results not reported.
ATS = American Thoracic Society.
Three of the included studies did not have the required data on pulmonary function for
the overall pleural-plaque and no-plaque groups, but did provide these data broken down by
another variable (exposure level or size of pleural plaque); for these three studies, data were
pooled across categories (weighted by number of individuals in each category) before inclusion
in the analysis (Zavalic and Bogadi-Sare, 1993; Jarvholm and Sanden, 1986; Ohlson et al.,
1985). Miller et al. (2013) used 1980 ILO guidelines but presented data for circumscribed
pleural plaques and plaques on the diaphragm separately. The combination (weighted average)
of these two groups was used in the analysis. Additionally, Ohlson etal. (1985) only reported
the overall number of individuals with and without pleural plaques, rather than numbers within
each category of exposure; thus, the number of individuals within each category was considered
proportional to the numbers in the entire study group.
Three studies did not provide SDs or standard errors for respiratory measures (Miller et
al.. 1992: Hilt etal.. 1987: Ohlson etal.. 1985). In addition, two studies (Weill et al.. 2011:
Oliver etal., 1988) reported overall SDs but did not present variance estimates for the
smoking-adjusted results; for these two studies, the smoking-adjusted results are used in the
meta-analysis (for Weill etal. (2011) results on males are used). For these five studies, SDs
were imputed as the linear average of reported SDs in other studies, weighted by sample size,
across the pleural-plaque and no-pieural-plaque groups. For Jarvholm and Larsson (1988), only
1-15
-------
data on smokers (in both the pieural-plaque and no-pleural-plaque groups) were used, because no
information was included on the number of former smokers and nonsmokers.
All of the x-ray studies used in these meta-analyses used the outcome of plaques as
defined by the 1980 ILO revision. The studies using HRCT, published between 1999 and 2011,
used a variety of descriptions to describe the pi eural-plaque group (see Supplemental Table I-A);
standardized guidelines for classification of pleural abnormalities identified using HRCT are not
currently available.
Data entry was performed independently by two people and any inconsistencies were
resolved by discussion and verification with the original study. All statistical analyses were
performed in R software; the R package Metafor (Viechtbauer, 2010) was used for conducting
the meta-analyses. Both x-ray and HRCT studies were included in the analysis. Analyses
stratified into these two groups were also conducted to investigate potential differences based on
detection method. HRCT has been reported to have greater sensitivity and specificity compared
to chest x-ray for the detection of pleural abnormalities [e.g., (Larson etal., 2014)]: only 50-80%
of cases of pleural thickening documented by HRCT are identified on x-ray (ATS, 2004). HRCT
is better able to differentiate such thickening from subpleural fat pads and identify parenchymal
abnormalities.
A random-effects model was used for both FVC and FEVi, as was done in a recent
meta-analysis (Wilken etal., 2011). That article examined the pulmonary effects of all types of
pleural abnormalities in combination, as well as the pulmonary effects of asbestos exposure in
the absence of any type of pleural abnormality. Summary estimates and the 95% confidence
intervals (CIs) are reported for each outcome.
All inferences are based on a comparison between exposed individuals with no
radiographic or HRCT abnormalities and exposed individuals with pleural plaques only (i.e.,
without any other radiographic or HRCT abnormalities). The outcomes are %predicted values
for FVC and FEVi, where predicted values are adjusted for age, gender, and height. The
potential confounding effects of smoking were addressed in various ways by 14 of the studies:
stratification (Oldenburg et al., 2001: Jarvholm and Larsson, 1988), adjustment (Clin etal., 2011:
Weill et al., 2011: Oliver et al., 1988), exclusion of ever smokers (Jarvholm and Sanden, 1986),
and indication that there was no or only a small difference in the smoking distribution between
groups (Rui et al., 2004: Di Lorenzo et al., 1996: Garcia-Closas and Christiani, 1995: Bresnitz et
al., 1993: Zavalic and Bogadi-Sare, 1993: Schwartz et al., 1990: Hilt etal., 1987: Ohlson et al.,
1985). Clin etal. (2011) and Weill et al. (2011) additionally controlled for the effects of body
mass index (BMI). One study (Ohlson et al., 1985) presented results stratified by exposure level,
and three studies (Clin etal., 2011: Di Lorenzo et al., 1996: Oliver etal., 1988) adjusted for a
cumulative asbestos exposure index or duration of exposure. These factors (smoking, BMI, and
asbestos exposure) were not measured in all studies, but the use of an internal comparison group
1-16
-------
(i.e., exposed workers) should have minimized differences in these factors when comparing
workers with no radiographic or HRCT abnormalities to workers with pleural plaques.
Among the studies identified for the meta-analyses, specific limitations pertaining to
participant selection, data collection, and analysis were noted as follows:
• Recruitment through clinic setting, or other attributes of recruitment, that may have
led to over selection of symptomatic individuals (Miller et al., 2013; Rui et al., 2004;
Singh et al.. 1999: Garcia-Closas and Christian!, 1995)
• Only one x-ray reader or different readers in different locations (without validation
sample), or lack of details about x-ray or HRCT reading protocol) (Miller et al., 2013;
Rui et al., 2004; Oldenburg et al., 2001; Singh et al., 1999; Soulatetal., 1999; Miller
et al., 1992; Jarvholm and Larsson, 1988; Jarvholm and Sanden, 1986; Ohlson et al.,
1985)
• Lack of blinding (or lack of reporting of blinding) of x-ray or HRCT readers to
asbestos exposure or medical history (Weill etal., 2011; Rui et al., 2004; Oldenburg
etal., 2001: Singh etal., 1999: Zavalic andBogadi-Sare, 1993: Jarvholm and
Larsson, 1988: Oliver etal., 1988: Hilt etal., 1987: Jarvholm and Sanden, 1986:
Ohlson etal., 1985)
Inadequate consideration of smoking as a potential confounder (Miller et al., 2013: van
Cleemput etal., 2001: Singh etal., 1999: SoulatetaL, 1999: Dujicetal., 1993: Miller et al.,
1992).
These 16 studies were not excluded from further consideration, but additional analyses
were conducted to evaluate the potential effect of these identified limitations on the results of the
meta-analyses.
1.2. RESULTS
1.2.1. Meta-Analyses of Cross-Sectional Studies
Figures 1-2 (FVC) and 1-3 (FEVi) show individual study results as well as the summary
effect estimates resulting from the meta-analyses. The summary effect estimates for both FVC
and FEVi are statistically significant, showing a change of-4.09 %pred (95% CI: -5.86, -2.31)
and -1.99 %pred (95% CI: -3.77, -0.22), respectively. The results of larger studies are very
consistent in showing a decrease in FVC (see Figure 1-2). In contrast, fewer large studies are
available for FEVi, and results are less consistent. The use of random-effect models was
supported for both pulmonary measures, as the tests for heterogeneity were statistically
significant, and the I2 was 80 and 57% for FVC and FEVi, respectively (where I2 represents the
proportion of the total variation across studies due to study heterogeneity instead of chance).
1-17
-------
DrA^rtit^ *1 do*?
Diczniiz i yyo
F^ i 1 f\ r*A n~rf\ *l QQ C^
ui Lorenzo iyyo
Dujic 1993
Garcia-Closas 1995
Hi It 1987
Javrholm 1986
Javrholm 1988
Miller 1992
Miller 2013
t~\Mf*r\n 'inQC
Unison lyoo
Oliver 1988
Schwartz 1 990
O I K-I/-I h 'IQQQ
oingn iyyy
Weill2011
Zavahc 1 993
Clin2011
/~*\\rl s*ivi^\i im OnO'l
uiQciiDurg ^uu i
Rui 2004
Ci-ii ilot -1 QQQ
oouiai iyyy
___,,
Van cieemput 2001
Summary Estimate
Xr*5\/ I •
-ray i •
Xnv I •
in r -i r n-i
-4.36 [ -5.99
1 o *-n r n ^ *~
\ -2.OU [ -o.oo
-3.80 [ -5.59
i -i no r ° °^
\ - \ .uy [ -o.oo
-6.00 [-10.20
• I 'i°nr|~'ip
• I 1 .oU [ -o. 1 o
• 1 n ~7r\ r p *~ ^
• 1 U./U [ -D.oo
-4.09 [ -5.86
2OQ 1
.00 J
2O~7 i
.Of \
-12.83]
, -1.13]
, 0.43]
, -2.47]
, 0.75]
, 0.01 ]
, -3.16]
IA r*o i
1 .DO J
, -1.01 ]
, -2.11 ]
8R-I 1
.O 1 J
, -2.73]
3OC 1
.00 J
, -2.01 ]
6KK: 1
.DO J
, -1.80]
7-7C 1
• ' " J
7QO 1
.yo J
, -2.31 ]
I I I I I I
-30.00 -20.00 -10.00 0.00 10.00 20.00
Mean Difference
Figure 1-2. Study-specific and summary effect estimates for change in
percentage predicted forced vital capacity (FVC) comparing
asbestos-exposed groups with and without localized pleural thickening (LPT)
or pleural plaques, x-ray and high resolution computed tomography (HRCT)
cross-sectional studies. Data are mean values; bars and values in brackets are
95% CI, size of data point is proportional to study weight.
1-18
-------
Breznitz 1993
J~\\ 1 f*\r^r\—rf*\ **! OOA
ui Lorenzo iyyo
Dujic 1993
Garcia-Closas 1995
Hilt 1987
Javrholm 1986
Javrholm 1988
Ohlson 1985
Oliver 1988
Zavalic 1993
Clin2011
Oldenburg 2001
n. • *~)f\r\ A
KUI 2004
c <*M il«^t **! odd
oouiai i yyy
Van Cleemput 2001
Summary Estimate
Xrn\/ I t
Xrnv I •
i ay i •
x-ray \-m-
x-ray | — • — |
x-ray !-•
x-ray | — • — |
x-ray |—
x-ray
x-ray \-m-
x-ray I — •
hrct W
Ki- - 1 I
hrct i
, ,
hrct I • I
L i I
hrct i
*
b i n ^n r r* ~* ^ ~*
1 i u.^u L ~o. / i , /
1 *i n ^ n r 0*1 0^5 n
l - 1 u.*HJ L ~£- I -^o , U
H -2.20 [ -5.37 , 0
-7. 10 [-11. 31 ,-2
H -0.80 [ -3.48 , 1
-6.20 [-11. 34, -1
•— I 0.50 [ -3.37 , 4
5.50 [ -0.21 ,
1 -2.10[ -5.00, 0
— I -0.30 [ -4.96 , 4
-4.00 [ -6.09 , -1
• I ^ OQ F Q ^O 1 Q
7 /-»/-» r HO CO H
.00 [ -12.53 , -1
4.20 [ -3.56 ,
-1.99[ -3.77, -0
H H n
11 J
46]
97]
89]
88]
06]
37]
O *1 1
21 ]
80]
36]
91 ]
oo 1
A "7 1
47]
96]
O*5 1
xJO 1
22]
I I I I I I
-30.00 -10.00 10.00
Mean Difference
Figure 1-3. Study-specific and summary effect estimates for change in
percentage predicted forced expiratory volume (FEVi) comparing
asbestos-exposed groups with and without localized pleural thickening (LPT)
or pleural plaques, x-ray and high resolution computed tomography (HRCT)
cross-sectional studies. Data are mean values; bars and values in brackets are
95% CI, size of data point is proportional to study weight.
1-19
-------
Analyses of x-ray and HRCT studies separately are shown in Figures 1-4 (FVC) and 1-5
(FEVi). For both measures of lung function, the results for x-ray and HRCT studies considered
separately are quite similar in magnitude to overall results (combining the two study types). For
FVC, results from both HRCT and x-ray studies considered as separate sets are statistically
significant: -3.30 %pred (95%CI: -5.25,-1.34) and-4.55 %pred (95% CI: -6.73,-2.38),
respectively. FEVi results for HRCT and x-ray studies considered separately were very similar
in magnitude to the combined results but are not statistically significant: -1.96 %pred (95% CI:
-6.01; 2.09) and -1.87 %pred (95% CI: -3.96, 0.23), respectively. Given that the overall
(combined) results for FEVi are statistically significant, this discrepancy is likely due to the
smaller sample sizes when x-ray and HRCT studies are separated.
1-20
-------
D t*A ~m\^^ "1 Q Q *3
Diczniiz iyyo
F^ i 1 f\ *"* n-rr\ *l Q Q C^
ui Lorenzo iyyo
Dujic 1993
Garcia-Closas 1995
Hilt 1987
Ja'vrholm 1986
Ja'vrholm 1988
Miller 1992
Miller2013
/^hiloi-in -1QQ£
unison iyoo
Oliver 1988
Schwartz 1990
O I r-i/-i h 'IQQQ
oingn iyyy
Weill2011
Zavalic 1993
Xn\/ i •
-lay i •
Xfn \ i i _
-ray i •
x-ray i • i
x-ray i — • — i
x-ray i •
x-ray i — • — i
x-ray i — •—
x-ray i •
x-ray m
X_ro\/ 1
lay i
x-ray i — • — i
x-ray I-BH
x-ray I-BH
Xr*5 \f i •
-lay 1 •
1 'J C.H F Q ^Q
1 -j.ou [ -y.oo
i Q on r on ~7~7
1 -y.^U [ ~^U. / /
-16.40 [-19.97,
-4.90 [ -8.67
i -2.60 [ -5.63
-6.50 [-10.53
H -2.30 [ -5.35
-3.00 [ -6.01
-5.00 [ -6.84
• i ^ ^n r -i '"*r!
• ' D.zU [ - 1 .zo ,
-4.30 [ -7.59
-4.40 [ -6.69
i ° ^n r -i r n-i
-4.36 [ -5.99
i o r r\ r *"* ^ r
1 -2.50 [ -8.35
2OQ
.00
2O~7
.O /
-12.83
, -1.13
, 0.43
, -2.47
, 0.75
, 0.01
, -3.16
1-1 CO
1 .DO
, -1.01
, -2.11
8R-I
.D I
, -2.73
3f\ f-
.35
Summary Estimate
-4.55[ -6.73 , -2.38]
i i i i i i
-30.00 -20.00 -10.00 0.00 10.00 20.00
Mean Difference
Clin2011
LJIQcMDUrg ZUU I
DI M onn/i
KUI ZUU'I-
Ci-ii ilot -1 QQQ
oouiai iyyy
van uieemput zuu i
Summary Estimate
hrct '-•-1
hi ct
hrct '
-3
i ~i
1 - 1
c
-D.
• i 1
• ' I
• i n
-3
.80
no
.uy
nn r
JU [
on
.oU
~7r\
. i\j
.30
[ -5.59
[R SO.
-o.oo
•i n on
- I U.ZU
[C -I C
-O. I D
[R I^1?
-D.OO
[ -5.25
,-2.01
6R^
.DO
•i sn
, - 1 .OU
7-7C
• i o
7 en
, / .yo
,-1.34]
I I I I I I
-15.00 -5.00 0.00 5.00 10.00
Mean Difference
Figure 1-4. Study-specific and summary effect estimates for change in
percentage predicted forced vital capacity (FVC) comparing
asbestos-exposed groups with and without localized pleural thickening (LPT)
or pleural plaques, for x-ray (top panel) and high resolution computed
tomography (HRCT) (bottom panel) cross-sectional studies. Data are mean
values; bars and values in brackets are 95% CI, size of data point is proportional
to study weight.
1-21
-------
Brezmiz i yyo
ui Lorenzo iyyo
Dujic 1993
Garcia-Closas 1995
Hi It 1987
Javrholm lyob
Javrholm 1988
Ohlcnn -1 QR^
unison lyoo
Oliver 1988
/.avanc i yyo
Summary Estimate
x-ray i— •-
x-ray i-l
x-ray i —
X_ra\/ L
x-ray i— •-
<^
• i n on r R 7-1
' i U.ZU [ -D.( 1
j -i n /in r o-i OR
1 - 1 U.4U [ -/i 1 ./iD
H -2.20 [ -5.37
-7.10 [-11.31
H -0.80 [ -3.48
.^0 [ -1 1 .o4
•— i 0.50 [ -3.37
• i ^ ^n r n o-i
H -2.1 0[ -5.00
i n T\ r A QR
1 i -u.ou [ -4.00
-1.87[ -3.96
~7 -\ -\ ^
, /Ml J
0/1R 1
.40 J
, 0.97]
, -2.89 ]
, 1.88]
.Ob J
, 4.37]
•\ -i o-i ^
\ \ .£. \ \
, 0.80]
4*?R ^
.00 j
, 0.23]
-30.00 -20.00 -10.00 0.00 10.00 20.00
Mean Difference
Clin2011
uiQenuury zuu i
DI M onn/i
KUI ZUU4
Ci-ii ilot -1 QQQ
oouiai iyyy
van oieempui zuu i
Summary Estimate
hrct '-•-'
hi ct
hrct '
hrct '
-^^M
-4.00 [ -6.09
, r no r o r;n
o.uy [ -y.ou
7 nn r -i o ^o
-( .UU [ - IZ.Oo
h i nonrRyio
• ' U.oU [ -D.'l-o
- -1.96 [-6.01
,-1.91 ]
-1 Q RQ 1
i y.oo j
1 47 1
1-1 QC 1
i .yo j
7 no i
.Uo J
, 2.09]
-20.00 -10.00 0.00 10.00
Mean Difference
20.00
Figure 1-5. Study-specific and summary effect estimates for change in
percentage predicted forced expiratory volume (FEVi) comparing
asbestos-exposed groups with and without localized pleural thickening (LPT)
or pleural plaques, for x-ray (top panel) and high resolution computed
tomography (HRCT) (bottom panel) cross-sectional studies. Data are mean
values; bars and values in brackets are 95% CI, size of data point is proportional
to study weight.
1-22
-------
There were no clear asymmetries in the examination of funnel plots for all the analyses
(although the HRCT analyses had few data points) suggesting that publication bias in not an
issue in these analyses. Exclusion of all of the studies with the limitations noted previously
(16 in the FVC meta-analysis and 12 in the FEVi analysis) resulted in more consistent results
(narrower CI despite a smaller number of studies) with a summary effect estimate of
-4.08 %pred (95% CI: -5.44; -2.71) for FVC (based on four studies: (Clinetal.. 2011: Di
Lorenzo et al.. 1996: Bresnitz et al.. 1993: Schwartz et al.. 1990)) and an effect for FEVi that is
almost doubled compared to the full set analysis (-3.87 %pred, 95% CI: -5.84; -1.90) (based
on three studies: (Clinetal., 2011; Di Lorenzo et al., 1996; Bresnitz et al., 1993)). In addition,
examination of the studies excluded because of analysis or reporting issues (see Table 1-4)
indicates that the results of this additional set of studies are also consistent with the pattern see in
Figures 1-2 and 1-3, with three of the five studies in Table 1-4 indicating a decrement in FVC in
the pieural-plaque group, compared with the no-pieural-plaque group (two studies did not state
whether there was a decrease or increase).
Of the five studies (Miller etal.. 2013: van Cleemput et al.. 2001: Dujicetal., 1993:
Zavalic and Bogadi-Sare, 1993; Oliver et al., 1988) that also reported diffusing capacity (DLco),
only two (Zavalic and Bogadi-Sare, 1993; Oliver etal., 1988) did not have potential limitations
related to adjustment for smoking. (Oliver et al., 1988) showed a borderline statistically
significant (p = 0.055) decrease in DLco (-4.9 %pred), while Zavalic and Bogadi-Sare (1993),
showed a slight (statistically nonsignificant) increase in DLco (1.2 %pred) for individuals with
pleural plaques relative to those without pleural plaques.
1.2.1.1. Relationship Between Pulmonary Function Measures and Extent of Pleural Plaques
Four cross-sectional studies also presented analyses of the extent of pleural plaques in
relation to degree of decrement in pulmonary function (Clin etal., 2011; van Cleemput et al.,
2001: Zavalic and Bogadi-Sare, 1993: Lilisetal., 1991b). Lilisetal. (1991b) is a publication
related to the Miller et al. (1992) study included in the meta-analysis, and so is not counted as a
separate primary study in the literature search results. In Clinetal. (2011), the decrease in FVC
seen with increasing maximum cumulative plaque extent was statistically significant, and for
FEVi the decrease was marginally significant (p = 0.06); there was a difference of approximately
-4 %pred in both FVC and FEVi when comparing the lowest to the highest plaque extent
category. In Lilis et al. (1991b), a higher index score (indicating increased pleural plaque size)
was significantly associated with a larger decrement of 5-10 %pred FVC (accounting for
smoking and time since first exposure) than was a low index score, van Cleemput et al. (2001)
reported a statistically nonsignificant decrease in both %predicted VC and %predicted FEVi with
increasing total surface area of pleural plaques; however, on average those with pleural plaques
had slightly better lung function than those without pleural plaques. Although van Cleemput et
al. (2001) concluded that neither the presence nor the extent of the plaques was correlated with
1-23
-------
pulmonary function parameters, this is a small study of only 73 workers compared to more than
2,000 workers in the study by Clin et al. (2011): these are both HRCT studies. Zavalic and
Bogadi-Sare(1993) reported that %predicted FVC and %predicted FEVi both tended to decrease
with increased plaque length. Additionally, the longitudinal study by Sichletidis et al. (2006)
demonstrated that after 15 years of follow-up, the total surface area of pleural plaques increased
twofold and pulmonary function was statistically significantly decreased over that period.
Although increased plaque surface area was not statistically significantly associated with the
observed reductions in %predicted FVC or %predicted FEVi, the reduction in total lung capacity
(TLC) was associated with plaque surface area (r = -0.486,/? = 0.041). Taken together, these
studies strongly suggest that the extent of the decrease in pulmonary function is associated with
the extent (size or total surface area) of pleural plaques.
1.2.1.2. Analysis by Categorical, Rather Than Continuous Measures of Pulmonary Function
Three studies presented analyses in terms of difference in the proportion of individuals
within a group below a specified value for the pulmonary function test or combination of tests.
In Oliver et al. (1988), the proportion with FVC <80 %pred was approximately doubled in the
pleural-plaque group (18.5%) compared with the group with no pleural plaques (9.0%) (relative
risk: 2.1, 95% CI: 1.1, 3.7); the smoking-adjusted mean difference between these two groups
was -4.3 %pred FVC. Restrictive disease was defined slightly differently in other studies.
Garcia-Closas and Christiani (1995) observed a statistically nonsignificant increase in the
proportion classified as having restrictive disease (defined as FVC <80 %pred and FEVi/FVC
>75%), from 3.9% in the group with no pleural plaques to 7.8% in the pleural plaques group. In
Dujic etal. (1993), the estimated relative risk for restrictive disease (defined as FVC <80 %pred
and FEVi/FVC >70%) in the group with pleural plaques, compared to the group with no pleural
plaques, was 2.6 (95% CI 1.7, 3.9); the results in terms of mean difference in %predicted FVC
between groups were notably larger than that of other studies in Figure 1-2. The relative risks for
obstructive disease in these studies were close to 1.0 (indicating no difference in those with
plaques compared to those without pleural plaques); obstructive disease was defined as
FEVi <80 %pred and either FEVi/FVC <70% (Duiic et aL 1993) or FEVi/FVC <75% (Garcia-
Closas and Christiani, 1995). However, the increase in the proportion of individuals with
mixed-pattern disease (FVC and FEVi <80 %pred, and 60%
-------
with any evidence of radiograph!c asbestosis (i.e., evaluated only those with ILO profusion
scores of 0/0). These included three x-ray studies (Garcia-Closas and Christian!, 1995; Dujic et
al., 1993; Oliver et al., 1988), with two additional studies presenting data for the 0/0 profusion
category separately (Miller et al., 2013; Miller et al., 1992). The resulting meta-analyses of FVC
(five studies) and FEVi (three studies) produced statistically significant summary estimates that
were noticeably larger than in the primary analysis: -6.66 %pred (95% CI: -11.37; -1.96) for
FVC and -3.46 %pred (95% CI: -6.37; -0.61) for FEVi.
FIRCT may be more sensitive than x-ray as a test used to exclude individuals with
parenchymal abnormalities [e.g., (Lebedova et al., 2003; Jankovic et al., 2002; Simundic et al.,
2002)]; note that in some cases, these studies identified parenchymal abnormalities detected
using HRCT even in the group with normal (i.e., 0/0) x-ray profusion scores. In a study of
162 subjects without radiographic (ILO 0/0) evidence of parenchymal fibrosis, Lebedova et al.
(2003) found parenchymal changes were detectable in the HRCT scans of 46.3% of the
participants. Asbestosis was found in 17 (10.5%) persons and suspected asbestosis in 58
(35.8%). Furthermore, parenchymal abnormalities were significantly more frequent in the
subjects with pleural lesions than in those without pleural lesions (67.0% versus 15.4%,
p < 0.0001). Analysis of HRCT studies alone showed that undetected parenchymal changes in
x-ray examinations (but which would be detectable using HRCT) are not likely to explain the
observed effects on pulmonary function. As shown in Figures 1-4 and 1-5, the decrease in FVC
observed in HRCT studies was somewhat smaller than that shown in x-ray studies (although still
statistically significant); for FEVi there was little difference in the effect size, although this
estimated effect was not statistically significant in the smaller set of HRCT studies.
1.2.2. Analysis of Longitudinal Studies
Longitudinal studies provide a basis for evaluating the progression of LPT or pleural
plaques over time, as seen by an increase in the extent of pleural plaques or thickening and a
corresponding increase in pulmonary function deficits with the passage of time. Only four
longitudinal studies were found in the literature search. The mean length of follow-up varied
among these studies from 3.7 to 15 years, with the longer follow-up periods providing evidence
supporting an association between pleural plaques and increased rate or degree of pulmonary
impairment (see Table 1-5). The presence of pleural plaques was not related to differences in
decline in FVC or FEVi measures in the studies with the shortest follow-up (3.7-4 years) (Rui et
al., 2004; Ohlson et al., 1985). In a case-control study with a 7-year follow-up, decreases in
FVC of 31 ± 12 (mean ± SE) and 15 ± 6 mL/year were seen in those with and without pleural
plaques, respectively, but this difference between groups was not statistically significant
(Ostiguy et al., 1995). In the small study of people with plaques only, but with the longest
follow-up period, the size of pleural plaques grew more than twofold (from 8.5 to 17.2 cm2) over
approximately 15 years (Sichletidis et al., 2006), and there was a large and statistically
1-25
-------
significant decrease of 14.6 %pred FVC and 4.3 %pred FEVi over the follow-up period. A
statistically significant association was not seen between the declines in FVC and FEVi and the
increase in plaque surface area, but was seen between TLC decline and plaque surface area. In
Sichletidis et al. (2006), the use of percentage predicted values accounts for the expected decline
due to increased age over the follow-up period. In addition, the observed pulmonary decrements
are unlikely to be the result of continued asbestos exposure. Ostiguy et al. (1995) stated that
additional exposure during the follow-up period was low, while Sichletidis et al. (2006) stated
that there was no additional exposure during the follow-up period.
1-26
-------
Table 1-5. Longitudinal studies examining forced vital capacity (FVC) or
forced expiratory volume (FEVi)
Reference, methods details
Ohlson et al. (1985)
Sweden
Asbestos cement plant
4 yr follow-up; no continuing exposure
Selection bias: 96% participation
Information bias: x-rays — one qualified
reader, blinding not reported;
spirometry — procedure reference not given,
some details
Confounding: internal comparison, adjusted
for covariates
Related reference: Ohlson etal. (1984)
Rui et al. (2004)
Italy
Referrals to an occupational medicine clinic
3. 7 yr follow-up
Selection bias: participation rate not reported;
participants had evidence of pleural plaques on
x-ray and subsequent referral for HRCT
Information bias: HRCT — one reader,
blinding not reported. Spirometry — procedure
reference, some details
Confounding: internal longitudinal
comparison, percentage predicted
Ostiguv et al. (1995)
Canada
Copper refinery; asbestos removal and
threshold limit values not exceeded over study
period
7 yr follow up
Selection bias: loss to follow-up not reported
Information bias: x-rays — two experienced
readers, blinded; spirometry — Renzetti (1979)
procedures, some details
Confounding: internal comparison
Sichletidis et al. (2006)
Greece (residential exposure); no continuing
exposure
1 5 yr follow-up
Selection bias: 78% follow-up
readers, blinding not reported; spirometry
procedure reference not given, some details
Confounding: internal comparison
Related reference: Sichletidis et al. (1992)
Results
From Table 6.
Adjusted percent decline (compared with baseline assessment)
FVC
FEVi
Pleural plaques
(« = 24)
6.34
6.43
No pleural
plaques (« = 50)
6.74
7.39
Mean difference
in amount of loss
0.40
0.96
Adjusted for height, age, tracheal area, cumulative exposure, and smoking.
From Table 2.
Mean (SD) percentage predicted, by group
VC
FEVi
First examination
No
Plaques plaques
(« = 36) (« = 67)
91(10) 97(10)
97(13) 103(12)
Mean
reduction
(95% CI)
for those
with
Last examination plaques3
No
Plaques plaques
(« = 36) (« = 67)
90(10) 96(11) -3.4
(-7.9,1.0)
95(14) 102(13) -1.5
(-7.1,4.0)
aAdjusted for smoking habit and seniority.
From Table 7.
Mean SEM annual loss (mL/yr)
FVC
Pleural plaques
(« = 51)
31 (12)
No pleural
plaques (« = 211)
15(6)
Mean difference
in rate of loss
16 mL/yr
From Table II.
Mean (SD) among people with pleural plaques (« = 18)
FVC %predicted
FEVi %predicted
1988
94.74 (17.98)
93.43(13.56)
2003
80.12(13.76)
89.1 (10.84)
Difference, 2003
minus 1988
-14.62
-4.33
SEM = standard error of the mean.
1-27
-------
1.3. DISCUSSION
This systematic review demonstrates statistically significant decrements of 4.09 %pred
FVC (95% CI: 2.31, 5.86) and 1.99 %pred FEVi (95% CI: 0.22; 3.77) in people exposed to
asbestos with pleural plaques relative to exposed people with no pleural plaques. This analysis
compared exposed persons with and without pleural plaques, so that while the total decrement in
either group could be due in part to asbestos exposure alone, the estimated difference between
the groups should primarily reflect the decrement due to pleural plaques. In the meta-analysis by
Wilken et al. (2011), asbestos-exposed workers without radiologic abnormalities showed
decrements in lung function (i.e., values below 100% of predicted: %pred FVC 95.7 and %pred
FEVi 93.6). Thus, the lung function decrements associated with pleural plaques in this analysis
are even more pronounced when compared to 100%pred, or normal lung function.
Cross-sectional studies suggest that an increased extent of pleural plaques is associated
with greater decrements in pulmonary function. Two smaller studies (van Cleemput et al., 2001;
Zavalic and Bogadi-Sare, 1993) both found a tendency for pulmonary function to decrease with
plaque size, although these associations were not statistically significant. The association
between plaque size and pulmonary function was statistically significant for %predicted FVC in
two larger studies (Clinetal.. 2011: Lilisetal.. 1991b).
Few longitudinal studies are available, and two of these had very short follow-up periods
of <5 years (Rui et al., 2004: Ohlson et al., 1985). In a case-control study with intermediate
follow-up (7 years) and subjects matched on age, Ostiguy et al. (1995) observed a tendency for a
more rapid decline in FVC in individuals with pleural plaques compared to those without pleural
plaques (31 ± 12 [mean ± SE] versus 15 ± 6 mL/year, respectively). The study with the longest
follow-up period [15 years, (Sichletidis et al., 2006)1 was conducted among people exposed to
asbestos in a community setting from whitewash material, rather than an occupational setting.
This small study observed a statistically significant decrease in percentage predicted FVC and
FEVi of 14.6 %pred and 4.3 %pred, respectively, among individuals with pleural plaques.
Although these decreases were not significantly associated with the increase in plaque surface
area, the reduction in TLC over the 15-year period was significantly associated with the increase
in plaque surface area (r = -0.486,/? = 0.041).
The analysis of HRCT studies alone showed similar results to both the overall
(combined) results, and to the x-ray studies alone. Thus, undetected parenchymal abnormalities
that could be detected by HRCT are unlikely to influence observed decrements. It is also
unlikely that the observed association between pleural plaques and decrements in pulmonary
function can be explained by the independent effects of asbestos exposure. The largest HRCT
study (Clin et al., 2011) controlled for cumulative exposure, as well as other potential
confounders, and demonstrated significant pulmonary function decreases consistent with our
summary effect estimate. Similar results were obtained in a large x-ray study (Oliver et al.,
1988) that controlled for duration of exposure. A smaller study that stratified for exposure
1-28
-------
observed a tendency for better lung function among workers with versus without pleural plaques
(Ohlson et al., 1985). Overall, these results indicate that differences in asbestos exposure are
unlikely to fully explain the observed differences in lung function. It is possible, however, that
people more sensitive to the effect of asbestos exposure, given the same level of exposure,
develop pleural plaques and also have a larger decrease in pulmonary function. In that case,
plaques may not be the cause of the decrease in pulmonary function, but are a marker for
susceptibility to pulmonary effects of asbestos.
Specific aspects of the design or analysis of these studies indicate that the demonstrated
association of pleural plaques and pulmonary function decrease are unlikely to be explained by
other causes of pulmonary function loss, such as demographic characteristics, smoking, or other
lung disease. Height, age, and gender were accounted for by use of percentage predicted values
that incorporate these variables. The sensitivity analysis addressed limitations or potential biases
noted through a systematic review of study methods conducted prior to evaluation of the results,
including limitations in the way in which smoking was addressed and lack of an explicit
statement that some kind of blinding procedure was used for the reading of the x-ray or HRCT.
In this sensitivity analysis, pulmonary decrements were essentially the same for FVC or
increased almost twofold for FEVi compared with the analysis that included all of the studies,
and the decrements remained statistically significant. Medical reasons for pulmonary decrease
were explicitly accounted for through exclusion of individuals with lung diseases in seven
studies (ClinetaL 2011; Rui et al., 2004; Singh etaL 1999; Zavalic and Bogadi-Sare, 1993;
Jarvholm and Larsson, 1988; Hilt et al., 1987; Jarvholm and Sanden, 1986). Because this type of
exclusion is a common practice, it may have been performed but not mentioned in some papers
because reporting of details of the participant recruitment and selection process was often
limited.
LPT was introduced as a term in the 2000 ILO guidance. LPT includes plaques on the
chest wall and at other sites (e.g. diaphragm). Plaques on the chest wall can be viewed either
face-on or in profile. A minimum width of about 3 mm is required for an in-profile plaque to be
recorded as present according to the 2000 ILO guidance. Although no studies reported results
for plaques with width of at least 3 mm (i.e., LPT), one large study Clin et al. (2011) (reported
results for plaques less than 2 mm and found that those with such plaques had at least 100%pred
FVC and FEVi. Thus, results of analysis for pleural plaques in this Appendix can be seen to
apply to LPT.
EPA considered the potential for BMI to affect the observed associations between pleural
plaques and lung function. Directionally consistent with this potential bias, a tendency for
greater FVC decrements in the x-ray studies (4.55%) relative to the HRCT studies (3.30%) was
seen. Given that the FVC loss is still observed in the HRCT studies; however, the associations
cannot be fully explained by the effects of BMI. Of the two studies that included BMI in their
analyses, one study, using x-rays, observed a slightly higher BMI in people with plaques (mean
1-29
-------
30.3 and 28.5 kg/m2, respectively for with and without plaques), and higher BMI and age were
significantly related to decrements in FVC (Weill et al., 2011): EPA used the BMI- and age-
adjusted results in its meta-analysis. The other study used HRCT, and observed similar mean
BMI between individuals with and without pleural plaques (27.7 and 27.4 kg/m2, respectively)
(Clin et al., 2011). More generally, the prediction of FEVi and FVC is not improved by
considering weight after taking into account height, age, race, and sex in cross-section analyses
of lung function (Hankinson et al., 1999). EPA does not believe large differences in BMI by
radiographic group are likely in the remaining studies examined, and overall, does not believe
that the observed associations between pleural plaques and lung function decrements are biased
by an effect of BMI.
With regard to fiber type, 13 studies did not report fiber type of asbestos exposure, four
reported mixed exposure, two reported mostly chrysotile exposure, and one reported LAA
exposure. Thus, EPA could not examine fiber characteristics in this analysis. However, EPA is
not aware of any studies of pleural plaques and lung function that indicated potential differences
in association by fiber type. Only study of exposure to LAA (Weill et al., 2011) showed
decrease in %pred FVC (this study did not report FEVi) comparable to overall result (4.36 vs.
4.09). Moreover, the results from the studies included in this meta-analysis did not display great
variability, although it is likely that study populations were exposed to different fiber types (or
mixtures).
The impact of the observed decrease in pulmonary function should be considered on both
the individual level and the population level. At the individual level, the decrement in FVC or
FEVi may or may not have a noticeable effect for a given patient. There have been such
statements regarding the impact of deficits associated with plaques as defined by the ILO 1980
guidelines. The American Thoracic Society (ATS, 2004) stated that "Although pleural plaques
have long been considered inconsequential markers of asbestos exposure, studies of large cohorts
have shown a significant reduction in pulmonary function attributable to the plaques, averaging
about 5% of FVC, even when interstitial fibrosis (asbestosis) is absent radiographically.
Decrements, when they occur, are probably related to early subclinical fibrosis." However, the
analyses of x-ray and HRCT studies individually (see Figures 1-4 and 1-5) suggest that
subclinical fibrosis does not fully explain the observed associations between pleural plaques and
pulmonary function decrements. The ATS document (ATS, 2004) went on to state that "There is
a significant but small association between the extent of circumscribed pleural plaques and FVC,
which is not seen with diffuse pleural thickening. Even so, most people with pleural plaques
alone have well-preserved pulmonary function." In addition to the ATS document, the American
College of Chest Physicians (ACCP: Banks et al., 2009) published a Delphi study conducted to
gauge consensus among published asbestos researchers, and found that these researchers
statistically rejected the statement that "Pleural plaques alter pulmonary function to a clinically
1-30
-------
significant degree" (although noting that some researchers strongly agreed with the statement,
and the response rate was relatively low at <40%).
At the population level, ATS (2000) stated that "any detectable level of permanent
pulmonary function loss attributable to air pollution exposure should be considered as adverse"
and that
"It should be emphasized that a small but significant reduction in a population
mean FEVi, or FEVo.75, is probably medically significant, as such a difference
may indicate an increase in the number of persons with respiratory impairment in
the population. In other words, a small part of the population may manifest a
marked change that is medically significant to them, but when diluted with the
rest of the population the change appears to be small."
Thus, even small changes in the average (mean) of a distribution of pulmonary function
parameters can result in a much larger proportion of the exposed population shifted down into
the lower "tail" of the pulmonary function distribution. In the study by Oliver et al. (1988), a
doubling (18.5% in pleural plaque group and 9% in no pleural plaque group, relative risk: 2.1,
95% CI: 1.1, 3.7) of the proportion of individuals with a <80 %pred FVC was seen among
people with pleural plaques; there was a group mean difference (smoking-adjusted) of 4.3 %pred
FVC. A similar situation is seen in the example of early childhood exposure to lead and
decrements in intelligence as measured by IQ (U.S. EPA, 2013). A mean deficit of 2 IQ points
would not be expected to be "clinically relevant" for an individual, but from a population
perspective, a downward shift of the entire IQ distribution by 2 IQ points would be quite
significant. By a similar argument, a shift in distribution of pulmonary function would result in a
considerable increase in the proportion of individuals with a significant degree of pulmonary
impairment below a clinically adverse level. At the same time, this shift would reduce the
proportion of individuals with high pulmonary function.
In summary, pleural plaques (and LPT) represent a persistent structural alteration of the
pleura. The statistical association between pleural plaques and LPT and decrements in
pulmonary function identified herein is consistent with plaques being an indicator of asbestos
exposure and indicate that pleural plaques (as defined by 1980 ILO) and LPT are associated with
declines in pulmonary function. Although the decrements in mean pulmonary function measures
associated with the presence of LPT may not be generally considered clinically significant, the
relation between plaque size and degree of decrement, and the increase in size over time indicate
these changes may be consequential even on the individual level. In addition, even small mean
differences can have a large impact on a population level.
1-31
-------
Supplemental Table I-A. Localized pleural thickening (LPT) pulmonary function studies evaluation: cross-sectional
studies, internal comparison (alphabetical within x-ray and high resolution computed tomography [HRCT] groups)
Reference, population3
Selection
X-ray or HRCT measures
Spirometry
Consideration of
exposure and smoking
Analysis
X-ray Studies
Bourbeau et al. (1990)
Quebec
Construction—insulators (union)
Mean (SD) age 44.3 (4.8) yra
Mean (SD) duration 17.3 (7.4) yr1
Mean (SD) TSFE 24.0 (5.6) yra
53% current smokers3
Mean pack-yr 22a
Percentage currently working not
reported
Invited for pulmonary
function tests in 1983-1984.
Included if:
Age 35 to <50 yr
Not receiving
compensation for
asbestosis in 1982
Within 30 km of Montreal
n = 110 out of 129 (85%)
participated
P-A view
Two B Readers
Blinding not described
FLO (1980)
Pleural plaques without
obliteration of costophrenic
angle and without small opacity
profusion >1/0 (« = 58, 52.5%)
Renzetti(1979)
procedures, best of three
values
One of two trained
technicians
Reference values not
reported (other than
Renzetti(1979) reference)
FEVi, FVC
Duration, not used in
analysis
Smoking data for pleural
plaque and no pleural
plaque groups,
respectively:
53 and 46% current
smokers; mean 22 and
15 pack-yr; adjusts for
smoking in analysis
Adjusted for age,
height, smoking,
parenchymal disease
(based on x-ray and
gallium-67 uptake
quantitation);
Table 5 excludes
costophrenic angle
obliteration and
profusion> 1/0
Bresnitz et al. (1993)
Philadelphia
Construction—elevator (union)
Mean (SD) age 52.2 (7.9) yr
Mean (SD) duration 27.1 (5.8) yr
TSFE not reported
36% current smokers
Pack-yr not reported
8 out of 91 retired
Screening program in 1988
through local chapter of the
International Union of
Elevator Constructors for
20+ yr. Eligibility based on
membership, regardless of
current employment status
n = 91 (n total eligible not
available)
P-A view
Two independent B Readers (+ 3rd
reader for consensus) (moderate
agreement between readers)
Blinded to exposure, medical
history, and other reading
FLO (1980)
Pleural thickening (15 bilateral,
5 unilateral)
DPT: none
Interstitial changes: none >1/0
ATS (1987) procedures, at
least three values. NIOSH
certified technician, sitting
position
86% of patients had three
acceptable curves and all
had at least one (none
excluded for
nonrepeatability)
Highest value used
Percentage predicted
based on Crapo et al.
(1981) equations
FEVi, FVC, FEVi/FVC,
FEF2575
Duration (job tenure),
not used in analysis
Authors noted no
association between
pleural abnormalities and
smoking
Table 2, internal and
external comparison
1-32
-------
Supplemental Table I-A. Localized pleural thickening (LPT) pulmonary function studies evaluation: cross-sectional
studies, internal comparison (alphabetical within x-ray and high resolution computed tomography [HRCT] groups)
(continued)
Reference, population3
Selection
X-ray or HRCT measures
Spirometry
Consideration of
exposure and smoking
Analysis
Pi Lorenzo et al. (1996)
Italy
Asbestos cement factory
Mean (SD) age 54.5 (6.5) yra
Mean (SD) duration 23.5 (7.4) yra
50% current smokers3
Pack-yr not reported
100% former workers3
Recruited through union,
n = 30 (out of 35)
participated
Eligibility criteria not
described
Referent: n = 9 male union
members, "never exposed to
respiratory irritant dust or
fumes"
P-A, lateral and oblique views
Two readers (radiologist and
occupational physician); 100%
concordance
Blinded to exposure status and to
other reading
FLO (1980)
Pleural plaques (n = 10)
Asbestosis (diffuse interstitial
fibrosis, > l/l,w = 11)
Healthy exposed (no bronchial,
parenchymal, or pleural disease,
w = 9)
Nonexposed (n = 9)
ATS (1987) procedures,
(details not reported)
Absolute value and
percentage predicted
based on reference values
from European
Community for Coal and
Steel (Quanjer and van
Zomeren, 1983)
FEVi, FVC, FEVi/FVC,
PEF, FEF25, FEF5o, FEF75
Duration, not used in
analysis
Authors indicated that
smoking distribution was
similar across groups
Percentage predicted,
by group (takes into
account age
differences) (see
Table 3)
Duiic et al. (1993)
Croatia
Asbestos cement factory
Mean (SD) age 58.2 (10.1) yra
TSFE not reported
Mean (SD) cumulative exposure
39.6 (12.3) f-yra
62% current smokers3
Mean (SD) 37.7 (29.4) pack-yra
83% current workers
Current and retired workers
at asbestos cement factory
eligible; n = 344 total (284
out of 309 current workers,
92%; 58 out of 112 retired
workers, 52%).
Excluded (n = 37):
Further exclusions:
isolated parenchymal
changes (profusion <1/1,
w=16)
combined pleural and
parenchymal disease
(« = 17)
DPT (n = 4)
Included:
isolated pleural plaques
(« = 55)
workers without any
radiographic change
(» = 255)
P-A view
Two ILO-trained readers
(radiologists)
Blinded to exposure, clinical, and
pulmonary function data
FLO (1980)
Plaque-like thickening at the lung
pleura interface along the lateral
thorax or either hemidiaphragm
was 2+ mm
ATS (1987) procedures,
best of three acceptable
values
Percentage predicted
based on Cotes (1975)
FEVi, FVC, FEV%,
FEF25-75, TLC, RV, DLco
Restriction: FVC <80%
andFEV%>70%
Obstruction: FEVi <80%,
FEV%<70%
Annual exposure data
from approximately
1960 to 1990, PCM fiber
counts. These data used
to calculate individual
level cumulative
exposure (1950s
exposures assumed to be
3 fibers/mL).
Smoking data for pleural
plaque and no pleural
plaque groups,
respectively:
62 and 38% current
smokers, mean 37.7 and
29.5 pack-yr;
quantitative adjusted
results not provided.
Table 2: Mean
difference in percentage
predicted by group,
unadjusted for smoking
and exposure; Table 4
and text: adjusted for
exposure and smoking
(reported as
significance level only)
1-33
-------
Supplemental Table I-A. Localized pleural thickening (LPT) pulmonary function studies evaluation: cross-sectional
studies, internal comparison (alphabetical within x-ray and high resolution computed tomography [HRCT] groups)
(continued)
Reference, population3
Selection
X-ray or HRCT measures
Spirometry
Consideration of
exposure and smoking
Analysis
Garcia-Closas and Christiani (1995)
Massachusetts
Construction—carpenters (union)
Mean (SD) age 51.9 (8.6)yra
Mean (SD) duration 28.0 (8.5) yra
TSFE not reported
24% current smokers*
Mean (SD) pack-yr 30.7 (21.6)a
98% currently working3
Invited to participate by
union, 1987-1988
618 out of 3,897 active
workers (16%) and 13 out of
375 retired workers (3%)
participated
« = 631
P-A view
Two B Readers
Blinded to exposure history
(mixed with 1,200 other x-rays)
FLO (1980)
Circumscribed plaque without
obliteration of costophrenic angle)
(n = 64, 10%)
Diffuse pleural thickening (n = 3,
0.5%)
Interstitial fibrosis (small
opacities 0/1,1,0 or higher)
(n = 43, (7%)
Other nonpneumoconiosis
abnormalities (n = 64,10%)
No abnormalities (n = 457, 72%)
ATS (1987) procedures,
multiple technicians, at
least three values;
nonreproducible results
and results with only one
value excluded
Percentage predicted
based on Crapo et al.
(1981) equations
FEVi, FVC, FEV%
Duration
Smoking data for pleural
plaque and no pleural
plaque groups,
respectively:
32 and 24% current
smokers, mean 30.7 and
22.3 pack-yr
Tables III unadjusted,
Tables IV-V adjusted
for yr in trade, smoking
status, pack-yr,
occupation (carpenter,
millwright, other), and
interstitial fibrosis
Hilt et al. (1987)
Norway
Asbestos-exposed workers
Mean (SD) age 67.3 (8.4) yra
Mean (SD) duration 3.6 (3.8) yra
Mean TSFE 37 yra
39% current smokersa
Pack-yr not reported
Percentage currently working not
reported
Other refs: (Hilt etal. (1986b); Hilt
etal. (1986a))
County-wide screening of
asbestos-exposed workers
(n = 21,483), referred for
reexamination if
abnormalities found on x-ray
(n= 1,431); 1,372 (96%)
participated
Exclusions:
141 with obstructive lung
disease, lung cancer, or
sarcoidosis, or other lung
diseases as primary
diagnosis
591 other (nonasbestos)
reasons for lung disease
w = 634
P-A + lateral views
Department radiologist followed
by one B Reader
Blinding not described
Reference for definitions not cited
Pleural plaques only (n = 363,
57%)
Fibrosis with or without plaques
(w = 83, 13%)
No abnormalities, previous
exposure reported (n = 98, 15%)
No abnormalities, no reported
exposure (n = 90,14%)
Procedure reference not
given; details not provided
(other than upright
position)
Reference values based on
asymptomatic men from
Oslo, based on study
using random sample of
Oslo population
FEVi, FVC, FEV%,
FVC <90 %pred,
FVC <80 %pred,
FEVi <80 %pred
4-level exposure:
uncertain, light moderate
heavy (not used in
analysis)
Smoking data for pleural
plaque and no pleural
plaque groups,
respectively:
39 and 55% current
smokers (higher in
pleural plaque group);
predicted values based
on sex, age, height, and
smoking habits.
Table IV (comparison
with predicted based on
reference values)
1-34
-------
Supplemental Table I-A. Localized pleural thickening (LPT) pulmonary function studies evaluation: cross-sectional
studies, internal comparison (alphabetical within x-ray and high resolution computed tomography [HRCT] groups)
(continued)
Reference, population3
Selection
X-ray or HRCT measures
Spirometry
Consideration of
exposure and smoking
Analysis
Jiirvholm and Sanden (1986)
Sweden (Gothenburg)
Shipping industry
Mean (SD) age 54.9 (5.8) yr
Mean duration 26 yr
0% current smokers
0 pack-yr
Percentage currently working not
reported (but likely to be high)
General screening of workers
in 1977-1979
(n = 3,904 participated; total
n not reported). Included if:
Men
Ages 40-65 yr
Never smoked
No other known or
suspected lung disease on
chest x-ray
No other asbestos exposure
before shipyards
No change of jobs during
employment at shipyards
>20 yr TSFE
Insufficient exposure data
(«=1)
« = 202
P-A + lateral views
One reader (from group of three
chest physicians)
Blinding not described
Thiringer et al. (1980) definition:
circumscribed thickening not
extending to the apices or with
connection to costophrenic
sinuses, or >3 mm thickness on
diaphragm if no calcification, or
<5 mm thick and no calcifications
with a marked edge at top and
bottom (n = 87)
Procedure reference not
given; best of three
values;
Trained nurses (n not
reported)
Tested before x-ray
Percentage predicted
based on Berglund et al.
(1963)
FEVi, FVC
4-level exposure (very
low, low, heavy, very
heavy); 7-level exposure
time (< once a yr to
>2hrperd).
Limited to never
smokers
Adjusted for age and
height
(see Table 1); Table II
vs. Table III and
Figures 1 and 2 include
stratification by
exposure (level or time)
Jiirvholm and Larsson (1988)
Sweden (Gothenburg)
Asbestos-exposed workers
62% ages 50-59 yra
43% current smokers*
89% >5 yr continuous exposure*
Percentage currently working not
reported (but likely to be high)
General screening of
asbestos-exposed workers in
1976 (n = 4,268). Included
if:
Men
Ages 40-65 yr
No other known or
suspected lung disease
No cardiac disease
n = 1,233
P-A + lateral views
One reader (from a group of
readers)
Blinding not described
Thiringer et al. (1980) definition:
calcifications typically localized
on the diaphragm or chest wall, or
typically localized elevations on
the diaphragm, >3 mm thick, with
a sharp edge, or well-demarcated
thickenings on chest wall >5 mm
wide(w= 130)
No pleural plaques (n = 1,103)
Procedure reference not
given; best of two values
A trained assistant
Percentage predicted
based on Berglund et al.
(1963)
FEVi
Smoking data for pleural
plaque and no pleural
plaque groups,
respectively:
43 and 39% current
smokers; analyses
stratified by smoking
status
Percentage predicted,
stratified by smoking
(see Table 5)
1-35
-------
Supplemental Table I-A. Localized pleural thickening (LPT) pulmonary function studies evaluation: cross-sectional
studies, internal comparison (alphabetical within x-ray and high resolution computed tomography [HRCT] groups)
(continued)
Reference, population3
Selection
X-ray or HRCT measures
Spirometry
Consideration of
exposure and smoking
Analysis
Miller et al. (1992)
United States and Canada
Insulation workers
Mean (SD) age 57
Mean (SD) TSFE 35
80% current and exsmokers
Mean (SD) pack-yr 40.6 (26.2)
Other ref: Lilisetal. (1991b): Lilis et
al. (1991c): Lilis et al. (1991aX Lilis et
al. (19921 Miller andZurlo (1996)
Cohort established 1967
(Selikoff and Hammond.
1979): 1981 to 1983
screening
Participation rate reported as
approximately 40%. No
difference in subsequent
mortality between
participants and
nonparticipants.
n = 2,611 (n = 2,270 with
duration >30 yr, plus 341
who joined with less than this
duration)
P-A and lateral views
One B Reader
Blinded to occupational and
medical history
ILO(1980)
Pleural plaques (circumscribed
and diffuse;
diffuse = costophrenic angle
obliteration)
Limited to O/- or 0/0 profusion
score): n = 203 no pleural
thickening, n = 121 circumscribed
pleural plaques, n = 1 diffuse
pleural plaques
ATS (1987) procedures;
standing position,
>3 acceptable readings
Percentage predicted
based on random sample
evaluated in the same
laboratory controlling for
smoking and age (Miller
etaL 1986)
FVC
Smoking data by group
not reported and not
included in analysis
Table 3, 0/0 and O/-
row, circumscribed vs.
no pleural thickening
Miller et al. T2013)
United States (four states)
Mean (SD) age 62.1 (9.5)yr
Mean (SD) duration 28.0 (10.6)
"vast majority" TSFE >15 yr
21% current smokers
Screening program through
unions, 1997-2004 (for
medico-legal evaluation)
Total n = 6,932; excluded
women, nonwhites, and those
missing smoking
information, x-ray,
spirometry, or diffusing
capacity data.
n = 4,003
P-A and lateral views
One B Reader
Blinded to occupational and
medical history
FLO (1980)
Circumscribed only (n = 290)
Diffuse only (n = 10)
Circumscribed and diffuse
(« = 16)
Diaphragm only (n = 83)
Costophrenic angle (n = 1)
ATS (1987) procedures
(details not reported but
equipment, techniques,
technicians noted to be
same as in teaching
hospitals)
Percentage predicted
based on Crapo et al.
(1981)
FVC
Smoking data by group
not reported and not
included in analysis
Table VI, pleural
abnormalities only
1-36
-------
Supplemental Table I-A. Localized pleural thickening (LPT) pulmonary function studies evaluation: cross-sectional
studies, internal comparison (alphabetical within x-ray and high resolution computed tomography [HRCT] groups)
(continued)
Reference, population3
Selection
X-ray or HRCT measures
Spirometry
Consideration of
exposure and smoking
Analysis
Ohlson et al. (1985) and Ohlson et
al. (1984): 1985 provides
quantitative results, but is smaller «;
Sweden
Asbestos cement plant
Mean age 59.1 yra
Meanfiber-yr20.9(range: 0-48)a
Mean pack-yr 40.6a
100% current workers3
Screening offered in 1976
(after plant closed),
participation rate 96%
Excluded if:
Retired
Former smokers
Female
<10 yr employment
Comparison group: workers
from other plants not using
asbestos (fertilizer, cement
products, wood products),
selected from same health
center, no x-ray signs of
chest disease
Original group n = 125
exposed workers and
76 referents.
At follow-up: n = 75
exposed, 56 referents.
6 cases and 3 referents had
died (cause of death for 5 of
the 6 cases known, not
related to asbestos), 32 cases
and 9 referents had changed
smoking status and were
excluded
P-A, lateral and oblique views
One qualified reader (member of
National Pneumoconiosis Panel)
Blinding not described
FLO (1980)
Pleural plaques (not defined)
(n = 42, 34%)
Procedure reference not
reported; sitting position;
best of three values
(within 5%)
One trained technician
Reference values from
Berglundet 31(19631)
FEVi, FVC
Estimated average
2fibers/mLinl950sand
1960s, 1 fiber/mLin
1970s; levels for specific
work areas estimated and
used for individual-level
cumulative exposure
Smoking data for pleural
plaque and no pleural
plaque groups,
respectively:
mean 40.6 and
33.4 pack-yr.
Table 4 (combining
exposure groups)
1-37
-------
Supplemental Table I-A. Localized pleural thickening (LPT) pulmonary function studies evaluation: cross-sectional
studies, internal comparison (alphabetical within x-ray and high resolution computed tomography [HRCT] groups)
(continued)
Reference, population3
Oliver et al. (1988)
Pennsylvania
Railroad workers
Mean age 65 yra
Mean duration 35 yr3
Mean TSFE 45 yra
26% current smokers (among full
sample)
Meanpack-yr30.8
Related reference: Oliver et al.
(1985)
Rosenstock et al. (1988)
United States (Washington)
Plumbers and pipefitters
Mean age 42. 1 yr
Mean duration 1,71 1 yr
TSFE not reported
33% current smokers
Schwartz et al. (1990)
Iowa
Sheet metal workers union,
Mean (SD) age 57.0 (8.0) yr
Mean (SD) duration 32.7 (6.7) yr
Mean (SD) TSFE 35.7 (6.5) yr
3 1 % current smokers
Mean pack-yr 28
72% currently working
Related reference: Broderick et al.
(1992)
Selection
Screening study, n = 383
n = 377 white men
Excluded if:
Interstitial fibrosis (>0/1,
n = 6)
Diffuse pleural thickening
(« = 10)
Unreadable x-rays (n = 2)
w = 359
Surveillance program
through unions, 1 982-1 984,
participation rates about 20
and 7% in Seattle and
Tacoma, respectively.
w = 681
12 union locals
1,223 out of 2,646 (46%)
participated;
Included if:
Employed >25 yr
n = 1,211 with x-rays
X-ray or HRCT measures
P-A and lateral views
One B Reader + one A or
B Reader
Blinding not described
FLO (1980)
Plaque-like thickening at the
lung-pleura interface along the
lateral chest wall tangentially or
along the en face rib margin,
>2 mm, or typical plaque-like
thickening along either
hemidiaphragm (n = 81, 23%)
No plaques(w = 278, 77%)
P-A view
Two trained readers
Blinded to clinical status
FLO (1980)
Validity test of 50 radiographs
read several mo later showed 98%
agreement within one category of
profusion
Pleural thickening: diffuse or
circumscribed, in absence of other
evident cause
Interstitial fibrosis: >1/0
profusion
P-A view
One experienced reader (+10%
validation study)
Blinded to exposure history
FLO (1980)
Circumscribed plaque, without
obliteration of costophrenic angle
(n = 260, 21.5%); includes 31%
with asbestosis >1/0
Diffuse (n = 74, 6%)
Normal (n = 877, 72%)
Spirometry
Renzetti(1979)
procedures, > three tests
Percentage predicted
based on Crapo et al.
(1981)
FEVi, FVC, DFco
Procedure reference not
reported. Best of
>3 values used; data on
reproducibility and impact
of nonreproducibility on
results given; no
exclusions based on
nonreproducibility.
Percentage predicted
based on Crapo et al.
(1981)
FEVi, FVC
Renzetti(1979)
procedures, seated,
without repeatability
requirement (18% would
have been excluded)
Average of the two largest
values
FVC (see
Table 9— Schwartz)
Consideration of
exposure and smoking
Duration, used as
stratification variable
Smoking data for pleural
plaque and no pleural
plaque groups,
respectively:
Mean 30. 9 and 2 1.2
pack-yr; adjusts for
smoking in the analysis
Smoking data by group
not reported and not
included in analysis
Duration, included in
adjusted analysis
Smoking data for pleural
plaque and no pleural
plaque groups,
respectively:
30.1 and 3 1.2% current
smokers, mean 29.9 and
25. 4 pack-yr. (These
data presented in table
that also includes
asbestosis); pack-yr
Analysis
Percentage predicted
adjusts for age and
height; also adjusted for
smoking status (in text)
Figure 4 (group O/- and
0/0)
Adjusted for age,
height, interstitial
fibrosis (ILO
profusion), pack-yr, in
sheet metal trade, (see
Table 4 in Broderick;
Tables 6-9 in Schwartz)
Table 9 in Schwartz
excludes interstitial
changes
1-38
-------
Supplemental Table I-A. Localized pleural thickening (LPT) pulmonary function studies evaluation: cross-sectional
studies, internal comparison (alphabetical within x-ray and high resolution computed tomography [HRCT] groups)
(continued)
Reference, population3
Selection
X-ray or HRCT measures
Spirometry
Consideration of
exposure and smoking
Analysis
included in adjusted
analysis
Singh et al. (1999)
Australia
Asbestos-exposed (various sources)
Mean (SD) age 64.1 (2.3) yr>
Duration not reported
TSFE not reported
8% current smokers3
Pack-yr not reported
Cohort seen in outpatient
clinic because of asbestos
exposure, 1994-1995
Excluded if:
Clinical or x-ray evidence
of asbestosis or other
interstitial lung disease,
asthma, emphysema, lung
cancer, pleural effusions,
neurologic or myopathic
disorder likely to weaken
respiratory muscles
n = 26
Views not reported
One experienced reader
Blinding not described
FLO (1980)
LPT = costal and/or
diaphragmatic plaques with no
involvement of costophrenic
angle (« = 12, 46%)
DPT = costophrenic angle
obliteration and thickening with
or without calcification of the
costal and/or diaphragmatic
pleura (n = 7,27%)
No abnormalities (n = 7,27%)
Reference not reported,
details not provided.
Percentage predicted
based on various
references
TLC, VC, RV
Smoking data for pleural
plaque and no pleural
plaque groups,
respectively:
8 and 0% current
smokers, based on single
individuals)
Percentage predicted
Weilletal. (2011)
Montana (Libby)
Community-based
Mean (SE) age 60.07 (0.53) yra
64% ever smokers
Community screening,
includes former workers at
vermiculite mine and mill,
family members, and other
area residents; n = 7,307
Excluded if:
No chest x-ray (n = 639)
Age <25 or >90 yr or
missing spirometry
(n = 817)
Other (nonvermiculite)
exposure likely
(«= 1,327)
No consensus x-ray
reading, missing smoking
data or missing exposure
pathway data (n = 127)
n = 4,397
P-A view
Two out of three B Readers
consensus
Blinding not described
ILO(1980)
Pleural abnormality excluding
DPT, costophrenic angle
obliteration, or interstitial disease
(profusion >1/0) (n = 482,11%)
DPT and costophrenic angle
obliteration, no interstitial disease
(n = 33, 1%)
Interstitial disease (profusion
>l/0)(w = 40,1%)
No abnormality (n = 4,065; 92%)
(total = 4,620, bigger than 4,397)
ATS (1995) procedures,
three acceptable (two
reproducible) tests or one
or two acceptable tests
Percentage predicted
based on Knudson et al.
(1983)
FEVi, FVC, FEVi/FVC
Stratified by smoking
status (ever/never) and
men and women
Table 4 (unadjusted)
and Table 6 stratified
by gender-smoking and
adjusted for age and
BMI
1-39
-------
Supplemental Table I-A. Localized pleural thickening (LPT) pulmonary function studies evaluation: cross-sectional
studies, internal comparison (alphabetical within x-ray and high resolution computed tomography [HRCT] groups)
(continued)
Reference, population3
Zavalic and Boaadi-Sare (19931
Croatia
Shipyard workers
Mean (SD) age 45.1 (5.2)yra
Mean (SD)duration 21 .5 (14. 1 ) yra
Mean (SD) TSFE 26.6 (17.2) yra
Smoking data not reported
Selection
Excluded 5 1 with other
confirmed diseases could
affect pulmonary function
X-ray or HRCT measures
P-A and oblique views
Agreement based on two out of
three readers (independent
readings; two occupational health
specialists and a radiologist)
Blinding not described
ILO (19801
No changes (n =101)
Pleural plaques only (n = 68)
Parenchymal librosis (n= 130;
n = 42 only parenchymal fibrosis)
No DPT, effusion, mesothelioma,
or lung cancer. All plaques were
bilateral.
Spirometry
Procedure reference not
provided. Best of three
values
Percentage predicted
based on
Ouanieretal. (19931
FEVi, FVC, FEVi/FVC,
MEF25, MEFso, MEF75
Consideration of
exposure and smoking
Authors indicated that
smoking distribution was
similar across groups
Analysis
Table 5 (pleural plaques
and parenchyma
category 0)
HRCT Studies
Clin et al. (20111
France
Exposed workers (retired or inactive)
Mean (SD) age 64.6 (5.4) yra
72% duration >30 yr1
TSFE not reported
6.4% current smokersa
Pack-yr not reported
Related ref: Paris et al. (2009)
Various recruitment
strategies (letters, union,
advertisements) for medical
surveillance program
4,812 recruited, excluded:
3 12 missing data;
873 inadequate CT quality,
57 extreme spirometry
values, 227 asbestosis or
other interstitial
abnormalities.
n = 2,743
Independent reading by two (out
of panel of seven) readers
Blinded to asbestos exposure and
smoking
Isolated pleural plaques (n = 403,
14.7%)
Normal (n = 1,802, 65.7%)
(excluding 123 with pleural
plaques with and other
nonspecific abnormalities [e.g.,
emphysema, bronchiectasis],
41 with diffuse pleural thickening,
and 374 with other nonspecific
abnormalities)
Procedure reference not
reported.
Multiple locations.
Percentage predicted
based on
Ouanieretal. fl 9931
European reference
equations.
Extreme values excluded
(« = 57)
Semiquantitative
exposure index: lifetime
job history questionnaire,
industrial hygienist
rating on 4 -level
exposure (passive, 0.01
to high, 10). Cumulative
index based on sum for
all jobs, divided into
quintiles (exposure
units-yr), included in
adjusted analysis.
Smoking data by pleural
plaque and no pleural
plaque group,
respectively: 6.4 and
6.0% current smokers,
included in adjusted
analysis.
Table 3
Adjusted for age,
gender, body mass
index, smoking,
location of pulmonary
function testing, yr
asbestos exposure,
cumulative exposure
index
1-40
-------
Supplemental Table I-A. Localized pleural thickening (LPT) pulmonary function studies evaluation: cross-sectional
studies, internal comparison (alphabetical within x-ray and high resolution computed tomography [HRCT] groups)
(continued)
Reference, population3
Selection
X-ray or HRCT measures
Spirometry
Consideration of
exposure and smoking
Analysis
Lebedova et al. r2003)
Czech Republic
Exposed workers (current or former)
Mean (SD) age 61.5 (9.2) yra
Mean (SD) duration 23.9 (9.9) yra
Mean (SD) TSFE 38.0 (10.8) yra
15.5% current smokers, 36.1%
exsmokersa
Mean(SD)pack-yr21.4 (17.7)
Mean(SD)BMI29.3(5.7)
Registry of current and
former asbestos processing
plant employees. 1,199
employees in registry, 2000
follow-up included those
(i) with documented
occupational exposure to
asbestos; (ii) absence of
parenchymal fibrosis
(profusion scores <0/1);
(iii) no history of disease
likely to bias chest
radiograph; (iv) no bronchial
asthma, n = 591.
Of those followed up in
2000, approximately 30%
were randomly selected from
profusion score groups
defined on the basis of x-rays
taken in 2000, n = 162.
HRCT:
Reading procedures not described.
Pleural lesions divided into
categories based on size of largest
plaque: 0 = none, 1 = small,
2 = medium, 3 = large, 4 = very
large.
Pleural plaques (n = 97, 59.9%)
Normal (n = 65, 40.1%)
X-ray (used to define and exclude
those with parenchymal fibrosis):
P-A view
Evaluated independently by one
radiologist and three physicians.
Blinding not described
ILO(1980)
European Respiratory
Society procedures used.
Reference equations for
percentage predicted not
reported.
Parenchymal changes recorded
were: thickened intralobular and
interlobular septal lines,
subpleural curvilinear lines,
parenchymal bands, ground glass
opacities, and honeycombing.
Definite asbestosis (n = 17,
10.5%)
Suspected asbestosis (n =58,
35.8%)
Exposure not considered
in analysis
Smoking data by pleural
plaque and no pleural
plaque group,
respectively: 15.5 and
27.2% current smokers,
36.land 23.1%
exsmokers, mean (SD)
21.4 (17.7) and 19.8
(14.5) pack-yr; smoking
habit included in
adjusted analysis
Table 5
Adjusted for smoking,
chronic bronchitis, BMI
and ischemic heart
disease, with an
interaction term
between fibrosis and
pleural lesions
1-41
-------
Supplemental Table I-A. Localized pleural thickening (LPT) pulmonary function studies evaluation: cross-sectional
studies, internal comparison (alphabetical within x-ray and high resolution computed tomography [HRCT] groups)
(continued)
Reference, population3
Selection
X-ray or HRCT measures
Spirometry
Consideration of
exposure and smoking
Analysis
Neri et al. (1996)
Italy
Shipyard factory workers (current, but
exposure ceased 11-14 yr prior to
examination)
Mean (SD) age 45.6 (6.5) yra
Mean (SD) duration 9.1 (5.5) yra
Mean (SD) duration of heavy
exposure 3.8 (4.^yr3
Mean (SD) TSFE 22.6 (5.2) yr3
Mean (SD) pack-yr 8.9 (10.1) yra
100% current workers
161 male 'blue-collar'
employees; included those
who (i) consented to exam;
(ii) were employed when
asbestos was being used;
(iii) were not occupationally
exposed to other mineral
dusts/welding fumes;
(iv) absence of small
irregular opacities (profusion
score >1/0) and/or clinical
symptoms of lung disease.
w=119
HRCT: by agreement of two
thoracic radiologists blinded to
exposure status
Pleural alterations were quantified
applying 1980 ILO criteria to the
reading of CT scans
Parenchymal abnormalities were
interpreted on the basis of
previous studies (AkiraetaL
1991; Akira et al., 1990; Lynch et
al., 1989)
Pleural plaques only (n = 50,
42.0%)
Normal (n = 31,26.1%)
American Thoracic
Society guidelines used
Reference values from
Paoletti et al. (1985);
Paoletti et al. (1986)
FEVi,FVC, TLC,
FEVi/FVC%, MEF25,
MEF5o, MEF75, MEF25-75
Estimated duration of
'heavy exposure' (based
on work tasks/location)
and total exposure (total
yr of employment ant
plant). Industrial
hygiene sampling
performed in 1977
showed averages ranging
from 6-18 f/cm3 at sites
near specific pieces of
equipment
No quantitative results
Oldenburg et al. GOOD
Germany
Exposed workers:
Mean age not reported
Mean duration 30.7 yr
TSFE not reported
76.2% of those with pleural plaques,
and 68.2% of those without plaques,
current or ex-smokers
(Additional study details provided in
personal communication from Xavier
Baurto L. Kopylev, 3/13/2014).
Registry of asbestos-exposed
workers (w~500,000);
included highly exposed
subjects with no other lung
disease, who had pleural
plaques or without pleural or
pulmonary asbestos-
associated changes.
Approximately 2/3 in registry
undergo periodic exams.
This study conducted in
Bochum area.
w = 43
HRCT: reading procedures not
described, blinding not reported.
Authors stated no subjects showed
signs of parenchymal
abnormalities.
Pleural plaques only (n = 2\,
48.8%)
Normal (n = 22, 51.2%)
Spirometry procedures
and references not
described
FEVi,FVC, FEVi/VC%,
MEF25, MEFjo, MEF75
Analysis stratified by
smoking status (current
and former smokers,
nonsmokers)
Table 1. Results
stratified by smoking
1-42
-------
Supplemental Table I-A. Localized pleural thickening (LPT) pulmonary function studies evaluation: cross-sectional
studies, internal comparison (alphabetical within x-ray and high resolution computed tomography [HRCT] groups)
(continued)
Reference, population3
Rui et al. (2004)
Italy
Mean (SD) age 53 (7) yra
Mean (SD) duration 30 (6) yra
TSFE not reported3
22% smokers (<15 pack-yr), 42%
smokers (>15 pack-yr), 36%
nonsmokersa
42% current workers
Additional study details provided in
personal communication from
Francesca Rui to L. Kopylev,
3/15/2014).
Soulat et al. (1999)
France
Nitrate fertilizer plant (asbestos
insulation) (former workers)
Mean (SE) age 65.2 (0.6) yr
Mean (SE) duration 12.9 (0.6) yr
Mean (SE) TSFE 38.9 (0.5) yr
1 9% current smokers
Mean (SE) 22.6 (1 .6) pack-yr
Selection
Workers referred to an
occupational medicine clinic
1991-2000; included those
with history of asbestos
exposure; had two spirometry
tests performed at least 1 yr
apart; had radiological
examination performed; no
signs of interstitial fibrosis,
emphysema, bronchiecstasis,
pleurisy, TB, or other
significant lung, cardiac,
skeletal or systemic disease.
Included only those workers
with pleural plaques on x-ray
who were further referred for
HRCT.
w = 103
350 exworkers identified
through retirement
association; 254 potentially
exposed, still living;
n = 170 participants
X-ray or HRCT measures
One reader for x-ray and HRCT,
blinding not reported.
Pleural plaques described by
location (unilateral/bilateral,
diaphragmatic) and presence of
calcification; defined as
"circumscribed areas of
thickening of the parietal pleura in
thoracic cage and/or diaphragm"
Pleural plaques only (n = 36,
35%)
Normal (n = 67, 65%)
One reader, blinded to patient
history and x-ray results
Pleural changes defined by size
and appearance: normal,
focalized, and diffuse thickening
(n = 84 without parenchymal
changes).
Parenchymal abnormalities were
interpreted on the basis of
previous studies (Aberle et al..
1988; Yoshimura et al., 1986)
n = 84 pleural thickening only;
No abnormalities (n = 51)
Spirometry
Spirometry procedures not
referenced.
Reference values from
CECA71
FEVi, VC, TLC
Spirometry procedures not
referenced.
Reference values from
Ouanieretal. (19931
Consideration of
exposure and smoking
Exposure duration data
by pleural plaque and no
pleural plaque group,
respectively: mean(SD)
30 (6) and 22 (6) yr;
exposure duration
included in adjusted
analysis
Smoking data by pleural
plaque and no pleural
plaque group,
respectively: 36 and
36% nonsmokers, 22 and
30% smokers
(<15 pack-yr), 42 and
34% smokers
(>1 5) pack-yr; smoking
"habit" included in
adjusted analysis (<15 or
>15 pack-yr)
Estimation of exposure
intensity (high, 65.9%;
moderate, 12.3%; low,
12.3%) but not used in
analysis of spirometry
results; Smoking data by
group not reported and
not included in analysis.
Analysis
Table 2: unadjusted,
cross-sectional analysis.
Table IV (unadjusted)
1-43
-------
Supplemental Table I-A. Localized pleural thickening (LPT) pulmonary function studies evaluation: cross-sectional
studies, internal comparison (alphabetical within x-ray and high resolution computed tomography [HRCT] groups)
(continued)
Reference, population3
Selection
X-ray or HRCT measures
Spirometry
Consideration of
exposure and smoking
Analysis
Staples et al. (1989)
California
Mean(SD)age59(ll)yr
Mean (SD) duration 20 (10) yr
Mean (SD) TSFE 34 (10) yr
38% current smokers
Mean pack-yr not reported
Percentage current workers not
reported
Selected from 400 exposed
workers. Included if:
Documented exposure to
asbestos
Latency >10yr
HRCT (and x-ray within
1 yr of HRCT)
Profusion <0/1 by x-ray
Excluded if:
DPT on x-ray or HRCT
HRCT indeterminate for
asbestosis
«=136
X-rays: FLO (1980). By
agreement of two readers, one of
which NIOSH-certified;
disagreement between 0/1 and 1/0
read by 3rd radiologist
HRCT: By agreement of two
radiologists
Blinded to x-ray and clinical data
Normal parenchyma (n = 76)
(divided into with and without
plaques; n per group not reported)
Suggestive of asbestosis (n = 57)
Procedure reference not
provided (details not
reported)
Percentage predicted
based (Crapoetal., 1981)
FEVi, TLC, VC, RV
Authors noted that
smoking distribution was
similar across groups.
Text, page 1,507;
"normal" group divided
into with and without
plaques; reported as
"not significantly
different" but
quantitative results not
reported
van Cleemput et al. (2001)
Belgium
Asbestos cement factory
Mean (SD) age 43.5 (2.2) yr
Mean (SD) duration 25.0 (1.4) yr
Mean (SD) cumulative exposure 26.3
(12.2)fiber-yr/mL
85% ever smokers
Mean pack-yr 10.9 yr
100% current workers
Included if:
Bom between 1945 and
1954
Hired between 1965 and
1969
Worked >2 yr
n = 73 (out of 88 identified
workers; 3 of
15 nonparticipants had
plaques)
X-rays: ILO (1980) three
independent readers, blinded to
exposure status
CT scans (reading protocol not
stated)
Pleural plaques seen in 26% of
exposed workers by x-ray, and in
70% by HRCT. None of the
exposed workers had asbestosis or
profusion scores above 1/0
European procedures
Ouanier etal. (1993)
(details not reported)
Percentage predicted
based on Ouanier et al.
(1993)
FEVi, FEVi/VC, PEF%,
MEF25, MEF5o, MEF75,
TLco (transfer fraction for
carbon monoxide)
Smoking data by group
not reported and not
included in analysis.
Table 3
NIOSH = National Institute for Occupational Safety and Health; PCM = phase contrast microscopy.
""Descriptive data for pleural plaque (or LPT) group; when not noted as such, data is for full study sample.
1-44
-------
Supplemental Table I-B. Localized pleural thickening (LPT)—pulmonary function studies evaluation:
longitudinal studies (alphabetical)
Reference, population
Selection
X-ray or HRCT measures
Spirometry
Exposure
Analysis
Ohlson et al. (1985)
Sweden
4 yr follow up of workers first
evaluated in 1976
Mean age 59.1 yr
Mean fiber-yr 20.9 (range: 0-48)
62% current smokers
Meanpack-yr40.6
100% current workers
Related reference: Ohlson et al.
(1984)
n = 15; male active workers at an
asbestos cement plant (production
ceased in 1976). Limited to
current and never smokers.
Referents: n = 56 workers from
three plants without exposure to
asbestos.
Original group was n = 125
exposed workers and 76 referents.
6 cases and 3 referents had died
(cause of death for 5 of the 6 cases
known, not related to asbestos),
32 cases and 9 referents had
changed smoking status and were
excluded.
P-A, lateral and oblique views
One qualified reader (member of
National Pneumoconiosis Panel)
Blinding not described
FLO (1980)
Exposed subjects had second
radiograph in 1980, referents only in
1976.
Pleural plaques (n = 24)
Procedure reference
not reported; sitting
position; best of three
values (within 5%).
One trained
technician
Reference values
from Berglund et al.
(1963)
FEVi, FVC
Estimated average
2fibers/mLinl950s
and 1960s,
1 fiber/mL in 1970s;
levels for specific
work areas estimated
and used for
individual-level
cumulative exposure
Table 6:
Longitudinal
decline among
those with and w/o
pleural plaques,
controlling for
height, age,
tracheal area, f/yr,
and smoking
Ostiguy et al. (1995)
Canada
Copper refinery
7 yr follow up of workers first
evaluated in 1983-1984
Mean (SE) age 46.6 (0.5) yr
Mean (SE) duration 20.6 (0.5) yr
28.7% current smokers
100% current workers
n = 396 original survey
(1983-1984) and 1991 follow-up
(n that did not have follow-up data
not reported); 262 included in
case-control study of pleural
plaques (four to five controls
selected per case)
P-A and lateral views
Two experienced readers (members
of Canadian Pneumoconiosis Reading
Panel), independent readings and then
consensus discussions
Blinded to asbestos exposure
ILO(1980)
Pleural thickening of the chest wall or
diaphragm, without costophrenic
angle obliteration; all plaques were
circumscribed and all readings of
lung parenchyma were in category 0
(n = 54 or 50?)
Costophrenic angle obliteration
(« = 4)
No pleural thickening (n = 440)
Renzetti(1979)
procedures, excluded
those (<1%) not
meeting repeatability
criteria
Same technician and
equipment for
baseline and
follow-up
Percentage predicted
based on Quanjer and
van Zomeren (1983)
FVC, FEVi, MMEF
Asbestos (used in
insulation materials)
gradually removed
from workplace in
the 1980s
Table 7, loss of
FVC by presence
or absence of
pleural plaques
1-45
-------
Supplemental Table I-B. Localized pleural thickening (LPT)—pulmonary function studies evaluation:
longitudinal studies (alphabetical; all x-ray-based) (continued)
Reference, population
Rui et al. (2004)
Italy
3.7 yr follow-up
Workers with pleural lesions:
Mean (SD) age 53 (7) yr
Mean (SD) duration 30 (6) yr
TSFE not reported
22% smokers (<15 pack-yr), 42%
smokers (>15 pack-yr), 36%
nonsmokers
42% current workers
Additional study details provided
in personal communication from
Francesca Rui to L. Kopylev,
3/15/2014.
Sichletidis et al. (2006)
Greece
Residential exposure
1 5 yr follow-up
14 men, 4 women,
Mean (SD) age at follow-up
72.7(6.5) yr
Smoking data not reported
Related reference: Sichletidis et
al. (1992)
Selection
Workers referred to an
occupational medicine clinic
1991-2000; included those with
history of asbestos exposure; had
two spirometry tests performed at
least 1 yr apart; had radiological
examination performed; no signs
of interstitial fibrosis, emphysema,
bronchiectasis, pleurisy, TB, or
other significant lung, cardiac,
skeletal or systemic disease.
Included only those workers with
pleural plaques on x-ray who were
further referred for HRCT.
w = 103
Recruited in seven villages in
northern Greece (asbestos used in
whitewash). Baseline survey in
1 988: 1 98 > age 40 with pleural
plaques (out of 818); 23 of these
had pulmonary function testing;
Follow-up survey in 2003,
126 survivors (18 with baseline
pulmonary function data, 78%)
n = 1 8 for longitudinal study
X-ray measures
One reader for x-ray and HRCT,
blinding not reported. Plaques
detected by x-ray, with HRCT used
to rule out parenchymal disease.
Pleural plaques described by location
(unilateral/bilateral, diaphragmatic)
and presence of calcification; defined
as "circumscribed areas of thickening
of the parietal pleura in thoracic cage
and/or diaphragm"
Pleural plaques only (n = 36, 35%)
Normal (n = 67, 65%)
Details not reported
Two experience physicians,
independent readings
ILOfl980)
Computer-based comparison of 2003
to 1,988 scans
Spirometry
Spirometry procedures
not referenced.
Reference values from
CECA71
FEVi, VC, TLC
Procedure reference
not reported.
Performed at
hospital-based
pulmonary function
laboratory
Percentage predicted
based on Crapo et al.
(1982).
FEVi, FVC,
FEVi/FVC, TLC, RV
Exposure
Exposure duration
data by pleural
plaque and no
pleural plaque
group, respectively:
mean (SD) 30 (6)
and 22 (6) yr;
exposure duration
included in adjusted
analysis
Smoking data by
pleural plaque and
no pleural plaque
group, respectively:
36 and 36%
nonsmokers, 22 and
30% smokers
(< 15 pack-yr), 42
and 34% smokers
(>1 5) pack-yr;
smoking "habit"
included in adjusted
analysis (<1 5 or >1 5
pack-yr)
Discussion notes
exposure had
ceased.
Analysis
Table 3:
longitudinal
analysis.
Generalized
estimating
equations used to
examine changes
in pulmonary
function over
time, adjusting for
presence/absence
of pleural plaques,
smoking habit,
and yr of
exposure.
Table II
(Also information
on progression of
plaques in
n = 126)
1-46
-------
Supplemental Table I-C. Localized pleural thickening (LPT)—pulmonary function studies evaluation:
external comparison (alphabetical within x-ray and high resolution computed tomography [HRCT] groups)
Reference, population
Selection
X-ray or HRCT measures
Spirometry
Exposure
Analysis
X-ray, studies
Ameille et al. (2004)
France (Paris, Normandy)
88% male
Mean (SD) age 58 (9.0) yr
Mean (SD) duration 25.4 (9.4) yr
Mean (SD) TSFE 33.5 (9.4) yr
20% current smokers
Pack-yr not reported
Percentage currently working not
reported
Fridriksson et al. (1981)
Sweden
Mean (SD) age 62.5 (9.8) yr
Mean (SD) duration
22.0 (14.4) yr
Mean (SD) TSFE 38.9 (9.95) yr
29% current smokers
Pack-yr not reported
Percentage current workers not
reported
Consecutive patients referred to
occupational medicine
departments in 1992-1994 for
suspected asbestos related pleural
fibrosis
n = 287 out of 365 with evidence
of pleural thickening (55 excluded
because of no pleural thickening;
18 excluded because of x-ray
quality)
General health survey, Uppsala,
Sweden, 1975-1976. Selected if
pleural plaques and no other
abnormality on x-ray and history
of asbestos exposure, no clinical
lung disease
n = 45 (five additional refusals)
P-A view
Three independent, experienced
readers (chosen from group of four)
Blinding not reported
ILO (2002)
Two definitions of DPT:
Definition 1 : pleural thickening of
the chest wall,
Associated and in continuity with
costophrenic angle obliteration
(11.8%)
Definition 2: pleural thickening at
least 5 mm wide and extending for
more than one quarter of the chest
wall (35. 5%).
Pleural plaques is any pleural
thickening not satisfying the DPT
definition (88.2 or 64.5%)
HRCT scans also used as "gold
standard"
X-ray details not specified
Total n = 45 divided into four
groups:
Grade 1 : pleural plaques only in the
flanks or flanks and diaphragm,
>5 mm thick, noncalcified (n = 7)
Grade 2: visible in P-A view,
noncalcified (n = 17)
Grade 3: Calcified pleural plaques
Grade 1 or 2 (n = 15)
Grade 4: Pleural plaques with
calcification
Grade 3 (n = 6)
ATS fl 987)
procedures,(details not
reported).
Expressed as
percentage predicted,
but reference
population not
specified
FEVi, FVC, TLC
Details not reported
Reference values from
same laboratory
(263 healthy men,
equations account for
age, height, weight,
smoking habits)
FEVi
Cumulative fiber
exposure estimated
for 1 52 patients
(Normandy group):
mean 255 f/cc-yr),
not used in analysis
Duration, 4 -level
intensity measure
(slight or
intermittent light,
more intense
intermittent,
continuous exposure
at moderate levels,
heavy daily
exposure)
(examined in
relation to extent of
pleural plaques)
External
comparison,
(percentage
predicted);
separate analysis
excluding 48 with
parenchymal
abnormalities
External
comparison,
Table 3:
percentage
predicted; also did
a matched analysis
(gender, age
within 4 yr,
3-level smoking
status)
1-47
-------
Supplemental Table I-C. Localized pleural thickening (LPT)—pulmonary function studies evaluation:
external comparison (alphabetical within x-ray and high resolution computed tomography [HRCT] groups)
(continued)
Reference, population
Selection
X-ray or HRCT measures
Spirometry
Exposure
Analysis
X-ray, studies
Hillerdal (1990)
Sweden
Men with history of asbestos
exposure
Mean (SD) age 57 (7) yra
Mean duration 29 yra
TSFE not reported
38% current smokers8
Pack-yr not reported
Percentage currently working not
reported
Hjortsbers et al. (1988)
Sweden (Malmo)
Railroad workers
Median age 57 yr
Median duration 30 yr
44% current smokers
Pack-yr not reported
"mostly" currently working
Clinic-based
Included if:
Age <70 yr
No known heart or other systemic
disease
Bilateral pleural lesions
Willing to participate
w = 23
Initial study 1977-1980 with chest
x-rays;
n = 87 with asbestos induced
pleural plaques selected
(excluding nine with ILO grading
1/1).
P-A + lateral views
Blinding not described
ILO standards (date not given)
Pleural plaques (bilateral), >5 mm
thick, well demarcated, without
obliteration of costophrenic angle
and without pulmonary fibrosis or
involvement of the visceral pleura
(n = 13, 57%); plus three unilateral
DPT (unilateral and bilateral fibrosis
(n = 10, 43%; two with asbestosis)
P-A + lateral views (+ oblique if
uncertain interpretation)
Two readers (trained radiologists)
Blinding not described
Thiringer et al. (1980) definition:
"Distinctly demarcated pleural
thickenings not reaching the apices
or costophrenic sinuses" (n = 87,
100%)
Procedure reference
not given; best of
3 FEVi values
Percentage predicted
based on equations
with smoking variable
using healthy people
not exposed to any
fibrosing agent,
normal x-ray, same
laboratory
FEVi, FEF5o
Procedure reference
not given; details not
provided (other than
sitting).
Reference equations
based on results from
200 nonsmoking men,
ages 20-70 (n = 40 per
decade); healthy, from
workplaces without
lung health hazards
FEVi
External
comparison,
Table 1 and
Figure 3
(comparison based
on reference
population)
External
comparison,
conditional
logistic regression
based on
hypothetical
matched controls
from reference
equations,
stratified by
smoking (see
Table III).
Table IV:
Predictors of
spirometry
(including
exposure)
1-48
-------
Supplemental Table I-C. Localized pleural thickening (LPT)—pulmonary function studies evaluation:
external comparison (alphabetical within x-ray and high resolution computed tomography [HRCT] groups)
(continued)
Reference, population
Kilburn and Warshaw (1990)
and
Kilburn and Warshaw (1991)
United States (20 sites)
Boilermakers and pipefitters
Mean (SD) age 63.3 (8.6) yr>
Duration >5 yr
TSFE>15yr
76% current smokers
McLoud et al. (1985)
United States (Boston)
Asbestos paper mill workers and
other high risk employees
HRCT Studies
Chow et al. (2009)
Sandrini et al. (2006)
Australia
Mean (SD) age 70 (4.23) yr>
42% exsmokers3
(not clear how much overlap
there is in participants)
Selection
General screening, union members
and other tradesmen. Recruitment
strategy not described, (total
eligible may be 4,572)
"Population comparisons" made to
group of 370 Michigan men
(random stratified population
sample) with and without
cardiorespiratory disease
n = 1,298
Screening of high risk workers
(n = 1,135) plus "clinic patients"
(n = 238)
n = 1,373
External controls: 717 university
employees (excluding beryllium
or asbestos exposure)
All men
Dust Disease Board (exposed
workers) and controls with no
asbestos exposure
X-ray or HRCT measures
P-A + lateral views
One reader
Blinding not described
FLO (1980)
Four groups:
(A and B) Pleural plaques only
(n = 45 calcified and n = 98
noncalcified)
(C) DPT without costophrenic angle
obliteration (n = 129)
(D) DPT with costophrenic angle
obliteration (n = 61)
(Groups A, B, and C = pleural
plaques)
P-A view
Two B Readers (for pleural findings)
Blinding not described
FLO (1971)
Plaques (circumscribed) (n = 227,
16.5%)
Diffuse pleural thickening (n= 185,
13.5%; 58 benign asbestos effusion;
47 confluent plaques
HRCT, details not provided
(referenced ATS, 2004)
Pleural plaques and diffuse pleural
thickening definition based on Jones
etal. (1988).
Pleural plaques (n = 26)
Diffuse pleural thickening (n= 16)
Asbestosis (n = 18)
Controls (n = 26)
Spirometry
Renzetti(1979)
procedures, standing
with nose clip (other
details not provided).
Percentage predicted
based on referent
group of 1 88 Michigan
men (random stratified
population sample)
without
cardiorespiratory
disease, adjusting for
height, age, and yr of
smoking.
Procedure reference
not given; details not
provided percentage
predicted based on
Korv etal. (1961)
ATS/ERS (2005)
(details not reported).
Percentage predicted
based on Cotes et al.
(1993)
Exposure
Duration (not used
in analysis)
Analysis
External
comparison
External
comparison, text
External
comparison,
Table 1
1-49
-------
Supplemental Table I-C. Localized pleural thickening (LPT)—pulmonary function studies evaluation:
external comparison (alphabetical within x-ray and high resolution computed tomography [HRCT] groups)
(continued)
Reference, population
Schneider et al. (2012)
Germany
Workers with asbestos disease
Mean (SD) age 55.9 (5.6) yra
Duration not reported
Cumulative exposure 66.7 (113.1)
fiber-yr3
TSFE not reported
27% current smokers*
Meanpack-yr22.1a
Selection
Selected from clinic treating
workers with compensated
asbestos diseases; consecutive
male patients
« = 154
X-ray or HRCT measures
HRCT read by single experienced
radiologist, blinded to clinical status
but aware of asbestos exposure
German (Hering et al., 2004) and
Japanese (Kusakaet al.. 2005) HRCT
guidelines
Pleural Plaques: "circumscribed and
discrete areas of hyaline or calcified
fibrosis localized on the parietal
pleura of the lateral chest wall, the
diaphragm or the mediastinum."
Parietal pleural plaques (n = 63)
Visceral pleural fibrosis (n = 10)
Asbestosis and parietal pleural
plaques (n = 39)
Asbestosis and visceral pleural
fibrosis (n = 42)
Spirometry
European Respiratory
Society procedures
(details not provided),
measures with two
highest attempts with
agreement within 5%
included. Adjusted for
body temperature and
pressure saturated with
water vapor.
Trained technicians
Percentage predicted
from various
references (all
European), including
Ouanieretal. (19931
FEVi, FVC,
FEVi/FVC, MEF5o,
DLco
Exposure
Analysis
Table 2, external
analysis
""Descriptive data for pleural plaque (or LPT) group; when not noted as such, data is for full study sample.
1-50
-------
1.4. REFERENCES
Aberle. PR: Gamsu. G: Ray. CS: Feuerstein. IM. (1988). Asbestos-related pleura! and parenchyma! fibrosis:
Detection with high-resolution CT. Radiology 166: 729-734.
http://dx.doi.0rg/10.1148/radiology.166.3.3340770
Akira. M: Yamamoto. S: Yokoyama. K: Kita. N: Morinaga. K: Higashihara. T: Kozuka. T. (1990). Asbestosis:
High-resolution CT-pathologic correlation. Radiology 176: 389-394.
http://dx.doi.org/10.1148/radiology. 176.2.2367652
Akira. M: Yokoyama. K: Yamamoto. S: Higashihara. T: Morinaga. K: Kita. N: Morimoto. S: Ikezoe. J: Kozuka. T.
(1991). Early asbestosis: Evaluation with high-resolution CT. Radiology 178: 409-416.
http://dx.doi.org/10.1148/radiology. 178.2.1987601
Ameille. J: Matrat M: Paris. C: Joly. N: RaffaeHi. C: Brochard. P: Iwatsubo. Y: Pairon. JC: Letourneux. M. (2004).
Asbestos-related pleural diseases: Dimensional criteria are not appropriate to differentiate diffuse pleural
thickening from pleural plaques. Am J Ind Med 45: 289-296. http://dx.doi.org/10.1002/ajim.10341
ATS (American Thoracic Society). (1987). Standardization of spirometry~1987 update. Am J Respir Crit Care Med
136: 1285-1298.
ATS (American Thoracic Society). (1995). Standardization of spirometry, 1994 update. American Thoracic Society.
Am J Respir Crit Care Med 152: 1107-1136. http://dx.doi.0rg/10.1164/airccm.152.3.7663792
ATS (American Thoracic Society). (2000). What constitutes an adverse health effect of air pollution? Am J Respir
Crit Care Med 161: 665-673. http://dx.doi.0rg/10.1164/airccm.161.2.ats4-00
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.20Q310-1436ST
ATS/ERS (American Thoracic Society and European Respiratory Society). (2005). ATS/ERS recommendations for
standardized procedures for the online and offline measurement of exhaled lower respiratory nitric oxide
and nasal nitric oxide, 2005. Am J Respir Crit Care Med 171: 912-930.
http://dx.doi.org/10.1164/rccm.200406-710ST
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.
Berglund. E: Birath. G: Bjure. J: Grimbv. G: Kjellmer. I: Sandqvist L: Soderholm. B. (1963). Spirometric studies in
normal subjects I: Forced expirograms in subjects between 7 and 70 years of age. J Intern Med 173: 185-
192. http://dx.doi.org/DOI: 10.1111/j.0954-6820.1963.tbl6520.x
Bourbeau. J: Ernst. P: Chrome. J: Armstrong. B: Becklake. MR. (1990). The relationship between respiratory
impairment and asbestos-related pleural abnormality in an active work force. Am Rev Respir Dis 142: 837-
842. http://dx.doi.0rg/10.1164/airccm/142.4.837
Bresnitz. EA: Oilman. MJ: Gracely. EJ: Airoldi. J: Vogel. E: Gefter. W. (1993). Asbestos-related radiographic
abnormalities in elevator construction workers. Am Rev Respir Dis 147: 1341-1344.
http://dx.doi.0rg/10.1164/airccrn/147.6 Pt 1.1341
Broderick. A: Fuortes. LJ: Merchant. JA: Galvin. JR: Schwartz. DA. (1992). Pleural determinants of restrictive lung
function and respiratory symptoms in an asbestos-exposed population. Chest 101: 684-691.
Chow. S: Campbell C: Sandrini. A: Thomas. PS: Johnson. AR: Yates. PH. (2009). Exhaled breath condensate
biomarkers in asbestos-related lung disorders. Respir Med 103: 1091-1097.
http://dx.doi.0rg/10.1016/i.rmed.2009.04.009
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
1-51
-------
Cotes. JE. (1975). Pulmonary function at different stages of life, including reference values. In JE Cotes (Ed.),
Pulmonary function (3rd ed., pp. 340-395). London, UK: Oxford Blackwell Scientific Publications.
Cotes. JE: Chinn. DJ: Quanjer. PH: Roca. J: Yernault. JC. (1993). Standardization of the measurement of transfer
factor (diffusing capacity). Report Working Party Standardization of Lung Function Tests, European
Community for Steel and Coal. Official Statement of the European Respiratory Society [Review]. Eur
Respir J6: 41-52. http://dx.doi.org/10.1183/09041950.041sl693
Crapo. RO: Morris. AH: Clayton. PD: Nixon. CR. (1982). Lung volumes in healthy nonsmoking adults. Bull Europ
PhysiolResp 18: 419-425.
Crapo. RO: Morris. AH: Gardner. RM. (1981). Reference spirometric values using techniques and equipment that
meet ATS recommendations. Am Rev Respir Dis 123: 659-664.
Di Lorenzo. L: Mele. M: Pegorari. MM: Fratello. A: Zocchetti. C: Capozzi. D. (1996). Lung cinescintigraphy in the
dynamic assessment of ventilation and mucociliary clearance of asbestos cement workers. Occup Environ
Med 53: 628-635. http://dx.doi.0rg/10.1136/oem.53.9.628
Dujic. Z: Eterovic. D: Tocilj. J. (1993). Association between asbestos-related pleural plaques and resting
hyperventilation. Scand J Work Environ Health 19: 346-351. http://dx.doi.org/10.5271/sjweh.1464
Fridriksson. HV: Hedenstrom. H: Hillerdal G: Malmberg. P. (1981). Increased lung stiffness of persons with pleural
plaques. Eur J Respir Dis 62: 412-424.
Garcia-Closas. M: Christiani. DC. (1995). Asbestos-related diseases in construction carpenters. Am J Ind Med 27:
115-125. http://dx.doi.org/10.1002/aiim.4700270111
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. http://dx.doi.0rg/10.1164/airccm.159.l.9712108
Hering. KG: Tuengerthal S: Kraus. T. (2004). Standardisierte CT/HRCT-klassifikation der bundesrepublik
deutschland fu r arbeits-und umweltbedingte thoraxerkrankungen. Radiologe 44: 500-511.
Hillerdal. G. (1990). Pleural and parenchymal fibrosis mainly affecting the upper lung lobes in persons exposed to
asbestos. Respir Med 84: 129-134. http://dx.doi.org/10.1016/S0954-6111(08)80015-7
Hilt B: Langard. S: Lund-Larsen. PG: Lien. JT. (1986a). Previous asbestos exposure and smoking habits in the
county of Telemark, Norway-a cross-sectional population study. Scand J Work Environ Health 12: 561-
566. http://dx.doi.org/10.5271/sjweh.2106
Hilt B: Lien. JT: Lund-Larsen. PG. (1987). Lung function and respiratory symptoms in subjects with asbestos-
related disorders: a cross-sectional study. Am J Ind Med 11:517-528.
Hilt B: Lien. JT: Lund-Larsen. PG: Lund. K: Langard. S. (1986b). Asbestos-related findings in chest radiographs of
the male population of the county of Telemark, Norway~a cross-sectional study. Scand J Work Environ
Health 12: 567-573. http://dx.doi.org/10.5271/sjweh.2105
Hjortsberg. U: Orbaek. P: Aborelius. M. Jr: Ranstam. J: Welinder. H. (1988). Railroad workers with pleural plaques:
I. Spirometric and nitrogen washout investigation on smoking and nonsmoking asbestos-exposed workers.
Am J Ind Med 14: 635-641. http://dx.doi.org/10.1002/aiim.4700140602
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
1-52
-------
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.
Jarvholm. B: Larsson. S. (1988). Do pleural plaques produce symptons? A brief report. J Occup Med 30: 345-347.
Jarvholm. B: Sanden. A. (1986). Pleural plaques and respiratory function. Am J Ind Med 10: 419-426.
http://dx.doi.org/10.1002/aiim.4700100409
Jones. RN: Mcloud. T: Rockoff. SD. (1988). The radiographic pleural abnormalities in asbestos exposure:
relationship to physiologic abnormalities [Review]. J Thorac Imaging 3: 57-66.
Kilburn. KH: Warshaw. RH. (1990). Abnormal pulmonary function associated with diaphragmatic pleural plaques
due to exposure to asbestos. Br J Ind Med 47: 611-614. http://dx.doi.0rg/10.1136/oem.47.9.611
Kilburn. KH: Warshaw. RH. (1991). Abnormal lung function associated with asbestos disease of the pleura, the
lung, and both: A comparative analysis. Thorax 46: 33-38. http://dx.doi.0rg/10.1136/thx.46.l.33
Knudson. RJ: Lebowitz. MD: Holberg. CJ: Burrows. B. (1983). Changes in the normal maximal expiratory flow-
volume curve with growth and aging. Am Rev Respir Dis 127: 725-734.
Kory. RC: Callahan. R: Boren. HG: Syner. JC. (1961). The Veterans Administration-Army cooperative study of
pulmonary function: I. Clinical spirometry in normal men. Am J Med 30: 243-258.
http://dx.doi.org/10.1016/0002-9343(61)90096-l
Kusaka. Y: Hering. KG: Parker. JE. (2005). International classification of HRCT for occupational and
environmental respiratory diseases. Tokyo, Japan: Springer.
Larson. TC: Franzblau. A: Lewin. M: Goodman. AB: Antao. VC. (2014). Impact of body mass index on the
detection of radiographic localized pleural thickening. Acad Radiol 21: 3-10.
http://dx.doi.0rg/10.1016/i.acra.2013.09.014
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
Lilis. R: Miller. A: Godbold. J: Benkert. S: Wu. X: Selikoff. IJ. (1992). Comparative quantitative evaluation of
pleural fibrosis and its effects on pulmonary function in two large asbestos-exposed occupational groups--
insulators and sheet metal workers. Environ Res 59: 49-66. http://dx.doi.org/10.1016/S0013-
9351(05)80225-7
Lilis. R: Miller. A: Godbold. J: Chan. E: Benkert. S: Selikoff. IJ. (1991a). The effect of asbestos-induced pleural
fibrosis on pulmonary function: Quantitative evaluation. AnnN Y Acad Sci 643: 162-168.
http://dx.doi.org/10.1111/i. 1749-6632.1991.tb24458.x
Lilis. R: Miller. A: Godbold. J: Chan. E: Selikoff. IJ. (199Ib). Pulmonary function and pleural fibrosis: Quantitative
relationships with an integrative index of pleural abnormalities. Am J Ind Med 20: 145-161.
http://dx.doi.org/10.1002/aiim.4700200203
Lilis. R: Miller. A: Godbold. J: Chan. E: Selikoff. IJ. (1991c). 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
Lynch. DA: Gamsu. G: Aberle. DR. (1989). Conventional and high resolution computed tomography in the
diagnosis of asbestos-related diseases. Radiographics 9: 523-551.
http://dx.doi.0rg/10.1148/radiographics.9.3.2727359
McLoud. TC: Woods. BO: Carrington. CB: Epler. GR: Gaensler. EA. (1985). Diffuse pleural thickening in an
asbestos-exposed population: Prevalence and causes. AJR Am J Roentgenol 144: 9-18.
http://dx.doi.0rg/10.2214/air.144.l.9
1-53
-------
Miller. A: Lilis. R: Godbold. J: Chan. E: Selikoff. IJ. (1992). Relationship of pulmonary function to radiographic
interstitial fibrosis in 2,611 long-term asbestos insulators: An assessment of the International Labour Office
profusion score. Am Rev Respir Dis 145: 263-270. http://dx.doi.0rg/10.1164/airccm/145.2 Pt 1.263
Miller. A: Thornton. JC: Warshaw. R: Bernstein. J: Selikoff. IJ: Teirstein. AS. (1986). Mean and instantaneous
expiratory flows, FVC and FEV1: prediction equations from a probability sample of Michigan, a large
industrial state. Bull Europ Physiol Resp 22: 589-597.
Miller. A: Widman. SA: Miller. JA: Manowitz. A: Markowitz. SB. (2013). Comparison of x-ray films and low-dose
computed tomographic scans: demonstration of asbestos-related changes in 2760 nuclear weapons workers
screened for lung cancer. J Occup Environ Med 55: 741-745.
http://dx.doi.org/10.1097/JOM.Ob013e3182954067
Miller. JA: Zurlo. JV. (1996). Asbestos plaques in a typical Veteran's hospital population. Am J Ind Med 30: 726-
729. http://dx.doi.org/10.1002/(SICI)1097-0274(199612)30:6<726::AID-AJIM9>3.0.CO:2-O
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
Ohlson. CG: Bodin. L: Rydman. T: Hogstedt C. (1985). Ventilatory decrements in former asbestos cement workers:
a four year follow up. Br J Ind Med 42: 612-616.
Ohlson. CG: Rydman. T: Sundell L: Bodin. L: Hogstedt C. (1984). Decreased lung function in long-term asbestos
cement workers: a cross-sectional study. Am J Ind Med 5.
Oldenburg. M: Degens. P: Baur. X. (2001). Asbest-bedingte Lungenfunktionseinschrankungen mit und ohne
Pleuraplaques. Atemwegs-undLungenkrankheiten27: 422-423.
Oliver. LC: Eisen. EA: Greene. R: Sprince. NL. (1988). Asbestos-related pleural plaques and lung function. Am J
Ind Med 14: 649-656. http://dx.doi.org/10.1002/aiim.4700140604
Oliver. LC: Eisen. EA: Greene. RE: Sprince. NL. (1985). Asbestos-related disease in railroad workers: A cross-
sectional study. Am Rev Respir Dis 131: 499-504.
Ostiguv. G: Vaillancourt C: Begin. R. (1995). Respiratory health of workers exposed to metal dusts and foundry
fumes in a copper refinery. Occup Environ Med 52: 204-210.
Paoletti P: Pistelli. G: Fazzi. P: Viegi. G: Pi Pede. F: Giuliano. G: Prediletto. R: Carrozzi. L: Polato. R: Saetta. M:
Zambon. R: Sapigni. T: Lebowitz. MD: Giuntini. C. (1986). Reference values for vital capacity and flow-
volume curves from a general population study. Eur Respir J 22: 451-459.
Paoletti. P: Viegi. G: Pistelli. G: Pi Pede. F: Fazzi. P: Polato. R: Saetta. M: Zambon. R: Carli. G: Giuntini. C.
(1985). Reference equations for the single-breath diffusing capacity: A cross-sectional analysis and effect
of body size and age. Am Rev Respir Dis 132: 806-813.
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
Quanjer. PH. Dalhuijsen. A.: van Zomeren. BC. (1983). Summary equations of reference values. Bull Euro
Physiopath Respir 19: 45-51.
Quanjer. PH: Tammeling. GJ: Cotes. JE: Pedersen. OF: Peslin. R: Yernault JC. (1993). Lung volumes and forced
ventilatory flows. Report Working Party Standardization of Lung Function Tests, European Community for
Steel and Coal. Official Statement of the European Respiratory Society. Eur Respir J 16: 5-40.
Renzetti. AD. Jr. (1979). Standardization of spirometry. Am Rev Respir Dis 119: 693-694.
1-54
-------
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
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
Rui F: De Zotti R: Negro. C: Bovenzi. M. (2004). [A follow-up study of lung function among ex-asbestos workers
with and without pleura! plaques]. Med Lav 95: 171-179.
Sandrini. A: Johnson. AR: Thomas. PS: Yates. DH. (2006). Fractional exhaled nitric oxide concentration is
increased in asbestosis and pleural plaques. Respirology 11:325-329. http://dx.doi.org/10.1111/i. 1440-
1843.2006.00852.x
Schneider. J: Arhelger. R: Raab. W: Hering. KG. (2012). The validity of static lung compliance in asbestos-induced
diseases. Lung 190: 441-449. http://dx.doi.org/10.1007/s00408-012-9388-6
Schwartz. DA: Fuortes. LJ: Calvin. J. R.: Burmeister. LF: Schmidt. LE: Leistikow. BN: Lamarte. FP: Merchant JA.
(1990). Asbestos-induced pleural fibrosis and impaired lung function. Am Rev Respir Dis 141: 321-326.
http://dx.doi.0rg/10.1164/airccm/141.2.321
Selikoff. IJ: Hammond. EC. (1979). Asbestos and smoking. JAMA 242: 458-459.
http://dx.doi.org/10.1001/iama.1979.03300050048029
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. (1992). Pleural plaques in a rural
population in central Macedonia, Greece. Med Lav 83: 259-265.
Simundic. I: Jankovic. T: Tocilj. J: Gjakun. K: Juric. I: Roglic. J: Captain. V. (2002). Diagnostic value of clinical,
radiologic and functional examinations of patients with pulmonary and pleural asbestosis. Acta Clin Croat
41:73-78.
Singh. B: Eastwood. PR: Finucane. KE: Panizza. JA: Musk. AW. (1999). Effect of asbestos-related pleural fibrosis
on excursion of the lower chest wall and diaphragm. Am J Respir Crit Care Med 160: 1507-1515.
http://dx.doi.0rg/10.1164/airccm.160.5.9806135
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
Staples. CA: Gamsu. G: Ray. CS: Webb. WR. (1989). High resolution computed tomography and lung function in
asbestos-exposed workers with normal chest radiographs. Am Rev Respir Dis 139: 1502-1508.
http://dx.doi.0rg/10.1164/airccm/139.6.1502
Thiringer. G: Blomqvist N: Brolin. I: Mattson. SB. (1980). Pleural plaques in chest x-rays of lung cancer patients
and matched controls (preliminary results). Eur J Respir Dis Suppl 107: 119-122.
U.S. EPA (U.S. Environmental Protection Agency). (2013). Integrated science assessment for lead [EPA Report].
(EPA/600/R-10/075F). Research Triangle Park, NC.
http://ofmpub.epa. gov/eims/eimscomm. getfile?p download id=514513
van Cleemput. J: de Raeve. H: Verschakelen. JA: Rombouts. J: Lacquet LM: Nemery. B. (2001). Surface of
localized pleural plaques quantitated by computed tomography scanning: No relation with cumulative
asbestos exposure and no effect on lung function. Am J Respir Crit Care Med 163: 705-710.
http://dx.doi.0rg/10.1164/airccm.163.3.2006017
1-55
-------
Viechtbauer. W. (2010). Conducting meta-analyses in R with the metafor package. J Stat Softw 36: 1-48.
Weill. D: Dhillon. G: Frevder. L: Lefante. J: Glindmever. H. (2011). Lung function, radiological changes and
exposure: Analysis of ATSDR data from Libby, MT, USA. EurRespir J 38: 376-383.
http://dx.doi.org/10.1183/09031936.00050210
Wilken. D: Velasco Garrido. M: Manuwald. U: Baur. X. (2011). Lung function in asbestos-exposed workers, a
systematic review and meta-analysis. J Occup Med Toxicol 6:21. http://dx.doi.org/10.1186/1745-6673-6-
21
Yoshimura. H: Hatakevama. M: Otsuii. H: Maeda. M: Ohishi. H: Uchida. H: Kasuga. H: Katada. H: Narita. N:
Mikami. R: Konishi. Y. (1986). Pulmonary asbestosis: CT study of subpleural curvilinear shadow. Work in
progress. Radiology 158: 653-658. http://dx.doi.0rg/10.1148/radiologv.158.3.3945733
Zavalic. M: Bogadi-Sare. A. (1993). Lung functions and chest radiographs in shipyard workers exposed to asbestos.
Arh Hig Rada Toksikol 44: 1-8.
1-56
-------
APPENDIX J.
DOCUMENTATION OF IMPLEMENTATION OF THE
2011 NATIONAL RESEARCH COUNCIL RECOMMENDATIONS
Background: On December 23, 2011, The Consolidated Appropriations Act, 2012, was
signed into law.l The report language included direction to EPA for the Integrated Risk
Information System (IRIS) Program related to recommendations provided by the National
Research Council (NRC) in its review of EPA's draft IRIS assessment of formaldehyde.2 The
report language included the following:
The Agency shall incorporate, as appropriate, based on chemical-specific data sets
and biological effects, the recommendations of Chapter 7 of the National
Research Council's Review of the Environmental Protection Agency's Draft IRIS
Assessment of Formaldehyde into the IRIS process.. .For draft assessments
released in fiscal year 2012, the Agency shall include documentation describing
how the Chapter 7 recommendations of the National Academy of Sciences (NAS)
have been implemented or addressed, including an explanation for why certain
recommendations were not incorporated.
The NRC's recommendations, provided in Chapter 7 of the review report, offered
suggestions to EPA for improving the development of IRIS assessments. Consistent with the
direction provided by Congress, documentation of how the recommendations from Chapter 7 of
the NRC report have been implemented in this assessment is provided in the tables below.
Where necessary, the documentation includes an explanation for why certain recommendations
were not incorporated.
The IRIS Program's implementation of the NRC recommendations is following a phased
approach that is consistent with the NRC's "Roadmap for Revision" as described in Chapter 7 of
the formaldehyde review report. The NRC stated that, "the committee recognizes that the
changes suggested would involve a multiyear process and extensive effort by the staff at the
National Center for Environmental Assessment and input and review by the EPA Science
Advisory Board and others."
The IRIS LAA assessment is in Phase 1 of implementation, which focuses on a subset of
the short-term recommendations, such as editing and streamlining documents, increasing
transparency and clarity, and using more tables, figures, and appendices to present information
. L. No. 112-74, Consolidated Appropriations Act, 2012.
2National Research Council, (2011). Review of the Environmental Protection Agency's Draft IRIS Assessment of
Formaldehyde.
J-l
-------
and data in assessments. Phase 1 also focuses on assessments near the end of the development
process and close to final posting. Chemical assessments in Phase 2 of the implementation will
address all of the short-term recommendations from Table J-l. The IRIS Program is
implementing all of the recommendations, but recognizes that achieving full and robust
implementation of certain recommendations will be an evolving process with input and feedback
from the public, stakeholders, and external peer review committees. Chemical assessments in
Phase 3 of implementation will incorporate the longer-term recommendations made by the NRC
as outlined below in Table J-2, including the development of a standardized approach to describe
the strength of evidence for noncancer effects.
In May 2014, the NRC released their report reviewing the IRIS assessment development
process. The NRC stated that the changes that EPA has proposed or implemented in the IRIS
process represented substantial improvements. As part of this review, the NRC made several
recommendations with respect to the systematic review and integration of scientific evidence for
chemical hazard and dose-response assessments that will be implemented in future assessments.
The NRC also stated that EPA should continue to improve its evidence-integration process
incrementally and enhance the transparency of its process. The NRC recommendations will
inform the IRIS Program's efforts in this area going forward.
J-2
-------
Table J-l. EPA's implementation of the 2011 National Research Council's
recommendations in the Libby Amphibole asbestos (LAA) assessment
NRC recommendations that the EPA is
implementing in the short term
Implementation in the LAA assessment
General recommendations for completing the IRIS formaldehyde assessment that the EPA will adopt
for all IRIS assessments (p. 152)
1. To enhance the clarity of the document, the
draft IRIS assessment needs rigorous editing to
reduce the volume of text substantially and
address redundancies and inconsistencies.
Long descriptions of particular studies should
be replaced with informative evidence tables.
When study details are appropriate, they could
be provided in appendices.
Implemented. The main body of the LAA
assessment has been rigorously edited by disciplinary
experts. The longer descriptions of the animal and
mechanistic studies are now contained in Appendix
D. Overall, the revised assessment contains 29 new
tables and seven new figures to improve transparency
and increase clarity. In direct response to a
recommendation from the external peer review by
EPA's Science Advisory Board (SAB), anew
appendix has been added to present a systematic
literature review of studies related to localized pleural
thickening and lung function. This appendix supports
the selection of the noncancer critical effect of
localized pleural thickening.
J-3
-------
Table J-l. EPA's implementation of the 2011 National Research Council's
recommendations in the Libby Amphibole asbestos (LAA) assessment
NRC recommendations that the EPA is
implementing in the short term
Implementation in the LAA assessment
Chapter 1 needs to be expanded to describe
more fully the methods of the assessment,
including a description of search strategies used
to identify studies with the exclusion and
inclusion criteria articulated and a better
description of the outcomes of the searches and
clear descriptions of the weight-of-evidence
approaches used for the various noncancer
outcomes. The committee emphasizes that it is
not recommending the addition of long
descriptions of the EPA guidelines to the
introduction, but rather clear concise statements
of criteria used to exclude, include, and
advance studies for derivation of the reference
concentrations (RfCs) and unit risk estimates.
Partially Implemented. Clear and concise
statements of criteria used to evaluate and advance
studies for the derivation of the RfC are in Section
5.2.1. At the recommendation of the SAB, EPA
conducted a systematic literature review of studies
related to localized pleural thickening and lung
function. A systematic literature search was done
covering the time span to December 2013; this search
is described in detail in Appendix I.
Statements of criteria used to evaluate and advance
studies for the literature review of localized pleural
thickening and lung functions are in Appendix I.
Standardized evidence tables for all health
outcomes need to be developed. If there were
appropriates tables, long text descriptions of
studies could be moved to an appendix or
deleted.
Partially Implemented. Long text descriptions of
laboratory animal and mechanistic studies were
moved to an appendix (Appendix D), and the tables
were formatted in response to the SAB peer review
panel recommendations. Standardized evidence
tables of studies related to localized pleural
thickening are included in the new Appendix I.
4. All critical studies need to be thoroughly
evaluated with standardized approaches that are
clearly formulated and based on the type of
research: for example, observational
epidemiologic or animal bioassays. The
findings of the reviews might be presented in
tables to ensure transparency.
Partially Implemented. All critical studies were
thoroughly evaluated in Section 4, with laboratory
animal study details presented in Appendix D. The
candidate studies for the derivation of the noncancer
critical effect were thoroughly reviewed with the
evaluation criteria and study findings presented in
tables in Section 5.
J-4
-------
Table J-l. EPA's implementation of the 2011 National Research Council's
recommendations in the Libby Amphibole asbestos (LAA) assessment
NRC recommendations that the EPA is
implementing in the short term
Implementation in the LAA assessment
The rationales for the selection of the studies
that are advanced for consideration in
calculating the RfCs and unit risks need to be
expanded. All candidate RfCs should be
evaluated together with the aid of graphic
displays that incorporate selected information
on attributes relevant to the database.
Implemented. Section 5.2.1 is devoted to the
selection of the principal study for the derivation of
the RfC. This section contains two tables outlining
the characteristics of the candidate studies and the
rationale for selecting the principal study. As
different subcohorts within the principal study were
also used to derive candidate RfCs, those findings are
presented for comparison in Table 5-11. NIOSH
provided data from the largest occupational cohort
study of workers exposed to LAA. The NIOSH data
set provided information on individuals' exposure
data with extended cancer mortality follow-up and
these data were used to derive the cancer unit risk.
Comparisons of the lung cancer and mesothelioma
unit risks from the LAA assessment with alternative
derivations from related analyses of the same
principal study are shown in Tables 5-54 and 5-55.
Strengthened, more integrative and more
transparent discussions of weight of evidence
are needed. The discussions would benefit
from more rigorous and systematic coverage of
the various determinants of weight of evidence,
such as consistency.
Partially Implemented. The weight of evidence
discussion in the LAA assessment has been expanded
to include a formal mode-of-action analysis (see
Section 4.6.2). This analysis includes discussion of
the available data supporting the weight of evidence
descriptor for cancer, with specific discussion of
dose-response concordance, temporal relationship
biological plausibility and human relevance. For
noncancer, a formal meta-analysis was used for
determining the critical effect of localized pleural
thickening (critical effect) on lung function measures
(Appendix I).
General Guidance for the Overall Process (seep. 164)
1. Elaborate an overall, documented, and
quality-controlled process for IRIS
assessments.
Ensure standardization of review and
evaluation approaches among contributors and
teams of contributors; for example, include
standard approaches for reviews of various
types of studies to ensure uniformity.
Implemented.
EPA has created discipline-specific workgroups
to formalize an internal process to provide
additional overall quality control for the
development of IRIS assessments. This initiative
uses a team approach to making timely, consistent
decisions about the development of IRIS
assessments across the Program. This team
J-5
-------
Table J-l. EPA's implementation of the 2011 National Research Council's
recommendations in the Libby Amphibole asbestos (LAA) assessment
NRC recommendations that the EPA is
implementing in the short term
9. Assess disciplinary structure of teams needed
to conduct the assessments.
Implementation in the LAA assessment
approach has been utilized for the development of
the LAA assessment. Additional objectives of the
teams are to help ensure that the necessary
disciplinary expertise is available for assessment
development and review, to provide a forum for
addressing peer review recommendations, and to
monitor progress in implementing the NRC
recommendations .
Evidence Identification: Literature Collection and Collation Phase (seep. 164)
10. Select outcomes on the basis of available
evidence and understanding of mode of action.
1 1 . Establish standard protocols for evidence
identification.
12. Develop a template for description of the
search approach.
13. Use a database, such as the Health and
Environmental Research Online (HERO)
database, to capture study information and
relevant quantitative data.
Implemented. The LAA assessment has detailed
discussions of mechanistic studies on the biological
response to LAA, including inflammation,
genotoxicity and cytotoxicity (see Section 4.4 and
Appendix D) and includes a formal mode of action
framework analysis for carcinogenicity (see
Section 4.6.2).
Partially Implemented. To update the LAA
assessment in response to the SAB recommendations,
a systematic literature search of the critical noncancer
health effect (localized pleural thickening) was done
covering the time span to December 2013; this search
is described in detail in Appendix I.
This is being fully implemented by the IRIS program
as part of Phase 2.
This is being fully implemented by the IRIS program
as part of Phase 2.
Implemented. HERO links were incorporated for all
citations.
Evidence Evaluation: Hazard Identification and Dose-Response Modeling (see p. 165)
14. Standardize the presentation of reviewed
studies in tabular or graphic form to capture
the key dimensions of study characteristics,
weight of evidence, and utility as a basis for
deriving reference values and unit risks.
Partially Implemented. The LAA Assessment
includes tables describing the available
epidemiological studies of LAA-exposed
populations. With respect to the key study, the
assessment also presents the characteristics of
workers and attributes of exposure in tabular and
graphical form.
J-6
-------
Table J-l. EPA's implementation of the 2011 National Research Council's
recommendations in the Libby Amphibole asbestos (LAA) assessment
NRC recommendations that the EPA is
implementing in the short term
Implementation in the LAA assessment
15. Develop templates for evidence tables, forest
plots, or other displays.
This is being fully implemented by the IRIS program
as part of Phase 2.
16. Establish protocols for review of major types
of studies, such as epidemiologic and
bioassay.
Partially Implemented. This assessment was
developed using standard protocols for evidence
evaluation that are provided in existing EPA
guidance.
Selection of Studies for Derivation of Reference Values and Unit Risks (seep. 165)
17. Establish clear guidelines for study selection.
a. Balance strengths and weaknesses.
b. Weigh human vs. experimental evidence.
c. Determine whether combining estimates
among studies is warranted.
Partially Implemented. As discussed above, the
text has been expanded to include more description of
the considerations made in selecting the study that
formed the basis for the quantitative cancer risk
estimates (see Section 5.2.1). The selection
considerations are also summarized in a table (see
Table 5-1 and Table 5-2).
Calculation of Reference Values and Unit Risks (see pp. 165-166)
18. Describe and justify assumptions and models
used. This step includes review of dosimetry
models and the implications of the models for
uncertainty factors; determination of
appropriate points of departure (such as
benchmark dose, no-observed-adverse-effect
level, and lowest observed-adverse-effect
level), and assessment of the analyses that
underlie the points of departure.
Implemented. The LAA assessment has a detailed
discussion of model selection for the epidemiological
data sets (see Section 5.2.2).
19. Provide explanation of the risk-estimation
modeling processes (for example, a statistical
or biologic model fit to the data) that are used
to develop a unit risk estimate.
Implemented. The LAA assessment has a detailed
discussion of model selection for the epidemiological
data sets (see Section 5.2.2).
20. Provide adequate documentation for
conclusions and estimation of reference values
and unit risks. As noted by the committee
throughout the present report, sufficient
support for conclusions in the formaldehyde
draft IRIS assessment is often lacking. Given
that the development of specific IRIS
assessments and their conclusions are of
interest to many stakeholders, it is important
that they provide sufficient references and
supporting documentation for their
conclusions. Detailed appendixes, which
might be made available only electronically,
should be provided, when appropriate.
Implemented. The LAA assessment includes
documentation of the estimates of the RfC in Section
5.2 and the IUR in Section 5.4.5. This includes clear
explanation of the methods used to develop the LAA
assessment, and description of the decisions made in
developing the hazard identification and dose-
response analyses. As recommended, supplementary
information and analyses are presented in appendices.
J-7
-------
Table J-2. National Research Council (2011) recommendations that the EPA is
generally implementing in the long term
NRC recommendations that the EPA is implementing
in the long term
Implementation in the LAA Assessment
Weight-of-Evidence Evaluation: Synthesis of Evidence
for Hazard Identification (seep. 165)
1. Review use of existing weight-of-evidence
guidelines.
2. Standardize approach to using weight-of-evidence
guidelines.
3. Conduct agency workshops on approaches to
implementing weight-of-evidence guidelines.
4. Develop uniform language to describe strength of
evidence on noncancer effects.
5. Expand and harmonize the approach for
characterizing uncertainty and variability.
6. To the extent possible, unify consideration of
outcomes around common modes of action rather
than considering multiple outcomes separately.
Partially implemented. As indicated above,
Phase 3 of EPA's implementation plan will
incorporate the longer-term recommendations
made by the NRC (2011). In addition, NRC
recently released a report on its review of
current methods for weight-of-evidence
analyses and EPA will implement revisions to
address the recommended approaches for
weighing scientific evidence for chemical
hazard identification in NRC (2014). In
addition, EPA held workshops in August 2013
and October 2014 on issues related to weight-
of-evidence to inform future assessments.
Calculation of Reference Values and Unit Risks (see
pp. 165-166)
1. Assess the sensitivity of derived estimates to model
assumptions and end points selected. This step
should include appropriate tabular and graphic
displays to illustrate the range of the estimates and
the effect of uncertainty factors on the estimates.
Implemented. The LAA assessment presents
derivation of multiple alternative RfC estimates
for different subcohorts of the principal study,
for different health endpoints, and for the
application of different uncertainty factors. The
LAA assessment also presents multiple
derivations of unit risk estimates for different
exposure-response models.
J-8
------- |