DRAFT—DO NOT CITE OR QUOTE EPA/635/R-11/002A
www.epa.gov/iris
v>EPA
TOXICOLOGICAL REVIEW
OF
LIBBY AMPHIBOLE ASBESTOS
In Support of Summary Information on the
Integrated Risk Information System (IRIS)
August 2011
(Note: This document is an assessment of the noncancer and cancer health effects
associated with the inhalation route of exposure only)
NOTICE
This document is an External Review draft. This information is distributed solely for the
purpose of predissemination peer review under applicable information quality guidelines. It has
not been formally disseminated by EPA. It does not represent and should not be construed to
represent any Agency determination or policy. It is being circulated for review of its technical
accuracy and science policy implications.
U.S. Environmental Protection Agency
Washington, DC
-------
-------
DISCLAIMER
This document is a preliminary draft for review purposes only. This document is
distributed solely for the purpose of predissemination peer review under applicable information
quality guidelines. It has not been formally disseminated by EPA. It does not represent and
should not be construed to represent any Agency determination or policy. Mention of trade
names or commercial products does not constitute endorsement or recommendation for use.
This document is a draft for review purposes only and does not constitute Agency policy.
iii DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS—TOXICOLOGICAL REVIEW OF LIBBY AMPHIBOLE ASBESTOS
LIST OF TABLES x
LIST OF FIGURES xiii
LIST OF ABBREVIATIONS AND ACRONYMS xiv
FOREWORD xviii
AUTHORS, CONTRIBUTORS, AND REVIEWERS xix
1. INTRODUCTION 1-1
1.1. RELATED ASSESSMENTS 1-3
1.1.1. IRIS Assessment for Asbestos (U.S. EPA, 1988a) 1-3
1.1.2. EPA Health Assessment for Vermiculite (1991b) 1-4
1.2. LIBBY AMPHIBOLE ASBESTOS-SPECIFIC HUMAN HEALTH
ASSESSMENT 1-4
2. LIBBY AMPHIBOLE ASBESTOS: GEOLOGY, USE, AND EXPOSURE
POTENTIAL 2-1
2.1. HISTORICAL BACKGROUND 2-1
2.2. GEOLOGY AND MINERALOGY OF LIBBY AMPHIBOLE ASBESTOS 2-3
2.2.1. Silicate Minerals 2-4
2.2.1.1. Mineralogy and Structure of Amphiboles 2-5
2.2.1.2. Amphibole Morphology 2-7
2.2.2. Vermiculite 2-12
2.2.3. The Mineralogy of Libby Amphibole Asbestos 2-14
2.2.3.1. Mineralogy 2-14
2.2.3.2. Morphology of the Libby Amphibole Asbestos 2-17
2.2.3.3. Dimensional Characteristics of Libby Amphibole Asbestos.... 2-18
2.3. EXPOSURE POTENTIAL 2-20
2.3.1. Libby Community 2-21
2.3.2. Communities near Vermiculite Expansion and Processing Plants 2-24
2.3.3. Exposures from Zonolite and Vermiculite for Homeowners,
Contractors, and Other Populations 2-24
3.FIBERTOXICOKINETICS 3-1
3.1. DEPOSITION OF FIBERS IN THE RESPIRATORY TRACT 3-3
3.2. CLEARANCE 3-8
3.2.1. Inhalation 3-8
3.2.1.1. Respiratory Tract 3-8
3.2.1.2. Pleural Cavity and Extrapulmonary Sites 3-13
3.2.2. Ingestion 3-14
3.2.3. Dermal Contact 3-14
3.3. SUMMARY 3-15
4. HAZARD IDENTIFICATION OF LIBBY AMPHIBOLE ASBESTOS 4-1
4.1. STUDIES IN HUMANS—EPIDEMIOLOGY 4-1
4.1.1. Studies of Libby, MT Vermiculite Mining Operation Workers 4-2
This document is a draft for review purposes only and does not constitute Agency policy.
iv DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS (continued)
4.1.1.1. Description of Mining and Milling Operations 4-3
4.1.1.2. Exposure Estimation 4-4
4.1.1.3. Cancer Mortality Risk 4-14
4.1.1.4. Noncancer Effects: Respiratory and Cardiovascular
Disease 4-20
4.1.2. Libby, MT Community Studies 4-28
4.1.2.1. Geographic Mortality Analysis 4-28
4.1.2.2. Community Screening—Respiratory Health 4-29
4.1.2.3. Other Reports of Asbestos-Related Disease Among Libby,
MT Residents 4-35
4.1.2.4. Summary of Respiratory Health Effects in Libby, MT
Community Studies 4-35
4.1.3. Marysville, OH Vermiculite Processing Plant Worker Studies 4-36
4.1.3.1. Summary of Marysville, OH Vermiculite Processing Plant
Worker Studies 4-44
4.1.4. Community Studies from Other Vermiculite Processing Plants 4-44
4.1.4.1. Summary of Community Studies from Other Vermiculite
Processing Plants 4-46
4.1.5. Case Reports 4-48
4.2. SUBCHRONIC AND CHRONIC STUDIES AND CANCER BIO AS SAYS
IN ANIMALS—ORAL, INHALATION AND OTHER ROUTES OF
EXPOSURE 4-48
4.2.1. Oral 4-49
4.2.2. Inhalation 4-49
4.2.3. Intratracheal Instillation Studies 4-54
4.2.4. Injection/Implantation Studies 4-56
4.2.5. Summary of Animal Studies for Libby Amphibole and Tremolite
Asbestos 4-58
4.3. OTHER DURATION OR ENDPOINT-SPECIFIC STUDIES 4-59
4.3.1. Immunological 4-59
4.4. MECHANISTIC DATA AND OTHER STUDIES IN SUPPORT OF THE
MODE OF ACTION 4-62
4.4.1. Inflammation and Immune Function 4-63
4.4.2. Genotoxicity 4-67
4.4.3. Cytotoxicity and Cellular Proliferation 4-69
4.5. SYNTHESIS OF MAJOR NONCANCER EFFECTS 4-71
4.5.1. Pulmonary Effects 4-71
4.5.1.1. Pulmonary Fibrosis (Asbestosis) 4-71
4.5.1.2. Other Nonmalignant Respiratory Diseases 4-73
4.5.2. Pleural Effects 4-73
4.5.3. Other Noncancer Health Effects (Cardiovascular Toxicity,
Autoimmune Effects) 4-74
4.5.4. Libby Amphibole Asbestos Summary of Noncancer Health Effects 4-75
This document is a draft for review purposes only and does not constitute Agency policy.
v DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS (continued)
4.5.5. Mode-of-Action Information (Noncancer) 4-76
4.6. EVALUATION OF CARCINOGENICITY 4-77
4.6.1. Summary of Overall Weight of Evidence 4-77
4.6.1.1. Synthesis of Human, Animal, and Other Supporting
Evidence 4-78
4.6.2. Mode-of-Action Information 4-80
4.6.2.1. Description of the Mode-of-Action Information 4-80
4.6.2.2. Application of the Age-Dependent Adjustment Factors 4-81
4.7. SUSCEPTIBLE POPULATIONS 4-81
4.7.1. Influence of Different Lifestages on Susceptibility 4-82
4.7.l.l.Lifestage Susceptibility 4-82
4.7.2. Influence of Gender on Susceptibility 4-88
4.7.3. Influence of Race or Ethnicity on Susceptibility 4-89
4.7.4. Influence of Genetic Polymorphisms on Susceptibility 4-89
4.7.5. Influence of Health Status on Susceptibility 4-91
4.7.6. Influence of Lifestyle Factors on Susceptibility 4-92
4.7.7. Susceptible Populations Summary 4-93
5. EXPOSURE-RESPONSE ASSESSMENT 5-1
5. LORAL REFERENCE DO SE(RfD) 5-1
5.2. INHALATION REFERENCE CONCENTRATION (RfC) 5-1
5.2.1. Choice of Principal Study and Critical Effect 5-1
5.2.1.1. Evaluation of Candidate Studies and Selection of Critical
Study 5-3
5.2.1.2. Evaluation of Exposure Paradigm in Candidate Studies 5-3
5.2.1.3. Evaluation of Exposure Assessment in Candidate Studies 5-12
5.2.1.4. Selection of Critical Effect 5-15
5.2.2. Evaluation of Radiographic Lesions as Potential Critical Effects 5-16
5.2.2.1. Health Effects of Parenchymal Changes as Small Opacities
Viewed on Standard Radiographs 5-16
5.2.2.2. Health Effects of Diffuse Pleural Thickening (DPT)
Viewed on Standard Radiographs 5-17
5.2.2.3. Health Effects of Localized Pleural Thickening (LPT)
Viewed on Standard Radiographs 5-18
5.2.3. Methods of Analysis 5-23
5.2.3.1. Exposure Data and Choice of Exposure Metric 5-23
5.2.3.2. Data Sets for Modeling Analyses 5-26
5.2.3.3. Statistical Modeling of the Sub-cohort 5-29
5.2.4. RfC Derivation—Including Application of Uncertainty Factors
(UFs) 5-35
5.2.5. Alternative Analyses of the Full Marysville Cohort 5-39
5.2.6. Previous Reference Concentration (RfC) Derivation 5-41
This document is a draft for review purposes only and does not constitute Agency policy.
vi DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS (continued)
5.3. UNCERTAINTIES IN THE INHALATION REFERENCE
CONCENTRATION (RfC) 5-41
5.3.1. Uncertainty in the Exposure Reconstruction 5-41
5.3.2. Uncertainty in the Radiographic Assessment of Localized Pleural
Thickening 5-44
5.3.3. Uncertainty Due to Time From First Exposure 5-45
5.3.4. Uncertainty in Background Rate of Localized Pleural Thickening 5-45
5.3.5. Uncertainty in Model Functional Form and Lagged Exposure 5-46
5.3.6. Uncertainty Due to Effect of Smoking 5-46
5.3.7. Sensitivity Analysis: Derivation of a POD for Lifetime Exposure
From the Cumulative Exposure Metric 5-48
5.3.8. Sensitivity Analysis for Choice of Critical Effect and Selection of
Benchmark Response (BMR) 5-50
5.4. CANCER EXPOSURE-RESPONSE ASSESSMENT 5-51
5.4.1. Overview of Methodological Approach 5-51
5.4.2. Choice of Study/Data—with Rationale and Justification 5-53
5.4.2.1. DescriptionoftheLibby Worker Cohort 5-54
5.4.2.2. Description of Cancer Endpoints 5-56
5.4.2.3. Description of Libby Amphibole Asbestos Exposures 5-58
5.4.2.4. Description of Libby Worker Cohort Work Histories 5-65
5.4.2.5. Estimated Exposures Based on Job-Exposure Matrix
(JEM) and Work Histories 5-66
5.4.3. Exposure-Response Modeling 5-73
5.4.3.1. Modeling of Mesothelioma Exposure Response in the
Libby Worker Cohort 5-74
5.4.3.2. Mesothelioma Mortality Analysis in the Libby Worker
Cohort 5-76
5.4.3.3. Modeling of Lung Cancer Exposure Response in the Libby
Worker Cohort 5-78
5.4.3.4. Lung-Cancer Mortality Analysis in the Libby Worker
Cohort 5-80
5.4.3.5. Summary of Mesothelioma and Lung Cancer Analysis of
Libby Worker Cohort 5-81
5.4.3.6. Analysis of Sub-Cohort of Employees Hired After 1959 5-83
5.4.4. Exposure Adjustments and Extrapolation Methods 5-97
5.4.5. Inhalation Unit Risk (IUR) of Cancer Mortality 5-97
5.4.5.1. Unit Risk Estimates for Mesothelioma Mortality 5-97
5.4.5.2. Unit Risk Estimates for Lung-Cancer mortality 5-100
5.4.5.3. IUR Derivation for Combined Mesothelioma and
Lung-Cancer Mortality 5-102
5.4.5.4. Applications of the Combined Mesothelioma and
Lung-cancer mortality IUR to Partial Lifetime
Environmental Exposure Scenarios 5-109
This document is a draft for review purposes only and does not constitute Agency policy.
vii DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS (continued)
5.4.6. Uncertainties in the Cancer Risk Values 5-111
5.4.6.1. Sources of Uncertainty 5-111
5.4.6.2. Summary 5-129
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 6-3
6.1.3. Noncancer Health Effects in Humans and Laboratory Animals 6-4
6.1.4. Carcinogenicity in Humans and Laboratory Animals 6-6
6.1.5. Susceptible Populations 6-7
6.1.6. Mode-of-Action Information 6-7
6.1.7. Weight-of-Evidence Descriptor for Cancer Hazard 6-8
6.2. EXPOSURE RESPONSE 6-9
6.2.1. Noncancer/Inhalation 6-9
Uncertainty and Sensitivity Analyses for RfC Derivation 6-12
6.2.2. Cancer/Inhalation 6-16
6.2.2.1. Background and Methods 6-16
6.2.3. Modeling of Mesothelioma Exposure Response 6-19
6.2.4. Unit Risk Estimates for Mesothelioma Mortality 6-20
6.2.5. Modeling of Lung Cancer Exposure Response 6-21
6.2.6. Unit Risk Estimates for Lung-Cancer Mortality 6-22
6.2.7. IUR Derivation Based on Combined Mesothelioma and
Lung-Cancer Mortality from Exposure to Libby Amphibole
Asbestos 6-24
6.2.7.1. Comparison with Other Published Studies of Libby, MT
Workers Cohort 6-25
6.2.8. Sensitivity Analysis 6-25
6.2.8.1. Sensitivity Analysis of Influence of High Exposures in
Early 1960s on Model Fit 6-25
6.2.8.2. Analysis of Potential Confounding of Lung Cancer Results
by Smoking in the Sub-cohort 6-26
6.2.9. Uncertainty in the Cancer Risk Values 6-26
6.3. APPLICATION OF THE LIBBY AMPHIBOLE ASBESTOS RfC AND IUR.... 6-28
6.3.1. Sites and Materials 6-28
6.3.2. Exposure Units for Libby Amphibole Asbestos 6-29
6.3.3. Applications to Early Lifetime and Partial Lifetime Environmental
Exposure Scenarios for IUR 6-30
6.3.4. Applications to Lifetime and Partial Lifetime Environmental
Exposure Scenarios for RfC 6-31
7. REFERENCES 7-1
This document is a draft for review purposes only and does not constitute Agency policy.
viii DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS (continued)
APPENDIX A: SUMMARY OF EXTERNAL PEER-REVIEW AND PUBLIC
COMMENTS AND DISPOSITION 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, MT; LOUISA COUNTY, VA; AND
PALABORA, REPUBLIC OF SOUTH AFRICA C-l
APPENDIX D: ANALYSIS OF SUBCHRONIC AND CHRONIC STUDIES AND
CANCER BIO AS SAYS IN ANIMALS AND MECHANISTIC
STUDIES D-l
APPENDIX E: EVALUATION OF EXPOSURE RESPONSE DATA FOR DISCRETE
PLEURAL THICKENING IN WORKERS FROM THE
MARYSVILLE, OH COHORT E-l
APPENDIX F: MARYSVILLE, OH WORKER OCCUPATIONAL EXPOSURE
RECONSTRUCTION F-l
APPENDIX G: EXTRA RISK AND UNIT RISK CALCULATION G-l
APPENDIX H: GLOSSARY OF ASBESTOS TERMINOLOGY H-l
This document is a draft for review purposes only and does not constitute Agency policy.
ix DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF TABLES
1-1. Derivation of the current IRIS inhalation unit risk for asbestos from the lifetime
risk tables in the AAHAU 1-4
2-1 Properties of vermiculite 2-13
2-2. Air sampling results for asbestos from Zonolite VAI in three homes 2-26
3-1. Factors influencing fiber deposition and clearance in the respiratory system 3-4
4-1. Exposure assessment methodologies used in evaluations of Libby, MT and
Marysville, OH worker cohorts 4-5
4-2. Source of primary samples for fiber measurements at the Libby mining and
milling operations 4-6
4-3. Dimensional characteristic of fibers from air samples collected in the vermiculite
mill and screening plant, Libby, MT 4-10
4-4. Respiratory (lung) cancer mortality and exposure-response analyses based on
studies of the vermiculite mine workers in Libby, MT 4-11
4-5. Mesothelioma mortality risk based on studies of the vermiculite mine workers in
Libby, MT 4-18
4-6. Nonmalignant respiratory mortality studies of the vermiculite mine workers in
Libby, MT 4-21
4-7. Chest radiographic studies of the Libby, MT vermiculite mine workers 4-24
4-8. Cancer mortality and nonmalignant respiratory disease mortality in the Libby, MT
community 4-30
4-9. Pulmonary function and chest radiographic studies in the Libby, MT community 4-31
4-10. Pulmonary function and chest radiographic studies of the Marysville, OH
vermiculite processing plant workers 4-38
4-11. Prevalence of pleural radiographic abnormalities according to quartiles of
cumulative fiber exposure in 280 participants 4-40
4-12. Prevalence of pleural thickening in 280 participants according to various cofactors.... 4-41
4-13. Description of study areas in ATSDR health consultations evaluating cancer
incidence and mortality 4-45
4-14. Incidence and mortality results for potential asbestos-related cancers (by cancer
site) in communities in the vicinity of vermiculite-processing facilities (with
ATSDR health consultations evaluating potential pathways of exposure) 4-47
4-15. In vivo data following exposure to Libby Amphibole asbestos 4-50
4-16. In vivo data following exposure to tremolite asbestos 4-52
4-17. Autoimmune-related studies in the Libby, MT community 4-61
4-18. In vitro data following exposure to Libby Amphibole asbestos 4-64
4-19. In vitro data following exposure to tremolite asbestos 4-65
5-1. Summary of candidate principal studies on Libby Amphibole asbestos for
reference concentration (RfC) derivation 5-4
5-2. Summary of rationale for identifying candidate principal studies on Libby
Amphibole asbestos for RfC development 5-8
5-3. Distribution of cases and time from first exposure (7) for cohort of Marysville
workers 5-28
This document is a draft for review purposes only and does not constitute Agency policy.
x DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF TABLES (continued)
5-4. Candidate models for association between cumulative Libby Amphibole asbestos
exposure in the Marysville sub-cohort and localized pleural thickening 5-33
5-5. Modeling of alternative endpoints in the Marysville worker cohort members
examined in 2002-2005 5-51
5-6. Demographic and exposure characteristics of the Libby worker cohort 5-55
5-7. Exposure intensity (fibers/cc) for each location operation from the beginning of
operations through 1982 [Amandus et al. (1987a); Table VII] 5-60
5-8. Demographic and exposure characteristics of the subset of the Libby worker
sub-cohort hired after 1959 5-65
5-9. Comparison of univariate model fit of various exposure metrics for mesothelioma
mortality in the full Libby worker cohort (n = 1,871) 5-77
5-10. Comparison of model fit of exposure metrics for mesothelioma mortality in the
sub-cohort hired after 1959. Only the model fits with information weights greater
than 0.010 are shown 5-85
5-11. Mesothelioma mortality exposure metrics fits, slopes, and credible intervals 5-86
5-12. Model fit comparison for different exposure metrics and lung-cancer mortality
associated with Libby Amphibole asbestos, controlling for age, gender, race, and
date of birth 5-88
5-13. Lung-cancer mortality exposure metrics fits, slopes, and confidence intervals for
all retained metrics from Table 5-12. 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 5-92
5-14. Sensitivity analysis of model fit comparison for different exposure metrics and
lung-cancer mortality associated with Libby Amphibole asbestos. Estimated
exposure intensities for all jobs during 1960-1963 were reduced by 50% 5-94
5-15. Sensitivity analysis of model fit comparison for different exposure metrics and
lung-cancer mortality associated with Libby Amphibole asbestos, 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 5-95
5-16. Mesothelioma mortality exposure metrics unit risks 5-98
5-17. Adjusted for underascertainment unit risks for the sub-cohort hired after 1959
corresponding to the different metrics 5-99
5-18. 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 5-101
5-19. Reasonable upper bound and lowest information criteria estimates of central risks
and unit risks, per fibers/cc, for mesothelioma mortality, lung-cancer mortality,
and the IUR for the combined mortality risk from mesothelioma and lung cancer .... 5-104
This document is a draft for review purposes only and does not constitute Agency policy.
xi DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF TABLES (continued)
5-20. 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 5-106
5-21. Mesothelioma regression results from different analyses of cumulative exposure
in the cohort of workers in Libby, MT. 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 5-110
6-1. Reasonable upper bound and lowest information criteria estimates of central risks
and unit risks, for mesothelioma mortality, lung-cancer mortality, and the IUR for
the combined mortality risk from mesothelioma and lung cancer (lURs are
presented in the units of excess cancers per fibers/cc, continuous lifetime
exposure) 6-24
This document is a draft for review purposes only and does not constitute Agency policy.
xii DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF FIGURES
2-1. Vermiculite mining operation on Zonolite Mountain, Libby, Montana 2-1
2-2. Expanded vermiculite and vermiculite attic insulation (VAI) shown in place
between ceiling joists 2-2
2-3. Nationwide distribution of Libby ore by county 2-3
2-4. Structure of the silicate minerals, illustrating silicate subclasses by the linking of
the basic silicon tetrahedron into more complex structures 2-6
2-5. Cross-section of amphibole fibers showing the silicon tetrahedrons that make up
each double-chain plate 2-8
2-6. Comparison of crystalline forms amphibole minerals 2-9
2-7. Vermiculite ore sample 2-13
2-8. Mineralogy of Libby Amphibole asbestos structures from samples taken from the
Zonolite Mountain site 2-15
2-9. Solution series linking tremolite, winchite, and richterite amphibole fibers 2-16
2-10. Scanning electron microscope image of amphibole mineral structures from the
Libby, MT mine 2-17
2-11. Fiber morphology of amphibole asbestos from the Libby, MT mine viewed under
a transmission electron microscope 2-18
2-12. Particle size of fibers in Libby ore and Libby air 2-19
3-1. General scheme for fiber deposition, clearance, and translocation of fibers from
the lung and GI tract 3-2
4-1. Location of 28 sites included in the Phase 1 community evaluations conducted by
ATSDR 4-45
5-1. Estimated and measured exposure concentrations in Marysville, OH facility 5-24
5-2. Graph of observed and estimated prevalence of localized pleural thickening
calculated using the Michaelis-Menten model with 10-year lagged exposure 5-35
5-3. Plot of the NIOSH j ob-exposure matrix for different j ob categories over time 5-59
5-4. Histogram showing the number of workers who experienced each incremental
number of different j obs among the 880 workers hired after 1959 5-66
This document is a draft for review purposes only and does not constitute Agency policy.
xiii DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF ABBREVIATIONS AND ACRONYMS
2-D two-dimensional
3-D three-dimensional
AAHAU Airborne Asbestos Health Assessment Update
AIC Akaike information criterion
AM amosite
APC antigen-presenting cells
ATSDR Agency for Toxic Substances and Disease Registry
BMC benchmark concentration
BMCL lower 95% confidence limit of the benchmark concentration
BMD benchmark dose
BMDL lower 95% confidence limit of the benchmark dose
BMI body mass index
BMR benchmark response
cc cubic centimeter
CDF cumulative distribution frequency
CE cumulative exposure
CHEEC cumulative human equivalent exposure for continuous exposure
CI confidence interval
COPD chronic obstructive pulmonary disease
COX-2 cyclooxygenase-2
CYP cytochrome P450
DHE dehydroergosterol
DIG deviance information criterion
DLCO single breath carbon monoxide diffusing capacity
DNA deoxyribonucleic acid
ECX effective concentration
ECSOD extracellular superoxide dismutase
EDX energy dispersive x-ray analysis
EPA U.S. Environmental Protection Agency
EPMA electron probe microanalysis
FEV1 forced expiratory volume in one second
FVC forced vital capacity
GSH glutathione
This document is a draft for review purposes only and does not constitute Agency policy.
xiv DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF ABBREVIATIONS AND ACRONYMS (continued)
GST glutathione S-transferase
HAEC human airway epithelial cells
HKNM human pleural mesothelial cells
HO heme oxygenase
HPRT hypoxanthine-guanine phosphoribosyltransferase
HRCT high resolution computed tomography
HTE hamster tracheal epithelial
IARC International Agency for the Research on Cancer
ICD International Classification of Diseases
ICRP International Commission Radiological Protection
IFN interferon
IH industrial hygiene
IL interleukin
ILO International Labour Organization
IRIS Integrated Risk Information System
IUR inhalation unit risk
JEM job exposure matrix
LAA Libby Amphibole asbestos
LECX lowest effective concentration
LDH lactate dehydrogenase
MCMC Markov chain Monte Carlo
MESA Mining Enforcement and Safety Administration
MIP-2 macrophage inflammatory protein-2
MnSOD manganese superoxide dismutase
MOA mode of action
mppcf million particles per cubic foot
MPPD multipath particle dosimetry
MSHA Mine Safety and Health Administration
NAT2 N-acetyl-transferase 2
NCHS National Center for Health Statistics
NDI National Death Index
Nf2 neurofibromatosis 2
NIEHS National Institute of Environmental Health Sciences
This document is a draft for review purposes only and does not constitute Agency policy.
xv DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF ABBREVIATIONS AND ACRONYMS (continued)
NIOSH National Institute for Occupational Safety and Health
NMRD non-malignant respiratory disease
NRC National Research Council
NTP National Toxicology Program
NVSR National Vital Statistics Report
Oggl 8-oxoguanine-DNA-glycosylase 1
OSHA Occupational Safety and Health Administration
PARP poly(ADP-ribose)polymerase
PBS phosphate buffered saline
PCM phase contrast microscopy
PCMe phase contrast microscopy equivalent
PHS Public Health Service
PM2.5 particulate matter 2.5 um diameter or less
POD point of departure
RfC reference concentration
RfD reference dose
RNA ribonucleic acid
ROS reactive oxygen species
RPM rat pleural mesothelial
RR relative risk
RT-PCR reverse transcription polymerase chain reaction
RTW residence time-weighted
SE Standard Error
SEER Surveillance, Epidemiology, and End Results
SEM scanning electron microscopy
SH spontaneously hypertensive
SHE Syrian hamster embryo
SHHF spontaneously hypertensive-heart failure
SIR standardized incidence ratio
SMR standardized mortality ratio
SOD superoxide dismutase
SPF specific-pathogen-free
SRR standardized rate ratio
This document is a draft for review purposes only and does not constitute Agency policy.
xvi DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF ABBREVIATIONS AND ACRONYMS (continued)
SSA/Ro52 autoantibody marker for apoptosis
STEM scanning transmission electron microscopy
SV40 simian virus 40
TEM transmission electron microscopy
TLC total lung capacity
TSFE time since first exposure
TWA time-weighted average
UCL upper confidence limit
UF uncertainty factor
USGS United States Geological Survey
VAI vermiculite attic insulation
WHO World Health Organization
WKY Wistar-Kyoto rat
XRCC1 X-ray repair cross complementing protein 1
XRD X-ray diffraction
This document is a draft for review purposes only and does not constitute Agency policy.
xvii DRAFT—DO NOT CITE OR QUOTE
-------
FOREWORD
The purpose of this Toxicological Review is to provide scientific support and rationale
for the hazard and dose-response assessment in IRIS pertaining to chronic inhalation exposure to
Libby Amphibole asbestos, a unique mixture of asbestos fibers originating from the vermiculite
mine near Libby, MT. It is not intended to be a comprehensive treatise on the agent or
toxicological nature of Libby Amphibole asbestos. The purpose of this document is to establish
a Libby Amphibole asbestos-specific reference concentration to address noncancer health effects
and to characterize the carcinogenic potential and establish an inhalation unit risk for Libby
Amphibole asbestos-related lung cancer and mesothelioma mortality.
The intent of Section 6, Major Conclusions in the Characterization of Hazard and
Exposure Response, is to present the significant conclusions reached in the derivation of the
reference dose, reference concentration, and cancer assessment where applicable, and to
characterize the overall confidence in the quantitative and qualitative aspects of hazard and dose
response by addressing the quality of data and related uncertainties. The discussion is intended
to convey the limitations of the assessment and to aid and guide the risk assessor in the ensuing
steps of the risk assessment process.
For other general information about this assessment or other questions relating to IRIS,
the reader is referred to EPA's IRIS Hotline at (202) 566-1676 (phone), (202) 566-1749 (fax), or
hotline.iris@epa.gov (e-mail address).
This document is a draft for review purposes only and does not constitute Agency policy.
xviii DRAFT—DO NOT CITE OR QUOTE
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS
CHEMICAL MANAGERS/AUTHORS
Thomas F. Bateson, ScD, MPH
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
Robert Benson, PhD
Region 8
Office of Partnerships and Regulatory Assistance
U.S. Environmental Protection Agency
Denver, CO
Danielle DeVoney, PhD, DABT, PE
Captain, U.S. Public Health Service
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
AUTHORS
Krista Yorita Christensen, PhD
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
Glinda Cooper, PhD
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
Rebecca Dzubow, MPH, MEM
Office of Children's Health Protection
U.S. Environmental Protection Agency
Washington, DC
Maureen R. Gwinn, PhD, DABT
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
This document is a draft for review purposes only and does not constitute Agency policy.
xix DRAFT—DO NOT CITE OR QUOTE
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS (continued)
Leonid Kopylev, PhD
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Washington, DC
CONTRIBUTING AUTHORS
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
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
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
This document is a draft for review purposes only and does not constitute Agency policy.
xx DRAFT—DO NOT CITE OR QUOTE
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS (continued)
CONTRACTOR SUPPORT
William Brattin, PhD
Syracuse Research Corporation
Denver, CO
ECFlex, Inc., Fairborn, OH
Heidi Glick
Debbie Kleiser
Crystal Lewis
Lana Wood
IntelliTech Systems, Inc., Fairborn, OH
Cris Broyles
Kathleen Secor
REVIEWERS
This document has been provided for review to EPA scientists and interagency reviewers from
other federal agencies and White House offices.
This document is a draft for review purposes only and does not constitute Agency policy.
xxi DRAFT—DO NOT CITE OR QUOTE
-------
1 1. INTRODUCTION
2 This document presents background information and justification for the Integrated Risk
3 Information System (IRIS) Summary of the hazard and exposure-response assessment of Libby
4 Amphibole asbestos,1 a mixture of amphibole fibers identified in the Rainy Creek complex and
5 present in ore from the vermiculite mine near Libby, MT. IRIS Summaries may include oral
6 reference dose (RfD) and inhalation reference concentration (RfC) values for chronic and other
7 exposure durations, and a carcinogenicity assessment. This assessment reviews the potential
8 hazards, both cancer and noncancer health effects, from exposure to Libby Amphibole asbestos
9 and provides quantitative information for use in risk assessments: an RfC for noncancer and an
10 inhalation unit risk addressing cancer risk. Libby Amphibole asbestos-specific data are not
11 available to support RfD or cancer slope factor derivations for oral exposures.
12 An RfC is typically defined as "an estimate (with uncertainty spanning perhaps an order
13 of magnitude) of a continuous inhalation exposure to the human population (including sensitive
14 subgroups) that is likely to be without an appreciable risk of deleterious effects during a
15 lifetime." In the case of Libby Amphibole asbestos, the RfC is expressed in terms of the lifetime
16 exposure in units of fibers per cubic centimeter of air (fibers/cc) in units of the fibers as
17 measured by phase contrast microscopy (PCM). The inhalation RfC for Libby Amphibole
18 asbestos considers toxic effects for both the respiratory system (portal-of-entry) and for effects
19 peripheral to the respiratory system (extrarespiratory or systemic effects) that may arise after
20 inhalation of Libby Amphibole asbestos. In this assessment, the estimates of hazard are derived
21 from modeling cumulative exposures from human data, and thus for exposures of less than a
22 lifetime the risk assessor should calculate a lifetime average concentration to compare to the
23 RfC.
24 The carcinogenicity assessment provides information on the carcinogenic hazard
25 potential of the substance in question, and quantitative estimates of risk from inhalation
26 exposures are derived. The information includes a weight-of-evidence judgment of the
27 likelihood that the agent is a human carcinogen and the conditions under which the carcinogenic
28 effects may be expressed. Quantitative risk estimates are derived from the application of a low-
1 The term "Libby Amphibole asbestos" is used in this document to identify the mixture of amphibole mineral fibers
of varying elemental composition (e.g., winchite, richterite, tremolite, etc.) that have been identified in the Rainy
Creek complex near Libby, MT. It is further described in Section 2.2.
This document is a draft for review purposes only and does not constitute Agency policy.
1 -1 DRAFT—DO NOT CITE OR QUOTE
-------
1 dose extrapolation procedure from human data. An inhalation unit risk (IUR) is typically
2 defined as a plausible upper bound on the estimate of cancer risk per ug/m3 air breathed for
3 70 years. For Libby Amphibole asbestos, the RfC is expressed as a Lifetime Daily Exposure in
4 fibers/cc (in units of the fibers as measured by PCM), and the IUR is expressed as cancer risk per
5 fibers/cc (in units of the fibers as measured by PCM).
6 Development of these hazard identification and exposure-response assessments for Libby
7 Amphibole asbestos has followed the general guidelines for risk assessment as set forth by the
8 National Research Council (1983). U.S. Environmental Protection Agency (EPA) Guidelines
9 and Risk Assessment Forum technical panel reports that may have been used in the development
10 of this assessment include the following: Guidelines for the Health Risk Assessment of Chemical
11 Mixtures (U.S. EPA. 1986c), Guidelines for Mutagenicity Risk Assessment (U.S. EPA. 1986b),
12 Recommendations for and Documentation of Biological Values for Use in Risk Assessment (U.S.
13 EPA, 1988b), Guidelines for Developmental Toxicity Risk Assessment (U.S. EPA, 199 la),
14 Interim Policy for Particle Size and Limit Concentration Issues in Inhalation Toxicity (U.S. EPA,
15 1994a), Methods for Derivation of Inhalation Reference Concentrations and Application of
16 Inhalation Dosimetry (U.S. EPA, 1994b), Use of the Benchmark Dose Approach in Health Risk
17 Assessment (U.S. EPA, 1995), Guidelines for Reproductive Toxicity Risk Assessment (U.S. EPA,
18 1996), Guidelines for Neurotoxicity Risk Assessment (U.S. EPA, 1998), Science Policy Council
19 Handbook: Risk Characterization (U.S. EPA, 2000c), Benchmark Dose Technical Guidance
20 Document (U.S. EPA, 2000a), Supplementary Guidance for Conducting Health Risk Assessment
21 of Chemical Mixtures (U.S. EPA, 2000d), A Review of the Reference Dose and Reference
22 Concentration Processes (U.S. EPA, 2002), Guidelines for Carcinogen Risk Assessment (U.S.
23 EPA, 2005a), Supplemental Guidance for Assessing Susceptibility from Early-Life Exposure to
24 Carcinogens (U.S. EPA, 2005b), Science Policy Council Handbook: Peer Review (U.S. EPA,
25 2006d), and A Framework for Assessing Health Risks of Environmental Exposures to Children
26 (U.S. EPA, 2006b).
27 The literature search strategy employed for this assessment is based on EPA's National
28 Center for Environmental Assessment's Health and Environmental Research Outline database
29 tool (which includes PubMed, MEDLINE, Web of Science, JSTOR, and other literature
30 sources). The key search terms included the following: Libby Amphibole, tremolite, asbestos,
31 richterite, winchite, amphibole, and Libby, MT. The relevant literature was reviewed through
This document is a draft for review purposes only and does not constitute Agency policy.
1 -2 DRAFT—DO NOT CITE OR QUOTE
-------
1 July 2011. Any pertinent scientific information submitted by the public to the IRIS Submission
2 Desk was also considered in the development of this document.
o
6
4 1.1. RELATED ASSESSMENTS
5 1.1.1. IRIS Assessment for Asbestos QJ.S. EPA, 1988a)
6 The IRIS assessment for asbestos was posted online in IRIS in 1988 and includes an IUR
7 of 0.23 excess cancers per 1 fiber/cc (U.S. EPA, 1988a) (this unit risk is given in units of the
8 fibers as measured by PCM). The IRIS IUR for general asbestos is derived by estimation of
9 excess cancers for a continuous lifetime exposure and is based on the central tendency—not the
10 upper bound—of the risk estimates (U.S. EPA, 1988a) and is applicable to exposures across a
11 range of exposure environments and types of asbestos (CAS Number 1332-21-4). Although
12 other cancers have been associated with asbestos (e.g., laryngeal, stomach, ovarian) (Straif et al.,
13 2009), the IRIS IUR for asbestos accounts for only lung cancer and mesothelioma. Additionally,
14 pleural and pulmonary effects from asbestos exposure (e.g., localized pleural thickening,
15 asbestosis, and reduced lung function) are well documented, though, currently, there is no RfC
16 for these noncancer health effects.
17 The derivation of the unit risk for general asbestos is based on the Airborne Asbestos
18 Health Assessment Update (AAHAU) (U.S. EPA, 1986a). The AAHAU provides various cancer
19 potency factors and mathematical models of lung cancer and mesothelioma mortality based on
20 synthesis of data from occupational studies and presents estimates of lifetime cancer risk for
21 continuous environmental exposures (0.0001 fiber/cc and 0.01 fiber/cc) (U.S. EPA, 1986a) (see
22 Table 6-3). For both lung cancer and mesothelioma, life-table analysis was used to generate risk
23 estimates based on the number of years of exposure and the age at onset of exposure. Although
24 various exposure scenarios were presented, the unit risk is based on a lifetime continuous
25 exposure from birth. The final asbestos IUR is 0.23 excess cancer per 1 fiber/cc continuous
26 exposure2 and was established by the EPA Carcinogen Risk Assessment Verification Endeavor
27 workgroup and posted on the IRIS database in 1988 (U.S. EPA. 1988a) (see Table 1-1).
28
2 An IUR of 0.23 can be interpreted as a 23% increase in lifetime risk of dying from mesothelioma or lung cancer
with each 1 fiber/cc increase in continuous lifetime exposure.
This document is a draft for review purposes only and does not constitute Agency policy.
1 -3 DRAFT—DO NOT CITE OR QUOTE
-------
Table 1-1. Derivation of the current IRIS inhalation unit risk for asbestos
from the lifetime risk tables in the AAHAU
Gender
Female
Male
All
Excess deaths per 100,000a
Mesothelioma
183
129
156
Lung cancer
35
114
74
Total
218.5
242.2
230.3
Risk
2.18 x 10
2.42 x 10
2.30 x 10
Unit risk
0.23
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
"Data are for exposure at 0.01 fibers/cc for a lifetime.
AAHAU = Airborne Asbestos Health Assessment Update.
Source: U.S. EPA (1988a).
1.1.2. EPA Health Assessment for Vermiculite (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 vermiculite from 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 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
Libby Amphibole asbestos is a complex mixture of amphibole fibers—both
mineralogically and morphologically (see Section 2.2). The mixture primarily includes
tremolite, winchite, and richterite fibers with trace amounts of magnesioriebeckite, 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
Libby Amphibole asbestos 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:
This document is a draft for review purposes only and does not constitute Agency policy.
1 -4 DRAFT—DO NOT CITE OR QUOTE
-------
1 Amandus et al., 1988; Amandus et al., 1987a: Amandus and Wheeler, 1987; Amandus et al.,
2 1987b: McDonald et al.. 1986a: McDonald et al.. 1986b: Lockevetal., 1984).
3 The IRIS database has an IUR3 for asbestos based on a synthesis of 14 epidemiologic
4 studies that included occupational exposure to chrysotile, amosite, or mixed mineral exposures
5 (chrysotile, amosite, crocidolite) (U.S. EPA, 1988a, 1986a). There is some uncertainty in
6 applying the resulting IUR for asbestos to exposure environments and minerals different from
7 those analyzed in the AAHAU (U.S. EPA. 1986a). There is currently no RfC, RfD, or oral slope
8 factor derived for asbestos on the IRIS database.
3 For purposes of this document, termed "IRIS IUR."
This document is a draft for review purposes only and does not constitute Agency policy.
1 -5 DRAFT—DO NOT CITE OR QUOTE
-------
1 2. LIBBY AMPHIBOLE ASBESTOS: GEOLOGY, USE, AND EXPOSURE POTENTIAL
2 2.1. HISTORICAL BACKGROUND
3 The term Libby Amphibole asbestos4 refers to various mineral forms of amphibole
4 asbestos found in the rocks and ore of Zonolite Mountain, 6 miles northeast of Libby, MT (see
5 Figure 2-1). Zonolite Mountain contains a large vermiculite deposit that has been mined since
6 the early 1920s for various commercial uses. Vermiculite miners, mill workers, and those
7 working in the processing plants were exposed to these amphibole fibers, which remain within
8 the vermiculite ore and product. As amphibole asbestos is present in the geological deposit from
9 which the vermiculite ore was being mined, workers were exposed to asbestos fibers during
10 various activities such as extracting ore from the mine, transporting ore and waste rock, milling
11 operations, and shipping the final product (Meeker et al., 2003; Amandus et al., 1987a:
12 McDonald et al., 1986a). Mortality and morbidity studies on the mine and mill workers from
13 Libby have reported adverse health effects in these workers including lung cancer, mesothelio-
14 ma, nonmalignant respiratory disease (NMRD; e.g., asbestosis), and pleural abnormalities
15 (McDonald et al.. 2004: Amandus and Wheeler. 1987: Amandus et al.. 1987b: McDonald et al..
16 1986a: McDonald et al.. 1986b): Sullivan. 2007, 709497;Larson, 2010, 711560;Moolgavkar,
17 2010,709457}. Pleural abnormalities and
18 signs of interstitial fibrosis have also been
19 reported in workers exfoliating and
20 processing expanded Libby vermiculite in
21 other facilities (Rohs et al., 2008: Lockey et
22 al.. 1984).
23 The primary commercial product from
24 the Zonolite mining operation was vermiculite
25 concentrate, which is produced by screening
26 and grading the ore to enrich for the raw
Figure 2-1. Vermiculite mining operation on
27 vermiculite mineral. The unexpanded mineral ^^^ Mountain, Libby, Montana.
4The term "Libby Amphibole asbestos" is used in this document to identify the mixture of amphibole mineral fibers
of varying elemental composition (e.g., winchite, richterite, tremolite, etc.) that have been identified in the Rainy
Creek complex near Libby, MT. It is further described in Section 2.2.
This document is a draft for review purposes only and does not constitute Agency policy.
2-1 DRAFT—DO NOT CITE OR QUOTE
-------
1 exhibits a sheetlike structure that is seen in related minerals (e.g., mica) (see Figure 2-2).
2
3
4
5
6
7
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Figure 2-2. Expanded vermiculite (a) and vermiculite attic insulation (b)
(VAI) shown in place between ceiling joists.
When heated to approximately 150°C, the vermiculite mineral expands like popcorn into
a light porous material. This process of expanding the mineral ore is termed "exfoliation" or
"popping" and occurs when the silicate sheets within the ore are rapidly dehydrated by applying
high heat. Libby Amphibole asbestos fibers were released during the energetic and other kinetic
processing of the ore and vermiculite concentrate, potentially exposing workers.
A portion of the vermiculite concentrate was exfoliated in Libby, MT and either used
locally or packaged and shipped for use elsewhere. However, most of the vermiculite
concentrate was transported across the country and elsewhere to expansion plants where it was
exfoliated and distributed. The Agency for Toxic Substances and Disease Registry (ATSDR,
2008b) has surveyed 28 of these facilities, identifying potential community exposures both to
amphibole asbestos fibers from the vermiculite concentrate before exfoliation, during exfoliation,
and during processing and in waste rock from the processing plants (see Section 4.1.4 and
Figure 2-3). Vermiculite from the Libby, MT mine was used commercially from the 1920s to
1990, and a review of company records from 1964-1990 indicates that approximately
6,109,000 tons of vermiculite concentrate was shipped to over 200 facilities (ATSDR, 2008b).
Expanded vermiculite from the Libby, MT site was used in numerous consumer and construction
products: including attic insulation, packing material, and soil conditioners, and in the production
This document is a draft for review purposes only and does not constitute Agency policy.
2-2 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
j-..t\ ij n:< K< by county
^P JDD.ODD or mars
% 2DD.ODO 10 299.599
0 1DD.ODD1D 1BB.SBB
O 1 CD SB.8M
Figure 2-3. 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 to 1990. EPA tabulated this shipping information
in a database.
Source: U.S. GAO (2007).
of gypsum wall board. There is also potential for exposure to Libby Amphibole asbestos in these
products (see Section 2.4).
2.2. GEOLOGY AND MINERALOGY OF LIBBY AMPHIBOLE ASBESTOS
A large vermiculite deposit is located on Zonolite Mountain, northeast of Libby, MT,
within a geologic unit known as the Rainy Creek complex. Geologic processes within the Rainy
Creek complex have resulted in the formation of fibrous amphiboles adjacent to igneous
intrusions into the complex (veins and dikes of alkaline granite, pegmatite, and quartz)
(Boettcher et al., 1996). The amphibole fibers identified fall within the tremolite-richterite-
This document is a draft for review purposes only and does not constitute Agency policy.
2-3 DRAFT—DO NOT CITE OR QUOTE
-------
1 magnesioriebecktite solid solution series (e.g., winchite, richterite, and tremolite) (Meeker et al.,
2 2003). An appropriate understanding of the mineralogy and geology of these materials is helpful
3 in defining the mineral fibers in Libby Amphibole asbestos.
4 Geological terms provide fiber and mineral definitions based on habit of formation and
5 fiber morphology. Conversely, the analytical methods that have been used to count fibers in air
6 samples, in both historical and current exposure environments, define microscopic fibers based
7 on dimensional characteristics and mineralogy (depending on the analytical method). Current
8 analytical methods do not have specific procedures for determining fiber morphology at the
9 microscopic level. Because the human and experimental animal data on adverse health effects of
10 asbestos rely on available analytical methods to document exposure, these definitions are
11 relevant to determining what constitutes a fiber for this health assessment. Therefore, available
12 data on the fiber morphology and fiber-size distribution of Libby Amphibole asbestos are
13 presented in the following sections.
14
15 2.2.1. Silicate Minerals
16 Silicate minerals are basically made up of oxygen and silicon, two of the most abundant
17 elements in the Earth's crust. Approximately 25% of known minerals and 40% of the common
18 minerals are silicates. Silicate minerals are hard, infusible, and have very low solubility in strong
19 mineral acids. Specific gravity ranges from fairly light to intermediate, luster is commonly
20 glassy, and most crush to a light powder even when the bulk specimen is black prior to crushing.
21 Silicates chiefly occur as components of rocks, segregations in rocks, or crystals lining cavities
22 in rocks. Most hard silicates are primary minerals (i.e., mineral forms that have not undergone
23 oxidative weathering). Secondary silicates have undergone oxidative weathering and contain
24 water of hydration (Dana et al., 1977). Silicate minerals can be defined by chemical structure,
25 crystal structure, trace minerals, and habit of formation.
26 The basic chemical unit of silicate crystalline structure is the [SiO/i]4 tetrahedron-shaped
27 anionic group. The basic unit consists of four oxygen molecules at the apices of a regular
28 tetrahedron surrounding and coordinated with one silicon ion (Si4+) at the center. The chemistry
29 is such that the oxygen molecules can bond to another silicon ion and, therefore, link one
30 [SiO/j]4" tetrahedron to another, and then another, and so forth by the process of polymerization.
31 The silicates can form as single tetrahedrons, double tetrahedrons, chains, sheets, rings and
This document is a draft for review purposes only and does not constitute Agency policy.
2-4 DRAFT—DO NOT CITE OR QUOTE
-------
1 framework structures (see Figure 2-4). More complex three-dimensional structures tectosilicates
2 (frameworks) may also form mineral fibers (e.g., erionite).
3 Each subclass of silicates has many mineral members. Specific minerals are defined by
4 the structure, chemistry, and morphology of the mineral. The minerals of interest in this
5 assessment are various forms of amphiboles (double-chain inosilicates) and vermiculite (a
6 phyllosilicate) (see Figure 2-4).
7
8 2.2.1.1. Mineralogy and Structure of Amphiboles
9 The mineralogy of amphiboles is important to understanding which mineral forms are
10 present in the Libby vermiculite mine, and, therefore, considered to be Libby Amphibole
11 asbestos. Amphibole minerals are double-chain inosilicates, meaning the chemical building
12 block for amphiboles is connected chains of the silicon tetrahedron (see Figure 2-4c).
13 Amphiboles form when edge-shared octahedra link two of the double-chain [SiO/i]4" plates (see
14 Figure 2-4d). The specific cations between the two double-chain plates define the elemental
15 composition of the mineral, while the ratio of these cations in each location is used to classify
16 amphiboles within a solid-solution series. The cation sites are designated as A, B, and C in
17 Eq. 2-1, which shows the general chemical formula for double-chain inosilicate amphiboles. The
18 Libby Amphibole asbestos is a complex mixture of mineral forms defined by the cation ratios in
19 each site (further discussed in Section 2.2.3).
20
21
22 ^o-i^CsrgO^COH, F, Cl)2 Eq. 2-1
23
24 where:
25
26 A= Na, K
27 B = Na, Li, Ca, Mn, Fe2+, Mg
28 C = Mg, Fe2+, Mn, Al, Fe3+, Ti
29 r=Si, Al.
30
This document is a draft for review purposes only and does not constitute Agency policy.
2-5 DRAFT—DO NOT CITE OR QUOTE
-------
(a) Nesosilicates or single tetrahedron.
The single tetrahedron comprises four oxygen
molecules covalently bound to the silicon, at
the center of the [SiO4]4~-tetrahedron.
(b) Inosilicates [ino (gr.) = thread] -
Single-chain silicates. Chain silicates are
realized by linking [SiO4]4~-tetrahedrons in a
way to form continuous chains. They can be
represented by a composition of [SiO3]2~. A
typical example is diopside CaMg[Si2O6], in
which the "endless" chains are also held
together by Ca2+ and Mg2+ ions.
1
2
3
4
5
6
(c) Inosilicates - Double-chain silicates.
Two silicate chains of the inosilicates are
linked at the corners, forming double-chains
and yielding [Si4On]6~ ions, as realized in
the tremolite-ferro-actinolite series
Ca2(Mg,Fe)5Si8O22(OH)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 [Si2O5]2~. Examples of
sheet silicates include chrysotile
Mg3Si2O5(OH) and vermiculite [(Mg,
Fe,A)3(Al,Si)2O10(OH)2 »4H2O].
A A
A A
Figure 2-4. Structure of the silicate minerals, illustrating silicate subclasses
by the linking of the basic silicon tetrahedron (a) into more complex
structures (b, c, or d).
This document is a draft for review purposes only and does not constitute Agency policy.
2-6 DRAFT—DO NOT CITE OR QUOTE
-------
1 The mineral subgroup within amphiboles is determined by the elemental composition.
2
o
J
4 • Tremolite subgroup (Ca amphiboles)
5 • Anthophyllite subgroup (Fe-Mg-Li orthoamphiboles)
6 • Richterite subgroup (Ca-Na amphiboles)
7 • Cummingtonite (Fe-Mg-Li clinoamphiboles)
8
9
10 A solid solution series includes a continuum of minerals with different cation
11 composition for each site. Solid solution series are defined by their end-members, where mineral
12 terminology can change as the proportion of cations changes within the crystalline structure. For
13 example, a solid solution series for the cation Site A will have one end-member with 100%
14 sodium ions and one end-member with 100% potassium ions. This series would include all
15 intervening ratios. Because each cation site has multiple possibilities, the elemental composition
16 of the amphibole silicates can be quite complex. It is the complexity of the amphiboles that
17 historically has given rise to a proliferation of mineral names with no systematic basis
18 (Hawthorne, 1981). Currently, amphiboles are identified by a clear classification scheme based
19 on crystal chemistry that uses well-established names based on the basic mineralogy, with
20 prefixes and adjective modifiers indicating the presence of substantial substitutions that are not
21 essential constituents of the end-members (Leake et al., 1997). The mineral classification system
22 does not designate certain amphibole mineral as asbestos. However, some mineral designations
23 have traditionally been considered asbestos (e.g., tremolite, anthophyllite.) Other commercial
24 forms of asbestos were known by trade names (i.e., amosite) rather than mineralogical
25 terminology (i.e., an amphibole mineral in the cummingtonite-grunerite solid solution series).
26
27 2.2.1.2. Amphibole Morphology
28 Mineral morphology is a function of the structural form of the silicate and the geologic
29 habit of formation, weathering and other mechanical processes. This discussion will focus on
30 morphology with respect to amphibole minerals.
31 The basic crystal structure of amphibole mineral is formed by the binding of a series of
32 double-chain plates (see Figure 2-5). Where the conditions are suitable, these crystals may form
This document is a draft for review purposes only and does not constitute Agency policy.
2-7 DRAFT—DO NOT CITE OR QUOTE
-------
3
4
5
6
7
- 7 Oxygon
— 4 Silicon
-4 Oxygon 1 Hydroxyl
- 7 Colion
-4 Oxygen 1 Hydroxyl
- 4 Silicon
7 Oxygen
Figure 2-5. Cross-section of amphibole fibers showing the silicon
tetrahedrons (A) 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 and Seidel (2010).
10 as elongated particles. The morphology of the elongated crystal structure is a function of the
11 temperature, pressure, local stress field and solution chemistry conditions during
12 crystallization—habit of formation. Thus, morphology at this level is described in terms of the
13 crystal forms which result from different habits of formation. Individual amphibole structures
14 may be described as acicular, prismatic, or a fibrous. A fiber would be an elongated crystal with
15 parallel sides, where acicular crystals are "needlelike" in appearance and prismatic crystals may
16 have several non parallel faces (e.g., varied, faceted faces). Asbestiform morphology is present
17 where the habit of formation allows crystals to form very long individual fibrils and fibers which
18 may become visible to the naked eye (see Figure 2-6). Thus, the amphibole crystalline structure
19 may result in a range of particle morphologies, including fibers. Where conditions are not
20 conducive to the formation of individual fibers and particles, the amphibole is described as
21 massive—appearing as a solid contiguous sample. Mechanical forces that break amphibole
22 crystals along the cleavage plane create smaller pieces or cleavage fragments. These fragments
This document is a draft for review purposes only and does not constitute Agency policy.
2-8 DRAFT—DO NOT CITE OR QUOTE
-------
(a)
3
4
5
6
7
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
Figure 2-6. Comparison of crystalline forms 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 (2006).
may be elongated, but differ from the crystals described above as at least one face of the
structure is the cleavage plane—not the face of a formed crystal.
With respect to classifying mineral field samples, geologists applied descriptive terms
appropriate for viewing samples simply or at low magnification (e.g., field glass). The geologic
terms for fiber morphology for classification of field samples is based on the macroscopic
appearance of the crystals and fibers (e.g., acicular "needle-like in form") (AGI, 2005). In this
framework, asbestos and asbestiform fibers are defined as long, slender, hair-like fibers visible to
the naked eye (see Figure 2-6). This is a hallmark of commercially mined asbestos which is
sought after for numerous applications because of its high tensile strength, heat resistance and in
some cases, can be woven. Although these terms were used to describe fibers in hand samples
and identify commercially valuable asbestos they are only applicable at the macroscopic level. It
is important to realize that material defined as commercial asbestos, mined, milled, and
manufactured into products not only contained these visible fibers, but many smaller fibers and
single crystals which were not visible to the naked eye (Dement and Harris, 1979). As further
This document is a draft for review purposes only and does not constitute Agency policy.
2-9 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
explained in Section 3, only these smaller fibers can enter the lung and transport to the pleura
where the health effects of asbestos are best characterized. Therefore, for the purposes of this
assessment (i.e., examining the health effects of asbestos fibers), consideration must be given to
how these microscopic fibers are defined. For this purpose, terms intended for describing field
samples may need to set aside, or redefined when applied at the microscopic level.
Currently there are several technologies commonly used to view and identify mineral
structures at high magnification using light microscopes or electron microscopy. As standard
analytical methods were developed for counting mineral fibers, structures and matrices using
these instruments, analytical definitions to describe fibers and structures were developed. Phase
contrast microscopy (PCM) was developed to detect fibers in occupational settings and has been
widely used to assess worker exposure (see Text Box 2-1). The definition of a PCM-fiber is
based purely on its dimensions. The standardization of the PCM method (i.e., NIOSH 7400) and
its importance in applying health standards in occupational settings, results the common usage of
the term 'fiber' to refer to those objects counted in the PCM analytical method (NIOSH, 1994a).
However, this method cannot define the material or morphology of the viewed fiber. Thus
PCM-fibers may be any material, and if they are mineral
fibers may be any fiber morphology. If the nature of the
fiber needs to be defined, NIOSH Method 7402 employs
electron microscopy to determine if the fibers viewed by
PCM are mineral fibers, and can establish the mineral
composition (NIOSH, 1994b). This method does not
recount the fibers, but, rather, it identifies what proportion
of the fibers are mineral fibers, with an elemental
composition consistent with asbestos, which is then used
to adjust the PCM-fiber count. Although the PCM-fiber
definition was not based on either mineralogy or an
understanding of which fibers might be biologically
relevant, this definition has become the basis of existing
health standards (e.g., MSHA. 2008: OSHA. 1994: U.S.
EPA. 1988a).
Text Box 2-1. Fibers Viewed by Light
Microscopy
The collection of fibers on an air filter, and
visually counted under a phase contrast
microscope (PCM), was first described in
1934 by the Dutch physicist Frits Zernike.
The specification of a fiber as >5 um in
length and length-to-diameter ratio (i.e.,
aspect ratio) of at least 3:1 resulted from this
method. As a light microscope technique, the
PCM method cannot distinguish mineral
fibers from other fibers.
The U.S. Public Health Service developed
and tested a standard air sampling method
based on PCM detection (i.e., National
Institute for Occupational Safety and Health
[NIOSH] Method #7400). The NIOSH
method specifies the analyst count fibers
>5 um in length with an aspect ratio of at
least 3:1. Results from PCM analysis are
reported as fibers per cubic centimeter of air
(fibers/cc.)
This document is a draft for review purposes only and does not constitute Agency policy.
2-10 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Electron microscopy can view objects at much higher magnification and can be coupled
with other techniques which can identify the mineralogy (see Text Box 2-2). X-ray diffraction
(XRD) may be used with the above techniques to differentiate crystalline structure of minerals in
solid materials and provides information on the availability of the total mineral present. Thus,
XRD can determine the mineral composition of the material analyzed, identifying its solid
solution series and classifying the mineral per standardized nomenclature for amphibole minerals
(see Section 2.1.1.1).
With the advent of the use of electron microscopy to identify mineral particles, there has
been an attempt to resolve the traditional dimensional fiber defmition(s), by describing the
particles examined by electron microscopy and
X-ray diffraction in terms that are both
geologically and mineralogically relevant.
Structures viewed by electron microscopy may
be described as having parallel sides, and
considered 'fibers'. Where long, thin, curving
fibers are viewed they may be described as
'asbestiform'. Structures with nonparallel sides
can be considered acicular or prismatic,
depending on their proportions. Thus, the
descriptive terms used by geologists have
migrated into the analytical field. However, the
habit of formation of a single structure viewed
by electron microscopy cannot be determined,
and, while descriptive, these terms may not
correlate to the geologic and commercial
definitions of these terms. Therefore, the use of
Text Box 2-2. Minerals Viewed by Electron
Microscopy
Electron microscopy employs electrons—rather
than light—to visualize the specimen. Furthermore,
instead of using glass lenses to focus the light
wavelengths, electromagnetic lenses are used to
focus electrons on the sample. The analytical
techniques included in electron microscopy for
asbestos testing are TEM, scanning electron
microscopy (SEM), and scanning transmission
electron microscopy (STEM). TEM produces
two-dimensional (2-D) images that generally use a
magnification factor of about 500 to 500,000 x.
SEM produces three-dimensional (3-D) images that
generally result in about 10 to 300,000x
magnification. STEM can produce both 2-D and
3-D images that generally result in about 10 to
500,000x magnification.
The ISO 10312 method for analyzing air filters,
enumerates structures much smaller than the PCM
fibers with a minimum length requirement of
0.5 um. Additionally, structures with an aspect ratio
of at least 5:1 are considered fibers, rather than 3:1,
as with PCM analysis. The ISO 10312 method also
defines other structures (fiber bundles, clusters, and
matrices) that are included in the structure count.
Therefore, the term "structure" rather than "fiber" is
used when presenting air sampling results from the
ISO 10312 method where structures per cc of air
(s/cc) are reported.
these definitions to describe individual particles
viewed by TEM can be problematic (Meeker et
al., 2003). Important characteristics such as crystal structure and surface chemistry cannot be
adequately categorized solely with visually determined definitions developed for the
classification of field samples.
This document is a draft for review purposes only and does not constitute Agency policy.
2-11 DRAFT—DO NOT CITE OR QUOTE
-------
1 The definition of 'fiber' and the appropriate application of other morphological terms is
2 an area of ongoing debate. From a public health and regulatory perspective, a PCM-fiber is the
3 fiber of interest (where confirmed as a mineral fiber with an elemental composition consistent
4 with asbestos). There is no requirement for a PCM-fiber to be asbestiform, and, in fact, the
5 method explicitly includes fibers with fairly low aspect ratios (i.e., as low as 3:1). Electron
6 microscopy identified a much broader range of fibers (having much greater resolution) and can
7 provide more specific identification of both mineralogy and the form of the structure.
8
9 2.2.2. Vermiculite
10 Vermiculite is the mineralogical name given to hydrated laminar
11 magnesium-aluminum-ironsilicate, which resembles mica in appearance [see Figure 2-7; (Mg,
12 Fe,A)3(Al,Si)2Oio(OH)2 •4H2O] (AGL 2005). Vermiculite is in the clay mineral group of the
13 phyllosilicates, which also includes kaolinite and montmorillonite. Mica, talc, and serpentine
14 (e.g., chrysotile asbestos) minerals are other well-known sheet silicates. These sheet-like
15 structures are produced by rings of tetrahedrons that are linked to other rings by shared oxygen
16 ions in a two-dimensional plane (see Figure 2-4d). The silicate sheet can extend broadly, and the
17 layered appearance of the mineral reflects this sheet-like structure. The symmetry of these
18 minerals is controlled primarily by the symmetry of the rings, which is usually altered to a lower
19 symmetry by other ions and other layers. Typically, crystals of this subclass are flat, platy, and
20 book-like, as in the mica group, and the sheets are then connected to each other by layers of
21 cations. These cation layers are weakly bonded and often have water molecules and other
22 neutral atoms or molecules trapped between the sheets. When subjected to heat, vermiculite has
23 the unusual property of exfoliating or expanding into "worm-like" pieces. The term vermiculite
24 is derived from the Latin vermiculare, which means to breed worms (The Vermiculite
25 Association, http://www.vermiculite.org). Vermiculite exfoliation occurs at approximately
26 150°C, producing a lightweight and highly absorbent material (AGL 2005). Additional
27 properties of vermiculite are listed in Table 2-1. Vermiculite ore is shown in Figure 2-7.
This document is a draft for review purposes only and does not constitute Agency policy.
2-12 DRAFT—DO NOT CITE OR QUOTE
-------
Table 2-1 Properties of vermiculite
Mineral class/subclass
Chemical formula
Crystal habit of formation
Hardness (Mohs scale)
Cleavage
Specific gravity
Mineral silicates/phyllosilicate
(Mg, Fe,A)3(Al,Si)2010(OH)2 ^I^O
Clay, scaly, aggregate
203
Perfect
2.4-2.7
3
4
9
10
11
12
13
14
15
16
17
18
19
Figure 2-7. Vermiculite ore sample. Brinton's Quarry, near West Chester,
Chester County, Pennsylvania, USA.
Source: Micaceous vermiculite book (http://www.excaliburmineral.com/cdintro.htm)
©Jeff Weissman/Photographic Guide to Mineral Species.
Vermiculite is mined across the world, including the United States (Virginia, South
Carolina, and Montana); South Africa; Uganda; China; Brazil; Russia; India; and Australia
(BGS, 2011). The specific mineralogy and geologic formation habit of vermiculite deposits
vary, and although amphibole minerals are consistent with the ultramafic rock formations
(composed chiefly of ferromagnesian igneous rock) that bear vermiculite, not all vermiculite
deposits contain amphibole asbestos.
This document is a draft for review purposes only and does not constitute Agency policy.
2-13 DRAFT—DO NOT CITE OR QUOTE
-------
1 2.2.3. The Mineralogy of Libby Amphibole Asbestos
2 2.2.3.1. Mineralogy
3 The amphibole mineral fibers within the vermiculite ore and product have historically
4 been reported as a sodium-rich tremolite (Amandus et al., 1987a: McDonald et al., 1986a: Leake,
5 1978; Boettcher, 1966; Larsen, 1942). More recently, various research groups have
6 characterized the more specific mineralogical composition of amphiboles from the Rainy Creek
7 deposit near Libby, MT (Gunter and Sanchez, 2009; Sanchez et al., 2008; Meeker et al., 2003;
8 Wvlie and Verkouteren, 2000: Rossetal.. 1993: Moatamed et al.. 1986).
9 EPA requested that the U.S. Geological Survey (USGS) design and conduct a study to
10 identify the amphibole minerals in the Libby vermiculite mine. Accordingly, USGS personnel
11 collected samples from different areas of the mine in an attempt to identify the range of materials
12 present both geographically, as well as collecting material which represented different habits of
13 formation (Meeker et al., 2003). Figure 2-8 shows data from 30 samples across the mine. The
14 mineral composition of each structure determines its mineral identity (Leake et al., 1997). Here,
15 the U.S. Geological Survey (USGS) used two different techniques to identify the mineral
16 composition of each structure (energy dispersive X-ray analysis [EDS] and electron probe
17 microanalysis [EPMA]). Similar mineral composition was determined by the two methods (see
18 Figure 2-8). Most amphibole structures are classified as winchite (84%), with lesser amounts
19 classified as richterite (11%) and tremolite (6%) (Meeker et al., 2003), based on the current
20 mineralogical nomenclature by Leake (1997). There are also trace amounts of
21 magnesioriebeckite, edenite, and magnesio-arfvedsonite present in Libby Amphibole asbestos
22 (Meeker et al., 2003). All of these minerals are within the mineral solid solution series for
23 tremolite-richterite- magnesioriebecktite. All of the amphiboles found at the mine site, with the
24 possible exception of magnesioriebeckite, can occur in fibrous habit. It was observed these
25 amphibole materials—even when originally present as massive material—can produce abundant,
26 extremely fine fibers by gentle abrasion or crushing (Meeker et al., 2003).
27
This document is a draft for review purposes only and does not constitute Agency policy.
2-14 DRAFT—DO NOT CITE OR QUOTE
-------
Edenite
Richterite
x° O*" ** ' •
:°iv./x£ *
Winchte
0 O.S
Meeker et al fiq. 6
1
Na(B)
H-
EDS EPMA
Mean Error (1a)
Magnesio-
arfvedsonite
Magnesio-
riebeckite
1.5
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Figure 2-8. Mineralogy of Libby Amphibole asbestos 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 Y-axis shows
the number of sites occupied by Na or K. The data shown are a composite of the analysis
fibers taken from 30 different field samples from various locations within the mine.
Notes: EDS is energy dispersive X-ray analysis; EPMA is electron probe microanalysis.
Source: Meeker et al. (2003).
Figure 2-9 shows the compositional variations between the predominate minerals found
in the Libby Amphibole asbestos (winchite, richterite, and tremolite). Although each structure
has as discrete mineral composition, when viewed as a population, fall within solid solution
series shown in Figure 2-8. For example, tremolite is one end-member of the solid solution
series. As calcium decreases and sodium increases, the fibers transition to richterite. Similarly,
as fibers have decreased magnesium and calcium with respect to tremolite, they are defined as
winchite. The sodium content that distinguishes these amphiboles has been redefined over time
in the International Mineralogical Association's mineral classification system, most recently in
1997 (Leake et al., 1997; Leake, 1978). As a result, some amphibole fibers previously defined as
tremolite prior to the new classification system are currently considered winchite based on
chemical composition (Leake et al., 1997).
This document is a draft for review purposes only and does not constitute Agency policy.
2-15 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Figure 2-9. Solution series linking tremolite, winchite, and richterite
amphibole fibers.
Source: Meeker et al. (2003).
The mineral composition of the fibers present is not classifiable to one distinct named
mineral category, but, rather, the composition spans several solid-solution series. However,
there seems to be a consistency in the range of elemental composition found within this material.
Libby Amphibole asbestos is not only made up of the end-members of these solid solution series,
but the spectrum of minerals along the solid solution series shown. Although the majority of
structures analyzed fell within these solid solution series, traces of other minerals were
identified. The term "Libby Amphibole" is used in this document to identify the mixture of
amphibole minerals, of varying elemental compositions (e.g., winchite, richterite, and tremolite),
which have been identified in the rocks and ore of the vermiculite mine near Libby, MT, and are
characteristic of the elongated structures commingled with the vermiculite mined at this location
(Meeker et al., 2003) (i.e., present in the ore vermiculite concentrate and processed materials).
Libby Amphibole Asbestos refers to those elongated structures of the Libby Amphibole mineral
mixture, which have been identified as amphibole fibers or structures, and have been associated
This document is a draft for review purposes only and does not constitute Agency policy.
2-16 DRAFT—DO NOT CITE OR QUOTE
-------
1 with health effects consistent with asbestos exposure (i.e., asbestosis, pleural abnormalities, lung
2 cancer and mesothelioma) (ATSDR. 2008b).
3
4 2.2.3.2. Morphology of the Libby Amphibole Asbestos
5 Mineral samples taken from the mine include veins of asbestiform amphibole and various
6 fiber morphologies in surrounding rock (Meeker et al., 2003). A sample viewed by scanning
7 electron microscope from the Zonolite Mountain mine illustrates the broad range of size and
8 morphologies for the mineral structures (see Figure 2-10). The USGS has described fibers
9 (including asbestiform), acicular and prismatic structures, and curved fibers all within the
10 minerals from the mine (Meeker et al., 2003). As individual fibrils and fiber bundles are viewed
11 under greater magnification under a transmission electron microscope, the range of fiber
12 morphologies can be more clearly seen (see Figure 2-11).
13
14
15
16 Figure 2-10. Scanning electron microscope image of amphibole mineral
17 structures from the Libby, MT mine. An evaluation of the textural
18 characteristics shows the material to include a range of morphologies from
19 prismatic crystals to fibers. Acicular and prismatic crystals, fibers bundles and
20 curved fibers are all present.
21
22 Source: Meeker et al. (2003).
23
This document is a draft for review purposes only and does not constitute Agency policy.
2-17 DRAFT—DO NOT CITE OR QUOTE
-------
I-1 Al_: 1 lit: CUMfHKrriUN Ol AMHUBOI IS I ROM IHi: RAINY
1
2
3
4
5
9
10
11
12
13
14
Figure 2-11. Fiber morphology of amphibole asbestos from the Libby, MT
mine viewed under a transmission electron microscope.
Source: Meeker et al. (2003).
2.2.3.3. Dimensional Characteristics of Libby Amphibole Asbestos
Cumulative particle-size-distribution frequencies (CDF) were developed for Libby ore
Grade 3, and Libby ore Grade 3 expanded by EPA Region 8 using the procedure described in
detail in Appendix C. As shown in Figure 2-12, the particle-size-distribution frequency for the
Libby Grade 3 ore, and the Libby Grade 3 ore expanded were similar to the
particle-size-distribution frequency in the ambient air monitoring samples in Libby, MT. Data
from ambient air monitoring in Libby are presented in Appendix B. The data to construct the
This document is a draft for review purposes only and does not constitute Agency policy.
2-18 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
1.0
0.9
0.8
0.7
0.6
5 0.5
Q 0.5
u
0.4
0,3
0.2
0.1
0.0
Particle Size Distributions of LA Particles • Libby #3 Ore (N = 320),
Libby #3 Ore Expanded (N = 108)
Length
Libby #3 Ore
Libby #3 Ore Expanded
LibbyAir
u
0.4
0.2
0-f
0.0
0.
_— -
31 0.1
Length (urn)
10
100
Width
Libby #3 Ore
Libby #3 Ore Expanded
LibbyAir
Width (urn)
10
10 *
n Q
0.8
n 7
0.6
/
Aspect Ratio
Libby #3 Ore
Libby #3 Ore Expanded
LibbyAir
10 100
Aspect Ratio
1000
Figure 2-12. 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 (2010b) (Provided as Appendix B.)
This document is a draft for review purposes only and does not constitute Agency policy.
2-19 DRAFT—DO NOT CITE OR QUOTE
-------
1 plot in Figure 2-11 are described in Appendices B and C. There are slight shifts towards longer
2 and thicker fibers in the ore samples compared to the air samples, with the aspect ratios being
3 almost identical in the ore and air samples. However, all of these differences are minor, and the
4 majority of these fibers are respirable.
5 Mineralogical characterization of the fibers from the Libby ore Grade 3 and the expanded
6 product using energy dispersive X-ray analysis (EDS) and selected area electron diffraction
7 (SAED) provided further confirmation of the similarity between the fibers from the Libby
8 Grade 3 ore and Libby Amphibole asbestos (methodology described in Section 2.3; see also
9 Appendix B). EDS spectra yielded an elemental fingerprint with sodium and potassium peaks
10 that were highly consistent with values reported for the winchite-richerite solution series
11 described for the Libby, MT ores (Meeker et al., 2003).
12 Based on these data, it is reasonable to conclude that the fibers from the Libby Grade 3
13 ore and expanded ore are similar in physical and mineralogical characteristics to the Libby
14 Amphibole asbestos fibers found in air samples from Libby, MT. The O.M. Scott facility in
15 Marysville, OH used Libby Grade 3 ore from about 1959 to 1980 (Moatamed et al.. 1986:
16 Lockeyetal., 1984). Therefore, the exposure and health effects information from the
17 Marysville, OH facility may be used to derive an RfC that can be applied to the Libby
18 community and other sites that received vermiculite ore from Libby, MT.
19 The Marysville, OH facility also used vermiculate ore from Virginia, South Africa, and
20 South Carolina. The Virginia and South African ores were tested for the presence of fibers as
21 described in Appendix C.5 As described in Appendix E, the Virginia and South African ores
22 released only a small quantity of amphibole fibers. EPA was unable to obtain an ore sample
23 from South Carolina. However, vermiculite ore from the Enoree mine in South Carolina is
24 known to contain amphibole fibers (see Appendix C) (U.S. EPA, 2000b: McDonald et al., 1988).
25
26 2.3. EXPOSURE POTENTIAL
27 Although the occurrence of Libby Amphibole asbestos is limited to a relatively small
28 geographic area, the potential for exposure to it has been greatly enhanced by the historical
29 mining, milling, and distribution of vermiculite operations in Libby, MT. Additionally, material
5 Dr. Lockey, University of Cincinnati, obtained samples of the Virginia and South Africa ores from the Marysville,
OH facility in 1980 and supplied these ores to the EPA for analysis.
This document is a draft for review purposes only and does not constitute Agency policy.
2-20 DRAFT—DO NOT CITE OR QUOTE
-------
1 was sent to processing plants across the nation where plant workers and community contacts may
2 have been exposed. Lastly, consumer products containing vermiculite mined near Libby contain
3 Libby Amphibole asbestos, and consumers may be exposed to Libby Amphibole asbestos while
4 using the products. For example, asbestos-contaminated vermiculite attic insulation from Libby
5 remains in homes today across North America, where there is the potential for residential
6 exposures. This section summarizes the potential for current exposures to the Libby Amphibole
7 asbestos in vermiculite in the Libby community, other communities potentially impacted by
8 processing plants, and from in-place Libby vermiculite attic insulation. Historical exposures for
9 the workers in Libby, MT, and other facilities are discussed in Section 4.1, where data are
10 available.
11 There are also lifestyle, activity, and lifestage factors, which may influence one's
12 exposure potential to asbestos. For example, children may spend more hours outside and engage
13 in activities that impact exposure level compared to adults (U.S. EPA, 2006b: NRC, 1993). In
14 general, children inhale more air per unit body weight (U.S. EPA, 2006b) and spend more time
15 outdoors than adults (Bateson and Schwartz, 2008; NRC, 1993), which could have resulted in
16 increased inhalation exposure to Libby Amphibole asbestos in children compared with adults. In
17 contrast, some adult activity patterns, such as gardening and home repair, may also result in
18 increased exposures where Libby Amphibole asbestos may be present. Thus for the various
19 environments where people may be exposed to Libby Amphibole asbestos, the potential
20 activities and pathways of exposure are discussed below, and where available, exposure
21 measurements are given for various exposure environments and activities.
22
23 2.3.1. Libby Community
24 The Libby community (the towns of Libby, Troy, and surrounding residences) defines the
25 area that may have been directly and indirectly impacted by mining/milling-activities. Many
26 individuals who worked in the mine lived in the surrounding areas. Facilities in the community
27 may have residual contamination from past milling and transport activities. Additionally,
28 expanded vermiculite, waste stoner rock (the waste material from exfoliation), and other
29 materials all potentially containing Libby Amphibole asbestos may have been transported off site
30 to residences and recreational areas. Taken together, there are numerous potential exposure
31 pathways for community residents, both historical and current.
This document is a draft for review purposes only and does not constitute Agency policy.
2-21 DRAFT—DO NOT CITE OR QUOTE
-------
1 During plant operations, individuals may have been exposed to materials inadvertently
2 transported from the workplace to vehicles, homes, and other establishments, typically on the
3 clothing, shoes, and hair of workers. This transport of material may result in "take-home
4 exposure" for the workers, their families, and other coresidents. The magnitude of these
5 exposures was not measured, so the levels to which individuals in the home might have been
6 exposed are not known. Based on studies of other industrial take-home exposures, individuals
7 doing laundry and cleaning house (often women) can be exposed to materials on workers'
8 clothing. Also, children who play on the floor might be more exposed than adults to dust from
9 take-home exposures (Kelly et al., 2006). The community health screening studies from Libby
10 showed that men were more likely to have both occupational and nonoccupational exposures,
11 while women were more likely to have household contact with exposed workers (Peipins et al.,
12 2003; ATSDR, 200Ib). There could also be gender differences in types of activities (e.g.,
13 household chores such as laundry and cleaning) or in intensity or duration of occupational and
14 recreational activities (Peipins et al., 2003).
15 Expanded vermiculite, as a finished product, was used as a soil amender and for attic
16 insulation. Community members may have been exposed and are possibly still exposed to these
17 consumer products. In a survey of Libby residents conducted by ATSDR in 2000-2001, almost
18 52% reported using vermiculite for gardening, 8.8% used vermiculite around the home, and
19 51% reported handling vermiculite attic insulation (Peipins et al., 2003). As vermiculite ore,
20 waste stoner rock, and product were present in the community; numerous activities may have
21 resulted in exposure. Individuals also reported exposures from the following activities:
22 participating in recreational activities along Rainy Creek Road, the road leading to the mine
23 (67%); playing at the ball field near the expansion plant (66%); playing in the vermiculite piles
24 (34%); heating the vermiculite to make it expand/pop (38%); or other activities in which there
25 was contact with vermiculite (31%) (Peipins et al., 2003). Memoranda from Christopher Weis
26 (Weis, 200la) state that asbestos mineral fibers were detected in outdoor sources (yard soil,
27 garden soil, driveway material, and assorted mine-waste materials) and indoor sources (dust and
28 vermiculite insulation) in Libby (Weis, 200la, b).
29 EPA has conducted more recent exposure sampling in the Libby community. Air
30 samples were taken in the community during activities considered appropriate for various
31 potential exposure scenarios. Personal air monitors were placed on the investigator conducting
This document is a draft for review purposes only and does not constitute Agency policy.
2-22 DRAFT—DO NOT CITE OR QUOTE
-------
1 the activity, and a second air sample was taken from a fixed location (area sample). Asbestos
2 fibers were collected on filters and counted by two different laboratory methods: (1) PCM and
3 (2) TEM. Although TEM analysis can count smaller fibers, results are shown here for PCM size
4 fibers used to estimate risk, called PCM equivalent fibers (PCMe).6
5 EPA continues to conduct air monitoring in the Libby community to support clean-up
6 and risk assessment activities. Ambient air monitoring conducted in 2006/2007 at 18 locations
7 across the area indicated that low levels of asbestos fibers are occasionally detected in the air,
8 even with no localized disturbance of asbestos-contaminated material (U.S. EPA, 2009b). Fibers
9 were counted by TEM, and structures7 >0.5 jim in length and with an aspect ratio >3 were
10 included (measured in structures per cc of air, s/cc). Average ambient air levels for the various
11 sampling locations ranged from 8 x 10~6 s/cc to 1.9 x 10~5 s/cc (U.S. EPA. 2009b). Both
12 ambient and activity-based air monitoring have been completed in five community schools (U.S.
13 EPA, 2010c). Outdoor activities conducted that were considered relevant to children's exposures
14 at the schools included playing sports, using playground equipment, and running/walking in
15 outdoor areas. Outdoor activities to assess exposure of the school maintenance workers included
16 digging/raking, power sweeping parking lots, and mowing and edging school lawns.
17 Additionally, ambient air samples were taken in each school (i.e., classrooms, cafeteria,
18 gymnasium, and hallways). Asbestos PCMe fibers were detected by TEM analysis in 5 of
19 63 outdoor activity-based samples, ranging from 0.0022 to 0.039 s/cc. No PCMe fibers were
20 detected in indoor air samples. However, 2 of 50 indoor area samples detected TEM asbestos
21 structures not considered to be PCMe fibers (5.1 x 10 4 s/cc and 5.9 x 10 4 s/cc), which are
22 within the range of analytical sensitivity for the indoor air samples (U.S. EPA, 2010c). It should
23 be noted that indoor air sampling did not include any activity-based sampling to assess student or
24 employee exposures.
25
6 These PCM equivalent fibers (PCMe fibers) are defined as those fibers viewed on TEM that meet the PCM
analytical requirements: >5 um in length and an aspect ratio of at least 3:1. Although the PCM methodology does
not specify a minimum fiber width, current PCM analytical methods reliably detect fibers of 0.25 um in width
(IPCS. 1986). which EPA employs to define PCMe fibers (U.S. EPA. 2008).
7 A single fiber, fiber bundle, cluster, or matrix as defined in the TEM analytical method ISO 10312.
This document is a draft for review purposes only and does not constitute Agency policy.
2-23 DRAFT—DO NOT CITE OR QUOTE
-------
1 2.3.2. Communities near Vermiculite Expansion and Processing Plants
2 Vermiculite from the Libby, MT mine was used commercially from the 1920s to 1990,
3 and a review of company records available from (1964-1990) indicates approximately
4 6,109,000 tons of vermiculite concentrate was shipped to over 200 facilities (ATSDR. 2008b).
5 The 2008 ATSDR (2008b) Summary Report on the 28 Libby vermiculite expansion and
6 processing facilities stated that household residents were exposed by contact with vermiculite
7 from the workers' clothes, shoes, and hair. Workers' personal vehicles likely contained
8 vermiculite dust from the facility emissions and from vermiculite that fell from their clothing and
9 hair on the drive home after work. The O.M. Scott Company (Marysville, OH) reported that
10 company policy was to launder work clothes for their employees and to make showers available
11 for use after work. These procedures, when implemented, should greatly reduce exposure
12 potential via household contact (ATSDR, 2005b). Whether other facilities made these services
13 available or how frequently they might have been used is unknown.
14 Communities near the expansion plants were subjected to some of the same exposure
15 pathways as for the Libby community. The 2008 ATSDR Summary Report observed that
16 individuals in the community could have been exposed through multiple avenues, such as living
17 near the plant and breathing emissions from the facility, disturbing waste-rock piles, having
18 direct contact with waste rock brought home, and living with indoor dust containing asbestos
19 brought in from outdoor sources (ATSDR. 2008b).
20
21 2.3.3. Exposures from Zonolite and Vermiculite for Homeowners, Contractors, and Other
22 Populations
23 Vermiculite was most notably used as attic insulation, as a soil amender for gardening,
24 and in the manufacturing of gypsum wallboard. EPA conducted a study to estimate the potential
25 for exposure to asbestos in homes containing VAI. Air samples were taken to define exposure
26 levels in the homes under various conditions: no activity (e.g., ambient air), as well as during
27 simulated remodeling activities and removal of the VAI (Versar, 2003). Samples were taken in
28 the living space of the homes as well as the attic space.
29 Air samples were collected in five occupied homes where Zonolite VAI was in place
30 (asbestos detected from trace levels to 1.54% by bulk analysis); no fibers were detected in the air
31 samples above 0.0016 PCMe fibers/cc in these homes. However, the air samples were taken
This document is a draft for review purposes only and does not constitute Agency policy.
2-24 DRAFT—DO NOT CITE OR QUOTE
-------
1 when the homes were empty, and there was no disturbance of the VAI or entry/exit into the attic
2 space. Therefore, EPA conducted a number of simulations under controlled conditions to
3 estimate exposures when VAI is disturbed during normal activities (e.g., moving boxes in an
4 attic), remodeling, and removal of the VAI. Structures were built within safe containment to
5 simulate attic space above living space, and VAI was installed in the simulated attics.
6 Remodeling activities resulted in personal exposures ranging from 0.50 to 1.841 fibers/cc PCMe.
7 Stationary samples of the attic air ranged from 0.008 to 0.203 fibers/cc PCMe. For those
8 simulations that included sampling in the 'living space' below the attic, asbestos fibers ranged
9 from 0.001 to 0.25 fibers/cc PCMe during renovations and from 0.001 to 0.035 fibers/cc PCMe
10 in the living space after renovations were complete (Versar, 2003). These data indicate that
11 exposures to asbestos fibers may occur when disturbing Libby Amphibole asbestos-containing
12 VAI in homes.
13 A second study on potential exposures to Zonolite VAI was conducted by an
14 environmental firm hired by attorneys representing individuals with VAI in their homes (Ewing
15 et al., 2010). This study was conducted in three homes containing Zonolite VAI, and air samples
16 were taken, representing ambient conditions (no disturbance of VAI), remodeling, activity in the
17 attic, and removal of the VAI by various methods (see Table 2-2). Disturbance of the
18 asbestos-containing VAI resulted in airborne asbestos levels, both in the personal air monitors
19 and area samples (Ewing etal., 2010).
This document is a draft for review purposes only and does not constitute Agency policy.
2-25 DRAFT—DO NOT CITE OR QUOTE
-------
Table 2-2. Air sampling results for asbestos from Zonolite 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
2.9-12.5d
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
4
5
6
7
8
l8
11
aAir sampling results reported as fibers analyzed by phased contrast microscopy (PCM).
bAir sampling results reported as structures, PCMe as analyzed by transmission electron microscope (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: Ewing et al. (2010).
This document is a draft for review purposes only and does not constitute Agency policy.
2-26 DRAFT—DO NOT CITE OR QUOTE
-------
1 3. FIBER TOXICOKINETICS
2 There are no published data on the toxicokinetics of Libby Amphibole asbestos.8
3 However, to help inform the reader as to the expected toxicokinetics of Libby Amphibole
4 asbestos, this section contains a general summary description of toxicokinetics of fibers. A more
5 detailed discussion of fiber toxicokinetics is beyond the scope of this document and is reviewed
6 elsewhere (NIOSH. 2011: ICRP. 1994).
7 The principal components of fiber toxicokinetics in mammalian systems are
8 (1) deposition at the lung epithelial surface, and (2) clearance from the lung due to physical and
9 biological mechanisms (including both translocation from the lung to other tissues [including the
10 pleura]), and elimination from the body (see Figure 3-1).
11 Libby Amphibole asbestos includes fibers with a range of mineral compositions
12 including amphibole fibers primarily identified as richterite, winchite, and tremolite (see
13 Section 2.2). Although the fiber size varies somewhat from sample to sample, a large percentage
14 (-45%) is less than 5 um long in bulk samples examined from the Libby mine site (Meeker et al.,
15 2003). Limited data from air samples taken in the workplace also document a large percentage
16 of fibers (including both respirable9 fibers as well as fibers <5 jim-long) (see Section 4.1.1.2 and
17 Table 4-3). The importance of the size of fibers and how they deposit following inhalation is
18 described below. Due to a lack of data specific to Libby Amphibole asbestos, these deposition
19 steps are discussed for general forms of asbestos. The main route of human exposure to mineral
20 fibers is through inhalation, although other routes of exposure play a role. Exposure of
21 pulmonary tissue to fibers via the inhalation route depends on the fiber concentration in the
22 breathing zone, the physical (aerodynamic) characteristics of the fibers, and the anatomy and
23 physiology of the respiratory tract. Ingestion is another pathway of human exposure and occurs
24 mainly through the swallowing of material removed from the lungs via mucociliary clearance or
25 drinking water contaminated with asbestos, or eating, drinking, or smoking in
26 asbestos-contaminated work environments (Condie, 1983). Handling asbestos can result in
27
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.
9 Respirable fibers are those that can be inhaled into the lower lung where gas exchange occurs and are defined by
their aerodynamic diameter (dz < 3 um; NIOSH) (2011).
This document is a draft for review purposes only and does not constitute Agency policy.
3 -1 DRAFT—DO NOT CITE OR QUOTE
-------
Dust in inspired and expired air
1
2
3
4
5
6
7
10
11
12
13
14
15
16
17
18
19
20
21
22
blood
nasopnaryngeal compartment
1 t
tracheo-bronchial compartment
1 t
3l\/gQlgi- compartmsnt
parenchyma! tissue
sputum
gastrointestinal
tract
feces
Figure 3-1. General scheme for fiber deposition, clearance, and translocation
of fibers from the lung and GI tract. General scheme for fiber deposition
(heavy arrows), clearance (light dotted arrows), and translocation (light arrows).
Diagram of Bignon et al. (1978) derived from International Commission on
Radiological Protection (ICRP) lung model by the Task Group on Lung
Dynamics (1966), as cited in ICRP (1994).
Source: ICRP (1994).
heavy dermal contact and exposure. Asbestos fibers could 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 of 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 than external dose, and it is determined both by fiber characteristics of the
exposure environment and the exposed population. Dose to the lung is a function of airway
anatomy, lung volume, ventilation rate, and clearance from the lung, as well as the fiber's
physical and chemical characteristics (U.S. EPA, 2004; Oberdorster, 1991). Many studies have
This document is a draft for review purposes only and does not constitute Agency policy.
3 -2 DRAFT—DO NOT CITE OR QUOTE
-------
1 examined the role of these physical and chemical characteristics in asbestos-induced disease in
2 the lung and are reviewed in more depth elsewhere (NIOSH, 2011; AT SDR, 200 la: Myojo and
3 Takava. 2001: Witschi and Last 1996: Lippmann. 1990: Merchant 1990: Yuetal.. 1986: Griffis
4 et al., 1983: Harris and Fraser, 1976: Harris and Timbrell, 1975). Factors influencing dose to
5 other tissues in the body (e.g., pleura, peritoneum, stomach, and ovaries) are not as well known,
6 but they are discussed below where data are available.
7
8 3.1. DEPOSITION OF FIBERS IN THE RESPIRATORY TRACT
9 The deposition of fibers in the respiratory tract is dependent on the aerodynamic
10 properties of the fiber (length, width, and density) and the anatomy and physiology of the
11 respiratory tract (NIOSH. 2011: ATSDR, 2004. 200la: Myoio and Takava. 2001: Witschi and
12 Last. 1996: Yuetal.. 1986: Griffis et al.. 1983: Harris and Fraser. 1976: Harris and Timbrell.
13 1975). The aerodynamic diameter of fibers is mostly determined by the geometric diameter and
14 density. In general, thicker fibers are deposited in the upper airways; thinner fibers are carried
15 deeper into the airways and alveolar regions. Fibers with aerodynamic diameters less than
16 approximately 3 jim meet the physical criteria necessary for deposition in the terminal
17 bronchioles and beyond to the alveoli. The site of fiber deposition within the respiratory tract
18 has implications related to lung retention and surface dose of fibers.
19 The respiratory tract encompasses the extrathoracic region (nasal passages, pharynx, and
20 larynx), thoracic region (the conducting airways [trachea bronchi, bronchioles]), and the
21 gas-exchange region of the lung (respiratory bronchioles, alveolar ducts, and alveoli). A full
22 review of the anatomy and architecture of the respiratory tract is beyond the scope of this
23 document but has been reviewed by ICRP (ICRP, 1994).
24 Fiber deposition occurs by five mechanisms: impaction, interception, sedimentation,
25 diffusion, and electrostatic precipitation (see Table 3-1):
26
27
28 1. Impaction: The momentum of the fiber causes it to directly impact the airway
29 surface as the airflow changes direction. This is the predominant method of
30 deposition in the nasopharyngeal region where airflow is swift and larger
31 fibers/particles are present.
This document is a draft for review purposes only and does not constitute Agency policy.
3 -3 DRAFT—DO NOT CITE OR QUOTE
-------
Table 3-1. Factors influencing fiber deposition and clearance in the respiratory system
Size of fiber
(aerodynami
c diameter)
5-30 urn
1-5 um
2 um or less
Area of
deposition in
respiratory
system
Extrathoracic
Region
(nasopharyngeal
region nasal
passages, pharynx,
larynx)
Thoracic Region
(trachea, bronchial
and bronchiolar
region)
O /
Gas-Exchange
Region
(respiratory
bronchioles,
alveolar ducts,
alveoli)
Predominant
method of
deposition
Impaction
Sedimentation
, impaction,
interception
r^
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
I
^§
I
a.
o
c
o
Source: Adapted from Witschi and Last (2001) in Casarett and Doull's Toxicology: The Basic Science of Poisons, 6th edition, p. 515.
-------
1 2. Interception: A special case of impaction where the edge of the fiber touches the
2 airway surface and is prevented from continuing along the airway. This
3 mechanism is important in the conducting airways (trachea and bronchi), where
4 the airflow is slower and laminar flow along the airway surface is conducive to
5 interception.
6 3. Sedimentation: Gravitational forces and air resistance cause fibers/particles to
7 settle out of the air column onto the airway surface. For sedimentation to occur,
8 air flow velocities must be low to allow the particle/fiber to settle, and this is a
9 predominant mechanism to the smaller conducting airways.
10 4. Diffusion: This method of deposition is predominant in the alveolar region where
11 air movement is negligible. Diffusion occurs from interactions of the fibers with
12 the movement of air molecules; this Brownian motion increases with decreasing
13 fiber size (<0.5-um diameter).
14 5. Electrostatic Precipitation: A special case of diffusion in which fiber motion
15 towards the airway surface is a function of static charge between the fiber and
16 airway surface. As with classic diffusion, this primarily occurs in the
17 gas-exchange region where airflow is negligible and electrostatic forces can
18 predominate.
19
20
21 Aerodynamic diameter (also called aerodynamic equivalent diameter) of fibers accounts
22 for the dimensional properties that influence the movement of the fiber's center of gravity
23 through the airways, so aerodynamic diameter is important in all depositional mechanisms. The
24 aerodynamic diameter is the diameter of a unit density (1 g/cm3) sphere that has the same
25 gravitational settling velocity as the particle of interest. Since the aerodynamic diameter informs
26 the deposition patterns of fibers, it is used in dosimetric modeling to determine the expected fiber
27 deposition in the respiratory tract. Impaction and interception, however, are also heavily
28 influenced by fiber length. Where the physical length of the fiber greatly exceeds the
29 aerodynamic diameter, impaction and interception can be underpredicted by modeling the center
30 of gravity of the fiber. Sedimentation is related to the mass of the fiber, as well as the
31 aerodynamic diameter, but generally occurs at lower velocities in smaller airways. Diffusion
32 occurs from interactions of the fibers with the movement of air molecules; this Brownian motion
33 increases with decreasing fiber size (<0.5-um diameter). Electrostatic precipitation occurs when
34 fiber charges induce opposite charges on the airway surfaces and the fiber is drawn to the airway
35 walls (Lippmann, 1990).
This document is a draft for review purposes only and does not constitute Agency policy.
3-5 DRAFT—DO NOT CITE OR QUOTE
-------
1 For high aspect ratio fibers, like asbestos, the shape factor often approaches one and the
2 equation reduces to the aerodynamic diameter that is approximately equal to the nominal fiber
3 diameter.10 Therefore, in employing the information from Table 3-1 to high aspect ratio fibers,
4 one may get an idea of the deposit!onal characteristic of fibers from the nominal diameter. By
5 definition, fibers have a greater aspect ratio than particles and as discussed, high aspect ratio
6 fibers may act significantly different than other particles with respect to some mechanisms of
7 deposition (e.g., impaction, interception, and electrostatic precipitation). Therefore, the
8 depositonal characteristics of fibers are not characterized completely by aerodynamic diameter.
9 No equivalent depositional model, however, is yet available for fibers in the dimensional range
10 of asbestos that takes into consideration the increased sedimentation and impaction for high
11 aspect ratio particles.
12 Fibers enter the respiratory tract along with airflow through the nasal and oral passages.
13 The nasal passage, from the nostril to the pharynx, serves as a filter for some fibers with
14 diameters 5-30 um. Clumps of fibers also could deposit in these regions. Many animal species,
15 including rats and mice, are obligate nose breathers, meaning that fibers pass only through the
16 nasal passages, and, therefore, are always subject to nasopharyngeal filtering. Humans,
17 monkeys, and dogs, among other species, breathe both orally and nasally (oronasal). Therefore,
18 larger fibers and clumps of fibers can bypass the upper respiratory tract filtering and be inhaled
19 directly into the larynx/trachea, especially during exertion (e.g., exercise or work), which may
20 further alter deposition by increased turbulence in the airways. This distinction is important
21 when comparing results of inhalation studies conducted in different species.
22 The conducting airways beyond the nasopharyngeal region include the trachea and
23 bronchi, which serially bifurcate into airways of decreasing internal diameters. The aerodynamic
24 diameter of fibers that can deposit in the tracheobronchial region is in the range of 1-5 um.
25 Fibers with aerodynamic diameter <1 um can deposit in the bronchioles and the alveoli (ICRP,
26 1994).
27 Generally, fibers with aerodynamic characteristics conducive to deposition in the
28 bronchioles and alveoli can cause pulmonary fibrosis and associated disease by either retention
29 in the alveoli or penetration into the peribronchiolar space. All fibers having an aerodynamic
10The physical properties of a fiber that determine its aerodynamic transport are combined and defined as the
aerodynamic diameter; one such property is the shape factor (ICRP. 1994).
This document is a draft for review purposes only and does not constitute Agency policy.
3 -6 DRAFT—DO NOT CITE OR QUOTE
-------
1 diameter that is less than approximately 2 jim, which includes Libby Amphibole asbestos, meet
2 the physical criteria necessary for deposition in the deeper regions of the respiratory tract at the
3 level of the terminal bronchioles or alveoli.
4 Deposition of fibers in the alveolar region of the lung is consistent with radiological
5 findings in humans of fibrosis in the lower lung fields at early stages of disease. Deposition of
6 fibers in the alveoli can become limited when fiber length approaches 40 jim (Morgan et al.,
7 1978). Alveolar deposition of fibers with high aspect ratios and length ranging from less than
8 1 |im to greater than 200 jim long, however, has been recorded (Morgan et al., 1978). In all
9 documented observations of fibers collected from either healthy or diseased individuals, short
10 fibers (<5 jim) were present in substantially greater numbers in lung tissue than were long fibers
11 (>5 jim) (Churg, 1982). Although information is limited on how fibers get to the pleura, fibers
12 observed in pleural tissue from mesothelioma cases are more likely to be short (<5 jim) (Suzuki
13 et al., 2005). These observations could be due in part to the increased deposition of smaller
14 fibers or the breakage of larger fibers over time (Bernstein et al., 1994; Davis, 1994).
15 The lung and nasal depositional differences are due in part to differences in airway
16 structure and breathing patterns across lifestages (i.e., children, adults), changing the depositional
17 pattern of different fiber sizes, possibly altering the site of action, and potentially resulting in
18 differential clearance and health effects (see Section 4.7).
19 Modeling of fiber deposition has been examined for various fiber types (e.g., refractory
20 ceramic fibers, chrysotile asbestos) (Sturm, 2009; Zhou et al., 2007; Lentz etal., 2003; Dai and
21 Yu. 1998: Yu et al.. 1997: Coin et al.. 1992). but not for Libbv Amphibole asbestos. In general,
22 the pattern of deposition for fibers is expected to have some similarities to the well-studied
23 deposition pattern for essentially spherical particles (reviewed in ICRP) (1994). For example,
24 the multipath particle dose model (Brown et al., 2005; Jarabek et al., 2005) uses information on
25 the physical properties of the particles (length and width [also called bivariate distribution] and
26 density), the anatomy and architectural features of the airways, airflow patterns that influence the
27 amount and the location of the deposition of the particles, and dissolution and clearance
28 mechanisms that are operative to estimate the retained dose in the target tissue.
29
This document is a draft for review purposes only and does not constitute Agency policy.
3 -7 DRAFT—DO NOT CITE OR QUOTE
-------
1 3.2. CLEARANCE
2 3.2.1. Inhalation
3 3.2.1.1. Respiratory Tract
4 Once fibers deposit on the surface of the respiratory tract, they may be removed (cleared)
5 from the lungs in several ways—including physical clearance, dissolution, phagocytosis, or
6 encapsulation. Some of these mechanisms, such as dissolution of the fibers or removal via the
7 mucociliary apparatus, can result in the fibers being cleared from the body (see Figure 3-1).
8 Other clearance mechanisms may remove fibers from the surface of the respiratory tract but
9 result in transport of the fibers to other tissues by translocation. Translocation of fibers from the
10 terminal bronchioles and alveoli into the peribronchiolar space, lymph nodes, and pleura has
11 been implicated in disease causation (e.g., pleural plaques, mesothelioma) (Dodson et al., 2001).
12 In human studies, the translocation of asbestos fibers following inhalation has been observed to
13 varying degrees throughout the pulmonary and extrapulmonary tissues of the respiratory system
14 (Dodson et al., 2005; Dodson etal., 2001; Kohyama and Suzuki, 1991; Suzuki and Kohyama,
15 1991; Armstrong et al., 1988), as well as other organs, including the brain, kidney, liver
16 (Miserocchi et al., 2008), and ovaries (Langseth et al., 2007). In many cases, the type of fiber
17 was not defined, and the individual exposure information is not available. Fibers that are not
18 cleared may remain at the epithelial surface or enter the parenchymal tissue of the lung.
19 Berry (1999) provided a review of the animal toxicity literature specifically for fiber
20 clearance. There are limited data on clearance patterns based on autopsy studies in humans.
21 Two studies estimated clearance half-life for amphibole asbestos (-20 years) as compared with
22 chrysotile asbestos (-10 years) (Finkelstein and Dufresne, 1999; Churg and Vedal, 1994): in
23 evaluating the data on lung fiber burden, Berry et al. (2009) estimated the range of the half-life
24 for crocidolite to be between 5 and 10 years. Generally, studies have focused on determining the
25 size and type of asbestos retained in specific tissues (Suzuki et al., 2005; McDonald et al., 2001;
26 Suzuki and Yuen, 2001: Dumortier et al., 1998: Gibbsetal., 1991: Dodson etal., 1990) and did
27 not discuss changes in fiber content since exposure. Sebastien et al. (1980) concluded that lung
28 fiber burden could not be used as an accurate reflection of pleural fiber burden.
29
This document is a draft for review purposes only and does not constitute Agency policy.
3-8 DRAFT—DO NOT CITE OR QUOTE
-------
1 3.2.1.1.1. Physical clearance of fibers
1 Fibers deposited in the nasal passages can be removed by physical clearance. When
3 breathing occurs through the nose, many fibers are filtered by the turbulent airflow in the nasal
4 passages, impacting against the hairs and nasal turbinates, as well as becoming entrained in
5 mucus in the upper respiratory tract where they can be subsequently removed by mucociliary
6 action or reflexive actions such as coughing or sneezing. The mucociliary escalator removes
7 fibers through ciliary movement of the sticky mucus lining (Wanner et al., 1996; Churg et al.,
8 1989). Fibers removed from the conducting airways through this mechanism are coughed out or
9 swallowed and enter the digestive tract where they may adversely affect the gastrointestinal
10 tissue, enter the blood stream, or be excreted. Clearance of fibers via mucociliary action is rapid
11 and is usually complete within minutes or hours. However, the mucociliary escalator extends
12 only down to the level of the terminal bronchioles and not to the alveoli. Therefore, particles
13 that reach the alveolar region of the lung cannot be cleared through this process. Fibers can also
14 translocate due to physical forces associated with respiration (Davis, 1989).
15 Some fibers are not cleared from the lung, leading to an accumulation with time (Case et
16 al., 2000; Finkelstein and Dufresne, 1999; Jones et al., 1988). The fibers that remain in the lung
17 may undergo a number of processes including translocation, dissolution, fragmentation, splitting
18 along the longitudinal axis, or encapsulation with protein and iron. Available data indicate
19 prolonged clearance from the lung of long (>5 jim) or short amphibole fibers (Coin et al., 1994;
20 Tossavainen et al., 1994). The prolonged clearance times for long amphibole fibers have led
21 some investigators to conclude that long versus short amphibole fibers are predominant in the
22 cause of disease despite the relatively small numbers of these longer fibers in comparison to
23 short fibers (Mossman et al.. 2011: ATSDR, 2003b). However, others argue that fibers of all
24 lengths induce pathological responses and urge caution in excluding, based on their length, any
25 population of fibers from consideration as possibly contributing to the disease process (Aust et
26 al.. 2011: Dodson et al.. 2003). Respirable-sized fibers of Libby Amphibole asbestos have been
27 identified in air samples from activity-based sampling from Libby, MT, and in airborne fibers
28 suspended from both Libby vermiculite concentrate and in the exfoliated product from that
29 concentrate. Based on fibers counted by the TEM analytical method (ISO 10312), the majority
30 of counted fibers are respirable (see Figure 2-12).
31
This document is a draft for review purposes only and does not constitute Agency policy.
3 -9 DRAFT—DO NOT CITE OR QUOTE
-------
1 3.2.1.1.2. Dissolution of fibers
1 Dissolution, or the chemical breakdown of fibers, is another method of removal of fibers
3 from the lung. This process varies, depending on the chemical composition of the fibers, as well
4 as the physiological environment. Dissolution can occur in the lung's extracellular fluids or in
5 the macrophage phagolysosome. Studies performed in vitro to determine dissolution rate of
6 fibers attempt to mimic the extracellular lung fluids and macrophage-phagolysosome system to
7 understand the length of time that fibers remain in the system (Rendall and Du Toit, 1994).
8 Studies have shown that dissolution occurs more rapidly for chrysotile fibers than for amphiboles
9 (Coffin etal., 1983). Fibers can also be physically diminished through splitting or breakage.
10 These smaller fragments are then more easily removed by phagocytosis or translocation.
11
12 3.2.1.1.3. Removal of fibers through phagocytosis
13 The principal clearance pathway for insoluble fibers deposited in the alveoli is through
14 phagocytosis by macrophages. Alveolar macrophages that have phagocytized insoluble fibers
15 migrate to the bronchoalveolar junctions where they enter onto the mucociliary escalator for
16 removal (Green, 1973). Alternatively, alveolar macrophages that have phagocytized insoluble
17 fibers can also migrate through the epithelial wall into the interstitial space and enter the
18 lymphatics (Green, 1973).
19 Alveolar macrophage cells engulf and transport deposited particles to the mucociliary
20 escalator or through the alveolar epithelium to the interstitial tissues, where they are removed or
21 translocated by the blood or lymphatics. Durable fiber impaction in these deeper regions also
22 stimulates activation of alveolar macrophage cells. In vitro and in vivo studies clearly indicate
23 that macrophage cells play a role in the translocation of fibers (Dodson et al., 2000a: Castranova
24 etal., 1996; Brody etal., 1981; Bignon etal., 1979). These studies have demonstrated the
25 presence of asbestos fibers in cell cytoplasm where they can be transported in association with
26 cytoskeletal elements to the proximity of the cell nucleus. Small chrysotile fibers can also
27 penetrate the nuclear membrane (Malorni etal., 1990).
28 A number of processes can disrupt the normal phagocytic function of the alveolar
29 macrophages. These processes include death or dysfunction of macrophages due to phagocytosis
30 of an excessive number of particles (often termed "overload") or highly reactive particles or an
31 attempted phagocytosis of fibers of lengths that exceed the dimensional capacity of the
This document is a draft for review purposes only and does not constitute Agency policy.
3-10 DRAFT—DO NOT CITE OR QUOTE
-------
1 macrophage (often termed "frustrated phagocytosis") (NIOSH, 2011). All of these processes can
2 induce inflammatory and fibrogenic responses. Limited inhalation laboratory animal studies
3 exist at nonoverloading concentrations of fibers or particles; therefore, there is insufficient
4 information to determine mechanisms at these lower doses reviewed in Mossman et al. (2011).
5
6 3.2.1.1.4. Encapsulation of fibers
7 Fibers that are too large to be easily engulfed by the alveolar macrophage can stimulate
8 the formation of "asbestos bodies." Asbestos bodies are fibers that, during prolonged residence
9 in the lung, have become coated with proteins, iron and calcium oxalate. Due to their iron
10 content, histological stains for iron have long been used to identify them in tissue; thus, they are
11 sometimes called "ferruginous bodies." The mechanisms that result in the formation of asbestos
12 bodies are poorly understood, although most appear to be formed around amosite fibers (Dodson
13 et al., 1996). The iron in the coating, however, is derived from the asbestos fiber, cells, or
14 medium surrounding the fiber and can remain highly reactive (Lund et al., 1994; Ohio et al.,
15 1992). Asbestos bodies can remain in the lung throughout the lifetime of the exposed individual.
16 Asbestos bodies comprise a minor portion of the overall fiber burden of the lung, and, after the
17 fiber is fully coated, these fibers might or might not participate directly in asbestos disease. The
18 presence of iron in the coating, however, could provide a source for catalysis of reactive oxygen
19 species similar to that observed with fibers.
20
21 3.2.1.1.5. Translocation to extrapulmonary tissues
22 Clearance from one tissue may involve translocation to another tissue. For example,
23 following fiber deposition in the respiratory tract, fibers may then clear via translocation to
24 extrapulmonary tissues like the pleura. The specific mechanism and translocation route depend
25 both on fiber characteristics and the tissue of deposition. Whether or not fibers are translocated
26 appears to depend on their physical-chemical characteristics, including two-dimensional size
27 (length and width); durability; solubility; and reactivity. This translocation is aided by high
28 durability and an inflammation-induced increase in permeability but is hindered by fibrosis.
29 Deposition occurs in the respiratory tract as described above; translocation from the respiratory
30 tract may, in turn, lead to fibers 'depositing' in extrapulmonary sites.
This document is a draft for review purposes only and does not constitute Agency policy.
3-11 DRAFT—DO NOT CITE OR QUOTE
-------
1 Apparent translocation of fibers throughout the respiratory tract is evident from
2 experimental animal research done by several investigators following exposure by both
3 intrapleural injection and inhalation (Miserocchi et al., 2008; Holt, 1982; Smith et al., 1980;
4 Bignon et al.. 1979: Smith et al.. 1979: Smith and Hubert. 1974). The data from most studies
5 show that fibers can—and do—translocate among tissues and organs and move by both
6 physiological and physical mechanisms (Holt 1983; Holt, 1982; Cook and Olson, 1979).
7 Conflicting results from another study, however, indicate no evidence of fiber translocation from
8 the central to peripheral compartments following inhalation exposure in rats, although this could
9 be due to the short duration of the study (29 days postexposure) (Coin etal., 1992).
10 Translocation of fibers to extrapulmonary tissues has been studied in multiple studies;
11 however, the mechanism is still unknown. This was more recently reviewed by Miserocchi et al.
12 (2008). Fibers have been measured in extrapulmonary tissues including pleural plaques and
13 mesothelial tissue (i.e., pleural or peritoneal) in miners, brake workers, insulation workers, and
14 shipyard workers (Roggli et al., 2002; Dodson et al., 2000b: Churg, 1994; Kohyama and Suzuki,
15 1991). These studies found fibers at all locations analyzed, with increased levels of amphibole
16 as compared to chrysotile in the parenchyma when subjects were exposed to a mixture of both
17 fiber types. Amphibole fibers, however, were less prevalent in the pleura and mesothelial tissues
18 (Kohvama and Suzuki. 1991: Sebastien et al.. 1989: Armstrong et al.. 1988: Churg. 1988:
19 Bignon etal., 1979). Few studies have examined the size distribution of fibers translocated to
20 specific tissues. For example, one early study suggested that the longer amphibole fibers
21 predominate in the lung while shorter chrysotile fibers are found in the pleura (Sebastien et al.,
22 1980): others showed that the fiber-length distribution was the same by fiber type regardless of
23 location (Kohvama and Suzuki, 1991: Bignon etal., 1979).
24 Transplacental transfer of both asbestos (chrysotile, tremolite, actinolite, and
25 anthophyllite) and nonasbestos fibers has been shown to occur in humans, as measured in the
26 placenta and in the lungs of stillborn infants (Hague et al., 1998: Hague etal., 1996: Hague et al.,
27 1992: Hague and Kanz, 1988). It is hypothesized that maternal health might influence the
28 translocation of fibers, as some of the mothers had preexisting health conditions (e.g.,
29 hypertension, diabetes, or asthma) (Hague et al., 1992). This group also measured transplacental
30 translocation in a mouse study and observed early translocation of crocidolite fibers through the
31 placenta in animals exposed via tail-vein injection (Hague et al., 1998). These studies did not
This document is a draft for review purposes only and does not constitute Agency policy.
3-12 DRAFT—DO NOT CITE OR QUOTE
-------
1 evaluate the source or levels of exposure, only the presence of fibers in the body during early
2 lifestages in mice and humans.
3 Sebastien et al. (1980) found chrysotile was the predominant fiber in parietal pleura of
4 autopsy cases, while the amphibole fibers found in the lungs ranged from 0 to 100% (mean
5 56%). Bignon et al. (1979) found similar distributions but also found increased amphibole fibers
6 in the associated lymph nodes. In this study, chrysotile and amphibole fibers were found
7 together in the lung parenchyma and alveolar spaces. Other studies show fewer amphibole fibers
8 at the site of diseased tissue in the pleura and mesothelial tissue than chrysotile (Kohyama and
9 Suzuki, 1991; Churg, 1988). Sebastien et al. (1989) examined fiber types in lungs of chrysotile
10 textile and mining workers from South Carolina and Quebec, respectively, to better understand
11 the unknown reason for differences in disease risk in each cohort. Both groups were exposed to
12 similar material, yet the South Carolina cohort had a much greater risk of respiratory cancer.
13 This study examined only lungs, although some of those exposed had nonpulmonary cancers.
14 Overall, the number of tremolite fibers retained in the lungs was higher than that of chrysotile
15 fibers retained in the lungs in both cohorts. Size distribution showed that most fibers measured
16 were 5.8-8.0 jim long, although measurements were not made for anything smaller than this.
17 Tremolite fibers had a greater mean diameter in both cohorts (0.35 jam) as compared to
18 chrysotile (0.10 jim), while chrysotile had more "Stanton" fibers (25.2-31.8%) as compared to
19 tremolite (5.9-6.3%). Stanton fibers are defined as >8 jim long and <0.25 jim in diameter
20 (Stanton et al. (1981)1, reviewed in Appendix D).
21
22 3.2.1.2. Pleura! Cavity and Extrapulmonary Sites
23 Studies have demonstrated fiber clearance from the respiratory tract may lead to
24 translocation to the pleural cavity and extrapulmonary sites. For example, in a study comparing
25 fiber burden in the lung, thoracic lymph nodes, and pleural plaques, Dodson et al. (1990)
26 observed that the average-length fiber found in the lung (regardless of type) was longer than
27 those found in the lymph nodes or plaques. Most fibers at all three sites were short (<5 jim). A
28 later study by this group (i.e., Dodson et al., 2000b) examined tissue from 20 individuals with
29 mesotheliomas, most with known asbestos exposures. Seventeen of the cases (85%) had
30 asbestos fibers in at least one other extrapulmonary site. The most prevalent type of asbestos in
31 the mesentery was amosite, and the second most prevalent was chrysotile. Tremolite was also
This document is a draft for review purposes only and does not constitute Agency policy.
3-13 DRAFT—DO NOT CITE OR QUOTE
-------
1 found, to some degree, in the mesentery and omentum, and in the lung. Dodson et al. (2005)
2 examined parenchymal lung tissue from a cohort of 54 mesothelioma patients and determined
3 the presence of asbestos in all patients analyzed. However, very little information is known
4 about the specific mechanisms of fiber clearance and/or translocation from the pleural cavity and
5 extrapulmonary sites, although many studies examining these tissues have observed fibers in
6 multiple tissue sites [reviewed in Aust et al. (2011), Case et al. (2011)]. Following intrapleural
7 injection of fibers in rats, Bignon et al. (1979) used transmission electron microscopic evaluation
8 following serial sacrifice to monitor migration of fibers from the pleural cavity to the lung
9 parenchyma.
10
11 3.2.2. Ingestion
12 Although ingestion is a potential route of exposure, limited research has examined
13 clearance (e.g., translocation) of fibers following ingestion, and no clearance studies are
14 available specific to Libby Amphibole asbestos. An early study to examine the tissue response
15 to asbestos fibers is not truly representative of a natural ingestion exposure, as the researchers
16 directly injected a suspension of amosite fibers into the duodenal wall (Meek and Grasso, 1983).
17 This study, however, also examined oral ingestion of amosite in healthy animals and those with
18 gastrointestinal ulcers to determine if translocation of fibers occurs through ulcers. Following
19 injection of amosite, granulomatous lesions were observed. Ingestion of the same material
20 resulted in no such lesions or in any other histopathological changes in either healthy or
21 compromised rats. Thus, no translocation was observed from either the healthy or the
22 compromised rat gastrointestinal tracts in this study. A later International Agency for Research
23 on Cancer study (Truhaut and Chouroulinkov, 1989) examined the effects of chrysotile and
24 crocidolite ingestion in Wistar rats. No translocation was observed. No further studies have
25 been found on clearance or translocation of fibers from the gastrointestinal tract.
26
27 3.2.3. Dermal Contact
28 No studies of dermal clearance or translocation have been reported in the published
29 literature.
30
This document is a draft for review purposes only and does not constitute Agency policy.
3-14 DRAFT—DO NOT CITE OR QUOTE
-------
1 3.3. SUMMARY
2 Although oral and dermal exposure to fibers does occur, inhalation is considered the main
3 route of human exposure to mineral fibers, and, therefore, it has been the focus of more fiber
4 toxicokinetic analyses. Exposure to Libby Amphibole asbestos is presumed to be through all
5 three routes of exposure; this assessment specifically focuses on the inhalation pathway of
6 exposure. Generally, fiber deposition in the respiratory tract is fairly well defined based on fiber
7 dimensions and density, although the same cannot be said for fiber translocation to
8 extrapulmonary sites (e.g., pleura). The deposition location within the pulmonary and
9 extrapulmonary tissues plays a role in the clearance of the fibers from the organism.
10 Fiber clearance from the respiratory tract can occur through physical and biological
11 mechanisms. Limited mechanistic information is available on fiber clearance mechanisms in
12 general, and no information specific to clearance of Libby Amphibole asbestos fibers is
13 available. Fibers have been observed in various pulmonary and extrapulmonary tissues
14 following exposure, suggesting translocation occurs to a variety of tissues. Studies have also
15 demonstrated fibers may be cleared through physical mechanisms (coughing, sneezing) or
16 through dissolution of fibers.
17 Multiple fiber characteristics (e.g., dimensions, density, and durability) play a role in the
18 toxicokinetics of fibers. For this reason, careful attention has been paid to these fiber
19 characteristics when analyzing research studies on Libby Amphibole asbestos and asbestiform
20 tremolite, an amphibole fiber that comprises part of Libby Amphibole asbestos (see
21 Appendix D). No toxicokinetic data are available specific to Libby Amphibole asbestos,
22 tremolite, richterite, or winchite. When available, this information is presented in the discussion
23 of each study in relation to the toxic endpoints described.
24
This document is a draft for review purposes only and does not constitute Agency policy.
3-15 DRAFT—DO NOT CITE OR QUOTE
-------
1 4. HAZARD IDENTIFICATION OF LIBBY AMPHIBOLE ASBESTOS
2 Several human studies are available that provide evidence for the hazard identification of
3 Libby Amphibole asbestos.11 This discussion focuses primarily on data derived from studies of
4 people exposed to Libby Amphibole asbestos—either at work or in the community. The adverse
5 health effects in humans are supported by the available Libby Amphibole asbestos experimental
6 animal and laboratory studies. Libby Amphibole asbestos contains winchite (84%), with lesser
7 amounts of richterite (11%) and tremolite (6%) with trace amounts of magnesioriebeckite,
8 edenite, and magnesio-arfvedsonite (Meeker et al., 2003) (see Section 2.2.3 for a more complete
9 discussion). Adverse health effects from tremolite exposure have been reported in both human
10 communities and laboratory animals; these effects are consistent with the human health effects
11 reported for Libby Amphibole asbestos. Studies examining the health effects of exposure to
12 winchite or richterite alone were not available in the published literature. The presentation of
13 noncancer and cancer health effects provides a comprehensive review of adverse health effects
14 observed from exposures to Libby Amphibole asbestos.
15
16 4.1. STUDIES IN HUMANS—EPIDEMIOLOGY
17 The Libby Amphibole asbestos epidemiologic database includes studies conducted in
18 occupational settings examining exposures to workers and community-based studies, which can
19 include exposures to workers, exposures to family members of workers, and exposures from
20 environmental sources. Occupational epidemiology studies exist for two worksites where
21 workers were exposed to Libby Amphibole asbestos. These worksites include the mine and mill
22 at the Zonolite Mountain operations near Libby, MT, and a vermiculite processing plant in
23 Marysville, OH. Worker cohorts from each site and the study results are described in
24 Section 4.1.1. Community-based studies include community health consultations for Libby, MT
25 conducted by the Agency for Toxic Substances and Disease Registry (ATSDR), including an
26 evaluation of cancer mortality data, and a health screening of current and former area
27 residents—including workers—that collected medical and exposure histories, chest X-rays, and
28 pulmonary function tests (ATSDR. 200Ib. 2000) (see Section 4.1.2). ATSDR, in conjunction
1: The term "Libby Amphibole asbestos" is used in this document to identify the mixture of amphibole mineral
fibers of varying elemental composition (e.g., winchite, richterite, tremolite, etc.) that have been identified in the
Rainy Creek complex near Libby, MT. It is further described in Section 2.2.
This document is a draft for review purposes only and does not constitute Agency policy.
4-1 DRAFT—DO NOT CITE OR QUOTE
-------
1 with state health departments, also conducted health consultations for 28 other communities
2 around vermiculite processing plants that were potentially exposed to Libby Amphibole asbestos
3 (see Section 4.1.4). These health consultations consisted of analyses of cancer incidence or
4 mortality data; results from nine of these studies are currently available.
5 No occupational studies are available for exposure to tremolite, richterite, or winchite
6 mineral fibers individually or as a mixture exposure, other than Libby Amphibole asbestos.
7 Communities, however, have been exposed to tremolite and other mineral fibers from natural
8 soils and outcroppings. Tremolite asbestos-containing soil has been used in whitewash in
9 interior wall coatings in parts of Turkey and Greece. Studies in these areas published as early as
10 1979 reported an increased risk of pleural and peritoneal malignant mesothelioma (Sichletidis et
11 al.. 1992: Baris et al.. 1987: Langer et al.. 1987: Baris et al.. 1979). More recent studies of
12 communities exposed to tremolite and chrysotile fibers report excess lung cancer and
13 mesothelioma (1.3- and 6.9-fold, respectively) (Hasanoglu et al., 2006). Other studies reported
14 pleural anomalies in residents exposed to naturally occurring asbestos, which includes actinolite,
15 tremolite, and anthophyllite (Metintas et al., 2005; Zeren et al., 2000). Clinical observations
16 include a bilateral increase in pleural calcification accompanied by restrictive lung function as
17 the disease progresses, a condition known as "Metsovo lung," named after a town in Greece
18 (Constantopoulos et al., 1985). In one community, the prevalence of pleural calcification was
19 46% (of 268 residents), increasing with age to 80% in residents over 70 (Langer et al., 1987).
20 Both tremolite and chrysotile were identified in bronchoalveolar lavage fluid of 65 residents
21 from different areas of Turkey who were environmentally exposed (Dumortier et al., 1998). The
22 health effects observed in communities with environmental and residential exposure to tremolite
23 are consistent with health effects documented for workers exposed to commercial forms of
24 asbestos.
25
26 4.1.1. Studies of Libby, MT Vermiculite Mining Operation Workers
27 Several studies of mortality from specific diseases among workers in the Libby, MT
28 mining operations have been conducted, beginning in the 1980s with the studies by McDonald
29 et al. (1986a) and Amandus and Wheeler. (1987). McDonald et al. (2004. 2002) published an
30 update with mortality data through 1999, and Sullivan (2007) updated the cohort originally
31 described by Amandus and Wheeler (1987) (referred to in this assessment as the Libby worker
This document is a draft for review purposes only and does not constitute Agency policy.
4-2 DRAFT—DO NOT CITE OR QUOTE
-------
1 cohort) with mortality data through 2001. Additionally, Larson et al. (201 Ob) reconstructed a
2 worker cohort and analyzed mortality through 2006 in this same study population, while another
3 study examined changes in lung abnormalities using X-rays taken between 1955 and 2004 of 88
4 workers (Larson et al., 2010a).
5
6 4.1.1.1. Description of Mining and Milling Operations
7 The vermiculite mining and milling operations have been described in considerable detail
8 (ATSDR, 2000; Amandus et al., 1987a). An open-pit vermiculite mine began limited operations
9 in 1923, and production increased rapidly between 1940 and 1950. This mine is located on
10 Zonolite Mountain, several miles east of Libby (ATSDR, 2000). The Kootenai River runs
11 between the town and the mine. The mining and milling operations continued until 1990
12 (ATSDR. 2008b. 2000).
13 The drilling and blasting procedures used in the strip-mining operations generated
14 considerable dust exposures, although the mining operations had lower intensity exposures
15 compared to the milling operations. Amandus et al. (1987a) noted that in 1970, a new drill with
16 a dust-control bagging system aimed at limiting workplace exposure was introduced to the
17 mining operations. Another aspect of the operations was the loading of ore for railroad
18 shipment. From 1935-1950, railroad box cars were loaded at a station in Libby. In 1950, the
19 loading station was moved to a loading dock on the Kootenai River, 7 miles east of town. Tank
20 cars were used from 1950-1959 and then switched to enclosed hopper cars in 1960.
21 The milling operations used a screening or sifting procedure to separate vermiculite
22 flakes from other particles and increase the concentration of vermiculite from approximately
23 20% in the bulk ore to 80-95% in the resulting product. A dry mill began operating in 1935, and
24 a wet mill began operating in the 1950s in the same building as the dry mill. One of the primary
25 changes in the conditions in the dry mill was the installation of a ventilation fan in 1964.
26 Exposure to asbestos inside the mill was estimated to be 4.6 times higher preceding this
27 installation (McDonald et al., 1986a). This ventilation fan resulted in higher amphibole fiber
28 exposures in the mill yard until 1968, when the exhaust stack for the fan was moved. Other
29 changes to the milling operations in the 1970s included replacement of hand bagging and sewing
30 with an automatic bagging machine (1972), pressurization of the skipper control room used for
31 transferring the ore concentrate from the mill to a storage site (1972), and construction of a new
This document is a draft for review purposes only and does not constitute Agency policy.
4-3 DRAFT—DO NOT CITE OR QUOTE
-------
1 wet mill (1974). Closing of the old dry and wet mills in 1976 had a substantial impact on
2 exposures at the worksite. In 1974, a new screening plant used to size-sort the ore concentrate
3 was constructed at the loading dock near the river. Two processing plants operated within the
4 town of Libby (ATSDR, 200 Ib). These expansion or exfoliation plants heated the ore
5 concentrate, resulting in additional release of the Libby Amphibole asbestos fibers in the area.
6
7 4.1.1.2. Exposure Estimation
8 In the early 1980s, two research groups conducted parallel studies of the mortality
9 experienced by workers in the Libby mining and milling operations. One study was undertaken
10 by the National Institute of Occupational Safety and Health (NIOSH) (Amandus et al.. 1987a:
11 Amandus and Wheeler, 1987; Amandus et al., 1987b) and the other by researchers from McGill
12 University (McDonald et al., 1986a: McDonald et al., 1986b). The exposure assessment
13 procedures used by the two groups relied on the same exposure measurements and used similar
14 assumptions in creating exposure estimates for specific job activities and time periods (see
15 Table 4-1). In brief, available air sampling data were used to construct a job-exposure matrix
16 assigning daily exposures (8-hour time-weighted average) for identified job codes based on
17 sampling data for specific locations and activities. Varying job codes and air exposures were
18 used for different time periods as appropriate to describe plant operations. Individual exposure
19 metrics (e.g., cumulative exposure) were calculated using the work history of each individual in
20 the study in conjunction with the plant job-exposure matrix. The specific study details for the
21 Libby, MT worker cohort are described in more detail below, with differences between the
22 research groups highlighted.
23 Before 1970, exposure estimates were based on midget impinger samples taken primarily
24 in the dry mill by state and federal inspectors. Total dust samples were measured as million
25 particles per cubic foot (mppcf) by the midget impinger method. Amandus et al. (1987a)
26 describe the period during which most of the midget impinger measurements were made as
27 1962-1967, and McDonald et al. (1986a) describe this period as 1962-1969, with a few
This document is a draft for review purposes only and does not constitute Agency policy.
4-4 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-1. Exposure assessment methodologies used in evaluations of Libby,
MT (see Section 4.1.1) and Marysville, OH (see Section 4.1.2) worker cohorts
Operation and study cohort
Asbestos fiber quantification and job-exposure
classification
Studies using methodology
Libby, MT mining and
milling operations; NIOSH
cohort
Exposure based on phase-contrast microscopy of
fibers >5 um long and aspect ratio >3:1
(1967-1982), and midget impinger data
(1956-1969).
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. Cumulative exposure
reported in units of fiber-years (equivalent to the
unit of fibers/cc-years EPA is using for all
studies).
Amandus et al. (1987a: 1987b);
Amandus and Wheeler (1987)
Libby, MT mining and
milling operations; NIOSH
cohort
Modification to Amandus et al. (1987a) job
classification: laborers and "unknown" jobs
assigned weighted-average exposure for all
unskilled jobs in work area (if known) during
calendar time period, rather than lower mill yard
exposure.
Weights based on the number of workers assigned
to unskilled jobs during same calendar time
period.
Sullivan (2007); Moolgavkar et
al. (2010)
Libby, MT mining and
milling operations; ATSDR
cohort assembled from W.R.
Grace & Co. records
Extension of Amandus et al. (1987a) exposure
data, with additional application of exposure
estimates to job titles from early 1980s through
1993.
Larson et al. (201 Ob; 2010a)
Libby, MT mining and
milling operations; McGill
University cohort
Similar to Amandus et al. (1987a). except with 28
"occupation locations," and conversion ratio = 4.6
for dry mill pre- and post 1964. Cumulative
exposure reported in units of fibers/mi-years
(equivalent to the unit of fibers/cc-years EPA is
using for all studies).
McDonald et al.
1986a; 1986b)
2002;
Marysville, OH
fertilizer production facility
using Libby, MT vermiculite
Libby, MT vermiculite ore used in the plant from
around 1960 to 1980.a
Industrial hygiene monitoring began 1972 (based
on fibers >5-um long, diameter <3 um, aspect
ratio >3:1). Breathing zone samples used after
1976. Fiber analysis by PCM.
Lockey et al.
(2008)
1984); Rohs et al.
1
2
o
3
4
5
6
7
aRohs et al. (2008) use 1963 as the beginning date of the use of Libby, MT vermiculite at the Marysville, OH plant,
based on information from ATSDR (2008b, 2005b). Lockey et al. (1984) used 1957 as the beginning date.
Subsequent to these publications, additional information was used to conclude that the beginning date for use of
Libby vermiculite ore was 1959 (see Appendix F).
NIOSH = National Institute for Occupational Safety and Health; PCM = phase contrast microscopy.
This document is a draft for review purposes only and does not constitute Agency policy.
4-5 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
additional measures in earlier years.12 The number of samples available before 1970 was
336 (Amandus et al., 1987a). 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 prior to 1974. The number of membrane-filter
samples available was 4,116. Air samples collected through membrane filters were analyzed by
phase contrast microscopy (PCM) to visually count fibers greater than >5-um long and having an
aspect ratio >3:1 (Amandus et al., 1987a).13 PCM methods from the 1960s allowed reliable
characterization of fibers with widths greater than approximately 0.4 um (Amandus et al., 1987a:
Rendall and Skikne, 1980). Further standardization of the PCM method provides better
visualization of thinner fibers, and 0.25 um width is considered the limit of resolution for fiber
width OPCS. 1986).
Table 4-2. Source of primary samples for fiber measurements at the Libby
mining and milling operations
Source
State of Montana
NIOSH
MESA/MSHAc'd
Company records
Unit of measurement
mppcf
fibers/ccb
fibers/cc
fibers/cc
Years
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 n and with an aspect ratio >3 were reported as fibers (see Section 2 for details).
°MESA: U.S. Mining and Enforcement and Safety Administration (former name of MSHA).
dMSHA: U.S. Mining and Safety Administration.
Source: Amandus et al. (1987a).
12Amandus et al. (1987a) indicates that one sample was available from 1942, and additional samples were available
after 1956; McDonald et al. Q986a) indicates that additional samples were available from 1944, 1956, and 1958.
13Amandus et al. (I987a) indicate (page 12, 4th full paragraph) that fibers >5-um long and with an aspect ratio >3
were measured. The actual value of the aspect ratio used by Amandus et al. could have been >3 because the
criterion for the NIOSH recommended exposure limit is based on an aspect ratio of >3, but EPA is reporting here the
information that was in the Amandus et al. (I987a) publication.
This document is a draft for review purposes only and does not constitute Agency policy.
4-6 DRAFT—DO NOT CITE OR QUOTE
-------
1 The samples taken from specific work locations within the plant were used to estimate
2 exposures in specific jobs and time periods based on professional consideration of temporal
3 changes in facilities, equipment, and job activities. The analysis by McDonald et al. (1986a) was
4 based on 28 occupation locations, while the work of Amandus et al. (1987a) was based on
5 25 occupation locations. These were defined to categorize tasks and locations across the mining,
6 milling, and shipping operations to group like tasks, with respect to exposure potential, for
7 evaluation. Both research groups established similar location operations for the Libby cohort.
8 For the years after 1968, data from filter samples were available for all locations, and NIOSH
9 researchers used the average (arithmetic mean) exposure when more than one sample was
10 available for a given location or job task and time period. McDonald et al. (1986a) used an
11 alternative procedure described by Oldham (1965) to estimate the mean of log-normal
12 distributions.
13 For exposures occurring prior to 1968, different procedures had to be used to estimate
14 exposures at the various locations because measures from sample filters were not available from
15 this earlier period. McDonald et al. (1986a) estimated pre-1968 exposure measurements for
16 26 location operations; assumptions were made and estimates based on data from later years or
17 related operations, although these assumptions are not stated by the authors. McDonald et al.
18 (1986a) did recognize the uncertainty in these calculations, and, for four areas, (drilling, ore
19 loading, river dock, and bagging plant), provided high and low estimates. Amandus et al.
20 (1987a) interviewed company employees, considered relative exposure levels between locations
21 post 1968 employing best available judgment to estimate task specific exposure levels.
22 Amandus et al. (1987a) expanded the procedures described in McDonald et al. (1986a) to
23 estimate pre-1968 exposures for four location operations (drilling, ore loading, river dock, and
24 bagging plant). "Low" and "high" estimates were generated using different assumptions; the
25 detailed results for the various assumptions were not presented, but the differences between them
26 were described by the authors as "slight," and the results presented were based on the high
27 estimate of exposure. Their decisions and specific assumptions are detailed (Amandus et al.,
28 1987a). The authors acknowledge there is uncertainty in exposure estimates prior to 1968 for
29 many of these locations. They do note that variability in sample results for the midget impinger
30 was low and that, in general, sample variability was low for fiber air-sampling results for areas
31 where the greatest numbers of employees worked (mill, service area, loading and bagging).
This document is a draft for review purposes only and does not constitute Agency policy.
4-7 DRAFT—DO NOT CITE OR QUOTE
-------
1 To estimate dry mill exposures prior to 1967, when fiber counts from phase contrast
2 microscopy air samples began to be used to measure exposures, Amandus et al. (1987a)
3 established a conversion factor from total dust counts (mmpcf) to fiber counts (fibers/cc). The
4 conversion ratio was based on a comparison of 336 impinger samples taken in 1965-1969 and
5 81 filter samples taken in 1967-1971. Both sets of samples were taken in the dry mill. Using
6 different subsets of the samples (i.e., different years) resulted in ratios that ranged from
7 1.9 fibers/cc: 1.0 mppcf to 11.5 fibers/cc: 1.0 mppcf The ratio based on the average fiber counts
8 from air samples (1967-1971) to the average total dust measurements in sample years
9 1965-1969 was 4.0 fibers/cc: 1.0 mppcf. This was the ratio used in the analyses in the NIOSH
10 studies (Amandus et al., 1987a: Amandus and Wheeler, 1987; Amandus et al., 1987b) because it
11 allowed for the use of the greatest amount of data from overlapping time periods, while
12 controlling for the reduced exposure levels after 1971 where fiber count based on phase contrast
13 microscopy—but not midget impinger data—were available. This dust-to-fiber conversion
14 factor was only used to estimate exposures in the dry mill. The resulting exposure
15 concentrations of 168 fibers/cc in 1963 and all prior years and 35.9 fibers/cc in 1964-1967 were
16 applied to dry mill exposures (Amandus et al., 1987a).
17 McDonald et al. (1986a) used a different procedure, based on the estimated reduction in
18 dust exposure with the installation of the ventilation system in 1964. Rather than develop a
19 direct dust-to-fiber conversion factor, they observed that total dust levels dropped approximately
20 4.6-fold after the installation of ventilation in the dry mill. Therefore, exposures in the dry mill
21 prior to 1965 were calculated as 4.6 times the fiber exposures measured by PCM between 1970
22 and 1974 (22.1 fibers/cc) resulting in estimated dry mill exposures of 101.5 fibers/cc prior to
23 1965 (McDonald et al.. 1986a).
24 Exposure estimates for each location operation derived from sampling data and history of
25 changes in control measures were used to develop a job-exposure matrix that estimated exposure
26 in fibers/cc for each job code during several calendar time periods. Jobs were mapped to
27 operation/location based on estimated time spent in different job tasks, thus estimating an 8-hour
28 time-weighted average exposure for each job during several calendar time periods. Job histories
29 from date of first employment to 1982 were used with the job-exposure matrix to develop
30 cumulative exposure estimates for each worker.
31
This document is a draft for review purposes only and does not constitute Agency policy.
4-8 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.1.1.2.1. Characteristics of historical fiber exposures
1 The resulting exposure estimates presented by both research groups, and the job-exposure
3 matrices used in calculating cumulative exposure for the cohort are based on fiber counts by
4 phase contrast microscopy analysis of air filters. As discussed in Section 2 (see Text Box 2-1),
5 phase contrast microscopy analysis does not distinguish between fiber mineralogy or
6 morphology and all fibers >5 um in length with an aspect ratio of 3:1 or greater are included.
7 Both researcher groups analyzed fibers available at the facility in order to identify the mineral
8 fibers in the air samples.
9 Transmission electron microscopy14 (TEM) analysis of airborne asbestos fibers indicated
10 a range of fiber morphologies—including long fibers with parallel sides, needlelike fibers, and
11 curved fibers (McDonald et al., 1986a). Of the fibers examined by TEM, >62% were >5 jim in
12 length and a wide range of dimensional characteristic were noted: length (1-70 jim), width
13 (0.1-2 |im), and aspect ratios from 3-100. Energy dispersive spectroscopy used to determine the
14 mineral analysis indicated that the fibers were in the actinolite-tremolite solid-solution series, but
15 sodium rich (McDonald et al., 1986a). This analysis is consistent with the current understanding
16 of amphibole asbestos found in the Libby mine (see Section 2.2.3).
17 At the time of their study, when exposure concentrations were reduced to generally less
18 than 1 fiber/cc, Amandus et al. (1987a) obtained eight air filters from area air samples collected
19 in the new wet mill and screening plant (provided by the mining company). These samples were
20 analyzed by phase contrast microscopy using the appropriate analytical method for the time
21 (NIOSH Physical and Chemical Analytical Method No. 239). From early method development
22 through current PCM analytical techniques, the Public Health Service, Occupational Safety and
23 Health Administration and NIOSH methods have defined a fiber by PCM analysis as having an
24 aspect ratio >3:1 (NIOSH. 1994a: Edwards and Lynch. 1968). Amandus et al. Q987a) reported
25 the dimensional characteristics of the fibers from these filters including aspect ratio, width, and
26 length (see Table 4-3). Data for 599 fibers from the 8 area air samples collected in the wet mill
27 and screening plant are provided. These data are limited in one sense by the minimum diameter
1 transmission electron microscopy (TEM) 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: energy dispersive spectroscopy (EDS) and selected area electron diffraction
(SAED).
This document is a draft for review purposes only and does not constitute Agency policy.
4-9 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-3. Dimensional characteristic 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
Percent
(%)
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
Percent
(%)
68
25
5
2
0
0
Aspect ratio
Range
5-10
10-20
20-50
50-100
>100
Total
counted
24
176
305
84
10
Percent
(%)
4
29
51
14
2
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
"Fibers were viewed and counted by Phase Contrast Microscopy.
Source: Amandus et al. (1987aX
and length cutoffs (>4.98-um long, >0.44-um wide, aspect ratio >3.0).15 Even with these greater
than 10:1, with 16% greater than 50:1 aspect ratio. Only 7% of the fibers had a width greater
than 0.88 jim, with one fiber reported of the 559 with a width greater than 1.76. It should be
noted that as NIOSH was examining PCM visible fibers, these data do not give the full fiber-size
distribution of Libby Amphibole asbestos fibers (see Section 2.2.3).
4.1.1.2.2. Descriptions of cohorts
The cohort studies conducted in the 1980s were similar in terms of exposure assessment
(as described in the previous section, Table 4-1), and other aspects of the study design (see
Table 4-4). 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. McDonald et al. O986a) included men hired before 1963 (n = 406) with
follow-up through 1983. A later analysis (McDonald et al., 2004) extended this follow-up
through 1999.
' See footnote 3, page 4-6.
This document is a draft for review purposes only and does not constitute Agency policy.
4-10 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-4. Respiratory (lung) cancer mortality and exposure-response analyses based on studies of the
vermiculite mine workers in Libby, MTa
Reference(s)
Inclusion criteria and design details
Standardized mortality
ratio (SMR) (95% CI)
Exposure-response analyses—lung cancer
I
'TS
s
•3
o
O
a
i
a,
o!
fel
H 1
o I
o I
H
Amandus and
Wheeler (19871
Men, hired before 1970, worked at least
one year, follow-up through 1982
(n = 575); 161 deaths (159 with death
certificates).
Mean duration: 8.3 years (0 worked less
than 1 year).
Mean fiber-years: 200.3.
12 female workers not included in this
analysis.
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 years since
first hire (latency):
Lung (n = 12)
SMR: 2.3 (p<0.05)
No exclusions:
Cumulative Exposure n
0.0-49 fibers/cc-yrs 6
50-99 fibers/cc-yrs 2
100-399 fibers/cc-yrs 2
>400 fibers/cc-yrs 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 of more years since first hire (20-year latency)
Cumulative Exposure
0.0-49 fibers/cc-yrs 2
50-99 fibers/cc-yrs 2
100-399 fibers/cc-yrs 1
>400 fibers/cc-yrs 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 a linear regression analysis of data with at least 20 years latency,
the results per fiber-year were: beta (standard error) = 0.60 (0.13)
and 0.58 (0.08) 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).
O
c
o
McDonald et al.
(2004): McDonald
et al. (1986a)
Men, hired before 1963, worked at least
one year (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 from 1983-1998 obtained from
National Death Index.
Mean duration: 8.7 years (0 worked less
than 1 year).
Meanfiber-yrs: 144.6.
Respiratory (n = 44)
SMR: 2.4 (1.7, 3.2)
Excluding first 10 years of follow-up:
Cumulative Exposure n
0.0-11.6 fibers/cc-yrs 5
11.7-25.1 fibers/cc-yrs 9
25.2-113.7 fibers/cc-yrs 10
>113.8 fibers/cc-yrs 163
per 100 fibers/cc-yrs increase
RR (95% Cir
1.0 (referent)
1.7 (0.58, 5.2)
1.9(0.63,5.5)
3.2(1.2,8.8)
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.
-------
Table 4-4. Respiratory (lung) cancer mortality and exposure-response analyses 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 day, follow-up 1960-2001
(n = 1,672); 767 deaths (95% with
known cause of death).
Mean duration: 4.0 years (808, -50%
worked less than 1 year).
Median fibers/cc -years: 8.7.
Underlying cause of death data from
death certificates or National Death
Index-Plus.
Inclusion criteria not described
(n — 1 S^'JV fr\11r\i*7 nr\ tlirr\nC\C\f\-
952 deaths (80% with known cause of
death).
Median duration: 0.8 years;
Median fibers/cc-yr =4.3.
Immediate and underlying cause of
death data (i.e., multiple cause of death)
from death certificates or National
Death Index-Plus.
Standardized mortality
ratio (SMR) (95% CI)
15 year exposure lag:
All cancer (n = 202)
SMR: 1.4 (1.2, 1.6)
Lung (n = 89)
SMR: 1.7(1.4,2.1)
Lung (n = 104)
SMR: 1.6 (1.3, 2.0)
Exposure-response analyses — lung cancer
15 year exposure lag:
Cumulative Exrjosure n SMR(95%CI)b SRR(95%CI)C
0.0-4.49 fibers/cc-yrs 19 1.5(0.9,2.3) 1.0 (referent)
4.5-22.9 fibers/cc-yrs 24 1.6(1.1,2.5) 1.1(0.6,2.0)
23. 0-99.0 fibers/cc-yrs 23 1.8(1.1,2.7) 1.4(0.7,2.7)
>100 fibers/cc-yrs 23 1.9(1.2,2.9) 1.5(0.8,2.8)
linear trend test (p < 0.00 1)
Duration n SMR(95%CI)b SRR(95%CI)C
10 years 14 2.5(1.4,4.3) 1.8(0.9,3.4)
20 year exposure lag:
Cumulative Exposure n SMR (95% CI)b RR (95% CI)e
0.0-<1. 4 fibers/cc-yrs 19 (not reported) 1.0 (referent)
1.4 to <8.6 fibers/cc-yrs 20 (not reported) 1.1(0.6,2.1)
8.6 to <44.0 fibers/cc-yrs 21 (not reported) 1.7(1.0,3.0)
>44.0 fibers/cc-yrs 38 (not reported) 3.2(1.8,5.3)
Per 100 fibers/cc-yrs increase 1.11 (1.05, 1.18)
(p = 0.006)
I
'TS
s
•3
o
§
o!
fel
H 1
H
O
c
o
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-years 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. (2010b). the RR is based on Cox proportional hazards modeling using an internal referent group.
SMR = standardized mortality ratio, CI = confidence interval, SRR = standardized rate ratio, RR = relative risk.
-------
1 A more recent analysis of the Libby, MT workers expanded the cohort to include all
2 workers, regardless of duration of employment (Sullivan, 2007). The total sample
3 (n= 1,672 white men) included 808 workers who had worked for less than 1 year. These
4 short-term workers had been excluded from the previous studies in Table 4-4. Analyses
5 presented in the report were based on follow-up from 1960-2001. This beginning point was
6 chosen because comparison rates for asbestosis, an outcome of interest, were not available before
7 1960 in the NIOSH Life-Table Analysis System, the analytic software used in the analysis
8 (Sullivan, 2007). Few deaths had occurred before 1960 (95 men dead or lost to follow-up before
9 1960 were excluded), so this exclusion criterion would not be expected to result in a substantial
10 loss of outcomes. Because mesothelioma was not coded separately until 1999, the mesothelioma
11 risk analysis is based on data from 1999-2001.
12 In the study by Sullivan (2007), comparison rates for standardized mortality ratio (SMR)
13 analyses were calculated from U.S. population cause-specific mortality data (limited to white
14 males) and adjusted for age and calendar year of follow-up (using 5-year groups). McDonald
15 et al. (2004) also used comparison rates from the U.S. population and included additional
16 analyses for the category of respiratory cancers using Montana population rates.
17 Larson et al. (201 Ob) reconstructed a worker cohort based on company records and
18 analyzed mortality risks through 2006. This study included 1862 workers; inclusion and
19 exclusion criteria are not stated, and, thus, it is not clear whether this analysis excluded females
20 or specific ethnic groups. The exposure assessment methodology was based on the methods
21 described by Amandus et al. (1987a)—without the modification used by Sullivan (2007).
22 Multiple causes of death (i.e., from any mention on the death certificate) were used, rather than
23 underlying cause of death. Because multiple causes of death are used, more than one cause of
24 death can be coded for an individual.
25 The studies of the Libby worker cohort by Amandus and Wheeler (1987), Sullivan
26 (2007), and Larson et al. (201 Ob) defined lung-cancer mortality based on more specific causes of
27 death codes compared to the broader classification of "all respiratory cancer" used by McDonald
28 et al. (2004; 1986a). For example, the International Classification of Diseases (ICD) codes used
29 for deaths due to cancers of the trachea, bronchus, and lung occurring during the applicable years
30 in the NIOSH cohort in Sullivan (2007) were ICD-7 162.0-162.1, 162.8, 163, ICD-8 162, and
31 ICD-9 162. In the first McDonald et al. Q986a) analysis, ICD-8 codes 160-163 for respiratory
This document is a draft for review purposes only and does not constitute Agency policy.
4-13 DRAFT—DO NOT CITE OR QUOTE
-------
1 cancer were used, which also included cancer of the larynx (ICD-8 code 161) and some types of
2 "other" respiratory cancers (ICD-8 code 160). The updated follow-up for 1999 included ICD-9
3 codes 160-165 for respiratory cancer, adding the "other" respiratory cancer group (ICD-9 codes
4 164 and 165). In the national Surveillance, Epidemiology, and End Results (SEER) cancer data
5 from 2003-2007, the age-adjusted mortality rate for cancer of the larynx was 1.2, compared to
6 52.5 per 100,000 person-years for lung and bronchial cancer (NCI, 2011). Thus, these additional
7 categories (larynx and "other" respiratory cancers) represent a relatively small proportion of
8 respiratory cancers, but they could be a source of some misclassification of the outcome if these
9 other cancers are not related to asbestos exposure.
10 The classification of mesothelioma was more difficult because of the lack of a unique
11 ICD code for mesothelioma prior to the 10th revision, implemented in the United States in 1999.
12 The updated NIOSH study by Sullivan (2007) identified 15 deaths for which mesothelioma was
13 mentioned on the death certificate. Only two of these deaths occurred between 1999 and 2001;
14 these were coded using the ICD-10 mesothelioma coding (C45). Larson et al. (201 Ob) classified
15 all death certificates listing mesothelioma as ICD-10 code C45. The updated McGill study
16 (McDonald et al., 2004) (with analysis through 1998) noted that the classification of
17 mesothelioma was based on a nosologist's review of death certificates; only 5 of the 12 cases
18 classified as mesothelioma had a cause of death listed as pleural cancer (ICD-9 code 163).
19
20 4.1.1.3. Cancer Mortality Risk
21 4.1.1.3.1. Lung cancer
22 The results within and among the papers in these two sets of studies (Larson et al., 201 Ob;
23 Sullivan. 2007: McDonald et al.. 2004: Amandus and Wheeler. 1987: McDonald et al.. 1986a)
24 show similar effects in terms of the increased risk seen for lung (or respiratory) cancer (see
25 Table 4-4). Exposure-response analyses from these studies demonstrated increasing mortality
26 with increasing exposure, using categorical and continuous measures of exposure, different lag
27 periods, and different exposure metrics. Because of the congruence in results and overlapping of
28 study participants among these studies, the most recent studies are discussed in detail below.
29 The analysis of McDonald et al. (2004) is limited to 406 male workers who were hired
30 before 1963 and who were employed for at least 1 year. The mean duration of work was
31 8.7 years. Cause of death data were obtained from the National Death Index for deaths from
This document is a draft for review purposes only and does not constitute Agency policy.
4-14 DRAFT—DO NOT CITE OR QUOTE
-------
1 1983 to 1998 and were based on ICD-8 coding by a nosologist using death certificates obtained
2 for deaths before 1983. Expected rates were based on age-, race- and sex- specific rates. A total
3 of 44 deaths due to respiratory cancers were observed, for an SMR = 2.4 (95% confidence
4 interval [CI]: 1.7, 3.2). A pattern of increasing mortality with increasing cumulative exposure
5 was seen, with relative risks (RRs) of 1.0 (referent), 1.7, 1.9, and 3.2 in categories of 0.0-11.6,
6 11.7-25, 25.2-113.7, and >113.8 fibers/cc-years, respectively (see Table 4-4). The estimated
7 linear increase in RR of respiratory cancer risk per 100 fibers/cc-years cumulative exposure was
8 0.36 (95% CI: 0.03, 1.2) (p = 0.02). McDonald et al. (2004) reported that similar results were
9 obtained with measures of exposure intensity and measures of residence-weighted exposure, but
10 the data were not presented in the paper.
11 Sullivan (2007) included 1,672 white male workers who were alive in 1960 or hired after
12 1960. There was no minimum duration of employment required for inclusion in this analysis,
13 and approximately 50% of the cohort (n = 808) had worked less than 1 year. Mortality follow-up
14 was conducted through 2001, with 767 identified deaths. The exposure assessment protocol was
15 based on that described by Amandus et al. (1987a), with a modification to the estimated intensity
16 of exposure to laborers and to those with "unknown" jobs. Sullivan (2007) assigned
17 weighted-average exposure for all unskilled jobs in a department (if known) during a calendar
18 time period, rather than lower mill yard exposure used by Amandus et al. (1987a). The weights
19 are based on the number of workers assigned to unskilled jobs during the same calendar time
20 period. In the Sullivan (2007) follow-up, SMRs, using underlying cause-of-death data (based on
21 death certificates) obtained through the National Death Index and from individual states, and
22 expected mortality based on national age-, race-, and sex-specific rates, were calculated. Using a
23 15-year exposure lag, SMRs were increased for lung cancer (n = 89, SMR = 1.7, 95% CI: 1.4,
24 2.1) and for all cancer mortality (n = 202, SMR = 1.4, 95% CI: 1.2, 1.6) (see Table 4-4).
25 Additionally, an internal referent group was used for analyses of risk in relation to cumulative
26 exposure and duration. The results of these internal analyses are presented as standardized rate
27 ratios (SRR) for white men, controlling for age group. Increasing risks across categories of
28 cumulative exposure and duration were observed with both types of analyses, indicating a
29 positive exposure-response relationship. The SMR estimates for lung-cancer mortality were 1.5,
30 1.6, 1.8, and 1.9 in the 1- to 4.49-, 4.5- to 22.9-, 23.0- to 99.0-, and >100 fibers/cc-year exposure
31 categories, respectively. The SRR estimates were 1.0, 1.1., 1.4, and 1.5, respectively, across
This document is a draft for review purposes only and does not constitute Agency policy.
4-15 DRAFT—DO NOT CITE OR QUOTE
-------
1 these same exposure categories (see Table 4-4). For comparison to the earlier work by
2 McDonald et al. (1986a), an SMR was provided for all respiratory cancer in those employed at
3 least 1 year (SMR 2.0, 95% CI: 1.5-2.5). For the full cohort employed at least 1 day, the SMR
4 for all respiratory cancer was 1.7 (95% CI: 1.4-2.1) (Sullivan. 2007).
5 Amandus and Wheeler (1987) provide some information on the smoking history of a
6 sample of 161 male workers employed during 1975-1982 with at least 5 years of employment in
7 the Libby cohort study and comparison data based on surveys conducted in the United States
8 from 1955-1978. Among the workers, 35% were current smokers, and 49% were former
9 smokers. This smoking information was obtained from questionnaires the company
10 administered to workers after 1975. Assuming the definitions are similar to those of the national
11 surveys, however, the prevalence of current smokers is similar in the worker cohort compared to
12 the U.S. white male population data (ranging from 37.5-41.9% current smokers between 1975
13 and 1978). The only year in this range with data on former smokers in the national survey is
14 1975, and, at that time, the prevalence of former smokers in the population data was 29.2%,
15 about 20% lower than among the workers. Using an estimated RR of lung cancer of 14 among
16 smokers, Amandus and Wheeler (1987) estimated that the difference in smoking rates between
17 workers and the comparison population could have resulted in a 23% increase in the observed
18 risk ratio and commented that the increased risk observed in the lower dose range
19 (<50 fiber-years) could be the result of confounding by smoking status.
20 Smoking patterns in the U.S. population changed considerably over the period
21 corresponding to the data reported by Amandus and Wheeler (1987). In the National Health
22 Interview Surveys conducted between 1974 and 1983, the prevalence of smoking in males
23 age 20 and older decreased from 42.1 to 35.5% (HHS. 1990). In addition, the prevalence of
24 former smokers can depend on the definition used. Based on 1986 survey data, the percentage of
25 adults age 17 and older classified as former smokers varied between 14.7 and 25.8% using
26 different definitions for time since last smoked (e.g., from quitting 5 or more years ago to
27 quitting within the past 3 months) (HHS, 1990). Thus, given the lack of information pertaining
28 to the period in which smoking information was collected and the specifics of the sources that
29 were used, EPA concludes there is considerable uncertainty regarding the evidence for
30 differences in smoking rates between the workers and the external comparison population.
This document is a draft for review purposes only and does not constitute Agency policy.
4-16 DRAFT—DO NOT CITE OR QUOTE
-------
1 Larson et al. (201 Ob) evaluated multiple causes of death, and, therefore, more than one
2 cause of death can be coded for an individual. A total of 104 lung or bronchus cancer deaths
3 were observed, for an SMR of 1.6 (95% CI: 1.3, 2.0) using an external comparison of United
4 States cause of death data from 1960 to 2002 (Larson et al., 201 Ob). A higher risk was seen in
5 the higher cumulative exposure categories using Cox proportional hazards modeling with an
6 internal referent group: relative risk 1.0 (referent), 1.1 (95% CI: 0.6, 2.1), 1.7 (95% CI: 1.0, 3.0),
7 and 3.2 (95% CI: 1.8, 5.3) respectively, for <1.4 (referent), 1.4 to <8.6, 8.6 to <44.0 and >44.0
8 fibers/cc-years. Larson et al. (201 Ob) used data from a health screening program conducted in
9 Libby by ATSDR in 2000-2001 (described in Section 4.1.2.2) pertaining to smoking history to
10 estimate that the proportion of smokers ranged from 50% to 66% in the unexposed group
11 (defined as exposure <8.6 fibers/cc-years) and between 66% and 85% among the exposed
12 (defined as >8.6 fibers/cc-years). Larson et al. (201 Ob) used these estimates in a Monte Carlo
13 simulation to estimate the potential bias in lung cancer risks that could have been introduced by
14 differences in smoking patterns. The bias-adjustment factor (RRunadjusted/RRadjusted = 1.3) reduced
15 the overall RR estimate for lung cancer from 2.4 to 2.0.
16
17 4.1.1.3.2. Mesothelioma
18 Data pertaining to mesothelioma risk from the available studies are summarized in
19 Table 4-5. McDonald et al. (2004) presented dose-response modeling of mesothelioma risk
20 based on 12 cases. Using Poisson regression, the mesothelioma mortality rate across increasing
21 categories of exposure was compared to the rate in the lowest exposure category. Note that the
22 referent group was also at excess risk of dying from mesothelioma; that is, one to three cases of
23 mesothelioma were observed in the referent group, depending on the exposure index. Three
24 exposure indices were used in analysis: average intensity over the first 5 years of employment,
25 cumulative exposure, and residence-weighted cumulative exposure. Because of the requirement
26 for 5 years of employment data, 199 individuals (including three mesothelioma cases) were
27 excluded from the analysis of average intensity. The residence-weighted cumulative exposure
28 was based on the summation of exposure by year, weighted by years since the exposure. This
29 metric gives greater weight to exposures that occurred a longer time ago. Although evidence of
30 an excess risk of dying from mesothelioma was seen in all groups, there was little evidence of
31 increasing RR with increasing average intensity or cumulative exposure. For the
This document is a draft for review purposes only and does not constitute Agency policy.
4-17 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-5. Mesothelioma mortality risk based on studies of the vermiculite
mine workers in Libby, MTa
Reference(s)
Inclusion criteria and design
details
Results
Amandus and
Wheeler (19871
Men, hired before 1970, worked at
least 1 year, follow-up through 1982
(n = 575); 161 deaths (159 with death
certificates).
Mean duration: 8.3 years (0 worked
less than 1 year). Mean fiber-years:
200.3. Twelve female workers not
included in this analysis.
2 mesothelioma deaths observed (hired in 1946, 33 years
latency, exposure >300 fibers/cc-years); 1.2% of all
deaths
McDonald et al.
(2004):
McDonald et al.
(1986a)
Men, hired before 1963, worked at
least 1 year (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 death for
deaths from 1983-1998 obtained
from National Death Index.
Mean duration: 8.7 years (0 worked
less than 1 year). Meanfiber-yrs:
144.6.
12 mesothelioma deaths observed; 4.
Excluding first 10 years of follow-up:
Cumulative Exposure n
0.0-11.6 fibers/cc-yrs 1
11.7-25.1fibers/cc-yrs 4
25.2-113.7 fibers/cc-yrs 3
>113.8 fibers/cc-yrs 4
per 100 fibers/cc-yrs increase
Intensity Category
0.0-11.6 fibers/cc-yrs
11.7-25.1 fibers/cc-yrs
25.2-113.7 fibers/cc-yrs
>113.8 fibers/cc-yrs
per 100 fibers/cc-yrs increase
Residence-weighted
0.0-25.1 fibers/cc-yrs
25.2-113.7 fibers/cc-yrs
>113.8 fibers/cc-yrs
per 100 fibers/cc-yrs increase
1
4
2
2
3
4
5
2% of all deaths
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)
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)
RR (95% CI)b
1.0 (referent)
1.57 (0.35, 7.07)
1.95(0.41,8.51)
0.03 (<0, 6.4)
(p > 0.20)
White men, enumerated in 1982,
alive in 1960 or hired after 1960,
worked at least 1 day, follow-up
1960-2001 (n = 1,672); 767 deaths
(95% with known cause of death).
Mean duration: 4.0 years (808, -50%
worked less than 1 year). Median
fibers/cc-years: 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.
15 mesothelioma deaths observed; 2% of all deaths
N = 2 for 1999-2001:
SMR: 15.1 (95% CI: 1.8,54.4)
Pleura! (n = 4)
SMR: 23.3 (95% CI: 6.3, 59.5)
This document is a draft for review purposes only and does not constitute Agency policy.
4-18 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-5. Mesothelioma mortality risk based on studies of the vermiculite
mine workers in Libby, MTa (continued)
Reference(s)
Larson et al.
(2Q10b)
Inclusion criteria and design
details
Inclusion criteria not described
(n = 1,862); follow-up through 2006;
952 deaths (80% with known cause
of death). Median duration: 0.8
years; Median fibers/cc-yr = 4.3.
Immediate and underlying cause of
death data (i.e., multiple causes of
death) from death certificates or
National Death Index-Plus.
Results
19 mesothelioma deaths observed
SMR: 94.8 (95% CI: 57, 248)
20 year exposure lag:
Cumulative Exposure n
<1.4fibers/cc-yrs 1
1.4 to <8.6 fibers/cc-yrs 2
8.6 to <440 fibers/cc-yrs 5
>44.0 fibers/cc-yrs 11
per 100 fibers/cc-yrs increase
RR (95% CI)C
1.0 (referent)
1.9(0.31, 13.6)
4.5 (0.8, 24.6)
17.1 (3.7,78.1)
1.15(1.03, 1.28)
(p= 0.0134)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
"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 et al. (2010b). the RR is based on Cox proportional hazards modeling using an internal referent group.
SMR = standardized mortality ratio, CI = confidence interval, SRR = standardized rate ratio, RR = relative risk.
residence-weighted cumulative exposure, an RR of 1.57 was observed among those with
500.1-1,826.8 fibers/cc-years exposure, and an RR of 1.95 was observed among workers with
higher residence-weighted cumulative exposure. Sullivan (2007) identified 15 deaths from
mesothelioma through a manual review of death certificates, with 14 classified as "pleural or
unspecified," and 1 classified as "peritoneal." Only two of these deaths occurred between 1999
and 2001, the period for which comparison data using the ICD-10 classification criteria were
available. Based on these two mesothelioma deaths, the SMR was 14.1 (95% CI: 1.8, 54.4).
Larson et al. (201 Ob) identified 19 mesothelioma deaths (coding any mention of mesothelioma
on the death certificate as the ICD-10 classification of C45). Comparison data were based on
multiple-causes-of-death data (1960 to 2002). The SMR for mesothelioma was 94.8 (95% CI:
57.0, 148.0), and an increasing risk was seen across quartiles of exposure (see Table 4-5). The
comparison rates for the SMR analysis are based on multiple cause of death data for the U.S.
population from 1960-2002; only a small portion of this period included the ICD-10 coding
scheme for mesothelioma. Thus, the expected rates could be underestimated, biasing the effect
estimates upward.
This document is a draft for review purposes only and does not constitute Agency policy.
4-19 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.1.1.3.3. Other cancers
1 Larson et al. (201 Ob) presented data on cancers other than respiratory tract and
3 mesothelioma. The category of malignant neoplasms of digestive organs and peritoneum
4 included 39 observed deaths, for an SMR of 0.8 (95% CI: 0.6, 1.1). No risk in relation to
5 asbestos exposure was seen with a 20-year lag. The potential for underascertainment of specific
6 causes of death should be noted, however, given the 10% loss to follow-up and missing cause of
7 death data for 9% of the identified deaths.
8
9 4.1.1.3.4. Summary of cancer mortality risk in Libby, MT vermiculite mining operation
10 workers
11 The studies conducted in the 1980s (Amandus and Wheeler, 1987; McDonald et al.,
12 1986a) as well as the extended follow-up studies published in more recent years (Larson et al.,
13 201 Ob: Sullivan, 2007; McDonald et al., 2004) provide evidence of an increased risk of
14 lung-cancer mortality and of mesothelioma mortality among the workers in the Libby
15 vermiculite mining and processing operations. The lung cancer analyses using an internal
16 referent group in the larger follow-up studies (Larson et al., 201 Ob: Sullivan, 2007; McDonald et
17 al., 2004) observed increasing risks with increasing cumulative exposure exposures when
18 analyzed using quartiles or as a continuous measure. Increased risks are also seen in the studies
19 reporting analyses using an external referent group (i.e., standardized mortality ratios) (Sullivan,
20 2007: Amandus and Wheeler, 1987: McDonald et al., 1986a).
21
22 4.1.1.4. Noncancer Effects: Respiratory and Cardiovascular Disease
23 4.1.1.4.1. Asbestosis and other nonmalignant respiratory disease mortality
24 The studies described previously also reported noncancer mortality data, with a specific
25 focus on respiratory diseases (see Table 4-6). In Sullivan (2007), the SMR for asbestosis
26 (ICD-9 code 501) was 166 (based on n = 22, underlying cause of death compared to a U.S. white
27 male referent group). In Larson et al. (201 Ob), the SMR was 143 (95% CI: 111,181), based on
28 69 observed asbestosis-related deaths using multiple-causes-of-death data. Increasing
29 cumulative exposure was observed to increase the risk for asbestosis mortality in both of these
30 analyses (see Table 4-6). A two- to threefold increase was also seen for other categories of
31 nonmalignant respiratory disease in Larson et al. (201 Ob), with an SMR of 2.4 (95% CI: 2.2, 2.6)
This document is a draft for review purposes only and does not constitute Agency policy.
4-20 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-6. Nonmalignant respiratory mortality studies of the vermiculite
mine workers in Libby, MTa
Reference(s)
Respiratory disease
(SMR, 95% CI)
Dose-response analyses:
Nonmalignant respiratory diseases and asbestosis
Amandus
and Wheeler
(1987)
(NIOSH)
No exclusions:
Nonmalignant respiratory
diseases (n = 20)
SMR: 2.4 (1.5, 3.8)
20 year latency:
Nonmalignant respiratory
diseases (n = 12)
SMR: 2.5 (p< 0.05)
No exclusions'. Nonmalignant respiratory diseases
Cumulative Exposure n
0.0-49 fibers/cc-yrs 8
50-99 fibers/cc-yrs 2
100-399 fibers/cc-yrs 3
>400 fibers/cc-yrs 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 years since first hire (latency): Nonmalignant respiratory
diseases
Cumulative Exposure n
0.0-49 fibers/cc-yrs 7
50-99 fibers/cc-yrs 2
100-399 fibers/cc-yrs 0
>400 fibers/cc-yrs 3
SMR (95%CI)b
3.3 (not reported, but;? < 0.05)
2.8 (not reported)
0 (not reported)
2.8 (not reported)
McDonald et
al. (2004):
McDonald et
al. (1986a)
(McGill)
Nonmalignant respiratory
diseases (n = 51)
SMR: 3.1(2.3, 4.1)
Excluding first 10 years of follow-up: Nonmalignant respiratory diseases
Cumulative Exposure
0.0-11.6 fibers/cc-yrs
11.7-25.1 fibers/cc-yrs
25.2-113.7 fibers/cc-yrs
>113.8fibers/cc-yrs
per 100 fibers/cc-yrs
n_ RR (95%CI)a
5 1.0 (referent)
13 2.5 (0.88, 7.2)
14 2.6 (0.93, 7.3)
19 3.1(1.2,8.4)
0.38(0.12, 0.96) (p = 0.0001)
Sullivan
(2007)
(NIOSH)
15 year 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)
15 year exposure lag: Asbestosis
Cumulative Exposure n
0.0-49.9 fibers/cc-yrs 3
50.0-249.9 fibers/cc-yrs 8
>250 fibers/cc-yrs 11
linear trend test
SMR (95% CI)b
37 (7.5, 122)
213 (91.6,433)
749 (373, 1,368)
SRR (95% CI)C
1.0 (referent)
7.3 (1.9,28.5)
25.3 (6.6, 96.3)
(p<0.01)
15year exposure lag: Nonmalignant respiratory diseases
n_ SMR(95%CI)b SRR (95% CI)C
Cumulative Exposure
0.0-4.49 fibers/cc-yrs
4.5-19.9 fibers/cc-yrs
20.0-84.9 fibers/cc-yrs
85.0-299.9 fibers/cc-yrs
>300 fibers/cc-yrs
18 1.8(1.1,2.8)
24 2.0(1.3,3.0)
26 2.2(1.5,3.3)
20 2.6(1.6,4.0)
23 4.8(3.1,7.3)
1.0 (referent)
1.2(0.6,2.3)
1.5(0.8,2.9)
1.4(0.7,2.7)
2.8(1.3,5.7)
This document is a draft for review purposes only and does not constitute Agency policy.
4-21 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-6. Nonmalignant respiratory mortality studies of the vermiculite
mine workers in Libby, MTa (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 year exposure lag: Asbestosis
Cumulative Exposure n SMR(95%CI)b RR(95%CI)e
<1.4fibers/cc-yrs 4 (not reported) 1.0 (referent)
1.4-<8.6fibers/cc-yrs 8 (not reported) 2.8(1.0,7.6)
86- <44.0 fibers/cc-yrs 25 (not reported) 8.0(3.2,19.5)
>44.0 fibers/cc-yrs 32 (not reported) 11.8(4.9,28.7)
Per 100 fibers/cc-yrs increase 1.18 (1.12, 1.23)
(p< 0.001)
20 'year exposure lag: Nonmalignant respiratory diseases
Cumulative Exposure n SMR(95%CI)b RR(95%CI)e
<1. 4 fibers/cc-yrs 43 (not reported) 1.0 (referent)
1. 4- <8.6 fibers/cc-yrs 46 (not reported) 1.4(0.9,2.1)
86- <44.0 fibers/cc-yrs 56 (not reported) 1.8(1.3,2.7)
>44.0 fibers/cc-yrs 58 (not reported) 2.5(1.7,3.6)
Per 100 fibers/cc-yrs increase 1.08 (1.03, 1.13)
(p = 0.0028)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
"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-years 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 internal referent group.
eln Larson et al. (2010b). the RR is based on Cox proportional hazards modeling using an internal referent group.
SMR = standardized mortality ratio, CI = confidence interval, SRR = standardized rate ratio, RR = relative
risk.
for all nonmalignant respiratory disease, and SMR = 2.8 (95% CI: 2.3, 3.4) for diseases other
than asbestosis, chronic obstructive pulmonary disease, and silicosis. These results are similar to
the nonmalignant respiratory disease mortality data from studies of this cohort using underlying
cause-of-death data. A markedly higher risk of nonmalignant respiratory disease mortality was
also observed in the cumulative exposure category of >300 or >400 fibers/cc-years, respectively
in Sullivan (2007) and Amandus and Wheeler (1987). Larson et al. (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
This document is a draft for review purposes only and does not constitute Agency policy.
4-22 DRAFT—DO NOT CITE OR QUOTE
-------
1 the cohort. The bias-adjustment factor (RRUnadjusted/RRadjusted = 1 -2) reduced the overall RR
2 estimate for nonmalignant respiratory mortality from 2.1 to 1.8.
o
J
4 4.1.1.4.2. Radiographic abnormalities
5 Respiratory disease risk is also evidenced by chest radiographs showing pleural and
6 parenchymal abnormalities in the Libby, MT worker cohorts (see Table 4-7). Two of these
7 studies were conducted in the 1980s and were based on X-rays of a subset of workers taken for
8 either an annual workplace screening (Amandus et al., 1987b) or as part of a study examination
9 (McDonald et al., 1986b). The subset of McDonald et al. (1986b) included 164 workers
10 currently employed at the Libby facility, 80 former employees, and 47 area residents without
11 known dust exposure. The subset selected by Amandus et al. (1987b) included workers with at
12 least 5 years tenure who had worked at Libby at some time during 1975-1982. The most recent
13 X-ray film for each worker, which NIOSH obtained from the Libby hospital that performed the
14 screening, was independently read by three qualified readers using the International Labor Office
15 (ILO) classification system. For the analysis, the classification indicating pleural abnormalities
16 by at least two of the three readers was used to determine the presence of pleural abnormalities,
17 while the median reading was used to determine the profusion category of small opacities. In the
18 McDonald et al. (1986b) study, all three readings agreed for about 90% of the chest X-rays that
19 showed evidence of pleural calcification, obliteration of the costophrenic angle, and pleural
20 thickening on the diaphragm. Similarly, all three readings agreed for about 80% of chest X-rays
21 that showed evidence of small opacities, pleural plaques, or diffuse thickening. Amandus et al.
22 (1987b) provided a more detailed breakdown of the correspondence between readers for the
23 rating of small opacities (by category). The prevalences of any opacities (category 1/0 or more)
24 were 10, 16, and 10% for Readers A, B, and C. This difference among raters was similar to that
25 seen in other studies. Other design details are described in Table 4-7.
26 Although both research groups utilized the ILO 1980 guidelines, McDonald et al (1986b)
27 reported pleural thickening on the chest wall (both pleural plaques and diffuse) but excluding
28 other sites. Amandus et al (1987b) report "any pleural change" (both pleural plaques and
29 diffuse, defined as "... any unilateral or bilateral pleural change, which included pleural plaque,
30 diffuse pleural thickening of the chest wall, diaphragm or other site, but excluded costophrenic
This document is a draft for review purposes only and does not constitute Agency policy.
4-23 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-7. Chest radiographic studies of the Libby, MT vermiculite mine
workers
Reference(s)
Inclusion criteria and design details
Results
McDonald et
al. (1986b)
Men employed on July 1, 1983 (n = 164).
Former male employees living within 200 miles; hired
before 1963 (n = 80), worked at least 1 year
(80 participants from 110 eligible); 43 had a previous
X-ray.
Men without known dust exposure (n = 47); X-rays
taken for other reasons (mostly employment related)
at same place during study period; 24 had a previous
X-ray.
Data from nine women employed on July 1, 1983 not
included in this report.
Pleura! thickening of the chest wall observed
in 15.9% of current employees and 52.5% of
past employees.
Small opacities (>1/0) observed in 9.1% of
current employees and 37.5% of past
employees.
Both abnormalities increased with age.
Age-adjusted and age-stratified (>60 years
old) analyses showed increasing risk of both
abnormalities with increasing cumulative
exposure.
Amandus et
al. (1987b)
Men, employed during 1975-1982 with at least
5 years tenure (n = 191); 184 with previous chest
X-rays; 121 with smoking questionnaires.
Annual radiographs taken since 1964; most recent
radiograph evaluated. Mean employment duration:
14 years. Mean fiber-years: 123 (all workers),
119 (workers with radiographs).
Pleura! thickening of the chest wall observed
in 13%.
Small opacities (>1/0) observed in 10%.
Both abnormalities increased with
increasing cumulative exposure.
Whitehouse
(2004)
n = 123 (86 former employees of W.R. Grace & Co.,
27 family members of employees, and 10 Libby
residents with only environmental exposures).
Average age: 66 years; 80% males. Fifty-six patients
had interstitial abnormalities at profusion category 0/1
or 1/0. Chest X-rays and/or HRCT scans; pulmonary
function tests (FVC, TLC, and DLCO).
Average yearly loss (n = 123):
FVC 2.2%
TLC 2.3%
DLCO 3.0%
Larson et al.
(2010a)
Men with 2 or more X-rays spanning a period of 4 or
more years. Most recent X-ray read independently by
each of 3 NIOSH B-readers; each series of X-rays (for
a given participant) then read by the panel for a
consensus determination of time of first appearance of
the detectable abnormality (n = 84).
Latency (time from hire to observed
change), median (25th, 75th percentile) years:
Localized pleura! thickening
8.6(1.4, 14.7)
Any pleura! calcification
17.5(8.1,24.2)
Diffuse pleural thickening
27.0(10.7,29.8)
4
5
6
7
DLCO = single breath carbon monoxide diffusing capacity; FVC = forced vital capacity; TLC = total lung capacity,
HRCT = high resolution computed tomography.
This document is a draft for review purposes only and does not constitute Agency policy.
4-24 DRAFT—DO NOT CITE OR QUOTE
-------
1 angle obliteration..."), which included all sites as well as a second category of "pleural
2 thickening of the chest wall."
3 Amandus et al. (1987b) reported pleural thickening of the chest wall in 13% and small
4 opacities (>1/0) in 9.1% of current employees. Similar data were reported by McDonald et al.
5 (1986b), with 15.9 and 10% with pleural thickening of the chest wall and small opacities,
6 respectively. In both studies, prevalence of these abnormalities increased with increasing
7 cumulative exposure. McDonald et al. (1986b) also included 80 former employees in their
8 study. The prevalence of pleural thickening of the chest wall (52.5%) and small opacities
9 (37.5%) was higher in these workers compared with current workers. These groups differed by
10 age, however, with only one of the 80 former workers < age 40 years compared with 80 of
11 164 current workers. Within the age category 40 to 59 years, the prevalences of chest wall
12 pleural thickening were 20.3 and 40.0% in current and former employees, respectively, and, in
13 the >60-years age group, the prevalences were 40.0 and 61.2%, respectively. The authors
14 attribute these differences in prevalence rates in current compared with former employees to
15 differences in cumulative exposure. Among the 47 area residents without known dust exposure
16 in an occupational setting in the study by McDonald et al. (1986b), the prevalence of pleural
17 thickening was 8.5% (n = 4), and the prevalence of small opacities was 2.1% (n = 1).
18 Both Amandus et al. (1987b) and McDonald et al. (1986b) provided categorical
19 exposure-response data as well as logistic models for various endpoints (e.g., small opacities,
20 pleural calcification, pleural thickening of the chest wall, and "any pleural change"). In
21 McDonald et al. (1986b), exposure and age were both predictive of pleural thickening along the
22 chest wall, and the regression coefficient for cumulative exposure (fibers-years/cc) was
23 0.0024 per unit increase in cumulative exposure for the log odds of the presence of pleural
24 thickening, adjusting for age and smoking. Exposure, age, and smoking status were all
25 predictive of small opacities, with a beta of 0.0035 per unit increase in cumulative exposure. In
26 contrast, although categorical analysis reported by Amandus et al. (1987b) indicated a positive
27 exposure response relationship for both "any pleural change" and pleural thickening along the
28 chest wall, exposure was not a significant predictor in regression analysis controlling for age
29 (regardless of smoking status). The estimated relationship between exposure and prevalence of
30 small opacities in Amandus et al. (1987b) was similar to that reported by McDonald et al.
31 Q986b).
This document is a draft for review purposes only and does not constitute Agency policy.
4-25 DRAFT—DO NOT CITE OR QUOTE
-------
1 Whitehouse (2004) examined changes in pulmonary function measures in 123 patients
2 seen in a pulmonary disease practice serving the Libby, MT area, with a mean follow-up time of
3 35 months. This study population included 86 former employees of W.R. Grace & Co.,
4 27 family members of employees, and 10 Libby residents with only environmental (i.e.,
5 nonoccupational, nonfamily-related) exposures. The average age at the time of the first
6 pulmonary study was 66 years, and 80% were male. Chest X-rays or high resolution computed
7 tomography scans revealed no evidence of interstitial changes in 67 (55%) of the 123 patients,
8 and 56 patients (45%) were found to have interstitial changes at profusion category 0/1 or 1/0.
9 Pulmonary function tests included forced vital capacity (FVC), total lung capacity (TLC), and
10 the single breath carbon monoxide diffusing capacity (DLCO). The average yearly loss was
11 2.2% for FVC, 2.3% for TLC, and 3.0% for DLCO. The subset of 94 patients who experienced a
12 loss of FVC was characterized as the group with worsening lung function. Among this group,
13 the average yearly loss was 3.2% for FVC, 2.3% for TLC, and 3.3% for DLCO.
14 Larson et al. (2010a) analyzed data from a subset of workers for whom pleural and/or
15 parenchymal abnormalities were seen on the most recently available X-ray and who had one or
16 more previous X-rays covering a span of at least 4 years available for comparison. Three
17 NIOSH B-readers independently reviewed the most recent of the available X-rays for each
18 individual in the study using ILO criteria (ILO, 2002). If pleural or parenchymal abnormalities
19 consistent with asbestos exposure were seen by each of the readers, the full series of X-rays for
20 that participant was evaluated to identify the time at which changes were first seen. For this set
21 of analyses, the readers worked as a consensus panel, examining each of the available X-rays in
22 reverse chronological order to determine the latency (i.e., length of time between first exposure,
23 as measured by date of hire and observed abnormality), and the degree of progression by type of
24 abnormality. Stored X-rays were found for 184 workers, and 84 were included in the analysis.
25 Exclusions were based on the following: 76 did not have at least two X-rays over the span of at
26 least 4 years, 20 declined to participate, unanimous classification of the most recent X-ray was
27 not reached for 3, and 1 worker did not have any detectable abnormality. Localized pleural
28 thickening was seen in 83 of these 84 workers who were known to have had pleural and/or
29 parenchymal abnormalities at a median latency of 8.6 years. Any pleural calcification was seen
30 in 37 workers, with a median latency of 17.5 years, and diffuse pleural thickening was seen in
31 12 workers (median latency: 27.0 years). The latency period increased with increasing profusion
This document is a draft for review purposes only and does not constitute Agency policy.
4-26 DRAFT—DO NOT CITE OR QUOTE
-------
1 categories, from a median of 18.9 years for >1/0, 33.3 years for progression to >2/l, and
2 36.9 years for progression to >3/2.
3
4 4.1.1.4.3. Cardiovascular-related mortality
5 Larson et al. (201 Ob) presents data on mortality due to cardiovascular diseases, with
6 SMRs of 0.9 (95% CI: 0.9, 1.0) seen for heart disease (n = 552) and 1.4 (95% CI: 1.2, 1.6) seen
7 for circulatory system diseases (n = 258). Deaths due to heart diseases were further categorized
8 into ischemic heart disease (n = 247) and other heart disease (n = 120, for pericarditis,
9 endocarditis, heart failure, and ill-defined descriptions and complications of heart disease), with
10 SMRs of 0.7 (95% CI: 0.6, 0.8) and 1.5 (95% 1.2, 1.8), respectively. Circulatory diseases
11 included hypertension without heart disease (n = 42), with an SMR of 1.7 (95% CI: 1.2, 2.4) and
12 diseases of arteries, veins, or lymphatic vessels (n = 136), SMR =1.6 (95% CI: 1.4, 2.0). The
13 combined category of cardiovascular-related mortality resulted in modestly increased risks
14 across quartiles of exposure, with RR of 1.0 (referent), 1.3 (95% CI: 1.0, 1.6), 1.3 (95% CI: 1.0,
15 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
16 >44.0 fibers/cc-years, respectively. Larson et al. (201 Ob) used a Monte Carlo simulation to
17 estimate the potential bias in cardiovascular disease risk that could have been introduced by
18 differences in smoking patterns between exposed and unexposed workers in the cohort. The
19 bias-adjustment factor (RRunadjusted/RRadjusted =1-1) reduced the overall RR estimate from 1.6 to
20 1.5. Because Larson et al. (201 Ob) analyzed multiple causes of death, the observed association
21 between exposure and cardiovascular disease-related mortality may reflect, at least in part, a
22 consequence of an underlying respiratory disease.
23
24 4.1.1.4.4. Summary ofnoncancer risk in Libby, MT vermiculite mining operation workers
25 The risk of mortality related to asbestosis and other forms of nonmalignant respiratory
26 disease is elevated in the Libby vermiculite mining and processing operations, with increasing
27 risk seen with increasing exposure to Libby Amphibole asbestos fibers in studies conducted in
28 the 1980s (Amandus and Wheeler, 1987; McDonald et al., 1986a) and in the extended follow-up
29 studies published in more recent years (Larson et al., 201 Ob: Sullivan, 2007; McDonald et al.,
30 2004). The analyses using an internal referent group in the larger follow-up studies (Larson et
This document is a draft for review purposes only and does not constitute Agency policy.
4-27 DRAFT—DO NOT CITE OR QUOTE
-------
1 al., 2010b: Sullivan, 2007; McDonald et al., 2004)16 observed increasing risks with increasing
2 cumulative exposure exposures when analyzed using tertiles or quartiles, or as a continuous
3 measure. Increased risks are also seen in the studies reporting analyses using an external referent
4 group, i.e., standardized mortality ratios (Sullivan, 2007; Amandus and Wheeler, 1987;
5 McDonald et al., 1986a). Radiographic evidence of small opacities (evidence of parenchymal
6 damage) and pleural thickening (both discrete and diffuse) has also been shown in studies of
7 Libby workers (Larson et al., 2010a: Whitehouse, 2004; Amandus et al., 1987b: McDonald et al.,
8 1986b).
9
10 4.1.2. Libby, MT Community Studies
11 In addition to worker exposures, the operations of the Zonolite Mountain mine are
12 believed to have resulted in both home exposures and community exposures. Potential pathways
13 of exposure (discussed below) range from release of airborne fibers into the community,
14 take-home exposure from mine workers (e.g., clothing), and recreational activities including
15 gardening and childhood play activities. Due to a potential for a broader community concern,
16 ATSDR conducted several studies and health actions responding to potential asbestos
17 contamination in the Libby, MT area.
18
19 4.1.2.1. Geographic Mortality Analysis
20 ATSDR conducted a location-specific analysis of mortality risks and a community health
21 screening for asbestos in the Libby area (see Table 4-8). The mortality analysis was based on
22 death certificate data from 1979-1998, with geocoding of current residence at time of death. The
23 six geographic areas used in the analysis were defined as the Libby city limits (1.1 square miles
24 around the downtown); the extended boundary of Libby (2.2 square miles around the
25 downtown); the boundary based on air modeling (16 square miles, based on computer modeling
26 of asbestos fiber distribution); the medical screening boundary (25 square miles, including the
27 town of Libby and areas along the Kootenai River); the Libby valley (65 square miles); and
28 central Lincoln County (314 square miles, based on a 10-mile radius around downtown Libby)
29 (ATSDR. 2000).
16See also reanalysis of Sullivan (2007) data by Moolgavar et al. (2010).
This document is a draft for review purposes only and does not constitute Agency policy.
4-28 DRAFT—DO NOT CITE OR QUOTE
-------
1 The 1990 population estimates were 2,531, 3,694, 4,300, 6,072, 8,617, and 9,512,
2 respectively, for these six areas. Age-standardized SMRs were calculated using underlying
3 cause-of-death information obtained from death certificates issued during the study period for
4 413 of 419 identified decedents, and Montana and U.S. populations were used as reference
5 groups. Increased SMRs were observed for both asbestosis and pulmonary circulation diseases
6 (see Table 4-8). The SMR for lung cancer ranged from 0.9-1.1 and 0.8-1.0 in the analyses for
7 each of the six geographic boundaries using Montana and U.S. reference rates, respectively. In
8 addition, four deaths due to mesothelioma were observed during the study period. These
9 analyses did not distinguish between deaths among workers and deaths among other community
10 members.
11
12 4.1.2.2. Community Screening—Respiratory Health
13 The ATSDR community health screening was conducted from July-November 2000 and
14 July-September 2001 with 7,307 total participants (ATSDR. 200Ib) (see Table 4-9). Eligibility
15 was based on residence, work, or other presence in Libby for at least 6 months before 1991. The
16 total population eligible for screening is not known; the population of Libby, MT in 2000 was
17 approximately 10,000. In addition to a standardized interview regarding medical history,
18 symptoms, work history, and other potential exposures, clinical tests included spirometry (forced
19 expiratory volume in one second [FEV1] and FVC) and chest X-rays (for participants aged
20 18 years and older). Moderate to severe restriction (defined by the researchers as FVC <70%
21 predicted value) was observed in 2.2% of the men and 1.6% of women but was not observed in
22 individuals less than age 18.
23 Two board-certified radiologists (B readers) examined each radiograph, and a third reader
24 was used in cases of disagreement. Readers were aware that the radiographs were from
25 participants in the Libby, MT health screening but were not made aware of exposure histories
26 and other characteristics (Peipins et al., 2004a: Price, 2004; Peipins et al., 2003). The
27 radiographs revealed pleural abnormalities in 17.9% of participants, with prevalence increasing
28 with increasing number of "exposure pathways" (defined on the basis of potential work and
29 residential exposure to asbestos within Libby and from other sources) (see Table 4-9). Detailed
30 results of an analysis excluding the former Libby workers cohort were not presented, but the
31 authors noted that the relationship between number of exposure pathways and increasing
This document is a draft for review purposes only and does not constitute Agency policy.
4-29 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-8. Cancer mortality and nonmalignant respiratory disease mortality
in the Libby, MT community
Reference(s)
Inclusion criteria and design details
Results
1979-1998, underlying cause of death
from death certificates; 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). Inquiries to
postmaster were required because of
P.O. Box address for 8% (n = 32);
information on 47 of 91 residents of
elderly care facilities resulted in
reclassification of 16 of 47 (34%) to
nonresidents of Libby.
U.S. Census data corresponding to the
same six geographic boundaries of
Libby, MT.
419 decedents identified, 418 death
certificates obtained, 413 with
geocoding.
Age-standardized SMRs based on
Montana and U.S. comparison rates.
Asbestosis SMRs were somewhat
higher using the U.S. referent group,
but choice of referent group had little
difference on SMRs for most diseases.
Four deaths from mesothelioma
observed in the study area.
Lung cancer (n = 82) SMR (95% CI)
Comparison area (Montana reference rates):
Libby city limits
Extended Libby boundary
Air modeling
Medical screening
Libby valley
Central Lincoln County
Pancreatic cancer (n = 10)
1.1 (0.8, 1.5)
1.1 (0.8, 1.5)
1.0 (0.8, 1.4)
0.9 (0.7, 1.2)
0.9 (0.7, 1.2)
0.9 (0.7, 1.1)
SMR (95% CI)
Comparison area (Montana reference rates):
Libby city limits
Extended Libby boundary
Air modeling
Medical screening
Libby valley
Central Lincoln County
Asbestosis (n = 11)
1.0 (0.5,2.1)
0.9 (0.4, 1.7)
0.7 (0.3,
0.7 (0.3,
0.6 (0.3,
0.5 (0.3,
1.4)
1.2)
1.0)
1.0)
SMR (95% CI)
Comparison area (Montana reference rates):
Libby city limits
Extended Libby boundary
Air modeling
Medical screening
Libby valley
Central Lincoln County
40.8 (13.2,95.3)
47.3 (18.9,97.5)
44.3 (19.1,87.2)
40.6 (18.5, 77.1)
38.7 (19.3,69.2)
36.3 (18.1,64.9)
Comparison area (U.S. reference rates):
Libby city limits
Extended Libby boundary
Air modeling
Medical screening
Libby valley
Central Lincoln County
Pulmonary circulation (n = 14) SMR (95% CI)
Comparison area (Montana 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)
Libby city limits
Extended Libby boundary
Air modeling
Medical screening
Libby valley
Central Lincoln County
2.3 (1.1,4.4)
1.9 (0.9,3.7)
1.8 (0.9,3.3)
1.6 (0.8,2.9)
1.6 (0.9,2.7)
1.5 (0.8,2.5)
4
5
This document is a draft for review purposes only and does not constitute Agency policy.
4-30 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-9. Pulmonary function and chest radiographic studies in the Libby,
MT community
Reference(s)
Inclusion criteria and design details
Results
Peipins et al.
(20031: ATSDR
(200 Ib)
Resided, worked, attended school, or participated in other
activities in Libby for at least 6 months before 1991
(including mine employees and contractors).
Health screening between July and November 2000.
Conducted interviews (n = 6,149, 60% of Libby residents
based on 2000 Census data) and chest X-rays (n = 5,590,
18 years and older), and determined spirometry—forced
expiratory volume in 1 second (FEV1), forced vital
capacity (FVC1), and ratio (FEV1/FVC).
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. Chest X-rays read by 1980
ILO classifications (3 views; posterior-anterior, right- and
left- anterior oblique). Peipins et al. (2003) similar to
(ATSDR. 200 Ib) except longer screening period
(July-November 2000 and July-September 2001).
Conducted interviews (n = 7,307) and chest X-rays
(n = 6,668).
Peipins (2003) and ATSDR (200Ib):
Pleura! abnormalities seen in 17.9%
of participants; increasing prevalence
with increasing number of exposure
pathways (6.7% among those with no
specific pathways, 34.6% among
those with 12 or more pathways).
ATSDR (200 Ib):
Moderate-to-severe FVC1 restriction
(FVC <70% predicted): 2.2% of men
>17 years old; 1.6% of women
>17 years old; 0.0% of men or
women <18 years old.
Also includes data on self-reported
lung diseases and symptoms.
Weilletal.
(2011)
Participants in the ATSDR community health screening
(see first row in table). Analysis limited to ages 25 to 90
years, excluding individuals with history of other asbestos-
related work exposures, with spirometry, consensus
reading of chest X-ray, smoking data, and exposure
pathway data (n = 4,397). Analysis based on five
exposure categories: (1) W.R. Grace worker, (2) other
vermiculite worker (contractor work), (3) other dusty
occupation, (4) household (combination of three household
categories), and (5) environmental ("no" to work and
household exposures in Categories 1-6). Chest X-rays
read by 1980 ILO classifications (frontal view).
Profusion DPT/
>1/0 Plaque CAO
Prevalence (%), ages 25 to 40 years:
1) W.R. Grace
2) Other
3) Dusty
4) Household
0.0
0.8
0.0
0.0
5) Environment 0.0
20.0
0.8
3.8
2.2
0.4
5.0
0.0
0.4
0.0
0.0
Prevalence (%), ages 41 to 50 years:
1) W.R. Grace
2) Other
3) Dusty
4) Household
0.0
0.5
0.0
0.0
5) Environment 0.0
26.2
7.8
2.8
11.1
1.9
5.0
1.0
0.9
0.4
0.2
Prevalence (%), ages 51 to 60 years:
1) W.R. Grace 3.2 34.9 3.2
2) Other 0.6 13.7 0.6
3) Dusty 0.6 12.6 0.0
4) Household 1.0 20.1 1.5
5) Environment 0.0 7.7 0.9
Prevalence (%), ages 61 to 90 years:
1)W.R. Grace 11.1 45.7 8.6
2) Other 0.6 24.8 8.5
3) Dusty 1.1 21.9 3.3
4) Household 2.4 38.3 5.7
5) Environment 1.3 12.7 2.2
This document is a draft for review purposes only and does not constitute Agency policy.
4-31 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-9. Pulmonary function and chest radiographic studies in the Libby,
MT community (continued)
Reference(s)
Inclusion criteria and design details
Results
Vinikoor et al.
(20101
Participants in the ATSDR community health screening
(see first row in table). Analysis limited to n = 1,003 ages
10-29 years at time of health screening (3
activities). Analysis of history of respiratory symptoms
and spirometry data (obstructive, restrictive, or mixed).
Little difference across exposure
levels in prevalence of
physician-diagnosed lung disease or
abnormal spirometry.
Odds Ratio (95% CI) seen between
>3 activities and
Usual cough 2.93 (0.93, 9.25)
Shortness of breath 1.32 (0.51, 3.42)
Bloody phlegm 1.49 (0.41, 5.43)
1
2
o
6
4
5
6
7
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
OR = odds ratio; DPT = diffuse pleural thickening; CAO = costophrenic angle obliteration.
prevalence of pleural abnormalities was somewhat attenuated with this exclusion. The
prevalence of pleural anomalies 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 definable exposure pathways, the prevalence of pleural anomalies was 6.7%,
which 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
underascertainment 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.
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.
From the 6,668 participants with chest X-rays, 1,327 individuals with a history of
asbestos-related work (other than with the Grace mining or related vermiculite operations) were
excluded, along with 817 excluded based on age (<25 or >90 years) or lack of spirometric data,
smoking data, or exposure pathway data. An additional 127 were excluded because a consensus
agreement (2 out of 3 readers) was not reached regarding the X-ray findings, leaving n = 4,397 in
the analysis. Analysis was based on five exposure categories: (1) Grace worker (n = 255),
(2) other vermiculite worker (e.g., secondary contractor worker for Grace or other jobs with
vermiculite exposure (n = 664), (3) other dusty occupation (e.g., plumber, dry wall finisher,
This document is a draft for review purposes only and does not constitute Agency policy.
4-32 DRAFT—DO NOT CITE OR QUOTE
-------
1 carpenter, roofer, electrician, welder, shipyard work or ship construction or repair (n = 831),
2 (4) household, including household with other vermiculite or dusty work (lived with a Grace
3 worker combination of three household categories) (n = 880), and (5) environmental ("no" to
4 work and household exposures in Categories 1-4) (n = 1,894). The frontal views (posterior-
5 anterior) of the chest X-rays were used in this analysis [in contrast to the use of frontal and
6 oblique views in Peipins et al. (2003)]. As expected, lung function (FEVi, FVC, and FEVi/FVC)
7 was lower among ever smokers compared with never smokers (within each age group) and
8 decreased with age (within each smoking category). The prevalence of X-ray abnormalities
9 (plaques, or diffuse pleural thickening, and/or costophrenic angle obliteration) also generally
10 increased with age (divided into 25-40, 41-50, 51-60, and 61-90 years) within each of the
11 exposure categories (see Table 4-9), with the highest prevalence seen among Grace workers. For
12 a given age, the prevalence among those with environmental exposure only (i.e., no household or
13 occupational exposures) was similar to the prevalence among those with non-Grace occupational
14 or household exposures in the next youngest age category. The prevalence among the household
15 contact category was similar or higher than the prevalence among the other vermiculite and dusty
16 job categories. This household contact category includes individuals who lived with a Grace
17 worker with no personal history of vermiculite or dust work (n = 594) and those who also had a
18 history of other vermiculite (n = 114) or dusty (n = 172) jobs. The authors noted the prevalence
19 rates were similar among these groups, and so the analysis was based on the combination of
20 these three groups. Mean FVCs (±SE) percentage predicted were 78.76 (±3.64), 82.16 (±3.34),
21 95.63 (±0.76), and 103.15 (±0.25), respectively, in those with diffuse pleural thickening and/or
22 costophrenic angle obliteration, profusion >1/0, other pleural abnormalities, and no pleural
23 abnormalties. The strongest effects of diffuse pleural thickening and/or costophrenic angle
24 obliteration on FVC were seen among men who had never smoked (-23.77,/? < 0.05), with
25 smaller effects seen among men who had smoked (-9.77, p < 0.05) and women who had smoked
26 (-6.73,p< 0.05).
27 Vinikoor et al. (2010) used the 2000-2001 health screening data to examine respiratory
28 symptoms and spirometry results among 1,224 adolescents and young adults who were 18 years
29 or younger in 1990 when the mining/milling operations closed. At the time of the health
30 screening, the ages in this group ranged from 10 to 29 years. Exclusion criteria for this analysis
31 included previous work for W.R. Grace, work for a contractor of W.R. Grace, exposure to dust at
This document is a draft for review purposes only and does not constitute Agency policy.
4-33 DRAFT—DO NOT CITE OR QUOTE
-------
1 other jobs, or exposure to vermiculite at other jobs. The total number of exclusions was 221,
2 leaving 1,003 in the analysis. The potential for vermiculite exposure was classified based on
3 responses to questions about six activities (handling vermiculite insulation, participation in
4 recreational activities along the vermiculite-contaminated gravel road leading to the mine,
5 playing at the ball fields near the expansion plant, playing in or around the vermiculite piles,
6 heating the vermiculite to "pop" it, and other activities involving vermiculite). The medical
7 history questionnaire included information on three respiratory symptoms: usually have a cough
8 (n= 108, 10.8%); troubled by shortness of breath when walking up a slight hill or when hurrying
9 on level ground (n = 145, 14.5%); coughed up phlegm that was bloody in the past year
10 (n = 59, 5.9%). A question on history of physician-diagnosed lung disease (n = 51, 5.1%) was
11 also included. The spirometry results were classified as normal in 896 (90.5%), obstructive in
12 62 (6.3%), restrictive in 30 (3.0%), and mixed in 2 (0.2%). Information on smoking history was
13 also collected in the questionnaire: 15.8% and 7.3% were classified as current and former
14 smokers, respectively. Approximately half of the participants lived with someone who smoked.
15 The analyses adjusted for age, sex, personal smoking history, and living with a smoker. For
16 usually having a cough, the odds ratios (ORs) were 1.0 (referent), 1.88 (95% CI: 0.71, 5.00),
17 2.00 (95% CI: 0.76, 5.28) and 2.93 (95% CI: 0.93, 9.25) for never, sometimes, frequently
18 participated in 1-2 activities, and frequently participated in >3 activities, respectively. For
19 shortness of breath, the corresponding ORs across those exposure categories were 1.0 (referent),
20 1.16 (95% CI: 0.55, 2.44), 1.27 (95% CI: 0.61, 2.63) and 1.32 (95% CI: 0.51, 3.42), and for
21 presence of bloody phlegm in the past year the ORs were 1.0 (referent), 0.85 (95% CI: 0.31,
22 2.38), 1.09 (0.41, 2.98), and 1.49 (95% CI: 0.41, 5.43). For history of physician-diagnosed lung
23 disease and abnormal spirometry results, there was little difference in the odds ratios across the
24 exposure categories: for lung disease, the ORs were 1.0 (referent), 1.95 (95% CI: 0.57, 6.71),
25 1.51 (95% CI: 0.43, 5.24) and 1.72 (95% CI: 0.36, 8.32) for the categories of never, sometimes,
26 frequently participated in 1-2 activities, and frequently participated in >3 activities, respectively.
27 For abnormal spirometry (i.e., obstructive, restrictive, or mixed, n = 94 cases), the ORs were
28 1.0 (referent), 1.34 (95% CI: 0.60, 2.96), 1.20 (95% CI: 0.53, 2.70) and 1.33 (95% CI: 0.42,
29 4.19) across these exposure groups.
30 Two other studies examining autoimmune disease and autoantibodies in residents of
31 Libby, Montana are described in Section 4.3.
This document is a draft for review purposes only and does not constitute Agency policy.
4-34 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.1.2.3. Other Reports of Asbestos-Related Disease Among Libby, MT Residents
1 Whitehouse et al. (2008) recently reviewed 11 cases of mesothelioma diagnosed between
3 1993 and 2006 in residents in or around Libby, MT (n = 9) and in family members of workers in
4 the mining operations (n = 2). Three cases were men who might have had occupational asbestos
5 exposure through construction work (Case 1), working in the U.S. Coast Guard and as a
6 carpenter (Case 5), or through railroad work involving sealing railcars in Libby (Case 7). One
7 case was a woman whose father had worked at the mine for 2 years; although the family lived
8 100 miles east of Libby, her exposure may have come through her work doing the family
9 laundry, which included laundering her father's work clothes. The other seven cases
10 (four women, three men) had lived or worked in Libby for 6-54 years, and had no known
11 occupational or family-related exposure to asbestos. Medical records were obtained for all
12 11 patients; pathology reports were obtained for 10 of the 11 patients. The Centers for Disease
13 Control estimated the death rate from mesothelioma, using 1999 to 2005 data, as approximately
14 14 per million per year (CDC, 2009), approximately five times higher than the rate estimated by
15 Whitehouse et al. (2008) for the Libby area population based on the estimated population of
16 9,500 for Lincoln County and 15 years (or 150,000 person-years) covered by the analysis.
17 Whitehouse et al. (2008) stated that a W.R. Grace unpublished report of measures taken in 1975
18 indicated that exposure levels of 1.1 fibers/cc were found in Libby, and 1.5 fibers/cc were found
19 near the mill and railroad facilities. Because the mining and milling operations continued to
20 1990, and because of the expected latency period for mesothelioma, Whitehouse et al. (2008)
21 suggests that additional cases can be expected to occur within this population.
22
23 4.1.2.4. Summary of Respiratory Health Effects in Libby, MT Community Studies
24 The geographic-based mortality analysis of 1997-1998 mortality data indicates that
25 asbestosis-related mortality is substantially increased in Libby, MT, and the surrounding area,
26 with rates 40 times higher compared with Montana rates and 60-70 times higher compared with
27 U.S. rates (ATSDR, 2000). These data provide evidence of the disease burden within the
28 community; however, because this analysis did not distinguish between deaths among workers
29 and deaths among other community members, it is not possible based on these data to estimate
30 the risk of asbestos-related mortality experienced by residents who were not employed at the
31 mining or milling operations. The community health screening studies provide more detailed
This document is a draft for review purposes only and does not constitute Agency policy.
4-3 5 DRAFT—DO NOT CITE OR QUOTE
-------
1 information regarding exposure pathways in addition to occupation (ATSDR, 200Ib). Data from
2 the ATSDR community health screening study indicate that the prevalence of pleural
3 abnormalities, identified by radiographic examination, increases substantially with increasing
4 number of exposure pathways (Peipins et al., 2003). In addition, the prevalence of some
5 self-reported respiratory symptoms among 10 to 29-year-old adolescents and young adults was
6 associated with certain exposure pathways. These participants were < age 18 in 1990 when the
7 mining/milling operations closed ("Vinikoor et al., 2010). A better understanding of the
8 community health effects and the examination of the potential progression of adverse health
9 effect in this community would benefit from additional research to establish the clinical
10 significance of these findings. The observation by Whitehouse et al. (2008) of cases of
11 mesothelioma among individuals with no direct occupational exposure to the mining and milling
12 operations indicates the need for continued surveillance for this rare cancer.
13
14 4.1.3. Marysville, OH Vermiculite Processing Plant Worker Studies
15 Libby vermiculite was used in the production of numerous commercial products,
16 including as a potting soil amender and a carrier for pesticides and herbicides. A Marysville, OH
17 plant that used Libby vermiculite in the production of fertilizer beginning around 1960 to 1980 is
18 the location of the two related studies described in this section.
19 The processing facility had eight main departments, employing approximately
20 530 workers, with 232 employed in production and packaging of the fertilizer and 99 in
21 maintenance; other divisions included research, the front office, and the polyform plant (Lockey,
22 1985). Six departments were located at the main facility (trionizing, packaging, warehouse,
23 plant maintenance, central maintenance, and front offices). Research and development and a
24 polyform fertilizer plant were located separately, approximately one-quarter mile from the main
25 facility. In the trionizing section of the plant, the vermiculite ore was received by rail or truck,
26 unloaded into a hopper, and transported to the expansion furnaces. After expansion, the
27 vermiculite was blended with other materials (e.g., urea, potash, herbicides), packaged, and
28 stored. Changes to the expander type and dust-control measures began in 1967, with substantial
29 improvement in dust control occurring throughout the 1970s.
30 Information about exposure assessment at the Marysville, OH plant is summarized in the
31 final row of Table 4-1. Industrial hygiene monitoring at the plant began in 1972. Lockey et al.
This document is a draft for review purposes only and does not constitute Agency policy.
4-36 DRAFT—DO NOT CITE OR QUOTE
-------
1 (1984) noted that the limited availability of data that would allow for extrapolation of exposures
2 for earlier time periods possibly resulted in the underestimation of exposures before 1974.17
3 Task-level air samples were conducted, and measurements were determined using scanning
4 electron microscopy and transmission electron microscopy (based on particles >5-um-long,
5 <3-um-diameter, and >3:1 aspect ratio).
6 Based on measurements and knowledge of plant operations, three categories of exposure
7 levels were defined. Group I was considered to be the nonexposed group and consisted of the
8 chemical processing, research, and front office workers. The chemical process plant was about a
9 quarter mile from the main vermiculite facility, but the same chemicals were used in both
10 locations. The 8-hour time-weighted average vermiculite exposure in this group, both before and
11 after 1974, was estimated as 0.049 fiber/cc (based on a single stationary sample taken outside the
12 main facility), which was characterized as similar to the background levels in the community.
13 Group II was the "low exposure" category and included central maintenance, packing, and
14 warehouse workers. The 8-hour time-weighted average vermiculite exposures in this group were
15 estimated as approximately 0.1-0.4 fibers/cc before 1974 and 0.03-0.13 fibers/cc in and after
16 1974. Group III was the "highest exposure" category, and included vermiculite expanders, plant
17 maintenance, and pilot plant workers. The 8-hour time-weighted average vermiculite exposures
18 in this group were approximately 1.2-1.5 fibers/cc before 1974 and 0.2-0.375 fibers/cc in and
19 after 1974. Cumulative fiber exposure indexes, expressed as fibers-year/cc, were derived for
20 each worker from available industrial hygiene data and individual work histories. Those with
21 less than 1 fiber/cc-year were assumed to be equivalent to a community population (in terms of
22 exposure) and were used as the comparison group. The estimated cumulative exposure for the
23 work force, including Group I workers, ranged from 0.01 to 28.1 fibers/cc-years using an 8-hour
24 workday and an assumed 365 days of exposure per year.18 Exposure was assumed to occur from
25 1957 to 1980 in this study. Exposure after work hours was assumed to be zero.
26 The first study of pulmonary effects in the Ohio plant workers was conducted in 1980
27 and involved 512 workers (97% of the 530 workers previously identified with past vermiculite
28 exposure) (Lockey etal., 1984) (see Table 4-10). Physical examination (for detection of
"Subsequent exposure assessment efforts by this team of investigators are described in Appendix F.
18Lockey et al. (1984) reported the maximum value for this group as 39.9 fibers/cc-years, but this estimate was later
corrected to exclude work from 1947 to 1956, prior to the use of vermiculite at the plant. Information provided in
personal communication from J. Lockey to Robert Benson, U.S. EPA, June 7, 2011.
This document is a draft for review purposes only and does not constitute Agency policy.
4-37 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-10. Pulmonary function and chest radiographic studies of the
Marysville, OH vermiculite processing plant workers
Reference(s)
Inclusion criteria and design details
Results
Lockey et al.
(1984): Lockey
(1985)a
1980, n = 512 (from 530 identified employees with
past vermiculite exposure; nonparticipants included
9 refusals and 9 unavailable due to illness or
vacation).
Smoking history, work history at the plant, and
other asbestos and fiber mineral work history data
were collected.
Chest exam (rales), nail clubbing, spirometry,
forced vital capacity, forced expiratory volume,
single-breath carbon monoxide diffusing capacity,
and chest X-rays (available for 502 participants)
were analyzed.
Mean employment duration: 10.2 yearsb
Three exposure groups, based on jobs and area:
Mean cumulative exposure13
Group I 0.45 fibers/cc-years
Group II 1.13 fibers/cc-years
Group III 6.16 fibers/cc -years
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.
Pleural thickening in 10 workers (2%);
bilateral, small opacities in 1 (0.2%).
Abnormality (combined outcomes)
increased with increasing cumulative
exposure.
Rohs et al.
2002-2005, interviews and chest X-rays conducted,
n = 298; 280 with interviews and readable chest
X-rays (from 431 workers in the 1980 study group,
of which, 513 were alive in 2004°; 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).
Age, smoking, asbestos exposure measure (at this
plant), and other asbestos exposure data used to
compare participants and nonparticipants.
Libby, MT vermiculite ore used in the plant from
1963-1980.
Pleural abnormalities in 80 workers
(28.7%).
Small opacities (>1/0) in 8 workers
(2.9%).
Increasing risk of pleura! abnormalities
with increasing cumulative fiber
exposure: odds ratios (adjusting for date
of hire, body mass index) by exposure
quartile were 1.0 (referent), 2.7, 3.5, and
6.9.
4
5
6
7
8
9
10
11
12
13
14
15
16
17
aLockey et al. (1984) is the published paper based on the unpublished thesis (Lockev. 1985).
bCalculated based on stratified data presented in Table 2 of Lockey et al. (1984).
°Rohs et al. (2008) identified one additional eligible worker from the original 512 employees identified in Lockey
et al. (1984).
pulmonary rales and nail clubbing), spirometry, 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. Radiographs were read independently by
two board-certified radiologists (B-readers), with a reading by a third reader when the initial
two readings did not agree. The number of workers within each exposure group was 112, 206,
and 194 in Groups I, II, and III, respectively. Approximately 44% were current smokers,
20% former smokers, and 35% lifetime nonsmokers, but smoking history (i.e., smoking status,
This document is a draft for review purposes only and does not constitute Agency policy.
4-3 8 DRAFT—DO NOT CITE OR QUOTE
-------
1 pack-years) did not differ by exposure group. Mean cumulative fiber estimates were 0.45, 1.13,
2 and 6.16 fibers/cc-years in Groups I, II, and III, respectively. An increased risk of costophrenic
3 angle blunting (n= 11), pleural, and parenchymal abnormalities (n = 11), or any of these
4 outcomes (n = 22) was observed in Group III compared with Group 1; the prevalence of any
5 radiographic change was 2.8% in Group I, 3.9% in Group II, and 5.8% in Group III. Using the
6 cumulative fiber metric, the prevalence of any radiographic change was 2.4% in the
7 <1 fiber/cc-year, 5.0% in 1-10 fibers/cc-year, and 12.5% in the >10 fibers/cc-year groups.
8 A follow-up study of this cohort was conducted in 2002-2005 (Rohs et al.. 2008) (see
9 Table 4-10). This study included 298 workers, of which 280 completed the study interview and
10 chest X-ray. Details of the reasons for nonparticipation rates are described in Table 4-10. The
11 evaluation of each worker included an interview to determine work and health history,
12 spirometry, pulmonary examination, and chest X-ray. The study interview included information
13 about smoking history and asbestos exposure at the Marysville, Ohio plant and other worksites.
14 Exposure was estimated using the procedure previously described using the data on fiber levels
15 (Lockey et al., 1984). Exposure was assumed to occur from 1963 to 1980 in this study,
16 assuming an 8-hour workday and 365 days of exposure per year (J. Lockey, University of
17 Cincinnati, personal communication to R. Benson, U.S. EPA, July, 2007). Each worker supplied
18 a detailed work history (start and end date for each area within the facility). The exposure
19 reconstruction resulted in a cumulative exposure estimate for each individual. The estimated
20 cumulative exposure for this follow-up study ranged from 0.01 to 19.03 fibers/cc-years
21 (mean = 2.48). The time from first exposure ranged from 23 to 47 years. Twenty-eight workers
22 reported previous occupational exposure to asbestos. Exposure outside of work was assumed to
23 be zero.
24 Three board-certified radiologists independently classified the radiographs using the ILO
25 classification system (ILO, 2002). Radiologists were blinded to all identifiers. Pleural
26 thickening (all sites) was reported as either localized pleural thickening or diffuse pleural
27 thickening. Diffuse pleural thickening of the chest wall may be reported as in-profile or face-on,
28 and is recorded on the lateral chest wall "only in the presence of and in continuity with, an
29 obliterated costophrenic angle" (ILO, 2002). Localized pleural thickening may also be viewed
30 in-profile or face-on and was described by Rohs et al. (2008) as "... (pleural) thickening with or
31 without calcification, excluding solitary costophrenic angle blunting" consistent with current
This document is a draft for review purposes only and does not constitute Agency policy.
4-39 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
o
J
4
5
6
7
8
9
10
11
ILO classification. Interstitial abnormalities were considered present if the reader identified
irregular opacities of profusion 1/0 or greater (ILO, 2002). For the analysis, a chest X-ray was
defined as positive for pleural abnormality and/or interstitial abnormality when the median
classification from the three readings was consistent with such effects. Radiographs classified as
unreadable were not used. Radiographic abnormalities found in the study population are
summarized in Tables 4-11 and 4-12.
Table 4-11. Prevalence of pleural radiographic abnormalities according to
quartiles of cumulative fiber exposure in 280 participants
Exposure
quartile
First
Second
Third
Fourth
Total
Exposure,
fiber-yr/cc,
and (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
(%)"
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)
12
13
14
15
16
17
18
19
20
21
22
23
24
25
aTwo observations in the second quartile and two in the third quartile had exact exposure values at the 50th percentile
cutoff point. Rounding put these four observations in the second quartile.
Significant trend, p < 0.001.
Typographical error in publication corrected.
The 80 workers with pleural thickening include 68 with localized pleural thickening (85%) and 12 with diffuse
pleural thickening (15%).
Source: Rohs et al. (2008). Table 3 and Figure 2; mean exposure levels and number of workers with parenchymal
abnormalities by quartile obtained from J. Lockey, University of Cincinnati (personal communication to Robert
Benson, U.S. EPA).
This document is a draft for review purposes only and does not constitute Agency policy.
4-40 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-12. Prevalence of pleural thickening in 280 participants according to
various cofactors
Variable
Hired on or before 1973
Hired after 1973
Body Mass Index,3 kg/m2
<24.9
25-29.9
>30
Ever smokedb
Yes
No
Age at time of interview
40-49
50-59
>60
Female
Male
Number
of
workers
186
94
28
101
110
184
96
55
116
109
16
264
Number with
pleural thickening
(%)
70 (37.6)
10(10.6)
8 (28.6)
31(30.7)
27 (24.5)
55 (29.9)
25 (26.04)
5(9.1)
28(24.1)
47(43.1)
1 (6.3)
79 (29.9)
Crude OR
5.07
Reference
Reference
1.11
0.81
1.21
Reference
Reference
3.18
7.58
Reference
6.40
95% CI
2.47-10.41
0.44-2.79
0.32-2.06
0.70-2.11
1.16-8.76
2.80-20.49
0.83-49.32
p- Value
<0.001
0.52
0.43
0.50
0.03
<0.001
0.07
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
a« = 239 for Body Mass Index 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)
Pleural thickening was observed in 80 workers (28.7%), and small opacities (>1/0) were
observed in 8 (2.9%). Six of the 8 participants with small opacities also had pleural thickening
(4 as LPT, 2 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 4-11). The range of exposures was estimated as
0.01-0.28, 0.29-0.85, 0.86-2.20, and 2.21-19.03 fiber/cc-years in the first, second, third, and
fourth quartiles, respectively (Rohs et al., 2008).
Pleural thickening was associated with hire on or before 1973 and age at time of
interview but was not associated with body mass index (BMI) or smoking history (ever smoked)
(see Table 4-12). Body mass index is a potentially important confounder because fat pads can
This document is a draft for review purposes only and does not constitute Agency policy.
4-41 DRAFT—DO NOT CITE OR QUOTE
-------
1 sometime be misclassified as localized pleural thickening. A hire date of on or before 1973 and
2 ages at time of interview are each highly correlated with cumulative exposure to fibers. The
3 small number of females (n= 16) in the cohort limits the analysis of the association with sex.
4 Modeling of odds ratios with cumulative fiber exposure and including various cofactors (age,
5 hired before 1973, or BMI) with the first exposure quartile as the reference was also conducted.
6 Each model demonstrated the same trend: increased prevalence of pleural thickening with
7 increasing cumulative exposure to fibers. Adjusting for age, date of hire, and body mass index
8 resulted in odds ratios of 2.7, 3.5, and 6.9 for the second, third, and fourth quartiles, respectively.
9 Age-adjusted and BMI-adjusted results were included in Table 4-11. There was no evidence of
10 significant interactions using this modeling.
11 There was potential coexposure to a number of herbicides, pesticides, and other
12 chemicals in the facility (personal communication to Robert Benson, EPA Region 8, from Ivan
13 Smith, The Scotts Company, June 7, 2007). The herbicides and pesticides used during the time
14 when Libby ore was used included atrazine, benomyl, bensulide, chloroneb, chlorothalonyl,
15 chlorpyrifos, 2,4-D, dacthal, diazinon, dicamba, dephenamid, disodium methanearsonate, dyrene,
16 ethoprop, linuron, MCPP, monuron, neburon, oxadiazon, terrachlor, pentachlorophenol,
17 phenylmercurie acetate, siduron, terrazole, thiophannate-methyl, thiram. Other chemicals used
18 included ammonium hydroxide, brilliant green crystals, caustic soda, corncobs, ferrous
19 ammonium sulfate, ferrous sulfate, florex RVM, frit-504, frit-505, hi sil, lime, magnesium
20 sulfate, mon-a-mon, potash, potassium sulfate, sudan orange, sudan red, sulfur, sulfuric acid,
21 UFC, urea, and Victoria green liquid dye. No quantitative information on exposure to these
22 chemicals is available. However, the addition of the other chemicals to the vermiculite carrier
23 occurred in a different part of the facility after expansion of the vermiculite ore. Industrial
24 hygiene monitoring in these areas showed very low levels of fibers in the air. In addition, none
25 of these other chemicals is volatile. Thus, it is unlikely that workers would be coexposed by
26 inhalation to these other chemicals. EPA has no information indicating that exposure to any of
27 these individual chemicals causes pleural thickening or evidence of small opacities typical of
28 those found in workers employed in the Marysville facility. The spectrum of radiograph!c
29 abnormalities observed in the lung and pleura are the same in the Marysville workers, the Libby
30 workers (see Section 4.1.1.4.2, Table 4-7), and the Libby community survey (including workers)
31 (see Section 4.1.2.2, Table 4-9).
This document is a draft for review purposes only and does not constitute Agency policy.
4-42 DRAFT—DO NOT CITE OR QUOTE
-------
1 This study demonstrates that exposure to Libby Amphibole asbestos can cause
2 radiographic evidence of pleural thickening and parenchymal abnormalities (small opacities) in
3 exposed workers. The prevalences of radiographic abnormalities involving the pleura were
4 28.7% in 2004 (80/280), compared to a 2% prevalence observed in 1984 (10/501). This apparent
5 increase in prevalence is most likely due to the additional time between the two studies giving
6 additional time for the abnormalities to become apparent in conventional X-rays. The follow-up
7 study also shows an increasing prevalence of pleural thickening with increasing cumulative
8 exposure to Libby Amphibole asbestos.
9 The influence of some potential sources of selection bias in Rohs et al. (2008) is difficult
10 to qualitatively or quantitatively assess. One type of selection is the loss due to the death of
11 84 of the 513 (16%) workers in the first study; this group may represent less healthy or more
12 susceptible population. Exclusion of the very sick or susceptible may imply that the population
13 of eligible participants was somewhat healthier that the whole population of workers; this
14 exclusion may result in an underestimation of risk. Another type of selection is the loss due to
15 nonparticipation among the 431 individuals identified as alive in 2004 (n= 135 refusals and
16 nonresponders; 31%). Participation rates in epidemiologic studies can be associated with better
17 health status, and participation is often higher among nonsmokers compared with smokers. This
18 type of selection of a relatively healthier group (among the living) could also result in an
19 underascertainment of the risk of observed abnormalities within the whole exposed population.
20 However, if participation was related differentially based on exposure and outcome (i.e., if
21 workers experiencing pulmonary effects and who were more highly exposed were more likely to
22 participate than the highly exposed workers who were not experiencing pulmonary effects), the
23 result would be to overestimate the exposure response. This latter scenario is less likely to occur
24 for asymptomatic effects (i.e., abnormalities detected by chest X-ray), such as those that are the
25 focus of this study than for symptoms such as shortness of breath or chest pain.
26 Some information is available on differences by participation status in the Rohs et al.
27 (2008) study. Although current age was similar (mean: 59.1 and 59.4 years, respectively, in
28 participants and living nonparticipant groups, p = 0.53), participants were more likely to have
29 been hired before or during 1973 (66.4 and 49.7%, respectively,/? = 0.001), and had higher mean
30 exposure levels (mean cumulative exposure: 2.48 and 1.76 fiber/cc-years, respectively,/? = 0.06).
31 Participants were also somewhat less likely to be ever smokers (58.6%) compared with the living
This document is a draft for review purposes only and does not constitute Agency policy.
4-43 DRAFT—DO NOT CITE OR QUOTE
-------
1 nonparticipants (66.2%). Using a conservative assumption that all living nonparticipants would
2 have had normal X-rays, resulted in estimated prevalences of pleural abnormalities of 3.7, 13.9,
3 18.5, and 38.3%, respectively, in the lowest-to-highest exposure quartile, with corresponding
4 odds ratios of 1.0 (referent), 4.19(95% CI: 1.34, 13.08), 5.91 (95% CL1.95, 17.93), and 16.15
5 (95% CI: 5.53, 47.17). This pattern is similar to that observed in the analysis that excludes the
6 living nonparticipants, indicating the observed trend with exposure was not an artifact of a bias
7 introduced by differences in participation rates among the workers.
8
9 4.1.3.1. Summary of Marysville, OH Vermiculite Processing Plant Worker Studies
10 The studies conducted in the 1980s (Lockey etal., 1984) and the follow-up of the cohort
11 (Rohs et al., 2008) indicate that pleural thickening can be seen among workers in this plant, with
12 increasing prevalence with increasing cumulative exposure. Radiographic evidence of small
13 opacities (interstitial changes in the lung) increased from 0.2% in the original study to 2.9% and
14 radiographic evidence of pleural thickening increased from 2 to 28.6% of participants in the
15 follow-up study. No effects on lung function were found in the original study (Lockey et al.,
16 1984). Lung function was not reported for the cohort follow-up, despite greater prevalence of
17 radiographic abnormalities (Rohs et al., 2008).
18
19 4.1.4. Community Studies from Other Vermiculite Processing Plants
20 ATSDR has completed community evaluations of 28 sites, in addition to Libby,
21 surrounding exfoliation plants that require further evaluation by EPA because of current
22 contamination or evidence (based on a database of invoices) that the plant processed more than
23 100,000 tons of vermiculite from the Libby, MT mine (see Figure 4-1). Nine of these
24 evaluations included analyses conducted in conjunction with state health departments using
25 death certificate data (see Table 4-13). These community-level evaluations do not address
26 individual exposures or residential histories; therefore, the evidence in these evaluations
27 pertaining to disease risk is somewhat limited.
28
This document is a draft for review purposes only and does not constitute Agency policy.
4-44 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
o
6
4
5
6
7
Figure 4-1. Location of 28 sites included in the Phase 1 community
evaluations conducted by ATSDR.
Source: ATSDR (2008a) http://www.atsdr.cdc.gov/asbestos/sites/national_map/.
Table 4-13. Description of study areas in ATSDR health consultations
evaluating cancer incidence and mortality"
Site, exposure period
Los Angeles, CA, 1950-1977
Newark, CA, 1967-1992
Santa Ana, CA, 1972-1993
West Chicago, IL, 1974-1996
Dearborn, MI, early 1950s-1989
St. Louis, Missouri, 1956-1988
Trenton, NJ, 1920s- 1990
Edgewater, NJ, not reported
Marysville, OH, 1963-1980C
Study area (n from 1990 census)
Incidence: census tract (n = 21,945)
Mortality: zip code (n = 57,615)
Incidence: census tract (n = 7,785)
Mortality: zip code (n = 37,861)
Census tract (35,000)
Mortality: zip code (n = 14,796)
City limits (n = 89,015)
Census tracts (n = 20, 1 12)
Census tracts and areas (n = 26,762)
Not reported
City limits (n = 9,656)
Year of report
2007
2005
2003
2003
2005
2006
2005
2005
2005
9
10
11
12
13
14
15
16
17
aAll incidence studies used Surveillance, Epidemiology, and End Results (SEER) data as comparison group except
New Jersey, which used New Jersey state rates. All mortality studies used U.S. rates from the National Center for
Health Statistics.
bThe ATSDR health consultation data presented incidence data from 1979-2000, but the 1986-1995 incidence data
and the mortality data were obtained from the report of the New Jersey Department of Health and Social Services
http://www.atsdr.cdc.gov/asbestos/sites/health consultations/index. html.
°The start date for the use of the Libby, MT vermiculite was given as variously described as 1963 or 1967 in the
ATSDR health consultation report (ATSDR. 2008b): the studies by Lockey et al. (1984) and Rohs et al. (2008)
used 1957 and 1963, respectively, as the start date.
This document is a draft for review purposes only and does not constitute Agency policy.
4-45 DRAFT— DO NOT CITE OR QUOTE
-------
1 The lung cancer standardized incidence ratios for these evaluations range from
2 0.74-1.07, and the SMRs range from 0.74-1.1, indicating little evidence of an increased risk of
3 lung cancer among these studies (see Table 4-14). As expected from the small number of
4 observations, the standardized incidence ratios for mesothelioma or the category of cancer of the
5 peritoneum, retroperitoneum, and pleura (excluding mesothelioma, but which could reflect some
6 misdiagnoses) are more variable, ranging from approximately 0.5-2.5. Breast and prostate
7 cancer were selected as negative controls (i.e., cancers that have not previously been associated
8 with asbestos exposure) in these evaluations. For breast cancer, the standard incidence ratios
9 (SIRs) ranged from 0.73 to 1.25, and for prostate cancer, the SIRs ranged from 0.58 to 1.11,
10 similar to the variability seen among the estimates for lung cancer. In summary, these studies do
11 not provide evidence of an increased risk of lung cancer in the communities surrounding plants
12 that processed vermiculite contaminated with Libby Amphibole asbestos; the small numbers of
13 mesothelioma cases and potential contribution of other asbestos-related sites in some areas make
14 it very difficult to interpret these data. A major limitation of these studies is the lack of
15 information on exposure. Selection of the study population is based on geographic area, with no
16 site-specific or individual-level assessment of relevant exposure pathways. Thus, the extent to
17 which community members were exposed around these facilities is unknown. The use of this
18 type of broad exposure characterization would be expected to result in considerable exposure
19 misclassification. As a result, more refined study designs are needed to evaluate risk to
20 individuals potentially exposed to Libby Amphibole asbestos in their community due to
21 operations at the expansion plants.
22 4.1.4.1. Summary of Community Studies from Other Vermiculite Processing Plants
23 The community-based mortality studies around the 28 exfoliation plants that processed
24 vermiculite contaminated with Libby Amphibole asbestos provide little evidence of an increased
25 risk of asbestos-related cancers in the surrounding communities. These studies are quite limited,
26 however, by the broad exposure classification and the inability to limit the analysis to individuals
27 who had resided in the specific areas during the relevant exposure periods. Additional studies
28 would be needed to more fully examine the potential risks associated with residential exposures
29 from these sources.
30
31
This document is a draft for review purposes only and does not constitute Agency policy.
4-46 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
Table 4-14. Incidence and mortality results for potential asbestos-related
cancers (by cancer site) in communities in the vicinity of
vermiculite-processing facilities (with ATSDR health consultations
evaluating potential pathways of exposure)
Study area0
Incidence"
Observed Expected0
SIR
(95% CI)
Mortality11
Observed
Expected0
SMR
(95% CI)
Lung and bronchus
Los Angeles, CAd
Newark, CAd
Santa Ana, CAd
West Chicago, IL
Dearborn, MI
St. Louis, MO
Trenton, NJ
Edgewater, NJ
Marysville, OH
100
29
79
-
757
-
496
35
-
117.4
27.2
95.4
-
764.4
-
671.0
30.7
-
0.85
1.07
0.83
-
0.99
-
0.74
1.14
-
(0.69, 1.04)
(0.71, 1.53)
(0.66, 1.03)
-
(0.92, 1.06)
-
(0.68,0.81)
(0.80, 1.59)
-
210
125
-
95
1,133
319
976
51
106
285.0
124.3
-
98.6
1,261.3
286.6
1,100.3
50
98.1
0.74
1.01
-
0.96
0.90
1.1
0.89
1.02
1.1
(0.64, 0.84)
(0.84, 1.2)
-
(0.78, 1.18)
(0.85, 0.95)
(1.0, 1.2)
(0.83, 0.94)
(0.76, 1.34)
(0.9, 1.3)
Mesothelioma
Los Angeles, CAd
Newark, CAd
Santa Ana, CAd
West Chicago, IL
Dearborn, MI
St. Louis, MO
Trenton, NJ
Edgewater, NJ
Marysville, OH
1
1
4
-
8
-
6
1
-
1.9
0.4
1.5
-
12.3
-
10.6
0.5
-
0.53
2.49
2.68
-
0.65
-
0.57
2.11
-
(0.01,2.96)
(0.03, 13.9)
(0.72, 6.87)
-
(0.28, 1.28)
-
(0.21, 1.24)
(0.03,11.7)
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
Peritoneum, retroperitoneum, and pleura
Excluding mesothelioma
Los Angeles, CAd
Newark, CAd
Santa Ana, CAd
West Chicago, IL
Dearborn, MI
St. Louis, MO
Trenton, NJ
Edgewater, NJ
Marysville, OH
1 3.1
3 0.7
6 2.7
-
16 19.1
-
10 16.7
1 0.8
-
0.32
4.06
2.24
-
0.84
-
0.60
1.28
-
(0.00, 1.78)
(0.82, 11.9)
(0.82, 4.87)
-
(0.48, 1.36)
-
(0.29, 1.10)
(0.02,7.13)
-
Including mesothelioma
0
0
-
1
9
o
3
18
0
0
2.1
0.9
-
0.8
9.6
2.3
8.3
0.2
0.8
0.0
0.0
-
1.28
0.93
1.3
2.17
0.0
0.0
-
(0,4.10)
-
(0.02,7.12)
(0.43, 1.77)
(0.3, 3.8)
(1.29, 3.43)
-
-
This document is a draft for review purposes only and does not constitute Agency policy.
4-47 DRAFT—DO NOT CITE OR QUOTE
-------
1 Table 4-14. Incidence and mortality results for potential asbestos related
2 cancers (by cancer site) in communities in the vicinity of vermiculite
3 processing facilities (with ATSDR health consultations evaluating potential
4 pathways of exposure) (continued)
5
6
7 aAll incidence studies used Surveillance, Epidemiology, and End Results (SEER) data as the comparison group
8 except New Jersey, which used New Jersey state rates; incidence period in all analyses was 1986-1995. An
9 additional analysis compared the Hamilton, NJ mesothelioma rates to SEER rates; standard incidence ratio (SIR)
10 was reported to be "increased slightly but remained under 1.0." Incidence data, ICD-10 (International
11 Classification of Diseases) codes: lung and bronchus, C340:C349; mesothelioma, M-9050:9053; peritoneum,
12 retroperitoneum, and pleura, C480:C488, C384; respiratory system and intrathoracic organs, C320:C399-excluding
13 mesothelioma; selective digestive organs, C150:C218, C260-C269-excluding mesothelioma.
14 bAll mortality studies used U.S. rates from the National Center for Health Statistics. Mortality period was
15 1989-1998 in the Los Angeles and Newark, CA analyses and was 1979-1998 in all analyses. Mortality data,
16 ICD-9 codes: lung and bronchus, 162.2-162.9; peritoneum, retroperitoneum, and pleura, 158, 163; respiratory
17 system and intrathoracic organs, 161-165; selective digestive organs, 150-154, 159.
18 'Expected values have been rounded.
19 dSimilar results were observed in the CA analyses using alternative methods to calculate standardized risk ratios for
20 incidence and mortality.
21
22 CI = confidence interval.
23
24 Source: Site-specific health consultations at
25 http://www.atsdr.cdc.gov/asbestos/sites/health consultations/index.html.
26
27
28 4.1.5. Case Reports
29 Progressive disease from exposure to Libby Amphibole was noted in a case report of fatal
30 asbestosis in an individual who died 50 years after working at a vermiculite processing plant for
31 a few months at about age 17 (Wright et al., 2002). In another case report, exposures that
32 stemmed from playing for a few years as a child in contaminated vermiculite waste materials
33 around a former Libby vermiculite processing facility was reportedly associated with the
34 development of asbestosis and fatal lung cancer (Srebro and Roggli, 1994).
35
36 4.2. SUBCHRONIC AND CHRONIC STUDIES AND CANCER BIOASSAYS IN
37 ANIMALS—ORAL, INHALATION AND OTHER ROUTES OF EXPOSURE
38 Laboratory animal studies with exposure to Libby Amphibole or tremolite asbestos show
39 effects similar to those observed in occupationally exposed human populations including pleural
40 pathology, mesothelioma, and lung cancer. Tremolite is an amphibole asbestos fiber that is a
41 component of Libby Amphibole asbestos (-6%). Also, in early studies Libby Amphibole
42 asbestos was defined as tremolite. Therefore, laboratory animal studies examining the effect of
43 tremolite exposure have been reviewed and are summarized below to potentially increase
This document is a draft for review purposes only and does not constitute Agency policy.
4-48 DRAFT—DO NOT CITE OR QUOTE
-------
1 understanding of the effects and mechanisms of Libby Amphibole asbestos. Detailed study
2 summaries can be found in Appendix D and summarized in Tables 4-15 and 4-16. No inhalation
3 studies have been performed for Libby Amphibole asbestos, but chronic intrapleural injection
4 studies in hamsters demonstrate carcinogenicity following exposure. The chronic inhalation and
5 intrapleural injection laboratory animal studies with tremolite asbestos demonstrated pleural
6 pathology and carcinogenicity in rats. These studies support the epidemiology studies of Libby
7 Amphibole asbestos exposure (see Section 4.1), and aid in informing the mechanisms of Libby
8 Amphibole asbestos-induced disease.
9
10 4.2.1. Oral
11 No studies in laboratory animals with oral exposure to Libby Amphibole were found in
12 the literature. However, one chronic cancer bioassay was performed following oral exposure to
13 tremolite. McConnell et al. (1983b) describe part of a National Toxicology Program study
14 (NTP, 1990b) performed to evaluate the toxicity and carcinogenicity of ingestion of several
15 minerals, including tremolite. The tremolite (Governeur Talc Co, Governeur, New York) used
16 was not fibrous. No significant tumor induction was observed in the animals with oral exposure
17 to tremolite animals. Although nonneoplastic lesions were observed in many of the aging rats,
18 these were mostly in the stomach and occurred in both controls and exposed animals. The
19 observed lesions included chronic inflammation, ulceration, and necrosis of the stomach
20 (McConnell et al., 1983b). McConnell et al. (1983b) suggested that nonfibrous tremolite could
21 account for the lack of toxicity following exposure in this group of animals. Also, oral studies of
22 asbestos, in general, show decreased toxicity and carcinogenicity as compared to inhalation and
23 implantation/injection studies (Condie, 1983).
24
25 4.2.2. Inhalation
26 There are no laboratory animal studies following inhalation exposure to Libby
27 Amphibole asbestos; however two studies have examined the effect of inhalation exposure to
28 tremolite in Wistar rats (Bernstein et al.. 2005: Bernstein et al.. 2003: Davis et al.. 1985). Davis
29 et al. (1985) performed a chronic inhalation study examining response in male Wistar rats
30 exposed in a chamber to 10 mg/m3 (-1,600 fibers/mL, >5 um) of commercially mined tremolite
31 over a 12-month period. Bernstein et al. (2005; 2003) exposed Wistar rats to tremolite
This document is a draft for review purposes only and does not constitute Agency policy.
4-49 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-15. In vivo data following exposure to Libby Amphibole asbestos
Species (sex)
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)
Spontaneously
Hypertensive (SH)
(n = 6/group)
SH Heart Failure
(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)
Libby Amphibole asbestos
(Six Mix) and crocidolite
Libby Amphibole asbestos
(Six Mix) and crocidolite
Libby Amphibole asbestos
(Six Mix)
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 Libby
Amphibole asbestos.
No differences were observed in histopathology.
Reference
Smith (1978) (W.R. Grace
study)
Putnam et al. (2008)
Smartt et al. (2010)
Shannahan et al. (20 11 a)
rs
s
3
TO
a
I
'TS
s
I
M
H
o
o
o
-------
Table 4-15. In vivo data following exposure to Libby Amphibole asbestos (continued)
Species (sex)
Spontaneously
Hypertensive (SH)
(M)
(n = 8/group)
Fisher 344 rats (M)
(n = 8/group)
Exposure route
Intratracheal instillation
(once)
4h, Id
1.0 mg deferoxamine
(DBF);
21 ug FeCl3; 0.5 mg
LA, 0.5mgFeLA;0.5
mg LA + 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 LA;
0.65 mg amosite in
250 uL saline
Fiber type
Libby Amphibole asbestos
(Six Mix)
Libby Amphibole asbestos
(Six Mix)
Amosite
Effects3
Statistically significant increases in neutrophils
was observed inBALF in animals exposed to LA,
FeLA and LA + DBF with the greatest increase
observed in the LA+DEF animals.
Statistically significant increases in inflammatory
markers were observed following exposure to LA
and amosite, including increased neutrophils and
inflammatory gene expression, with the greatest
increase in amosite-exposed rats.
Reference
Shannahan et al. (20 lib)
Padilla-Carlin et al.
QQ11)
rs
s
3
TO
a
I
'TS
s
O
I"
M
H
aWhen available, results are shown as number of animals with tumors/total number of animals examined.
o
o
o
-------
Table 4-16. In vivo data following exposure to tremolite asbestos
Species (sex)
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)
Sprague-Dawley
and Wistar rats
(n = 32 Wistar rats
(Sample A); 48
Sprague-Dawley
rats [Samples B
andC])
Exposure route
Oral
1% bw in feed pellets;
lifetime exposure
starting in dam
Inhalation
10 mg/m3 (7 h each
day, 5 days per week,
total of 224 days)
Intraperitoneal injection
10 mg/2 mL PBS;
single exposure
Intrapleural injection
10 or 25 mg
Intrapleural injection
20 mg/rat
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)
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
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 and
B;
Sample C: 14/47
Reference
McConnell et al. Q983b)
Davis et al. (1985)
Davis et al. (1991)
Smith et al. (1979)
Wagner et al. (1982)
rs
s
3
TO
a
I
'TS
s
I
M
H
H 8.
oy
O
O
-------
Table 4-16. In vivo data following exposure to tremolite asbestos (continued)
Species (sex)
Osborne-Mendel
rats
(w = 28/group)
Wistar rats (F)
(n = 40/group)
Wistar rats (M)
(n = 56)
C57B1/6 mice (F)
(n = 10/group)
Exposure route
Hardened gelatin
technique
40 mg
Intraperitoneal injection
1 x 3.3 and 1 x 15 mg,
lifetime observation
Inhalation (flow-past
nose only)
100 fibers/cm3 longer
than 20 um, 5 days,
follow-up 1 year later
Intratracheal instillation
Two doses of 60 ug
each given 1 week apart
in the first and second
week of a 7-month
experiment
Fiber type
Tremolite (Two samples)
Tremolite
Tremolite
Tremolite and wollastonite
Effects3
Sample 1: 21/28 pleura! sarcomas
Sample 2: 22/28 pleural sarcomas
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 granule ma development (1 day
post exposure);
Slight interstitial fibrosis observed at 90 and 180
days 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
Stanton et al. (1981)
Roller et al. (1997. 1996)
Bernstein et al. (2005;
20031
Pfau et al. (2008)
rs
s
3
TO
a
I
'TS
s
I
O i£- aWhen available, results are shown as number of animals with tumors/total number of animals examined.
O ^ bAsbestiform types led to mesothelioma in most if not all exposed animals in this study.
H
O
O
O
-------
1 (100 fibers/cm3) and chrysotile for 13 consecutive weeks (6 hours per day, 5 days per week) with
2 1-year follow-up. The results of these inhalation studies produced pronounced inflammation and
3 very high levels of pulmonary fibrosis. Davis et al (1985) also demonstrated an increase in
4 carcinomas and mesotheliomas following exposure to tremolite, with no pulmonary tumors
5 observed in the controls. These results show that Wistar rats exposed to tremolite exhibited
6 increased numbers of pulmonary lesions and possibly tumors.
7
8 4.2.3. Intratracheal Instillation Studies
9 Intratracheal instillation has been used to examine the effect of exposure to Libby
10 Amphibole (Padilla-Carlin et al.. 2011: Shannahan et al.. 2011 a: Shannahan et al.. 20lib: Smartt
11 etal.. 2010: Putnam et al.. 2008) and tremolite asbestos (Blake et al.. 2008: Pfau et al.. 2008:
12 Sahuetal.. 1975). These studies exposed C57B1/6 mice (100 ug/mouse), Wistar Kyoto (WKY)
13 rats (0.25 or 1 mg/rat) or Fisher 344 rats (0.65 or 6.5 mg/rat) once to Libby Amphibole asbestos
14 and analyzed the results up to 3 month postexposure. Putnam et al. (2008) observed nonsta-
15 tistically significant increases in collagen following exposure to Libby Amphibole asbestos, as
16 well as gene expression alterations related to membrane transport, signal transduction, epidermal
17 growth factor signaling, and calcium regulation. Smartt et al. (2010) followed up this study by
18 analyzing specific genes by quantitative RT-PCR for genes involved in collagen accumulation
19 and scar formation (Coll Al, CollA2, Col3Al). Libby Amphibole asbestos exposure led to
20 increased gene expression of Coll A2 at 1 week postinstillation and Col3Al at 1 month post
21 exposure. Both studies observed increased inflammation, however, Libby Amphibole asbestos
22 exposure demonstrated minimal inflammation that did not progress in the time points examined.
23 These studies demonstrate that exposure to Libby Amphibole asbestos may lead to inflammation
24 and fibrosis. Shannahan et al. (2011 a) exposed two rat models of human cardiovascular disease
25 to Libby Amphibole asbestos to determine if the preexisting cardiovascular disease in these
26 models would impact the lung injury and inflammation following exposure. Healthy WKY rats
27 were compared to spontaneously hypertensive (SH) and spontaneously hypertensive heart failure
28 rats following exposure. All rats (male only) were exposed to 0, 0.25, or 1.0 mg/rat via
29 intratracheal instillation and were examined at 1 day, 1 week and 1 month postexposure. No
30 changes were observed histopathologically, however, changes were observed in markers of
31 homeostasis, inflammation and oxidative stress. While inflammation and cell injury were
This document is a draft for review purposes only and does not constitute Agency policy.
4-54 DRAFT—DO NOT CITE OR QUOTE
-------
1 observed in all strains, no strain-related differences were observed following exposure to Libby
2 Amphibole asbestos (Shannahan et al., 201 la). In a follow-up study to further examine the role
3 of iron in the inflammatory response to Libby Amphibole asbestos exposure, Shannahan et al.
4 (20lib) exposed SH rats to Libby Amphibole asbestos alone and with bound Fe as well as with
5 an iron chelator (deferoxamine, DEF). Exposure to Libby Amphibole asbestos led to statistically
6 significant increases in inflammatory markers (e.g., neutrophils, IL-8) with the greatest increase
7 occurring in the presence of DEF. Iron bound to Libby Amphibole asbestos was not released
8 following instillation except in the presence of DEF as supported by the lack of increase in
9 BALF iron. These results suggest that chelation of iron bound to Libby Amphibole asbestos as
10 well as endogenous proteins increases the toxicity of Libby Amphibole asbestos in vivo.
11 Padilla-Carlin et al. (2011) exposed Fisher 344 rats (male only) to Libby Amphibole
12 asbestos (0.65 or 6.5 mg/rat) or amosite (0.65 mg/rat; positive control) by intratracheal
13 instillation to examine inflammatory response for 3 months post-exposure. Libby Amphibole
14 asbestos exposure led to statistically significant increases of neutrophils in BALF as early 1 day
15 post-exposure, with other inflammatory markers (e.g., protein, LDH, GGT) increased statistically
16 significantly at different timepoints during the 3 month period post-exposure. However, on a
17 mass basis, amosite produced a greater inflammatory response as measured by inflammatory
18 markers (e.g., neutrophil influx, gene expression changes) and histopathological analysis
19 demonstrating interstitial fibrosis. These studies demonstrate a statistically significant increase
20 in inflammatory response to Libby Amphibole asbestos in mice and rats as measured in BALF
21 by cytology, histopathology and gene expression analysis. Follow-up studies are needed to
22 inform the chronic effects of exposure to Libby Amphibole asbestos.
23 Laboratory animal studies of tremolite intratracheal instillation exposure have been
24 performed in mice in doses ranging from 60 jig to 5 mg. Male Swiss albino mice exposed to
25 tremolite (5 mg) via intratracheal instillation demonstrated histological changes (Sahu et al.,
26 1975). Microscopic results following exposure to tremolite showed acute inflammation of the
27 lungs at 7 days post exposure, including macrophage proliferation and phagocytosis similar to
28 that observed with amosite and anthophyllite. Limited progression of fibrotic response was
29 observed at 60 and 90 days post exposure, with no further progression of fibrotic response.
30 Blake et al. (2008) and Pfau et al. (2008) examined the role of asbestos in autoimmunity. Blake
31 et al. (2008) performed in vitro assays with Libby Amphibole asbestos (see Section 4.4), and
This document is a draft for review purposes only and does not constitute Agency policy.
4-55 DRAFT—DO NOT CITE OR QUOTE
-------
1 both studies performed the in vivo assays with tremolite. C57BL/6 mice were instilled
2 intratracheally for a total of two doses each of 60-ug saline and wollastonite or Korean tremolite
3 sonicated in sterile phosphate buffer saline (PBS,) given 1 week apart in the first 2 weeks of a
4 7-month experiment. Sera from mice exposed to tremolite showed antibody binding colocalized
5 with SSA/Ro52 on the surface of apoptotic blebs (Blake et al.. 2008). In Pfau et al. (2008). by
6 26 weeks, the tremolite-exposed animals had a significantly higher frequency of positive
7 antinuclear antibody tests compared to wollastinate and saline. Most of the tests were positive
8 for dsDNA and SSA/Ro52. Serum isotyping showed no major changes in immunoglobulin
9 subclasses (IgG, IgA, IgM), but serum IgG in tremolite-exposed mice decreased overall.
10 Further, IgG immune complex deposition in the kidneys increased, with abnormalities suggestive
11 of glomerulonephritis. No increased proteinuria was observed during the course of the study.
12 Local immunologic response was further studied on the cervical lymph nodes. Although total
13 cell numbers and lymph-node size were significantly increased following exposure to tremolite,
14 percentages of T- and B-cells did not significantly change.
15
16 4.2.4. Injection/Implantation Studies
17 There are no laboratory animal studies examining intraperitoneal injection or
18 implantation of Libby Amphibole asbestos. Biological effects following exposure to tremolite
19 have been examined in five intraperitoneal injection studies (Roller et al., 1997, 1996; Davis et
20 al.. 1991: Wagner etal.. 1982: Smith etal.. 1979: Smith. 1978) and one implantation study
21 (Stanton et al.. 1981).
22 Studies by Smith and colleagues (Smith etal.. 1979: 1978). Wagner et al. (1982). Davis
23 et al. (1991) and Roller et al. (1997, 1996) demonstrated that intrapleural injections of tremolite
24 asbestos19 is associated with an increase in pleural fibrosis and mesothelioma in hamsters and
25 rats compared to controls or animals injected with less fibrous materials. Doses ranged from
26 10-25 mg/animal for each study, and although carcinogenesis was observed in these studies
27 there was a variable level of response to the different tremolite forms examined. Although these
28 studies clearly show the carcinogenic potential of Libby Amphibole or tremolite asbestos fibers,
29 intrapleural injections bypass the clearance and dissolution of fibers from the lung after
19 Smith (1978) used tremolite from Libby, MT; Smith et al. (1979) may also have used tremolite from Libby, MT
(i.e., Libby Amphibole asbestos).
This document is a draft for review purposes only and does not constitute Agency policy.
4-56 DRAFT—DO NOT CITE OR QUOTE
-------
1 inhalation exposures. Further, limited information was provided confirming the presence or
2 absence of particles or fibers less than 5 um in length in these studies, limiting the interpretation
3 of results.
4 There is one laboratory animal study that examined the effect of tremolite exposure
5 following implantation of fibers in the pleural cavity. Stanton et al. (1981) also examined
6 tremolite and describe a series of studies on various forms of asbestos. Fibers, embedded in
7 hardened gelatin, were placed against the lung pleura. As an intrapleural exposure, results might
8 not be comparable to inhalation exposures, as the dynamics of fiber deposition and pulmonary
9 clearance mechanisms are not accounted for in the study design. Studies using two tremolite
10 asbestos samples from the same lot were described as being in the optimal size range for
11 carcinogenesis; the fibers were distinctly smaller in diameter than the tremolite fibers Smith et al.
12 (1979) used. These samples both had a high number of fibers in the size range (>8-um long and
13 <0.25-um diameter; i.e., "Stanton fibers"). Exposure to both tremolite samples led to
14 mesotheliomas in 21 and 22 of 28 rats exposed. The Stanton et al. (1981) study also used talc
15 that did not lead to mesothelioma production.
16 There are no studies currently available in laboratory animals exposed to Libby
17 Amphibole asbestos by inhalation. However, the chronic intraperitoneal injection study in
18 hamsters (Smith et al., 1979; Smith, 1978) demonstrated tumor formation following exposure to
19 tremolite obtained from the Libby, MT mine. No other chronic studies of Libby Amphibole
20 asbestos are available. A recent study in rats examining the impact of preexisting cardiovascular
21 disease on pulmonary inflammation demonstrated an increase in inflammatory markers
22 following exposure to Libby Amphibole asbestos via intratracheal instillation in SH rats as
23 compared to normal healthy controls exposed to the same dose (Shannahan et al., 201 Ib). More
24 recent studies examined gene expression changes (Hillegass et al., 2010; Putnam et al., 2008)
25 and early protein markers of fibrosis (Smartt et al., 2010) in mice exposed to Libby Amphibole
26 asbestos via intraperitoneal injection. These studies demonstrated an increase in gene and
27 protein expression related to fibrosis following exposure to Libby Amphibole asbestos.
28 Tremolite fibers, although obtained from different locations throughout the world, consistently
29 led to pulmonary lesions and/or tumor formation with various routes of exposure (inhalation,
30 injection, instillation) and in multiple species (rats, hamsters, and mice) (Bernstein et al., 2005;
31 Bernstein et al.. 2003: Roller et al.. 1997. 1996: Davis et al.. 1985: Wagner et al.. 1982: Stanton
This document is a draft for review purposes only and does not constitute Agency policy.
4-57 DRAFT—DO NOT CITE OR QUOTE
-------
1 et al., 1981). Although comparing potency of the various forms of tremolite is difficult given the
2 limited information on fiber characteristics and study limitations (e.g., length of follow-up
3 postexposure), these results show potential increased risk for cancer (lung and mesothelioma)
4 following exposure to tremolite asbestos.
5 The results of the studies described above show the fibrogenic and carcinogenic potential
6 of Libby Amphibole and tremolite asbestos. Further, the more recent studies by Blake et al.
7 (2008) and Pfau et al. (2008) support human studies demonstrating potential autoimmune effects
8 of asbestos exposure (see Section 4.3.1).
9
10 4.2.5. Summary of Animal Studies for Libby Amphibole and Tremolite Asbestos
11 Tables 4-15 and 4-16 summarize the studies described in this section, with full study
12 details available in Appendix D. Limited in vivo studies have been performed exposing
13 laboratory animals to Libby Amphibole asbestos. One intrapleural injection study using
14 tremolite from the Libby, MT area is included in this section under Libby Amphibole asbestos
15 since earlier terminology for Libby Amphibole asbestos was often tremolite (Smith, 1978).
16 Hamsters in this study exposed to Libby Amphibole asbestos developed fibrosis and
17 mesothelioma following exposure. Subchronic studies in mice (Smartt et al., 2010; Putnam et
18 al., 2008) demonstrated gene and protein expression changes related to fibrosis production
19 following exposure to Libby Amphibole asbestos. Finally, short-term studies in rats
20 demonstrated an increase in inflammatory markers following exposure to Libby Amphibole
21 asbestos (Padilla-Carlin et al.. 2011: Shannahan et al.. 2011 a: Shannahan et al.. 201 Ib).
22 Because tremolite is part of Libby Amphibole asbestos, results from tremolite studies
23 were also described. In general, fibrous tremolite has been shown to cause pulmonary
24 inflammation, fibrosis and/or mesothelioma or lung cancer in rats (Bernstein et al., 2005;
25 Bernstein et al.. 2003: Davis etal.. 1991: Davis etal.. 1985: Wagner etal.. 1982) and hamsters
26 (Smith et al., 1979). The single short-term study on mice showed limited response to tremolite
27 (Sahuet al., 1975). The one chronic-duration oral study (McConnell et al., 1983b) did not show
28 increased toxicity or carcinogenicity; this study, however, used only nonfibrous tremolite, which
29 later studies showed to be less toxic and carcinogenic than fibrous tremolite (Davis et al., 1991).
30 Chronic inflammation is hypothesized to lead to a carcinogenic response through the
31 production of reactive oxygen species and increased cellular proliferation (Hanahan and
This document is a draft for review purposes only and does not constitute Agency policy.
4-58 DRAFT—DO NOT CITE OR QUOTE
-------
1 Weinberg, 2011). Although limited, the data described in Section 4.2 suggest an increase in
2 inflammatory response following exposure to Libby Amphibole asbestos and tremolite asbestos
3 similar to that observed for other durable mineral fibers [reviewed in Mossman et al. (2007)1.
4 Whether this inflammatory response then leads to cancer is unknown. Studies examining other
5 types of asbestos (e.g., crocidolite, chrysotile, and amosite) have demonstrated an increase in
6 chronic inflammation as well as respiratory cancer related to exposure [reviewed in Kamp and
7 Weitzman (1999)1. Chronic inflammation has also been linked to genotoxicity and mutagenicity
8 following exposure to some particles and fibers (Driscoll etal., 1997; 1996; 1995). The evidence
9 described above suggests chronic inflammation is observed following Libby Amphibole asbestos
10 and tremolite asbestos exposure; however, the role of inflammation and whether it leads to lung
11 cancer or mesothelioma following exposure to Libby Amphibole asbestos is unknown.
12 ROS production has been measured in response to both Libby Amphibole asbestos and
13 tremolite asbestos exposure. Blake et al. (2007) demonstrated an increase in the production of
14 superoxide anion following exposure to Libby Amphibole asbestos. Blake et al. (2007) also
15 demonstrated that total superoxide dismutase was inhibited, along with a decrease in intracellular
16 glutathione, both of which are associated with increased levels of ROS. These results are
17 supported by a recent study in human mesothelial cells (Hillegass et al., 2010) (described in
18 Section 4.4 and Appendix D). Increased ROS production was also observed in human airway
19 epithelial cells following exposure to Libby Amphibole asbestos (Duncan et al., 2010) (described
20 in Section 4.4 and Appendix D). This increase in ROS and decrease in glutathione are common
21 effects following exposure to asbestos fibers and particulate matter. Although ROS production is
22 relevant to humans, based on similar human responses as compared to animals, information on
23 the specifics of ROS production following exposure to Libby Amphibole asbestos is limited to
24 the available data described here. Therefore, the role of ROS production in lung cancer and
25 mesothelioma following exposure to Libby Amphibole asbestos is unknown.
26
27 4.3. OTHER DURATION OR ENDPOINT-SPECIFIC STUDIES
28 4.3.1. Immunological
29 Two epidemiology studies have examined the potential role of Libby Amphibole asbestos
30 and autoimmunity. Noonan et al. (2006) used the data from the community health screening to
31 examine self-reported history of autoimmune diseases (rheumatoid arthritis, scleroderma, or
This document is a draft for review purposes only and does not constitute Agency policy.
4-59 DRAFT—DO NOT CITE OR QUOTE
-------
1 lupus) in relation to the asbestos exposure pathways described above (see Table 4-17). To
2 provide more specificity in the self-reported history of these diseases, a follow-up questionnaire
3 was mailed to participants to confirm the initial report and obtain clarifying information
4 regarding the type of disease, whether the condition had been diagnosed by a physician, and
5 whether the participant was currently taking medication for the disease. Responses were
6 obtained from 208 (42%) of the 494 individuals who had reported these conditions. Of these
7 208 responses, 129 repeated the initial report of the diagnosis of rheumatoid arthritis, and
8 161 repeated the initial report of the diagnosis of one of the three diseases (rheumatoid arthritis,
9 scleroderma, or lupus). Among people aged 65 and over (n = 34 rheumatoid arthritis cases,
10 determined using responses from the follow-up questionnaire), a two- to threefold increase in
11 risk was observed in association with several measures reflecting potential exposure to asbestos
12 (e.g., asbestos exposure in the military) or specifically to Libby Amphibole asbestos (e.g., past
13 work in mining and milling operations, use of vermiculite in gardening, and frequent playing on
14 vermiculite piles when young). Restricted forced vital capacity, presence of parenchymal
15 abnormalities, playing on vermiculite piles, and other dust or vermiculite exposures were also
16 associated with rheumatoid arthritis in the group younger than 65 (n = 95 cases). Restricted
17 forced vital capacity was defined as FVC <80% predicted and a ratio of FEV1 to
18 FVC >70% predicted. For all participants, an increased risk of rheumatoid arthritis was observed
19 with increasing number of exposure pathways. RRs of 1.0, 1.02, 1.79, 2.51, and 3.98 were
20 observed for 0 (referent), 1, 2-3, 4-5, and 6 or more pathways, respectively (trend/? < 0.001,
21 adjusting for restrictive spirometry, parenchymal abnormalities, and smoking history). Although
22 the information gathered in the follow-up questionnaire and repeated reports of certain diagnoses
23 decreased the false-positive reports of disease, considerable misclassification (over-reporting and
24 under-reporting) is likely, given the relatively low confirmation rate of self-reports of
25 physician-diagnosed rheumatoid arthritis (and other autoimmune diseases) seen in other studies
26 (Karlson et al.. 2003: Rasch etal.. 2003: Ling et al.. 2000).
27 Another study examined serological measures of autoantibodies in 50 residents of Libby,
28 MT, and a comparison group of residents of Missoula, Montana (Pfau et al., 2005): (see
29 Table 4-17). The Libby residents were recruited for a study of genetic susceptibility to
30
This document is a draft for review purposes only and does not constitute Agency policy.
4-60 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-17. Autoimmune-related studies in the Libby, MT community
Reference(s)
Inclusion criteria and design details
Results
Noonan et al.
(2006)
Nested case-control study among 7,307 participants in
2000-2001 community health screening. Conducted
interviews, gathered self-reported history of rheumatoid
arthritis, scleroderma, or lupus.
Follow-up questionnaire mailed to participants concerning
self-report of "physician-diagnosis" of these diseases and
medication use.
Association with work in Libby
mining/milling operations (ages
65 and older):
Rheumatoid arthritis
OR: 3.2 (95% CI: 1.3,8.0)
Rheumatoid arthritis, lupus,
scleroderma
OR: 2.1 (95% CI: 0.90, 4.1)
Risk increased with increasing
number of asbestos exposure
pathways.
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 sex-matched to Libby
participants.
Serum samples obtained; IgA levels, prevalence of
antinuclear, anti-dsDNA antibodies, anti-RF antibodies,
and anti-Sm, RNP, SS-A, SS-B, and Scl-70 antibodies
determined.
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-Ro (SSA), and
anti-La (SSB) antibodies observed
in Libby sample.
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
asbestos-related lung disease, and the Missoula residents were participants in a study of immune
function The Libby sample exhibited an increased prevalence (22%) of high-titer (>1:320)
antinuclear antibodies when compared to the Missoula sample (6%), and similar increases were
seen in the Libby sample for rheumatoid factor, anti-RNP, anti-Scl-60, anti-Sm, anti-Ro (SSA),
and anti-La (SSB) antibodies. Although neither sample was randomly selected from the
community residents, an individual's interest in participating in a gene and lung disease study
likely would not be influenced by the presence of autoimmune disease or autoantibodies in that
individual.
Hamilton et al. (2004). Blake et al. (20081 and Pfau et al. (2008) examined the role of
asbestos in autoimmunity in laboratory animal or in vitro studies. Blake et al. (2008) performed
in vitro assays with Libby Amphibole asbestos (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 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
This document is a draft for review purposes only and does not constitute Agency policy.
4-61 DRAFT—DO NOT CITE OR QUOTE
-------
1 higher frequency of positive antinuclear antibody tests compared to wollastinate and saline.
2 Most of the tests were positive for dsDNA and SSA/Ro52. Serum isotyping showed no major
3 changes in immunoglobulin subclasses (IgG, IgA, IgM), but serum IgG in tremolite-exposed
4 mice decreased overall. Further, IgG immune complex deposition in the kidneys increased, with
5 abnormalities suggestive of glomerulonephritis. No increased proteinuria was observed during
6 the course of the study. Local immunologic response was further studied on the cervical lymph
7 nodes. Although total cell numbers and lymph-node sizes were significantly increased following
8 exposure to tremolite, percentages of T- and B-cells did not significantly change. Hamilton et al.
9 (2004) investigated the ability of Libby Amphibole, crocidolite, and PlV^.s (collected over a 6
10 month period in Houston, TX, from EPA site 48-201-1035) to alter the antigen-presenting cell
11 (APC) function was altered in cultured human alveolar macrophages. Asbestos exposure
12 (regardless of type) and PM2.5 up-regulated a THi lymphocyte derived cytokine, interferon
13 gamma (IFNy), and the Tn2 lymphocyte-derived cytokines interleukin-4 (IL-4) and
14 interleukin-13 (IL-13). There was, however, extreme variation among subjects in the amount of
15 response. In addition, there was no correlation between an individual's cells' response to
16 asbestos versus PM, suggesting that more than one possible mechanism exists for a
17 particle-induced APC effect and individual differential sensitivities to inhaled bioactive particles.
18 Although limited number of studies, these results suggest a possible effect on
19 autoimmunity following exposure to Libby Amphibole asbestos. Further studies are needed to
20 increase understanding of this potential effect.
21
22 4.4. MECHANISTIC DATA AND OTHER STUDIES IN SUPPORT OF THE MODE OF
23 ACTION
24 In vitro analysis of fibers depends on the characteristics of the fibers and cell types used
25 for the studies. Therefore, in reviewing the literature it is important to pay attention to cell types
26 used, particularly related to the ability to internalize fibers and produce an oxidative stress
27 response. Results from in vitro studies have demonstrated potential biological mechanisms of
28 oxidative stress and inflammation in response to exposure to Libby Amphibole and tremolite
29 asbestos. These studies are summarized below and in Tables 4-18 and 4-19, with detailed study
30 descriptions available in Appendix D.
This document is a draft for review purposes only and does not constitute Agency policy.
4-62 DRAFT—DO NOT CITE OR QUOTE
-------
1 Limited in vitro studies have been conducted with Libby Amphibole asbestos from the
2 Zonolite Mountain mine. These studies demonstrated an effect of Libby Amphibole asbestos on
3 inflammation and immune function (Duncan et al., 2010; Blake et al., 2008; Blake et al., 2007;
4 Hamilton et al., 2004), oxidative stress (Hillegass et al., 2010), and genotoxicity (Pietruska et al.,
5 2010). Similar endpoints have been examined in vitro following exposure to tremolite asbestos
6 (Okavasu et al., 1999: Wvlieetal., 1997: Suzuki and Hei, 1996: Athanasiou et al., 1992: Wagner
7 etal., 1982).
8
9 4.4.1. Inflammation and Immune Function
10 Hamilton et al. (2004) showed an increase in TH1 and TH2 cytokines following exposure
11 to both asbestos and paniculate matter, suggesting a similar effect of exposure to both materials
12 on immune function. Analysis of these results is limited, as the use of primary cells in culture
13 that led to an extremely variable response. Two studies by Blake et al. (2008: 2007) further
14 examined the effect of Libby Amphibole asbestos on immune response in murine macrophages.
15 These studies demonstrated that Libby Amphibole asbestos was internalized, and this
16 internalization resulted in an increase in reactive oxygen species (ROS). These studies also
17 showed a variable cytotoxic response, as Libby Amphibole asbestos exposure did not result in a
18 statistically significant increase in cytotoxicity, while crocidolite did. DNA damage also was
19 increased in crocidolite-exposed cells—but not in Libby Amphibole asbestos exposed-cells. An
20 increase (relative to controls) in autoantibody formation following exposure to Libby Amphibole
21 asbestos also was observed. Studies that examined cellular response to tremolite also found that
22 fiber characteristics (length and width) play a role in determining ROS production, toxicity, and
23 mutagenicity (Okavasu et al., 1999: Wagner etal., 1982).
24 Mechanisms of oxidative stress following exposure to Libby Amphibole asbestos were
25 also studied in human mesothelial cells (Hillegass et al., 2010). Gene expression changes
26 following exposure to 15 x io6 |im2/cm2 Libby Amphibole asbestos20 as compared to the
27 nonpathogenic control (75 x IO6 |im2/cm2 glass beads) in the human mesothelial cell line
28 LP9/TERT-1 for 8 and 24 hours. Gene ontology of these results demonstrated alterations in
29 genes related to signal transduction, immune response, apoptosis, cellular proliferation,
20 Libby Amphibole asbestos samples were characterized for this study with analysis of chemical composition and
mean surface area (Meeker et al.. 2003). Doses were measured in surface area and described based on viability
assays as either the -nontoxic (15 * io6 um2/cm2) or the toxic dose (75 * io6 um2/cm2).
This document is a draft for review purposes only and does not constitute Agency policy.
4-63 DRAFT—DO NOT CITE OR QUOTE
-------
Table 4-18. In vitro data following exposure to Libby Amphibole asbestos
Test system
Primary human alveolar
macrophages and
lymphocytes
Murine macrophages
(primary and
RAW264.7)3
Murine macrophages
(primary and
RAW264.7)
Human lung epithelial
cells (wild-type and
XRCC1 -deficient)
Human mesothelial cells
(LP9/TERT-1 and
HKNM-2)
Primary human airway
epithelial cells (HAECs)
Fiber type
Libby Amphibole asbestos or
crocidolite
Libby Amphibole asbestos and
crocidolite
Libby Amphibole asbestos or
crocidolite
Libby Amphibole asbestos or
crocidolite
Libby Amphibole asbestos or
crocidolite
Libby Amphibole asbestos
(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 24 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 * 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 TH1 and TH2 cytokines (IFNy,
IL-4, IL-13)
Internalized Libby Amphibole asbestos 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,
particularly SOD2
Increases in pro-inflammatory gene expression
and ROS production
Reference
Hamilton et al.
(2004)
Blake et al.
(2007)
Blake et al.
(2008)
Pietruska et al.
(2010)
Hillegass et al.
(2010)
Duncan et al.
(2010)
o
s
3
a
>!
a
I
'TS
s
a
a
H 1
o I
H S.
~ ,rs
o
c
o
"All results for RAW264.7. Data not shown for primary cells though authors state similar response to RAW264.7.
PBS = phosphate buffer saline, ROS = reactive oxygen species, GSH = glutathione, DNA = deoxyribonucleic acid, LDH = lactic dehydrogenase,
BGL = (3-glucuronidase, SHE = Syrian hamster ovary, HTE = hamster trachea! epithelial, RPM = rat pleura! mesothelial, NIEHS = National Institute of
Environmental Health Sciences, HPRT = hypoxanthine-guanine phosphoribosyltransferase.
-------
Table 4-19. 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
A[L] cells (hamster
hybrid cells containing
human chromosome 11)
HTE and RPM cell lines
A[L] cells (hamster
hybrid cells containing
human chromosome 11)
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)
Tremolite, erionite, RCF-1
Dose/exposure duration
0, 50, 100, and 150 ug/mL
18 h
TA98, TA100, and TA102:
0-500 ug/per plate
2 days
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).
0-400 ug/mL
24 h
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 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.
No significant increase in HPRT mutations for
these three fibers;
Dose-dependent induction of mutations in CD59
did occur for erionite and tremolite.
Reference
Wagner et al.
(1982)
Athanasiou et al.
(1992)
Suzuki and Hei
(1996)
Wylie et al.
(1997)
Okayasu et al.
(1999)
o
S
S
a
>!
a
I
S
1
a
a
H §
§1
-
H
H S.
~ ,rs
o
C
o
PBS = phosphate buffer saline, ROS = reactive oxygen species, GSH = glutathione, DNA = deoxyribonucleic acid, LDH = lactic dehydrogenase,
BGL = (3-glucuronidase, SHE = Syrian hamster ovary, HTE = hamster tracheal epithelial, RPM = rat pleural mesothelial, NIEHS = National Institute of
Environmental Health Sciences, HPRT = hypoxanthine-guanine phosphoribosyltransferase.
-------
1 extracellular matrix, cell adhesion and motility, and only in one gene related to reactive oxygen
2 species processing. Oxidative stress was observed as both dose- and time-dependent in cells
3 exposed to Libby Amphibole asbestos but was increased following exposure to the higher dose
4 of Libby Amphibole asbestos (statistical analysis not possible). Glutathione (GSH) levels were
5 transiently depleted following 2-8 hours exposure to the higher dose of Libby Amphibole
6 asbestos, with a gradual recovery up to 48 hours in LP9/TERT-1 cells (HKNM-2 not analyzed).
7 These studies demonstrate that Libby Amphibole asbestos exposure leads to increases in
8 oxidative stress as measured by ROS production, gene expression, protein and functional
9 changes in oxidative stress proteins (SOD), and GSH level alterations in human mesothelial
10 cells.
11 Gene expression alterations of interleukin-8 (IL-8), cyclooxygenase-2 (COX-2), heme
12 oxygenase (HO)-l as well as other stress-responsive genes as compared to amosite (Research
13 Triangle Institute) was observed in primary human airway epithelial cells (HAEC) following
14 exposure to Libby Amphibole asbestos. Comparisons were made with both fractionated
15 (aerodynamic diameter <2.5 jim) and unfractionated fiber samples (Duncan et al., 2010).
16 Crocidolite fibers (UICC) were also included in some portions of this study for comparison.
17 Primary HAECs were exposed to 0, 2.64, 13.2, and 26.4 jig/cm2 of crocidolite, amosite (AM),
18 amosite 2.5 (fractionated), Libby Amphibole asbestos, or Libby Amphibole asbestos
19 2.5 (fractionated) for 2 or 24 hours in cell culture. Cytotoxicity was determined by measurement
20 of lactate dehydrogenase (LDH) from the maximum dose (26.4 jig/cm2) of both amosite and
21 Libby Amphibole asbestos samples, with less than 10% LDH present following exposure to all
22 four samples. Minimal increases in gene expression of IL-8, COX-2, or HO-1 were observed at
23 2 hours postexposure to all five fiber types; at 24 hour postexposure, however, a dose response
24 was observed following exposure to all fiber types with the results showing a pro-inflammatory
25 gene expression response (Duncan et al., 2010). These results support a limited cytotoxicity of
26 both amosite and Libby Amphibole asbestos under these concentrations and time frames.
27
This document is a draft for review purposes only and does not constitute Agency policy.
4-66 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.4.2. Genotoxicity
2 Genotoxicity and, more specifically, mutagenicity, are associated with tumor formation
3 through alterations in genetic material.21 Mutagenicity refers to a permanent effect on the
4 structure and/or amount of genetic material that can lead to heritable changes in function, while
5 genotoxicity is a broader term including all adverse effects on the genetic information (Eastmond
6 et al., 2009). Results of standard mutation assays like the Ames test, which analyze for point
7 mutations, have found asbestos and other mineral fibers to be negative or only marginally
8 positive (Walker et al., 1992). Several other studies, however, have shown that asbestos
9 exposure can result in a variety of chromosomal alterations, which are briefly discussed below.
10 Genotoxicity following exposure to asbestos fibers has been described as the result of
11 two distinct mechanisms, either ROS production leading to direct DNA damage, or physical
12 interference of mitosis by the fibers. For both DNA damage and mitotic interference, the fibers
13 must first enter the cell. Some studies have shown that a direct interaction between fibers and
14 cellular receptors might also lead to increased ROS production. ROS production is likely to be a
15 key event in fiber-induced direct DNA damage, as observed following exposure to other forms
16 of asbestos, while the indirect DNA damage requires fiber interaction with cellular components
17 (e.g., mitotic spindle, chromosomes).
18 ROS production and genotoxicity (micronuclei induction) following exposure to Libby
19 Amphibole asbestos has been demonstrated in XRCC1-deficient human lung epithelial
20 H460 cells (Pietruska et al., 2010). XRCC1 is involved in the repair mechanisms for oxidative
21 DNA damage, particularly single strand breaks. Micronuclei induction was measured following
22 treatment of cells by controls (positive, hydrogen peroxide; negative, paclitaxel) and by
9
23 5 jig/cm fibers or TiO2 particles for 24 hours. Spontaneous micronuclei induction was increased
24 in XRCC1-deficient cells in a dose-dependent manner following exposure to crocidolite and
25 Libby Amphibole asbestos as compared to control. These results support a potential genotoxic
26 effect of exposure to both crocidolite and Libby Amphibole asbestos.
21 Genotoxicity: a broad term and 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 which provide an indication of induced damage to DNA (but not direct evidence of mutation) via
effects such as unscheduled DNA synthesis, sister chromatid exchange, or mitotic recombination, as well as tests for
mutagenicity; Mutagenicity: refers to the induction of permanent transmissible changes in the amount or structure of
the genetic material of cells or organisms. These changes, "mutations," may involve a single gene or gene segment,
a block of genes, or whole chromosomes. Effects on whole chromosomes may be structural and/or numerical (as
defined in the European Union Technical Guidance on Risk Assessment (CEC. 1996).
This document is a draft for review purposes only and does not constitute Agency policy.
4-67 DRAFT—DO NOT CITE OR QUOTE
-------
1 Athanasiou et al. (1992) performed a series of experiments to measure genotoxicity
2 following exposure to tremolite, including the Ames mutagenicity assay, micronuclei induction,
3 chromosomal aberrations, and gap-junction intercellular communication. Although a useful test
4 system for mutagenicity screening for many agents, the Ames assay is not the most effective test
5 to detect mutations induced by mineral fibers. Mineral fibers can cause mutation through
6 generation of ROS or direct disruption of the spindle apparatus during chromatid segregation.
7 Fibers do not induce ROS in the Ames system, however, and the Salmonella typhimurium strains
8 do not endocytose the fibers. Only one study was found in the published literature that used the
9 Ames assay to measure mutagenicity of tremolite. Metsovo tremolite asbestos has been shown
10 to be the causative agent of endemic pleural calcification and an increased level of malignant
11 pleural mesothelioma (see Section 4.1). To measure the mutagenicity of Metsovo tremolite,
12 S. typhimurium strains (TA98, TA100, and TA102) were exposed to 0-500 ug/plate of asbestos
13 (Athanasiou et al., 1992). Metsovo tremolite did not yield a statistically significant increase in
14 revertants in the Ames assay, including in the TA102 Salmonella strain, which is generally
15 sensitive to oxidative damage. This study demonstrated clastogenic effects of tremolite,
16 including chromosomal aberrations and micronuclei induction. Tremolite exposure in Syrian
17 hamster embryo (SHE) cells did lead to a dose-dependent increase in chromosome aberrations
18 that was statistically significant at the highest doses tested (1.0-3.0 ug/cm2) (p < 0.01)
19 (Athanasiou et al., 1992). A statistically significant dose-dependent increase in levels of
20 micronuclei was demonstrated following tremolite exposure at concentrations as low as
r\
21 0.5 ug/cm (p < 0.01) in BPNi cells after 24-hour exposure. Literatures searches did not find
22 tremolite tested for clastogenicity in other cell types, but the results of this study suggest
23 interference with the spindle apparatus by these fibers. No analysis was performed to determine
24 if fiber interference of the spindle apparatus could be observed, which would have supported
25 these results. No effect on the gap-junctional intercellular communication following tremolite
26 exposure was observed in both Chinese hamster lung fibroblasts (V79) and Syrian hamster
27 embryo BPNi cells, which are sensitive to transformation (Athanasiou et al., 1992).
28 Okayasu et al. (1999) analyzed the mutagenicity of Metsovo tremolite, erionite, and the
29 man-made ceramic (RCF-1) fiber. Human-hamster hybrid A(L) cells contain a full set of
30 hamster chromosomes and a single copy of human chromosome 11. Mutagenesis of the CD59
31 locus on this chromosome is quantifiable by antibody complement-mediated cytotoxicity assay.
This document is a draft for review purposes only and does not constitute Agency policy.
4-68 DRAFT—DO NOT CITE OR QUOTE
-------
1 The authors state that this is a highly sensitive mutagenicity assay, and previous studies have
2 demonstrated mutagenicity of both crocidolite and chrysotile (Hei et al., 1992). The cytotoxicity
3 analysis for mutagenicity was performed by exposing 1 x 105 A(L) cells to a range of
4 concentrations of fibers as measured by weight (0-400 ug/mL or 0-80 ug/cm2) for 24 hours at
5 37°C. CD59 mutant induction showed a dose-dependent increase in mutation induction for
6 erionite and tremolite, but RCF-1 did not.
7 In summary, one in vitro study examined genotoxicity of Libby Amphibole asbestos by
8 measuring DNA adduct formation following exposure via murine macrophages (primary and
9 immortalized) (Blake et al., 2007). The data showed no increase in adduct formation as
10 compared to unexposed controls. A second study observed increases in micronuclei induction in
11 both normal human lung epithelial cells and XRCC1-deficient cells for both Libby Amphibole
12 and crocidolite asbestos (Pietruska et al., 2010). Two studies of tremolite examined
13 genotoxicity. The first found no significant increase in revertants in the Ames assay (Athanasiou
14 et al., 1992), which is similar to results obtained for other forms of asbestos. This study did find,
15 however, that tremolite exposure led to a dose-dependent increase in chromosome number and
16 micronuclei formation, which has also been described for other asbestos fibers [as reviewed in
17 Hei et al. (2006) and Jaurand and Levy (1997)1. Hei and colleagues (Okavasu et al.. 1999)
18 performed mutation analysis with tremolite and found a dose-dependent increase in mutations in
19 CD59 in hamster hybrid cells. Genotoxicity analysis in humans, following exposure to Libby
20 Amphibole asbestos or tremolite, has not been measured, although other types of asbestos fibers
21 have led to increases in genotoxicity in primary cultures and lymphocytes (Dopp et al., 2005;
22 Poser et al., 2004). In general, these studies have examined genotoxicity with a focus on ROS
23 production as a key event. Although Libby Amphibole asbestos- and tremolite-specific data are
24 limited to in vitro studies, given the similarities in response to other forms of asbestos, there is
25 some evidence to suggest genotoxicity following exposure to Libby Amphibole and tremolite
26 asbestos. However, the potential role of this genotoxicity in lung cancer or mesothelioma
27 following exposure to Libby Amphibole asbestos is unknown.
28
29 4.4.3. Cytotoxicity and Cellular Proliferation
30 The initial stages of turnorigenicity may be an increased cellular proliferation at the site
31 of fiber deposition, which can increase the chance of cancer by increasing the population of
This document is a draft for review purposes only and does not constitute Agency policy.
4-69 DRAFT—DO NOT CITE OR QUOTE
-------
1 spontaneous mutations, thereby affording genotoxic effects an opportunity to multiply.
2 Increased cell proliferative regeneration is also a hallmark of tumor clonal expansion and
3 generally occurs in response to increased apoptosis.
4 Wagner et al. (1982) examined the in vitro cytotoxicity of three forms of tremolite used
5 in their in vivo studies. LDH and p-glucuronidase were measured in the medium following
6 incubation of unactivated primary murine macrophages to 50, 100, and 150 ug/mL of each
7 sample for 18 hours. The Korean tremolite (Sample C) produced results similar to the positive
8 control: increased toxicity of primary murine macrophages, increased cytoxicity of Chinese
9 hamster ovary (CHO) cells, and increased formation of giant cells from the A549 cell line. The
10 tremolite sample from Greenland (Sample B) did result in increased toxicity over controls;
11 although to a lesser degree (statistics are not given). Although differential toxicity of these
12 samples was noted on a mass basis, data were not normalized for fiber content or size. The
13 inference is that differential results may be due, at least in part, to differential fiber counts.
14 Wylie et al. (1997) examined the mineralogical features associated with cytotoxic and
15 proliferative effects of asbestos in hamster tracheal epithelial (HTE) and rat pleural mesothelial
16 (RPM) cells with a colony-forming efficiency assay. HTE cells are used because they give rise
17 to tracheobronchial carcinoma, while RPM cells give rise to mesotheliomas. The results of the
18 analysis with fiber exposure by mass (ug/cm2) show elevated colonies in HTE cells following
19 exposures to both asbestos fibers (p < 0.05) at the lowest concentrations, while significant
20 decreases were observed for both asbestos fibers at the higher concentrations (0.5 ug/cm2,
21 p< 0.05) (Wylie et al., 1997). No proliferation was observed for either chrysotile or crocidolite
22 asbestos fibers in RPM cells, but cytotoxicity was observed at concentrations greater than
r\
23 0.05 ug/cm (p < 0.05). All talc samples were less cytotoxic in both cell types. Analyzing the
24 data for cytotoxicity and proliferation based on the exposure measurement demonstrated
25 differences in response depending solely on how the fibers were measured: by mass, number, or
26 surface area. These results show variability in interpreting the results of the same assay based on
27 the defined unit of exposure. Most early studies used mass as the measurement for exposure,
28 which can impact how the results are interpreted. When possible, further analysis of fiber
29 number and surface area would help elucidate the role of these metrics, particularly for in vivo
30 studies.
This document is a draft for review purposes only and does not constitute Agency policy.
4-70 DRAFT—DO NOT CITE OR QUOTE
-------
1 Tremolite and Libby Amphibole asbestos exposure led to increases in both fibrosis and
2 tumorigenicity in all but one animal study, supporting a possible role for proliferation in
3 response to these fibers. However, there are limited data to demonstrate that increased
4 cytotoxicity and cellular proliferation following exposure to Libby Amphibole asbestos leads to
5 lung cancer or mesothelioma.
6 Summary. The review of these studies clearly highlights the need for more controlled
7 studies examining Libby Amphibole asbestos in comparison with other forms of asbestos and for
8 examining multiple endpoints—including ROS production, DNA damage, and pro-inflammatory
9 gene expression alterations—to improve understanding of mechanisms involved in cancer and
10 other health effects. Data gaps still remain to determine specific mechanisms involved in Libby
11 Amphibole asbestos-induced disease. Studies that examined cellular response to tremolite also
12 found that tremolite exposure may lead to increased ROS production, toxicity, and genotoxicity
13 (Okayasu et al., 1999; Wagner et al., 1982). As with the in vivo studies, the definition of fibers
14 and how the exposures were measured varies among studies.
15
16 4.5. SYNTHESIS OF MAJOR NONCANCER EFFECTS
17 The predominant noncancer health effects observed following inhalation exposure to
18 Libby Amphibole asbestos are effects on the lungs and pleural lining surrounding the lungs.
19 Recent studies have also examined noncancer health effects following exposure to Libby
20 Amphibole asbestos in other systems, including autoimmune effects and cardiovascular disease.
21 These effects have been observed primarily in studies of exposed workers and community
22 members and are supported by laboratory animal studies.
23
24 4.5.1. Pulmonary Effects
25 4.5.1.1. Pulmonary Fibrosis (Asbestosis)
26 Asbestosis is the interstitial pneumonitis and fibrosis caused by inhalation of asbestos
27 fibers and is characterized by a diffuse increase of collagen in the alveolar walls (fibrosis) and
28 the presence of asbestos fibers, either free or coated with a proteinaceous material and iron
29 (asbestos bodies). Fibrosis results from a sequence of events following lung injury, which
30 includes inflammatory cell migration, edema, cellular proliferation, and accumulation of
31 collagen. Asbestosis is associated with dyspnea, bibasilar rales, and changes in pulmonary
This document is a draft for review purposes only and does not constitute Agency policy.
4-71 DRAFT—DO NOT CITE OR QUOTE
-------
1 function: a restrictive pattern, mixed restrictive-obstructive pattern, and/or decreased diffusing
2 capacity (ATS, 2004). Radiographic evidence of small opacities in the lung is direct evidence of
3 scarring of the lung tissue and as the fibrotic scarring of lung tissue consistent with mineral dust
4 and mineral fiber toxicity. The scarring of the parenchymal tissue of the lung contributes to
5 measured changes in pulmonary function, including obstructive pulmonary deficits from
6 narrowing airways, restrictive pulmonary deficits from impacting the elasticity of the lung as
7 well as decrements in gas exchange.
8 Workers exposed to Libby Amphibole asbestos from vermiculite mining and processing
9 facilities in Libby, MT, as well as plant workers in Marysville, OH, where vermiculite ore was
10 exfoliated and processed, have an increased prevalence of small opacities on chest X-rays, which
11 is indicative of fibrotic damage to the parenchymal tissue of the lung (Rohs et al., 2008;
12 Amandus et al.. 1987b: McDonald et al.. 1986b: Lockev et al.. 1984). These findings are
13 consistent with a diagnosis of asbestosis, and the studies are described in detail in
14 Section 4.1.1.4.2. Significant increases in asbestosis as the primary cause-of-death have been
15 documented in studies of the Libby worker cohort report (see Table 4.6 for details) (Larson et al.,
16 2010b: Sullivan. 2007: Amandus and Wheeler. 1987: McDonald et al.. 1986a). For both
17 asbestosis mortality and radiographic signs of asbestos (small opacities), positive exposure-
18 response relationships are described where these effects are greater with greater cumulative
19 exposure to Libby Amphibole asbestos.
20 Deficits in pulmonary function consistent with pulmonary fibrosis have been reported in
21 individuals exposed to Libby Amphibole asbestos. The initial study of the Marysville, OH
22 cohort measured but reported no change in pulmonary function (Lockey et al., 1984).
23 Pulmonary function was not reported for the cohort follow-up, although prevalence of pleural
24 and parenchymal abnormalities was increased (Rohs et al., 2008). Although studies of the
25 occupational Libby worker cohort do not include assessment of pulmonary function (Amandus et
26 al., 1987b: McDonald et al., 1986b) data from the ATSDR community screening, which included
27 workers, provide support for functional effects from parenchymal changes. The original report
28 of the health screening data indicated moderate-to-severe pulmonary restriction in 2.2% of men
29 (Peipins et al.. 2003: ATSDR. 200 Ib). A recent reanalysis of these data show that for study
30 participants with small opacities viewed on the radiographs (grade 1/0 or greater), and DPT the
31 mean FVC is reduced to 78.76 (±3.64), 82.16 (±3.34), respectively of the expected value (Weill
This document is a draft for review purposes only and does not constitute Agency policy.
4-72 DRAFT—DO NOT CITE OR QUOTE
-------
1 et al., 2011). A mean FVC of 95.63 (±0.76) was reported for those with other pleural
2 abnormalities versus 103.15 (±0.25) in participants with no radiographic abnormalities. The
3 strongest effects of diffuse pleural thickening and/or costophrenic angle obliteration on FVC
4 were seen among men who had never smoked (-23.77,/> < 0.05), with smaller effects seen
5 among men who had smoked (-9.77, p < 0.05) and women who had smoked (-6.73, p < 0.05).
6 Laboratory animal and mechanistic studies of Libby Amphibole asbestos are consistent with the
7 noncancer health effects observed in both Libby workers and community members. Pleural
8 fibrosis was increased in hamsters after intrapleural injections of Libby Amphibole asbestos
9 (Smith, 1978). More recent studies have demonstrated increased collagen deposition consistent
10 with fibrosis following intratracheal instillation of Libby Amphibole asbestos fibers in mice
11 (Padilla-Carlin et al.. 2011: Shannahan et al.. 2011 a: Shannahan et al.. 20lib: Smartt etal.. 2010:
12 Putnam et al., 2008). Pulmonary fibrosis, inflammation, and granulomas were observed after
13 tremolite inhalation exposure in Wistar rats (Bernstein et al., 2005: Bernstein et al., 2003) and
14 intratracheal instillation in albino Swiss mice (Sahu et al., 1975). Davis et al. (1985) also
15 reported pulmonary effects after inhalation exposure in Wistar rats including increases in
16 peribronchiolar fibrosis, alveolar wall thickening, and interstitial fibrosis.
17
18 4.5.1.2. Other Nonmalignant Respiratory Diseases
19 Mortality studies of the Libby workers indicate that there is increased mortality, not only
20 from asbestosis, but other respiratory diseases. Deaths attributed to chronic obstructive
21 respiratory disease and deaths attributed to "other" nonmalignant respiratory disease were
22 elevated more than twofold (see Table 4-6) (Larson et al.. 201 Ob: Sullivan. 2007). These
23 diseases are consistent with asbestos toxicity, and the evidence of a positive exposure-response
24 relationship for mortality from all nonmalignant respiratory diseases, supports this association.
25
26 4.5.2. Pleural Effects
27 Pleural thickening that is caused by mineral fiber exposure includes two distinct
28 biological lesions: discrete pleural plaques in the parietal pleura and diffuse pleural thickening of
29 the visceral pleura. Both forms of pleural thickening can be viewed on standard radiographs.
30 However, the two are not always clearly distinguishable on X-rays, and smaller lesions may not
31 be detected. High resolution computed tomography is a method that can distinguish between the
This document is a draft for review purposes only and does not constitute Agency policy.
4-73 DRAFT—DO NOT CITE OR QUOTE
-------
1 lesions, as well as detect smaller lesions than are visible on X-rays. Pleural thickening may
2 restrict lung function, increase breathlessness with exercise, and contribute to chronic chest pain.
3 The potential for health effects and severity of health effects are increased with the extent and
4 thickness of the pleural lesions.
5 Data from the ATSDR community health screening study indicate that the prevalence of
6 pleural abnormalities, identified by radiographic examination, increases substantially with
7 increasing number of exposure pathways (Peipins et al., 2003). A reanalysis of these data also
8 considered age, smoking history, and types of exposures. Increased pleural thickening is
9 reported for Libby workers, those with other vermiculite work and those in "dusty trades."
10 Increased LPT is reported in both those exposed only as househole contacts or through
11 environmental exposure pathways, with greater incidence by age (38.3 and 12.7%, respectively,
12 in the 61-90 age group) (Weill et al.. 2011). DPT is reported at lower rates with 5.9 and 2.2%,
13 respectively, in these exposure groups in the highest age bracket evaluated (age 61-90).
14 Increased pleural thickening is reported for both of the studied worker cohorts, with
15 evidence of positive exposure response relationships (Larson et al., 2010a: Rohs et al., 2008;
16 Amandus et al.. 1987b: McDonald et al.. 1986b: Lockev et al.. 1984). Both McDonald etal.
17 (1986b) and Amandus et al. (1987b) indicate age is also a predictor of pleural thickening in
18 exposed individuals, which may reflect the effects of time from first exposure. Smoking data
19 were limited on the Libby workers and analyses do not indicate clear relationships between
20 smoking and pleural thickening (Amandus et al., 1987b: McDonald et al., 1986b). Pleural
21 thickening in workers at the Scott Plant (Marysville, OH) was associated with hire on or before
22 1973 and age at time of interview but was not associated with BMI or smoking history (ever
23 smoked) (Rohs et al.. 2008).
24
25 4.5.3. Other Noncancer Health Effects (Cardiovascular Toxicity, Autoimmune Effects)
26 There is limited research available on noncancer health effects occurring outside the
27 respiratory system. Larson et al. (201 Ob) examined cardiovascular disease-related mortality in
28 the cohort of exposed workers from Libby (see Section 4.1.1.4.3). Mechanistic studies have
29 examined the potential role of iron and the associated inflammation for both the respiratory and
30 cardiovascular disease (Shannahan et al., 201 Ib). Two studies examined the association between
31 asbestos exposure and autoimmune disease (Noonan et al., 2006) or autoantiboides and other
This document is a draft for review purposes only and does not constitute Agency policy.
4-74 DRAFT—DO NOT CITE OR QUOTE
-------
1 immune markers (Pfau et al., 2005) (see Table 4-17). Limitations in the number, scope, and
2 design of these studies make it difficult to reach conclusions as to the role of asbestos exposure
3 in either cardiovascular disease or autoimmune disease.
4
5 4.5.4. Libby Amphibole Asbestos Summary of Noncancer Health Effects
6 The studies in humans summarized in Section 4.1 have documented an increase in
7 mortality from nonmalignant respiratory disease, including asbestosis, in workers exposed to
8 Libby Amphibole asbestos (Larson et al.. 201 Ob: Sullivan. 2007: McDonald et al.. 2004:
9 Amandus and Wheeler, 1987). Radiographic evidence of pleural thickening and interstitial
10 damage (small opacities) are also well documented among employees of the Libby vermiculite
11 mining operations (Larson et al., 2010a: Amandus et al., 1987b: McDonald et al., 1986b).
12 Additional studies have documented an increase in radiographic changes in the pleura and
13 parenchyma among employees of a manufacturing facility in Marysville, OH that used Libby
14 vermiculite ore contaminated with Libby Amphibole asbestos (Rohs et al., 2008: Lockey et al.,
15 1984). Positive exposure-response relationships for these health effects for both occupational
16 cohorts studied, as well as the observed latency, support an association between exposure to
17 Libby Amphibole asbestos and these pleuro-pulmonary effects. Studies of community members
18 exposed to Libby Amphibole asbestos have documented similar pleural abnormalities and
19 pulmonary deficits consistent with parenchymal damage (Weill etal., 2011; Whitehouse, 2004;
20 Peipins et al., 2003). Although limited, animal studies support the toxicity of Libby Amphibole
21 asbestos to pleural and pulmonary tissues. Developing research supports a role of inflammatory
22 processes in the toxic action of Libby Amphibole asbestos, consistent with the observed health
23 effects (Duncan et al., 2010; Hamilton et al., 2004). Taken together, the strong evidence in
24 human studies, defined exposure response relationships, and supportive animal studies provide
25 compelling evidence that exposure to Libby Amphibole asbestos causes nonmalignant
26 respiratory disease, including asbestosis, pleural thickening, and deficits in pulmonary function
27 associated with mineral fiber exposures. Existing data regarding cardiovascular effects and the
28 potential for autoimmune disease are limited.
29
This document is a draft for review purposes only and does not constitute Agency policy.
4-75 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.5.5. Mode-of-Action Information (Noncancer)
2 The precise mechanisms causing toxic injury from inhalation exposure to Libby
3 Amphibole asbestos have not been established. However, nearly all-durable mineral fibers with
4 dimensional characteristics that allow penetration to the terminal bronchioles and alveoli of the
5 lung have the capacity to induce pathologic response in the lung and pleural cavity (ATSDR,
6 200la: Witschi and Last 1996). The physical-chemical attributes of mineral fibers are important
7 in determining the type of toxicity observed. Fiber dimension (width and length), density, and
8 other characteristics such as chemical composition, surface area, solubility in physiological
9 fluids, and durability all play important roles in both the type of toxicity observed and the
10 biologically significant dose. Fibrosis results from a sequence of events following lung injury,
11 which includes inflammatory cell migration, edema, cellular proliferation, and accumulation of
12 collagen. Fibers do migrate to the pleural space, and it has been hypothesized that a similar
13 cascade of inflammatory events may contribute to fibrotic lesions in the visceral pleura.
14 Thickening of the visceral pleura is more often localized to lobes of the lung with pronounced
15 parenchymal changes, and it has also been hypothesized that the inflammatory and fibrogenic
16 processes within the lung parenchyma in response to asbestos fibers may influence the fibrogenic
17 process in the visceral pleura. The etiology of parietal plaques is largely unknown with respect
18 to mineral fiber exposure.
19 There is currently insufficient evidence to establish the noncancer mode of action for
20 Libby Amphibole asbestos. Limited in vitro studies have demonstrated oxidative stress
21 following Libby Amphibole asbestos exposures in various cell types (Duncan et al., 2010;
22 Hillegass et al.. 2010: Pietruska et al.. 2010: Blake et al.. 2007). Libby Amphibole asbestos
23 fibers increased intracellular ROS in both murine macrophages and human epithelial cells
24 (Duncan etal., 2010; Blake et al., 2007). Surface iron, inflammatory marker gene expression
25 was increased following exposure to Libby Amphibole asbestos in human epithelial cells
26 (Shannahan etal.. 201 Ib: Duncan et al.. 2010: Pietruska et al.. 2010) (see Table 4-18).
27 Tremolite studies demonstrate cytotoxicity in various cell culture systems (see Table 4-19).
28 The initial stages of any fibrotic response involve cellular proliferation, which may be
29 compensatory for cell death due to cytotoxicity. Analysis of cellular proliferation has
30 demonstrated both increases and decreases following exposure to asbestos fibers in vitro and in
31 vivo depending on the specific fiber or cell type (Mossman et al., 1985; Topping and Nettesheim,
This document is a draft for review purposes only and does not constitute Agency policy.
4-76 DRAFT—DO NOT CITE OR QUOTE
-------
1 1980). Other studies have focused on the activation of cell-signaling pathways that lead to
2 cellular proliferation following exposure to asbestos (Scapoli et al., 2004; Shuklaetal., 2003;
3 Dingetal.. 1999: Zanella et al.. 1996).
4 Although slightly increased compared to controls, cytotoxicity in murine macrophage
5 cells exposed to Libby Amphibole asbestos was decreased compared to other fiber types (Blake
6 et al., 2008). Cytotoxicity was slightly, but statistically significantly, increased compared to an
7 unexposed control at 24 hours post exposure to Libby Amphibole asbestos, while crocidolite
8 exposure resulted in even higher levels of cytotoxicity. No other in vitro study examined
9 cytotoxicity following exposure to Libby Amphibole asbestos, although an increase in apoptosis
10 was demonstrated in this same cell system (Blake et al., 2008). Recent studies in mice exposed
11 to Libby Amphibole asbestos demonstrated increased collagen deposition and collagen gene
12 expression, markers of fibrosis (Smartt et al., 2010; Putnam et al., 2008). Short-term studies in
13 rats also demonstrated an increased inflammatory response (Padilla-Carlin et al., 2011;
14 Shannahan et al., 201 la: Shannahan et al., 201 Ib). Tremolite and Libby Amphibole asbestos
15 exposure led to increases in both fibrosis in all but one animal study, supporting a role for
16 proliferation in response to these fibers. Taken together with studies on other asbestos fibers,
17 these data suggest that a cytotoxicity and cell proliferation may play a role in the noncancer
18 health effects following exposure to Libby Amphibole asbestos.
19 Although continued research demonstrates that the Libby Amphibole asbestos has
20 biologic activity consistent with the inflammatory action and cytotoxic effects seen with other
21 forms of asbestos, the data are not sufficient to establish a mode of action for the
22 pleura-pulmonary effects of exposure to Libby Amphibole asbestos.
23
24 4.6. EVALUATION OF CARCINOGENICITY
25 4.6.1. Summary of Overall Weight of Evidence
26 Under the EPA Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), Libby
27 Amphibole asbestos is carcinogenic to humans following inhalation exposure based on
28 epidemiologic evidence that shows a convincing association between exposure to Libby
29 Amphibole asbestos fibers and increased lung cancer and mesothelioma mortality (Larson et al.,
30 201 Ob: Moolgavkar et al., 2010: Sullivan, 2007: McDonald et al., 2004: Amandus and Wheeler,
31 1987: McDonald et al., 1986a). These results are further supported by animal studies that
This document is a draft for review purposes only and does not constitute Agency policy.
4-77 DRAFT—DO NOT CITE OR QUOTE
-------
1 demonstrate the carcinogenic potential of Libby Amphibole asbestos fibers and tremolite fibers
2 in rodent bioassays. As a durable mineral fiber of respirable size, this conclusion is consistent
3 with the extensive published literature that documents the carcinogenicity of amphibole fibers.
4 U.S. EPA's Guidelines for Carcinogenic Risk Assessment (U.S. EPA, 2005 a) indicate
5 that for tumors occurring at a site other than the initial point of contact, the weight of evidence
6 for carcinogenic potential may apply to all routes of exposure that have not been adequately
7 tested at sufficient doses. An exception occurs when there is convincing information (e.g.,
8 toxicokinetic data) that absorption does not occur by other routes. Information on the
9 carcinogenic effects of Libby Amphibole asbestos via the oral and dermal routes in humans or
10 animals is absent. The increased risk of lung cancer and mesothelioma following inhalation
11 exposure to Libby Amphibole asbestos has been established by studies in humans, but these
12 studies do not provide a basis for determining the risk from other routes of exposure.
13 Mesothelioma occurs in the pleural and peritoneal cavities and, therefore, is not considered a
14 portal-of-entry effect. However, the role of indirect or direct interaction of asbestos fibers in
15 disease at these extrapulmonary sites is still unknown. There is no information on the
16 translocation of Libby Amphibole asbestos to extrapulmonary tissues following either oral or
17 dermal exposure, and limited studies have examined the role of these routes of exposure in
18 cancer. Therefore, Libby Amphibole asbestos is considered carcinogenic to humans by the
19 inhalation route of exposure.
20
21 4.6.1.1. Synthesis of Human, Animal, and Other Supporting Evidence
22 Libby, MT workers have been the subject of multiple mortality studies demonstrating an
23 increased cancer mortality in relation to estimated fiber exposure. Occupational studies
24 conducted in the 1980s (Amandus and Wheeler, 1987) (McDonald et al., 1986a) as well as the
25 extended follow-up studies published in more recent years (Larson et al., 201 Ob: Sullivan, 2007;
26 McDonald et al., 2004) and additional analyses of the extended follow-up (Moolgavkar et al.,
27 2010) provide evidence of an increased risk of lung-cancer mortality and of mesothelioma
28 mortality among the workers exposed to Libby Amphibole asbestos in the Libby vermiculite
29 mining and processing operations. This pattern is seen in the lung cancer analyses using an
30 internal referent group in the larger follow-up studies (Larson et al., 201 Ob: Sullivan, 2007;
31 McDonald et al., 2004), with cumulative exposure analyzed using quartiles or as a continuous
This document is a draft for review purposes only and does not constitute Agency policy.
4-78 DRAFT—DO NOT CITE OR QUOTE
-------
1 measure, and in the studies reporting analyses using an external referent group [i.e., standardized
2 mortality ratios (Sullivan, 2007; Amandus and Wheeler, 1987; McDonald et al., 1986a).
3 McDonald et al. (2004) also reported increasing risk of mesothelioma across categories of
4 exposure; the more limited number of cases available in earlier studies precluded this type of
5 exposure-response analysis. This association is also supported by the case series of
6 11 mesothelioma patients among residents in or around Libby, MT, and among family members
7 of workers in the mining operations (Whitehouse et al., 2008).
8 Although experimental data in animals and data on toxicity mechanisms are limited for
9 Libby Amphibole asbestos, tumors were observed in tissues similar to those in humans (e.g.,
10 mesotheliomas, lung cancer) indicating the existing data are consistent with the cancer effects
11 observed in humans exposed to Libby Amphibole asbestos. Smith (1978) reported increased
12 incidence of mesotheliomas in hamsters after intrapleural injections of Libby Amphibole
13 asbestos. Additionally, studies in laboratory animals (rats and hamsters) exposed to tremolite via
14 inhalation (Bernstein et al., 2005; Bernstein et al., 2003; Davis et al., 1985), intrapleural injection
15 (Roller et al.. 1997. 1996: Davis etal.. 1991: Wagner et al.. 1982: Smith etal.. 1979) or
16 implantation (Stanton et al., 1981) have shown increases in mesotheliomas and lung cancers.
17 Tremolite from various sources was used and varied in fiber content and in potency (see
18 Section 4.2, Appendix D). Although McConnell et al. (1983b) observed no increase in
19 carcinogenicity following oral exposure to nonfibrous tremolite, the ability of this study to
20 inform the carcinogenic potential of fibrous tremolite through inhalation is unclear, and these
21 study results contribute little weight to the evaluation of the carcinogenicity of fibrous Libby
22 Amphibole asbestos.
23 The available mechanistic information suggests Libby Amphibole asbestos induces
24 effects that may play a role in carcinogenicity (see Section 4.3.4, Appendix D). Several in vitro
25 studies have demonstrated oxidative stress and genotoxicity following Libby Amphibole
26 asbestos exposures in various cell types (Duncan etal., 2010: Hillegass etal., 2010: Pietruska et
27 al., 2010: Blake et al., 2007). Libby Amphibole asbestos increased intracellular ROS in both
28 murine macrophages and human epithelial cells (Duncan etal., 2010: Blake et al., 2007).
29 Additionally, surface iron, inflammatory marker gene expression and aneugenic micronuclei
30 were increased following exposure to Libby Amphibole asbestos in human epithelial cells
31 (Duncan et al., 2010: Pietruska et al., 2010). Tremolite studies demonstrate cytotoxic and
This document is a draft for review purposes only and does not constitute Agency policy.
4-79 DRAFT—DO NOT CITE OR QUOTE
-------
1 clastogenic effects (e.g., micronucleus induction and chromosomal aberrations) of the fibers in
2 various cell culture systems.
3 In summary, the epidemiologic data demonstrate an association between exposure to
4 Libby Amphibole asbestos and increased cancer risk. Supporting evidence of carcinogenic
5 potential was observed in the limited number of laboratory animal studies exposed to Libby
6 Amphibole asbestos or tremolite (see Tables 4-15 and 4-16 summarizing in vivo studies).
7 Overall, the available evidence supports the conclusion that Libby Amphibole asbestos is
8 carcinogenic to humans.
9
10 4.6.2. Mode-of-Action Information
11 4.6.2.1. Description of the Mode-of-Action Information
12 EPA guidance provides a framework for analyzing the potential mode(s) of action by
13 which physical, chemical, and biological information is evaluated to identify key events in an
14 agent's carcinogen!city (U.S. EPA, 2005a). Agents can work through more than one mode of
15 action (MOA), and MOA can differ for various endpoints (e.g., lung cancer versus
16 mesothelioma). Reasonably, the analysis of a MOA would start with some knowledge of an
17 agent's biological activity that leads to cellular transformation resulting in carcinogenicity.
18 Although early steps in the process often can be identified, carcinogenicity is a complex process
19 resulting from multiple changes in cell function. Due to the limited data available specific to
20 Libby Amphibole asbestos, the mode of action of Libby Amphibole asbestos for lung cancer and
21 mesothelioma following inhalation exposure cannot be established.
22 Research on various types of mineral fibers supports the role of multiple biologic
23 responses following exposure to asbestos in general (i.e., chronic inflammation, generation of
24 ROS, direct genotoxicity, and cytotoxicity and cellular proliferation) in the carcinogenic
25 response to mineral fibers. However, the complexities of fiber toxicity make it difficult to define
26 modes of action for asbestos, in general [as reviewed in Aust et al. (2011): Mossman et al.
27 (2011): Huang et al. (2011): Bunderson-Schelvan et al. (2011): Broaddus et al. (2011)1. Further,
28 limitations in early study design and presentation of the results hinder understanding of mode
29 and mechanism of action for specific fiber types. Most studies lack information on the
30 characterization of fibers and cell types used, hindering understanding of the mode(s) of action.
31 Particularly of importance is the route of exposure utilized in the in vivo studies, as results
This document is a draft for review purposes only and does not constitute Agency policy.
4-80 DRAFT—DO NOT CITE OR QUOTE
-------
1 obtained from nonphysiologically relevant routes of exposure (i.e., intraperitoneal injection,
2 gelatin implant) may not accurately reflect the response in occupational inhalation exposures.
3 Occupational studies demonstrate human health effects (e.g., lung cancer, mesothelioma)
4 following exposure to Libby Amphibole asbestos. Although the limited mechanistic data
5 demonstrate biological effects similar to those of other mineral fibers following exposure to
6 Libby Amphibole asbestos, the existing literature are insufficient to establish a mode of action
7 for Libby Amphibole asbestos for lung cancer or mesothelioma. These biological effects
8 following exposure to Libby Amphibole asbestos and/or tremolite are demonstrated in a limited
9 number of laboratory animal and in vitro studies. Multiple key events for one particular MOA
10 have not been identified; therefore, the mode of action for Libby Amphibole asbestos
11 carcinogenicity cannot be established.
12
13 4.6.2.2. Application of the Age-Dependent Adjustment Factors
14 As described above, the mode of action for Libby Amphibole asbestos is unknown. The
15 weight of evidence does not support a mutagenic mode of action for Libby Amphibole asbestos
16 carcinogenicity. Therefore, according to EPA's Supplemental Guidance for Assessing
17 Susceptibility from Early-Life Exposure to Carcinogens (U.S. EPA, 2005b), the application of
18 the Age-Dependent Adjustment Factors is not recommended.
19
20 4.7. SUSCEPTIBLE POPULATIONS
21 Certain populations may be more susceptible to adverse health effects from exposure to
22 Libby Amphibole asbestos. Because the adverse health effects resulting from exposure to Libby
23 Amphibole asbestos have been, for the most part, studied in occupational cohorts of adult white
24 men (see Sections 4.1.1 and 4.1.3), there is limited information on the effects to a broader
25 population. A few studies, however, have examined health effects resulting from
26 nonoccupational exposure in other age groups, in other genders (i.e., females), and in different
27 race or ethnicity groups. The data from these studies could inform whether any differential risk
28 exists for these groups (see Sections 4.1.2 and 4.1.4). However, it should be noted that the
29 ability to distinguish true differences from chance variation in effect estimates is related to the
30 sample size and statistical power, which, in most cases, is quite limited in these studies. In
31 addition, genetic polymorphisms, preexisting health conditions, and differences in nutritional
This document is a draft for review purposes only and does not constitute Agency policy.
4-81 DRAFT—DO NOT CITE OR QUOTE
-------
1 status may alter an individual's response to Libby Amphibole asbestos. Finally, coexposures to
2 other substances (e.g., tobacco smoke or paniculate matter) may increase an individual's risk of
3 adverse health effects from exposure to Libby Amphibole asbestos. Where data are available,
4 each of these factors is discussed below with respect to increased susceptibility to noncancer
5 effects and cancer from exposure to Libby Amphibole asbestos, and where information specific
6 to Libby Amphibole asbestos is not available, the general literature on the toxicity of mineral
7 fibers is briefly referenced.
8 There are also factors that may influence one's exposure potential to asbestos based on
9 lifestage or other defined population. For example, children spend more hours outside and may
10 engage in activities which impact exposure level compared to adults (U.S. EPA, 2006b: NRC,
11 1993). Because lifestage and activity patterns can increase the potential for health effects from
12 exposure, these factors define those who may be more susceptible to health effects due to greater
13 exposure. Section 2.3 discusses this exposure potential, including how children workers,
14 household contacts and residents may be exposed to Libby Amphibole asbestos.
15
16 4.7.1. Influence of Different Lifestages on Susceptibility
17 Individuals at different lifestages differ from one another physiologically, anatomically,
18 and biochemically. Individuals in early and later lifestages differ markedly from adulthood in
19 terms of body composition, organ function, and many other physiological parameters, which can
20 influence the toxicokinetics and toxicodynamics of chemicals and their metabolites in the body
21 (Guzelian et al., 1992). This also holds true for mineral fibers, including asbestos fibers (see
22 Section 3). This section presents and evaluates the literature on how individuals in early or later
23 lifestages might respond differently and thus potentially be more susceptible to adverse health
24 effects of Libby Amphibole asbestos exposure.
25
26 4.7.1.1. Lifestage Susceptibility
27 Humans in early lifestages (i.e., conception through adolescence) can have unique
28 susceptibilities compared to those in later lifestages because they undergo rapid physiological
29 changes during critical periods of development (Selevan et al., 2000). Furthermore, they are
30 often exposed to xenobiotics via unique exposure pathways (i.e., transplacental transfer and
31 breast milk ingestion) (U.S. EPA. 2006b: NRC. 1993). Although no data exist for Libby
This document is a draft for review purposes only and does not constitute Agency policy.
4-82 DRAFT—DO NOT CITE OR QUOTE
-------
1 Amphibole asbestos, limited observations in stillborn infants indicate occurrence of
2 transplacental transfer of tremolite (Hague et al., 1998; 1996) and other asbestos and nonasbestos
3 fibers (Hague et al.. 1998: Hague et al.. 1996: Hague et al.. 1992: Hague et al.. 1991). Haque et
4 al. (1992) hypothesized that maternal health conditions might influence the translocation of
5 fibers, as some of the mothers had preexisting health conditions. Transplacental transfer of
6 asbestos also has been demonstrated in animals following maternal exposure by gavage (Haque
7 et al., 2001) or injection (Haque and Vrazel, 1998; Cunningham and Pontefract 1974) (see
8 Section 3). These studies did not evaluate sources or levels of exposure, and injection studies are
9 a less relevant route of exposure than inhalation. Based on these studies, Libby Amphibole
10 asbestos fibers may be transferred through the placenta, resulting in prenatal exposure at any
11 stage of fetal development.
12 Increased lung deposition of fibers in children compared with adults has been observed
13 (Bennett et al.. 2008: Isaacs and Martonen, 2005: Asgharian et al.. 2004: Phalen and Oldham,
14 2001: Oldham et al.. 1997: Schiller-Scotland et al.. 1994: Phalen et al.. 1985). Nasal deposition
15 of particles was shown to be lower in children compared to adults—particularly during exercise
16 (Becquemin et al., 1991). The lung and nasal depositional differences are due in part to
17 structural differences across lifestages, which can change the depositional pattern of different
18 fiber sizes and possibly alter the site of action and result in differential clearance and subsequent
19 health effects. It is unclear, however, whether the lung surface, body weight, inhalation volume,
20 or exposure patterns are most determinative of dose. One study reported that the ratio of lung
21 surface area to body weight does not differ considerably for a 10-month old, a 9-year old, and an
22 adult (Short, 1952). Another study suggested that deposition of fine particles (2-|im mass
23 median aerodynamic diameter, which is in the size range of those for Libby Amphibole asbestos
24 reported in Table 2-2) in the lung is increased for overweight (>95* percentile BMI) children
25 who breathe more at rest compared to underweight children (<25th percentile BMI) (Bennett and
26 Zeman, 2004).
27 There are few studies analyzing noncancer outcomes in children exposed to Libby
28 Amphibole. A Libby medical screening program collected data on 7,307 participants, including
29 600 children aged 10-17 years old, representing 8.2% of the cohort (Peipins et al., 2003).
30 Pulmonary function tests showed that none of these children had moderate or severely restricted
31 lung function (ATSDR, 2002, 2001b). This study also studied chest radiographs for those
This document is a draft for review purposes only and does not constitute Agency policy.
4-83 DRAFT—DO NOT CITE OR QUOTE
-------
1 18 years old or older (Noonan et al.. 2006: Peipins et al.. 2003: ATSDR, 200 Ib), but X-rays were
2 not conducted on children. In addition, the prevalence of some self-reported respiratory
3 symptoms among 10-29-year-old adolescents and young adults was associated with certain
4 exposure pathways. These participants were < age 18 in 1990 when the mining/milling
5 operations closed (Vinikoor et al., 2010). Understanding of the community health effects and the
6 examination of the potential progression of adverse health effect in this community would
7 benefit from additional research to establish the clinical significance of these findings. No other
8 studies of noncancer outcomes in early lifestages of humans or experimental animals exposed to
9 Libby Amphibole asbestos have been reported.
10 For exposure to other types of asbestos, studies have reported noncancer outcomes in
11 early lifestages. Those in the very young include reports of stillbirth (Hague et al., 1998: 1996)
12 and death among infants (age 1-27 months) due to sudden infant death syndrome and
13 bronchopulmonary dysplasia (Hague and Kanz, 1988). These studies found higher levels of
14 asbestos in the lungs of those who died compared to controls. In the infant study, the authors
15 speculate that either there was a preexisting abnormal lung physiology in these children that may
16 contribute to a reduced ability to clear fibers from the lung, or that the children could have an
17 increased exposure to asbestos (Hague and Kanz, 1988). Those in older children include reports
18 of pleural and diaphragmatic calcifications (Epler et al., 1980) and altered immune and
19 respiratory conditions (Shtol' et al., 2000).
20 In experimental animals, offspring of rats exposed to tremolite had decreased body
21 weight gain at weaning and 8-weeks-old compared to controls (NTP, 1990b: McConnell et al.,
22 1983b). This was also observed in some similar studies of other forms of asbestos (NTP, 1990a,
23 1988. 1985: McConnell et al.. 1983b) but not in others (McConnell et al.. 1983a: NTP. 1983).
24 Embryonic toxicity was observed in a few experimental animal studies. Crocidolite injected into
25 pregnant mice resulted in altered limb differentiation in cultured embryos [Krowke et al. (1983),
26 abstract], and chrysotile in drinking water given to pregnant mice resulted in decreased
27 postimplantation survival in cultured embryos (Schneider and Maurer, 1977): however, pregnant
28 mice exposed to chrysotile in drinking water did not affect in vivo embryonic survival
29 (Schneider and Maurer, 1977).
30 It is possible that early lifestage exposure may increase the risk of noncancer outcomes in
31 adulthood compared to adult exposure. After tremolite exposure during childhood, one study
This document is a draft for review purposes only and does not constitute Agency policy.
4-84 DRAFT—DO NOT CITE OR QUOTE
-------
1 reported altered immunity in adulthood (Zerva et al., 1989), and one study described a case
2 report of asbestosis in adulthood (Voisin et al., 1994). Another study also reported an increased
3 risk of asbestosis after childhood exposure to asbestos from parental occupational exposure to
4 asbestos (Kilburn etal., 1985). To address the potential for increased susceptibility to cancer
5 from early lifetime exposures, one needs to consider if there is evidence of differential health
6 effects such as increased potency from early lifetime exposure, decreased latency based on the
7 age of exposure, or cancers observed with early lifetime exposures not seen with adult exposures.
8 There are no published reports that can directly answer these questions for exposure to Libby
9 Amphibole asbestos.
10 While cancers in adults have been documented following exposure to Libby Amphibole
11 asbestos, similar reports describing childhood cancers resulting from this exposure have not been
12 identified. Few cancers occurring in children have been documented in children exposed to any
13 form of asbestos. Examples of cases include a 17-year old exposed to chrysotile and tremolite
14 (Andrion et al., 1994) and a 3-year old exposed to chrysotile (Lieben and Pistawka, 1967), both
15 of whom developed mesothelioma. However, childhood mesothelioma, in particular, may have
16 an etiology that is different from that of the disease that is seen in adults (Cooper et al., 1989).
17 No cancer bioassays have been performed in juvenile animals exposed to Libby Amphibole
18 asbestos.
19 Of the 11 Libby Amphibole asbestos-related mesothelioma cases described by
20 Whitehouse et al. (2008), 2 reported potential exposure scenarios that were limited to childhood,
21 and both of these were diagnosed at a relatively young age at diagnosis (48, compared with 52 to
22 82 years of age for the other nine cases). Although these case studies support the link between
23 exposure to Libby Amphibole asbestos and mesothelioma, it is unclear if children are more
24 susceptible than adults.
25 Case reports of exposure to tremolite during childhood, and subsequent diagnosis of
26 mesothelioma in adulthood (Senyigit et al., 2000; Schneider et al., 1998; Sakellariou et al., 1996;
27 Rev etal., 1993; Magee etal., 1986), support the limited data summarized above for Libby
28 Amphibole asbestos. Additional case studies of mesothelioma after childhood exposure to other
29 types of asbestos are available (Yano et al., 2009; Ascoli et al., 2003; Magnani et al., 2001; Rom
30 etal., 2001: Schneider et al., 1996a: Schneider et al., 1996b: Schneider et al., 1995: Roguin et al.,
31 1994: Cazzadori et al., 1992: Inase et al., 1991: Kane et al., 1990: Li et al., 1989: Martensson et
This document is a draft for review purposes only and does not constitute Agency policy.
4-85 DRAFT—DO NOT CITE OR QUOTE
-------
1 al.. 1984: Wassermann et al.. 1980: Lietal.. 1978: Anderson et al.. 1976: Wagner etal.. 1960).
2 These studies, however, do not clarify whether exposure during childhood yields different
3 adverse health effects compared with exposure during adulthood.
4 In experimental studies, the offspring of rats orally exposed to nonfibrous tremolite did
5 not demonstrate an increase in tumors compared to controls (NTP, 1990b: McConnell et al.,
6 1983b). Similar studies of other forms of asbestos did report an increase of various neoplasms in
7 the offspring (NTP. 1990a, 1988. 1985: McConnell et al.. 1983a: McConnell et al.. 1983b), but
8 another study reported none (NTP, 1983).
9 Studies of exposure to other types of asbestos have attempted to determine if exposure to
10 asbestos in early life results in an increased risk of developing cancer. An early study in the
11 United Kingdom described occupational exposure to chrysotile, crocidolite, and amosite for a
12 group of 900 women. First exposure from ages 15-24 years led to a higher relative mortality
13 risk for lung and pleural cancer compared with women who were first exposed at older ages
14 (SMR 30 based on 12 observed and 0.4 expected, SMR 8 based on 4 observed and 0.5 expected,
15 and SMR 6.7 based on 6 observed and 0.9 expected in the first exposure at ages 15-24, 25-34,
16 and >35 years, respectively) (Newhouse et al., 1972). A study by Hansen et al. (1998) in
17 Wittenoom, Western Australia examined 27 individuals diagnosed with mesothelioma who had
18 been environmentally exposed to crocidolite (i.e., residents of the town but not directly employed
19 in the area's crocidolite mining and milling industry); 11 of these subjects were children
20 <15 years old at the time of exposure. One-third of all the subjects were less than 40 years old
21 when diagnosed, but the authors found no increase in mesothelioma mortality rates when
22 analyzed by age at first exposure. However, risk was significantly increased based on time from
23 first exposure, duration of exposure, and cumulative exposure (Hansen et al., 1998). Additional
24 studies of this cohort found that the mesothelioma mortality rate was lower for those first
25 exposed (based on age residence in the area began) to crocidolite at ages <15 years (n = 24;
26 mesothelioma mortality rate 47 per 100,000 person-years) compared with those first exposed at
27 ages >15 years (« = 43; mesothelioma mortality rate 112 per 100,000 person-years) (Reid et al.,
28 2007). The hazard ratio for age at first residential exposure of >15 years compared with
29 <15 years was 3.83 (95% CI: 2.19, 6.71), adjusting for cumulative exposure, gender, and an
30 interaction term for gender and cumulative exposure.
This document is a draft for review purposes only and does not constitute Agency policy.
4-86 DRAFT—DO NOT CITE OR QUOTE
-------
1 Based on these very limited and inconclusive studies on other forms of asbestos, no
2 conclusions can be drawn about differential risk of adverse health effects after early lifestage
3 exposure to Libby Amphibole asbestos compared to exposure during adulthood. It is unknown
4 whether early lifestage exposure compared to adult exposure increases susceptibility for adult
5 cancers, as measured by increased incidence, severity, or disease progression, or by decreased
6 latency.
7 Later lifestage is generally defined as >65 years old. Because pulmonary function
8 (volume and rate of breathing) decreases with age (Weiss, 2010), increased deposition of fibers
9 in the lung from exposures in later lifestages is unlikely. Clearance of fibers from the lung might
10 be reduced, however, as older adults have a less effective cough reflex and strength and the cilia
11 are less able to move mucus up and out of the airway (U.S. EPA, 2006a). Older adults could be
12 more susceptible to the effects of Libby Amphibole asbestos due to the gradual age-related
13 decline in physiological processes. Additionally, decreased immune function, increased genetic
14 damage, and decreased DNA repair capacity can result in increased susceptibility with age (U.S.
15 EPA, 2006a). These age-associated alterations could decrease fiber-induced DNA damage repair
16 but might also reduce the incidence of fiber-induced DNA damage due to decreased
17 phagocytosis or inflammation. Specific data pertaining to age-varying effects of Libby
18 Amphibole asbestos on these processes are not available.
19 Because the risk of many types of noncancer effects increases with age, an increasing rate
20 of specific diseases with increasing age can be expected among individuals exposed at some
21 point in their lives to Libby Amphibole asbestos. Radiographic tests among those exposed to
22 Libby Amphibole show that older age, which may be highly correlated with time since first
23 exposure in some occupational settings, is one of the factors most associated with pleural or
24 interstitial abnormalities (Rohs et al., 2008; Horton et al., 2006; Muravov et al., 2005; Peipins et
25 al.. 2003: AT SDR. 200 Ib: Amandus et al.. 1987b: McDonald et al.. 1986b: Lockevetal., 1984).
26 Abnormal radiographs also increase with age in general population studies (Pinsky et al., 2006).
27 In the community health screening study, an increased risk of rheumatoid arthritis among
28 individuals ages >65 years was observed in relation to several measures reflecting exposure to
29 Libby Amphibole asbestos (e.g., worked for W.R. Grace, used vermiculite for gardening)
30 (Noonan, 2006). However, the available studies do not provide a basis for evaluating the timing
31 of the exposure in relation to these outcomes. No conclusions can be drawn about differential
This document is a draft for review purposes only and does not constitute Agency policy.
4-87 DRAFT—DO NOT CITE OR QUOTE
-------
1 risk of noncancer after later lifestage exposure to Libby Amphibole compared to exposure earlier
2 in life.
3 No studies assessing the carcinogenic effect of exposures occurring in older age groups
4 are available for Libby Amphibole asbestos. It should be noted that observed health effects
5 among individuals exposed to Libby Amphibole asbestos are likely to increase with increasing
6 age due to the long latency period for the exposure response for asbestos and lung cancer and
7 other chronic diseases. However this type of observation would not directly address the question
8 of whether exposures at older ages have a stronger or weaker effect compared with exposures at
9 younger ages.
10
11 4.7.2. Influence of Gender on Susceptibility
12 A discussion of gender-related differences in risk from asbestos exposure raises several
13 important issues, such as gender-related differences in exposure patterns, physiology, and
14 dose-response (Smith, 2002). For example, nasal breathing filters out particles, and men tend to
15 breathe less through their nose during exercise than women do (Bennett et al., 2003). Bennett
16 et al. (1996) showed a gender difference in fractional deposition (defined as the ratio of particles
17 not exhaled to total particles inhaled) of particles 2 jim in mass median aerodynamic diameter.
18 This particle diameter is within the range of Libby Amphibole asbestos particles reported in
19 Table 2-2. This study found that, in general, women had a greater retention of particles
20 compared to men because men had higher ventilation rates compared to women; however, the
21 overall deposition rate was higher in the men (Bennett et al., 1996).
22 Most occupational studies for Libby Amphibole asbestos have examined the effects of
23 exposure only in men (Moolgavkar et al., 2010; Sullivan, 2007; McDonald et al., 2004;
24 Amandus et al., 1988; Amandus et al., 1987a: Amandus and Wheeler, 1987; McDonald et al.,
25 1986a: McDonald et al., 1986b). There is limited information specifically on women exposed to
26 Libby Amphibole asbestos. In the Libby, MT community studies, no gender-related trends in
27 mortality due to lung or digestive cancer were observed (ATSDR, 2000). These limited data do
28 not provide a basis for drawing conclusions regarding gender-related differences in adverse
29 health effects from Libby Amphibole asbestos.
30
This document is a draft for review purposes only and does not constitute Agency policy.
4-88 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.7.3. Influence of Race or Ethnicity on Susceptibility
2 Race and ethnicity often are used in medical and epidemiological studies to define
3 various groups of the population. These categories could be surrogates for differences in
4 exposure (e.g., occupation, socioeconomics, behavior) or biology (e.g., physiology, genetics), in
5 which case these factors may play a role in susceptibility as well. Nasal structure and lung
6 architecture can influence the depositional patterns for both particles and fibers. One study of
7 18 Caucasians (ages 8 to 30 years) and 14 African Americans (ages 8 to 25 years) reported
8 increased ventilation rates during exercise in the African Americans (matched on sex, age,
9 height, and weight) (Cerny, 1987). Another study (11 Caucasians and 11 African Americans,
10 ages 18 to 31 years) reported decreased nasal deposition efficiency (for particle sizes of 1-2 jim,
11 which is in the range of those for Libby Amphibole asbestos reported in Table 2-2) in African
12 Americans compared to Caucasians (Bennett and Zeman, 2005). Furthermore, nasal breathing
13 during exercise occurred less in Caucasians compared to African Americans in this study
14 (Bennett et al.. 2003).
15 Of the occupational and residential studies for Libby Amphibole asbestos, the vast
16 majority of subjects with known race were white, precluding the ability to conduct an analysis of
17 racial and ethnicity-related differences in the mortality risks within the Libby worker cohort. In
18 a study of occupational exposure to chrysotile asbestos in a textile factor, lung-cancer mortality
19 risk in relation to exposure was lower in nonwhite males (0.84, 95% CI: 0.52-1.27) compared to
20 white males (2.34, 95% CI: 1.94-2.79), although a statistically significant increase in SMR was
21 observed for nonwhite males at high exposure levels (>120 fiber-years/mL) (Hein et al., 2007).
22 This observed difference could be due to a lower prevalence of smoking among nonwhite
23 compared with white males (Hein et al., 2007).
24
25 4.7.4. Influence of Genetic Polymorphisms on Susceptibility
26 XRCC1 is a DNA damage repair gene. A recent study demonstrated that
27 XRCC1-deficient cells exposed to Libby Amphibole or crocidolite asbestos demonstrated
28 increased levels of micronuclei induction (Pietruska et al., 2010). Two other studies examined
29 XRCC1 polymorphisms in relation to disease risk with other types of asbestos exposure. Zhao
30 et al. (2006) found no association between XRCC1 polymorphisms and asbestosis in
31 asbestos-exposed workers. A study by Dianzani et al. (2006), however, did find an association
This document is a draft for review purposes only and does not constitute Agency policy.
4-89 DRAFT—DO NOT CITE OR QUOTE
-------
1 between XRCC1 and asbestos-induced lung disease in a population exposed to asbestos
2 pollution. Further work is necessary, with clear definitions of patient populations and their
3 exposure levels, so that these studies and others can be compared to determine if XRCC1
4 polymorphisms increase susceptibility to adverse health effects following exposure to Libby
5 Amphibole asbestos.
6 SODs are free radical scavengers that dismutate superoxide anion to oxygen and
7 hydrogen peroxide. SODs are expressed in most cell types exposed to oxygen. Several common
8 forms of SODs occur and are named by the protein cofactor: copper/zinc, manganese, iron, or
9 nickel. A recent study observed no significant alterations in levels of intracellular SOD
10 following a 3 hour exposure to Libby Amphibole asbestos in mice (Blake et al., 2007). Other
11 studies in humans and mice have examined SOD expression in relation to other types of asbestos
12 exposure. Manganese superoxide dismutase activity was elevated in biopsies of human
13 asbestos-associated malignant mesothelioma, although no genotypic differences were found to
14 be related to this change in activity (Hirvonen et al., 2002). Other studies have focused on the
15 role of extracellular superoxide dismutase (EcSOD) and asbestos-induced pulmonary disease
16 (Kliment et al.. 2009: Gao et al.. 2008: Fattman et al.. 2006: Tan et al.. 2004). These studies
17 have suggested a protective effect of EcSOD, with mice that lack this form of SOD having
18 increased sensitivity to asbestos-induced lung injury (Fattman et al., 2006). Familial studies
19 showing unusually high incidence of mesothelioma suggest that genetic factors might play a role
20 in the etiology of mesothelioma (Ugolini et al., 2008; Huncharek, 2002; Roushdy-Hammady et
21 al., 2001), although whether a genetic factor or a common environmental element leads to the
22 similar responses in these families is difficult to determine. Increased interest in the role of
23 genetic factors in asbestos-related health outcomes has led to several analytical studies on
24 specific genetic polymorphisms. A review of 24 published reports (19 studies) discusses the
25 current state of knowledge regarding genetic susceptibility associated with asbestos-related
26 diseases (in particular, malignant pleural mesothelioma). Results from several studies
27 demonstrated an association between asbestosis-related diseases and GSTMl-null
28 polymorphism, whereas results for other polymorphisms were conflicting (Neri et al., 2008).
29 Some polymorphisms discussed in Neri et al. (2008) are in genes for 7V-acetyl-transferase 2;
30 glutathione-s-transferases (GSTs); SOD; CYP1A1, CYP2D6; neurofibromatous 2 (Nf2); p53;
This document is a draft for review purposes only and does not constitute Agency policy.
4-90 DRAFT—DO NOT CITE OR QUOTE
-------
1 and XRCC1. Although occupational asbestos exposure was assessed, the type of asbestos is
2 generally unknown in these studies.
3 Limited animal studies have examined the role of genetic variations related to asbestos
4 exposure, including specific signaling pathways (Shukla et al., 2007), DNA damage repair (Lin
5 et al., 2000; Ni et al., 2000), and tumor suppressor genes (Vaslet et al., 2002; Kleymenova et al.,
6 1997; Marsella et al., 1997). Genetic alterations of particular interest for mesothelioma include
7 those involved in tumor suppression (p53, Nf2) and oxidative stress (SOD, GSTs). Nf2 and p53
8 are frequently altered in mesotheliomas, but no consistent mutations have been found (Cheng et
9 al., 1999; Mayall et al., 1999; Bianchi et al., 1995). Alterations in expression of antioxidant
10 enzymes like SOD and GST in mesothelioma can yield cells more resistant to oxidative stress as
11 compared to normal cells due to increased antioxidant activity (Ramos-Nino et al., 2002;
12 Rahman and MacNee, 1999). No studies that examine the role of cell-cycle control genes were
13 found following exposure to Libby Amphibole asbestos. Additionally, no information on other
14 genetic polymorphisms in relation to disease risk among those exposed to Libby Amphibole
15 asbestos was identified in the available literature.
16
17 4.7.5. Influence of Health Status on Susceptibility
18 Preexisting health conditions could potentially alter the biological response to asbestos
19 exposure. Mesothelioma risk has been hypothesized to be related to immune impairment
20 (Bianchi and Bianchi, 2008) and simian virus 40 exposure in humans (Carbone et al., 2007;
21 Kroczynska et al., 2006: Cristaudo et al., 2005: Foddis et al., 2002: Bocchetta et al., 2000:
22 Mavall et al., 1999). Coexposure to asbestos and SV40 has been associated with p53-related
23 effects in vitro (Foddis et al., 2002: Bocchetta et al., 2000: Mavall et al., 1999), and cell signaling
24 aberrations in vivo (Kroczynska et al., 2006: Cristaudo et al., 2005). However, the influence on
25 cancer risk is unknown, as these lines of research are not fully developed and have not been
26 applied specifically to Libby Amphibole asbestos.
27 Obesity can compromise inhalation exposure, as increased particle deposition in the lungs
28 of overweight children (Bennett and Zeman, 2004) and adults (Graham et al., 1990) has been
29 observed. Individuals with respiratory diseases could have compromised lung function that
30 alters inhalation exposure to Libby Amphibole asbestos. For example, individuals with chronic
31 obstructive pulmonary disease have increased inhalation volume (Phalen et al., 2006) and
This document is a draft for review purposes only and does not constitute Agency policy.
4-91 DRAFT—DO NOT CITE OR QUOTE
-------
1 increased fine particle deposition (Phalen et al., 2006; Bennett et al., 1997; Kim andKang, 1997)
2 and retention (Regnis et al., 2000). Similarly, studies have reported an increase in coarse particle
3 (aerodynamic diameter >5 jim) deposition in individuals with cystic fibrosis (Brown and
4 Bennett 2004; Brown et al., 2001). For people exposed to Libby Amphibole asbestos, an
5 increased risk for interstitial lung abnormalities was observed for those with a history of
6 pneumonia (Peipins et al., 2003). In another study, bronchial asthma was examined as a
7 potential confounding variable for asbestos-related effects on pulmonary function, although no
8 confounding was observed (Whitehouse, 2004).
9
10 4.7.6. Influence of Lifestyle Factors on Susceptibility
11 No studies were identified that examined lifestyle factors specifically with respect to
12 Libby Amphibole asbestos. Lifestyle factors such as exercise, nutritional status, and smoking
13 habits could affect the biological effects of asbestos exposure through various mechanisms. For
14 example, those with more physically demanding jobs or those who regularly engage in vigorous
15 exercise might experience increased lung deposition from fine particles or fibers compared to
16 those with a more sedentary lifestyle (Phalen et al., 2006; Becquemin et al., 1991). Randomized
17 controlled trials of vitamin supplementation (beta-carotene and retinol) have been conducted for
18 asbestos-related lung cancer, but results do not support a protective effect (Cullen et al., 2005)
19 For lung cancer, a synergistic relationship between cigarette smoking and asbestos
20 exposure has been demonstrated (Wraith and Mengersen, 2007; Hammond et al., 1979; Selikoff
21 and Hammond, 1979). Research has suggested that asbestos fibers might also enhance the
22 delivery of multiple carcinogens in cigarette smoke, and that cigarette smoking decreases the
23 clearance mechanisms in the lungs and could, therefore, lead to an increase in fiber presence in
24 the lungs (Nelson and Kelsey, 2002). Smoking likely causes genetic alterations associated with
25 lung cancer (Landi et al., 2008) that might increase the carcinogenic risk from exposure to
26 asbestos. Benzo(a)pyrene, a component of tobacco, also has been observed to enhance the
27 carcinogenic effects of asbestos (Loli et al., 2004; Kimizuka et al., 1987; Mossman et al., 1984;
28 DiPaolo et al.. 1983: Mossman et al.. 1983: Reissetal., 1983).
29
This document is a draft for review purposes only and does not constitute Agency policy.
4-92 DRAFT—DO NOT CITE OR QUOTE
-------
1 4.7.7. Susceptible Populations Summary
2 A very limited amount of information is available on exposure to Libby Amphibole
3 asbestos early in life that could lead to increased risk of asbestos-induced disease later in life.
4 Due to the long latency period of some diseases in relation to asbestos exposure in general,
5 adverse effects may be more likely to be observed with an increase in age. This assumption
6 requires further investigation. The number of women who have been occupationally exposed to
7 Libby Amphibole asbestos is very small, and health risks have not been evaluated specifically
8 for this group. Differences between men and women in residential sources and types of exposure
9 (e.g., types of activities done in the household) also preclude the possibility of drawing
10 conclusions regarding the relative susceptibility of women compared with men to health effects
11 of exposure to Libby Amphibole asbestos. Similarly, sufficient data are not available to draw
12 conclusions regarding racial or ethnic variation in susceptibility to diseases caused by exposure
13 to Libby Amphibole asbestos. In addition, the potential modifying effects of genetic
14 polymorphisms, preexisting health conditions, nutritional status, and other lifestyle factors have
15 not been studied, specifically as related to exposure of Libby Amphibole asbestos and health
16 outcomes.
This document is a draft for review purposes only and does not constitute Agency policy.
4-93 DRAFT—DO NOT CITE OR QUOTE
-------
1 5. EXPOSURE-RESPONSE ASSESSMENT
2 5.1. ORAL REFERENCE DOSE (RfD)
3 Data are unavailable to characterize the toxic effects of Libby Amphibole asbestos22
4 following oral exposure. Thus, an oral reference dose is not derived.
5
6 5.2. INHALATION REFERENCE CONCENTRATION (RfC)
7 5.2.1. Choice of Principal Study and Critical Effect
8 Studies in humans have shown radiographic evidence of health effects on the lung and
9 pleura (a thin tissue surrounding the lung and lining the chest cavity) such as pleural thickening
10 and fibrosis of the lung and pleura in exposed workers (Larson etal., 2010a: Rohs et al., 2008;
11 Amandus et al., 1987b: McDonald et al., 1986b: Lockey etal., 1984) as well as community
12 studies (Weill etal..2011: Muravov et al.. 2005: Peipins et al.. 2004b: Whitehouse. 2004:
13 Peipins et al.. 2003) (see Sections 4.1.1.4 and 4.1.2). Five cohort mortality studies of workers
14 who mined, milled, and processed Libby vermiculite (henceforth described as the Libby
15 workers) identified increased risk of mortality from noncancer causes including nonmalignant
16 respiratory disease—especially asbestosis, chronic obstructive pulmonary disease, and silicosis
17 (Larson etal.. 201 Ob: Sullivan. 2007: McDonald et al.. 2004: Amandus and Wheeler. 1987:
18 McDonald et al., 1986a) as well as cardiovascular disease (Larson et al., 201 Ob). Additionally,
19 there is a potential for autoimmune effects following inhalation exposure to Libby Amphibole
20 asbestos (Noonan et al., 2006: Pfau et al., 2005) (see Section 4.3). The overall noncancer hazard
21 identification for exposure to Libby Amphibole asbestos is summarized in Section 4.5. A
22 reference concentration (RfC) is intended to define an exposure level that is likely to be without
23 an appreciable risk of adverse health effects; studies that relate these health effects to exposure
24 levels are necessary for RfC derivation23. Quantitatively, study characteristics preferred for RfC
25 derivation include adequate exposure-response information, ideally with analyses based on
22The 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.
23An RfC is defined as "An estimate (with uncertainty spanning perhaps an order of magnitude) of a continuous
inhalation exposure to the human population (including sensitive subgroups) that is likely to be without an
appreciable risk of deleterious effects during a lifetime."
This document is a draft for review purposes only and does not constitute Agency policy.
5-1 DRAFT—DO NOT CITE OR QUOTE
-------
1 estimates including assignment of quantitative exposure estimates to distinguish exposure levels
2 in the study subjects.
3 Of the available human studies, only the worker mortality and morbidity studies provide
4 exposure estimates suitable for quantitative analysis to derive benchmark concentration estimates
5 or NOAELs/LOAELs and, thus, would allow for consideration for use in RfC derivation (Larson
6 etal.. 2010b: Rohs et al.. 2008: Sullivan. 2007: McDonald et al.. 2004: Amandus and Wheeler.
7 1987: Amandus et al.. 1987b: McDonald et al.. 1986a: McDonald et al.. 1986b: Lockev et al..
8 1984). Although there are data that define exposures from some activities in the community (see
9 Section 2.3), these data do not address all potential exposures nor are data available on activity
10 patterns, which would be needed to provide individual exposure measurements. There are no
11 studies in laboratory animals on the inhalation route of exposure suitable for derivation of an
12 RfC because available animal studies lack adequate exposure-response information and are of a
13 short-term duration. Therefore, only the worker studies that include adequate exposure
14 assessment and identify health effects are considered for RfC derivation.
15 Five cohort mortality studies of Libby workers identified increased risk of mortality from
16 noncancer causes (Larson et al., 201 Ob: Sullivan, 2007: McDonald et al., 2004: Amandus and
17 Wheeler, 1987: McDonald et al., 1986a). These studies were not considered as candidates for
18 RfC derivation because the radiographic parenchymal and pleural abnormalities are more
19 sensitive than the corresponding mortality causes. An RfC is intended to be a level at which no
20 category of adverse health outcome would occur.
21 Although one study (i.e., Larson et al., 201 Ob) has reported an increase in mortality from
22 various cardiovascular diseases, no studies have been conducted in a population exposed to
23 Libby Amphibole asbestos on cardiovascular endpoints other than mortality. The reported
24 excess mortality specific to vascular effects is unique, and further substantiation of this finding is
25 needed. Thus, the mortality represents a more severe health effect from related pulmonary and
26 pleural endpoints. The less severe indicator of the first radiographic changes is the preferred
27 endpoint for RfC derivation.
28 Several morbidity studies examined the quantitative association between exposure to
29 Libby Amphibole asbestos and lesions in the lung or surrounding pleura in exposed human
30 populations; two are studies in Libby workers (Amandus et al., 1987b: McDonald et al., 1986b),
31 and two are studies in workers from the Marysville, OH facility (Rohs et al., 2008: Lockev et al.,
This document is a draft for review purposes only and does not constitute Agency policy.
5-2 DRAFT—DO NOT CITE OR QUOTE
-------
1 1984). Rohs et al. (2008) was a follow-up study to Lockey et al. (1984) on a subset of the same
2 cohort and reported a higher prevalence of adverse effects following the longer time from first
3 exposure. These four studies, all of which demonstrate an association between Libby Amphibole
4 asbestos exposure and increased risk of effects on the lung and pleura, were considered for
5 selection as the principal study to serve as the basis for the derivation of the RfC.
6 All four candidate principal studies (Rohs et al., 2008; Amandus et al., 1987b: McDonald
7 et al., 1986b: Lockey et al., 1984) have adequate reporting of the studied populations, methods of
8 analysis, statistical analyses, and results. Each of the four candidate studies reports radiographic
9 signs of nonmalignant respiratory effects, which may be considered as endpoints for an RfC
10 derivation, specifically pleural thickening (localized and/or diffuse) and small opacities
11 (indicative of parenchymal damage) (ILO, 2002, 1980, 1971). Table 5-1 summarizes the four
12 candidate principal studies. See Sections 4.1.1.4 and 4.1.3 for detailed study information.
13
14 5.2.1.1. Evaluation of Candidate Studies and Selection of Critical Study
15 The candidate studies were evaluated in terms of quality attributes that would support
16 their use as a principal study in the derivation of an RfC. When selecting among candidate
17 principal studies, there were several factors, summarized in Table 5-2, that were generally
18 considered.
19
20 5.2.1.2. Evaluation of Exposure Paradigm in Candidate Studies
21 Each of the studies provided estimates of cumulative Libby Amphibole asbestos exposure
22 (in fibers/cc-year), rather than mean or peak exposure. However, there were differences in
23 exposure intensity. In contrast to vermiculite facility workers in Libby, MT, the workers at the
24 O.M. Scott Plant in Marysville, OH, were generally exposed at lower levels (see Table 5-1), and
25 were primarily exposed in the workplace. Because of showering and changing into civilian
26 clothes at the end of the work shift for most employees, nonoccupational exposure in the
27 Marysville workers was minimal. Despite the uncertainty in the magnitude of pre-1972
28 exposures (discussed below), the available data indicate worker exposures in the Marysville
29 plant did not generally include the high intensity exposures observed for the Libby worker
30 cohort, with Rohs et al. (2008) reporting a mean exposure of 2.48 fibers/cc-year. The lower
This document is a draft for review purposes only and does not constitute Agency policy.
5-3 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-1. Summary of candidate principal studies on Libby Amphibole asbestos for reference concentration
(RfC) derivation
Cohort and
reference
Study population
Outcome assessment
Radiographic endpoints
evaluated
Exposure assessment
Exposure characteristics
Libby Worker Cohort
I
§
I
a,
McDonald et al.
(1986b)
244 employees, comprising
164 "current" workers (as
of July 1, 1983) and 80
"past" workers
Age at exam (years):
"current" "past"
<39 80 1
40-59 69 30
>60 15 49
No job tenure information;
[10.7 years as reported by
Armstrong et al. (1988)1
Radiographs taken at time
of cohort assembly
(1983)
Films independently read
by three experienced
readers using 1980 ILO
standards
Film quality:
Good: 56%
Fair: 36%
Poor: 7%
Unreadable: 0.4%
1) Parenchymal changes
(small opacities >1/0)
2) Pleural changes (pleural
thickening on chest wall,
pleural calcification)
Individual work histories and
exposure levels for specific
work locations were used to
estimate cumulative exposures
for cohort members.
1935-1967: Exposure
estimated based on
professional judgment. For
mill locations only
(1950-1967), exposure
estimated using dust-to-fiber
conversion and interviews
with plant employees.
1968-1982: Air samples
analyzed for fibers by PCM
analysis.
Mean cumulative exposure
"current" 40.1 fibers/cc-yr
"past" 118.9 fibers/cc-yr
Exposure categories:
<10 fibers/cc-yr (n = 92)
10-<20 fibers/cc-yr
(n = 64)
20-<100 fibers/cc-yr
(n = 53)
100-<200 fibers/cc-yr
(n = 16)
> = 200 fibers/cc-yr
(n = 19)
§
o
c
o
-------
p
° ^
2; ^
o^
H S
W|
V
O
c
O
Table 5-1. Summary of candidate principal studies on Libby Amphibole asbestos for reference concentration
(RfC) derivation (continued)
Cohort and
reference
Study population
Outcome assessment
Radiographic endpoints
evaluated
Exposure assessment
Exposure characteristics
I
I
a.
Amandus et al.
(1987b)
184 men employed
1975-1982, with at least 5
years job tenure
Mean (SD), years:
Age at exam: 44 (12)
Job tenure: 14 (8)
Company radiographs
Source year:
1981-1982 (72.8%)
1976-1980 (26.6%)
<1975 (1 worker)
Films independently read
by three readers using
1980 R.O standards
Film quality (by reader):
Excellent: 22.8, 24.4,
47.9%
Acceptable: 60.9, 60.9,
29.3%
Poor: 16.3, 14.7, 22.8%
Unreadable: None
1) Parenchymal changes
(small opacities >1/0)
2) Pleural changes ("any
pleura! change"3, pleura!
calcification, pleura!
thickening on chest wall
only)
Individual work histories and
exposure levels for specific
work locations were used to
estimate cumulative exposures
for cohort members.
1935-1967: Exposure
estimated based on
professional judgment. For
mill locations only
(1950-1967), exposure
estimated using dust-to-fiber
conversion and interviews
with plant employees.
1968-1982: Air samples
analyzed for fibers by PCM
analysis.
Exposure categories:
0-15 fibers/cc-year
(n = 63)
16-30 fibers/cc-year
(n = 29)
31-85 fibers/cc-year
(n = 44)
>86 fibers/cc-year (n = 48)
-------
Table 5-1. Summary of candidate principal studies on Libby Amphibole asbestos for reference concentration
(RfC) derivation (continued)
Cohort and
reference
Study population
Outcome assessment
Radiographic endpoints
evaluated
Exposure assessment
Exposure characteristics
O.M. Scott Plant Cohort, Marysville, OHb
Lockey et al.
(1984)
512 plant employees
Mean (range), years:
Age at exam: 37.5 (19-66)
Mean (SE), years:
Job tenure by exposure
group and smoking status
(NS=nonsmoker,
EX=former smoker,
CS=current smoker)
Low, NS: 6.6 (1.1)
Low, EX: 11.3(1.6)
Low, CS: 10.5 (1.2)
Medium, NS: 8.4 (1.0)
Medium, EX: 13.3(1.3)
Medium, CS: 8.9 (0.7)
High, NS: 12.2 (0.9)
High, EX:13.0 (1.1)
High, CS: 10.7 (0.9)
Posterior-anterior chest
radiographs taken in 1980
Films independently read
by 2 board-certified
radiologists (B -readers)
using modification of
1971 ILO standards. A
third B -reader was used
to resolve any difference
in diagnosis.
1) Parenchyma! changes
(only one small opacity
recorded [grade 1/1],
unclear if opacities graded
1/0 or 0/1 would have been
reported)
2) Pleural changes (pleural
plaque, pleural thickening,
pleural calcification)
3) Costophrenic angle
blunting only
Self-reported individual work
histories and exposure levels
for specific work locations
were used to estimate
cumulative exposures for
cohort members.
1957-1971: Exposure
estimated based on interviews
with plant employees and
post- 1972 air measurements.
Some workplace exposure
control measures were taken
prior to 1972.
1972-1980: Air samples
analyzed for fibers by PCM
analysis. The exposure
reconstruction in the original
study was based on limited
data, and air sampling data
from 1972 on were not
available for all jobs. Where
data were not available, the
earliest available sampling
data informed early exposures
(Lockev. 1985).
Exposure categories:
<1 fibers/cc-year
(n = 253)
1-10 fibers/cc-year
(n = 200)
>10 fibers/cc-year (n = 48)
I
I
a.
o
c
o
-------
Table 5-1. Summary of candidate principal studies on Libby Amphibole asbestos for reference concentration
(RfC) derivation (continued)
Cohort and
reference
Study population
Outcome assessment
Radiographic endpoints
evaluated
Exposure assessment
Exposure characteristics
I
§
I
a,
Rohs et al. (2008)
280 plant employees
[follow-up of cohort
described in Lockey et al.
(1984)1
Mean (SD), range (years):
Age: 59.1 (10.5), 44-87
Mean (SD), median (years):
Years since first exposure
No pleura! changes
(n = 200): 32.1 (5.5), 31.0
Pleura! changes present
(n = 80): 36.8 (4.9), 37.9
Posterior-anterior chest
radiographs taken
2002-2005
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
1) Parenchyma! changes
(small opacities, profusion
score >1/0)
2) Pleural changes
(localized pleural
thickening [any pleural
thickening excluding
costophrenic angle
blunting], diffuse pleural
thickening [any pleural
thickening with
costophrenic angle
blunting], pleural
calcification)
Exposure assessment from
Lockey et al. (1984) with
change in start date to 1963.
Exposure categories:
0.01-0.28 fibers/cc-year
(n = 70)
0.29-0.85 fibers/cc-year
(n = 72)
0.86-2.20 fibers/cc-year
(n = 68)
2.21-19.03 fibers/cc-year
(n = 70)
"Amandus et al. (1987b p. 28) define "any pleural change" as "... 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...."
bln addition to the exposure information used by Lockey et al. (1984) and Rohs et al. (2008). the University of Cincinnati augmented and refined these exposure
estimates using additional exposure data, which included industrial hygiene measurements not previously available and measurements using industrial hygiene
data from the facility to determine estimates of exposure after 1980.
H
H S
W|
V
o
c
o
-------
Table 5-2. Summary of rationale for identifying candidate principal studies
on Libby Amphibole asbestos for 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 most relevant environmental exposure scenarios are
expected to address chronic exposure scenarios (potentially including both
continuous exposure from ambient conditions and episodic activity-related
exposures).
Measures of cumulative exposure are a widely used metric to address asbestos risk.
It is consistent with the expectation that toxic responses will reflect an accumulative
effect of asbestos inhaled and deposited in tissues over time. Additionally mean
exposure, exposure duration, and time from first exposure (TSFE) have all been
reported as predictors of health effects from asbestos exposure. Cumulative
exposure has the advantage that it reflects both duration and intensity (e.g., mean
level) of asbestos exposure.
Relatively lower exposure intensities that may represent conditions more similar to
' ' preferred as there may t
lower exposure levels.
Relatively lower exposure intensities that may represent conditions more s
environmental exposures are preferred as there may be less uncertainty in
extrapolation of the results to lower exposure levels.
Results from studies with high exposure intensity or cumulative exposure are, other
things being comparable, judged less relevant for environmental risk assessment
compared to studies defining effects at lower levels of exposure. Some biological
processes (e.g., potential decrease in effectiveness of particle clearance processes)
may more strongly influence responses at very high levels of exposure and be less
relevant at lower levels. Thus, exposure conditions with lower level exposures may
remove some of the uncertainty in estimating health effects from environmental
exposures.
Study design characteristics
Sufficient follow-up time for outcomes to develop (which can depend on the health
outcome being addressed).
Study size and participation rates that are adequate to detect and quantify health
outcomes being studied are preferred, with no indications of bias in study population
selection.
Use of a study design or analytic approach, which adequately addresses the relevant
sources of potential confounding, including age, sex, smoking, and exposure to other
risk factors (such as non-Libby asbestos).
This document is a draft for review purposes only and does not constitute Agency policy.
5-8 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-2. Summary of rationale for identifying candidate principal studies
on Libby Amphibole asbestos for RfC development (continued)
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 appropriate exposure information, and individual, rather
than area samples are preferred where available. 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, 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 do 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.
Measurement of effect(s)
Emphasis is placed on the more sensitive health outcome endpoints that are
available. For 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 (HRCT) and independently evaluated multiple
qualified readers according to ILO standards.
Evaluation of radiographs should not be influenced by knowledge of exposure status.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
intensity exposures for the Marysville cohort and corresponding lower cumulative exposures are
advantages of this study, considering there are uncertainties inherent in exposure-response data
and extrapolating from the high intensity occupation exposures to lower level exposures often
seen in community and environmental exposures.
5.2.1.2.1. Evaluation of study design in candidate studies
The candidate principal studies differed in the study populations, in terms of follow-up
time, study size and participation, and available information (see Table 5-1). The study sizes are
similar for the two Libby worker studies (n= 184 and n= 244, respectively) (Amandus et al.,
1987b: McDonald et al.. 1986b) and the Marysville update (n = 280) (Rohs et al.. 2008).
Adequate follow-up time allows for the health effect to manifest prior to sampling. In the
case of pleural abnormalities, there is some variability with latency based on intensity of
This document is a draft for review purposes only and does not constitute Agency policy.
5-9 DRAFT—DO NOT CITE OR QUOTE
-------
1 exposure as well as the nature of the pleural lesion where discrete pleural plaques have a shorter
2 latency than diffuse thickening of the visceral pleura. Larson et al. (2010a) studied the latency
3 for individuals in the Libby worker cohort, reporting a median latency of 8.6 years for localized
4 pleural thickening versus 27 years for diffuse pleural thickening and 19 years for minimal signs
5 of small opacities (parenchymal changes).24 Lockey et al. (1984) report the mean employment
6 duration for their exposure groups from 6.6 to 13.3 years at the time of their study (but do not
7 assess time since first exposure (TSFE); thus, it is unclear whether in the first examination these
8 workers had sufficient follow-up to assess the radiographic changes, especially diffuse pleural
9 thickening and small opacities. The Rohs et al. (2008) report includes 24 more years of
10 follow-up time and is preferred over the early Lockey et al. (1984) study on this basis.
11 Both studies of the Libby workers report duration of employment and average age of the
12 participants, but not TSFE. The McDonald et al. (1986b) study included both current and former
13 workers—these former workers likely have longer time from first exposure compared with
14 current workers. The study included all current plant employees (164 men, 9 women).
15 However, there was a lower participation rate in former employees (80 of 110 eligible former
16 employees agreed to provide chest radiographs). Additionally, X-rays for all study participants
17 were taken in the same year, providing similar quality X-rays between past and current
18 employees. In contrast, Amandus et al. (1987b) only considered workers employed during 1975
19 to 1982 and relied on available radiographs regardless of year (radiographs were available for
20 93% of employees). Because workers terminated prior to 1975 were excluded from the study,
21 older individuals, and individuals with longer TSFE were less likely to be included than in the
22 study by McDonald et al. (1986b), which included former workers. Both Libby worker studies
23 do report radiographic abnormalities, so the follow-up is adequate for some effects to be
24 documented; however, compared with the Rohs et al. (2008) study, the Libby worker studies
25 have shorter follow-up times.
24 Individual latency for visible LPT in Libby exposed workers was evaluated in 84 workers with radiographic
evidence of pleural and/or parenchymal changes (Larson et al.. 2010aX By examining historical radiographs,
researchers were able to identify the first appearance of the lesions, although it is recognized that retrospective
design of this study likely identified lesions at earlier time points, as the readers were aware of the later X-rays
(Larson et al.. 2010aX It is acknowledged that some of the workers at Libby may have been exposed through the
community prior to working, and in fact, one individual had the first pleural change noted at 9 years of age, prior to
occupational exposure (Larson et al.. 2010a'). Where data on prior exposures were available, workers with no prior
exposure had an average latency of 9.4 years versus 5.1 years for workers with potential exposures prior to hire
(N= 63 and 31, respectively).
This document is a draft for review purposes only and does not constitute Agency policy.
5-10 DRAFT—DO NOT CITE OR QUOTE
-------
1 Among Marysville workers, there were very few employees who declined to participate
2 in the earlier study by Lockey et al. (1984), where 512 out of 530 employees were included, but
3 there is potential for selection bias in the follow-up by Rohs et al. (2008), where only
4 280 employees out of the original cohort were evaluated. Rohs et al. (2008) state that employees
5 hired in 1973 or earlier (when exposure estimates were more uncertain) were more likely to
6 participate compared to employees hired after 1973, and while the range of cumulative Libby
7 Amphibole asbestos exposure was similar between participants and nonparticipants, participants
8 did have higher mean cumulative exposure estimates. While it is accurate that exposure levels
9 were uncertain before sampling began at Marysville in 1972, it is also accurate that exposures
10 were much lower beginning in 1974, when additional industrial hygiene controls were
11 implemented. Thus, persons hired <1973 had higher exposure (if less perfectly measured), while
12 those hired >1974 had lower exposure, and likely less disease (under an assumption of an
13 exposure-response effect). Thus, we might assume that the prevalence rates in nonparticipants
14 are likely lower than in participants. The self-selection to participate in the study is dependent
15 on the exposure, thus leading to dependent censoring and potential selection bias (see
16 Section 4.1.3 for a discussion of this potential selection bias). However, Rohs et al. (2008)
17 conducted a sensitivity analysis assuming that all living nonparticipants had no pleural changes
18 and report a similar significant trend of increased pleural changes by exposure quartile. In
19 contrast, participation rates for the Libby worker studies were much higher (see above), and there
20 is no indication of potential bias in selection of these study participants (Amandus et al., 1987b:
21 McDonald et al., 1986b).
22 Both studies of Libby workers also evaluated age and smoking as potential confounders
23 of the association between Libby Amphibole asbestos exposure and radiographic abnormalities.
24 McDonald et al. (1986b) report that both age and cumulative exposure are significant predictors
25 of small opacities and pleural abnormalities in the study of current and former workers,
26 providing regression coefficients for cumulative exposure, age, and smoking status. Amandus et
27 al. (1987b) report that although cumulative exposure and age are both significant predictors for
28 small opacities, cumulative exposure was not significantly related to pleural abnormalities when
29 age is included in the model, thus limiting the usefulness of these data for RfC derivation based
30 on pleural abnormalities. Neither study of Libby workers addressed gender, body mass index
This document is a draft for review purposes only and does not constitute Agency policy.
5-11 DRAFT—DO NOT CITE OR QUOTE
-------
1 (BMI), or time from first exposure, although both studies excluded workers with other
2 asbestos/dusty trade occupations.
3 With respect to the Marysville, OH worker cohort, Lockey et al. (1984) only matched on
4 age in their analysis. The follow-up examination by Rohs et al. (2008) included information on
5 several important covariates, including age, gender, hire date, prior exposure to asbestos, BMI,
6 and smoking history. Hire date and age were significantly associated with the prevalence of
7 pleural abnormalities, and results are presented considering these covariates.
8
9 5.2.1.3. Evaluation of Exposure Assessment in Candidate Studies
10 For both the O.M. Scott facility in Marysville, OH and the Libby, MT facilities, exposure
11 estimates rely primarily on fiber counts using phase contrast microscopy (PCM) and
12 reconstruction of earlier exposures from company records, employee interviews, and the
13 professional judgment of the researchers estimating historical exposures (Amandus et al., 1987a:
14 McDonald et al., 1986a: Lockeyetal., 1984). Work histories for the Libby worker cohort were
15 extracted from company employment records, while work histories for the Marysville cohort
16 were self-reported.
17 The two studies of workers in Libby, MT used similar exposure estimation, based on the
18 same fiber measurements and work records (Amandus et al., 1987b: McDonald et al., 1986a).
19 As discussed in Section 4.1.1.2, exposures prior to 1968 are not based on fiber measurements by
20 PCM and, thus, are more uncertain that later exposure estimates.25 The study population of
21 McDonald et al. (1986b) included current and former workers, with 26% of participants over 60
22 and 40% of participants between 40-59 years of age at the time of their X-ray in 1983.
23 Although tenure and dates of employment are not reported, exposure estimates for this study
24 group would include the less-certain exposure estimates prior to 1968 (McDonald et al., 1986a).
25 However, Amandus et al. (1987b) studied workers still employed during 1975-1982 (i.e.,
26 excluding those terminated prior to 1975) who had at least 5 years of employment. The average
27 tenure of the study participants was 14 years. Although both studies have the limitation of
28 less-certain exposure estimates prior to 1968, based on study design, the Amandus et al. (1987b)
25 Exposures in the dry mill at Libby, MT, prior to 1967 were estimated from total dust measurements based on
site—specific conversion ratios. Exposures for all other location operations prior to 1968 were estimated because no
air sampling data were available (Amandus et al.. 1987a: McDonald et al.. 1986bX
This document is a draft for review purposes only and does not constitute Agency policy.
5-12 DRAFT—DO NOT CITE OR QUOTE
-------
1 study group includes a greater proportion of more recent workers. However, neither researcher
2 assessed these uncertainties nor the impact of early exposure estimates on the apparent
3 exposure-response relationship.
4 Another source of uncertainty in exposure estimates for this cohort is possible
5 community/nonoccupational exposures. Members of the Libby worker cohort may have lived in
6 Libby prior to/after employment and resided in Libby and surrounding areas during employment.
7 In both cases, there may have been community exposures to Libby Amphibole asbestos that are
8 not captured in occupational-based cumulative exposure metrics. This unmeasured
9 nonoccupational exposure may be low relative to the estimated occupational exposures, but is,
10 nevertheless, a source of uncertainty in estimating the exposure-response relationship.
11 The quality of the exposure assessment also changed over time in the Marysville cohort
12 (Rohs et al., 2008; Lockey, 1985). Industrial hygiene measurements based on PCM analysis are
13 available for the O.M. Scott facility beginning in 1972, although personal breathing zone
14 samples were not available until 1976 (Rohs et al., 2008). Thus, exposure levels for all job tasks
15 prior to 1972 are estimates from later sampling events. Additionally, air sampling data were not
16 available for several job tasks until the late 1970s. For example, air-sampling data were only
17 available for two of seven job tasks in the trionizing department beginning in 1973 (expander
18 and dryer). All others have dates of 1976 or later [see Table 10, Lockey (1985)]. The
19 installation of exposure control equipment in 1974 adds to the uncertainty in early exposures
20 estimated from sampling in later years. There is uncertainty when the Libby ore was first used in
21 the facility. Company records indicated that the date was between 1957 and 1960, and the
22 University of Cincinnati used the best-available information from focus group interviews to
23 assign the first usage of Libby ore in 1959 (see Appendix F).
24 EPA has collaborated with the University of Cincinnati research team to better evaluate
25 historical exposures at the O.M. Scott facility in Marysville, OH (see Appendix F). Although no
26 air-sampling results were found prior to 1972, additional information on plant processes from
27 other records and employee interviews has resulted in updated exposure estimates (see
28 Section 5.2.3.1). These refined estimates of the historical exposure improve exposure
29 characterization for the Marysville worker cohort over previous publications.
30
This document is a draft for review purposes only and does not constitute Agency policy.
5-13 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.2.1.3.1. Evaluation of outcome assessment in candidate studies
1 In all four candidate studies, outcomes were assessed using chest radiographs
3 independently evaluated by multiple readers. However, there were differences in the standards
4 used for evaluation of radiograph!c changes, as well as timing and quality of the radiographs.
5 The two studies in Libby workers (Amandus et al., 1987b: McDonald et al., 1986b) used similar
6 outcome-assessment procedures, with radiographs evaluated by three readers according to 1980
7 ILO standards. Two different sets of standards were used to evaluate radiographs in the
8 Marysville cohort. The first study used modified 1971 ILO standards (modifications not
9 stipulated) (Lockey etal., 1984), while the follow-up study used the updated 2000 ILO standards
10 (Rohs et al.. 2008).
11 Radiograph quality may also impact outcome assessment. In McDonald et al. (1986b),
12 which used radiographs taken in 1983 specifically for the study, 7% of films were classed as
13 "poor quality" (some technical defect impairing the pneumoconiosis classification) and 0.4% as
14 "unreadable." Amandus et al. (1987b), which used available radiographs taken over a wide time
15 period (1975 to 1982), report that the proportion of films rated as "poor quality" ranged from
16 14.7% to 22.8% depending on the reader. In the Marysville cohort, Lockey et al. (1984) state
17 that"... radiographs that could not be interpreted because of poor quality were repeated" (p. 953).
18 Rohs et al. (2008) do not report the percentage of films rated as "poor quality" but do note that
19 7 out of 298 (2.3%) radiographs taken were considered unreadable.
20
21 5.2.1.3.2. Selection of principal cohort
22 Based on the criteria set out in Table 5-2 and the above evaluation, the update of the
23 Marysville, OH worker cohort (Rohs et al., 2008) is the preferred cohort. The main advantages
24 of the Marysville, OH worker cohort over the two studies of pleural and lung abnormalities in
25 the workers in Libby, MT are:
26
27
28 1) Adequate follow-up time and the availability of time from first exposure data for
29 evaluation,
30 2) Minimal exposure to Libby Amphibole asbestos outside of the workplace,
This document is a draft for review purposes only and does not constitute Agency policy.
5-14 DRAFT—DO NOT CITE OR QUOTE
-------
1 3) Better quality radiographs, and use of the most recent ILO reading guidelines in the
2 cohort update,
3 4) Data are more appropriate for low-dose extrapolation—a lower range of cumulative
4 exposures for the study participants (n = 280), compared to Libby workers,
5 5) The data allow consideration of more covariates and potential confounders (e.g.,
6 BMI, smoking status, age),
7 6) The presence of a demonstrated exposure-response relationship for Libby amphibole
8 asbestos exposure and radiographic abnormalities—in contrast to the study by
9 Amandus et al. (1987b), which does not support an exposure-response relationship
10 for pleural abnormalities based on the cumulative exposure metric (when age is
11 included as a covariate).
12
13
14 The disadvantages of the Marysville, OH cohort compared to the two studies of pleural
15 and lung abnormalities in the workers in Libby, MT are:
16
17
18 1) Approximately 70% of the Marysville, OH cohort were hired before 1972 when there
19 were no measured exposure data [Rohs et al. (2008), and Lockey et al. (1984) study].
20 2) Participants in Rohs et al. (2008) were self-selected, with greater participation among
21 older employees and those who began work prior to 1973 when exposures were
22 relatively higher. This is a potential source of bias in study population selection
23 analyzed by Rohs et al. (see Section 4.1.3).
24 3) Exposure estimates are based on self-reported work histories. In this case, there is
25 some uncertainty in the employment history, and some individuals had extensive
26 overtime work. Employment history was self-reported during interviews with each
27 individual for the original study (i.e., Lockey et al., 1984), and errors in this process
28 could affect assigned Libby Amphibole asbestos exposure estimates for this cohort.
29
30
31 5.2.1.4. Selection of Critical Effect
32 There are several endpoints that are suitable for consideration for the derivation of an
33 RfC for Libby Amphibole asbestos where health effects data and exposure information are
34 available in the principal study (Rohs et al., 2008; Lockey et al., 1984): (1) parenchymal changes
35 viewed as small opacities in the lung; (2) blunting of the costophrenic angle (measured between
36 the rib cage and the diaphragm); or (3) pleural thickening (both localized and diffuse). Each of
37 these effects is an irreversible pathological lesion (ATS, 2004). As the available epidemiologic
This document is a draft for review purposes only and does not constitute Agency policy.
5-15 DRAFT—DO NOT CITE OR QUOTE
-------
1 studies describe these endpoints as viewed on standard X-rays (see Text Box 5-1), it is important
2 to understand the distinction between what is viewed on the radiograph versus the underlying
3 biologic lesion. The following discussion reviews the health effects associated with each of
4 these radiographic abnormalities observed in workers exposed to Libby Amphibole asbestos.
5
6
Text Box 5-1. Radiographic Abnormalities of the Lung and Pleura
Parenchymal changes in the lung (small opacities): The small opacities viewed within the lung
(interstitial changes) are indicative of pneumoconiosis and are associated with exposure to not only
mineral fibers, but also mineral dust and silica. The radiographic signs of pneumoconiosis begin as small
localized areas of scarring in the lung tissue and can progress to significant scarring and lung function
deficits. The ILO standards provide a scheme for grading the severity of the small opacities; the size,
shape, and profusion of the small opacities are recorded, as well as the affected zone of the lung (ILO.
2002).
Obliteration of the costophrenic angle: The costophrenic angle (CPA) is measured as the angle between
the ribcage and the diaphragm on a posterior anterior-viewed radiograph (the costophrenic recess). When
CPA blunting or obliteration is noted on a radiograph, it is recorded as present or absent (ILO. 2002).
Obliteration of the CPA may occur in the absence of other radiographic signs.
Pleural thickening: The pleura! 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" (ILO. 2002). Localized
pleural thickening may also be viewed in-profile or face-on and is generally a pleural plaque (parietal).
Calcification is noted where present (ILO. 2002).
9 5.2.2. Evaluation of Radiographic Lesions as Potential Critical Effects
10 5.2.2.1. Health Effects of Parenchymal Changes as Small Opacities Viewed on Standard
11 Radiographs
12 Radiographic evidence of small opacities in the lung is evidence of fibrotic scarring of
13 lung tissue consistent with mineral dust and mineral fiber toxicity. The scarring of the
14 parenchymal tissue of the lung contributes to measured changes in pulmonary function,
15 including obstructive pulmonary deficits from narrowing airways, restrictive pulmonary deficits
16 from impacting the elasticity of the lung as well as decrements in gas exchange. However,
17 although data across the mineral fiber literature strongly support a finding of functional deficits
18 where small opacities are visible on radiographs, the data also indicate that deficits in pulmonary
19 function (consistent with interstitial fibrosis) are seen before these changes are detected by
This document is a draft for review purposes only and does not constitute Agency policy.
5-16 DRAFT—DO NOT CITE OR QUOTE
-------
1 radiographic examination. Thus, changes in lung function may occur before the fibrotic lesions
2 can be detected on standard radiographs (ATS, 2004; Broderick et al., 1992). For example,
3 decreased Carbon monoxide (CO) diffusion is a sign of reduced gas exchange in the pulmonary
4 region of the lung and is observed in workers exposed to other types of asbestos even when small
5 opacities are absent on radiographs. Similarly, obstructive deficits in lung function may be
6 observed without radiographic signs for fibrotic lesions of small opacities. As decreased
7 diffusion and obstructive deficits are mechanistically linked to changes in the parenchymal tissue
8 these data suggest radiographs may not be sensitive enough to detect and protect against small
9 localized lesions in parenchymal tissue of the lung. Radiographic evidence of small opacities
10 indicates interstitial damage of the lung paremchyma, is associated with decreased pulmonary
11 function and considered evidence of an adverse health effect. Thus, small opacities are an
12 appropriate endpoint for RfC derivation. However, as there is evidence of functional changes in
13 lung function from lesions not detectable on conventional radiographs, more sensitive endpoints
14 should be considered.
15
16 5.2.2.2. Health Effects of Diffuse Pleural Thickening (DPT) Viewed on Standard
17 Radiographs
18 DPT is a fibrotic lesion (often described as a basket weave of collagen) in the visceral
19 pleura that encases each lobe of the lungs. The fibrotic lesion restricts the ability of the lung to
20 expand mechanically, as well as by reducing the available volume (where thickening has
21 progressed) (Jones et al., 1988) and DPT is strongly associated with reduced lung function (ATS,
22 2004). There are consistent reports of impaired lung function associated with DPT in
23 asbestos-exposed populations (Broderick et al., 1992; Kilburn and Warshaw, 1991; Bourbeau et
24 al., 1990). A cross-sectional study of men (n = 1,298) exposed to asbestos through various
25 trades (e.g., boiler makers, welders, plumbers/pipefitters) included chest radiographs and
26 spirometry (Kilburn and Warshaw, 1991). When considering the effect of DPT (with
27 costophrenic angle [CPA] blunting) on radiographic function, FVC, FEV1, and FEF25-7526 were
28 all significantly reduced (85, 79, and 66% of predicted values, respectively) as compared with
29 individuals with calcification or plaques only in men with no signs of small opacities (ILO
26Forced Vital Capacity (FVC); Forced Expiratory Volume in 1 second (FEV1) and Percent FVC
(FEV%) = [(100 x FEV1) - FVC, FEF25-75, is the expiratory flow between 25% and 75% of the FEV]
This document is a draft for review purposes only and does not constitute Agency policy.
5-17 DRAFT—DO NOT CITE OR QUOTE
-------
1 profusion score of 0/0 or 0/1) (p < 0.0001). The relationship between pleural fibrosis and FVC
2 was studied in asbestos-exposed sheet metal workers (N = 1,211) where not only the type of
3 thickening (discrete versus diffuse) (ILO, 1980) but also CPA involvement and the location of
4 the thickening were taken into consideration (Broderick et al., 1992). Univariate analysis
5 indicated FVC was decreased by both DPT (with CPA blunting) and circumscribed thickening,
6 diaphragm involvement, CPA involvement, and the extent of the thickening (Broderick et al.,
7 1992). Multivariate linear regression, allowing for control of potential confounders, found
8 decreased FVC was significantly related to DPT, plaques, CPA involvement, and extent of the
9 thickening, but not diaphragmatic involvement (Broderick et al., 1992).
10 The mechanisms for reduced lung volume in individuals with asbestos-related DPT have
11 been examined by measuring lung function and changes in diaphragm length, rib-cage
12 dimensions, and subphrenic volume in 26 patients during breathing (Singh et al., 1999). DPT
13 reduced both total lung capacity and FVC with corresponding decreases in rib-cage expansion
14 and movement of the diaphragm, consistent with the restrictive nature of these lesions, which
15 may encase part of the lung (Singh et al., 1999). These direct measurements of the effect of DPT
16 chest wall and diaphragmatic motion illustrate the role of DPT in reducing lung volume,
17 contributing to restrictive deficits in pulmonary function. Taken together, the epidemiologic
18 evidence and the mechanistic information that support a restrictive effect of fibrotic lesion in the
19 visceral pleura, substantiate the associations between DPT and decreased pulmonary function.
20 As such, the observation of DPT on standard radiographs is representative of pathological
21 changes directly related to reduced lung function and is, therefore, an indication of adversity,
22 and, can serve as an appropriate health endpoint for consideration in RfC derivation.
23
24 5.2.2.3. Health Effects of Localized Pleural Thickening (LPT) Viewed on Standard
25 Radiographs
26 Localized pleural thickening (LPT) viewed on a standard radiograph may include both
27 pleural plaques and pleural thickening that does not involve blunting of the costophrenic angle
28 (ILO, 2002). Thus, both parietal plaques and localized thickening of the visceral pleura may be
29 designated as LPT. Thickening of the parietal pleura is due to an acellular collagen plaque
30 (basket weave of collagen fibers) between the parietal pleura and the ribcage (or along the
31 diaphragm) often described as discrete or circumscribed pleural plaques (ATS, 2004; Jones,
This document is a draft for review purposes only and does not constitute Agency policy.
5-18 DRAFT—DO NOT CITE OR QUOTE
-------
1 2002). Thickening of the visceral pleural is a fibrosis with diffuse borders and may extend into
2 the lung parenchyma (ATS, 2004; Jones, 2002). The pathology and health effects of the
3 different lesions are evaluated here in the characterization of the health significance of LPT.
4 Costal parietal plaques occur between the thoracic cage and parietal pleura, which is
5 normally adherent to the thoracic cage (ATS, 2004; Jones, 2002). Costal parietal plaques have
6 been described as collagen deposits with ragged irregular edges and up to 1 cm in depth and may
7 be calcified. These parietal plaques have been associated with constricting pain in the thoracic
8 cavity (Mukherjee et al., 2000). The parietal pleura is well innervated by the intercostal and
9 phrenic nerves and is considered very sensitive to painful stimuli (Jones, 2002). With respect to
10 parietal plaques, pain during exertion or exercise could result in restrained chest wall motion
11 during exertion or exercise. Thus, Bourbeau et al. (1990) hypothesized that the dyspnea and
12 changes in pulmonary function noted in individuals with pleural plaques may be due to physical
13 irritation and perhaps a constricting action where parietal plaques are well progressed or
14 numerous and impact a large proportion of the parietal surface.
15 Kouris et al. (1991) examined the presence of dyspnea, and measures of pulmonary
16 function (i.e., FVC, FEV1, and FEV%27) in asbestos-exposed workers (n = 913) in relation to
17 radiographic signs of lung and pleural anomalies. Radiographs were contemporary to the study
18 and read in accordance with ILO (1980) guidelines. Pleural plaques were associated with
19 reduced FVC and FEV1.0 (87.6% and 84.1% of predicted, respectively, p< 0.0005), although
20 deficits associated with diffuse thickening were greater (76.4% and 73.9%, p < 0.0005) (Kouris
21 et al., 1991). Correspondingly odds ratios for decreased FVC and FEV1.0 (80% decrement)
22 were increased by the presence of both plaques and diffuse thickening (1.5 for plaques and
23 4.2 and 4.7 for diffuse thickening, respectively). Interestingly, when history of lung disease was
24 considered, pleural plaques had a greater effect in individuals without previous lung disease
25 (OR of 2.1 for FVC and 1.7 for FEV1.0).
26 Pleural thickening in general is associated with decreased pulmonary function (Petrovic
27 et al., 2004; Wang et al., 2001; Miller et al., 1994)and this association is strengthened as the
28 severity of the pleural thickening increases (Lilis et al., 1991). Few available studies have
29 examined the relationship between pleural plaques identified on standard radiographs (ILO,
27Forced Vital Capacity (FVC); Forced Expiratory Volume in 1 second (FEV1) and Percent FVC
(FEV%) = [(100 x FEV1) - FVC].
This document is a draft for review purposes only and does not constitute Agency policy.
5-19 DRAFT—DO NOT CITE OR QUOTE
-------
1 1980) and pulmonary function without including DPT in the analysis and adequately controlling
2 for the presence of small opacities (indicative of parenchymal damage)28.
3 Lilis et al. (1991) examined pulmonary function in long-term asbestos insulation workers,
4 and found that one measure (FVC) decreased significantly as the severity of pleural fibrosis (all
5 types, as indicated by a pleural index) increased. This decrease was more dramatic when
6 including parenchymal changes (small opacities) or if DPT was viewed separately. A second
7 analysis focusing on participants with pleural plaques found an inverse relationship between
8 severity of the pleural plaques and FVC (p < 0.0001), when adjusting for the independent effects
9 of duration, smoking and presence of small opacities (Lilis et al., 1991). This finding supports a
10 view that pleural plaques, when extensive, may contribute to restrictive lung deficits, but the
11 analysis included individuals with known small opacities (e.g., lung fibrosis). The authors do not
12 address the potential that the pleural index may also correspond to increased severity of
13 parenchymal changes, potentially confounding the analysis where accounting for small opacities
14 (profusion scores of 1/0 or greater) may not adequately control for asbestos-related parenchymal
15 damage.
16 Oliver et al. (1988) studied the relationship between pulmonary function and pleural
17 plaques in asbestos-exposed railway workers (n = 383). Case selection included exclusion of
18 workers with DPT (ILO, 1980) and exclusion of any indication of small opacities (only
19 profusion scores of 0/0 were included). Standard spirometry was conducted to evaluate
20 restrictive and obstructive pulmonary deficits. Additionally, single-breath diffusing capacity
21 (DLCO) was measured which would indicate parenchymal defects. The DLCO was similar in
22 subjects with and without circumscribed plaques, suggesting little or no subradiographic
23 parenchymal damage, which corresponded to the presence of pleural plaques. Pleural plaques
24 were associated with both decreased FVC and pulmonary restriction (p = 0.03 and 0.04,
25 respectively) where the diagnostic certainty for the plaques was considered 'definite', and there
26 was an association between level of diagnostic certainty and these pulmonary deficits (p = 0.02)
27 (Oliver et al., 1988). Quantitative pleural score, based on the number and extent of plaques, was
28It is difficult to control for effects subradiographic parenchymal fibrosis on lung function, where it may not have
progressed to visible small opacities, and it has been suggested that reduced lung function, which has been
associated with circumscribed plaques in some studies, may be reflecting the effects of subradiographic
parenchymal changes, rather than a direct effect of DPP (ATS. 2004: Erdinc et al.. 2003: Miller and Zurlo. 1996:
Brodericketal.. 1992).
This document is a draft for review purposes only and does not constitute Agency policy.
5-20 DRAFT—DO NOT CITE OR QUOTE
-------
1 also associated with decreased FVC and pulmonary restriction (p = 0.0135 and 0.0126,
2 respectively) (Oliver et al., 1988). Of the available studies that assess pleural thickening with
3 standard radiographs, this study best controls for the possibility of subradiographic parenchymal
4 damage and is, therefore, strong evidence that circumscribed pleural plaques independently
5 impact pulmonary function. The observed restrictive pulmonary deficit is consistent with the
6 potential for pleural plaques to restrict chest wall motion or the elasticity of the diaphragm.
7 Three high-resolution computed tomography (HRCT) studies were conducted specifically
8 to assess the potential for parietal plaques to impact lung function. Staples et al. (1989) report no
9 difference in lung function or diffusing capacity between participants (n = 76) with and without
10 pleural plaques. Soulat et al. (1999) found no difference in FEV1 or FVC between
11 asbestos-exposed insulators with (n = 84) and without (n = 51) pleural plaques in the absence of
12 any parenchymal changes. As severity of pleural thickening has been shown to be positively
13 associated with decrease measures of pulmonary function, Van Cleemput et al. (2001) not only
14 examined the effect of HRCT defined pleural plaques on pulmonary function, but also assessed
15 the extent of the pleural plaques. Neither the presence nor extent of pleural plaques were
16 associated with lung function parameters (diffusing capacity or normalized spirometric values)
17 (van Cleemput et al., 2001). Where pleural plaques and diffuse thickening (visceral pleura) were
18 both identified by HRCT and correlated to pulmonary function, diffuse visceral thickening—but
19 not plaques—were associated with decreased lung volume and FVC (Copley et al., 2001).
20 Although CPA involvement was not independently assessed, several scoring systems for severity
21 were compared which included CPA involvement, and as in other studies, increased severity
22 correlated to greater decrements.
23 The mechanisms for reduced lung volume in individuals with asbestos-related pleural
24 plaques and DPT have been examined by measuring lung function and changes in diaphragm
25 length, rib-cage dimensions and subphrenic volume in 26 patients during breathing (Singh et al.,
26 1999). Pleural plaques alone did not reduce any of the measures of lung function in this study,
27 but there were indications of reduced diaphragm movement (Singh etal., 1999). This may be an
28 indication that diaphragmatic plaques in the parietal pleura have the potential to attenuate the
29 movement of the diaphragm during breathing. Because this study is relatively small (N= 26)
30 and a distinction was not made between costal and diaphragmatic plaques by the study authors,
This document is a draft for review purposes only and does not constitute Agency policy.
5-21 DRAFT—DO NOT CITE OR QUOTE
-------
1 additional work is needed to better understand the direct effects of pleural plaques on lung
2 function.
3 Although some researchers have questioned that pleural plaques alone directly impact
4 pulmonary function, a critical review of the literature from 1965-1999 concludes: "1)
5 Individuals with asbestos-induced pleural plaques may have alterations in pulmonary function
6 and /or clinical symptoms that are independent of smoking and radiographic parenchymal
7 fibrosis and, 2) the respiratory changes dues to asbestos-induced pleural plaques are generally
8 less severe than those caused by pleural thickening" (Rockoff et al., 2002). Therefore, although
9 the evidence is mixed, pleural plaques may be independently associated with reduced pulmonary
10 function.
11 No studies correlating pulmonary function to radiographic signs of localized pleural
12 thickening (LPT) using the ILO (ILO, 2002) guidelines could be located. However, several
13 researchers employed similar classification schemes, modifying earlier ILO classification
14 systems, such that DPT was diagnosed only in conjunction with blunting of the CPA. This
15 modification potentially includes cases of diffuse pleural thickening (without CPA blunting) in
16 their analysis of pleural plaques, making their findings somewhat applicable to the current
17 classification of LPT (Garcia-Closas and Christian!, 1995; Broderick et al., 1992). Pleural
18 thickening (without CPA blunting) was associated with mixed respiratory impairment in a study
19 of asbestos-exposed construction carpenters (n = 631) (OR of 3.7 [95% Confidence Interval (CI):
20 1.4-12.3]) but was only weakly associated when the outcome was restrictive deficit specifically
21 (1.3 [95% CI: 0.4-3.9]) (Garcia-Closas and Christiani. 1995). Broderick et al. (1992) found
22 decreased FVC was not only significantly associated with "diffuse thickening" (with CPA
23 blunting) but also with "pleural plaques" (which included all pleural thickening without CPA
24 blunting). The severity of pleural thickening (both as width or percentage of lateral wall) and
25 calcification was associated with reduced FVC as well (Broderick et al., 1992). Kilburn and
26 Warshaw (1991) assessed pulmonary function in individuals with "plaques only," "diffuse
27 thickening only," and "diffuse thickening with CPA blunting," showing progressive deficits
28 across these categories in FVC, FEV1, and mid-expiratory flow (e.g., FEV1: 90.5, 86.2, and
29 49.4% \p < 0.05], respectively). Again, there is a trend that diffuse thickening has a greater
30 impact on lung function parameters, although an independent effect of plaques cannot be ruled
31 out by these data.
This document is a draft for review purposes only and does not constitute Agency policy.
5-22 DRAFT—DO NOT CITE OR QUOTE
-------
1 In summary, the radiographic classification of localized pleural thickening (LPT) under
2 current ILO guidelines may include both parietal plaques (in the pleura lining the interior of the
3 ribcage) and diffuse visceral thickening (without CPA obliteration) (ILO, 2002). The two
4 lesions (parietal plaques and localized visceral thickening) are distinct and may contribute
5 independently to observed health effects. Parietal plaques are known to induce chronic
6 constricting chest pain that increases in severity as the extent of the plaques increases. Pleural
7 thickening in general is associated with reduced lung function parameters with increased effect
8 correlating with increased severity of the pleural thickening (Petrovic et al., 2004; Wang et al.,
9 2001: Miller et al.. 1994: Lilisetal.. 1991). There is clear evidence from HRCT studies that the
10 presence and extent of visceral thickening does impair lung function, although, when evaluated
11 independently, parietal plaques were not statistically correlated with decreased pulmonary
12 function (Copley et al., 2001: Schwartz et al., 1993). Specifically considering the designation of
13 LPT, lung function impairment has been demonstrated in several studies where pleural
14 thickening without CPA involvement has been studied (Garcia-Closas and Christian!, 1995:
15 Broderick et al., 1992: Kilburn and Warshaw, 1991). Thus, the radiographic classification of
16 localized pleural thickening (LPT) (ILO, 2002) includes pleural lesions associated with chronic
17 chest pain, decreased lung volume, and decreased measures of lung function. Therefore, EPA
18 considers LPT an adverse effect and an appropriate endpoint for RfC derivation.
19
20 5.2.3. Methods of Analysis
21 5.2.3.1. Exposure Data and Choice of Exposure Metric
22 EPA collaborated with a research team at the University of Cincinnati to update the
23 exposure reconstruction for use in the job-exposure matrix (JEM) for all workers in the
24 Marysville, OH cohort, taking into account additional industrial hygiene data that were not
25 available for previous studies conducted in this cohort. As discussed in detail in Appendix F,
26 exposure estimates for each worker in the O.M. Scott Marysville, OH plant were developed
27 based on available industrial hygiene data from the plant. Figure 5-1 shows the average
28 exposure concentrations of fibers in air (PCM fibers/cc)29 of each department from 1957 to 2000,
29
29PCM, where fibers are viewed and counted by light microscopy, does not identify the composition of the fiber.
Thus, the mineralogy of fibers identified under PCM cannot be determined.
This document is a draft for review purposes only and does not constitute Agency policy.
5-23 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Estimated
Measured
^^ Trionizing
-^Plant Maintenance
-^Central Maintenance
^^Background
0.0
1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000
Year
Figure 5-1. Estimated and measured exposure concentrations in Marysville,
OH facility3
aTrionizing is a term used in the Marysville, OH facility and includes unloading of rail cars
containing vermiculite ore (track), using conveyers to move the vermiculite ore into the expander
furnaces, separation of the expanded vermiculite from sand, blending in of lawn care chemicals,
and drying and packaging of the final product. As no unexpanded ore was used in pilot plant,
research, polyform, office, packaging, or warehouse, jobs in these categories were assigned as
background. Workers assigned to plant maintenance activities spent 50% of their time in
trionizing areas and 50% of their time in areas assigned as plant background. Workers assigned to
central maintenance spend 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).
This document is a draft for review purposes only and does not constitute Agency policy.
5-24 DRAFT—DO NOT CITE OR QUOTE
-------
1 indicating the time periods when fiber measurements were not available ('Estimated') and were
2 available ('Measured').
3 In brief, the starting point for the JEM was the measured or estimated concentration of
4 fibers in air (fibers/cc) of each department from 1957-2000. The distribution of exposure by
5 department is summarized in Figure 5-1. Using available data on the year of hire and the
6 departments in which each person worked, the cumulative exposure (fibers/cc-year) for each
7 worker for each year since the date of hire was estimated. Each worker's cumulative exposure
8 was then adjusted to a cumulative human equivalent exposure for continuous exposure (CHEEC;
9 fibers/cc-year) to represent exposure 24 hours/day and 365 days/year (assuming that any
10 exposure off site was zero) for the full duration of employment. Adjustments for different
11 inhalation rates in working versus nonworking time periods were incorporated in this analysis.
12 The calculated value is similar to what EPA usually refers to as continuous human equivalent
13 exposure (U.S. EPA, 1994b). These calculations are somewhat more complex than the usual
14 conversions to equivalent continuous exposure concentrations that EPA makes in the analysis of
15 occupational studies. Conversions for noncancer effects are usually made using an adjustment
16 factor of 240 days - 365 days x 10 m3 + 20 m3 (U.S. EPA. 1994b). However, the adjustment
17 factor in this current assessment takes into account the extensive seasonal overtime for some job
18 codes at the Marysville facility, as well as other annual periods when work hours were reduced
19 (see Appendix F). The estimated CHEEC was used to represent Libby Amphibole asbestos
20 exposure in all subsequent analyses because it combines aspects of both intensity of exposure
21 and duration of exposure.30 For Libby Amphibole asbestos, the exposure metric is calculated as
22 cumulative exposure (fibers/cc-year). Cumulative exposure is a commonly evaluated exposure
23 metric in occupational studies, especially for mineral fibers, where fiber retention may be
24 relevant to toxicity. It should be noted that discrete parietal plaques have often been associated
25 with other exposure metrics (e.g., mean exposure, TSFE) (i.e., Paris et al., 2008; Jakobsson et al.,
26 1995: Ehrlich et al.. 1992: Copes etal.. 1985). Paris et al. (2008) show significant
27 exposure-response relationships for both mean and cumulative exposure metrics for pleural
28 plaques (identified by HRCT) among workers with mixed fiber exposures, when accounting for
29 age, smoking, and TSFE. Mean exposure provided a better overall fit (Paris et al., 2009). Thus,
30 EPA has conducted an uncertainty assessment for the RfC derivation from the sub-cohort by also
30The University of Cincinnati used the term CHEEC in its report (see Appendix F).
This document is a draft for review purposes only and does not constitute Agency policy.
5-25 DRAFT—DO NOT CITE OR QUOTE
-------
1 exploring alternative methods to weight the BMCLio in units of cumulative exposure, to
2 represent the average exposure needed for RfC derivation (see Section 5.3.7).
3 Because localized pleural thickening does not generally occur immediately after exposure
4 and requires some time to develop to the state that it can be detected on a conventional chest
5 X-ray, exposures that occur close to the time of X-ray may not contribute to the occurrence of
6 observable disease and may obscure the exposure-response relationship. Accordingly, a lagged
7 exposure (i.e., cumulative exposure discounting the most recent time period) may be the most
8 appropriate measure to use. Therefore, exposure estimates with various lags were investigated
9 (lags of 0, 5, 10, 15, and 20 years). For example, a CHEEC value based on a lag of 5 years
10 excludes all exposures that occurred within 5 years of the date of X-ray. Looking at the
11 occurrence of the outcome for various categories of time elapsed since first exposure, the first
12 localized pleural thickening was detected -10 years after the first exposure.
13
14 5.2.3.2. Data Sets for Modeling Analyses
15 The individual health outcome data for all workers who participated in the Lockey et al.
16 (1984) study and the follow-up study by Rohs et al. (2008) were used for exposure-response
17 modeling. To avoid any bias from previous occupational exposure to asbestos, only the data
18 from those who did not report any previous occupational exposure to asbestos were used. The
19 data from Lockey et al. (1984) and Rohs et al. (2008) were combined for the full cohort to
20 provide a greater range in time from first exposure (described below). Outcome assessments,
21 i.e., chest X-rays, were performed at two different time points, 1980 and 2002-2005. While the
22 evaluation approaches were generally similar (independent readings by three certified
23 B-readers), it is important to note that X-ray readings were performed by different individuals,
24 under a different reading protocol in 1980 (modified 1971 ILO standards) compared to 2000s
25 [ILO (2002) standards], leading to some uncertainty in statistical analyses that combine these
26 data sets. An additional consideration is human body composition—in some cases, difficulty in
27 distinguishing fat pads from true pleural thickening may lead to misclassification of the outcome.
28 BMI measurements are available for the latter study but not for the 1980 evaluation; the effect of
29 BMI was investigated and is discussed below.
30 Radiographs were evaluated by two B-readers with a consensus evaluation by a third
31 reader in the case of disagreement in the original study by Lockey et al. (1984). In the follow-up
This document is a draft for review purposes only and does not constitute Agency policy.
5-26 DRAFT—DO NOT CITE OR QUOTE
-------
1 by Rohs et al. (2008), a radiographic reading was considered positive "when the median
2 classification from the three independent B readings was consistent with pleural and/or
3 interstitial changes" (p. 631). Because the ILO criteria were updated in 2000, the reader forms
4 from Lockey et al. (1984) showing pleural changes were evaluated for consistency with the ILO
5 2000 criteria. This reevaluation did not result in any change in the diagnosis for any individual
6 from the 1980 reading.31 In addition, no difference in reported X-ray quality was noted between
7 the Lockey et al. (1984) data and the follow-up by Rohs et al. (2008).
8 The full data set of the exposure-response relationship for localized pleural thickening
9 was as follows. The radiographic data from Lockey et al. (1984) (n = 513) and Rohs et al.
10 (2008) (n = 280), were combined for a total of 793 X-ray evaluations (this includes repeated
11 X-rays on the same individual). X-rays obtained from workers who reported exposure to
12 asbestos at other locations were excluded from consideration (n = 793 - 105 = 688 X-ray
13 evaluations).
14 For workers who were X-rayed in both Lockey et al. (1984) and Rohs et al. (2008), one
15 of the observations was excluded so that there were no repeat observations for individual
16 workers in the data set used for modeling. For workers who were negative for localized pleural
17 thickening in Lockey et al., the (1984) study data were excluded, and the Rohs et al. (2008) data
18 were retained. For workers who were positive for localized pleural thickening in Lockey et al.
19 (1984) and also in Rohs et al. (2008), the 1984 study data were retained. One worker was
20 positive in 1984 and negative in 2008 (removing this worker from the analysis did not change
21 results). The 2008 study data were retained for this worker. This procedure resulted in « = 688
22 X-rays - 252 duplicates = 436 X-rays, representing 436 individual workers.
23 Two workers from Lockey et al. (1984) were excluded because the start day and the
24 X-ray date were the same (n = 436 - 2 = 434). For each worker, the estimated cumulative
25 exposure corresponded to the date of the X-ray retained for analysis—if the 1980 X-ray was
26 used, the individual's cumulative exposure estimate covered the period from start of work
27 through the X-ray date in 1980. If the 2002-2005 X-ray was used, cumulative exposure covered
28 the period from start of work through the date of job stop or 2000, whichever occurred earlier.
31Personal communication (e-mail) from Dr. James Lockey, University of Cincinnati, to Dr. Robert Benson in
March 2011 reports that a review of the 1980 B-reader forms using the ILO 2000 guidelines would not result in
changes in individual diagnosis for study participants.
This document is a draft for review purposes only and does not constitute Agency policy.
5-27 DRAFT—DO NOT CITE OR QUOTE
-------
1 The Marysville cohort data comprise 434 workers who were not previously exposed to
2 asbestos and had at least one X-ray observation. Because the concentration of Libby Amphibole
3 asbestos in workplace air was estimated rather than measured for all years prior to 1972, this data
4 set was stratified into two subsets: (1) workers hired in 1972 or after (for whom all exposure
5 values are measured), and (2) workers hired before 1972 (for whom some of the exposure values
6 are estimated). Distributions of cases and TSFE (7) at each outcome assessment are shown in
7 Table 5-3.
9
10
11
12
Table 5-3. Distribution of cases and time from first exposure (7) for cohort
of Marysville workers
Examined 1980 (Lockey et
al., 1984)
Examined 2002-2005 (Rohs
etaL 20081
Marysville cohort
(n = 434, examination in
either 1980 or
2002-2005)
All participants"
Cases/Total
5/434
57/252
61/434
Range of T
0.42-23.43
23.14-47.34
0.42-47.34
First exposed before 1972
Cases/Total
4/236
45/133
48/236
Range of T
8.75-23.43
31.07-47.34
8.75-47.34
First exposed 1972 or later
Cases/Total
1/198
12/119
13/198
Range of T
0.42-8.42
23.14-32.63
0.42-32.63
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
aThe 252 individuals examined in 2002-2005 were also examined in 1980. Note that there were originally
513 individuals in the Lockey et al. (19841 cohort; of these, 77 had previous asbestos exposure and were excluded
(n = 436). Two individuals were excluded because their X-ray date was the same as their employment start date
(n = 434). These exclusions are also reflected in the Rohs et al. (2008) cohort.
Source: Rohs et al. (2008) and Lockey et al. (1984).
The more accurate exposure data are considered to be those from 1972 and later, as these
data were based on analytical measurements. Due to the longer follow-up time and additional
covariate information, the most informative outcome data come from the 2002-2005
examination. Based on these considerations, a sub-cohort of the Marysville workers, which
includes data from workers in the 2002-2005 examination, and who began work in 1972 or later
This document is a draft for review purposes only and does not constitute Agency policy.
5-28 DRAFT—DO NOT CITE OR QUOTE
-------
1 (12 cases of localized pleural thickening and 106 unaffected individuals32) (Rohs et al., 2008),
2 was chosen as the preferred analysis to develop a point of departure (POD) for localized pleural
3 thickening to serve as the basis for the RfC. Additionally, sample POD estimates based on
4 statistical analyses of results from the full cohort [Lockey et al. (1984) and Rohs et al. (2008)
5 combined, as described above] were included for comparison.
6
7 5.2.3.3. Statistical Modeling of the Sub-cohort
8 EPA performed analyses of study results for the sub-cohort whose exposures began on or
9 after 1/1/1972 when workplace PCM measurements were available, reducing uncertainties
10 associated with exposure assessment. Localized pleural thickening (LPT), as diagnosed from a
11 standard radiograph (ILO, 2002), was selected as the critical effect based on the health effects
12 associated with pleural thickening specific to this diagnosis (see Section 5.2.2.3). Alternative
13 critical effects were not considered for the sub-cohort analysis given the limited number of cases
14 (one case of DPT and no cases of small opacities). Epidemiologic methods were used to analyze
15 the exposure-response data, and benchmark concentration (BMC) methodology was used to
16 estimate PODs. In this approach, the available data are fit to a set of mathematical
17 exposure-response models to determine an appropriate empirical representation of the data.
18 General model fit is evaluated to determine whether the model form appropriately represents the
19 data; here, this was done using the Hosmer-Lemeshow test (a form of the Pearson %2
20 goodness-of-fit statistic). Among models with adequate general fit, a recommended model form
21 is then determined; commonly, this is the model with the best fit as measured by Akaike's
22 Information Criterion (AIC) value among these model forms judged to provide an appropriate
23 and statistically adequate representation of the data. For inhalation data, the BMC is defined as
24 the exposure level, calculated from the best-fit model, which results in a specified benchmark
25 response (BMR). The RfC is derived from the lower 95% confidence limit of the BMC, referred
26 to as the BMCL, which accounts for statistical uncertainty in the model fit to the data. All
32There was one individual whose radiographic examination indicated diffuse pleural thickening, who was excluded
from further analyses of the preferred sub-cohort. Diffuse pleural thickening represents a more severe outcome than
the selected critical effect of LPT—including this individual as a case would not be appropriate given that the
critical effect is selected to represent a most sensitive endpoint, and the subsequent selection of a benchmark
response in modeling efforts. Diffuse pleural thickening is considered separately as an endpoint (with appropriate
benchmark response) in sensitivity analyses of alternative outcomes in the larger group of workers examined in
2002-2005 (see Section 5.3.8).
This document is a draft for review purposes only and does not constitute Agency policy.
5-29 DRAFT—DO NOT CITE OR QUOTE
-------
1 analyses were performed using SAS® statistical software v. 9.1. BMCLs were obtained by the
2 profile likelihood method as recommended by Crump and Howe (1985) using the NLMIXED
3 (nonlinear mixed modeling) procedure in SAS (Wheeler, 2005) (see Appendix E for details).
4 For models where a background parameter is included, a 1% risk of localized pleural
5 thickening was assumed. Establishing a background rate for LPT prevalence is problematic for
6 several reasons. Little data exist to define background rates for LPT, as this designation is more
7 recent, and the majority of the published data use earlier ILO guidelines, which define discrete
8 pleural plaques (DPP). Secondly, it is difficult to define a population without exposure to
9 asbestos in any setting. As environmental and community exposures can increase pleural
10 thickening (Weill etal.. 2011: Luoet al.. 2003: Hiraoka et al.. 1998: Zittingetal.. 1996) the
11 question arises, Is there a true background rate? Also, in general, pleural thickening increases
12 with both age and TSFE in a population. There is a study that reports the LPT in Libby
13 community members with no reported pathways of exposure (Weill et al., 2011). LPT
14 prevalence is reported at 0.4% in participants age 25-40, and 1.4% in participants age 41-50
15 (based on X-rays taken in 2000). Older study participants (61-90) had a LPT prevalence of
16 12.7%, likely influenced by high historical exposures, as well as the increased TSFE. In two
17 studies of persons not known to be previously exposed to asbestos, Anderson et al. (1979) and
18 Castellan et al. (1985) report DPP estimated prevalence of 1.2% (4/326) and 0.2% (3/1,422),
19 respectively. In cross-sectional studies, which may include persons with occupational exposure
20 to asbestos, Rogan reported DPP prevalence estimates of 1.2% in the National Health and
21 Nutrition Examination (NHANES) I study (1971-1975) (Rogan et al.. 1987) and 3.9% in the
22 NHANES II study (Rogan et al., 2000). Among military populations, two studies have reported
23 an estimated DPP prevalence of 2.3% (Muller et al.. 2005: Miller and Zurlo. 1996). Based on
24 these reports, the 1% background rate was chosen as representing the prevalence among persons
25 without occupational exposure to asbestos in the age range of the Rohs et al. (2008) study
26 population. As there is some uncertainty regarding the true background rate for LPT, a
27 sensitivity analysis was performed where the model includes the background rate as an estimated
28 parameter rather than using the set value of 1%. There was little change in the resulting model
29 fits or BMCLs (see Section 5.3.4).
30 In the absence of agent-specific information to assist in identifying a BMR, a 10% extra
31 risk was judged to be a minimally biologically significant level of change, and is also
This document is a draft for review purposes only and does not constitute Agency policy.
5-30 DRAFT—DO NOT CITE OR QUOTE
-------
1 recommended for standard reporting purposes (U.S. EPA, 2000a). LPT is an irreversible
2 pathological change and associated with health effects including chronic pain, dyspnea, and
3 deficits in pulmonary function (see Section 5.2.2.3). The likelihood and severity of these health
4 effects increases with increased extent and severity of the pleural thickening. However, as the
5 data from the critical study do not provide information on the severity of the lesions, we cannot
6 assess the relative likelihood of any of these health effects. Thus, the observed LPT prevalence
7 may include a range of lesions from minimally adverse to severe. The biology of more severe
8 lesions (i.e., DPT and small opacities) could justify lower BMRs; however, there are not enough
9 cases to model these endpoints in this sub-cohort. A sensitivity analysis was conducted using the
10 data set included in Rohs et al. (2008) to examine the impact of choice of BMR and critical
11 effect on the POD (see Section 5.3.8).
12
13 5.2.3.3.1. Statistical model evaluation and selection
14 Dichotomous statistical models describing the probability of individual response as a
15 function of cumulative exposure (represented by CHEEC in units of fibers/cc-year) were used.
16 In order to investigate the key explanatory variables for analysis, a forward-selection process was
17 used to evaluate the association of each of the potential covariates with the risk of localized
18 pleural thickening, controlling for Libby Amphibole asbestos exposure. Covariates considered
19 for inclusion in the model were TSFE (7), age at X-ray, gender, smoking history, and BMI. This
20 initial modeling was done using a standard logistic regression model, as is commonly applied in
21 analysis of epidemiological data. The base model was a logistic regression model with
22 cumulative Libby Amphibole asbestos exposure (natural log transformed) as the independent
23 variable. This model provided an adequate fit to the data (Hosmer-Lemeshow/>-value of 0.64),
24 and the exposure variable was statistically significantly associated with the outcome
25 (beta = 0.5676, standard error, [SE] = 0.2420 increase in log odds for every unit increase in
26 CHEEC, />-value = 0.02). Covariates were evaluated according to whether inclusion of the
27 covariate improved model fit as assessed by the AIC, and statistical significance of the covariate.
28 When controlling for Libby Amphibole asbestos exposure, none of these covariates were
29 associated with odds of localized pleural thickening: T: />-value = 0.89; age at X-ray:
30 />-value = 0.77; gender: />-value = 0.78; smoking history: />-value = 0.17; BMI: p-value = 0.41.
31 The inclusion of each of the covariates with the exception of smoking increased the AIC for the
This document is a draft for review purposes only and does not constitute Agency policy.
5-31 DRAFT—DO NOT CITE OR QUOTE
-------
1 model, and the improvement in model fit with the addition of smoking was marginal (decrease of
2 0.1 AIC units). Therefore, only cumulative Libby Amphibole asbestos exposure (CHEEC) was
3 included in further analyses, although sensitivity analyses were performed to investigate the
4 potential impact of smoking (see Section 5.3.6 and Appendix E).
5 The candidate models (see Table 5-4 for model forms) were logistic (with CHEEC
6 considered as continuous, and continuous with a natural logarithm transformation), probit (with
7 CHEEC considered as continuous, and continuous with a natural logarithm transformation),
8 3-parameter log-logistic, dichotomous Hill, and dichotomous Michaelis-Menten models (with
9 only CHEEC for the latter three models). These are statistical models used to evaluate
10 dichotomous data that were considered appropriate here given the supralinear nature of the
11 observed relationship between Libby Amphibole asbestos exposure and prevalence of localized
12 pleural thickening. For each of the candidate models, exposure lags of 0, 5, 10, 15, and 20 years
13 were investigated. Although zero lag exposures are not likely to be biologically relevant (i.e.,
14 some lag is expected for development of LPT), these models were included for completeness and
15 for comparison of relative model fits. Similarly, although we explored models with exposure
16 lagged by 20 years, there were cases of localized pleural thickening in the full cohort with fewer
17 than 20 years since first exposure; therefore, using such a long lag (which necessitates the
18 assumption that these are background cases) was not judged to be appropriate, and the results are
19 not further considered; these models are indicated by gray shading in Table 5-4. Further details
20 of these analyses are included in Appendix E.
21 All of the candidate models had adequate fit. Models were compared using the AIC—
22 values were quite similar among the candidate models, ranging from 74.0 to 77.8 (see
23 Table 5-4). The model with the lowest AIC was the Michaelis-Menten model with 10-year
24 lagged exposure (AIC = 74.0). For this model form, the AIC values did not vary much for lags
25 of 5 to 15 years, but the 10-year lagged exposure provided the lowest AIC and was selected as
26 the preferred exposure metric. There were several models that had similar model fits (within
27 2 AIC units, a proximity that can be considered to be a range that cannot clearly differentiate
28 between models) (Burnham and Anderson, 2002) as the best-fitting model, including the logistic
29 and probit models with the natural log of CHEEC as the exposure metric (lags of 5, 10, and
30 15 years), the 3-parameter log-logistic model (lags of 5, 10, and 15 years), the Dichotomous Hill
31 model (lag of 10 years), and the Michaelis-Menten model with exposure lagged by 5 or 15 years.
This document is a draft for review purposes only and does not constitute Agency policy.
5-32 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
Table 5-4. Candidate models for association between cumulative Libby
Amphibole asbestos exposure in the Marysville sub-cohort and localized
pleural thickening
Model
Logistic
CHEEC, lag 5
CHEEC, lag 10
CHEEC, lag 15
CHEEC, lag 20
Logistic
CHEEC, lag 5
CHEEC, lag 10
CHEEC, lag 15
CHEEC, lag 20
Probit model
CHEEC, lag 5
CHEEC, lag 10
CHEEC, lag 15
CHEEC, lag 20
Probit model
CHEEC, lag 5
CHEEC, lag 10
CHEEC, lag 15
CHEEC, lag 20
3 -parameter
log-logistic
CHEEC, lag 5
CHEEC, lag 10
CHEEC, lag 15
CHEEC, lag 20
Exposure
metric
CHEEC
In(CHEEC)
CHEEC
In(CHEEC)
In(CHEEC)
Form"
P(LPT) = 1/[1 + exp(-a -6*CHEEC)]
P(LPT)= !/[! +
exp(-a-6*ln(CHEEC))]
P(LPT)=<$>(a + 6*CHEEC)
P(LPT) = (a + 6*ln(CHEEC))
P(LPT) = bkg + (1 - bkg)/[l + exp(-a
-6*ln(CHEEC))]
AIC
77.7
77.5
77.4
77.6
77.8
75.5
75.2
74.6
74.7
75.4
77.2
77.0
77.0
77.2
77.4
76.0
75.7
75.2
75.0
75.7
74.9
74.6
74.1
74.3
75.2
Hosmer-Leme
show GOF
/7-value
0.7423
0.6914
0.6751
0.6474
0.8800
0.6537
0.5454
0.5708
0.6620
0.8152
0.7698
0.7146
0.6864
0.6645
0.8884
0.6041
0.4967
0.5385
0.6166
0.7945
0.7030
0.4894
0.5853
0.7238
0.8277
BMC
~
1.5245
1.4734
1.4510
~
~
0.2281
0.2028
0.1686
~
~
1.3773
1.3336
1.3148
~
~
0.2066
0.1843
0.1544
~
~
0.3096
0.2696
0.2193
~
BMCL
~
0.8836
0.8540
0.8242
~
~
0.0601
0.0591
0.0463
~
~
0.8481
0.8048
0.7776
~
~
0.0502
0.0496
0.0441
~
~
0.0979
0.0888
0.0693
~
This document is a draft for review purposes only and does not constitute Agency policy.
5-33 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-4. Candidate models for association between cumulative Libby
Amphibole asbestos exposure in the Marysville sub-cohort and localized
pleural thickening (continued)
Model
Dichotomous Hillb
CHEEC, lag 5
CHEEC, lag 10
CHEEC, lag 15
CHEEC, lag 20
Michaelis-Menten0
CHEEC, lag 5
CHEEC, lag 10d
CHEEC, lag 15
CHEEC, lag 20
Exposure
Metric
In(CHEEC)
In(CHEEC)
Form*
P(LPT) = bkg + (Plateau -
bkg)*CHEECb/[exp(-a) + CHEECb]
P(LPT) = bkg + (Plateau -
bkg)*CHEEC/[exp(-a) + CHEEC]
AIC
76.9
76.5
76.0
76.2
77.2
74.9
74.5
74.0
74.3
76.0
Hosmer-Leme
show GOF
/7-value
0.6040
0.3598
0.4244
0.6659
0.8277
0.5243
0.3351
0.4163
0.5664
0.5610
BMC
~
0.3083
0.2640
0.2112
-
~
0.3096
0.2642
0.2097
-
BMCL
~
0.1015
0.0923
0.0724
-
~
0.1352
0.1177
0.0898
-
1 abkg indicates background rate, fixed at 1%.
2 bFor statistical modeling, the equivalent model form was used: P(PT) = bkg + (Plateau - bkg)/[l + exp(-a -
3 b*ln(CHEEC))].
4 Tor statistical modeling, the equivalent model form was used: P(PT) = bkg + (Plateau - bkg)/[l + exp(-a -
5 ln(CHEEC))].
6 Parameter estimates for the best-fitting models are as follows:
7 intercept = -0.1801 (SE = 1.0178), plateau = 0.5577 (SE = 0.3568, ^-value = 0.1207).
8
9
10 The range was relatively narrow among these similarly fitting models (BMCLs ranging from
11 0.0441 to 0.1352), with the lowest BMCL -2.7 times lower than the BMCL for the
12 Michaelis-Menten model, with exposure lagged by 10 years.
13 The potential confounding effect of covariates was reexamined in the best-fitting model.
14 As in the initial assessment, after controlling for the effect of exposure (CHEEC, lagged by
15 10 years), there was no association between risk of LPT and TSFE (p-value = 0.997), age at
16 X-ray (p-value = 0.87), gender (p-value = 0.55) or BMI (p-value = 0.38), and inclusion of each
17 of these covariates increased the AIC (with the exception of BMI, due to missing information for
18 some individuals). The variable representing smoking history did not meet the alpha = 0.05
19 criterion for statistical significance (p-value = 0.08), although inclusion of this variable decreased
20 the AIC from 74.0 in the best-fitting model, to 72.3. Smoking was not considered further in the
This document is a draft for review purposes only and does not constitute Agency policy.
5-34 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
o
J
\
5
6
7
10
11
12
13
14
15
16
17
18
19
20
21
22
23
derivation of the RfC due to the lack of statistical significance at the alpha = 0.05 level.
However, because inclusion of the smoking variable did improve model fit, it is investigated
further as a sensitivity analysis (see Section 5.3.6 and Appendix E).
The Michaelis-Menten model using the 10-year lagged exposure had a/?-value for fit of
0.42, an AIC value of 74.0, and an estimated intercept = -0.1801 (SE = 1.0178) and plateau of
0.5577 (SE = 0.3568) (see Figure 5-2). This model yielded a BMCio of 0.2642 fibers/cc-year,
and corresponding BMCLio of 0.1177 fibers/cc-year for a 10% increase in prevalence of
localized pleural thickening. This BMCLio of 0.1177 fibers/cc-year is the preferred POD
estimate to support development of an RfC for Libby Amphibole asbestos.
Prevalence of localized pleural thickening
p p p p p p
o ^ is> L*J 4^ Ln b>
• • •••••• Estimated prevalence by
.•* 10 years
•
t
." quartilesof unlagged
•* cumulative exposure amon§
.•* cases
f
12468
CHEEC (flbers/cc-yr)
I by
10
Figure 5-2. Graph of observed and estimated prevalence of localized pleural
thickening calculated using the Michaelis-Menten model with 10-year lagged
exposure.
5.2.4. RfC Derivation—Including Application of Uncertainty Factors (UFs)
Among the available studies that could provide exposure-response data for the
relationship between Libby Amphibole asbestos exposure and risk of localized pleural
thickening (LPT), consideration of study attributes led to the selection of a study of the
Marysville, OH worker cohort as the primary data set for RfC derivation (Rohs et al., 2008) (see
Section 5.2.1). An updated job-exposure matrix is available for this follow-up of the original
This document is a draft for review purposes only and does not constitute Agency policy.
5-35 DRAFT—DO NOT CITE OR QUOTE
-------
1 cohort described by Lockey et al. (1984). The updated job-exposure matrix provides a more
2 refined understanding of exposure to Libby Amphibole asbestos throughout plant operation (see
3 Section 5.2.3.1 and Appendix F). However, due to remaining uncertainties in exposures prior to
4 1972, EPA elected to model a sub-cohort of plant employees that consisted of individuals who
5 began their employment in 1972 or later (see Section 5.2.3.2). It is acknowledged that although
6 this provides a sub-cohort with less potential for exposure misclassification, there is reduced
7 power due to fewer individuals and fewer observed cases. Therefore, EPA provides a supporting
8 analysis using the combined results for the Marysville plant workers as reported in both the
9 original study and in the update (Rohs et al.. 2008: Lockey etal.. 1984)1 (Section 5.2.5).
10 LPT is an irreversible pathological change associated with constricting chest pain,
11 dyspnea, and decreased pulmonary function and, therefore, it is selected as the critical effect in
12 the sub-cohort. The Michaelis-Menten model, with a 10-year lag for exposure, provided the best
13 model fit for the sub-cohort data (AIC = 74.0, see Table 5-4). Using a 10% BMR for LPT, a
14 BMC of 0.2642, and a BMCLio of 0.1177 (fibers/cc)-years were calculated (see Table 5-4). As
15 this POD is in units of cumulative exposure, and the RfC is given in continuous lifetime
16 exposure, the POD was adjusted to 70 years of exposure, lagged by 10 years (nonoccupational,
17 lifetime exposure). Thus the adjusted lifetime BMCLio is 1.96 x 10"3 fibers/cc (as derived
18 below), and is the POD for RfC derivation.
19
20
21 Lifetime-BMCLio = BMCLio ^ (lifetime exposure duration)
22 = [0.1177 (fibers/cc) x year] + [70 - 10 years]
23 = 1.96 x 10"3 fibers/cc
24
25
26 Following EPA practices and guidance (U.S. EPA, 2002, 1994b), application of the
27 following uncertainty factors was evaluated resulting in a composite UF of 100.
28
29
30 • An interspecies uncertainty factor, UFA, of 1 is applied for extrapolation from animals
31 to humans because the critical effect used as the basis for the RfC was observed in
32 humans.
This document is a draft for review purposes only and does not constitute Agency policy.
5-36 DRAFT—DO NOT CITE OR QUOTE
-------
1 • An intraspecies uncertainty factor, UFn, of 10 was applied to account for human
2 variability and potentially susceptible individuals in the absence of quantitative
3 information to assess the toxicokinetics and toxicodynamics of Libby Amphibole
4 asbestos in humans. Only adults sufficiently healthy for full-time employment were
5 included in the principal study and the study population was primarily male.
6 • A LOAEL to NOAEL uncertainty factor, UFL, of 1 was applied because the current
7 approach is to address this factor as one of the considerations in selecting a BMR for
8 BMC modeling. In this case, a BMR of 10% extra risk was considered to be
9 minimally biologically significant.
10 • A subchronic-to-chronic uncertainty factor, UFs, of 1 was applied because the
11 selected POD is from a study population including chronic exposure (Rohs et al.,
12 2008). The average employment duration for the sub-cohort corresponding for the
13 RfC derivation is 18.7 years (SD = 8.6; range = 0.3-29.0).
14 • A database uncertainty factor, UFo, of 10 was applied to account for database
15 deficiencies in the available literature for the health effects of Libby Amphibole
16 asbestos. Although there is a large database for asbestos in general, only three study
17 populations exist for Libby Amphibole asbestos specifically: the Marysville, OH
18 worker cohort, the Libby worker cohort and the ATSDR community screening (which
19 includes some Libby worker cohort participants). Limitations of these studies are
20 described below.
21 1. Evidence exists for an association between exposure to Libby Amphibole asbestos
22 and other noncancer health effects with no exposure-response information.
23 Without additional data, it is unknown if a lower POD or RfC would be derived
24 for these effects.
25 a. Two studies have found a possible increased prevalence of autoimmune
26 disease and biological markers for autoimmune disease in Libby residents
27 (Noonan et al., 2006: Pfau et al., 2005), although these studies do not
28 indicate whether the autoimmune effects would be observed at exposures
29 lower than that observed for localized pleural thickening. Subsequent
30 animal studies have indicated that exposure to Libby Amphibole asbestos
31 does induce auto-antibodies in mice (Blake et al., 2008).33
32 b. A mortality analysis for the Libby worker cohort also found associations
33 between occupational exposures to Libby Amphibole asbestos and
34 mortality due to cardiovascular disease (Larson et al., 201 Ob).
33It is unknown if autoimmune effects are secondary to the chronic inflammatory response expected from exposure
to mineral fibers. However, one study of individuals in a community exposed to tremolite found changes in immune
parameters in exposed individuals without localized pleural thickening, and that additional immune markers,
including autoantibodies, increased in individuals with localized pleural thickening (ZervaetaL 1989).
This document is a draft for review purposes only and does not constitute Agency policy.
5-37 DRAFT—DO NOT CITE OR QUOTE
-------
1 c. Deficits in pulmonary function have been documented in those exposed to
2 Libby Amphibole asbestos occupationally or in the community. However,
3 exposure data are lacking to define an exposure response relationship on
4 this sensitive endpoint (Weill etal., 2011; Whitehouse, 2004).
5 2. There are no data in laboratory animals or humans on general systemic effects for
6 Libby Amphibole asbestos. However, it is known that inhaled asbestos fibers
7 migrate out of the lung and into other tissues (see Section 3.1), lending
8 uncertainty to any assumptions that other effects would not be expected.
9 3. Although data do exist to define an exposure-response relationship for
10 radiographic abnormalities in the Marysville, OH worker cohort, these data are
11 limited by the dates of the available radiographs. The data for the sub-cohort of
12 workers exposed post-1972 allowed for assessing prevalence of LPT up to
13 approximately 30 years after first exposure (Mean = 28.2 years,
14 range = 23.2-32.7 years). However, there is evidence to indicate that the
15 prevalence of pleural plaques and pleural thickening in general is likely to
16 continue to increase more than 30 years after first exposure (Paris et al., 2009;
17 Paris et al.. 2008; Jakobsson et al.. 1995; Hillerdal 1994; Ehrlich et al.. 1992;
18 Jarvholm. 1992; Lilis et al.. 1991; Merchant 1990; McDonald et al.. 1986b). As
19 the RfC is intended for a lifetime of exposure, and pleural thickening is known to
20 progress across the lifetime (even with less-than-lifetime exposures), the lack of
21 health data assessed at end of lifetime is a data gap.
22
23
24 The derivation of the RfC from the morbidity studies of the Marysville, OH worker
25 cohort [i.e., Rohs et al. (2008)1 was calculated from a POD, lifetime-BMCLio of 1.96 x io~3
26 fibers/cc for localized pleural thickening, (adjusted to 70 years of exposure, lagged by 10 years
27 (nonoccupational, lifetime exposure), and dividing by a composite uncertainty factor (UF) of
28 100.
29 As derived below, the chronic RfC is 2 x 10 5 fibers/cc for Libby Amphibole asbestos
30 and was calculated by dividing the lifetime-POD by a total UF of 100:
31
32
33 Chronic RfC = Lifetime-BMCLio - UF
34 = 1.96 x 10'3 fibers/cc - 100
35 = 1.96 x io~5 fibers/cc, rounded to 2 x io~5 fibers/cc
36
37
This document is a draft for review purposes only and does not constitute Agency policy.
5-38 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.2.5. Alternative Analyses of the Full Marysville Cohort
2 Modeling of the full cohort was also conducted utilizing the full data set for localized
3 pleural thickening from the Marysville cohort. Since the full cohort includes data combined
4 from Lockey et al. (1984) and Rohs et al. (2008), there were individuals who had more than one
5 observation. As described in Section 5.2.3.2, for those workers X-rayed in both 1980 (Lockey et
6 al.. 1984) and 2004-2005 (Rohs et al.. 2008). one of the observations was excluded so that there
7 are no repeat observations for individual workers in the data used for the modeling.
8 Time from first exposure to X-ray (the variable T, in this model) is an important variable
9 in understanding the full Marysville data set, as can be seen by the much higher prevalence of
10 localized pleural thickening in the 2000s compared to the 1980 assessment, an increase which
11 cannot be fully explained by the increases in cumulative exposure occurring with continued
12 exposure. Consequently, in looking at the full cohort, Tis a strong predictor of localized pleural
13 thickening. Study T-values are measures of the time from first exposure to the event that an
14 X-ray was taken that detected an abnormality. As such, these values in themselves are not
15 measures of biological latency—an abnormality may be present for some time before the event
16 that an X-ray is taken. Given the occurrence of higher exposures in earlier years in this study,
17 higher T-values correspond to individuals who likely experienced the early higher intensity
18 exposures. This may lead to some uncertainty in the estimated models because uncertainty in the
19 estimated exposures can influence the apparent relationship between T and lesion prevalence. A
20 similar approach as described in Section 5.2.3.3.1 was used to evaluate candidate models for the
21 full cohort. Details are provided in Appendix E. However, as time from first exposure (7) was
22 an important covariate for these analyses, further efforts were needed to develop a model
23 incorporating T along with cumulative exposure. The logistic and probit models including
24 CHEEC as a continuous exposure had inadequate model fit as evaluated using the
25 Hosmer-Lemeshow test (p-values of 0.003 for both) and so were not considered for further
26 analysis. The remaining candidate models (logistic and probit with the natural logarithm of
27 CHEEC, 3-parameter log-logistic, dichotomous Hill, and dichotomous Michaelis-Menten) had
28 adequate fit. Among these models, the AIC values ranged from 327.9 (Michaelis-Menten)
29 to 346.8 (logistic with the natural logarithm of CHEEC) (see Appendix E). Based on these
30 results, the Michaelis-Menten model was selected for further evaluation, and different
This document is a draft for review purposes only and does not constitute Agency policy.
5-39 DRAFT—DO NOT CITE OR QUOTE
-------
1 approaches were investigated to represent T along with cumulative exposure to Libby Amphibole
2 asbestos using this model form.
3 The approach taken to incorporate Twas through modification of the plateau term in the
4 Michaelis-Menten model to allow the plateau for the exposure-response relationship to change
5 for different values of T. After investigating various forms for the plateau (described in
6 Appendix E), the plateau term used took the form: Plateau = Background + (1-background) x
7
-------
1 Another alternative analysis is based on projection of risks using the full cohort model for
2 a "lifetime" time from first exposure of 70 years. Note that none of the workers had a
3 T> 50 years; therefore, this modeling represents a mathematical extrapolation beyond available
4 data. A BMCLio of 0.0042 fibers/cc-year was calculated using the Cumulative Normal
5 Michaelis-Menten model. This POD combined with a lag time of 5 years and a total UF
6 of 30 was used to derive an alternative RfC of 2.1 x 10 6 fibers/cc, or rounding to one significant
7 digit, 2 x 10 6 fibers/cc. See Appendix E for details.
8 Each of the candidate PODs (analyses from both the sub-cohort and full cohort) has
9 strengths and weaknesses. A major strength of the preferred analysis (Marysville sub-cohort) is
10 that by limiting the data set to those individuals hired in 1972 or later, the exposure
11 reconstruction relies only on data supported by industrial hygiene measurements in the facility.
12 The exposures were also lower after 1972 as compared to previous years. However, this
13 approach reduces the number of individuals in the data set from 434 to 119 and reduces the
14 number of cases from 61 to 12. In addition, this approach narrows the range in the time from
15 first exposure to 23.15-32.65 years (see Table 5-3). The analyses of the full cohort have the
16 strength of using all of the data available on the Marysville cohort and of using a model that
17 incorporates both cumulative exposure and time from first exposure as relevant explanatory
18 variables. One weakness of the full cohort analyses is that the exposure reconstruction relies on
19 estimates of the exposure conditions in the Marysville facility before industrial hygiene data
20 were available in 1972.
21
22 5.2.6. Previous Reference Concentration (RfC) Derivation
23 There is no previous RfC derivation for Libby Amphibole asbestos.
24
25 5.3. UNCERTAINTIES IN THE INHALATION REFERENCE CONCENTRATION
26 (RfC)
27 5.3.1. Uncertainty in the Exposure Reconstruction
28 As in all epidemiologic studies, there are uncertainties in the exposure reconstruction. In
29 this case, there is some uncertainty in the employment history, and some individuals had
30 extensive overtime work. Employment history was self-reported during interviews with each
31 individual for the original study (Lockey et al., 1984), and errors in this process could affect
This document is a draft for review purposes only and does not constitute Agency policy.
5-41 DRAFT—DO NOT CITE OR QUOTE
-------
1 assigned Libby Amphibole asbestos exposure estimates. As stated previously, fiber
2 measurements started in the Marysville plant in 1972; exposures prior to this time were estimated
3 by University of Cincinnati scientists, based on focus group interviews with 15 long-term former
4 workers and the times when engineering changes were made to control dust in the facility (see
5 Appendix F). Exposure estimates for the period prior to 1972, can, thus, be considered as
6 semiquantitative rather than directly based on industrial hygiene data. The University of
7 Cincinnati analysis assumed that early exposure levels in the plant are twice those measured in
8 1972 (see Appendix F). The greater uncertainty of the pre-1972 exposure estimates led to EPA's
9 decision to focus the analysis on the post-1972 group of workers rather than the full cohort.
10 Although it is generally true that the use of more data is an advantage for statistical analyses
11 because it allows for the computation of more statistically precise effect estimates, this increased
12 precision may be offset by a negative impact on the accuracy of the effect estimate if an increase
13 in sample size is accompanied by greater exposure misclassification or other biases.
14 While the uncertainties related to a lack of quantitative measurements are not relevant to
15 the sub-cohort analysis, it is important to recognize that exposure assessment post-1972 also has
16 some limitations. The main sources of uncertainty are incomplete exposure measurements for
17 some of the occupations/tasks before industrial hygiene improvements that started about 1973 or
18 1974 and continued throughout the 1970s (see Appendix F, Figure F-l).
19 There is uncertainty when the Libby ore was first used in the facility. Company records
20 indicated that the date was between 1957 and 1960, and the University of Cincinnati used the
21 best-available information from focus group interviews to assign the first usage of Libby ore in
22 1959 (see Appendix F). There is also uncertainty in the data regarding asbestos content in other
23 ore sources before and after Libby ore use. In 1957 and 1958, only ore from South Carolina was
24 used. From 1959 to 1971, ores from Libby and South Carolina were used. From 1972 to 1980,
25 ores from Libby, South Carolina, South Africa, and Virginia were used with Libby being the
26 major source. Libby ore was not used in the facility after 1980. However, industrial hygiene
27 measurements collected after 1980 showed low levels of fibers in the facility. PCM analysis
28 does not determine the mineral/chemical make-up of the fiber, and, thus, cannot distinguish
29 between different kinds of asbestos.
30 As reported in Appendix C, the EPA analysis of bulk ores from Virginia and South
31 Africa showed the presence of only a few or no Amphibole asbestos fibers; EPA could not obtain
This document is a draft for review purposes only and does not constitute Agency policy.
5-42 DRAFT—DO NOT CITE OR QUOTE
-------
1 a sample of ore from South Carolina. However, the South Carolina ore is known to contain
2 fibers [see Appendix F; U.S. EPA (2000b); McDonald et al. (1988)1. Using the industrial
3 hygiene data, the University of Cincinnati estimated that the fiber content of the South Carolina
4 ore was about 10% of that of the Libby ore (see Appendix F). This result is consistent with data
5 comparing South Carolina and Libby ores from samples tested in 1982 (U.S. EPA, 2000b). EPA
6 believes that the overwhelming exposure to fibers in the Marysville facility is from the Libby
7 ore. Therefore, EPA has attributed all of the adverse health effects to exposure to fibers from
8 Libby ore from 1957 to 1980 and from the post-1980 exposure. However, because the
9 concentration of fibers in the workplace was near background after 1980, the post-1980 time
10 period makes only a small contribution to an individual's cumulative exposure.
11 There was potential coexposure to other chemicals in the Marysville facility (see
12 Section 4.1.3). These other chemicals were used after expansion of vermiculite ore in another
13 area of the facility. Industrial hygiene data showed very low levels of fibers in the areas where
14 the additional chemicals were added to the expanded vermiculite. In addition, none of these
15 chemicals are volatile. The most likely route of exposure to these chemicals is through dermal
16 contact. It is unlikely that any coexposure to these particular chemicals would alter the
17 exposure-response relationship of Libby Amphibole asbestos in the respiratory system (see
18 Sections 4.1.3 and 5.3.1).
19 The University of Cincinnati Research Team assumed that there was no exposure to
20 Libby Amphibole asbestos outside of the workplace. The interviews with the Marysville
21 workers revealed that about 10% of the workers reported bringing raw vermiculite home. These
22 interviews also revealed that changing to street clothes from work-supplied coveralls was
23 standard practice at the end of the shift, and approximately 64% of the workers showered before
24 leaving the workplace. For these workers, it is likely that additional exposure outside the
25 workplace was minimal. However, for the remainder of the workers, it is reasonable to assume
26 that additional exposure could have occurred at home. Additional data collected by the
27 University of Cincinnati Research Team document that no increased prevalence of pleural or
28 parenchymal change consistent with asbestos exposure has been observed in household contacts
29 of the workers from the Marysville facility (J. Lockey, University of Cincinnati, personal
30 communication to Robert Benson, U.S. EPA, 2011).
31
This document is a draft for review purposes only and does not constitute Agency policy.
5-43 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.3.2. Uncertainty in the Radiographic Assessment of Localized Pleural Thickening
2 The use of conventional radiographs to diagnose pleural thickening has several
3 limitations. The localized thickening must be of sufficient size and thickness to be viewed on the
4 X-ray; small lesions may exist but not be reported. More severe and larger lesions are more
5 reliably detected on radiographs. There are also potential interferences. Fat pads may be
6 mistaken as pleural plaques as they generally occur against the ribcage in a similar location
7 (Gilmartin, 1979): this is one source of uncertainty between readers. Although generally related
8 to mineral fiber exposure, pleural plaques may also be a result of trauma to the chest, and pleural
9 thickening may appear after an active TB infection. Often signs of trauma (e.g., fractured ribs)
10 and radiographic signs of past TB infection can be seen and are noted by the reader. In these
11 cases, LPT would not be diagnosed. There is a certain amount of subjectivity when viewing the
12 X-rays determining which features are representative of pleural thickening and if signs of
13 alternative etiology can be noted; thus, several certified readers are generally consulted, and a
14 consensus of opinions determines the diagnosis. Regardless, there is still potential for outcome
15 misclassification. For example, one of the workers in the Marysville cohort had a positive X-ray
16 in the 1980 evaluation but a negative X-ray at the 2002-2005 evaluation (excluding this worker
17 from the analysis did not change results). However, uncertainty in the presence or absence of
18 localized pleural thickening in each individual is considered minimal due to the use of three
19 highly qualified chest radiologists evaluating the radiographic films and the use of the majority
20 vote of the readers for the diagnosis.
21 BMI was investigated as a potential explanatory variable because fat pads can sometimes
22 be misdiagnosed as pleural thickening. BMI was not measured in the 1980 examination but was
23 available for most participants of the 2000s examination. To address whether fat deposits may
24 affect outcome classification, EPA considered the effect of adding BMI as a covariate in the
25 model. However, BMI did not display an association with odds of localized pleural thickening
26 in this population (see Appendix E). While these covariates were not associated with the risk of
27 localized pleural thickening in the sub-cohort after adjusting for exposure, it was not possible to
28 evaluate this relationship in the full cohort. In the general U.S. population, BMIs have increased
29 between 1980 and the 2000s, so one cannot necessarily assume the relationships will be the same
30 for the two examination periods.
31
This document is a draft for review purposes only and does not constitute Agency policy.
5-44 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.3.3. Uncertainty Due to Time From First Exposure
2 There is some uncertainty associated with the length of follow-up of the Marysville
3 cohort. The observed range of TSFE to X-ray in the full cohort is 0.4-47 years, and 23.2-32.7
4 years in the preferred sub-cohort (see Table 5-3). It is anticipated that the prevalence of
5 localized pleural thickening in the study population—and in the post 1972 exposure cohort—
6 may continue to show some increase with passage of time. In this case, the modeling approach
7 may not accurately reflect the exposure-response relationship that would be seen with a longer
8 follow-up time. However, a recent study by Larson et al. (2010a) examined serial radiographs
9 conducted on a group of Libby vermiculite workers with pleural or parenchymal changes. They
10 found that among those workers with localized pleural thickening, all cases were identified
11 within 30 years, and that the median time from hire to the first detection of localized pleural
12 thickening was 8.6 years. Albeit the retrospective evaluation of radiographs is a different and
13 more sensitive procedure, these findings indicate that the range of follow-up time in the
14 Marysville sub-cohort is likely sufficient to support the exposure-response modeling developed
15 in this current assessment. Note that the likelihood that prevalence of localized pleural
16 thickening may further increase beyond 30 years after first exposure is a principal rationale cited
17 for the selection of a database UF of 10 in this current assessment.
18
19 5.3.4. Uncertainty in Background Rate of Localized Pleural Thickening
20 In the derivation of the RfC, a background rate of 1% for localized pleural thickening was
21 used. As discussed in Section 5.2.3.3, there is uncertainty in estimating the value of this
22 parameter. However, in statistical modeling of the Marysville sub-cohort, potential uncertainty
23 in the background rate of localized pleural thickening has little impact on the estimated POD.
24 The best-fitting model (Michaelis-Menten with 10-year lagged exposure) was rerun, allowing the
25 background rate to be estimated as a parameter rather than fixed, with a resulting estimated
26 background rate of 3.12% (SE = 2.84%). Both the fixed and estimated values are in the range of
27 estimates from previous studies described above, and the difference in the POD when the
28 background rate is fixed at 1% versus when it is estimated is -15% (0.1177 compared with
29 0.1349 fibers/cc-year, and it does not affect the proposed RfC (after rounding to one significant
30 digit).
31
This document is a draft for review purposes only and does not constitute Agency policy.
5-45 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.3.5. Uncertainty in Model Functional Form and Lagged Exposure
2 A number of model forms were explored in the initial stages of analysis (see Appendix E)
3 before selecting the Michaelis-Menten model. In this application, the ratio of the BMCio to the
4 BMCLio (0.2642 ^ 0.1177 = 2.2) was reasonable given the size of the available data set,
5 indicating acceptable statistical precision in the BMC estimate. In addition, BMCs and BMCLs
6 estimated from other candidate models for the post-1972 exposure sub-cohort were in a similar
7 range to the selected model. Finally, the complementary analysis with the full cohort (utilizing a
8 time from first exposure of 30 years, which was selected to be consistent with time since first
9 exposure values within the sub-cohort) provided similar results to the sub-cohort analysis. A
10 second model-based uncertainty is the choice of lag for cumulative exposure. The RfC
11 derivation is based on the exposure lagged by 10 years, since this lag yielded the lowest AIC.
12 However, if other lags (with similar AICs) are used, the difference in POD may fluctuate to be
13 approximately 20% higher or approximately 55% lower. Thus, the choice of lag does not affect
14 the proposed RfC (after rounding to one significant digit).
15
16 5.3.6. Uncertainty Due to Effect of Smoking
17 Smoking is an important variable to consider when evaluating respiratory health
18 outcomes. Although data are mixed, a few studies suggest smoking may affect risk of
19 developing pleural thickening or timing of pleural thickening development among persons
20 exposed to asbestos. However, no studies were identified that assessed the relationship between
21 LPT specifically and any measure of smoking status. Discrete pleural plaques as defined in
22 earlier ILO classification systems have not been associated with smoking in asbestos-exposed
23 workers (Mastrangelo et al., 2009; Paris et al., 2009; Koskinen et al., 1998), but there is evidence
24 that small opacities (asbestosis) and diffuse pleural thickening may be associated with smoking
25 in asbestos-exposed individuals.34 As the current classification of LPT includes cases of diffuse
34Studies among populations exposed to general 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
thickening, including both pleural plaques and diffuse pleural thickening (McMillan etal. 1980X 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—or possible—associations when examining risk of pleural changes (Paris et
al.. 2008: Dement etal.. 2003: Zitting et al.. 1996: YanoetaL 1993: Lilisetal.. 1991: Baker etal.. 1985) or no
association with smoking (Soulat et al.. 1999: Neri et al.. 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
This document is a draft for review purposes only and does not constitute Agency policy.
5-46 DRAFT—DO NOT CITE OR QUOTE
-------
1 pleural thickening where the CPA is not involved, investigation of the potential for smoking to
2 modify the effect of asbestos exposure on the prevalence of LPT is warranted.
3 Each of the four candidate studies considered for RfC derivation considered smoking in
4 their analytic approach. In the Libby workers cohort, McDonald et al. (1986b) assessed pleural
5 thickening of the chest wall (both discrete and diffuse regardless of CPA involvement) and found
6 smoking status (current, former, or never smoker) was of borderline statistical significance
7 (p = 0.10) in a regression model, controlling for Libby Amphibole asbestos exposure and age.
8 This is consistent with the broader asbestos literature, addressing all pleural thickening or all
9 pleural abnormalities. Amandus et al. (1987b) evaluated radiographic abnormalities consistent
10 with the current LPT designation; the authors took a different analytic approach to assess
11 smoking effects, constructing separate models for the full cohort and restricting to current and
12 former smokers. The parameter estimates were not significant for the two models, although the
13 coefficients corresponding to Libby Amphibole asbestos exposure were slightly higher for the
14 full cohort model.
15 In the Marysville workers cohort, smoking was characterized using pack-years in the
16 original study (Lockey et al., 1984) and as ever/never smoking in the follow-up (Rohs et al.,
17 2008). Lockey et al. (1984) reported that the pack-years variable was significantly associated
18 with risk of all radiographic changes using discriminate analysis (any pleural thickening, small
19 opacities, and blunting of the CPA) but did not present results for effect of smoking controlling
20 for Libby Amphibole asbestos exposure. Rohs et al. (2008) did not find a difference in smoking
21 prevalence among those with and without any radiographic changes but also did not report
22 results controlling for Libby Amphibole asbestos exposure, or for LPT specifically.
23 Therefore, EPA explored the effect of smoking on the critical endpoint. In analyses for
24 RfC derivation, the variable representing smoking history (ever smoker vs. never smoker) was of
25 borderline significance in the best-fitting model (p = 0.08) and improved model fit (see
26 Appendix E). The limited sample size (only three cases were never smokers) and limited nature
27 of the smoking information precluded use of the smoking variable for RfC derivation. However,
28 the model including smoking was examined as a sensitivity analysis. In this analysis, BMCs and
29 BMCLs estimated separately for smokers and nonsmokers differed by approximately sixfold,
categories of ever/never or former/current/never, while others used pack-years) and differences in the specific
outcome studied.
This document is a draft for review purposes only and does not constitute Agency policy.
5-47 DRAFT—DO NOT CITE OR QUOTE
-------
1 suggesting that smokers may be at a higher risk for LPT from exposure to Libby Amphibole
2 asbestos than nonsmokers. Thus, an estimated BMCL for smokers would be lower than the POD
3 used for RfC derivation (0.04 fibers/cc-year for smokers versus 0.12 fibers/cc-year for the entire
4 sub-cohort). Conversely, a BMCL for nonsmokers would be slightly higher
5 (0.25 fibers/cc-year). These sensitivity analyses indicate a need for further research on the effect
6 of smoking in relation to LPT risk among asbestos-exposed populations.
7
8 5.3.7. Sensitivity Analysis: Derivation of a POD for Lifetime Exposure From the
9 Cumulative Exposure Metric
10 Exposure-response modeling for LPT in the Marysville sub-cohort used the cumulative
11 exposure (CE) metric (represented as CHEEC, described in Section 5.2.2.1) providing a POD in
12 fibers/cc-years. In order to derive an RfC in the units of continuous air concentration for a
13 lifetime (i.e., fibers/cc), the POD from the CE metric was weighted across a lifetime exposure.
14 Thus, the lifetime BMCLio is 1.96 x 10"3 (0.1177 fibers/cc-years - 60 years35). This procedure is
15 one way to account for the duration of exposure in the occupational study being less than
16 lifetime. There is some uncertainty as to whether and how to take account for less-than-lifetime
17 exposure in the occupational cohort. The cohort participants had a wide range of exposure
18 durations, all of which are less than lifetime36. As there are other reasonable alternatives to
19 derive a lifetime RfC, a sensitivity analysis was conducted to examine if RfC derivation was
20 greatly impacted by the method chosen to convert the POD in units of cumulative exposure, to
21 an air concentration for lifetime exposure.
22 Use of the CE metric adjusted based on ventilation rates and work schedule to a
23 continuous air concentration is consistent with EPA guidance (represented as CHEEC in this
24 assessment) (U.S. EPA, 2002, 1994b). Guidelines also recommend that if the human study is a
25 less-than-lifetime study, additional adjustment may be needed, depending on the nature of the
26 observed health effect for an RfC applicable to lifetime exposure (U.S. EPA, 1994b). Although
27 cumulative exposure is often associated with asbestosis (small opacities) and DPT, many other
28 studies have found pleural plaques are better predicted by other exposure metrics (e.g., average
35Because the best-fitting model had a 10-year lag, the lag is applied to the weighting across a lifetime as well. Sixty
years represent lifetime exposure of 70 years; 70 years - 10 years for the lag in exposure.
36This is especially true for the RfC derived from the sub-cohort hired after 1972, which had a more limited range of
employment duration (mean=18.7 years [SD=8.6]; range=0.3-29.0).
This document is a draft for review purposes only and does not constitute Agency policy.
5-48 DRAFT—DO NOT CITE OR QUOTE
-------
1 intensity, mean exposure, duration). The use of a measure of average exposure (averaged over
2 the period of exposure) is consistent with previous studies (asbestos in general) that report
3 associations of the prevalence of pleural plaques with mean or average asbestos exposure (Paris
4 et al.. 2008: Jakobsson et al.. 1995: Ehrlich et al.. 1992). The first alternative method was to
5 weight the POD across duration of exposure in the sub-cohort, rather than a full lifetime. The
6 second alternative is model the exposure-response relationship for LPT against average exposure
7 (a measure of the cumulative exposure for each worker averaged over the individual worker's
8 duration of work exposure).
9 The first sensitivity analysis is calculated by dividing the modeled POD for the
10 sub-cohort (0.1177 fibers/cc-years [30-year BMCLio]) by the average employment duration for
11 the sub-cohort of 18.7 years. Therefore, the POD expressed as the equivalent concentration for
12 the mean worker exposure duration for the sub-cohort is 6.3 x 10~3 (fibers/cc, continuous air
13 concentration) ([0.1177 fib ers/cc-y ears] + 18.7 years).
14 For the second analysis, the average exposure was calculated for each participant
15 (AvgExp = CHEEH for each worker + duration of exposure for each worker). The
16 exposure-response relationship was defined using the best-fitting model for the sub-cohort
17 (Michaelis-Menten). The average exposure metric also provided an adequate fit to the data for
18 the preferred sub-cohort (Hosmer-Lemeshow GOF37; P = 0.72) and provided a slightly better—
19 but similar—fit to the CE metric [AIC = 72.2 versus 74.0). The Michaelis-Menten model
20 provided a BMC of 1.8 x 10"2 fibers/cc, and a BMCLio of 8.5 x 10"3 fibers/cc for the average
21 work-duration exposure metric. This BMCLio is about 4-fold higher than the lifetime-BMCLio
22 above from the primary analysis (1.96 x 10"3 fibers/cc).
23 The three methods provide PODs that vary by a factor of up to 4 (2.0 x 10"3, 6.1 x 10"3, or
24 8.5 x 10"3 fibers/cc) when expressed as a continuous air concentration. The primary analysis
25 assumes duration contributes to risk and thus calculates a concentration across a lifetime that
26 would yield the POD CE. The second analysis is consistent with assuming duration contributes
27 to risk but estimating the concentration only for the mean duration in the observed database. The
37General model fit was evaluated with the Hosmer-Lemeshow (2000) test (a form of the Pearson %2 goodness-of-fit
[GOF] statistic). This is a goodness-of-fit test that compares observed and expected events. Observations are sorted
in increasing order of estimated probability of the event occurring and then divided into ~10 groups; the test statistic
is calculated as the Pearson %2 statistic of observed and expected frequencies in these groups.
This document is a draft for review purposes only and does not constitute Agency policy.
5-49 DRAFT—DO NOT CITE OR QUOTE
-------
1 third analysis assumes duration does not contribute to risk and models the average work duration
2 continuous exposure equivalent for each worker.
3 The difference comes in whether the critical study is considered of adequate duration to
4 inform health effects from a lifetime exposure, or if further adjustment is needed across time.
5 The primary analysis provides this adjustment to a full lifetime. This sensitivity analysis
6 indicates that the approach taken to average the POD based on the CE metric (CHEEC) across a
7 lifetime was a reasonable approach, as similar results are obtained using different approaches
8 (i.e., within 4-fold).
9
10 5.3.8. Sensitivity Analysis for Choice of Critical Effect and Selection of Benchmark
11 Response (BMR)
12 The critical effect selected for RfC derivation is localized pleural thickening. Alternative
13 endpoints were not modeled using the preferred sub-cohort due to small numbers—there were
14 five cases of bilateral LPT, only one case of diffuse pleural thickening, and no individuals with
15 interstitial changes. As a sensitivity analysis, these three alternative endpoints (along with all
16 LPT) were modeled among all Marysville workers not previously exposed to other forms of
17 asbestos, with X-rays performed in 2002-2005 (n = 250). These analyses were performed using
18 the Michaelis-Menten model with a background rate of 1% and unlagged CHEEC as the
19 exposure metric. BMRs of 1, 5, and 10% were investigated (see Table 5-5). Use of the 10%
20 BMR for these alternative endpoints allows for comparison with a POD based on the selected
21 critical effect of LPT. In this larger cohort, the POD for a 10% increase in LPT was
22 0.06 fibers/cc-years (in comparison with the POD derived from the sub-cohort and used in RfC
23 derivation of 0.118 fib ers/cc-y ears). Results for all pleural thickening (LPT and DPT) did not
24 differ from results for LPT. Bilateral localized pleural thickening was included as a rough
25 indication of increased severity within LPT, and as expected results in higher PODs at each
26 BMR than LPT. The resulting BMCLs for DPT and small opacities (1.17 and 2.89
27 fib ers/cc-y ears, respectively, 10% BMR) are higher than the POD for LPT (0.06 fib ers/cc-y ears).
28 Thus, use of an alternative endpoint at the same BMR would provide a higher POD, and
29 corresponding higher RfC.
30 However, a 10% BMR is not appropriate for more severe endpoints and BMCLs are
31 calculated at 1 and 5% BMRs as well. If DPT is used as a critical effect, PODs of 0.081 and
This document is a draft for review purposes only and does not constitute Agency policy.
5-50 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-5. Modeling of alternative endpoints in the Marysville worker
cohort members examined in 2002-2005
BMR
BMR = 1%
BMR = 5%
BMR = 10%
Parameter
AIC
Alpha (SE)
Plateau (SE)
BMC
BMCL
BMC
BMCL
BMC
BMCL
Bilateral localized pleural
thickening (n = 33) vs. no
abnormalities (n = 181)
164.6
0.2670 (0.5420)
0.4120 (0.0962)
0.0193
0.0097
0.1075
0.0552
0.2501
0.1337
Diffuse pleural thickening
(n = 10) vs. no abnormalities
(n = 181)
64.1
-2.8434 (1.6617)
0.6166 (0.6307)
0.2849
0.0814
1.5259
0.4728
3.3494
1.1715
Interstitial changes
(« = 7) vs. no
abnormalities (n = 181)
45.9
-4.0674 (0.5014)
1.0000 (--)
0.5899
0.2425
3.0739
1.3158
6.4894
2.8923
4
5
6
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
0.473 fibers/cc-years would be calculated for a 1% and 5% BMR, respectively. If small
opacities are used as a critical effect, the PODs are higher at both a 1% and a 5% BMR
(0.243 and 1.32, respectively). In summary, the use of more severe alternative endpoints (with
appropriate BMRs) results in PODs higher than that estimated using the critical effect of LPT
(0.06 fibers/cc-year, BMR 10%), and all are higher than the POD used in RfC derviation, with
the exception of DPT at a 1% BMR (0.0814 fibers/cc-year). BMCLs for these more severe
endpoints using a 1% BMR were within -twofold of the preferred POD (0.0814 and
0.2425 fibers/cc-year for diffuse pleural thickening and interstitial changes, respectively). There
is uncertainty associated with these estimates due to the inclusion of individuals hired before
1972, when no quantitative exposure measurements were available. Thus, a choice of alternative
critical effects (even with lower BMRs) would not result in an RfC appreciably lower than the
proposed RfC based on LPT and a 10% BMR.
5.4. CANCER EXPOSURE-RESPONSE ASSESSMENT
5.4.1. Overview of Methodological Approach
The objective of this human health assessment is to derive a cancer estimate for
inhalation exposure to Libby Amphibole asbestos. 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
This document is a draft for review purposes only and does not constitute Agency policy.
5-51 DRAFT—DO NOT CITE OR QUOTE
-------
1 for asbestos are given in fibers/cc of air as counted by PCM, since they are based on health
2 effects observed in occupational cohorts and this is the standard for measuring fiber exposures in
3 an occupational environment (OSHA, 1994; U.S. EPA, 1988a). Similarly, when examining the
4 available health effects data for Libby Amphibole asbestos, the best available exposure metric at
5 this time is fibers/cc counted by PCM (see Section 4.1.1.2). Therefore, for Libby Amphibole
6 asbestos, the IUR represents the lifetime risk of mortality from either mesothelioma or lung
7 cancer in the general U.S. population from chronic inhalation exposure to Libby Amphibole
8 asbestos at a concentration of 1 fiber/cc of air.
9 lURs are based on human data when appropriate epidemiologic studies are available.
10 The general approach to developing an IUR from human epidemiologic data is to quantitatively
11 evaluate the exposure-response relationship (slope) for that agent to derive a specific estimate of
12 its cancer potency in the studied population. For this current assessment, the first step was to
13 identify the most appropriate data set available, which in this case can be used to quantitatively
14 estimate the effects of Libby Amphibole asbestos exposure on cancer mortality. Once the
15 relevant data describing a well-defined group of individuals along with their exposures and
16 health outcomes were selected, an appropriate statistical model was selected that adequately fit
17 the data, and then individual-level exposures were modeled using a variety of possible exposure
18 metrics (see Section 5.4.2). The available epidemiologic data allowed for modeling of the
19 effects of estimated ambient occupational exposures to Libby Amphibole asbestos on the
20 observed cancer mortality risk in workers. Exposure-response modeling was conducted for each
21 cancer mortality endpoint individually, and in some cases, the statistical model and the specific
22 metric of exposure used for each cancer endpoint may have been different. For example, the
23 exposure metric that best describes the exposure-response relationship for mortality from
24 mesothelioma attributable to occupational exposure to Libby Amphibole asbestos was found to
25 be different from the exposure metric that best describes mortality from lung cancer (see
26 Section 5.4.3). Potential covariates that may also be important predictors of cancer mortality are
27 included in the statistical models. These models were then statistically evaluated to determine
28 which exposure metric representing estimated ambient occupational exposures provided the best
29 statistical fit to the epidemiologic data.
30 This cancer potency (slope) estimate derived from the epidemiologic data is then applied
31 to the general U.S. population to determine the exposure level that would be expected to result
This document is a draft for review purposes only and does not constitute Agency policy.
5-52 DRAFT—DO NOT CITE OR QUOTE
-------
1 in 1% extra cancer mortality risk over a lifetime of continuous exposure. For epidemiologic
2 studies, which may be based on larger numbers of individual observations, smaller response
3 levels that are closer to the background response levels are considered appropriate. Extra risk is
4 defined as equaling (Rx - RQ) ^ (1 - RO), where Rx is the lifetime cancer mortality risk in the
5 exposed population, and R0 is the lifetime cancer mortality risk in an unexposed population (i.e.,
6 the background risk). For example, if the expected lifetime risk of lung-cancer mortality in the
7 unexposed general U.S. population is 5%, then this human health assessment seeks to estimate
8 the level of exposure to Libby Amphibole asbestos that would be expected to result in a lifetime
9 risk of lung-cancer mortality of 5.95%; this lifetime risk of mortality is equivalent to a 1% extra
10 risk: (0.0595-0.05) H- (1-0.05) = 0.01. For mesothelioma mortality, an absolute risk of 1% was
11 considered, rather than extra risk, for two reasons. First, because mesothelioma is very rare in
12 the general population (Hillerdal 1983), and second, because mesothelioma is almost
13 exclusively caused by exposure to asbestos, including Libby Amphibole asbestos.
14 A life-table analysis (see Appendix G for details) was used to compute the 95% lower
15 bound on the lifetime exposure to Libby Amphibole asbestos that corresponds to a 1% extra risk
16 of cancer mortality in the general U.S. population using age-specific mortality statistics and the
17 exposure-response relationships for each cancer endpoint as estimated in the studied population.
18 This lower bound on the level of exposure serves as the POD for extrapolation to lower
19 exposures and for deriving the unit risk. Details of this analysis are presented in Section 5.4.5.
20 A cancer-specific unit risk was obtained by dividing the extra risk (1%) by the POD. The
21 cancer-specific unit risk estimates for mortality from either mesothelioma or lung cancer were
22 then statistically combined to derive the final IUR (see Section 5.4.5.3). Uncertainties in this
23 cancer assessment are described in detail in Section 5.4.6.
24
25 5.4.2. Choice of Study/Data—with Rationale and Justification
26 This human health assessment is specific to Libby Amphibole asbestos. This current
27 assessment does not seek to evaluate quantitative exposure-response data on cancer risks from
28 studies of asbestos that did not originate in Libby, MT.
29 The available sources of data included the cohort of workers employed at the vermiculite
30 mining and milling operation in and around Libby, MT. This cohort has been the subject of
31 several epidemiologic analyses (Larson et al., 201 Ob: Moolgavkar et al., 2010; Sullivan, 2007;
This document is a draft for review purposes only and does not constitute Agency policy.
5-53 DRAFT—DO NOT CITE OR QUOTE
-------
1 Amandus and Wheeler, 1987; McDonald et al., 1986a) (and described in detail in Section 4.1).
2 There have also been published reports on cases of mesothelioma in the Libby, MT area
3 (Whitehouse et al., 2008) and mortality data published by the Agency for Toxic Substances and
4 Disease Registry (ATSDR, 2000). However, published mortality data on Libby, MT residents
5 (Whitehouse et al., 2008; ATSDR, 2000) could not be used in exposure-response modeling due
6 to lack of quantitative exposure data.
7 The other available cohort of workers exposed to Libby Amphibole asbestos was from an
8 Ohio vermiculite processing plant (see Section 4.1.3) (Rohs et al., 2008; Lockey et al., 1984).
9 Pleural changes were evaluated; however, no data were available pertaining to cancer incidence
10 or mortality in the Ohio cohort. No other worker cohorts exposed to Libby Amphibole asbestos
11 with cancer incidence or mortality data were available.
12 The most appropriate available data set with quantitative exposure data for deriving
13 quantitative cancer mortality risk estimates based on Libby Amphibole asbestos exposure in
14 humans is the cohort of workers employed at the vermiculite mining and milling operation in and
15 around Libby, MT (hereafter referred to as the Libby worker cohort). These data are considered
16 the most appropriate to inform this human health assessment for several reasons: (1) these
17 workers were directly exposed to Libby Amphibole asbestos, (2) detailed work histories and
18 job-specific exposure estimates are available to reconstruct estimates of each individual's
19 occupational exposure experience, (3) the cohort is sufficiently large and has been followed for a
20 sufficiently long period of time for cancer to develop (i.e., cancer incidence) and result in
21 mortality, and (4) the broad range of exposure experiences in this cohort provided an
22 information-rich data set, which allowed evaluation of several different metrics of exposure.
23 Uncertainties in these data are discussed in Section 5.4.6.
24
25 5.4.2.1. Description of the Libby Worker Cohort
26 The Libby worker cohort has been extensively studied. McDonald et al. published three
27 studies on a subset of the cohort (2004. 2002: 1986a). Scientists from NIOSH conducted two
28 epidemiologic investigations, resulting in several published reports on different subsets of the
29 cohort (Sullivan. 2007: Amandus et al.. 1988: Amandus and Wheeler. 1987). Larson etal.
30 (201 Ob) analyzed an ATSDR reconstruction of the Libby worker cohort from company records
31 with exposure estimates obtained from NIOSH with mortality follow-up through 2006.
This document is a draft for review purposes only and does not constitute Agency policy.
5-54 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Moolgavkar et al. (2010) reanalyzed the Sullivan (2007) data with mortality follow-up through
2001 using a different statistical approach.
According to Sullivan (2007), nearly all of these study subjects were workers at the
Libby, MT vermiculite mine, mill, and processing plant. Although the mine was several miles
from Libby, MT, some of the study subjects worked in the town (see Section 4.1.1.1). 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' 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 5-6 shows the demographic and
exposure characteristics of this cohort.
Table 5-6. Demographic 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
Mean year of birth
Mean year of hire
Mean age at hire (years)
Mean person-years of follow-up (no lag)
Total person-years of follow-up (no lag)
Mean employment duration (years)
Mean cumulative exposure (fiber/cc-year)
Median cumulative exposure (fiber/cc-year)
Range of cumulative exposures (no lag) (fiber/cc-year)a
All workers
1,871
1,009
18
111
1929
1959
30.2
35.9
67,101
3.7
96.0
9.8
0-1722
"According to the work histories and JEM, there were 26 workers who had zero exposure. These
individuals (7 men and 19 women) all worked at the office downtown.
This document is a draft for review purposes only and does not constitute Agency policy.
5-55 DRAFT—DO NOT CITE OR QUOTE
-------
1 For the purposes of this current assessment, vital status follow-up was completed by
2 NIOSH through 2006 using the National Death Index [NDI-Plus; Bilgrad (1999)1. Workers
3 known to be alive on or after January 1, 1979 (the date NDI began tracking deaths nationwide),
4 but not found in the NDI search, were assumed to have been alive on December 31, 2006
5 (Sullivan. 2007). Nearly 54% of workers in the cohort (n = 1,009) had died by
6 December 31, 2006. NIOSH researchers obtained death certificates from across the United
7 States (while exposure occurred in and around Libby, deaths could have occurred elsewhere) for
8 deaths prior to 1979 and coded to the International Classification of Diseases (ICD) revision in
9 effect at the time of death by a single National Center for Health Statistics-trained nosologist.
10 After 1979, ICD codes were obtained from the NDI-Plus. For workers known to be deceased,
11 the underlying cause of death was determined from death certificates and coded to the ICD codes
12 using the rubrics of the ICD revision in effect at the time of death [ICD-5 (WHO. 1938). ICD-6
13 (WHO. 1948). ICD-7 (WHO. 1957). ICD-8 (WHO. 1967). ICD-9 (WHO. 1977). or ICD-10
14 (WHO. 1992)1
15 Basic demographic information on the occupational cohort members was largely
16 complete. However, when data were missing, they were imputed by NIOSH based on the
17 following assumptions regarding gender, race, and date of birth. Seven workers with unknown
18 gender were assumed to be male because 96% of the workforce was male, and NIOSH review of
19 names did not challenge that assumption (Sullivan, 2007). Workers of unknown race (n = 935)
20 were assumed to be white because workers at this facility were known to be primarily white, and
21 U.S. Census Bureau data indicate that 90-95% of the local population identify themselves as
22 white (Sullivan, 2007). For four workers with unknown birth dates, date of birth was estimated
23 by subtracting the mean age at hire for the cohort from the worker's hire date. The potential
24 impact of this imputation procedure on the analytic results is discussed in Section 5.4.6.
25
26 5.4.2.2. Description of Cancer Endpoints
27 This human health assessment of Libby Amphibole asbestos focuses on two cancer
28 endpoints: mesothelioma and lung cancer. The endpoint for both mesothelioma and lung cancer
29 was mortality, not incidence. Incidence data are not available for the Libby worker cohort.
30 However, there is evidence that other cancer endpoints may also be associated with exposure to
31 asbestos. The International Agency for Research on Cancer (IARC) concluded that there was
This document is a draft for review purposes only and does not constitute Agency policy.
5-56 DRAFT—DO NOT CITE OR QUOTE
-------
1 sufficient evidence in humans that other types of asbestos (chrysotile, crocidolite, amosite,
2 tremolite, actinolite, and anthophyllite) were causally associated with mesothelioma and lung
3 cancer, as well as cancer of the larynx and the ovary (Straif et al., 2009). Among the entire
4 Libby worker cohort, only two deaths were found to be due to laryngeal cancer, and there were
5 no deaths from ovarian cancer among the 84 female workers. The EPA did not evaluate these
6 other outcomes as part of this current assessment. The limited number of female workers in this
7 cohort is discussed later as a source of uncertainty in the derived estimates (see Section 5.4.6).
8 Mesothelioma did not have a distinct ICD code prior to introduction of the 10* revision
9 (ICD-10), which was not implemented until 1999. Therefore, for deaths in the Libby worker
10 cohort occurring from 1979 to 1998, death certificates were obtained if the NDI identified the
11 death as being from one of the possible mesothelioma codes identified by Marsh et al. (2001), or
12 from respiratory cancer, nonmalignant respiratory disease, digestive cancer, or unspecified
13 cancer. Death certificates (1940-1998) were reviewed by the NIOSH principal investigator
14 (Sullivan, 2007) to identify any mention of mesothelioma on the death certificate, as is the
15 standard procedure for assessing mesothelioma mortality and as has been used in other analyses
16 of Libby worker cohort mesothelioma mortality (Larson et al., 201 Ob: McDonald et al., 2004).
17 In total, 18 mesothelioma deaths occurring from 1979 to 2006 were identified by NIOSH using
18 these methods, which serve as the basis for this current assessment; 19 mesothelioma deaths
19 were identified by Larson et al. (201 Ob) for the same cohort from death certificates for all causes
20 of death rather than the more targeted set of causes identified by Marsh et al. (2001) or Sullivan
21 (2007).
22 Whitehouse et al. (2008) identified four mesothelioma cases among workers that were
23 not included in Sullivan (2007) with mortality follow-up through 2001; no other information was
24 provided. Most likely, three mesothelioma cases from these four were accounted for during the
25 update of the NIOSH cohort to 2006, which serves as the basis for this current assessment.
26 Whitehouse et al. (2008) also provided detailed information on 11 residential cases, but this
27 information could not be used in exposure-response analyses for this current assessment because
28 there is no quantitative exposure information for these cases and no information defining or
29 enumerating the population from which these cases arose.
30 Mortality records (and death certificates) may not always reflect the true cause of death
31 for various reasons (e.g., misdiagnosis, improper recording on the death certificate, or miscoding
This document is a draft for review purposes only and does not constitute Agency policy.
5-57 DRAFT—DO NOT CITE OR QUOTE
-------
1 of the cause of death). For mesothelioma, the undercounting of cases (underascertainment) is a
2 particular concern given the limitations of the ICD classification systems used prior to 1999
3 [detection rates varied from 12% from ICD-9 codes alone to 83% from manual inspection of
4 death certificates (Davis etal., 1992)1: recent studies demonstrated that ICD-10 coding has
5 detection rates similar to the latter rate above (Camidge et al., 2006; Pinheiro et al., 2004). The
6 appropriate procedure for pre-ICD-10 codes is not to use ICD codes alone but to manually
7 inspect death certificates, as was done by Sullivan (2007). There is also evidence that the
8 detection rate of peritoneal mesothelioma is much lower than pleural mesothelioma (Selikoff and
9 Seidman, 1992). This current assessment has accounted for the impact of this
10 underascertainment on the final IUR (see Section 5.4.5.1.1).
11 Lung-cancer mortality was based on the underlying cause of death identified by the ICD
12 code on death certificates according to the ICD version in use at the time of death. Based on
13 these different ICD codes, lung-cancer mortality included malignant neoplasms of the trachea,
14 bronchus, and lung, and was identified by the following codes: ICD-5 code '047' (excluding
15 '47c, Cancer of unspecified respiratory organs'), ICD-6 codes '162' or '163,' ICD-7 codes '162'
16 or '163' (excluding '162.2, Cancer of the pleura'), ICD-8 and ICD-9 code '162', and ICD-10
17 codes 'C33' or 'C34'. In all, there were 111 deaths, with an underlying cause attributed to lung
18 cancer. All deaths after 1960 were coded as bronchus or lung because the ICD versions in use as
19 that time distinguished malignant neoplasms of the trachea as distinct from bronchus and lung.
20 Other investigators of this cohort have used different definitions of lung cancer or used different
21 follow-up periods, as described in Section 4.1.1.2.2 (Description of Cohorts).
22
23 5.4.2.3. Description ofLibby Amphibole Asbestos Exposures
24 The mining, milling, and processing operations at the mine and in and around Libby,
25 conditions of exposure, and j ob-specific estimates of exposure intensity have been thoroughly
26 described in Section 4.1 (Sullivan. 2007: Amandus et al.. 1987a: McDonald et al.. 1986a).
27 Briefly, miners extracted vermiculite ore from an open-pit mine that operated on Zonolite
28 Mountain outside the town ofLibby, MT. The ore was processed in a dry mill (1935-1974)
29 and/or two wet mills (1950-1974 and 1974-1990). The resulting concentrate was transported by
30 railroad to processing plants around the United States where the vermiculite was expanded for
31 use in loose-fill attic insulation, gardening, and other products (see Section 2.1).
This document is a draft for review purposes only and does not constitute Agency policy.
5-58 DRAFT—DO NOT CITE OR QUOTE
-------
Exposure
Intensity
(fibers cc)
Year
•Selected Jobs
3
4
5
Figure 5-3. Plot of the 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.
8 EPA adopted the JEM developed and used by Sullivan (2007) (see Figure 5-3), which
9 was, in turn, based on that used in the earlier NIOSH study for jobs through 1982 (Amandus et
10 al., 1987a: Amandus and Wheeler, 1987). As discussed in more detail in Section 4.1, Amandus
11 et al. (1987a) defined 25 location operations to which they assigned exposure intensity based on
12 available information (see Table 5-7). A job category may have involved more than one location
13 operation, and the 8-hour time-weighted average exposure (8-hour TWA) for each job category
14 in the JEM was calculated from the exposure intensity and time spent at each location operation
15 (Amandus et al.. 1987a).
16 For the later data in Table 5-7 from 1967 through 1982, over 4,000 air samples analyzed
17 for fibers by PCM analysis were available to inform the exposure intensity for the 25 location
18 operations (see Table 5-7). Therefore, the JEM for 1968-1982 is based on direct analytic
This document is a draft for review purposes only and does not constitute Agency policy.
5-59 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-7. Exposure intensity (fibers/cc) for each location operation from the beginning of operations through
1982 [Amandus et al. (1987a); 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
Bagging plant
Tails belt
Year
<50
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
12.9
7.3
50-59
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
12.9
7.3
60-63
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
12.9
7.3
64-67
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
12.9
7.3
68-70
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
12.9
7.3
71
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
-
12.9
7.3
72-74
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
-
4.3
7.3
75-76
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
-
1.2
0.7
77-79
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
-
1.2
0.7
80-82
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
-
1.2
0.7
I
I
a.
i*
H S
W|
V
o
c
o
-------
Table 5-7. Exposure intensity (fibers/cc) for each location operation from the beginning of operations through
1982 (continued)
Location operation
Screen plant
Drilling
Ore loading
River dock
Bagging plant
High
Low
High
Low
High
Low
High
Low
Year
<50
-
23
6.7
82.5
24
116.9
38
12.9
4.6
50-59
-
23
6.7
27.7
15
42.5
19
12.9
4.6
60-63
-
23
6.7
10.7
9
17
6.4
12.9
4.6
64-67
-
23
6.7
10.7
9
17
6.4
12.9
4.6
68-70
-
9.2
6.7
3.2
3.2
17
5.1
12.9
4.6
71
-
9.2
9.2
3.2
3.2
5.1
5.1
12.9
4.6
72-74
-
9.2
9.2
3.2
3.2
5.1
5.1
4.3
4.3
75-76
0.5
0.6
0.6
0.2
0.2
0.5
0.5
1.2
1.2
77-79
0.5
0.6
0.6
0.2
0.2
0.5
0.5
1.2
1.2
80-82
0.5
0.6
0.6
0.2
0.2
0.5
0.5
1.2
1.2
I
§
I
a,
H
H S
W|
V
o
c
o
-------
1 measurements in air for each location operation (Amandus et al., 1987a). With the exception of
2 the dry mill, no air samples were available for other location operations at the mine and
3 processing facilities prior to 1967. In order to estimate exposures that occurred before that time,
4 the NIOSH researchers interviewed plant employees and based estimates of exposure intensities
5 on known changes in operations over the years and professional judgments regarding the relative
6 intensity of exposure; exposure intensity for 23 of the pre-1967 location operations was
7 extrapolated from post-1967 measurements based on reasoned assumptions for each location
8 operation (Amandus et al., 1987a).
9 However, the amount and quality of measurement data in the facility in earlier years were
10 much more limited (Amandus et al., 1987a). A total of 40 dust samples were taken, exclusively
11 in the dry mill, over the years 1950-1964. Using these measurements, much higher exposures
12 were inferred to occur prior to 1964 than those measured in later years. Although air sampling
13 for fibers by PCM was available beginning in 1967, average fiber concentrations (dry mill)
14 differed rather widely between limited data sets from different investigators up through the early
15 1970s: 1967-1968, NIOSH data, 65 fibers/cc (n = 14); 1970, company data, 11 fibers/cc
16 (n = 15); 1971, Mine Safety and Health Administration (MSHA) data, 31 fibers/cc (n = 52);
17 1972, MSHA and company data, 15 fibers/cc (n = 45). Thus, estimated exposure levels continue
18 to be uncertain during the period when fiber concentration measurements by PCM became
19 available in 1967.
20 Air samples collected by the State of Montana were available for the dry mill
21 from 1956-1969, but these were analyzed for total dust, not asbestos fibers. Total dust samples
22 (collected by a midget impinger) were examined by light microscopy, but no distinction was
23 made between mineral dusts, debris, and asbestos fibers. All objects were counted and reported
24 in the units of million particles per cubic foot of air (mppcf). Amandus et al. (1987a) developed
25 a relationship between total dust and asbestos fiber counts based on the comparison of
26 contemporaneous air sampling in the dry mill (see Section 4.1.1.2). The conversion ratio of
27 4.0 fibers/cc per mppcf was used to estimate exposure intensity for two location operations in the
28 dry mill for the years prior to 1967.
29 The exposure intensity (fibers/cc) for each of the location operations (see Table 5-7) was
30 used to calculate an estimate of daily occupational exposure for each job category in the JEM
31 (see Figure 5-3). For each job, the time spent at each location operation and the exposure
This document is a draft for review purposes only and does not constitute Agency policy.
5-62 DRAFT—DO NOT CITE OR QUOTE
-------
1 intensity for each location operation were averaged to derive an estimate of the 8-hour TWA.
2 The resulting JEM available for this current assessment and previous epidemiologic studies of
3 the Libby worker cohort is based on the air concentration of fibers as enumerated by PCM,
4 which measures fibers longer than 5 um with an aspect ratio >3:1 [i.e., the fiber size regulated
5 under the OSHA standard (OSHA, 2006)]. Additionally, only fibers that are wide enough to be
6 viewed on PCM can be detected with this method. Amandus et al. (1987a) considered fibers
7 >0.44 um in diameter to be visible by PCM in the historical filter analysis. More recent
8 techniques have refined the PCM method, and fibers greater than 0.25 um in diameter are now
9 considered PCM fibers (IPCS, 1986).
10 There was one important limitation of the NIOSH work history data. In the earlier study
11 (Amandus and Wheeler, 1987), workers with "common laborer" job assignments and some
12 workers with unknown job assignments hired between 1935 and 1959 were assigned the
13 relatively low exposure levels estimated for the mill yard (Sullivan, 2007). Of the 991 workers
14 hired before 1960, 811 workers had at least one job with an unknown job assignment, with
15 706 having all department and job assignments prior to 1960 listed as unknown. In the more
16 recent study by Sullivan (2007), these workers were assigned the same relatively high time
17 weighted average estimated exposure intensity (absolute majority of these workers were assigned
18 66.5 fibers/cc) for all jobs during that time period. The lack of information on specific job
19 assignments for such a large portion of these early workers when exposures were higher resulted
20 in the misclassification of the exposure and effectively yielded exposure metrics that were
21 differentiated only by the duration of each worker's employment. Because of the lack of more
22 specific measured fiber exposure data during this early period, the EPA experienced difficulties
23 in identifying an adequate exposure-response model fit for the complete cohort. These
24 difficulties are described in detail in Section 5.4.3.5.
25 As a result, the IUR analyses were based on the subset of workers hired after 1959 (i.e.,
26 on or after January 1, 1960) and consisted of 880 workers. Of these 880 workers hired after
27 1959, 28 workers had at least one job with an unknown job assignment with 9 having all job and
28 department assignments between 1960-63 listed as unknown. These workers were assigned a
29 time-weighted average estimated exposure intensity of 66.3 fibers/cc. In addition, reabstracting
30 work histories for the more recent study (Sullivan, 2007) identified several job assignments not
31 mentioned in the earlier publications. Sullivan (2007) estimated exposure for the additional job
This document is a draft for review purposes only and does not constitute Agency policy.
5-63 DRAFT—DO NOT CITE OR QUOTE
-------
1 and calendar time period-specific combinations based on professional experience and review of
2 exposure records from earlier studies of the Libby worker cohort (Amandus et al., 1987a:
3 Amandus and Wheeler, 1987; McDonald et al., 1986a). Uncertainties in the exposure
4 assessment for this sub-cohort are described in Section 5.4.6.1.2.4. While the Sullivan (2007)
5 study was limited to the white male workers, EPA's analysis includes all workers regardless of
6 race or gender. Table 5-8 shows the demographic and exposure characteristics of the sub-cohort
7 hired after 1959. Figure 5-3 shows a three-dimensional representation of the job-exposure
8 matrix used by Sullivan (2007) and in this current assessment. Not all jobs were included; thus,
9 the figure is not comprehensive but rather illustrative. The three axes show the intensity of fiber
10 exposure as an 8-hour TWA (fibers/cc, vertical axis) for selected job categories over time
11 (horizontal axes). For several jobs, the estimated 8-hour TWA was greater than 100 fibers/cc for
12 the decades prior to 1963. Figure 5-3 shows the variability in exposures across jobs and over
13 time. From 1967-1982, all exposure measurements that inform the JEM are based on
14 location-specific air samples analyzed for fibers by PCM. As stated above, pre-1968 exposures
15 in the dry mill were based on the measurement of dust levels from 1956-1967 that were
16 converted to PCM by Amandus et al. (1987a) and extrapolated backwards in time. Pre-1968
17 exposures for all other locations within the JEM were extrapolated from post-1967 fiber levels
18 based on reasoned assumptions (Amandus et al., 1987a).
19 Amandus et al. (1987a) recognized the uncertainty in the pre-1968 exposures assigned to
20 the cohort. Although there is some uncertainty in the dust-to-fiber conversion, this conversion
21 (4.0 fibers/cc per mppcf) was based on contemporaneously collected dust and fiber data collected
22 in the dry mill and only applied to the dry mill environment. Amandus et al. (1987a) considered
23 a range of possible conversion factors (1.2-11.5 fibers/cc per mppcf). Greater uncertainty may
24 lie with the reasoned assumptions used to extrapolate exposures to the early decades for all
25 which Amandus et al. estimated a range of possible exposure intensities: drilling, ore loading, the
26 river dock, and the bagging plant, where intensity of exposure may vary as much as threefold
27 between the low and high estimates (see Table 5-8). Finally, some workers were employed after
28 1982 through 1993 when demolition of the facilities was completed (Larson et al., 201 Ob).
29 These exposures were not evaluated by Sullivan (2007) and were not included in the NIOSH
30 JEM. However, only 148 sub-cohort workers were still employed on May 31, 1982, according
31 to the NIOSH records. Because exposure concentrations in 1982 (see Table 5-7) were generally
This document is a draft for review purposes only and does not constitute Agency policy.
5-64 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-8. Demographic and exposure characteristics of the subset of the
Libby worker sub-cohort hired after 1959
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Characteristic
Number of workers
Number of deaths from all causes
Number of deaths from mesothelioma
Number of deaths from lung cancer
Mean year of birth
Mean year of hire
Mean age at hire (years)
Mean person-years of follow-up (no lag)
Total person-years of follow-up (no lag)
Mean employment duration (years)
Mean cumulative exposure (fiber/cc-year)
Median cumulative exposure (fiber /cc-year)
Range of cumulative exposures (no lag) (fiber/cc-year)a
Sub-cohort hired after 1959
880
230
7
32
1942
1971
28.6
32.2
28,354
3.3
19.2
3.4
0-462
""According to the work histories and JEM, there were 21 sub-cohort workers who had zero cumulative
exposure. These 21 individuals all worked at the office downtown.
location operations considered. For example, there were four location operations for 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. Uncertainties in all aspects of JEM are
described in Section 5.4.6.1.2.
5.4.2.4. Description of Libby Worker Cohort Work Histories
NIOSH staff abstracted demographic data and work history data from company personnel
and payroll records, including W-4 federal tax forms. An individual's work history was
determined from job change slips, which recorded 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 with known termination dates before September
1953. The majority of workers in this cohort as a whole and among those hired on or after
January 1, 1960, worked at multiple jobs; many of the workers switched jobs repeatedly or had
This document is a draft for review purposes only and does not constitute Agency policy.
5-65 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
o
J
4
5
6
7
the estimated exposure for a job change from one year to the next. Of the 880 workers hired in
1960 or afterwards in the sub-cohort, the mean number of times a worker's exposure level
changed according to the JEM was 5, the median was 2, and the maximum number of changes
was 32 (see Figure 5-4; see also Figure 5-3 for a depiction of job-exposure intensities for
different jobs overtime).
Number of Different Jobs Held Among Workers Hired After 1959
30(tf
9
10
11
12
13
14
15
16
17
18
a 11 13 15 17 IE 21 2} 15 17 Z9 31
Number of Jobs Held by Worker
Figure 5-4. Histogram showing the number of workers who experienced
each incremental number of different jobs among the 880 workers hired
after 1959.
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):
This document is a draft for review purposes only and does not constitute Agency policy.
5-66 DRAFT—DO NOT CITE OR QUOTE
-------
1 After identification of an occupational hazard one of the goals of occupational
2 epidemiology is to quantify the risks by determining the dose-response relations
3 for the toxic agent. In many circumstances little is known about the dose received
4 by target tissues; the data available usually pertain only to exposure to various
5 concentrations of the toxic material in the workplace. The calculation of dose
6 requires additional physiological and chemical information relating to absorption,
7 distribution, biochemical reactions, retention, and clearance.
8
9
10 In asbestos epidemiology the usual measure of exposure is the product of the
11 concentration of asbestos dust in the air (fibers or particles per ml) and the
12 duration of exposure to each concentration summed over the entire duration of
13 exposure (years); this measure is the cumulative exposure....
14
15
16 Cumulative exposure has been the traditional method of measuring exposure in
17 epidemiologic analyses of many different occupational and environmental exposures and was the
18 exposure metric applied to the risk of lung-cancer mortality in the Integrated Risk Information
19 System (IRIS) assessment for general asbestos (U.S. EPA, 1988a). Two alternative approaches
20 to developing exposure metrics to describe the effects of concentrations of asbestos dust in the
21 air on the risks of mortality have also been proposed. The first alternative was proposed by
22 Jahr (1974), who studied silica-induced pneumoconiosis and suggested that exposures to
23 occupational dusts could be weighted by the time since exposure. This yields an exposure metric
24 that gives greater weight to earlier exposures. Berry et al. (1979) subsequently suggested the
25 application of exposure metrics that allowed for the clearance of dust or fibers by using a decay
26 term on exposures. For the evaluation of mortality risk from mesothelioma, U.S. EPA (1988a)
27 used a different exposure metric than was used for lung-cancer mortality, which factored in the
28 time since first exposure. As observed in U.S. EPA (1988a), it is important to note that different
29 characterizations of estimated ambient exposures may be reasonably expected to be associated
30 with different endpoints.
31 Most studies of asbestos-related mortality have evaluated either cumulative exposure,
32 exposure concentration, or the duration of employment as exposure metrics. Many studies have
33 been limited in the availability of detailed exposure data—especially at the individual level. In
34 the Libby worker cohort data developed by NIOSH and used in this current assessment, detailed
35 work histories, together with job-specific exposure estimates, allowed for the reconstruction of
36 each individual's estimated occupational exposure over time to define multiple exposure metrics.
This document is a draft for review purposes only and does not constitute Agency policy.
5-67 DRAFT—DO NOT CITE OR QUOTE
-------
1 From this information-rich, individual-level data set from NIOSH, EPA constructed a
2 suite of the different metrics of occupational exposure, which had been proposed in the asbestos
3 literature or used in the IRIS asbestos assessment (U.S. EPA, 1988a). This suite of models was
4 defined a priori to encompass a reasonable set of proposed exposure metrics to allow sufficient
5 flexibility in model fit to these data. These exposure metrics were evaluated in
6 analytic-regression models to test which exposure metrics were the best empirical predictors of
7 observed cancer mortality, and the better fitting models were advanced for consideration as the
8 basis of the exposure-response relationship for the IUR. The types of exposure metrics evaluated
9 were intended to allow for variations of the classic metric of cumulative exposure, allowing for
10 more or less weight to be placed on earlier or later exposures. These simulated exposure metrics
11 were derived mathematically to approximate underlying processes that are not well understood
12 (see Section 5.4.6). Thus, the fit of exposure metrics is evaluated on the basis of maximizing the
13 likelihood for the Libby worker cohort, and the estimated parameters do not necessarily have
14 biological interpretations.
15 The first exposure metric—cumulative exposure (CE)—is a simple addition of each day
16 of exposure across time (see Eq. 5-1). CE has been widely used in modeling risk of cancer in
17 occupational epidemiology and has been used for modeling lung cancer (Larson etal., 201 Ob:
18 Moolgavkar et al., 2010; Sullivan, 2007; McDonald et al., 2004) and mesothelioma (McDonald
19 et al., 2004) in the Libby worker cohort. When using this exposure metric in the risk model, all
20 exposures have equal weight regardless of when they occurred and lead to the same estimated
21 cancer risk whether exposure happened early or later in life.
22 EPA calculated each individual's occupational CE to Libby Amphibole asbestos over
23 time from their date of hire until the date they ceased to be employed in the Libby operations or
24 until the date NIOSH collected the work history data, if still employed in May 1982. Workers
25 were assumed to remain at their final occupational CE level until death or the end of the
26 follow-up period on December 31, 2006. Each worker's CE at any time point (daily increment)
27 since their date of hire was computed as the sum of their exposure intensity (fibers/cc) on each
28 specific occupational day (xt) from day 1 through day k. Mathematically, this was defined as
29
30
This document is a draft for review purposes only and does not constitute Agency policy.
5-68 DRAFT—DO NOT CITE OR QUOTE
-------
1
2 v^
3 CE at time 4= zLxti (Eq. 5-1)
7=1
4 Where
5
6 Xf. = the estimated job-specific exposure intensity for the day tj, and
7 tk = the day on which the exposure is estimated.
8
9
10 A second exposure metric—residence time-weighted (RTW) exposure—gives additional
11 weight to early exposures. By doing so, the RTW exposure metric allows the possibility that
12 early exposures are more influential on cancer mortality predictions in the model. Unlike many
13 chemicals that are rapidly metabolized in the body and excreted, asbestos fibers are durable, and
14 some may remain in the body for years. Fibers that remain in the lung may continue to damage
15 lung cells and tissue until they are removed or cleared (see Section 3.2). Similarly, fibers that
16 translocate to the pleura may damage cells as long as they remain in this tissue. Therefore, a
17 fiber exposure may not only damage tissue during the exposure, but fibers may remain in these
18 tissues, with cellular and tissue damage accumulating over time.
19 The RTW exposure metric in this current assessment is sometimes called the cumulative
20 burden, or the area under the curve. A type of RTW metric was proposed for modeling of
21 mesothelioma mortality by Newhouse and Berry (1976) based on a general understanding of the
22 relationship between tumor incidence rate and time to cancer (Cook et al., 1969) as well as
23 animal models of mesothelioma (Berry and Wagner, 1969). Similar types of RTW metrics were
24 applied to the insulators asbestos cohort by Peto et al. (1982), discussed by Finkelstein (1985),
25 and applied in the derivation of the IUR in the IRIS assessment for asbestos (U.S. EPA, 1988a).
26 McDonald et al. (2004) and Moolgavkar et al. (2010) used RTW-type metrics for modeling
27 mesothelioma in the Libby worker cohort, and McDonald et al. (2004) applied an RTW metric
28 for modeling lung-cancer mortality in the Libby worker cohort.
29 In calculating RTW, each day's exposure is multiplied by the time since the exposure
30 occurred (see Eq. 5-2). RTW CE was calculated as a cumulative function of each time-interval's
31 CE such that earlier exposures contribute greater weight.
32
This document is a draft for review purposes only and does not constitute Agency policy.
5-69 DRAFT—DO NOT CITE OR QUOTE
-------
k j
X"1 X"1
2 RTW CE at time & = Zj Zj *',• (Eq. 5-2)
j=i 1=1
3
4 Where
5
6 x/. = the estimated job-specific exposure intensity for the day tt, and
7 tk = the day on which the exposure is estimated.
8
9
10 The CE and RTW exposure metrics result in increasing or sustained metrics of exposure
11 across time. However, it is known that some cellular and genetic damage may be repaired over
12 time, which could decrease cancer risk from exposure over time. Additionally, asbestos fibers
13 are cleared (removed) from the lung through natural processes and translocated to other tissues
14 (see Section 3.2.1.1). Therefore, when considering lung cancer, it is possible that removal of
15 asbestos fibers from the lung would reduce lung cancer risk over time. Although less is known
16 about removal of asbestos from the pleura, there may be clearance mechanisms operative in that
17 tissue as well (see Section 3.2.1.2). As noted earlier, Berry et al. (1979) proposed the use of
18 exposure metrics based on occupational exposures, which addressed the issue of clearance
19 through a mathematical decay term that modified measured ambient exposures. For
20 mesothelioma, modeling a decay term on exposure has been proposed by Berry (1999). Based
21 on this proposal, several recent papers applied a decay term to modeling mesothelioma mortality
22 (Berry et al.. 2009: Reid et al.. 2009: Barone-Adesi et al.. 2008: Gasparrini et al.. 2008: Clements
23 et al., 2007: Hodgson et al., 2005: Berry et al., 2004). Similarly, recent publications indicate that
24 the relative risk of lung cancer due to asbestos exposure declines 15-20 years after the cessation
25 of exposure to asbestos (Magnani et al., 2008: Hauptmann et al., 2002).
26 Mathematically allowing for the magnitude of earlier exposures to diminish with
27 advancing time was considered to be a method of giving less weight in the analyses to earlier
28 exposures compared to the previous two exposure metrics. Therefore, two additional exposure
29 metrics were considered, where a decay rate was applied to the CE and RTW exposure metrics
30 (see Eq. 5-3 and 5-4).
This document is a draft for review purposes only and does not constitute Agency policy.
5-70 DRAFT—DO NOT CITE OR QUOTE
-------
1 For each exposure metric, the application of a half-life was calculated by depreciating
2 each time-interval's (tj-i'Jj) exposure according to a model of exponential decay with various
3 half-lives (Ti/2) of 5, 10, 15, and 20 years. Note that the particular kinetics of Libby Amphibole
4 asbestos fibers are not fully understood, and the relevance of these particular half-lives was
5 determined from the statistical fit of these exposure metrics to the risk of cancer mortality, rather
6 than the biological half-life of the fibers. For a very large half-life, decay is very slow, and these
7 metrics would be very similar to the CE and RTW exposure metrics.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
1 Xfj *exfl
7=1
T
(Eq. 5-3)
Where
the estimated job-specific exposure intensity for the day tj, and
the day on which the exposure is estimated.
RTW with half-life at time tk=
k j
exp
(Eq. 5-4)
In addition to the exposure metrics used in the lung-cancer mortality analysis, modeling
of mesothelioma mortality (see Section 5.4.3.1) included the exposure model used in the IRIS
assessment for asbestos (U.S. EPA, 1988a), originally proposed in Peto et al. (1982):
= C-Q-KM
(Eq. 5-5)
Where
Im
C
= the observed deaths from mesothelioma/person-years,
= the average concentration of asbestos in the air,
This document is a draft for review purposes only and does not constitute Agency policy.
5-71 DRAFT—DO NOT CITE OR QUOTE
-------
1 KM = an estimated slope describing the relationship between Libby Amphibole asbestos
2 exposure and mesothelioma mortality, and
3 Q = the function of the time since first exposure (f) and the duration of employment
4 (d):
5 For t< 10, Q = 0
6 VorW d+10 = t-
8
9
10 The asbestos IUR (U.S. EPA, 1988a) metric (see Eq. 5-5) was originally fit to aggregate
1 1 cohort data and was based on a function of average cohort exposure, time since first exposure,
12 and duration of employment. The analysis here of individual data for Libby Amphibole asbestos
13 is, therefore, a different application of this exposure metric, and its fit to the mesothelioma
14 mortality of the Libby worker cohort is evaluated in this current assessment.
15 In addition to the use of these methods of describing exposure metrics representing
16 estimated ambient exposure to Libby Amphibole asbestos dust for use in predicting the risk of
17 mortality, there is the important issue of potentially modifying the exposure metrics to account
18 for cancer latency. Without knowledge of the specific timing of etiologically relevant exposure
19 that may initiate and promote cancers ultimately resulting in mortality, any exposure metric may
20 include exposures during some time period that do not have bearing on the risk of mortality. In
21 the absence of such information on the specific cancer latency associated with a specific
22 exposure, Rothman (1981) suggested that the most relevant exposure period could be identified
23 by comparing the fit of exposure metrics across multiple lag periods to allow for the
24 identification of the optimal latency period as an expression of a lag time between exposure and
25 mortality. This has since become a standard practice in occupational and environmental
26 epidemiology. Accordingly, exposure estimates for all exposure metrics were adjusted to
27 account for the time period between the onset of cancer and mortality. The lag period defines an
28 interval before death, or end of follow-up, during which, any exposure is excluded from the
29 calculation of the exposure metric. Cohort members who died or were lost within the initial
30 years of follow-up were assigned lagged exposure values of zero if they had not been followed
31 for longer than the lag time. The various exposure metrics were lagged at 10, 15, and 20 years to
32 account for different potential cancer latencies within the limitations of the available data.
This document is a draft for review purposes only and does not constitute Agency policy.
5-72 DRAFT— DO NOT CITE OR QUOTE
-------
1 Metrics without a lag were fit for comparison purposes but were not considered to be
2 biologically reasonable, given that the outcome under analysis is cancer mortality (specifically,
3 mesothelioma and lung cancer), for which latency periods of 10 years or more have been
4 established for asbestos (U.S. EPA, 1988a). Consequently, metrics that were not adjusted by
5 lagging exposure in the final years before mortality (or the end of follow-up) were not
6 considered further in the development of an IUR for Libby Amphibole asbestos.
7
8 5.4.3. Exposure-Response Modeling
9 Sufficient biological information to select models for the epidemiology data on the basis
10 of biological mechanism (see Section 3) is not available. In this situation, EPA's practice is to
11 investigate a range of model forms to determine how to best empirically model the
12 exposure-response relationship in the range of the observed data. For Libby Amphibole
13 asbestos, possible exposure metrics were explored for model fit to the chosen models. The
14 exposure metric options were selected to provide a range of shapes that was sufficiently flexible
15 to allow for a variety of ways that time and duration might relate to cancer risk in the data being
16 modeled. EPA then evaluated how well the models and exposure metric combinations fit the
17 data being modeled. Metrics that did not fit the data well were rejected. For purposes of
18 calculating a reasonable upper bound on the risk per exposure, two different types of uncertainty
19 were accounted for. The first uncertainty is in the estimated slope for each exposure metric, and
20 this was accounted for by using the upper bound estimated using the statistical variance of the
21 estimated slope. EPA accounted for the second uncertainty that stemmed from the choice of
22 exposure metrics among the set that fit the data by using the exposure metric (among those few
23 with a reasonable fit) that estimated the highest risk (because formal estimation of an upper
24 bound was not possible). This is explained in more detail below and in Section 5.4.5.
25 The risk estimates are based on epidemiological analysis of the primary NIOSH data
26 (Libby worker cohort). The rationale for selection of the Libby worker cohort is presented in the
27 previous section (see Section 5.4.2). Analysis of this primary epidemiologic database allows the
28 comparison of multiple metrics of exposure to quantify the exposure-response relationship. This
29 approach is intended to support the empirical representation of the exposure-response
30 relationship of estimated ambient occupational exposure to Libby Amphibole asbestos with
31 observed cancer mortality risk. The exposure-response modeling may be influenced by
This document is a draft for review purposes only and does not constitute Agency policy.
5-73 DRAFT—DO NOT CITE OR QUOTE
-------
1 uncertainties in the magnitude and time course of the exposure estimates and, therefore, may not
2 necessarily reflect the biologic disposition of inhaled fibers (see Section 5.4.6).
3 The following sections provide information about modeling of the full cohort first, the
4 difficulties in identifying adequately fitting models to these data, and the decision to base the
5 analysis on a sub-cohort of workers that did allow for identifying adequately fitting models.
6
7 5.4.3.1. Modeling of Mesothelioma Exposure Response in the Libby Worker Cohort
8 The background incidence of mesothelioma is extremely rare (Hillerdal, 1983). Since
9 there is a very low background risk, the exposure-response model applied here examines the
10 relationship of the absolute risk of mesothelioma mortality that is attributable to Libby
11 Amphibole asbestos exposure because there is not a true background risk of mesothelioma
12 mortality among people who were truly unexposed to Libby Amphibole asbestos (as opposed to
13 the relative risk model, which is used for lung-cancer mortality; see Section 5.4.3.3). Poisson
14 regression models are employed here for estimating the absolute risk of mesothelioma, as the
15 Poisson distribution is an appropriate model for use with data that are counts of a relatively rare
16 outcome, such as observed mesothelioma deaths in the Libby worker cohort. Other analyses of
17 mesothelioma mortality in the Libby worker cohort have also used the Poisson regression model
18 (Moolgavkar et al., 2010; McDonald et al., 2004). In the Poisson regression model, probability
19 of k events is specified as
20
21
22 P(k) = ^—
k\
23 (Eq. 5-6)
24
25
26 where X is parameterized with the exposure metric (defined in Section 5.4.2.5). Then, life-table
27 analysis is used to estimate risks in the general U.S. population for the derivation of the unit risk
28 of mesothelioma mortality (see Section 5.4.5.1).
29 Estimation of the exposure-response relationship for mesothelioma mortality using the
30 Poisson regression model was performed using a Monte Carlo Markov Chain (MCMC) Bayesian
31 approach with an uninformative or diffuse prior [WinBUGS Version 1.4 (Spiegelhalter et al.,
32 2003)]. Use of diffuse priors is a standard procedure in Bayesian analysis, in situations like this
This document is a draft for review purposes only and does not constitute Agency policy.
5-74 DRAFT—DO NOT CITE OR QUOTE
-------
1 one, when there is no prior knowledge about the toxicity of Libby Amphibole asbestos under a
2 particular model. Since this analysis focuses only on the Libby worker cohort and does not try to
3 factor in data from other sources in estimating potency, use of a diffuse prior is considered
4 appropriate for this analysis.
5 The benefit of using the WinBUGS software is its computational ease and that it provides
6 a posterior distribution of P (the mesothelioma slope factor) rather than just a point estimate. A
7 diffuse (high variance) Gaussian distribution, truncated to exclude negative parameter values, is
8 used as a diffuse prior. With such a prior, results of MCMC analysis are expected to be similar
9 to maximum likelihood estimation in a non-Bayesian analysis. Standard practices of MCMC
10 analysis were followed for verifying convergence and sensitivity to the choice of initial values.
11 The posterior distribution is based on three chains with a burn-in of 10,000 (i.e., the first
12 10,000 simulations are dropped so that remaining samples are drawn from a distribution close
13 enough to the true stationary distribution to be usable for estimation and inference) and thinning
14 rate of 10 (i.e., only each 10* simulation is used—thus reducing autocorrelation) such that
15 3,000 total simulations constitute the posterior distribution of p. The mean of the posterior
16 distribution served as a central estimate, and the 90% credible interval38 defined the 5* percentile
17 and the 95th percentile of the distribution, which served as bounds for the 95th lower and upper
18 one-sided confidence intervals, respectively.
19 Multiple metrics of exposure (see Section 5.4.2.5) as well as exposure intensity, duration
20 of employment, age at death or loss to follow-up, and time since first exposure were compared
21 using the Deviance Information Criterion (DIG). The DIG (Spiegelhalter et al., 2002) is used in
22 Bayesian analysis and is an analogue of the AIC, with smaller values indicating a better
23 statistical fit to the data. Use of the DIG and AIC is standard practice in comparing the fit of
24 nonnested models to the same data set with the same dependent outcome variable but different
25 independent covariates. According to Burnham and Anderson (2002), "These methods allow the
26 data-based selection of a "best" fitting model and a ranking and weighting of the remaining
27 models in a predefined set." Because of the small number of deaths from mesotheliomas in
28 absolute terms, only uni- and bi-variate models (with age or time since first exposure as the
29 second covariate) were considered. Sex and race were not used as covariates since all
30 mesotheliomas were observed in men assumed to be white (Sullivan, 2007). Each exposure
38A credible interval is the Bayesian analogue of a confidence interval.
This document is a draft for review purposes only and does not constitute Agency policy.
5-75 DRAFT—DO NOT CITE OR QUOTE
-------
1 metric was lagged by 0, 10, 15, or 20 years. The use of a lag period aims to account for the
2 latency period between the onset of mesothelioma (which occurs some time before clinical
3 diagnosis) and mesothelioma mortality.
4
5 5.4.3.2. Mesothelioma Mortality Analysis in the Libby Worker Cohort
6 For the full Libby worker cohort (n= 1,871), the duration of employment provided a
7 considerably better univariate model fit than the other possible exposure metrics, indicating that
8 this exposure metric was the best single predictor of mesothelioma mortality in the full Libby
9 worker cohort. The bivariate model, which included duration of employment and age at death or
10 censoring, provided the overall best fit (DIG = 196). The inclusion of information on the
11 concentration of exposure beyond the duration of employment resulted in a degradation in model
12 fit (see Table 5-9). The metric used in the IUR for asbestos (U.S. EPA. 1988a) (see Eq. 5-5) had
13 a much higher DIG of 233.7 in the analysis here. It is likely that the poorer fit seen when using
14 information on exposure concentration is the result of the fact that duration of employment is
15 measured with comparatively little error, while derivation of specific exposure concentrations
16 may be subject to a sizable measurement error. Moreover, as described in Section 5.4.2.3, for
17 706 of 991 (71%) workers hired from 1935 to 1959, only the duration of employment was
18 known, but not the job category or department code, and, thus, the same time-weighted average
19 estimated exposure intensity for that time period had been assigned to 653 of these workers39
20 (Sullivan, 2007). It is likely that because of the potential for particularly large exposure
21 measurement error among more than two thirds of the workers hired prior to 1960 who were
22 assigned the same exposure intensity, this resulted in the duration of employment being the best
23 predictor of mesothelioma mortality. Additionally, estimates of exposure intensity prior to 1968
24 have greater uncertainty associated with them than more recent exposure measurements, which
25 are based on fiber counts in air samples analyzed by PCM. For the majority of job locations
26 (23 of 25), no exposure measurements were available prior to 1968, and exposures were
27 estimated based on employee interviews (in 1982) and what was known about major changes in
28 operations between 1935 and 1967. For two exposure locations, the dust-to-fiber conversion
29 ratio is based on measurements taken in the late 1960s, so extrapolations from the mid-1960s to
39Note that Sullivan (2007) analyzed the population of 1,672 white male workers rather than all 1,871 workers so the
numbers of workers with missing job category and department information were different.
This document is a draft for review purposes only and does not constitute Agency policy.
5-76 DRAFT—DO NOT CITE OR QUOTE
-------
1 the early 1960s is likely to be more certain than extrapolation further back in time. The fact that
2 the metric using only duration of employment fit best and the additional incorporation of
3 exposure intensity information worsened the fit indicates that it is unlikely that IUR estimates
4 can be developed using the full cohort data because exposure values were not predictive of
5 mesothelioma mortality.
6
7
8 Table 5-9. Comparison of univariate model fit of various exposure metrics
9 for mesothelioma mortality in the full Libby worker cohort (« = l,871)a'b
10
Variable
Duration of employment
Age at death or censoring
CE lagged 15 yr
CE lagged 10 yr
RTW lagged 10 yr with 5-yr !/2 life
CE lagged 10 yr with 20-yr !/2 life
RTW with 5-yr !/2 life
RTW with 10-yr !/2 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
11
12 aSince one of the mesothelioma deaths occurred less than 20 years from start of the exposure, lag 20 metrics
13 assigned no exposure to this case, which resulted in the very poor fit of exposure metrics lagged 20 years.
14 bLower DIC values represent better fits. Models with DIC within 10 units of the DIC of the model with the lowest
15 DIC are shown.
16
17 DIC = Deviance Information Criterion.
18
19
20 The DIC values for models that included lag and/or half-life adjustments to the exposure
21 metrics were not penalized in the regression analyses for including these extra parameters
22 because those factors were not represented as covariates but rather were embedded in the
23 exposure metrics. While these results were obtained using each instance with lag and/or half-life
24 as a separate model fit, it may be appropriate to penalize the DIC values from these results for
25 inclusion of these parameters. Note that if the DIC values from the lag and/or half-life models
26 were penalized, this would serve to improve the relative fit of the model using only duration as a
27 parameter in comparison with the lag and/or half-life models because the DICs for the penalized
28 models would increase while the DIC for the unpenalized models would be unchanged.
This document is a draft for review purposes only and does not constitute Agency policy.
5-77 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.4.3.3. Modeling of Lung Cancer Exposure Response in the Libby Worker Cohort
1 To develop an exposure-response relationship for lung cancer, the lung-cancer mortality
3 data were modeled as a function of the historical exposure data for the Libby worker cohort. The
4 mesothelioma mortality data were modeled to estimate the absolute risk because it is very rare in
5 the general population (Hillerdal, 1983). Lung-cancer mortality does have a known background
6 risk, and, thus, modeling of lung-cancer mortality is based on the relative risk rather than the
7 absolute risk. As such, there are different analytic methods available that can use information on
8 time-varying exposures. The NIOSH-developed individual-level exposure data for the Libby
9 worker cohort are very detailed, with start and stop dates for each of the workers' jobs and
10 estimated fiber exposures for 25 specific location-operations (Amandus et al., 1987a). It is,
11 therefore, important to find a model that makes efficient and effective use of these
12 time-dependent data.
13 The Cox proportional hazards model (Cox, 1972) is one of the most commonly used
14 statistical models for the epidemiologic analysis of survival and mortality in cohort studies with
15 extensive follow-up (Larson et al., 201 Ob: Moolgavkar et al., 2010). In the Cox proportional
16 hazards model, the conditional hazard function, given the covariate Z, is assumed to have the
17 form
18
19
20 Mt | Z) = A0(Oe>q3(/?rZ) (Eq. 5-7)
21
22
23 where P is the vector of regression coefficients, Xo(0 denotes the baseline hazard function, and T
24 denotes transposition of the vector. One of the strengths of this model is that knowledge of the
25 baseline risk function is not necessary, and no particular shape is assumed for the baseline
26 hazard; rather, it is estimated nonparametrically. The contributions of covariates to the hazard
27 are multiplicative. When Z represents exposure and ffZ is small, the Cox proportional hazards
28 model is consistent with linearity of dose response for low doses.
29 When the proportional hazards assumption holds, it is possible to estimate the hazard
30 ratio of exposure (relative risk) without estimating the hazard function in the unexposed (or in
31 the lowest exposures seen within the study group) since this baseline hazard function drops out
32 of the calculations. The Cox proportional hazards model assumes that a function of covariates
This document is a draft for review purposes only and does not constitute Agency policy.
5-78 DRAFT—DO NOT CITE OR QUOTE
-------
1 (i.e., exposures) result in risks that are a constant multiple of the baseline hazard in unexposed
2 individuals over some timescale, typically calendar time or age. This proportionality is assumed
3 to be constant across the range of observed exposures, given the set of modeled covariates, and
4 can be evaluated across time.
5 The Cox proportional hazards model was chosen to represent the lung-cancer mortality
6 data for several reasons. Of primary importance is that it takes statistical advantage of the
7 extensive exposure data collected for the cohort on time-varying exposures to Libby Amphibole
8 asbestos. There are no other standard model formulations that allow for the analysis of
9 time-varying exposures in the manner achieved by the Cox proportional hazards model. The
10 exposure-response relationship (proportional hazards ratio) determined in this model intrinsically
11 takes into account the effects of other causes of mortality that are unrelated to exposure (i.e.,
12 independent censoring). Further, all comparisons are made within the cohort by comparing the
13 mortality experience of people with different exposures within the same cohort population. The
14 issue of competing risks that are dependent on exposure (e.g., asbestosis or nonmalignant
15 respiratory disease) is an acknowledged uncertainty for this type of analysis (see Section 5.4.6).
16 Other methods common to occupational epidemiology, such as the use of standardized
17 mortality ratios typically rely upon comparisons of the mortality experience in an exposed
18 population group compared to that in the general population. However, the comparison
19 population may not always be appropriate due to differences in general health status (e.g., the
20 healthy worker effect) and differences in exposure to other risk factors for a specific disease
21 (e.g., smoking history). The lack of comparability between the study population and the
22 comparison population can lead to confounding by other measured or unmeasured
23 characteristics, which may be statistically associated with both the exposure of interest and the
24 endpoint. The Cox proportional hazards model controls for such potentially confounding
25 characteristics by using a comparison group from within the study population (i.e., internal
26 controls). Internal controls are a statistically appropriate comparison group because they are
27 expected to be more similar in potentially confounding characteristics to the remainder of the
28 cohort, thereby controlling for both measured and unmeasured confounding and helping ensure
29 that comparisons are more statistically valid.
30
This document is a draft for review purposes only and does not constitute Agency policy.
5-79 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.4.3.4. Lung-Cancer Mortality Analysis in the Libby Worker Cohort
1 As described in the previous section, quantitative exposure-response relationships for
3 lung-cancer mortality were evaluated using the Cox proportional hazards model. Cox
4 proportional hazards models of this type require the specification of a timescale. Age is typically
5 the time-related variable with the strongest relationship to cancer mortality and was used as the
6 timescale in these analyses. Use of age as the timescale in a time-varying Cox proportional
7 hazards model controls for age as a risk factor by design rather than by parametric modeling and
8 effectively rules out age as a potential confounder. Individual covariates available to EPA in the
9 complete analytic data set compiled from the NIOSH data were evaluated for their ability to
10 explain the lung-cancer mortality. These included sex, race, birth year, age at hire, and various
11 exposure-related variables including TWA workplace intensity of exposure in fibers/cc, j ob type,
12 and the start and stop date of each different job. These data allowed for the computation of
13 cumulative exposure, cumulative exposure with application of a half-life, and RTW cumulative
14 exposure, with and without application of a half-life (see Section 5.4.2.5). Each exposure metric
15 was also lagged by 0, 10, 15, or 20 years. The use of a lag period aims to account for the latency
16 period between the onset of lung cancer (which occurs some time before clinical diagnosis) and
17 lung-cancer mortality.
18 All lung-cancer mortality analyses were conducted using SAS software version 9.1 (SAS,
19 Gary, NC). EPA fit the extended Cox proportional hazards model (Tableman and Kim, 2004;
20 Kleinbaum and Klein, 1996), which included both time-independent factors such as sex, race,
21 and date of birth, as well as time-dependent measures of Libby Amphibole asbestos exposure
22 over the entire time course of each individuals' lifetime from their date of hire until death or loss
23 to follow-up. This method allows for control of potential confounding by age by design rather
24 than through multivariate covariate modeling. The inclusion of date of birth in these analyses
25 controls for any potential birth cohort effect.
26 EPA's analyses of time-dependent exposure data included goodness-of-fit testing of the
27 proportionality assumption for the Libby worker cohort. Because Cox proportional hazard
28 models rely on the assumption that the hazard rate among the exposed is proportional to the
29 hazard rate among the unexposed, it is important to evaluate the model against this assumption.
30 Therefore, analyses of extended Cox proportional hazards models tested this assumption using a
31 Wald test on the model interaction term between the Libby Amphibole asbestos exposure metric
This document is a draft for review purposes only and does not constitute Agency policy.
5-80 DRAFT—DO NOT CITE OR QUOTE
-------
1 and the timescale (i.e., age). As a general rule, a nonzero slope that is either increasing or
2 decreasing indicates a violation of the proportional hazards assumption. Wald tests for the
3 complete cohort consistently showed that the interaction term was a statistically significant
4 predictor of lung-cancer mortality (p < 0.05) and was interpreted as evidence that the hazards did
5 not remain proportional over time. The cause of the lack of proportionality is unknown, but
6 several likely explanations are discussed in Section 5.4.3.5 below and in the discussion of
7 uncertainties in Section 5.4.6.1.
8
9 5.4.3.5. Summary of Mesothelioma and Lung Cancer Analysis ofLibby Worker Cohort
10 Several possible explanations exist for the finding that duration of employment was the
11 best fitting exposure metric for mesothelioma mortality, as well as the finding of the lack of
12 proportionality of hazards in the lung-cancer mortality modeling.
13
14
15 1) Duration of employment, but neither department code nor job category, was known for
16 706 of 991 (71%) workers hired from 1935 to 1959. Without knowledge of the job
17 category, the same exposure concentration had been assigned to almost all of these
18 workers, likely resulting in a particularly large measurement error for exposure in
19 approximately one third of the total cohort of 1,871 workers. This is a very likely
20 explanation for the superior fit for duration of employment in modeling of mesothelioma
21 mortality relative to the other exposure metrics based on measured exposures. Assigning
22 the same exposure concentration to so many of the workers hired before 1960, regardless
23 of job, likely resulted in significant exposure misclassification. Random error in
24 exposure measurements generally attenuates the strength of epidemiologic associations
25 between exposure and observed effect, weakening the predictive ability of any of the
26 exposure-based metrics compared to duration of employment, which was more accurately
27 determined for all workers in the cohort.
28 2) Even where the job category was identified, few exposure data exist prior to 1968. For
29 the majority of job locations (23 of 25), no exposure measurements were available prior
30 to 1967, and so exposures were estimated based on employee interviews (conducted in
31 1982) to determine what was known about major changes in operations between 1935
32 and 1967. For two job locations, dust-to-PCM extrapolations are based on measurements
33 taken in the late 1960s, so extrapolating from the mid-1960s to the early 1960s is likely to
34 be more certain than extrapolating further back in time. Random error in these exposure
35 measurements would also generally attenuate the strength of association between
36 exposure and observed effect during the earlier years of mine operation and, thus, a
37 greater degree of measurement error in the earlier years could have resulted in the lack in
38 proportionality of the hazard ratios for lung cancer over time. A greater degree of
39 measurement error in the earlier years could also provide an explanation for the worse fit
40 of the mesothelioma models that incorporated these exposure measures.
This document is a draft for review purposes only and does not constitute Agency policy.
5-81 DRAFT—DO NOT CITE OR QUOTE
-------
1 3) Another explanation for the lack of proportional hazards in modeling lung-cancer
2 mortality may be that this cohort has an anomalous age structure due to the hiring of
3 much older individuals during the time of the Second World War. Among those workers
4 in the cohort hired prior to 1960, 9% were older than 50 years at the time of hire, and
5 22% were older than 40 years. Among those workers hired in 1960 or afterwards, only
6 4% were older than 50 years, and 14% were older than 40 years. Older workers differ
7 from younger workers in several potentially important ways that could alter their
8 response to exposures. Older workers were born in a different era, with different
9 nutritional and public health standards which may influence mortality patterns.
10 4) The lack of proportional hazards in modeling lung-cancer mortality may also be a
11 reflection of confounding or effect modification, which can change in magnitude over
12 time. The most likely candidate for confounding or effect modification is smoking.
13 NIOSH records show that of the 1,871 workers in the full Libby workers cohort,
14 1,121 workers (60%) were missing smoking status data, while 750 (40%) had data with
15 values "S" (Smoker), "Q" (Former Smoker), or "N" (Nonsmoker). Given this high
16 percentage of missing values, EPA did not consider these smoking data to be adequate
17 for use in the evaluation of confounding or effect modification.
18 5) Smoking rates, over time, among the sub-cohort of workers hired after 1959 are likely to
19 have been more similar since smoking rates change more slowly over shorter periods of
20 time than over longer ones. This restriction in time period of hiring would also result in
21 less variation by birth year cohort, which is strongly related to smoking patterns as people
22 of different generations developed different smoking rates. Thus, this restriction in the
23 time period of hiring may make the cohort members more similar to each other, thereby
24 reducing the potential impact of any smoking-related confounding. Further discussion of
25 the relevance of smoking can be found in the section on uncertainties (see Section 5.4.6).
26
27
28 When the assumption of proportionality is not met, the potential influence of
29 confounding factors in the full-cohort analysis is of concern. Additionally, the lack of job
30 category information for 69% of the workers hired prior to 1960 and greater measurement error
31 in early exposures may result in significant random exposure measurement error, which may bias
32 the observed exposure-response relationships towards the null.
33 Although duration of employment was the best exposure metric for modeling
34 mesothelioma mortality in the full cohort, it made quantitatively estimating an exposure-response
35 relationship difficult. In addition, violation of the underlying statistical assumptions adversely
36 impacted modeling of lung-cancer mortality in the full cohort. Therefore, EPA chose to
37 undertake a sub-cohort analysis.
This document is a draft for review purposes only and does not constitute Agency policy.
5-82 DRAFT—DO NOT CITE OR QUOTE
-------
1 In particular, because uncertainty in retrospective assessment of workplace exposures is
2 reduced in the later years, EPA decided to analyze a sub-cohort of all the workers with as late a
3 starting employment date as possible, while still maintaining a sufficient number of lung cancer
4 and, especially, mesothelioma mortalities. Nearly all of the workers with completely missing
5 data on job category or department code and only duration of employment available were hired
6 before 1960, and so EPA developed a sub-cohort analysis by dividing the total cohort into those
7 hired prior to 1960 (n = 991) and those hired after 12/31/1959 (n = 880). This cut point roughly
8 divided the cohort in half. For the sub-cohort of those workers hired after 1959, there were
9 sufficient numbers of both mesothelioma and lung cancer mortalities to apply the Poisson and
10 Cox proportional hazards model, correspondingly. EPA initially examined the fit of these
11 models using several exposure metrics to predict mortality from mesothelioma and found that in
12 this sub-cohort, the exposure metrics that included information on exposure concentration
13 provided superior statistical fits to the exposure metrics based only on employment duration. In
14 this same sub-cohort, the assumptions of the Cox proportional hazards model were also satisfied
15 for the modeling of time-varying exposure.
16 While it is generally true that the use of more data is an advantage in statistical analyses
17 because it allows for the computation of more statistically precise effect estimates, this advantage
18 could not be utilized, because of the difficulty in deriving risks from the full cohort analysis (see
19 also Section 5.4.6 on uncertainties remaining in the sub-cohort).
20
21 5.4.3.6. Analysis of Sub-Cohort of Employees Hired After 1959
22 The reasons stated in Section 5.4.2 for choice of Libby worker cohort data are still valid
23 for the sub-cohort. In particular, (1) these workers were directly exposed to Libby Amphibole
24 asbestos, (2) detailed work histories and job-specific exposure estimates are available to
25 reconstruct estimates of each individual's occupational exposure experience with only 9 workers
26 completely missing job and department codes during the period when relatively high average
27 time-weighted estimated exposure intensity was assigned, (3) the sub-cohort is still sufficiently
28 large and has been followed for a sufficiently long period of time for cancer to develop (i.e.,
29 cancer incidence) and result in mortality, and (4) the broad range of exposure experiences in the
30 sub-cohort provided an information-rich data set.
31
This document is a draft for review purposes only and does not constitute Agency policy.
5-83 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.4.3.6.1. Results of analysis of mesothelioma mortality in the sub-cohort
1 Of the 880 workers hired after 1959, 230 (26%) had died by December 31, 2006. The
3 number of mesothelioma deaths in the sub-cohort is 7 (2 deaths coded in ICD-10 and 5 deaths
4 coded in ICD-9), and the mesothelioma death rate of 24.7 per 100,000 person-years for the
5 sub-cohort is similar to the mesothelioma death rate of 26.8 per 100,000 person-years for the full
6 cohort (18 mesothelioma deaths), with a difference of less than 10%.
7 Table 5-10 shows the relative fit of various exposure metrics for mesothelioma mortality
8 in the sub-cohort hired after 1959, including only those exposure metrics whose information
9 weight was greater than 0.01. Information weights are computed from the DICs (Burnham and
10 Anderson, 2002). As discussed below, metrics with higher DICs and lower information weights
11 are unlikely to provide a good fit and are, thus, not included in Table 5-10. Information weights
12 are commonly used in Bayesian analyses. Information weights can be computed by first
13 assessing the differences between the best DIC and each of the others (A Did).
14
15
( 1 \ 1^ ( \ \
16 DIC w, = exp -- A DIG, / J]exp -- A DIG, (Eq. 5-8)
v 2 )J r=1 \_ 2 )
17
18
19 The other exposure metrics that were fit included those metrics used in the full cohort
20 analysis [duration of employment, time since first exposure, age at death or censoring, RTW
21 metrics, CE with lag metrics, and IRIS IUR (1988a) metric], but all of them fit worse than any of
22 the metrics in Table 5-10, irrespective of possible penalization for extra parameters as discussed
23 in the analysis of the full cohort. The two metrics with cumulative exposure lagged 15 and
24 10 years, both with 5-year half life, provided the two best fits as indicated by their lower DIC
25 values and higher information weights (see Table 5-10). Cumulative exposures lagged 10 or
26 15 years, both with 10-year half life, provided the next two best fits according to DIC values, but
27 models including each of these metrics exhibited noticeably lower information weights than the
28 best metric. All metrics in Table 5-10 contain a decay term and have the same number of
29 parameters in their corresponding model, allowing for a direct comparison of the DIC values
30 (DICs are similar to AICs in what is considered an important difference) and information
31 weights. It is important to note that the suite of exposure metrics that were applied in this current
32 assessment to modeling mesothelioma mortality encompass the range of choices described in the
This document is a draft for review purposes only and does not constitute Agency policy.
5-84 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
o
J
4
Table 5-10. Comparison of model fit of exposure metrics for mesothelioma
mortality in the sub-cohort hired after 1959a'b. Only the model fits with
information weights greater than 0.010 are shown
Exposure metric
CE with 5-year !/2 life
CE with 5-year !/2 life
CE with 10-year !/2 life
CE with 10-year !/2 life
CE with 10-year !/2 life
CE with 5-year !/2 life
CE with 15-year !/2 life
CE with 15-year !/2 life
CE with 15-year !/2 life
CE with 20-year !/2 life
CE with 20-year !/2 life
CE with 20-year !/2 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
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
aLower DIC values represent better fits.
bSince one of mesothelioma deaths occurred in less than 20 years 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.
DIC = Deviance Information Criterion.
asbestos literature including CE, RTW, and decay metrics as well as the IRIS IUR (U.S.
EPA, 1988a) metric.In the sub-cohort hired after 1959, the DIC value for mesothelioma using the
IRIS IUR (U.S. EPA. 1988a) metric (see Eq. 5-5) is substantially higher (DIC = 98.4) than for
any of the metrics in Table 5-10. This indicates that the IRIS IUR (U.S. EPA. 1988a) metric
does not provide as good a fit for the Libby Amphibole asbestos worker cohort, using the
estimated historical exposure levels, as the other metrics in Table 5-10. Setting the exponents in
the IRIS IUR (U.S. EPA. 1988a) metric to the values of 2 and 4, as suggested by EPA (1986aX
did not improve the fit of the metric to the Libby Amphibole asbestos worker cohort data (results
not shown). A substantial difference of this analysis from the IRIS IUR (U.S. EPA. 1988a)
modeling is that this analysis is based on individual-level data, whereas the IRIS IUR (U.S. EPA,
1988a) application was to aggregate data. Also, cohorts used in the IRIS IUR (U.S. EPA. 1988a)
did not include cohorts exposed to Libby Amphibole asbestos. Alternately, the relative fit of this
model may have been affected by uncertainties in the estimated exposure described in detail in
Section 5.4.6.
This document is a draft for review purposes only and does not constitute Agency policy.
5-85 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
o
J
4
5
6
7
8
9
10
11
12
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
"time since first exposure" did not improve the fit, as measured by DIG (results not shown).
As described in Section 5.4.2.5, only metrics with nonzero lag were retained for
derivation of unit risks. Table 5-11 shows slopes and credible intervals for all retained metrics
from Table 5-10. The units of the slopes are fiber/cc-year. These slopes and credible intervals
represent calendar year continuous environmental exposure as described above and define the
Table 5-11 Mesothelioma mortality exposure metrics fits, slopes, and
credible intervals
Exposure metric
CE - 5-yr !/2 life
CE - 5-yr !/2 life
CE - 10-yr 1A life
CE - 10-yr !/2 life
CE-15-yr!/2life
CE-15-yr!/2life
CE - 20-yr !/2 life
CE - 20-yr !/2 life
Lag years
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% CI for slope x 10~5
(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)
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
CI = credible interval.
"Exposed Hazard Rate" in the life-table procedure when multiplied by the exposure level (see
Appendix G for details).
Based on the results from the exposure metric with the lowest DIC (cumulative exposure
with a 5-year half life for decay and a 15-year lag for cancer mortality latency), the slope was
2.06 x io~4 per fiber/cc-year based on a 365-day calendar year, and the 95% upper bound on the
slope was 3.43 x io~4 per fiber /cc-year. This point estimate and 95% upper bound represent the
relative risk (including statistical uncertainty within the exposure metric) of mesothelioma
mortality observed from exposure to Libby Amphibole asbestos fibers in the worker cohort for
this exposure metric. Issues related to uncertainty in the choice of exposure metric are described
further in the section on the derivation of the combined IUR of mesothelioma and lung cancer
(see Section 5.4.5.3).
This document is a draft for review purposes only and does not constitute Agency policy.
5-86 DRAFT—DO NOT CITE OR QUOTE
-------
1
2 5.4.3.6.2. Results of the analysis of the lung-cancer mortality in the sub-cohort
3 EPA based its final analyses for lung-cancer mortality on the subset of workers hired
4 after 1959. Thus, this analysis is based on 32 deaths from lung cancer40 (ICD-8: two deaths with
5 the code 162.1; ICD-9: one death with the code 162.2, 20 deaths with the code 162.9; ICD-10:
6 nine deaths with the code C349) out of 230 deaths that occurred in the sub-cohort of 880
7 workers.
8 All multivariate Cox proportional hazards models with time-varying exposures were
9 initially fit, using one exposure metric at a time, to the sub-cohort hired after 1959 with
10 covariates for sex, race, and date of birth. Lung-cancer mortality was modeled using CE and
11 RTW exposure, where each metric was potentially modified by four different half-lives (5, 10,
12 15, or 20 years). Each of these exposure metrics was also evaluated with four different lag
13 periods to allow for cancer latencies of 0, 10, 15, or 20 years. The lag period is defined as
14 immediately prior to observed cancer death, where exposure is not considered to be causally
15 related to mortality. In all, 40 exposure response multivariate models were evaluated for the
16 adequacy of the exposure metric to fit the epidemiologic data. Each exposure metric and the
17 comparative model fit statistics are presented in Table 5-12.
18 The assumptions of the Cox proportional hazards model were reevaluated for the
19 sub-cohort. Restricting the cohort addressed each of the previously listed potential explanations
20 for the lack of hazard proportionality (see Section 5.4.3.3). First, measurement error for
21 exposures is likely to have been smaller after 1959 for several reasons. One reason is that the
22 706 workers for whom job category and department code information was missing during all of
23 their employment prior to 1960 were removed from the analysis. Also, beginning in 1968, fiber
24 concentrations by PCM analysis of site-specific air samples were available for all location
25
40 Note that in the full cohort, it was unclear whether there were cases of trachea! cancer included in the definition of
lung cancer as many of the recorded ICD codes on death certificates did not provide sufficient detail to distinguish
tracheal cancer cases from lung cancer cases. However, among the sub-cohort 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 there were no deaths from tracheal cancer.
This document is a draft for review purposes only and does not constitute Agency policy.
5-87 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
o
6
4
5
Table 5-12. Model fit comparison for different exposure metrics and
lung-cancer mortality associated with Libby Amphibole asbestos, controlling
for age, gender, race, and date of birth
Ordered by exposure metric
Exposure metric
CE
CE
CE
CE
CE 20-yr !/2 life
CE 20-yr !/2 life
CE 20-yr !/2 life
CE 20-yr !/2 life
CE15-yr!/2life
CE15-yr!/2life
CE15-yr!/2life
CE15-yr!/2life
CE 10-yr !/2 life
CE 10-yr !/2 life
CE 10-yr !/2 life
CE 10-yr !/2 life
CE 5-yr !/2 life
CE 5-yr !/2 life
CE 5-yr !/2 life
CE 5-yr !/2 life
RTW
RTW
RTW
RTW
RTW 20-yr !/2 life
RTW 20-yr !/2 life
RTW 20-yr !/2 life
RTW 20-yr !/2 life
RTW 15-yr !/2 life
RTW 15-yr !/2 life
RTW 15-yr !/2 life
RTW 15-yr !/2 life
RTW 10-yr !/2 life
RTW 10-yr !/2 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
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
365.001
364.662
362.283
364.041
Ordered by model fit
Exposure metric
CE 10-yr !/2 life
CE 5-yr !/2 life
CE 15-yr !/2 life
CE 20-yr !/2 life
CE 5-yr !/2 life
CE 10-yr !/2 life
CE
CE 20-yr !/2 life
CE 15-yr !/2 life
CE 15-yr !/2 life
CE 20-yr !/2 life
RTW 5-yr !/2 life
CE
CE 10-yr !/2 life
RTW 10-yr !/2 life
RTW 15-yr !/2 life
RTW 20-yr !/2 life
CE
RTW 5-yr !/2 life
CE 5-yr !/2 life
RTW
RTW 10-yr !/2 life
CE 5-yr !/2 life
RTW 15-yr !/2 life
CE 10-yr !/2 life
RTW 20-yr !/2 life
RTW
CE 15-yr !/2 life
RTW 20-yr !/2 life
RTW 15-yr !/2 life
CE 20-yr !/2 life
RTW 10-yr !/2 life
RTW 5-yr !/2 life
RTW 5-yr !/2 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
20
20
15
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
364.703
364.719
364.768
364.831
Multivariate
model /7-value
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
0.1014
0.1021
0.1041
0.1067
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
0.5530
0.6188
0.6021
0.6884
This document is a draft for review purposes only and does not constitute Agency policy.
5-88 DRAFT—DO NOT CITE OR QUOTE
-------
Table 5-12. Model fit comparison for different exposure metrics and lung-cancer mortality
associated with Libby Amphibole asbestos, controlling for age, gender, race, and date of
birth (continued)
Ordered by exposure metric
Exposure metric
RTW 10-yr !/2 life
RTW 10-yr !/2 life
RTW 5-yr !/2 life
RTW 5-yr !/2 life
RTW 5-yr !/2 life
RTW 5-yr !/2 life
Lag
(yr)
15
20
0
10
15
20
AIC
364.962
364.719
361.593
363.224
364.768
364.831
Ordered by model fit
Exposure metric
RTW
RTW 10-yr !/2 life
CE
RTW
RTW 15-yr !/2 life
RTW 20-yr !/2 life
Lag
(yr)
10
15
20
15
15
15
AIC
364.835
364.962
364.964
364.990
365.001
365.011
Multivariate
model /7-value
0.1069
0.1124
0.1125
0.1136
0.1141
0.1146
Exposure
/7-value
0.6586
0.8173
0.8204
0.8809
0.9100
0.9599
1
2
o
J
4
5
6
7
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
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.
AIC: Akaike Information Criterion.
operations to inform the JEM. Prior to 1968, the exposure intensity for 23 of 25 location
operations was estimated based on reasoned assumptions informed by employee interviews in
the early 1980s. It is likely the uncertainty of these reasoned assumptions increased the farther
back in time that exposures were estimated, making the earliest exposure estimates (1940s and
1950) less certain than those only a few years before fiber count data were available. Finally,
between 1956 and 1967, 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. (1987a), dust-to-fiber conversions are likely to
be less uncertain than extrapolations further backwards 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.
In addition, by focusing on the more homogeneous age distribution of workers hired after
1959, concerns about differential cancer mortality latency were diminished. Third, smoking
rates among this more narrowly defined sub-cohort are likely to have been more homogeneous,
and, thus, restricting analysis to this sub-cohort 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 sub-cohort hired post-1959. There was no evidence
This document is a draft for review purposes only and does not constitute Agency policy.
5-89 DRAFT—DO NOT CITE OR QUOTE
-------
1 to reject the hypothesis of proportionality, and the exposure models demonstrated adequate fits to
2 the data, with statistically significant effect estimates. In each of the Cox proportional hazards
3 model analyses with time-varying exposures—across all the exposure metrics and across all the
4 lag lengths—no violations of the assumption of proportionality of hazards were found.
5 As the exposure-response models cannot strictly be considered to be nested, a standard
6 measure of fit called the Akaike Information Criterion [AIC; Burnham and Anderson (2002)]
7 was used for comparison of goodness of fit across models based on the same data set. In their
8 text on model selection, Claeskens and Hjort (2008) state that".. .for selecting a model among a
9 list of candidates, Akaike's information criterion (AIC) is among the most popular and versatile
10 strategies." Claeskens and Hjort (2008) also state that the model yielding the smallest AIC is
11 judged the best one and it is a common practice in environmental epidemiology to simply select
12 the single model with the best statistical fit (i.e., the lowest AIC) among the models that were
13 evaluated. Smaller AIC values generally indicate a better fitting model relative to larger AIC
14 values. While large differences in AIC values can reveal important differences in model fit,
15 small differences are less conclusive. For example, models differing in AIC by 2 or less units
16 can be considered to have a substantial level of empirical support [Burnham and Anderson
17 (2002): p. 70].
18 Table 5-12 shows the models and exposure metrics ordered by fit. Of interest is whether
19 there are models with distinct exposure metrics that adequately fit these data (as measured by
20 statistical significance of the model p-va\ue) and then, a measure of relative fit among these
21 adequately fitting models. Of the 40 exposure-response metrics, 14 demonstrated an adequate fit
22 to the data as measured by the overall model fit, with the likelihood ratio test being statistically
23 significant (p < 0.05), as well as having statistically significant exposure metrics (p < 0.05).
24 However, note that only the nine models that demonstrated adequate model and exposure metric
25 fit and incorporated a lag period to account for lung-cancer mortality latency were advanced for
26 potential use in developing a unit risk. While metrics that did not include an adjustment for lag
27 on the exposure metric to account for cancer mortality latency were fit to these data for the sake
28 of completeness, they were dropped from further consideration because they implicitly assume
29 no passage of time between the initiation of cancer, subsequent promotion of that cancer, and
30 mortality.
This document is a draft for review purposes only and does not constitute Agency policy.
5-90 DRAFT—DO NOT CITE OR QUOTE
-------
1 Several general patterns were discernable with respect to which exposure metrics best
2 predicted lung-cancer mortality when comparing AICs for relative model fit. The data show that
3 lagging exposure by 10 years best predicts lung-cancer mortality compared to other lags. This
4 trend is seen across both the cumulative exposure without decay and the various half-life
5 cumulative exposure metrics where a 10-year lag of exposure best predicts lung-cancer mortality
6 for all cumulative exposure metrics compared to other lags; metrics with 15-year lags were
7 generally the next best in terms of fit. Another conclusion is that the models that included RTW
8 exposure metrics, regardless of half-life or lag, were less suitable than the models that employed
9 cumulative exposure and its variants.
10 Among the 40 exposure metric models that were evaluated, the exposure model with the
11 lowest AIC value was for cumulative exposure with a 10-year half life for decay and a 10-year
12 lag for cancer mortality latency and had a model p-va\ue of 0.0071 (see Table 5-12). This
13 multivariate model controlled for age, gender, race, and date of birth. This model estimated a
14 slope (beta) of 1.26 x 10 2 per fiber/cc-year based on a 365-day calendar year,41 and the
15 95th percentile upper bound on this parameter was 1.88 x 10 2 per fiber/cc-year. The/>-value for
16 the Libby Amphibole asbestos regression coefficient (slope) was <0.001, indicating that this
17 parameter was statistically significantly greater than zero. Table 5-13 shows the slopes and
18 confidence intervals for all retained metrics from Table 5-12.
19 According to the model results presented in Table 5-12, there were other exposure
20 metrics that predicted lung-cancer mortality and exhibited statistically significant effect
21 estimates. Several other metrics were considered to fit nearly as well as the model with the
22 smallest AIC since their AIC values were within two units of the exposure model with the lowest
23 AIC, a proximity that can be considered to be a range that cannot clearly differentiate between
24 models (Burnham and Anderson, 2002). As each of the other exposure metrics was based on a
25 different reorganization of the same exposure data, the different slopes are not directly
26 comparable, but all adequately fitting lagged models also produce statistically significant slopes
27 for the exposure-response relationship (p < 0.05). Of particular note are the results of the
28 cumulative exposure model, with a 10-year lag for latency, but without a decay function, since it
29 showed the lowest AIC among nondecay models.
41The two-sided 90% confidence interval is (6.00 x 10~3, 1.88 x 10~2); the two-sided 95% confidence interval is
(5.12 x ID'3, 2.00 x 1Q-2).
This document is a draft for review purposes only and does not constitute Agency policy.
5-91 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
Table 5-13. Lung-cancer mortality exposure metrics fits, slopes, and
confidence intervals for all retained metrics from Table 5-12. 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 l/i life
CE 5-yr !/2 life
CE15-yr!/2life
CE 20-yr !/2 life
CE 5-yr !/2 life
CE 10-yr !/2 life
CE
CE15-yr!/2life
CE 20-yr !/2 life
Lag
years
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
(Beta)
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)
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
CI = confidence interval
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.
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 5-12, the models are treated as having the same number of parameters since each model
represents the same exposures in a different way but with a single exposure parameter in the
regression models and are, therefore, equally penalized in the software's AIC calculation.
Because an argument can be made that exposure metrics that do not include a decay function
with their 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
This document is a draft for review purposes only and does not constitute Agency policy.
5-92 DRAFT—DO NOT CITE OR QUOTE
-------
1 of the CE models would be somewhat improved in terms of their comparison with the values in
2 Table 5-12 (AICs are generally penalized 2 units for each additional parameter).
3 Table 5-13 displays the lagged exposure-response models and metrics with adequate
4 model fit (p < 0.05) to the epidemiologic data that were further considered. The units of the
5 slopes are fiber/cc-year. These slopes and confidence intervals represent calendar year
6 continuous environmental exposure as described above and define the "Exposed Hazard Rate" in
7 the life-table procedure when multiplied by the exposure level (see Appendix G for details).
8
9 5.4.3.6.3. Summary of results of the analysis of the lung-cancer mortality in the sub-cohort
10 As presented in Table 5-13, the CE model with 10-year half life and lag provided an
11 adequate fit to the data (p < 0.05) and had the lowest AIC value. The cumulative exposure
12 model with a 10-year lag also yielded a statistically adequate fit to these data (p < 0.05), as did
13 several decay models with a 15-year lag. These results demonstrate reasonable uncertainty in the
14 metric of exposure such that no single exposure model can be definitively selected based on
15 goodness of fit alone, because IUR is based on the plausible upper bound of the effect estimate.
16 Based on the results from the lowest AIC multivariate model (i.e., cumulative exposure with a
17 10-year half life for decay and a 10-year lag for cancer mortality latency), the slope was
18 1.26 x icf2 per fiber/cc-year based on a 365-day calendar year, and the 95% upper bound on the
19 slope was 1.88 x icf2 per fiber/cc-year. This point estimate and 95% upper bound represent the
20 relative risk (including statistical uncertainty within exposure metric) of lung-cancer mortality
21 observed from exposure to Libby Amphibole asbestos fibers in the worker cohort for this
22 exposure metric. Issues related to uncertainty in the choice of exposure metric are described
23 further in the section on the derivation of the combined IUR of mesothelioma and lung cancer
24 (see Section 5.4.5.3).
25
26 5.4.3.6.4. Sensitivity analysis of the influence of high exposures in early 1960s on the model
27 fit in the sub-cohort
28 As discussed in Section 5.4.2.5, the comparison of model fit between various exposure
29 metrics is an empirical process and does not necessarily reflect either a specific biological or
30 other factor as an underlying cause for model fit. Although data do not exist to evaluate
31 biological bases for model fit, other potential factors can be explored where data allow. For
This document is a draft for review purposes only and does not constitute Agency policy.
5-93 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
o
J
4
5
6
7
10
11
12
13
14
15
16
17
18
example, because of concerns that very high (>100 fibers/cc) 8-hour TWA exposures during
1960-1963 (see Table 5-7) 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, there was one change in
the relative fit of 3r and 4* best fit decay models, but the observation that exposure metrics
including decay fit better than exposure metrics without decay was unchanged (see Table 5-14).
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 5-14. The revised data set DIG for the model used in IRIS IUR (U.S. EPA. 1988a) was
97.9.
Table 5-14. Sensitivity analysis of model fit comparison for different
exposure metrics and lung-cancer mortality associated with Libby
Amphibole asbestos. Estimated exposure intensities for all jobs during
1960-1963 were reduced by 50%.
Exposure Metric
CE 5-yr !/2 life
CE 5-yr !/2 life
CE 10-yr !/2 life
CE 10-yr !/2 life
CE15-yr!/2life
CE15-yr!/2life
CE 20-yr !/2 life
CE 20-yr !/2 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 5-11)
DIC
70.6
72.8
73.9
74
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
19
20
21
22
23
24
25
CE = Cumulative Exposure with exponential decay modeled with different half-lives; DIC = Deviance Information Criterion.
For modeling lung-cancer mortality on this revised data set, there was no difference in
the order of the relative fit between the same exposure models that fit the sub-cohort of workers
hired after 1959 and included the exposures as estimated by Amandus et al. (1987a) during
This document is a draft for review purposes only and does not constitute Agency policy.
5-94 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
o
6
4
5
6
7
8
9
10
11
1960-1963 (see Table 5-15). The models based on the revised data set fit marginally better
based on AIC.
Table 5-15. Sensitivity analysis of model fit comparison for different
exposure metrics and lung-cancer mortality associated with Libby
Amphibole asbestos, controlling for age, gender, race, and date of birth.
Estimated exposure intensities for all jobs during 1960-1963 were reduced by
50%. Lung cancer models presented include those with statistically significant
multivariate model p-va\ue and nonzero lag in exposure.
Exposure
metric
CE 10-yr !/2 life
CE 5-yr !/2 life
CE15-yr!/2life
CE 20-yr !/2 life
CE 5-yr !/2 life
CE 10-yr !/2 life
CE
CE15-yr!/2life
CE 20-yr !/2 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 5-13)
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
Multivariate
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
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
CE = Cumulative Exposure with or without exponential decay modeled with different half-lives.
AIC = Akaike Information Criterion.
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.
5.4.3.6.5. Additional analysis of the potential for confounding of lung cancer results by
smoking in the sub-cohort of workers hired after 1959
In the full cohort analysis, the proportional hazard assumption was not found to hold, and
it was possible that one of the reasons for this failure was the 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. By restricting the dates of hire in the sub-cohort, those
This document is a draft for review purposes only and does not constitute Agency policy.
5-95 DRAFT—DO NOT CITE OR QUOTE
-------
1 workers in the sub-cohort may be made more similar to each other in ways that would reduce the
2 potential for confounding by smoking and, in this sub-cohort, the proportional hazards
3 assumption was found to hold, thus statistically eliminating concern regarding confounding by
4 smoking (because smoking, in general, is known as a very strong confounder).
5
6 As an additional check on the potential for confounding, a new method was evaluated to
7 test for confounding by smoking in occupational cohorts that do not have data on smoking.
8 Confounding, which can bias observed results when there is an uncontrolled variable, which is
9 correlated with both the explanatory variable and the outcome variable, is a distinct concept from
10 effect-measure modification (i.e., synergy), which might reflect different observed effects of
11 exposure to Libby Amphibole asbestos among smokers as compared to nonsmokers. The extent
12 of effect-measure modification cannot be assessed without adequate data on smoking; however,
13 the issue is discussed in Section 5.4.6.
14 A method has been described by Richardson (2010) to determine if an identified
15 exposure relationship with lung cancer is confounded by unmeasured smoking in an occupational
16 cohort study. Richardson (2010) demonstrated that an exposure of interest (i.e., Libby
17 Amphibole asbestos) can be used to predict an outcome other than lung cancer such as chronic
18 obstructive pulmonary disease (COPD), which is known to be caused by smoking, but not
19 thought to be related to the exposure of concern.42 If a positive relationship is identified where
20 no causal association is suspected, this would suggest that smoking and the exposure metric
21 (Libby Amphibole asbestos) were positively correlated and that the identified exposure-response
22 relationship was, in fact, confounded by smoking. EPA implemented this methodology to model
23 the potential effects of Libby Amphibole asbestos on the risk of COPD mortality on the
24 sub-cohort of workers hired after 1959. Using the exposure metric defined as cumulative
25 exposure with a 10-year lag, the extended Cox proportional hazards model with time-varying
26 exposures estimated a slope (beta) for COPD of -0.056 per fiber/cc-year based on a 365-day
27 calendar year. The/>-value for the coefficient (slope) was 0.102, indicating that this parameter
28 was not statistically significantly different from zero. Using the exposure metric defined as
29 cumulative exposure with a 10-year half life for decay and a 10-year lag for cancer latency, the
42Richardson (2010) cited articles by Rushton (2007a, b) with possible associations between asbestos and COPD
which, if true, would have explained a positive association among the Libby workers cohort but should not detract
from the use of the Richardson method as applied to these Libby workers, where a negative association is found.
This document is a draft for review purposes only and does not constitute Agency policy.
5-96 DRAFT—DO NOT CITE OR QUOTE
-------
1 extended Cox proportional hazards model with time-varying exposures estimated a slope (beta)
2 of -0.135 per fiber/cc-year based on a 365-day calendar year. The/>-value for the coefficient
3 (slope) was 0.116, indicating that this parameter was not statistically significantly different from
4 zero.
5 Summarizing these findings, EPA used the method described by Richardson (2010) to
6 evaluate whether exposures to Libby Amphibole asbestos predicted mortality from COPD as an
7 indication of potential confounding by smoking and found a nonsignificant negative relationship,
8 which was inconsistent with confounding by smoking in the sub-cohort of workers hired after
9 1959.
10
11 5.4.4. Exposure Adjustments and Extrapolation Methods
12 The estimated exposures based on JEM and work histories are discussed in
13 Section 5.4.2.5. Note that all slopes presented with units of fiber/cc-year are for calendar year
14 and not for occupational year.
15
16 5.4.5. Inhalation Unit Risk (IUR) of Cancer Mortality
17 The derivation of the unit risk estimates, defined as the lifetime risk of mortality from
18 either mesothelioma or lung cancer from chronic inhalation of Libby Amphibole asbestos at a
19 concentration of 1 fiber/cc of air, is presented in the following subsections. Note that all slopes
20 are presented as per fiber/cc-year for a 365-day calendar year rather than for an occupational
21 year. Also, note that while the slopes are not adjusted for differences in breathing rates and the
22 number of hours of exposure in an occupational (8-hour) day as compared to a whole (24-hour)
23 day, the central risk and unit risk estimates do incorporate this adjustment.
24
25 5.4.5.1. Unit Risk Estimates for Mesothelioma Mortality
26 Computational details of the methodology and tables for deriving the unit risk for
27 mesothelioma mortality are presented in Appendix G. The modeling analysis presented above
28 showed that metrics including lag and half-life parameters provided the best empirical fit to the
29 Libby worker sub-cohort data. Although there is uncertainty in applying these models for
30 occupational mortality to estimation of risks for different exposure levels and time patterns (see
31 Section 5.4.6), following the recommendations of the Guidelines for Carcinogen Risk
This document is a draft for review purposes only and does not constitute Agency policy.
5-97 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
Assessment (U.S. EPA, 2005a), a linear low-dose extrapolation below the POD was used because
the mode of action for Libby Amphibole asbestos for mesothelioma is largely unknown. Using
the results of the cumulative exposure model with best-fitting lag and decay parameters, the
LECoi for the adult-only-exposures was determined to be 0.245 fibers/cc, which yielded an
adult-based unit risk of mesothelioma mortality of 0.041 (POD of 1% divided by the LECoi),
which when scaled by 70/54 to encompass the whole lifespan, yielded a lifetime unit risk of
0.053 per fibers/cc. The value of the risk corresponding to the measure of central tendency
involves ECoi rather than LECoi. The ECoi for the adult-only-exposures was determined to be
0.406 per fibers/cc, which when divided into a POD of 1%, yielded an adult-based central
estimate for mesothelioma mortality of 0.025, which when scaled by 70/54 to encompass the
whole lifespan, yielded a lifetime central estimate of 0.032 per fibers/cc.
The mesothelioma unit risks for model results presented in Table 5-11 and discussed in
Section 5.4.3.6.1 are presented in Table 5-16. All of the metrics in Table 5-16 are CE metrics
lagged 10-15 years (the fit of 20-year lag models was much worse since 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).
Table 5-16. Mesothelioma mortality exposure metrics unit risks
Exposure metric
CE - 5-yr !/2 life
CE - 5-yr !/2 life
CE - 10-yr l/i life
CE - 10-yr V* life
CE-15-yr!/2life
CE-15-yr!/2life
CE - 20-yr !/2 life
CE - 20-yr !/2 life
Lag years
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
Information weight
0.428
0.143
0.082
0.078
0.033
0.028
0.020
0.016
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
23
24
This document is a draft for review purposes only and does not constitute Agency policy.
5-98 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
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. Kopylev et
al. (2011) reviewed the literature on this underascertainment and developed 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
sub-cohort. As the number of peritoneal mesotheliomas is partially known in the Libby worker
sub-cohort, the appropriate adjustment factor for the sub-cohort is 1.39 [Kopylev et al. (2011),
TableS].
The adjusted mesothelioma central risk (based on the ECoi), corresponding to the best-fit
metric, was 0.044 (0.032 x 1.39) per fibers/cc, and adjusted mesothelioma mortality unit risk was
0.074(0.053 x 1.39) per fibers/cc. Mesothelioma mortality-adjusted unit risks are listed in
Table 5-17 along with their information weights.
Table 5-17. Adjusted for underascertainment unit risks for the sub-cohort
hired after 1959 corresponding to the different metrics
Exposure metric
CE - 5-yr !/2 life
CE - 5-yr !/2 life
CE - 10-yr !/2 life
CE - 10-yr V* life
CE-15-yr!/2life
CE-15-yr!/2life
CE - 20-yr V* life
CE - 20-yr V* life
Lag years
15
10
10
15
10
15
10
15
Information weight
0.428
0.143
0.082
0.078
0.033
0.028
0.020
0.016
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
25
26
This document is a draft for review purposes only and does not constitute Agency policy.
5-99 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.4.5.2. Unit Risk Estimates for Lung-Cancer mortality
2 Computational details of the methodology and tables for deriving the unit risk for
3 lung-cancer mortality are presented in Appendix G. Although there is uncertainty in applying
4 these models for occupational mortality to the estimation of risks for different exposure levels
5 and time patterns (see Section 5.4.6), following the recommendations of the Guidelines for
6 Carcinogen Risk Assessment (U.S. EPA, 2005a), a linear low-dose extrapolation below the POD
7 was used because the mode of action for Libby Amphibole asbestos for lung cancer is
8 undetermined. The nine exposure-response models retained from Table 5-12 (shown in
9 Table 5-13) all had reasonably similar goodness of fits. No single model stands out as clearly
10 statistically superior; however, there is a range of quality of fit within the set that could be
11 considered adequate. The lung-cancer mortality unit risks are shown in Table 5-18.
12 Using the results of the exposure model with the lowest AIC value (i.e., cumulative exposure
13 with a 10-year half life for decay and a 10-year lag for cancer latency) alone, the LECoi for the
14 adult-only-exposures was determined to be 0.333 fibers/cc, which yielded an adult-based unit
15 risk of lung-cancer mortality of 0.0300 (POD of 1% divided by the LECoi), which when scaled
16 by 70/54 to encompass the whole lifespan, yielded a lifetime unit risk of 0.0389 per fibers/cc.
17 The value of the risk that would correspond to the measure of central tendency involves ECoi
18 rather than LECoi. The ECoi for the adult-only exposures was determined to be 0.499 per
19 fibers/cc, which when divided into a POD of 1%, yielded an adult-based central estimate for
20 lung-cancer mortality of 0.0200, which when scaled by 70/54 to encompass the whole lifespan,
21 yielded a lifetime central estimate of 0.0260 per fibers/cc.
22 Using the results of the exposure model based on cumulative exposure with a 10-year lag
23 for cancer latency, the LECoi for the adult-only-exposures was determined to be 0.191 fibers/cc,
24 which yielded an adult-based unit risk of lung-cancer mortality of 0.0524 (POD of 1% divided
25 by the LECoi), which when scaled by 70/54 to encompass the whole lifespan, yielded a lifetime
26 unit risk of 0.0679 per fibers/cc. The ECoi for the adult-only exposures was determined to be
27 0.325 per fibers/cc, which when divided into a POD of 1%, yielded an adult-based central
28 estimate for lung-cancer mortality of 0.0308, which when scaled by 70/54 to encompass the
29 whole lifespan, yielded a lifetime central estimate of 0.0399 per fibers/cc.
30
31
This document is a draft for review purposes only and does not constitute Agency policy.
5-100 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
Table 5-18. 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
CE 10-yr l/i life
CE 5-yr !/2 life
CE15-yr!/2life
CE 20-yr !/2 life
CE 5-yr !/2 life
CE 10-yr !/2 life
CE
CE15-yr!/2life
CE 20-yr !/2 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 EC0i)
0.0260
0.0195
0.0300
0.0326
0.0167
0.0231
0.0399
0.0258
0.0280
Unit risk
(based on LEC0i)
0.0389
0.0293
0.0455
0.0501
0.0260
0.0375
0.0679
0.0434
0.0486
5
6
7
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
The resulting unit risks in Table 5-18 ranged from 0.0260 to 0.0679 fibers/cc. This
shows that the unit risk (i.e., 0.0389 per fibers/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 5.4.3.6.2), 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 AIC for
the CE models may reasonably be thought to be somewhat lower than through the standard
calculation of AIC. The CE metric with a 10-year lag does fit these data, 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.
Issues related to uncertainty in the choice of exposure metric are described in the section
on the derivation of the combined IUR of mesothelioma and lung cancer below.
This document is a draft for review purposes only and does not constitute Agency policy.
5-101 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.4.5.3. IVR Derivation for Combined Mesothelioma and Lung-Cancer Mortality
1 Before risks can be combined, it is important to understand several concepts that are
3 pertinent to the evaluation and comparison of the cancer-specific mortality unit risks that will be
4 combined. First, there is statistical uncertainty in the potency estimate within the
5 exposure-response model defined by each exposure metric. This within-metric uncertainty is
6 accounted for by the Bayesian credible interval around the potency estimates (slopes) for
7 mesothelioma mortality (see Table 5-11) and by the confidence interval around the potency
8 estimates (slopes) for lung-cancer mortality (see Table 5-13). Next, there is uncertainty in the
9 choice of metrics for developing an IUR (called cross-metric uncertainty, described below).
10 Finally, when unit risks corresponding to metrics are chosen accounting for uncertainty, these are
11 statistically combined into the IUR. Details are provided below.
12 For this current assessment, EPA obtained the best available demographic, exposure, and
13 vital status data from NIOSH. Subsequently, the best-fitting statistical models were identified,
14 which were then applied to derive central estimates of the lifetime combined mesothelioma and
15 lung-cancer mortality risk in the general population exposed to a continuous concentration of
16 1 fiber/cc of Libby Amphibole asbestos. Then, the individual exposure metric-specific risks
17 were calculated as the statistical (95%) upper confidence bounds on these central estimates. Use
18 of the upper confidence bound accounts for uncertainty in the effect estimate for each metric—
19 otherwise referred to as the within-metric uncertainty.
20 Another source of uncertainty is the choice of the appropriate exposure metric among a
21 set of results that appear to fit the data similarly well. This uncertainty is referred to as the
22 between-metric or cross-metric uncertainty. For the Libby worker cohort data, the best-fit
23 (lowest information criterion values) metrics lead to estimates of risks that are more like
24 mid-range estimates among the other metrics (see Tables 5-17 and 5-18) with sufficiently close
25 information criterion values, rather than upper bound estimates. While the lung cancer unit risk
26 computed from the model with the lowest AIC appears to be robust, Table 5-18 shows that there
27 is a range of possible unit risk values from the set of models with adequate fit (as measured by a
28 statistically significants-value for the exposure metric term) and similar goodness of fit.
29 Likewise, for mesothelioma mortality, among the models with adequate fit shown in Table 5-17,
30 there is a range of possible unit risk values.
This document is a draft for review purposes only and does not constitute Agency policy.
5-102 DRAFT—DO NOT CITE OR QUOTE
-------
1 The IUR should be a reasonable upper bound on the extra risk. As is clear from
2 Tables 5-17 and 5-18 in the preceding sections, the unit risks based on the metrics with the
3 lowest information criterion values provide a lower estimate of cancer mortality risk than some
4 other similarly fitting metrics. While the models with the lowest information criterion values
5 have the greatest statistical support, other models that yield higher unit risks are also statistically
6 plausible. This current assessment selected the upper bound unit risk among the plausible
7 exposure metrics (regardless of the small residual differences in quality of fit) to account for
8 cross-metric uncertainty. Because there were few metrics with unit risks higher than the best
9 fitting metric's unit risk for each cancer mortality endpoint, this method effectively selects the
10 highest unit risk among those considered for each cancer mortality endpoint.
11 Once the cancer-specific mortality unit risks are selected, the two are then combined.
12 Because each of the unit risks is itself an upper bound estimate, summing such upper bound
13 estimates across mesothelioma and lung-cancer mortality is likely to overstate the overall risk.
14 Therefore, following the recommendations of the Guidelines for Carcinogen Risk Assessment
15 (U.S. EPA, 2005a), a statistically appropriate upper bound on combined risk was derived in order
16 to gain an understanding of the overall risk of mortality resulting from mesothelioma and from
17 lung cancers. It is important to note that this estimate of overall potency describes the risk of
18 mortality from cancer at either of the considered sites and is not just the risk of both cancers
19 simultaneously.
20 Because the estimated risk for both mesothelioma and lung-cancer mortality was derived
21 using Poisson and Cox proportional hazards models, correspondingly, it follows from statistical
22 theory that each of these estimates of risk is approximately normally distributed. For
23 independent normal random variables, a standard deviation for a sum is easily derived from
24 individual standard deviations, which are estimated from confidence intervals: standard
25 deviation = (unit risk - central risk) H- Zo.95, where Zo.95 is a standard normal quantile equal
26 to 1.645. For normal random variables, the standard deviation of a sum is the square root of the
27 sum of the squares of individual standard deviations.
28 The upper bound among the mesothelioma mortality unit risks was 0.122 per fibers/cc.
29 The upper bound among the computed lung-cancer mortality unit risks was 0.0680 per fibers/cc.
30 The central estimate of risk was 0.075 for mesothelioma mortality per fibers/cc and 0.0399 per
31 fibers/cc for lung-cancer mortality (see Tables 5-17 and 5-18, respectively).
This document is a draft for review purposes only and does not constitute Agency policy.
5-103 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
o
6
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
In order to combine the unit risks, one first obtains an estimate of standard deviation of
the sum of the individual unit risks as
V[ [[(0.122 - 0.075) -H 1.645]2 + (0.068 - 0.0399) - 1.645 ]2 ] = 0.033 per fibers/cc (Eq. 5-9)
Then, the combined central estimate of risk of mortality from either mesothelioma or
lung cancer is 0.0399 + 0.075 = 0.115 per fibers/cc, and the combined IUR is
0.115 + 0.033 x 1.645 = 0.169 per fibers/cc.
Selecting the upper bound unit risk estimates for use in combining unit risks accounts for
many potential uncertainties. It accounts for uncertainty in the effect estimate (i.e., the
within-metric uncertainty) and the uncertainty attributable to the choice of exposure metric (i.e.,
the cross-metric uncertainty). The combined IUR from the best fitting mesothelioma and
lung-cancer mortality models (using two different model selection criteria) can be computed for
comparison with Tables 5-17 and 5-18, respectively, by the same steps as above, and the results
are shown in Table 5-19.
Table 5-19. Reasonable upper bound and lowest information criteria
estimates of central risks and unit risks, per fibers/cc, for mesothelioma
mortality, lung-cancer mortality, and the IUR for the combined mortality
risk from mesothelioma and lung cancer
Model
Reasonable upper bound3
Lowest information criteria13
Mesothelioma
Central
estimate
0.075
0.044
Unit risk
0.122
0.074
Lung cancer
Central
estimate
0.040
0.026
Unit risk
0.068
0.040
Combined mesothelioma and
lung cancer
Central
estimate
0.115
0.070
IUR
0.169
0.103
25
26
27
28
29
30
31
32
33
Tor mesothelioma, the selected model parameterized exposure as cumulative exposure with exponential decay
half-life of 5 years and a 15-year lag. For lung cancer, the selected model parameterized exposure as cumulative
exposure without decay and a 10-year lag.
bFor mesothelioma, the selected model parameterized exposure as cumulative exposure with exponential decay
half-life of 5 years and a 10 -year lag. For lung cancer, the selected model parameterized exposure as cumulative
exposure with exponential decay half-life of 10 years and a 10-year lag.
This document is a draft for review purposes only and does not constitute Agency policy.
5-104 DRAFT—DO NOT CITE OR QUOTE
-------
1 Compared to the combined IUR from the best fitting exposure models, the EPA's
2 selected combined IUR of mesothelioma and lung-cancer mortality accounts for both the
3 demonstrated cross-metric uncertainty as well as several additional potential uncertainties, which
4 could have resulted in underestimates of the mesothelioma and lung-cancer mortality risks from
5 the epidemiologic data. These additional uncertainties are discussed in Section 5.4.6. The IUR
6 value of 0.169 per fibers/cc accounts for important quantitative uncertainties in the selection of
7 the specific exposure metric that may have remained in an IUR that might otherwise have been
8 based on the best fitting exposure models alone.
9
10 5.4.5.3.1. Comparison with other published studies ofLibby workers cohort
11 For lung cancer, two alternative analytic approaches to the use of EPA's extended Cox
12 proportional hazards models could have been used for the calculation of a unit risk of
13 lung-cancer mortality. All of the choices are based on different analyses of the Libby worker
14 cohort; however, inclusion criteria differ among the analyses as does the length of mortality
15 follow-up. Each of the two approaches has two options to estimate the slope of the
16 exposure-response relationship in place of the regression slope estimated from the Cox
17 proportional hazards model and follow through with the same life-table procedure to calculate
18 the unit risk of lung-cancer mortality.
19 The first approach would be to use the published categorical results based on Sullivan
20 (2007). The first option in this approach was for EPA to estimate a slope to those categorical
21 data. The second option was to use the slope estimated in a published reanalysis of categorical
22 data of the Sullivan (2007) cohort by Berman and Crump (2008). The second approach would
23 be to use the published regression results of other researchers who modeled the underlying
24 continuous data. The first option in this approach was to use the slope estimated by Larson et al.
25 (201 Ob). The second option was to use the slope estimated by Moolgavkar et al. (2010).
26 For comparison purposes, the lung cancer unit risk from these alternatives is computed,
27 however, as all analyses are based upon different subsets of the Libby workers cohort and used
28 different analytic methods, the results are not necessarily interchangeable. Table 5-20
29 summarizes lung cancer risks derived from these studies.
30
This document is a draft for review purposes only and does not constitute Agency policy.
5-105 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
o
J
4
5
Table 5-20. 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)
Moolgavkar et al.
(2010V3
Herman and
Crump (2008)a
Larson et al.
(2010b)
Cohort definition
Hired post-1959
Exposures 1960-1982
Still alive post-1959
White males
Exposures 1960-1982
Still alive post-1959
White males
Exposures 1960-1982
Still alive post-1959
White males
Exposures 1960-1982
Full cohort
Exposures 1935-1993
Follow-up
2006
2001
2001
2001
2006
Lung
cancer
cases/TV
32/880
99/1,672
95/1,662
93/1,672
98/1,862
Slope per
fiber/cc-year
xlO~3
(calendar year)
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
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
"Sullivan (2007) and reanalysis of Sullivan (2007) state slightly different number of lung cancers. It is impossible to
reconcile these numbers from published information.
bReanalysis of Sullivan (2007).
The first alternative analytic approach to estimating the extra risk from a linear regression
of individual mortality data was to use a standard technique used 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 rate 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 who have 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
This document is a draft for review purposes only and does not constitute Agency policy.
5-106 DRAFT—DO NOT CITE OR QUOTE
-------
1 to the lung-cancer mortality rates in the lowest exposure category. The weights used for the
2 SRRs were the inverses of the variances. Midpoints of the exposure intervals were used, and for
3 the unbounded interval, the midpoint was assumed to be twice the starting point of that interval.
4 Using this approach, a regression coefficient of 4.2 x 1CT3 per fiber/cc-year
5 ([SE] = 7.7 x 10 4 per fiber/cc-year, p = 0.03) was obtained from the weighted linear regression
6 of the categorical SRR results. Because the data from Sullivan (2007) were already adjusted for
7 the length of an occupational year (240 days) to the length of a calendar year (365 days), only the
8 standard adjustment for inhaled air volume was performed. The concentration estimate obtained
9 using this regression modeling and the life-table analysis procedure was LECoi = 0.272 fibers/cc,
10 resulting in the lung cancer unit risk of 0.0368 per fibers/cc.
11 The Berman and Crump (2008) reanalysis was based on the Sullivan (2007) summary
12 results except they used a lag of 10 years (Sullivan, 2008, personal communication to Berman
13 and Crump). They fit the IRIS IUR (U.S. EPA. 1988a) lung cancer model to aggregate data
14 using an extra multiplicative parameter a (in this model, the relative risk at zero exposure is
15 estimated a rather than 1). In this model, the relative risk at zero exposure is a rather than 1
16 (unity). With a = 1, their model did not fit, and with a estimated, the fit was satisfactory.
17 Berman and Crump (2008) chose the central estimate of the slope from the fit with a estimated,
18 but constructed an "informal" 90% confidence interval by the union of two confidence intervals
19 (this upper bound is shown in see Table 5-20). This was done to address uncertainty in the
20 estimated parameter a, similar to what is done in this current assessment with estimated lag and
21 decay. Note also, that Berman and Crump (2008) also provide an UF to adjust for several
22 sources of uncertainty in exposures, resulting in an upper bound risk of 0.3162.
23 The second alternative analytic approach to estimating the extra risk of lung cancer from
24 a Cox regression with time-dependent covariates of individual mortality data was to use the
25 results published by Larson et al. (201 Ob), with cumulative exposure and a 20-year lag. This
26 analysis of lung-cancer mortality was based on the full cohort of 1,862 workers updated until
27 2006 and using the same model form as the current EPA analysis (the extended Cox proportional
28 hazards model). Larson et al. (201 Ob) reported a regression coefficient of 1.06 x io~3 per
This document is a draft for review purposes only and does not constitute Agency policy.
5-107 DRAFT—DO NOT CITE OR QUOTE
-------
1 fiber/cc-year (SE = 3.1 x 10 4 per fiber/cc-year,/> = 0.0006).43 EPA assumed that the cumulative
2 exposures reported by Larson et al. (201 Ob) were based on years of occupational exposure
3 (240 days per year) during a 365-day calendar year. In order to account for exposure on every
4 day of the year for a calculation of unit risk, an adjustment for exposures during the length of an
5 occupational year (240 days) to the length of an calendar year (365 days) and an adjustment for
6 the volume of inhaled air were performed to match EPA's analyses. The concentration estimate
7 obtained using the Larson et al. (201 Ob) regression modeling and the life-table analysis
8 procedure was LECoi = 1.26 fibers/cc, resulting in a lung cancer unit risk of 0.0103 per fibers/cc.
9 Moolgavkar et al. (2010) also used the Cox proportional hazards model with
10 time-dependent covariates for analysis of the Sullivan (2007) cohort with a 15-year lag. The
11 parameter in this study estimates 1.11 x ICT3 per fiber/cc-year (SE = 2.5 x 10~4 per
12 fiber/cc-year), which is very close to Larson et al. (201 Ob), and, therefore, the lung cancer unit
13 risk based on their analysis would be very close to Larson et al. (201 Ob). Comparison with
14 McDonald et al. (2004) is difficult, since their outcome is defined as respiratory cancer (ICD-9
15 160-165), which is more expansive than other researchers' definitions of the outcome as lung
16 cancer, and their sub-cohort of 406 white men employed before 1963—a time period when
17 exposure assessment was less reliable and more likely to include exposure-measurement error;
18 nonetheless, the parameter estimate resulting from the Poisson analysis by McDonald et al.
19 (2004) was 3.6 x 10~3 per fiber/cc-year.
20 EPA based their analyses on the exposures that occurred after 1959, while the Sullivan
21 (2007), Larson et al. (201 Ob), and Moolgavkar et al. (2010) analyses were based on the cohort
22 including those hired before 1960, and McDonald et al. (2004) included only workers hired
23 before 1964. As explained in detail in the discussion (see Section 5.4.6) on uncertainty in the
24 exposure assessment, there were only several measurements from the 1950s and one from 1942,
25 and most of the exposure estimation for the early years of the cohort's experience were based on
26 estimates of the ratio of dust to fibers estimated in the late 1960s and extrapolated backwards in
27 time for several decades. Moreover, 706 of the workers hired before 1960 (not necessarily
28 short-term) did not have an exposure measurement assigned to them at all, leading to much
29 larger measurement error. These limitations in the underlying exposure assessment for the years
43Note that EPA results based on the sub-cohort hired after 1959 were from the same model form but based on the
cumulative exposure with a 10-year lag and had a slope of 5.81 x 10"3 per fibers/cc-year (SE = 2.48 x 10"3 per
fiber-cc/year, p = 0.018).
This document is a draft for review purposes only and does not constitute Agency policy.
5-108 DRAFT—DO NOT CITE OR QUOTE
-------
1 prior to 1968 likely resulted in exposure measurement error that could have attenuated the
2 analytic regression results, thereby yielding a smaller effect estimate for the whole cohort
3 compared to the sub-cohort hired after 1959. It appears the differences in results are mostly
4 attributable to the time periods of analysis and corresponding to the time period measurement
5 errors rather than the analytic approach. The small discrepancy between observed lung cancer
6 deaths between this current assessment and Larson et al. (201 Ob), described in Section 4.1.1.1, is
7 unlikely to play a role in the difference between risk estimates. Moreover, for the sub-cohort
8 hired after 1959, all deaths are included in the Larson et al. (201 Ob) lung cancer-counting rules.
9 None of the approaches used by McDonald et al. (2004), Sullivan (2007), nor Larson et
10 al. (201 Ob) could have been appropriately used for the unit risk of mesothelioma as they are not
11 based on absolute risk metrics of association, and the current assessment considered the relevant
12 metric of association to be the absolute risk. Berman and Crump (2008) did not evaluate risk of
13 mesothelioma. Moolgavkar et al. (2010) used an absolute risk model for mesothelioma. These
14 results are summarized in Table 5-21. The upper bound results for the full cohort presented by
15 Moolgavkar et al. (2010) are about 80% of the IRIS IUR (U.S. EPA, 1988a) estimate of
16 mesothelioma slope factor in a similar RTW-type metric, leading to an approximately 80%
17 estimate of the mesothelioma unit risk, as dependence is linear in the mesothelioma slope factor
18 (see Eq. 5-5). This is very close to this current assessment's estimate based on the sub-cohort,
19 which is also about 80% of the IRIS IUR (U.S. EPA, 1988a) estimate of mesothelioma risk.
20 Duration of employment is the best metric for the full cohort, and it does not support
21 exposure-response estimation.
22
23 5.4.5.4. Applications of the Combined Mesothelioma and Lung-cancer mortality IUR to
24 Partial Lifetime Environmental Exposure Scenarios
25 In the application of the IUR, scenarios other than lifetime environmental exposure are often of
26 interest to risk assessors. The life-table analysis in the (general) IRIS IUR for asbestos (U.S.
27 EPA, 1988a) predicts risk increases as the age of the first exposure decreases. The authors of
28 that analysis recommended the life-tables in that analysis be consulted when assessing partial
29 lifetime exposures (U.S. EPA, 1986a). In 2008, EPA (Office of Solid Waste and Emergency
30 Response) provided guidance for calculating risk estimates for less-than-lifetime exposures
31 based on the source life-table analysis (U.S. EPA, 2008). The age-at-onset of exposure and
This document is a draft for review purposes only and does not constitute Agency policy.
5-109 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
o
J
4
5
Table 5-21. Mesothelioma regression results from different analyses of
cumulative exposure in the cohort of workers in Libby, MT. 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)
Moolgavkar et al.
(2010V
Larson et al. (201 Ob)
Herman and Crump
(2008)a
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 fibers/cc)
Upper Bound = 0.12
Central = 0.08
No estimates of absolute
risk
Upper Bound -0.1 3
Central ~ 0.08
No estimates of absolute
risk
No estimates provided
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
"Reanalysis of Sullivan (2007).
duration-dependent unit risks reflect the influence of the time-cubed function in the
mesothelioma model (see Eq. 5-5) (U.S. EPA, 2008, 1986a) used in the 1986 assessment.
Because the time-cubed mesothelioma model, or parameterization of exposure metrics, did not
fit the data for mesothelioma mortality from exposure to the Libby Amphibole asbestos, the
approach to estimating risk of partial life exposure recommended by EPA when applying the
general IRIS IUR for asbestos (U.S. EPA, 2008) is not appropriate when applying the Libby
Amphibole asbestos-specific IUR.
Thus, this current assessment recommends that estimates of the risks of less-than-lifetime
exposures be computed by simple calculations of average lifetime exposure concentration
multiplied by the IUR. This recommendation is consistent with standard Superfund guidance
where exposures are estimated and averaged across a lifetime exposure, and the IUR is simply
applied to calculate excess cancer risk (U.S. EPA, 2008, 2001b).
This document is a draft for review purposes only and does not constitute Agency policy.
5-110 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.4.6. Uncertainties in the Cancer Risk Values
2 It is important to consider uncertainties in the derivation of the mesothelioma and
3 lung-cancer mortality risks in this assessment in the context of uncertainties in animal-based
4 health assessments. This assessment does not involve extrapolation from high doses in animals
5 to low doses in humans. This assessment is based on a well-documented and well-studied cohort
6 of workers with adequate years of follow-up to evaluate mesothelioma and lung-cancer mortality
7 risks with PODs within the range of the data. The discussions below explore uncertainty in the
8 derivation of the IUR in order to provide a comprehensive and transparent context for the
9 resulting cancer mortality risk estimates.
10
11 5.4.6.1. Sources of Uncertainty
12 Sources of uncertainty in this assessment include
13
14
15 1) Uncertainty in low-dose extrapolation,
16 2) Uncertainty in exposure assessment, including analytical measurements
17 uncertainty,
18 3) Uncertainty in model form,
19 4) Uncertainty in selection of exposure metric,
20 5) Uncertainty in assessing mortality corresponding to the cancer endpoints,
21 6) Uncertainty in control of potential confounding in modeling lung-cancer
22 mortality,
23 7) Uncertainty due to potential effect modification,
24 8) Uncertainty due to length of follow-up,
25 9) Uncertainty in use of life-tables to calculate cancer mortality unit risks,
26 10) Uncertainty in combining of mortality risks to derive a composite cancer
27 mortality IUR,
28 11) Uncertainty due to extrapolation of findings in adults to children.
29
30
This document is a draft for review purposes only and does not constitute Agency policy.
5-111 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.4.6.1.1. Uncertainty in low-dose extrapolation
1 A common source of uncertainty in quantitative cancer risk assessments generally derives
3 from extrapolating from high doses in animals to low doses in humans. Compared to
4 assessments based on animal data, the uncertainty from low-dose extrapolation in this
5 assessment employing occupational epidemiology data is considered to be somewhat reduced for
6 the following reasons. The NIOSH worker cohort developed by Sullivan (2007) includes
7 410 workers employed less than 1 year among the 880 workers hired on or after January 1, 1960.
8 Although short-term workers, on average, experience a mean exposure intensity per day worked
9 greater than workers employed more than a year (Sullivan, 2007), the cohort nevertheless
10 includes many short-term workers with relatively low cumulative occupational exposures.
11 Further, inclusion of salaried workers in the NIOSH cohort (Sullivan, 2007) adds many workers
12 with lower workplace exposure. Thus, while occupational exposure concentrations may be
13 generally higher than typical ongoing environmental concentrations, the low-dose exposures in
14 this occupational database may be representative of nonoccupational exposures.
15 While many occupational epidemiology studies are based on relatively high exposure
16 levels that are beyond the range of common environmental exposures, many in the Libby
17 workers cohort experienced exposures that were near or below the PODs derived from the
18 life-table analysis. The POD for the selected lung-cancer mortality exposure metric was
19 0.191 fibers/cc. The POD for the selected mesothelioma mortality exposure metric was
20 0.106 fibers/cc. Among the workers hired after 1959 who had at least 1 year of occupational
21 exposure (n = 470; 20 lung cancer deaths), there were 19 (4%) with average occupational
22 exposure concentrations of less than 0.3 fibers/cc, including 1 lung cancer death (5%).
23 Although data might have been modeled down to a very low cumulative exposure level,
24 the Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a) recommend defining a POD
25 for low-dose extrapolation in order to increase the stability of the IUR estimate at lower
26 exposures, where fewer cancers might be expected. Thus, the uncertainty associated with
27 low-dose extrapolation is somewhat mitigated since the linear extrapolations from the dose
28 associated with the POD from the life-table analyses of each cancer endpoint were encompassed
29 within the observed data range. Nonetheless, some uncertainty remains in the extrapolation from
30 occupational exposures to lower environmental exposures when using a POD.
31
This document is a draft for review purposes only and does not constitute Agency policy.
5-112 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.4.6.1.2. Uncertainty in exposure assessment
1 Accurate exposure assessment is generally considered to be a major challenge for
3 occupational epidemiologic studies and is a challenge that is well recognized by the NIOSH
4 investigators (Amandus et al., 1987a). As stated previously in Section 5.4.3.3, while it is
5 generally true that the use of more data is an advantage in statistical analyses because it allows
6 for the computation of more statistically precise effect estimates, this advantage in precision may
7 be offset by a negative impact on the accuracy of the effect estimate if an increase in sample size
8 is accompanied by greater exposure misclassification or other biases. Therefore, EPA decided to
9 base this Libby Amphibole asbestos-specific human health risk assessment upon the mortality
10 experience of workers hired on or after January 1, 1960. EPA's use of the sub-cohort analysis is
11 based on the belief that it is important to accurately estimate the true underlying
12 exposure-response relationships by relying on the most accurate exposure data. The use of this
13 sub-cohort greatly reduces the uncertainty in exposure error compared to evaluations based on
14 the entire cohort. More specifically,
15
16
17 a) Job category and department codes were completely unknown for 706 of the
18 991 workers' jobs from 1935 to 1959 (71% of the cohort for this time period). These
19 workers were assigned the same exposure concentration (66.5 fibers/cc) for all years
20 without this information. Examination of the post-1959 cohort removes this
21 significant source of exposure misclassification (only 9 of 880 sub-cohort workers did
22 not have department code and job category information).
23 b) Using the more recently hired cohort minimizes the uncertainty in estimated worker
24 exposures based on the JEM, which was informed by air sampling data available in
25 1956 and later years. Although there are still uncertainties in the task-specific
26 exposure estimates from 1960-1967, uncertainty in the assessment of earlier
27 exposure levels is considerably greater.
28 c) Exposure measurements were collected from the area samples and represented
29 exposures for all the workers with the same job code. Statistically, this causes
30 Berkson measurement error effect, which is described later in this section.
31
32 As the EPA exposure-response modeling for mesothelioma and lung-cancer mortality is
33 based on the post-1959 sub-cohort, the remaining discussion of uncertainty in exposure
34 measurement will address these data.
35
This document is a draft for review purposes only and does not constitute Agency policy.
5-113 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.4.6.1.2.1. Sources of uncertainty in job history information
1 Worker exposures for the EPA exposure-response modeling were calculated based on job
3 histories and the JEM from 1960 through 1982 (see Figure 5-3). Overall, there is little
4 uncertainty in the job history information. Regarding exposure estimation for the occupational
5 cohort, the NIOSH investigators (Amandus et al., 1987a) conducted a detailed retrospective
6 exposure assessment to estimate the individual worker exposures. NIOSH used extensive
7 occupational exposure data to construct the time-specific JEM, spanning decades (Amandus et
8 al., 1987a). These data were reabstracted from the workers' employment records for quality
9 assurance (Sullivan, 2007). NIOSH records on work histories and job-specific exposure
10 extended from the 1930s through May 1982. But, the vermiculite mining and milling operation
11 continued on for several years, and some workers were retained through 1993 for plant close-out
12 activities. Only 148 members of the post-1959 cohort (n = 880) were employed as of the May
13 1982 employment records when the cohort was enumerated by NIOSH (Sullivan, 2007).
14 Because exposure concentrations in 1982 (see Table 5-7) were generally below 1 fiber/cc with
15 only two locations having concentrations of 1.2 fibers/cc, it is unlikely that these workers'
16 exposures were significantly underestimated.
17
18 Sources of uncertainty in exposure intensity for the identified location operations
19 The available exposure data that inform the JEM include over 4,000 air samples, the
20 majority of which were collected after 1967 (see Table 4-1). All of the job location exposure
21 estimates (see Table 5-7) from 1968-1982 were directly informed from air samples collected on
22 membrane filters and analyzed for fibers by PCM. The availability of site- and task-specific air
23 samples for these years provides a good basis for the exposure estimates. However, there are
24 some uncertainties in estimating asbestos exposures using air samples analyzed by PCM.
25
26
27 1) PCM analysis does not determine the mineral or chemical make-up of the fiber: The
28 PCM method defines and counts fibers based on the size (aspect ratio and length) of the
29 particle without regard for the material that makes up the fiber being viewed. The PCM
30 method was developed for use in occupational environments where asbestos was present,
31 and the nature of the fibers should be further evaluated to confirm the fibers viewed
32 under PCM are asbestos. McGill University researchers evaluated the fibers collected on
33 membrane filters in the early 1980s and confirmed the presence of asbestos fibers in the
34 tremolite-actinolite solution series consistent with the Libby Amphibole asbestos
This document is a draft for review purposes only and does not constitute Agency policy.
5-114 DRAFT—DO NOT CITE OR QUOTE
-------
1 (McDonald et al., 1986a). NIOSH researchers confirmed the presence of tremolite
2 asbestos in bulk dust samples but not in air samples from the facility (Amandus et al.,
3 1987a). Although less specific to fibers, 60-80% of the airborne dust in the mills in 1968
4 was tremolite, further supporting the presence of asbestos in the air (based on State of
5 Montana air sampling, and X-ray diffraction analysis by the Public Health Service [PHS
6 correspondence, October 17, 1968]). However, although the presence of mineral fibers in
7 the actinolite-tremolite series was confirmed in the work environment, it is possible that
8 there were also fibers counted by PCM from other materials (such as textiles from clothes
9 and packaging materials). Therefore, it is unknown from these data what proportion of
10 the counted PCM fibers was mineralogically asbestos, or other materials present in the
11 workplace.
12 2) PCM defines fibers as particles with an aspect ratio greater than 3:1: There is an ongoing
13 debate in the literature on asbestos toxicity regarding the influence of aspect ratio on
14 relative toxicity. Specifically, in mining environments, it has been speculated that a
15 larger proportion of low aspect ratio fibers from mineral dusts may significantly impact
16 the apparent cancer potency of the measured PCM fibers in those environments (Berman,
17 2010; U.S. EPA, 1988a). There are few data available to understand fiber morphology
18 and fiber aspect ratios in the Libby cohort working environment. Considering the
19 post-1959 cohort, PCM fiber size distribution and aspect ratio data only exist for a set of
20 eight air samples (599 fibers) collected from the wet mill and screening operations and
21 analyzed by the NIOSH researchers (Amandus et al., 1987a). For these air samples, over
22 96% of the fibers viewed by PCM had an aspect ratio greater than 10:1 (Table 4-2)
23 (Amandus et al., 1987a)44' However, because these samples were provided by the
24 company in the early 1980s, they do not represent conditions in the old wet mill or dry
25 mill operations, which were significantly dustier environments (Amandus et al., 1987a).
26 It is possible that prior to industrial hygiene (IH) modifications in 1974, the dry and old
27 wet mills generated proportionally more mineral dusts than screening and new wet mill
28 operations after IH modifications. No data are available for the mining environment,
29 which would also be expected to generate a range of mineral dusts. Therefore, there is a
30 significant uncertainty about the size and aspect ratio of fibers included in PCM fiber
31 counts for the maj ority of the post-1960 workers cohort.
32 3) The resolution of visible PCM fibers: Current analytical instruments used for PCM
33 analysis have resulted in a standardization of minimum fiber width considered visible by
34 PCM between 0.2 and 0.25 |im. Historical PCM analysis (1960s and early 1970s)
35 generally had less resolution, and fibers with minimum widths of 0.4 or 0.44 jim were
36 considered visible by PCM (Amandus et al., 1987a: Rendall and Skikne, 1980).
37 McDonald et al. (1986a) compared fibers viewed by PCM and TEM and estimated that
38 approximately 1/3 of the total fibers could be viewed by the optical microscope. Because
39 38% of the fibers were <5 jim in length, this implies approximately 30% were not
40 viewable by optical microscopy for other reasons, such as width. However, it is
41 unknown what proportion of that 30% would be viewed with the minimum width
44Although Amandus et al. (1987a) report the sizing of PCM fibers, the details of the methodology are not given
regarding how these fibers were identified. No method is cited, and it is unclear if the sizing was done by PCM or
TEM for fibers in the reported size categories.
This document is a draft for review purposes only and does not constitute Agency policy.
5-115 DRAFT—DO NOT CITE OR QUOTE
-------
1 resolution of 0.25 jim for later optical microscopy. It is likely that early PCM counts
2 were underestimated relative to the later data for the cohort but by less than a factor of 2.
o
J
4
5 Prior to 1968, no air sampling data were available for 23 of the 25 job location operations
6 (see Table 4-2), and the exposure estimates were extrapolated from later air sampling data.
7 Amandus et al. (1987a) recognized there is significant uncertainty in the extrapolation of
8 available air sampling data to previous time periods. The researchers took into account major
9 changes in operations and interviewed employees in the early 1980s regarding previous years of
10 operation. The assumptions used to make these extrapolations are clearly stated for each of the
11 plant operations. For four operations, high and low estimates of pre-1968 exposures were
12 provided based on different sets of exposure assumptions (see Table 5-7). For ore loading, there
13 were negligible differences in the exposure estimates for the period from 1960-1967 (10.7
14 versus 9 fibers/cc). For drilling, the river dock, and the bagging plant, there were 3.4-, 2.6-, and
15 2.8-fold differences, respectively, between the high and low estimates of exposure between 1960
16 and 1968.
17 Dry mill exposures between 1960 and 1968 were informed by air sampling for total dust
18 collected in the dry mill facility from 1956-1969 (where total dust was collected by midget
19 impingers). Amandus et al. (1987a) derived a conversion factor of 4.0 fibers/cc per mppcf to
20 apply to the two location operations in the dry mill during these years. There was a range of
21 conversion factors considered for the dry mill depending on how the dust and fiber air samples
22 (PCM) were grouped and averaged (1.2 to 11.5 fibers/cc per mppcf). A subset of dust and fiber
23 samples available over the same time period (1967-1968) resulted in a ratio of 8.0 fibers/cc per
24 mppcf. In contrast, a ratio of 1.9 fibers/cc resulted when total dust samples from 1969 were
25 compared with fiber samples from 1970. However, both of these subsets had limited numbers of
26 samples available. Therefore, the conversion factor of 4.0 fibers/cc per mppcf was selected
27 based on using the maximum samples available over a time period when the dry mill exposures
28 were considered similar: dust samples (1965-1969) and fiber samples (1967-1971).
29
30 5.4.6.1.2.2. Sources of uncertainty in the calculation of the job-exposure matrix (JEM)
31 The exposures in the JEM (see Figure 5-3) were calculated from the exposure intensities
32 of the various task-specific exposure intensities shown by job location operation (see Table 5-7).
This document is a draft for review purposes only and does not constitute Agency policy.
5-116 DRAFT—DO NOT CITE OR QUOTE
-------
1 The uncertainties in the exposure intensity for the job location operations will impact the JEM.
2 Additionally, for each of the job categories in the JEM, NIOSH researchers defined which tasks
3 (job location operations) were conducted and for what proportion of the work day. A TWA
4 exposure for each job category across time was calculated based upon these assumptions and the
5 task-specific exposure estimates. There is a measure of uncertainty in these assumptions for
6 each job category. Additionally, there is inter-individual variation within the job categories.
7 These uncertainties are common to exposure reconstruction for epidemiological cohorts.
8
9 5.4.6.1.2.3. Uncertainty in the exposure metric
10 The PCM measurement is the available exposure metric for analysis of Libby worker
11 cohort at this time. Currently, there is no optimal choice of the best dose metric for asbestos, in
12 general, and, in particular, for Libby Amphibole asbestos, even if a TEM-based dose-response
13 JEM was available. Uncertainties related to PCM analytical method are discussed in Section 2.
14 Briefly, PCM cannot distinguish between asbestos and nonasbestos material or differentiate
15 between specific types of asbestos. Further, due to limitations of this methodology, PCM does
16 not take into account fibers smaller than 5 jim in length.
17
18 5.4.6.1.2.4. Evaluation of the effects of uncertainties in exposure measurement
19 An understanding of the effects of exposure measurement error on the risks estimated
20 from epidemiologic analyses is important to place these possible exposure measurement errors in
21 context. The effect of exposure measurement error on estimates of the risk of mesothelioma or
22 lung-cancer mortality attributable to exposure depends upon the degree to which that error may
23 be related to the likelihood of mesothelioma or lung-cancer mortality. Exposure measurement
24 error that is similar in pattern among workers who died of lung cancer to exposure measurement
25 error in people who did not die of lung cancer is a nondifferential exposure measurement error.
26 Differential exposure measurement error that is associated with the outcome can cause bias in an
27 effect estimate towards or away from the null, while nondifferential exposure error typically
28 results in bias towards the null (Rothman and Greenland, 1998). From the above evaluation of
29 uncertainties, there is no indication that the uncertainties in job history information, exposure
30 estimates for specific tasks, or calculation of the JEM would be differential based on the cancer
31 health outcome data. Therefore, these uncertainties are considered nondifferential, and the
This document is a draft for review purposes only and does not constitute Agency policy.
5-117 DRAFT—DO NOT CITE OR QUOTE
-------
1 general result is likely to be an attenuation in risk estimates towards the null (that is, the addition
2 of random noise to a clear signal tends to reduce the clarity of the observed signal and the
3 avoidance of random noise—here from poor quality exposure measurements—results in a
4 stronger observed signal).
5 Generally speaking, if the exposure concentrations estimated by NIOSH were
6 systematically too high, then the associated risks of exposure estimated in the regression analysis
7 would be low since the same actual risk would be spread across a larger magnitude of exposure.
8 Similarly, if the exposure concentrations estimated by NIOSH were systematically too low, then
9 the associated risks of exposure estimated in the regression analysis would be too high. From the
10 above evaluation, the majority of the sources of uncertainty are not systematic. There are a few
11 areas of uncertainty that may be classified as biased:
12
13
14 1) High- and low-exposure estimates for four job location operations were provided
15 between 1960 and 1967. Amandus et al. (1987a) chose the high estimates of
16 exposure for these job location operations when calculating the JEM. Therefore,
17 there will be a bias towards the high end for the job categories informed by these
18 data. There was a 1.1 - to 3.4-fold difference between the high and low estimates.
19 This difference will be less pronounced where these exposure concentrations are
20 averaged with other job location operations in the JEM and across multiple jobs for
21 the majority of the workers (see Figure 5-3).
22 2) Current PCM analysis would count more fibers relative to early PCM methods based
23 on minimum fiber width resolution. For example, Amandus et al. (1987a) used a
24 minimum width cutoff of 0.44 in their review of PCM fibers in the 1980s, which may
25 have resulted in as much as a twofold underestimate compared to current PCM
26 methods with a width resolution of 0.25 |im. Additionally, as PCM methodology has
27 developed over time, it is unknown when PCM results from company records would
28 be considered relatively standard to a minimum width resolution between 0.2 and
29 0.25 |im. Also, prior to standardization of PCM to 0.25-|im minimum width, there
30 was inter-laboratory variability as well. Therefore, the size distribution of PCM
31 fibers (e.g., minimum width) reported in the JEM may have changed over time.
32 Although theoretically a systematic bias, given the years for which PCM data are
33 available, this is likely an insignificant effect.
34 3) Asbestos was a contaminant of vermiculite that was the primary object of production.
35 Mine, old dry mill, and wet mill ambient air may have contained material other than
36 asbestos that could have contributed to PCM fiber count. The exposures in the old
37 dry and wet mills and mine location may have included a greater proportion of dust to
38 fibers than tasks using the ore and refined vermiculite after the new wet mill became
39 operational. It is possible there is a systematic over-count of fibers in the dusty
This document is a draft for review purposes only and does not constitute Agency policy.
5-118 DRAFT—DO NOT CITE OR QUOTE
-------
1 environment due to interference from mineral fragments. This likely impacts the
2 exposure intensity for 23 of 25 job location operations within the mine and old dry
3 mill. Estimated exposures from job categories that include these operations may be
4 biased upwards.
5
6
7 Nondifferential measurement error in a continuous exposure can be of the classical or
8 Berkson type and typically arises in environmental and occupational settings as a mixture of the
9 two forms (Zeger et al., 2000). Classical measurement error occurs when true exposures are
10 measured with additive error (Carroll et al., 2006) and the average of many replicate
11 measurements, conditional on the true value, equals the true exposure (Armstrong, 1998). This
12 error is statistically independent of the true exposure that is being measured and attenuates true
13 linear effects of exposure, resulting in effect estimates in epidemiologic studies that are biased
14 towards the null (Heid et al., 2004; Zeger et al., 2000; Armstrong, 1998). Such errors occur
15 when the mean values of multiple local air samples are used.
16 Berkson measurement error is independent of the surrogate measure of exposure (Heid et
17 al., 2004; Berkson, 1950) and is present when the average of individuals' true exposures,
18 conditional on the assigned measurement, equals the assigned measurement. Berkson
19 measurement error can arise from the use of local area mean sampled exposures to represent the
20 individual exposures of people in that area—even when the estimated area mean is equal to the
21 true underlying mean (i.e., no classical measurement error). Examples of random variability in
22 personal behavior that may produce Berkson measurement error in personal exposure estimates
23 include the volume of air breathed per day among the workers and the effectiveness of an
24 individual's nasal filtration at removing contaminants. In general, Berkson measurement error is
25 not thought to bias effect estimates but rather increases the standard errors of effect estimates
26 (Zeger et al., 2000). However, some epidemiologic studies have suggested that Berkson
27 measurement error can produce a quantitatively small bias towards the null in some analyses
28 (Bateson and Wright 2010; Kim et al., 2006; Reeves etal., 1998; Burr, 1988).Uncertainties in
29 the levels and time course of asbestos exposure for the Libby workers also adds uncertainty to
30 the evaluation of the relative fit of different exposure metrics.
31
This document is a draft for review purposes only and does not constitute Agency policy.
5-119 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.4.6.1.2.5. Exposure to other kinds of asbestos and residential exposure
1 Another source of uncertainty in the estimation of exposures in the Libby workers cohort
3 is the potential contribution of nonoccupational or residential exposures as well as exposures to
4 other kinds of asbestos in employment before or after working in Libby.
5 Many of the workers resided in Libby, MT, before and/or after their employment at the
6 mining and milling facilities ended. The vermiculite from the mine had been used at numerous
7 sites around the town, including baseball fields around the expansion plant and as filler in
8 gardens (U.S. EPA, 2010a, 200la). Exposure to asbestos could have occurred among individuals
9 outside of the workplace, particularly through activities with the potential of stirring up of dirt or
10 other materials that had been mixed with the vermiculite (Weis, 200la). The results of
11 community sampling indicated that even 10 years after mill operations ceased during some
12 activities, asbestos fiber concentrations in the air could exceed OSHA standards established for
13 the protection of workers (Weis, 200 la).
14 Therefore, the workers' actual personal exposures as the sum of occupational and
15 nonoccupational exposures are likely to have been underestimated by the use of estimated
16 Libby-related occupational exposure alone. The difficulty stems from the lack of data on
17 residential exposures and lack of information on pre- and postemployment residence of the
18 Libby workers. Nonoccupational exposures were likely to have been smaller in magnitude than
19 the occupational exposures, but workers may have lived in and around Libby, MT, for many
20 more years than they were exposed occupationally. The impact of residential exposure could be
21 more prominent for workers with lower occupational exposure who resided in Libby for a long
22 time. Whitehouse et al. (2008) has reported several cases of mesothelioma among residents of
23 the Libby, MT region who were not occupationally exposed. However, since the report by
24 Whitehouse et al. (2008) details only the cases and does not define or enumerate the population
25 from which those cases were derived, computed relative risks from nonoccupational exposures
26 were not available. ATSDR (2000) reported higher relative risks of mesothelioma among the
27 population of Libby, MT, including former workers residing in Libby, but did not provide
28 relative risk for nonoccupational exposure. Instead, the ATSDR report on mortality (2000)
29 grouped cases among the former workers with nonoccupationally exposed cases. Therefore, it is
30 not clear what the magnitude of the contribution of workers' nonoccupational exposures was to
31 their overall risk.
This document is a draft for review purposes only and does not constitute Agency policy.
5-120 DRAFT—DO NOT CITE OR QUOTE
-------
1 Some of the occupational workers with lower exposures, such as short-term workers, may
2 have either been high school or college students working during the summer or may have been
3 transient workers who may not have stayed for a long time in Libby. Sullivan (2007) analyzed
4 differences between short- and long-term workers and reported little difference between the
5 groups except for age at hire. As the short-term workers were younger on average, this
6 supported the suggestions that some of the short-term workers may have been college students
7 working during the summer. This population of short-term workers is not well defined;
8 however, it is possible that short-term transient workers could potentially have been exposed to
9 other kinds of asbestos or other lung carcinogens in their non-Libby occupational career, which
10 might have affected their pre- and post-Libby risk profile for asbestos exposure. While their
11 occupational histories other than working in Libby are unknown, it is very unlikely that they
12 include exposures of the magnitude that were encountered in the Libby mine and mill. The
13 impact of these uncertainties on regression slopes is difficult to evaluate. However the slope
14 may be somewhat underestimated as an observed increase in risk would be attributed to a larger
15 exposure differential than might have been present due to the addition of nonoccupational
16 exposures. There will also be a downward bias from random exposure measurement error with
17 lower occupational exposure affected disproportionately; however, the magnitude of this bias
18 would be expected to be small.
19
20 5.4.6.1.2.6. Conclusion regarding uncertainty in exposure assessment
21 Overall, there are likely to be multiple sources of uncertainty attributable to exposure
22 measurement error. It is possible that systematic error may have been introduced into the
23 exposure intensities assigned to several of the job location operations discussed above. In each
24 case, these errors in estimating exposures were overestimates. The magnitude of the potential
25 overestimates of drilling and dry and old wet mill exposures is uncertain. The dust-to-fiber
26 conversion ratio applied to the dry mill during 1960-1967 could be an over or underestimate by
27 as much as twofold. Random error in the measurement of dust or fibers would likely have
28 produced an underestimation of risk. There is no known bias in the assumptions to extrapolate
29 exposure to pre-1968 location operations outside of the dry mill, and random bias would also
30 likely have produced an underestimation of risk.
31
This document is a draft for review purposes only and does not constitute Agency policy.
5-121 DRAFT—DO NOT CITE OR QUOTE
-------
1 5.4.6.1.3. Uncertainty in model form
1 For mesothelioma mortality, the Poisson regression model is commonly used for rare
3 outcomes and has been applied by McDonald et al. (2004) and Moolgavkar et al. (2010) to
4 model mesothelioma risk in the Libby worker cohort. For lung-cancer mortality, the Cox
5 proportional hazards model is a well-established method that is commonly used in cohort studies,
6 including by Larson et al. (201 Ob) and Moolgavkar et al. (2010) for the Libby worker cohort,
7 because this type of survival analysis takes into account differences in follow-up time among the
8 cohort. Larson et al. (201 Ob) conducted Poisson regression analyses and reported that their lung
9 cancer results using this different model form were similar to those from their extended Cox
10 proportional hazards models, but those results were not shown.
11 Both of these model forms allow for the evaluation and control of important potential
12 confounding factors such as age, sex, and race, and for the modeling of exposure as a continuous
13 variable. Both model forms yielded exposure-response results with good fit to the occupational
14 exposure data. The default assumption of the extended Cox proportional hazards model as well
15 as the Poisson regression model is that all censoring (due to death or loss to follow-up) is
16 assumed to be independent of exposure to the Libby Amphibole asbestos (e.g., death in an
17 automobile accident or moved to Canada). However, exposure to Libby Amphibole asbestos
18 may be causing deaths from other causes such as asbestosis or nonmalignant respiratory disease
19 (Larson et al., 201 Ob), which is referred to as dependent censoring. The concern is that the
20 observation of lung-cancer mortality may be precluded by mortality from other causes.
21 In the cohort of 880 workers hired after 1959, 32 died of lung cancer, while 10 died of
22 asbestosis, and 21 died of nonmalignant respiratory disease. The mean length of follow-up from
23 the date of hire until death for the workers who died of lung cancer was 24.9 years. However,
24 the mean length of follow-up for the workers who died of asbestosis or nonmalignant respiratory
25 disease was 30.4 years, so it does not appear that early deaths from other causes associated with
26 exposure to the Libby Amphibole asbestos (Larson et al., 201 Ob) would have precluded many
27 cases of lung cancer. This implies that any potential bias in the lung cancer risk estimates due to
28 dependent competing risks is small.
29 With respect to mesothelioma mortality, it should be noted that the exposure-response
30 modeling is limited by the number of deaths. However, dependent censoring, as described
31 above, is not accounted for in the Poisson regression model and likely causes a downward bias in
This document is a draft for review purposes only and does not constitute Agency policy.
5-122 DRAFT—DO NOT CITE OR QUOTE
-------
1 the estimation of risk. The mean length of follow-up for the workers who died of mesothelioma
2 was 30.1 years, and there is some evidence that early deaths from other exposure-related causes
3 precluded an individual's risk of death from mesothelioma; only lung cancer exhibited a shorter
4 average follow-up time compared to mesothelioma, and in 419 cases of mesothelioma,
5 mesothelioma and lung cancer were never coidentified (Roggli and Vollmer, 2008).
6
7 5.4.6.1.4. Uncertainty in selection of exposure metric
8 There is uncertainty about what metric should be used for modeling exposure to Libby
9 Amphibole asbestos. The previous IRIS IUR assessment for asbestos (U.S. EPA, 1988a) found
10 that cumulative exposure with a 10-year lag was the best metric for lung-cancer mortality, and a
11 more complicated model (see Eq. 5-5) based on average cohort exposure intensity, average
12 cohort time since first exposure, and average duration of employment was the best metric for
13 mesothelioma mortality. This current assessment evaluated these models, but also models that
14 include unlagged and lagged cumulative exposure with and without a half-life of various lengths,
15 and RTW exposure with and without a half-life. In the analysis of comparative model fit, lagged
16 cumulative exposure with a half-life provided the best fits for both mesothelioma and
17 lung-cancer mortality associated with Libby Amphibole asbestos. However, evaluation of
18 20-year lag and longer lag times for mesothelioma was not possible, as the earliest mesothelioma
19 death happened less than 20 years from the start of the exposure, and, hence, exposure was
20 zeroed out, and the fit of any model with 20-year lag was very poor. Latency time for
21 mesothelioma may be as long as 60-70 years [e.g., Bianchi and Bianchi (2009)], so the precise
22 lag time is uncertain.
23 In evaluating the data on lung fiber burden, Berry et al. (2009) estimated the range of the
24 half-life for crocidolite to be between 5 and 10 years. That range is consistent with the finding of
25 a 5 to 10-year half life with 10-15 years lag that provided the best fit to the Libby workers cohort
26 mesothelioma mortality data. Similarly, recent publications indicate that the relative risk of lung
27 cancer due to asbestos exposure declines 15-20 years after the cessation of exposure to asbestos
28 (Magnani et al., 2008; Hauptmann et al., 2002). The marginally best fit for the Libby workers
29 cohort lung-cancer mortality data was for CE models with a 5 to 20-year half life and 10-year
30 lag. However, the precise lag and half-life times are somewhat uncertain. Sensitivity analysis
31 that excluded people with high exposure during 1960-1963 (see Section 5.4.3.6.4) provides
This document is a draft for review purposes only and does not constitute Agency policy.
5-123 DRAFT—DO NOT CITE OR QUOTE
-------
1 further evidence that distinguishing between various lags and decays may be difficult with these
2 data. A limitation of this sensitivity analysis is the decrease in the number of cases, especially
3 for mesothelioma. Resolving this uncertainty would require longer follow-up time, which would
4 allow for a sub-cohort analysis of workers hired in 1967 or afterwards (when exposure estimates
5 began to be based on PCM measurements) until a sufficient number of cases would be available
6 for additional analysis.
7 These simulated decay models were derived mathematically to approximate underlying
8 biological processes that are not well understood, and their better fit is based on maximizing the
9 likelihood for the workers cohort and may not necessarily apply to the environmental exposure
10 patterns. Nonetheless, while the mode of action for carcinogenicity is unknown, the models
11 incorporating a half-life in the exposure metric were clearly preferable for mesothelioma
12 mortality, and the goal of the regression modeling effort was to identify the best fitting exposure
13 model for the Libby worker cohort.
14 The selection of the exposure metric is a source of cross-metric variability discussed in
15 Section 5.4.5.3, and the IUR incorporates this variability. The difference between this value and
16 the value derived from the best fitting exposure model describes the quantitative uncertainty,
17 which is less than twofold.
18
19 5.4.6.1.5. Uncertainty in assessing of mortality corresponding to the cancer-specific endpoints
20 As well established in the literature, mortality rates calculated from death certificates are
21 lower than the true rate of death due to both lung cancer, and to a larger degree, mesothelioma
22 [lung cancer sensitivity: ranging from 86% in an asbestos cohort (Selikoff and Seidman, 1992) to
23 95% in general (Percy et al., 1981): mesothelioma sensitivity: ranging from 40% for ICD-9
24 (Selikoff and Seidman. 1992) to about 80% for ICD-10 (Camidge et al.. 2006: Pinheiro et al..
25 2004). This underestimation of the true rate will result in a lower estimated risk compared with
26 that which would be estimated based on the true rate. The underestimation of risk is much more
27 pronounced for the absolute risk model (mesothelioma) than for the relative risk model (lung
28 cancer). Misdiagnosis rates would need to be quite disparate in the cohort and the comparison
29 population to impact relative risks, and this is unlikely for internal controls that were used in the
30 lung cancer analysis using the Cox proportional hazards model. Therefore, EPA considered use
31 of a procedure to adjust risks for mesothelioma—but not for lung cancer—underascertainment
This document is a draft for review purposes only and does not constitute Agency policy.
5-124 DRAFT—DO NOT CITE OR QUOTE
-------
1 (see Section 5.4.5.1.1). This procedure makes certain assumptions, in particular, that an
2 adjustment factor derived for the full cohort applies to the sub-cohort hired after 1959, and that
3 the rate of misdiagnosis of peritoneal mesotheliomas has not improved recently, and that the
4 proportion of peritoneal mesotheliomas in the cohort is estimated from the available information
5 on the type of mesothelioma in one-third of mesothelioma cases. However, overall uncertainty
6 in this adjustment is low, and the application of the adjustment reduces the bias associated with
7 the diagnostic underascertainment.
8 The endpoint for both mesothelioma and lung cancer was mortality, not incidence. The
9 latter is generally desirable, but median survival with lung cancer and, especially, mesothelioma
10 is not very long, so uncertainty related to the endpoint being death and not incidence is low.
11 There is evidence that other cancer endpoints may also be associated with exposure to the
12 commercial forms of asbestos. IARC concluded that there was sufficient evidence in humans
13 that commercial asbestos (chrysotile, crocidolite, amosite, tremolite, actinolite, and
14 anthophyllite) was causally associated with lung cancer and mesothelioma, as well as cancer of
15 the larynx and the ovary (Straif et al., 2009). Among the entire Libby workers cohort, only
16 2 deaths were found to be due to laryngeal cancer, and there were no deaths from ovarian cancer
17 among the 24 deaths of 84 female workers. The lack of sufficient number of workers to estimate
18 risk of ovarian cancer is an uncertainty in an overall cancer health assessment.
19 The remaining uncertainties attributed to assessing mortality corresponding to the cancer
20 endpoints are considered to be low.
21
22 5.4.6.1.6. Uncertainty in control of potential confounding in modeling lung cancer
23 It is well known that smoking is a strong independent risk factor for lung cancer and may
24 have a synergistic effect with asbestos exposure (Wraith and Mengersen, 2007). In contrast,
25 smoking is not considered a risk factor for mesothelioma (Selikoff and Lee, 1978; Anderson et
26 al.. 1976).
27 As an important potential confounder of the lung-cancer mortality analysis, the possible
28 effect of smoking on the estimated risk of lung-cancer mortality associated with exposure to
29 Libby Amphibole asbestos needs to be evaluated to the fullest extent possible. This
30 consideration was discussed in Amandus and Wheeler (1987) and in Section 4.1.1.3.
This document is a draft for review purposes only and does not constitute Agency policy.
5-125 DRAFT—DO NOT CITE OR QUOTE
-------
1 Additionally, W.R. Grace and Co. instituted a smoking ban on the property in 1979
2 (Peacock, 2003). Information is not available as to the effect of this smoking ban at work on
3 smoking patterns outside of the work environment. About 30% of the sub-cohort was still
4 employed in 1979 and all of the post-1959 cohort had been terminated by May 1982, so the
5 impact of a workplace smoking ban on cohort smoking history may explain the higher proportion
6 of former smokers in the Amandus and Wheeler (1987) data. Lung cancer risks in ex-smokers
7 decrease over time compared to lung cancer risks in continued smokers. A reduction of smoking
8 in the Libby worker population may lead to fewer observations of lung cancer deaths in later
9 years of the cohort study than would have occurred in the absence of the smoking restrictions.
10 Changes in smoking behavior during the course of the epidemiological observation period would
11 lead to changes in the observed time course of lung cancer death rates. This issue is related to
12 potential effect modification of lung-cancer mortality described in Section 5.4.6.1.7.
13 Without high-quality individual-level data on smoking that could be used to control for
14 potential confounding, it is still possible to comment upon the likelihood and potential magnitude
15 of confounding and the impact any confounding would be expected to have on the lung-cancer
16 mortality risk estimates. Confounding can be controlled for in a number of ways including by
17 modeling and by restriction. Restriction of the study population can reduce any potential
18 confounding by making the resulting population more similar. For instance, there can be no
19 confounding by gender when a study population is restricted to only men. This assessment
20 restricted the study population to those workers hired after 1959. Smoking habits have changed
21 over time, and it can reasonably be assumed that the range of smoking habits among those hired
22 after 1959 is less variable than that among the whole cohort, particularly because of the narrower
23 range of birth cohorts represented in this sub-cohort. This should have the effect of reducing
24 some of the potential for confounding. Analytic examinations of potential confounding are
25 discussed below.
26 Additionally, the extended Cox proportional hazards models controlled for date of birth,
27 which effectively controls for any secular trends in confounders over time (Tableman and Kim,
28 2004). Amandus and Wheeler (1987) cite data from the U.S. Public Health Service (HEW.
29 1979) showing a steady decrease in the prevalence of current smoking from 52.9% in 1964 when
30 the U.S. Surgeon General's report on smoking was released to 42.3% in 1970 and 37.5% in 1978
31 (HHS, 1990). If current smoking were a meaningful confounder, such a reduction in smoking
This document is a draft for review purposes only and does not constitute Agency policy.
5-126 DRAFT—DO NOT CITE OR QUOTE
-------
1 rates over time should have produced a noticeable distortion in the proportionality of the hazards
2 as the magnitude of confounding by smoking changes with smoking prevalence. No violation of
3 the proportional hazards assumption was observed in the context of the Cox proportional hazards
4 model; hence, there is no evidence of confounding by smoking in the analyses of workers hired
5 after 1959.
6 Lastly, Richardson (2010) describes a method to determine if an identified exposure
7 relationship with lung cancer is confounded by unmeasured smoking in an occupational cohort
8 study. EPA implemented this methodology to model the potential effects of Libby Amphibole
9 asbestos on the risk of COPD mortality on the sub-cohort of workers hired after 1959 (see
10 Section 5.4.3.6.5). Summarizing these findings, EPA used the method described by Richardson
11 (2010) to evaluate whether exposures to Libby Amphibole asbestos predicted mortality from
12 COPD as an indication of potential confounding by smoking and found a nonsignificant negative
13 relationship, which was inconsistent with confounding by smoking.
14
15 5.4.6.1.7. Uncertainty due to potential effect modification
16 Among the 32 deaths from lung cancer in workers hired after 1959 that were used to
17 estimate the unit risk of lung-cancer mortality (see Section 5.4.5.2), data on smoking listed 16 as
18 smokers, 4 as former smokers, and 12 of the 32 had missing data. Thus, data to support an
19 estimate of the risk of Libby Amphibole asbestos among known nonsmokers were not available.
20 It is theoretically possible that the risk of lung-cancer mortality estimated in this current
21 assessment is a reflection of a positive synergy between smoking and asbestos, and that the
22 adverse effect of Libby Amphibole asbestos among the potentially nonsmoking workers has been
23 overestimated. The unit risk of the lung cancer estimate herein and the combined mesothelioma
24 and lung-cancer mortality IUR would then be health protective for any population that had a
25 lower prevalence of smoking than that of the Libby worker cohort. However, if the smoking ban
26 did diminish the effect of smoking, then any overestimation would be somewhat mitigated.
27
28 5.4.6.1.8. Uncertainty due to length of follow-up
29 There is some potential uncertainty regarding the length of follow-up for cancer
30 mortality, even more so with the restriction of the cohort to those workers hired after 1959. The
31 hire dates among this subset of the cohort ranged from January 1960 to November 1981 (the
This document is a draft for review purposes only and does not constitute Agency policy.
5-127 DRAFT—DO NOT CITE OR QUOTE
-------
1 mean date of hire was May 1971). Follow-up continued until the date of death or
2 December 31, 2006, whichever occurred first. Therefore, the range of follow-up was from 25 to
3 46 years, with a mean of more than 35 years.
4 However, for mesothelioma mortality, the length of the latency period is considerably
5 longer. Suzuki (2001) reviewed 1,517 mesothelioma cases from 1975 through 2000 and was
6 able to estimate the latency for 800. Suzuki (2001) reported 17% of cases had a latency of less
7 than 30 years with 52% of cases with a latency of less than 40 years. Bianchi and Bianchi (2009)
8 estimated the mesothelioma latency in 552 cases and reported mean latency periods of 35 years
9 among insulators, 46 years among various industries, and 49 years among shipyard workers.
10 The effect of insufficient length of follow-up for mesothelioma mortality would be to
11 underestimate the risk of exposure since there would be workers who may eventually die of
12 mesothelioma that are not counted in this current assessment. Because the risk of mesothelioma
13 mortality is evaluated as an absolute risk, the unit risk of mesothelioma mortality may reasonably
14 be expected to rise with time moderated by the increase in person-years of follow-up. According
15 to the results of Suzuki (2001) and of Bianchi and Bianchi (2009), a mean length of follow-up of
16 35 years may only have captured half of all eventual mesothelioma mortality cases among the
17 Libby workers hired after 1959. If this were so, then the unit risk of mesothelioma mortality
18 could be larger than was estimated from existing data, depending on the relationship between the
19 number of additional deaths and increase in person-years.
20
21 5.4.6.1.9. Uncertainty in use of life-tables to calculate cancer mortality IUR
22 The life-table procedure computes the extra risk of death from birth up to 85 years of age,
23 in part, because this is how national cancer incidence and mortality rate data that are one basis of
24 the life-tables are made available (see 2003-2007 SEER Table 15.10, age-specific U.S. death
25 rates). Because the prevalence of cancer mortality is a function of increasing age, this cut-off at
26 age 85 ignores a small additional risk of lung-cancer mortality among a small percentage of
27 people who have the higher background risk. This has the effect of slightly underestimating the
28 IUR that would be derived if the life-table were extended for an additional period of time,
29 accounting for longer life spans. Extension of the life-table analysis to people over the age of
30 85 requires an additional assumption. Assuming that having attained the age of 85 years, the
31 additional life expectancy is 5 years, then the lung-cancer mortality unit risk based on the LECoi
This document is a draft for review purposes only and does not constitute Agency policy.
5-128 DRAFT—DO NOT CITE OR QUOTE
-------
1 would be somewhat larger—on the order of 5-10%—slightly more than the additional
2 mesothelioma mortality risk if the life-tables were extended.
o
J
4 5.4.6.1.10. Uncertainty in combining of risk for composite cancer IUR
5 For the purpose of combining risks, it is assumed that the unit risks of mesothelioma and
6 lung-cancer mortality are normally distributed. Since risks were derived from a large
7 epidemiological cohort, this is a reasonable assumption supported by the statistical theory, and
8 uncertainty related to it is low.
9
10 5.4.6.1.11. Uncertainty in extrapolation of findings in adults to children
11 The analysis of lung-cancer mortality specifically tested and confirmed the assumption
12 that the relative risk of exposure is independent of age within the age range of the occupational
13 sub-cohort hired after 1959. However, no comparable data are available to estimate the lifetime
14 risk from early life exposures. The life-table procedure is conducted so as to initiate exposure at
15 age 16 to represent adult exposures. Then, the adult-only-exposure IUR estimates derived from
16 the life-table analysis need to be rescaled to a 70-year lifespan in order to yield the standard
17 lifetime IUR, allowing risk estimate calculations involving less-than-lifetime exposure scenarios,
18 in the standard manner. After reseating, the resulting "adult-based" IUR estimate (in contrast to
19 the unsealed "adult-only-exposure" IUR estimate obtained from the life-table calculations) can
20 be employed seamlessly by the end-user in the same manner as for an adult-based IUR estimate
21 derived from a rodent bioassay. Lack of published information on risks associated with Libby
22 Amphibole asbestos-specific exposure during childhood is the uncertainty associated with the
23 proposed extrapolation. If such information is subsequently published, the extrapolation
24 procedure can be updated.
25
26 5.4.6.2. Summary
27 In the discussion of the overall uncertainty in the IUR, it is important to distinguish
28 between uncertainty that encompasses both the direction and the magnitude from uncertainty
29 with known directional effects on the IUR but of unknown magnitude. In this summary, only the
30 latter uncertainties, which may result in underestimated or overestimated risk, are listed below.
This document is a draft for review purposes only and does not constitute Agency policy.
5-129 DRAFT—DO NOT CITE OR QUOTE
-------
1 Uncertainties that are not thought to alter the estimated magnitude of the risk in a systematic
2 direction are not included in this summary.
3 The sources of uncertainty that could lead to a likely underestimation of the cancer risk
4 value include the following:
5
6
7 • Use of historical PCM exposure measurements. Because asbestos was a
8 component of vermiculite that was the primary object of production, mine and dry
9 and old wet mill ambient air may have contained material other than asbestos that
10 could have contributed to fibers counted by PCM. Therefore, it is possible that
11 exposure estimates for some, or possibly a large portion of the cohort, are
12 overestimated, and, therefore, the resulting IUR may be underestimated.
13 • Measurement error in exposure assessment and assignment. This current
14 assessment showed that unit risk results from analysis of the lung-cancer
15 mortality in the full cohort (see Table 5-21) compared to the sub-cohort hired
16 after 1959 may have been attenuated as much as 2-6 times (see Section
17 5.4.6.1.2.4). By excluding those cohort members hired before 1960 for whom
18 there was insufficient work history information to estimate their exposures, the
19 unit risk for lung cancer was less attenuated due to exposure measurement error.
20 However, exposure measurements from the 1960s are also imperfect and include
21 a lesser degree of exposure measurement error, which could have led to
22 underestimated risk even in the sub-cohort hired after 1959.
23 • Limited length of follow-up. Absolute risk is used for mesothelioma; therefore,
24 the unit risk of mesothelioma mortality could be larger than was estimated from
25 existing data, depending on the relationship between the number of additional
26 deaths and an increase in person-years.
27 • Use of life-tables to calculate the IUR based on cancer mortality. The
28 lung-cancer mortality unit risk based on the LECoi would be somewhat larger,
29 about 5-10%, and the mesothelioma unit risk would be slightly less (about 3%)
30 than that if the life-tables were extended from 85 to 90 years to account for longer
31 life spans.
32 • Small number of women and ovarian cancer. While asbestos is causally
33 associated with increased risks of ovarian cancer (Straif et al., 2009), there were
34 only 84 women in the whole cohort, and there were no deaths from ovarian cancer
35 among 24 total deaths. To the extent that there was an increased risk of ovarian
36 cancer in the Libby workers cohort due to inhalation exposures that was
37 unobserved, then the IUR would be somewhat underestimated. However, it was
38 not possible to estimate the magnitude of this underestimation on the total cancer
39 risk.
This document is a draft for review purposes only and does not constitute Agency policy.
5-130 DRAFT—DO NOT CITE OR QUOTE
-------
1 • Dependent competing risks. Competing risk of mortality from other diseases
2 related to exposure may have resulted in underestimates of the risk of mortality
3 from either mesothelioma or lung cancer. The mean length of follow-up for the
4 Libby workers who died of mesothelioma was to 30.1 years, and evidence exists
5 (Bianchi and Bianchi, 2009; Suzuki and Yuen, 2001) that early deaths from other
6 exposure-related causes could have precluded an individual's risk of death from
7 mesothelioma. However, it was not possible to estimate the magnitude of this
8 effect on the total cancer risk.
9
10
11 The sources of uncertainty that could lead to a likely overestimation of the cancer risk
12 value include the following:
13
14
15 • Potential residual confounding and effect modification. The unit risk of
16 lung-cancer mortality estimated herein, and the combined mesothelioma and
17 lung-cancer mortality IUR, would over-estimate the risk in any population that
18 had a lower prevalence of smoking than that of the Libby worker cohort. Since
19 the Libby worker cohort had a large prevalence of smokers and ex-smokers and
20 no known nonsmokers developed lung cancer, it is also possible that estimated
21 risk for lung cancer is actually risk for an interaction of lung cancer and smoking,
22 and effects of smoking and asbestos are known to be between additive and
23 multiplicative (see Section 4).
24
This document is a draft for review purposes only and does not constitute Agency policy.
5-131 DRAFT—DO NOT CITE OR QUOTE
-------
1 6. MAJOR CONCLUSIONS IN THE CHARACTERIZATION OF HAZARD AND
2 EXPOSURE RESPONSE
3 Libby Amphibole asbestos,45 present in vermiculite from the mine near Libby, MT, is a
4 complex mixture of amphibole fibers—both mineralogically and morphologically (see
5 Section 2.2). The mixture primarily includes tremolite, winchite, and richterite amphibole
6 minerals which exhibit a range of fiber morphologies (e.g., asbestiform, acicular, prismatic)
7 (Meeker et al., 2003). Given the exposure potential to Libby Amphibole asbestos—and its
8 characteristic mineral composition—a hazard characterization and cancer exposure-response
9 assessment are presented.
10 As discussed in Section 1, there is currently no reference concentration (RfC) for
11 asbestos, and the U.S. Environmental Protection Agency (EPA) Integrated Risk Information
12 System (IRIS) inhalation unit risk (IUR) for asbestos is based on a synthesis of 14 epidemiologic
13 studies that included occupational exposure to chrysotile, amosite, or mixed mineral fibers
14 (chrysotile, amosite, and crocidolite) (U.S. EPA, 1988a). There is uncertainty in applying the
15 resulting IUR to environments and minerals that are not included in the studies considered for
16 the asbestos IUR derivation (U.S. EPA, 1988a). Published mortality studies on the Libby, MT
17 worker cohort have become available since the derivation of the IRIS asbestos IUR [i.e.,
18 McDonald et al. (2004: 1986a): Amandus and Wheeler (1987): Sullivan (2007): Larson et al.
19 (201 Ob)]. This assessment documents noncancer and cancer health effects from inhalation
20 exposure to Libby Amphibole asbestos. Data are not available to support derivation of either a
21 reference dose (RfD) or a cancer oral slope factor (OSF) following oral exposures to Libby
22 Amphibole asbestos.
23
24 6.1. HUMAN HAZARD POTENTIAL
25 6.1.1. Exposure
26 Vermiculite ore mined near Libby, MT, contained Libby Amphibole asbestos, which
27 remained in the vermiculite concentrate (VC) and exfoliated product shipped from the facilities
28 (see Section 2). Vermiculite from the Libby, MT mine was used commercially from the 1920s to
45 The term "Libby Amphibole asbestos" is used in this document to identify the mixture of amphibole mineral
fibers of varying elemental composition (e.g., winchite, richterite, tremolite, etc.) that have been identified in the
Rainy Creek complex near Libby, MT. It is further described in Section 2.2.
This document is a draft for review purposes only and does not constitute Agency policy.
6-1 DRAFT—DO NOT CITE OR QUOTE
-------
1 1990, and a review of company records available from (1964-1990) indicates approximately
2 6,109,000 tons of VC was shipped to over 200 facilities (ATSDR. 2008b). Vermiculite was
3 most notably used as attic insulation, a soil amender for gardening, and in the manufacturing of
4 gypsum wallboard. The exposure potential to Libby Amphibole asbestos includes historical
5 exposures (both occupational and community), as well as the potential for ongoing exposures to
6 waste materials, contaminated soils and vegetation, and consumer products (e.g., vermiculite
7 attic insulation; see Section 2.3) (ATSDR. 2008b. 2001b).
8 There are many ways in which workers and residents in Libby, MT, and the surrounding
9 communities may have been exposed while the mining and milling operations were active.
10 Historical routes of exposure include (1) occupational exposure; (2) take-home exposure for
11 household contacts of the workers (including children); (3) dust/fiber emissions to the
12 community from the milling and exfoliating facilities; (4) distribution of waste material into the
13 community as fill (including yards and recreational areas); (5) use of vermiculite attic insulation
14 in homes; (6) use of vermiculite in gardening/horticulture; and (7) children playing in the waste
15 stoner rock piles (Peipins et al., 2003). Other than documentation of dust and fiber exposure
16 levels for mine and mill workers, there are few data to inform the levels of exposure to
17 household contacts and community members during mine and mill operations. Although no
18 historical exposure measurements are available from the homes of the workers, the EPA has
19 conducted sampling to determine exposure levels from vermiculite and waste materials that
20 remain in the community (U.S. EPA, 2006c; Weis, 200la, b) (see Appendix B). These data
21 provide information useful to understand what historical exposures might have been for similar
22 activities. More recently, EPA has characterized exposures for various exposure pathways in the
23 community and continues to evaluate exposure potential in the ongoing efforts for cleanup (U.S.
24 EPA. 2010a).
25 Outside of Libby, MT, vermiculite concentrate and exfoliated product was shipped to
26 271 domestic sites that served as processing facilities (U.S. GAP, 2007). These sites included
27 exfoliation plants (e.g., for the production of vermiculite insulation) as well as nonexfoliation
28 facilities (e.g., production of gypsum wallboard). The vermiculite concentrate was exfoliated by
29 heat-induced expansion resulting in vermiculite produced for commercial purposes. Both the
30 commercial vermiculite and the waste stoner rock (i.e., residual waste stoner rock from
31 exfoliation) contained Libby Amphibole asbestos fibers. Potential exposure routes in these
This document is a draft for review purposes only and does not constitute Agency policy.
6-2 DRAFT—DO NOT CITE OR QUOTE
-------
1 communities located around the country parallel the exposures in Libby, MT, including
2 occupational exposures, take-home exposures from workers, and children playing in the piles of
3 waste stoner rock near the facility (ATSDR. 2008b. 2005b. 2003a). Waste materials (expanded
4 vermiculite and waste stoner rock) from some of these facilities were also used for fill in local
5 communities, potentially creating additional exposure pathways based on an Agency for Toxic
6 Substances and Disease Registry (ATSDR) review of 28 facilities, and a survey of the Western
7 Minerals Plant, MN (ATSDR, 2008b, 2003a). Few historical samples are available from these
8 facilities that could be used to quantify the exposure potential for workers or for the surrounding
9 communities (ATSDR, 2008b, 2005a, 2003a). Air modeling conducted for one exfoliating
10 facility in Minnesota does provide support for the potential of dust/fiber emissions from
11 exfoliating plants to impact ambient air quality in the vicinity of the plant (ATSDR, 2003 a).
12 While the mine was active, there was potential exposure to commercial products
13 containing vermiculite from Libby, MT, especially in gardening soils and vermiculite attic
14 insulation. No studies have evaluated the potential for consumer exposure when vermiculite
15 from Libby, MT, was employed as a soil amender, but air sampling at one facility where this was
16 produced (O.M. Scott facility in Marysville, OH) demonstrated that workers handling this
17 material during manufacture were exposed to Libby Amphibole asbestos fibers (see
18 Section 5.2.3.1). There is potential for exposure in homes that contain vermiculite attic
19 insulation from Libby, MT, where residents and workers might enter attics for various uses,
20 repairs, and renovations (see Section 2.3.3).
21
22 6.1.2. Fiber Toxicokinetics
23 There is no specific information available on the fiber toxicokinetics of Libby Amphibole
24 asbestos. However, as a mineral fiber, the characteristics that define the deposition, clearance,
25 and translocation of other amphibole fibers might apply to Libby Amphibole asbestos. As
26 discussed in Section 3, the specific fiber dimensions and density of Libby Amphibole asbestos
27 will determine the probable pattern of deposition in the respiratory tract and other tissues (e.g.,
28 pleura, peritoneum). Based on the fiber-size profile of airborne Libby Amphibole asbestos
29 fibers, deposition is expected throughout the respiratory tract including the alveolar regions.
30 Less is known about mineral fiber translocation to other target tissues in general, and, to date, no
This document is a draft for review purposes only and does not constitute Agency policy.
6-3 DRAFT—DO NOT CITE OR QUOTE
-------
1 studies have specifically examined translocation following exposure to Libby Amphibole
2 asbestos.
3 As with other mineral fibers, clearance is likely to occur via the mucociliary apparatus in
4 the upper respiratory tract and the mucociliary escalator for those fibers deposited in the trachea
5 and bronchioles. This clearance is enhanced by macrophage action, which may transport some
6 of the fibers from the alveolar sac to the mucociliary system. Fibers may also be dissolved in
7 lung fluids or through the more aggressive action of alveolar macrophages. In general,
8 amphibole asbestos is considered more persistent and less likely to dissolve than other natural
9 mineral fibers, including serpentine asbestos (i.e., chrysotile) fibers. However, no data are
10 available for Libby Amphibole asbestos specifically, and it is unknown if Libby Amphibole
11 asbestos fibers would split or break in the pulmonary compartment as has been shown with some
12 amphibole fibers (e.g., ferroactinolite) (Coffin etal., 1983).
13 Any fibers deposited in the respiratory tract and not cleared via the mucociliary system,
14 or not dissolved, can remain in the lung or can be transported to other tissues. Although data
15 specific to Libby Amphibole asbestos are not yet available, other asbestos fiber types can
16 translocate from the lung via macrophage action and transport through the lymph system, or
17 direct migration may occur through tissues from the mechanical action of the lung. Pleural and
18 peritoneal effects documented in Libby Amphibole asbestos-exposed individuals support the
19 potential for translocation of Libby Amphibole asbestos into the pleura.
20
21 6.1.3. Noncancer Health Effects in Humans and Laboratory Animals
22 Noncancer health effects identified in humans following inhalation exposure to Libby
23 Amphibole asbestos include pleural abnormalities, asbestosis, and reduced lung function as well
24 as increased mortality from noncancer causes. Two cohorts of workers exposed to Libby
25 Amphibole asbestos have been studied: workers at the mine and related operations in Libby, MT
26 and employees in the O.M. Scott plant in Marysville, OH, where the vermiculite product was
27 exfoliated and used as an inert carrier in lawn care products. Radiographic assessments of study
28 participants in both cohorts indicate radiographic abnormalities consistent with asbestos-related
29 disease, specifically pleural thickening (localized [LPT] and diffuse [DPT]) and small opacities
30 (indicative of interstitial fibrosis) (Rohs et al., 2008; Amandus et al., 1987b: McDonald et al.,
31 1986b: Lockey et al., 1984). These studies provided quantitative exposure estimates and were
This document is a draft for review purposes only and does not constitute Agency policy.
6-4 DRAFT—DO NOT CITE OR QUOTE
-------
1 considered suitable for exposure-response analysis to support an RfC derivation. Additionally,
2 five cohort mortality studies of Libby, MT workers identified increased risk of mortality from
3 noncancer causes, including nonmalignant respiratory disease (e.g., asbestosis) (Larson et al.,
4 201 Ob: Sullivan. 2007: McDonald et al.. 2004: Amandus and Wheeler. 1987: McDonald et al..
5 1986a) and cardiovascular disease (Larson et al., 201 Ob).
6 ATSDR conducted health screening of community members in and around Libby, MT
7 (including past workers), and identified an increase in radiographic abnormalities with an
8 increased number of exposure pathways (Peipins et al., 2004a: Peipins et al., 2003: ATSDR,
9 2001b). Other researchers have also used these data to identify the increased prevalence of
10 respiratory symptoms in children (Vinikoor et al., 2010) and to evaluate the prevalence of
11 radiographic abnormalities and reduced lung function in nonworker participants (Weill et al.,
12 2011). Radiographic abnormalities were more prevalent in mine/mill workers versus other
13 exposure categories (i.e., household contacts, dusty trades, and community-only exposures)
14 (Weill et al., 2011). Pleural thickening (LPT or DPT) increased with age, within each exposure
15 group. Decreased pulmonary function (as percent of the predicted forced vital capacity) are
16 reported for participants with radiographic abnormalities (small opacities, DPT, and LPT) with
17 greater effects seen in participants with small opacities and DPT (Weill et al., 2011). A nested
18 case-control study based on this study group also identified a potential for increased prevalence
19 of autoimmune disease (Pfau et al., 2006), although additional research is needed to explore this
20 potential health outcome.
21 Although laboratory animal data and experimental data on toxicity mechanisms are
22 limited for Libby Amphibole asbestos, the existing data are consistent with the health effects
23 observed in both workers and community members exposed to Libby Amphibole asbestos.
24 Experimental animal studies have demonstrated increased collagen deposition consistent with
25 fibrosis following intratracheal instillation of Libby Amphibole asbestos fibers in C57B16 mice
26 (Smarttetal., 2010: Putnam et al., 2008) and Fisher 344 rats (Padilla-Carlin et al., 2011) as well
27 as increased markers of pulmonary inflammation in a rat model for human cardiovascular
28 disease (Shannahan et al., 201 la: Shannahan et al., 201 Ib). Pulmonary fibrosis, inflammation,
29 and granulomas were observed after tremolite, which comprises approximately 6% of the fiber
30 mixture in Libby Amphibole asbestos, inhalation exposure in specific-pathogen-free (SPF) male
31 Wistar rats (Bernstein et al., 2005: Bernstein et al., 2003), and intratracheal instillation in male
This document is a draft for review purposes only and does not constitute Agency policy.
6-5 DRAFT—DO NOT CITE OR QUOTE
-------
1 albino Swiss mice (Sahu et al., 1975). Davis et al. (1985) also reported pulmonary effects after
2 inhalation exposure to tremolite in SPF male Wistar rats including increases in peribronchiolar
3 fibrosis, alveolar wall thickening, and interstitial fibrosis.
4
5 6.1.4. Carcinogenicity in Humans and Laboratory Animals
6 There is convincing evidence of a causal association between exposure to Libby
7 Amphibole asbestos mesothelioma and lung cancer in workers from the Libby, MT vermiculite
8 mining and milling operations (Larson et al., 201 Ob: Sullivan, 2007; McDonald et al., 2004;
9 Amandus et al., 1988; Amandus and Wheeler, 1987; McDonald et al., 1986a). No other
10 occupational cohort with exposures to Libby Amphibole asbestos has been studied with respect
11 to mortality risks. Whitehouse et al. (2008) documented 11 mesothelioma cases in nonworkers
12 exposed to Libby Amphibole asbestos in Libby, MT. Increased lung cancer and mesothelioma
13 deaths are also reported for worker cohorts exposed to other forms of amphibole fibers (amosite
14 and crocidolite) (de Klerk et al., 1989; Seidman et al., 1986; Henderson and Enterline, 1979).
15 These findings are consistent with the increased cancers reported for communities exposed to
16 various rocks and soils containing tremolite fibers (Hasanoglu et al., 2006; Sichletidis et al.,
17 1992: Baris et al.. 1987: Langer et al.. 1987: Baris et al.. 1979: Yazicioglu, 1976). Although
18 potency, fiber dimension, and mineralogy differ between amphiboles, these studies are
19 supportive of the hazard identification of Libby Amphibole asbestos fibers described in this
20 assessment.
21 Although there is a limited laboratory animal database, the studies that are available
22 support the determination of carcinogenicity of Libby Amphibole asbestos fibers. Smith (1978)
23 demonstrated mesotheliomas in hamsters given a single intrapleural injection of Libby
24 Amphibole asbestos material (see Table 4-15). Tremolite is also carcinogenic in studies in rats,
25 hamsters, and mice, resulting in pleural mesothelioma, peritoneal mesothelioma, and lung cancer
26 depending on the route of exposure (see Table 4-16) (Bernstein et al., 2005; Bernstein et al.,
27 2003: Roller et al., 1996: Davis et al., 1991: Davis et al., 1985: Stanton et al., 1981). Although
28 comparing the potency of the tremolite used in these studies is difficult given the lack of
29 information on fiber characteristics and other study limitations, these results demonstrate an
30 increased risk for lung cancer and mesothelioma following exposure to tremolite asbestos.
31
This document is a draft for review purposes only and does not constitute Agency policy.
6-6 DRAFT—DO NOT CITE OR QUOTE
-------
1 6.1.5. Susceptible Populations
2 Certain populations could be more susceptible than the general population to adverse
3 health effects from exposure to Libby Amphibole asbestos. In general, factors that may
4 contribute to increased susceptibility from environmental exposures include lifestage, gender,
5 race/ethnicity, genetic polymorphisms, health status, and lifestyle. However, little data exist to
6 address the potential of increased susceptibility to cancer or noncancer effects from exposure to
7 the Libby Amphibole asbestos.
8 Most occupational studies of workers exposed to Libby Amphibole asbestos have
9 examined the effects only in men because this group represents the vast majority of workers in
10 these settings (Moolgavkar et al.. 2010: Sullivan. 2007: McDonald et al.. 2004: Amandus et al..
11 1988: Amandus et al., 1987a: Amandus and Wheeler, 1987: Amandus et al., 1987b: McDonald
12 et al., 1986a: McDonald et al., 1986b). The analysis presented here includes all workers,
13 however, there were few women in the cohort, and therefore no determination can be made
14 regarding increased susceptibility to lung cancer or mesothelioma by gender. Gender-related
15 differences in exposure patterns, physiology, and dose-response are some of the factors that may
16 contribute to gender-related differences in risk from asbestos exposure (Smith, 2002). The
17 limited data available from community-based studies (ATSDR, 2000) do not provide a basis for
18 drawing conclusions regarding gender-related differences in carcinogenic effects from Libby
19 Amphibole asbestos. Racial diversity among workers exposed to Libby Amphibole asbestos is
20 also limited, and data on ethnic groups are absent, precluding the ability to examine racial and
21 ethnicity-related differences in the mortality risks within the Libby, MT worker cohort. Finally,
22 the potential modifying effects of genetic polymorphisms, pre-existing health conditions,
23 nutritional status, and other lifestyle factors have not been studied sufficiently to determine their
24 potential contribution to variation in risk in the population.
25
26 6.1.6. Mode-of-Action Information
27 Due to the limited data that are available specific to Libby Amphibole asbestos, the mode
28 of action (MO A) of Libby Amphibole asbestos for lung cancer and mesothelioma following
29 inhalation exposure cannot be established. Laboratory animal studies of mice (Smartt et al.,
30 2010: Putnam et al.. 2008). hamsters (Smith. 1978) or rats (Padilla-Carlin et al.. 2011:
31 Shannahan et al., 201 la: Shannahan et al., 201 Ib) exposed to Libby Amphibole asbestos suggest
This document is a draft for review purposes only and does not constitute Agency policy.
6-7 DRAFT—DO NOT CITE OR QUOTE
-------
1 a similar type of inflammatory response to that observed with other mineral fibers; however, no
2 inhalation studies were available in the published literature. In vivo studies in rats, hamsters, or
3 mice exposed to tremolite (Roller et al., 1997, 1996; Davis etal., 1991; McConnell etal., 1983b:
4 Wagner etal., 1982; Stanton et al., 1981; Smith etal., 1979) show results similar to other
5 amphibole asbestos fibers including lung cancer and mesothelioma, with limited inhalation
6 studies (Bernstein et al., 2005; Bernstein et al., 2003; Davis et al., 1985). In vitro studies
7 demonstrate that the uptake of Libby Amphibole asbestos fibers by macrophage, mesothelial,
8 and lung epithelial cell lines may lead to an increase in oxidative stress as measured by reactive
9 oxygen species production, gene expression changes or genotoxicity (Hillegass et al., 2010;
10 Pietruska etal.. 2010: Blake et al.. 2007). Thus, the available data indicate that Libby
11 Amphibole asbestos induces biological responses similar to other forms of asbestos such as
12 oxidative stress, chronic inflammation, genotoxicity, and increased cell proliferation. These
13 biological effects following exposure to Libby Amphibole asbestos and/or tremolite are
14 demonstrated in a limited number of laboratory animal and in vitro studies. Multiple key events
15 for one particular toxicity pathway or MOA have not been identified and adequately supported;
16 therefore, the MOA for Libby Amphibole asbestos carcinogenicity cannot be established.
17
18 6.1.7. Weight-of-Evidence Descriptor for Cancer Hazard
19 Under the EPA Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), Libby
20 Amphibole asbestos is carcinogenic to humans following inhalation exposure based on
21 epidemiologic evidence that shows convincing evidence of a causal association between
22 exposure to Libby Amphibole asbestos fibers and increased lung cancer and mesothelioma
23 mortality (Larson et al.. 201 Oa: Moolgavkar et al.. 2010: Sullivan. 2007: McDonald et al.. 2004:
24 Amandus and Wheeler, 1987; McDonald et al., 1986a). These results are further supported by
25 animal studies that demonstrate the carcinogenic potential of Libby Amphibole asbestos fibers
26 and tremolite fibers in rodent bioassays. As a durable mineral fiber of respirable size, this
27 conclusion is consistent with the extensive published literature that documents the
28 carcinogenicity of amphibole fibers.
29 U.S. EPA's Guidelines for Carcinogenic Risk Assessment (U.S. EPA, 2005a) indicate
30 that for tumors occurring at a site other than the initial point of contact, the weight of evidence
31 for carcinogenic potential may apply to all routes of exposure that have not been adequately
This document is a draft for review purposes only and does not constitute Agency policy.
6-8 DRAFT—DO NOT CITE OR QUOTE
-------
1 tested at sufficient doses. An exception occurs when there is convincing information (e.g.,
2 toxicokinetic data) that absorption does not occur by other routes. Information on the
3 carcinogenic effects of Libby Amphibole asbestos via the oral and dermal routes in humans or
4 animals is absent. The increased risk of lung cancer and mesothelioma following inhalation
5 exposure to Libby Amphibole asbestos has been established by studies in humans, but these
6 studies do not provide a basis for determining the risk from other routes of exposure.
7 Mesothelioma occurs in the pleural and peritoneal cavities and, therefore, is not considered a
8 portal-of-entry effect. However, the role of indirect or direct interaction of asbestos fibers with
9 tissues at extrapulmonary sites is still unknown. There is no information on the translocation of
10 Libby Amphibole asbestos to extrapulmonary tissues following either oral or dermal exposure,
11 and limited studies have examined the role of these routes of exposure in cancer. Therefore,
12 Libby Amphibole asbestos is considered carcinogenic to humans by the inhalation route of
13 exposure.
14
15 6.2. EXPOSURE RESPONSE
16 This assessment contains a derivation of an RfC for noncancer effects and an IUR for
17 cancer based on epidemiologic data. It does not contain an RfD or OSF.
18
19 6.2.1. Noncancer/Inhalation
20 Of the observed noncancer health effects from exposure to Libby Amphibole asbestos,
21 data that provide exposure-response information are only available for increased pleural
22 thickening (localized and diffuse) and signs of interstitial fibrosis (i.e., small opacities) in the two
23 worker cohorts (i.e., Libby worker cohort and Marysville worker cohort). Both cohorts provide
24 individual exposure estimates, and document increased hazard of pleural and parenchymal
25 effects. As detailed in Section 5.2.1, each of the available studies has strengths and weaknesses.
26 The cohort of Marysville, OH workers [Lockey et al. (1984) and the follow-up by Rohs et al.
27 (2008)] was selected as the principal cohort over the Libby worker cohort for five reasons:
28 (1) lack of confounding by residential and community exposure; (2) information on important
29 covariates (e.g., BMI); (3) exposure-response relationship defined for lower cumulative exposure
30 levels (in the post-1972 sub-cohort); (4) adequate length of follow-up; and (5) use of more recent
31 criteria for evaluating radiographs (ILO, 2002) (see Section 5.2.1 for details). Of the observed
This document is a draft for review purposes only and does not constitute Agency policy.
6-9 DRAFT—DO NOT CITE OR QUOTE
-------
1 radiographic abnormalities in exposed workers, localized pleural thickening (LPT) was selected
2 as the critical effect due to its higher prevalence relative to the other outcomes, minimal
3 adversity (compared with other effects), and specificity for durable mineral fiber exposure. LPT
4 is an irreversible pathological change associated with constricting chest pain, dyspnea, and
5 decreased pulmonary function and considered adverse (see Section 5.2.1.4). For an RfC
6 derivation, analyses focused on the cohort of Marysville, OH workers described by Rohs et al.
7 (2008). Specifically, the RfC was derived from the sub-cohort of the Marysville, OH workers
8 who started employment after 1972, due to the greater certainty in exposure assessment in this
9 group.
10 Benchmark dose (BMC) modeling, with a benchmark response of 10% extra risk, was
11 used to derive the point of departure (POD). A Michaelis-Menten regression model was the
12 best-fitting model for the sub-cohort and used to estimate the exposure-response relationship for
13 Libby Amphibole asbestos and LPT. Cumulative exposure with a lag of 10 years was selected as
14 the exposure metric, based on evidence for biological latency and model fit considerations. A
15 background rate of LPT of 1% was assumed based on a limited number of published estimates.
16 The resulting BMCio under these modeling assumptions was 0.2642 fibers/cc-year; the
17 corresponding lower 95% confidence limit of the BMCio (BMCLio) is 0.1177 fibers/cc-year as a
18 cumulative lifetime exposure. The RfC is for continuous exposure (i.e., 24 hours/day,
19 365 days/year, with exposure beginning at birth and continuing for 70 years). Thus, the modeled
20 BMCLio as CE was adjusted to 70 years of exposure, lagged by 10 years (non-occupational,
21 lifetime exposure) resulting in a value of 60 years (see Section 5.2.4).
22
23
24 POD = BMCLio + (lifetime exposure duration)
25 = [0.1177 (fibers/cc) x year] + [70 - 10 years]
26 = 1.96 x 10"3 fibers/cc
27
28
29 The RfC is obtained by applications of uncertainty factors as needed. Two uncertainty
30 factors (UF) have been applied for a composite UF of 100 (intraspecies uncertainty factor,
31 UFA =10; database uncertainty factor, UFD; = 10) (see Section 5.2.4). As shown below, the
This document is a draft for review purposes only and does not constitute Agency policy.
6-10 DRAFT—DO NOT CITE OR QUOTE
-------
1 chronic RfC is 2 x 10 5 fibers/cc for Libby Amphibole asbestos; it was calculated by dividing the
2 lifetime-POD by a composite UF of 100:
3
4
5 Chronic RfC = POD-UF
6 = 1.96 x 10~3 fibers/cc- 100
7 = 1.96 x io~5 fibers/cc, rounded to 2 x io~5 fibers/cc
8
9
10 Modeling was also conducted in the full cohort of workers described in Lockey et al.
11 (1984) and Rohs et al. (2008). These analyses used a different modeling approach, due to the
12 wider range of exposures and time from first exposure. A modified Michaelis-Menten model
13 provided the best fit to the full cohort data, which incorporated time from first exposure via the
14 plateau term for the model. For a time from first exposure of 30 years and exposure lag of
15 10 years, the BMC and BMCL corresponding to a 10% extra risk of LPT were 0.1477 and
16 0.0580 fibers/cc-year, respectively. This BMC and BMCL are quite similar to the values
17 obtained in the analysis for the RfC and provide important support for the selected modeling
18 approach. When time from first exposure is set at 40 years, the calculated RfC is
19 4 x io~6 fibers/cc.
20 Confidence in the principal study is considered medium. The data used are human,
21 epidemiological data which are preferred to animal bioassays, and the principal study is
22 conducted in a population of occupationally exposed workers with long-term, relatively low
23 intensity exposures. However, use of the sub-cohort resulted in a smaller data set, and fewer
24 cases to model. Additionally there are weaknesses in the primary study. Exposure estimates are
25 based on self-reported job histories. The study used a cross-sectional design and may be
26 negatively biased as individuals with more severe disease could have left employment or may
27 have died and not been included in the follow-up study, resulting in an underestimation of
28 overall toxicity. However, for a less severe effect, such as LPT, this bias should be minimal. As
29 discussed in Sections 4.1.3 and 5.2.1.3.2, there may have been potential for selection bias due to
30 exposure-dependent censoring in this population, based on information provided by Rohs et al.
31 (2008) regarding the higher average exposure in participants compared to nonparticipants. In
32 terms of sensitivity of the study to detect a health effect, it is known that high-resolution
This document is a draft for review purposes only and does not constitute Agency policy.
6-11 DRAFT—DO NOT CITE OR QUOTE
-------
1 computed tomography can identify mineral fiber-related lesions in the respiratory tract, which
2 cannot be identified by standard radiographs (Muravov et al., 2005; ATS, 2004; Staples et al.,
3 1989). Thus, the technology employed for determining the prevalence of radiograph!c changes
4 in the Marysville cohort may underestimate the actual prevalence of localized pleural thickening.
5 Confidence in the database is low-to-medium. The database contains long-term mortality
6 and morbidity studies in humans exposed via inhalation to Libby Amphibole asbestos. The
7 morbidity studies do provide appropriate data for RfC derivation for pleural and lung
8 abnormalities. However, although decreased pulmonary function, a potential for autoimmune
9 effects, and cardiovascular disease are noted in exposed individuals, data do not provide an
10 exposure-response relationship. It is known that inhaled asbestos fibers migrate out of the lung
11 and into other tissues (see Section 3.1), lending uncertainty to any assumptions that other effects
12 would not be expected. There are no data in laboratory animals or humans on general systemic
13 effects. Therefore, overall confidence in the RfC is low-to-medium, reflecting medium
14 confidence in the principal study and low-to-medium confidence in the database.
15 Uncertainty and Sensitivity Analyses for RfC Derivation: It is important to consider the
16 sources of uncertainties in the derivation of the RfC for Libby Amphibole asbestos. These
17 include the following:
18 Measurement error in exposure assessment and assignment. The estimated exposure for
19 each individual relied on self-reported employment history, which may be subject to recall error.
20 Only data from 1972 and later were used for an RfC derivation, based on lack of fiber
21 measurements prior to this date; although better there remains some uncertainty in exposures
22 prior to installation of IH controls (1974). There is also uncertainty in the post-1972 data
23 regarding asbestos content in other ore sources (Virginia, South Carolina, and South Africa).
24 Although Libby Amphibole asbestos was not used in the facility after 1980, industrial hygiene
25 measurements collected after 1980 showed low levels of fibers. However, because the
26 concentration of fibers in the workplace was near background after 1980, this exposure makes
27 only a small contribution to an individual's cumulative exposure estimate. Similarly, any
28 exposure to Libby Amphibole asbestos outside of the workplace is not likely to contribute
29 significantly to cumulative exposure 10% of workers reported bringing raw vermiculite
30 home, and the majority showered and changed clothes before leaving the workplace.
This document is a draft for review purposes only and does not constitute Agency policy.
6-12 DRAFT—DO NOT CITE OR QUOTE
-------
1 Radiographic assessment of localized pleural thickening. Conventional radiographs—
2 rather than the more sensitive high-resolution computed tomography—were used to determine
3 the health outcome. Localized pleural thickening may be difficult to detect on these radiographs,
4 leading to the potential for outcome misclassification. However, uncertainty in the detection of
5 LPT in each individual is considered minimal due to the use of a team of highly qualified chest
6 radiologists evaluating the radiographic films and the use of consensus diagnosis.
7 Length of follow-up. Time from first exposure to X-ray was 23.2-32.7 years in the
8 preferred sub-cohort (mean of 28.2 years). The literature suggests that the prevalence of LPT
9 may increase with time, beyond this observed range of time from first exposure. The lack of
10 observed data beyond -30 years after first exposure (on average) is a source of uncertainty when
11 characterizing the exposure-response relationship for a full lifetime of exposure (e.g., 70 years).
12 This likelihood that the prevalence of localized pleural thickening may increase further with time
13 beyond 30 years after first exposure, and lack of data to support characterization of the
14 exposure-response curve outside this range, is a principal rationale cited for the selection of a
15 database UF of 10 for an RfC derivation.
16 Background rate of localized pleural thickening. In the derivation of the RfC, a
17 background rate of 1% for LPT was used. Previous studies have reported a range of prevalence
18 estimates (0.02 to 3.9%) in populations not known to be occupationally exposed to asbestos.
19 However, in statistical modeling of the Marysville, OH sub-cohort, uncertainty in the
20 background rate of localized pleural thickening is very low. The difference in the POD when the
21 background rate is fixed at 1% versus when it is estimated (estimated background rate of 3.12%)
22 is -15% (0.1177 compared to 0.1349 fibers-year/cc), and it does not affect the proposed RfC
23 (after rounding to one significant digit).
24 Model Form. A number of model forms were explored in the initial stages of analysis
25 (see Appendix E) before selecting the Michaelis-Menten model. The BMC and the BMCL
26 estimated from other candidate models for the sub-cohort, as well as those obtained in modeling
27 from the full cohort were in a similar range to the selected model. A second model-based
28 uncertainty is the choice of lag for cumulative exposure. The RfC derivation is based on the
29 exposure lagged by 10 years, since this lag yielded the lowest Akaike Information Criterion
30 (AIC) value, and indication of superior fit. However, if other lags (with similar AICs) are used,
31 the difference in POD may fluctuate to be approximately 20% higher or approximately 55%
This document is a draft for review purposes only and does not constitute Agency policy.
6-13 DRAFT—DO NOT CITE OR QUOTE
-------
1 lower. However, the choice of lag does not affect the proposed RfC (after rounding to one
2 significant digit).
3 Effect of smoking. Information on ever/never smoking was available for the preferred
4 sub-cohort. This individual variable did not meet statistical significance in the best-fitting
5 model, although inclusion did improve model fit (see Appendix E). When including smoking in
6 the best-fitting model, BMCs and BMCLs estimated separately for smokers and nonsmokers
7 differed by approximately sixfold. Smoking was not included in the model selected for RfC
8 derivation due to the lack of statistical significance, limited sample size (only three cases were
9 never smokers out of a total of 12 cases), and lack of detailed information on smoking history,
10 but these sensitivity analyses indicate a need for further research on the effect of smoking in
11 relation to LPT risk among asbestos-exposed populations.
12 Sensitivity analysis for the derivation of a POD for lifetime exposure from the CE metric
13 of the worker cohort. Exposure-response modeling for LPT in the Marysville sub-cohort used
14 the cumulative exposure (CE) metric (represented as CHEEC, described in Section 5.2.3.1)
15 providing a POD in fibers/cc-years. In order to derive an RfC in the units of continuous air
16 concentration for a lifetime (i.e., fibers/cc), the POD from the CE metric was weighted across a
17 lifetime exposure. Thus, the lifetime BMCLio is 1.96 x 1CT3 [0.1177 (fibers/cc)-years ^
18 60years]. This procedure is one way to account for the duration of exposure in the occupational
19 study being less than lifetime. There is some uncertainty as to whether—and how—to take
20 account for less-than-lifetime exposure in the occupational cohort. A sensitivity analysis was
21 done to consider other procedures for this averaging. The primary analysis assumes duration
22 contributes to risk and thus calculates a concentration across a lifetime that would yield the POD.
23 The second analysis is consistent with assuming duration contributes to risk but estimating the
24 concentration only for the mean duration in the observed database. The third analysis assumes
25 duration does not contribute to risk and models the average work duration continuous exposure
26 equivalent for each worker. This sensitivity analysis indicates that the approach taken to average
27 the POD based on the CE metric (CHEEC) across a lifetime was a reasonable approach, as
28 similar results are obtained using different approaches (i.e., within 4 fold).
29 Choice of critical effect. The critical effect selected for RfC derivation is localized
30 pleural thickening. Alternative endpoints were not modeled using the preferred sub-cohort due
31 to small numbers—there were five cases of bilateral LPT, only one case of diffuse pleural
This document is a draft for review purposes only and does not constitute Agency policy.
6-14 DRAFT—DO NOT CITE OR QUOTE
-------
1 thickening, and no individuals with interstitial changes. As a sensitivity analysis, these three
2 alternative endpoints (along with all LPT) were modeled among all Marysville workers not
3 previously exposed to other forms of asbestos, with X-rays performed in 2002-2005 (n = 250).
4 These analyses were performed using the Michaelis-Menten model with a background rate of 1%
5 and unlagged CHEEC as the exposure metric. BMRs of 1, 5, and 10% were investigated (see
6 Table 5-5). Use of the 10% BMR for these alternative endpoints allows for comparison to a
7 POD based on the selected critical effect of LPT. In this larger cohort, the POD for a
8 10% increase in LPT was 0.06 fibers/cc-years (in comparison with the POD derived from the
9 sub-cohort and used in RfC derivation of 0.1177 fibers/cc-years). Results for all pleural
10 thickening (LPT and DPT) did not differ from results for LPT. Bilateral localized pleural
11 thickening was included as a rough indication of increased severity within LPT, and as expected
12 results in higher PODs at each BMR than LPT. The resulting BMCLs for DPT and small
13 opacities (1.17 and 2.89 fibers/cc-years respectively, 10% BMR) are higher than the POD for
14 LPT (0.06 fibers/cc-years). Thus, use of an alternative endpoint at the same BMR would provide
15 a higher POD, and corresponding higher RfC.
16 However, a 10% BMR is not appropriate for more severe endpoints and BMCLs are
17 calculated at 1 and 5% BMRs as well. If DPT is used as a critical effect, PODs of 0.081 and
18 0.473 fib ers/cc-y ears would be calculated for a 1% and 5% BMR respectively. If small opacities
19 are used as a critical effect, the PODs are higher at both a 1% and a 5% BMR (i.e., 0.243 and
20 1.32, respectively). In summary, the use of more severe alternative endpoints (with appropriate
21 BMRs) results in PODs higher than that estimated using the critical effect of LPT (i.e.,
22 0.06 fib ers/cc-y ear, BMR 10%), and all are higher than the POD used in RfC derviation, with the
23 exception of DPT at a 1% BMR (0.0814 fib ers/cc-y ear). BMCLs for these more severe
24 endpoints using a 1% BMR were within ~2-fold of the preferred POD (0.0814 and
25 0.2425 fib ers/cc-y ear for diffuse pleural thickening and interstitial changes, respectively). There
26 is uncertainty associated with these estimates due to the inclusion of individuals hired before
27 1972, when no quantitative exposure measurements were available. Thus, a choice of alternative
28 critical effects—even with lower BMRs—would not result in an RfC appreciably lower than the
29 proposed RfC based on LPT and a 10% BMR.
30
This document is a draft for review purposes only and does not constitute Agency policy.
6-15 DRAFT—DO NOT CITE OR QUOTE
-------
1 6.2.2. Cancer/Inhalation
2 6.2.2.1. Background and Methods
3 The most appropriate data set for deriving quantitative cancer risk estimates based on
4 Libby Amphibole asbestos exposure in humans is the cohort of workers employed at the
5 vermiculite mining and milling operation near Libby, MT (see Section 4.1). No data were
6 available pertaining to cancer incidence or mortality in the Marysville, OH cohort, and mortality
7 and exposure data for other populations exposed to Libby Amphibole asbestos are very limited.
8 Whitehouse et al. (2008) provided detailed information on 11 mesothelioma cases among
9 nonworkers, but this information could not be used in exposure-response analyses for this
10 assessment, because there is no quantitative exposure information for these cases and no
11 information on the population from which these cases arose.
12 The Libby, MT worker cohort has been the focus of two epidemiologic investigations by
13 National Institute for Occupational Safety and Health (NIOSH) scientists. A database created by
14 NIOSH in the 1980s contains demographic data, work history, and vital status at the end of May
15 of 1982 for 1,881 workers at the vermiculite mine, mill, and processing plant in Libby, MT (see
16 Section 4.1.1.1). Vital status follow-up was completed by NIOSH through 2006 using the
17 National Death Index (Bilgrad, 1997). Nearly 54% of workers in the cohort (n = 1,009) had died
18 by December 31, 2006. The data from this update (provided by NIOSH) is the basis of the EPA
19 exposure-response modeling.
20 EPA does not have sufficient biological information to select models for the
21 epidemiology data on the basis of biological mechanism (see Section 5). In this situation, EPA's
22 practice is to investigate a range of model forms to determine how to best empirically model the
23 exposure-response relationship in the range of the observed data. In this case, different exposure
24 metrics were explored for model fit in the analytic models. The exposure metric options were
25 selected to provide a range of shapes that was sufficiently flexible to allow for a variety of ways
26 that time and duration might relate to cancer risk in the data being modeled. EPA then evaluated
27 how well the models and exposure metric combinations fit the data being modeled. Metrics that
28 did not fit the data well were rejected. For purposes of calculating a reasonable upper-bound on
29 the risk per exposure EPA accounted for uncertainty in the choice of exposure metrics by using
30 the exposure metric (among those of reasonable fit) that estimated the highest risk. This is
31 explained in more detail below and in Sections 5.4.3-5.4.5. However, there are other
This document is a draft for review purposes only and does not constitute Agency policy.
6-16 DRAFT—DO NOT CITE OR QUOTE
-------
1 uncertainties in the modeling of the epidemiological data that may impact the IUR and these are
2 described in detail in Section 5.4.6.
3 Cumulative exposure has been the traditional method of measuring exposure in
4 epidemiologic analyses of many different occupational and environmental exposures and was the
5 exposure metric applied for to the risk of lung-cancer mortality in the EPA general asbestos
6 evaluation (U.S. EPA, 1988a). Two alternative approaches to developing exposure metrics to
7 describe the effects of air concentrations of asbestos dust in the air on the risks of mortality have
8 also been proposed. The first alternative was proposed by Jahr (1974) who studied
9 silica-induced pneumoconiosis. He also suggested that exposures to occupational dusts could be
10 weighted by the time since exposure yielding an exposure metric which gives greater weight to
11 earlier exposures. Berry et al. (1979) subsequently suggested the application of exposure metrics
12 that allowed for the clearance of dust or fibers by using a decay term on exposures. For the
13 evaluation of mortality risk from mesothelioma, U.S. EPA (1988a) used a different exposure
14 metric than was used for lung-cancer mortality, which factored in the time since first exposure.
15 It is important to note that different characterizations of ambient exposures may be reasonably
16 expected to be associated with different endpoints (i.e., lung cancer or mesothelioma).
17 In the Libby, MT worker cohort data developed by NIOSH and used by the EPA in this
18 assessment, detailed work histories, together with job-specific exposure estimates, allowed for
19 the reconstruction of each individual's occupational exposure experience over time to define
20 multiple exposure metrics. From this information-rich individual-level data set from NIOSH, the
21 EPA constructed a suite of the different metrics of occupational exposure which had been
22 proposed in the asbestos literature or used in the EPA health assessment on general asbestos
23 exposures (U.S. EPA, 1988a). This suite of models was defined a priori to encompass a
24 reasonable set of proposed exposure metrics to allow sufficient flexibility in model fit to these
25 data. These exposure metrics were evaluated in analytic-regression models to test which
26 exposure metrics were the best empirical predictors of observed cancer mortality and the better
27 fitting models were advanced for consideration as the basis of the exposure-response relationship
28 for the IUR. The types of exposure metrics evaluated were intended to allow for variations of
29 the classic metric of cumulative exposure, allowing for more or less weight to be placed on
30 earlier or later exposures. These simulated exposure metrics were derived mathematically to
31 approximate underlying processes that are not well understood, and their fit is evaluated on the
This document is a draft for review purposes only and does not constitute Agency policy.
6-17 DRAFT—DO NOT CITE OR QUOTE
-------
1 basis of maximizing the likelihood for the workers cohort and estimated parameters does not
2 necessarily have biological interpretation (see Section 5.4.2.5 for details).
3 Exposure estimates for all exposure metrics were adjusted to account for the time period
4 between the onset of cancer and mortality. The lag period defines an interval before death, or
5 end of follow-up, during which, any exposure is excluded from the calculation of the exposure
6 metric. Modeling of mesothelioma mortality included two additional exposure metrics: duration
7 of exposure and the exposure metric including a cubic function of time (see Eq. 5-5), originally
8 proposed in Peto et al. (1982) and employed in derivation of the IUR for asbestos (U.S. EPA,
9 1988a,
10 Analyses of mesothelioma mortality were conducted using a Poisson model with a
11 Markov chain Monte Carlo (MCMC) Bayesian approach, whereas analyses of lung-cancer
12 mortality were conducted using the Cox proportional hazards model with time-varying
13 exposures. There was one important limitation of the NIOSH job exposure matrix (JEM). Of
14 the 991 workers hired before 1960, 706 workers with unknown department code and unknown
15 job assignments hired between 1935 and 1959 were assigned the same average estimated
16 exposure intensity. The lack of information on specific job assignments for such a large portion
17 of these early workers when exposures were higher resulted in the misclassification of the
18 exposure and effectively yielded exposure metrics that were differentiated only by the duration
19 of each worker's employment. For this reason and because there was little measured fiber
20 exposure data during the earlier period, identifying an adequate exposure-response model fit was
21 unsuccessful. The two biggest problems were that the duration of employment was the
22 best-fitting metric for modeling mesothelioma and that the Cox model assumptions were violated
23 in modeling lung-cancer mortality (see Section 5.4.3.5). As a result, this assessment developed a
24 sub-cohort analysis by dividing the whole cohort into two groups: those hired prior to 1960 and
25 those hired after 1959. This removed all but nine cohort members with missing department code
26 and job category information and lessened the effect of estimates of early exposures where no air
27 sampling data were available. For the sub-cohort of those hired after 1959, those two biggest
28 problems were resolved: the assumptions of the Cox model were satisfied, and a lagged
29 cumulative exposure with a decay (rather than duration of exposure, as for the full cohort) was
30 the best-fitting metric for mesothelioma.
This document is a draft for review purposes only and does not constitute Agency policy.
6-18 DRAFT—DO NOT CITE OR QUOTE
-------
1 Of the 880 workers hired after 1959, 230 (26%) had died by December 31, 2006. The
2 number of mesothelioma deaths in the sub-cohort is relatively small (n = 7, 2 deaths coded in
3 ICD-10 and 5 deaths coded in ICD-9), but the rate of mesothelioma mortality was very similar in
4 the subcohort (24.7 per 100,000 person-years vs. 26.8 per 100,000 person-years for the full
5 cohort [18 mesothelioma deaths], a difference of less than 10%).
6
7 6.2.3. Modeling of Mesothelioma Exposure Response
8 A Poisson model is employed for estimating the absolute risk of mesothelioma following
9 exposure to Libby Amphibole asbestos, as the Poisson distribution is an appropriate model for
10 use with data that are counts of a relatively rare outcome, such as observed mesothelioma deaths
11 in the Libby, MT worker cohort. Estimation of the exposure-response relationship for
12 mesothelioma using the Poisson model was performed in WinBUGS software by a MCMC
13 Bayesian approach with an uninformative (diffuse) prior. The model was run to fit the mortality
14 data to exposure data for various exposure metrics described above. To comparatively evaluate
15 how much better one model fits than another, the Deviance Information Criterion (DIG) was
16 used. DIG is used in Bayesian analysis and is an analogue of AIC (Burnham and Anderson,
17 2002). Use of the DIG and AIC is standard practice in comparing the fit of nonnested models to
18 the same data set with the same dependent outcome variable but different independent
19 covariates.
20 Two cumulative exposure metrics with decay provided the best model fits. Both metrics
21 had a common 5-year half life, with lag times of either 10 or 15 years. In the sub-cohort hired
22 after 1959, the DIG value for mesothelioma using the IRIS IUR (U.S. EPA. 1988a) metric (see
23 Eq. 5-5) is substantially higher (DIG = 98.4) than for any of the metrics in Table 5-10, where the
24 lowest DIG is 70.6. This difference of over 20 DIG units, is an indication that the model used for
25 mesothelioma in the U.S. EPA (1988a) IUR derivation (see Eq. 5-5), does not fit these data from
26 the Libby, MT work cohort, compared to other exposure metrics presented (see Table 5-10). It
27 should be noted that the data modeled here are very different from the data on which the IRIS
28 assessment for asbestos (U.S. EPA, 1988a) is based—and one does not necessarily expect the
29 same model to fit different data sets—this is why EPA goes through a process to determine the
30 best-fitting model in each case. One difference with the IRIS IUR (U.S. EPA. 1988a) modeling
31 is that the analysis in this assessment is based on individual-level data, whereas the IRIS IUR
This document is a draft for review purposes only and does not constitute Agency policy.
6-19 DRAFT—DO NOT CITE OR QUOTE
-------
1 (U.S. EPA, 1988a) application was to aggregate data. Also, cohorts used in the IRIS IUR (U.S.
2 EPA, 1988a) did not include cohorts exposed to Libby Amphibole asbestos and Libby
3 Amphibole asbestos may be different from other types of asbestos. Alternately, the relative fit of
4 this model may have been affected by uncertainties in the estimated exposure described in detail
5 in Section 5.4.6.
6 As it is less likely that exposure during the last few years before death were contributory
7 to the development of the cancer and cancer mortality, the zero lag metrics were dropped from
8 further consideration. All eight models retained for derivation of IUR include a decay half-life in
9 the exposure metric. For the sub-cohort hired after 1959, the best-fitting exposure metric was
10 cumulative exposure with a 5 year half-life and a 15 year lag time with a central estimate for the
11 P of 2.07 x l(T4 with 95% upper confidence limit (UCL) of 3.42 x 1(T4.
12
13 6.2.4. Unit Risk Estimates for Mesothelioma Mortality
14 The increased risk of mesothelioma mortality attributable to continuous fiber exposure
15 was estimated using a life-table procedure based on the general U.S. population. The life-table
16 procedure involved the application of the estimated Libby Amphibole asbestos toxicity to a
17 structured representation of the general U.S. population in such a manner as to yield age-specific
18 risk estimates for cancer mortality in the presence or absence of exposure to Libby Amphibole
19 asbestos (see Section 5.4.5; Appendix G).
20 A default linear low-dose extrapolation below the POD was used because the mode of
21 action by which Libby Amphibole asbestos causes mesothelioma cannot be established. The
22 lower limit on the effective concentration (LECoi) for adult-only exposures was determined to be
23 0.245 fibers/cc, which yielded an adult-based unit risk for mesothelioma mortality of 0.053 per
24 fiber/cc (POD of 1% divided by the LECoi).
25 The value of the effective concentration (EC) that would correspond to the measure of
26 central tendency is the ECoi. This value is used in the derivation of a combined risk of
27 mesothelioma and of lung cancer. The ECoi was determined to be 0.406 per fiber/cc, which
28 when divided into a POD of 1% and scaled (by 70/54) to encompass the whole lifespan, gives a
29 lifetime central estimate value of 0.032 per fiber/cc.
30 For mesothelioma, the undercounting of cases (underascertainment) is a particular
31 concern given the limitations of the International Classification of Diseases (ICD) classification
This document is a draft for review purposes only and does not constitute Agency policy.
6-20 DRAFT—DO NOT CITE OR QUOTE
-------
1 systems used prior to 1999. In practical terms, this means that some true occurrences of
2 mortality due to mesothelioma are missed on death certificates and in almost all administrative
3 databases such as the National Death Index. Even after introduction of special ICD code for
4 mesothelioma with introduction of ICD-10 in 1999, detection rates are still imperfect (Camidge
5 et al., 2006; Pinheiro et al., 2004) and the reported numbers of cases typically reflect an
6 undercount of the true number. Kopylev et al. (2011) reviewed the literature on this
7 underascertainment and developed methods to account for the likely numbers of undocumented
8 mesothelioma deaths.
9 To compensate for mesothelioma underascertainment attributable to ICD coding, the
10 mesothelioma mortality unit risk was further adjusted following the analysis of Kopylev et al.
11 (2011). The adjusted mesothelioma central (i.e., maximum likelihood estimate) risk,
12 corresponding to the best-fit metric, was 0.044 per fiber/cc, and the adjusted mesothelioma
13 mortality unit risk was 0.074 per fiber/cc. The adjusted mesothelioma mortality unit risks from
14 all eight exposure parameterization models with adequate fit produced a range of unit risk values
15 (see Table 5-17) from 0.044 to 0.122 Thus, there is uncertainty in mesothelioma risks generated
16 from similar-fitting models from different exposure metrics (see details in Section 5.4.6.1.3).
17
18 6.2.5. Modeling of Lung Cancer Exposure Response
19 All multivariate extended Cox models were fit to the sub-cohort hired after 1959 with
20 covariates for sex, race, and date of birth, and exposure. Exposure for each of the 40 exposure
21 parameterizations was calculated independently and fit of these exposure metrics was evaluated
22 one at a time. As the exposure-response models cannot strictly be considered to be nested, a
23 standard measure of fit, the AIC (Burnham and Anderson, 2002), was used for comparison of
24 model fit with smaller values of AIC, indicating better goodness of fit. Of the
25 40 exposure-response metrics, 14 demonstrated an adequate fit to the data as measured by the
26 overall model fit with the likelihood ratio test (p < 0.05) as well as having statistically significant
27 exposure metrics (p < 0.05). However, only the nine models that demonstrated adequate model
28 and exposure metric fit and incorporated a lag period to account for cancer latency were
29 considered further in the development of the IUR (see Table 5-18).
30 Lagging exposure by 10 years was a better predictor of lung-cancer mortality compared
31 to other lags. As it is less likely that exposure during the last few years before death were
This document is a draft for review purposes only and does not constitute Agency policy.
6-21 DRAFT—DO NOT CITE OR QUOTE
-------
1 contributory to the development of the cancer and cancer mortality, the zero lag metrics were
2 dropped from further consideration. The residence time-weighted cumulative exposure, both
3 with and without decay of the exposure metric, did not fit these lung-cancer mortality data well
4 compared to the other models (see Table 5-12); this form of exposure metric does not
5 demonstrate evidence of an empirical fit to these epidemiologic data.
6 The model with the smallest AIC was for cumulative exposure with a 10-year half-life for
7 decay and a 10-year lag for cancer latency. The extended Cox model estimated a slope (beta) of
8 1.26 x 1CT2 per fiber/cc-year based on a 365-day year, and the 95 percentile upper bound was
9 1.88 x 1CT2 per fiber/cc-year. Thep-va\ue for the Libby Amphibole asbestos regression
10 coefficient (slope) was <0.001. The slopes and confidence interval for the other exposure
11 metrics, which had similar fits to these data are reported in Table 5-13. Uncertainty in the choice
12 of the exposure metric (cross-metric uncertainty) is considered in the derivation of the final unit
13 risk (see details in Section 5.4.5.3), representing the range of unit risks that are derived from
14 these similarly fitting metrics. The model results that were ultimately selected to reflect the
15 upper-bound among the range of results were based on the cumulative exposure with a 10-year
16 lag exposure metric (CE10). The extended Cox model estimated a slope (beta) of
17 5.28 x io~3 per fiber/cc-year based on a 365-day year, and the 95th percentile upper bound was
18 1.00 x 1CT2 per fiber/cc-year.
19
20 6.2.6. Unit Risk Estimates for Lung-Cancer Mortality
21 The increased risk of lung-cancer mortality attributable to continuous fiber exposure was
22 estimated using a life-table procedure based on the general U.S. population. The life-table
23 procedure involved the application of the estimated Libby Amphibole asbestos-specific toxicity
24 to a structured representation of the general U.S. population in such a manner as to yield
25 age-specific risk estimated for cancer mortality in the presence or absence of exposure to Libby
26 Amphibole asbestos (see Section 5.4.5; Appendix G).
27 The nine exposure-response models retained in Table 5-13 all had reasonably similar
28 goodness of fits. No single model stands out as clearly statistically superior; however, there is a
29 range of quality of fit within the set that could be considered to have adequate fit. The
30 lung-cancer mortality unit risks are shown in Table 5-18.
This document is a draft for review purposes only and does not constitute Agency policy.
6-22 DRAFT—DO NOT CITE OR QUOTE
-------
1 Using the results of the exposure model with the lowest AIC value (i.e., cumulative
2 exposure with a 10-year half-life for decay and a 10-year lag for cancer latency) alone, the LECoi
3 for the adult-only exposures was determined to be 0.333 fibers/cc. This yields an adult-based
4 unit risk of lung-cancer mortality of 0.0300 (POD of 1% divided by the LECoi). This estimate
5 was then scaled by 70/54 to encompass the whole lifespan; it yielded a lifetime unit risk of
6 0.0389 per fiber/cc. The value of the concentration that would correspond to the measure of
7 central tendency was based on the ECoi rather than LECoi. The ECoi for the adult-only
8 exposures was determined to be 0.499 per fiber/cc, which, when divided into a POD of 1%,
9 yielded an adult-based central estimate for lung-cancer mortality of 0.0200. This estimate was
10 then scaled by 70/54 to encompass the whole lifespan to, yielded a lifetime central estimate of
11 0.0260 per fiber/cc.
12 Using the results of the exposure model based on cumulative exposure with a 10-year lag
13 for cancer latency, the LECoi for the adult-only exposures was determined to be 0.191 per
14 fibers/cc, yielding an adult-based unit risk of lung-cancer mortality of 0.0524 (POD of 1%
15 divided by the LECoi). When scaled by 70/54 to encompass the whole lifespan, it yielded a
16 lifetime unit risk of 0.0679 per fiber/cc. The value of the risk that would correspond to the
17 measure of central tendency involves the ECoi rather than the LECoi. The ECoi for the
18 adult-only exposures was determined to be 0.325 per fiber/cc, which, when divided into a POD
19 of 1%, yielded an adult-based central estimate for lung-cancer mortality of 0.0308. This estimate
20 was then scaled by 70/54 to encompass the whole lifespan to, yielded a lifetime central estimate
21 of 0.0399 per fiber/cc.
22 The resulting unit risks in Table 5-18 ranged from 0.0260 to 0.0679 per fibers/cc, for a
23 lifetime continuous exposure. This shows that the unit risk based on the exposure metric with
24 the lowest AIC value (i.e., cumulative exposure with a 10-year half-life for decay and a 10-year
25 lag for cancer latency) is in the center of this range (i.e., 0.0389 per fiber/cc). This estimate is in
26 the middle of the range of possible unit risks and does not capture the uncertainty across metrics
27 with similar goodness of fit (see details in Section 5.4.6.1.3).
28 The model results selected to represent the upper bound risk among the range of
29 reasonable results are based on CE10 metric with a 10-year lag. The model results selected to
30 reflect the upper-bound among the range of results are based on the CE10 exposure metric with a
31 10-year lag, providing an IUR of 0.0679 per fibers/cc.
This document is a draft for review purposes only and does not constitute Agency policy.
6-23 DRAFT—DO NOT CITE OR QUOTE
-------
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
6.2.7. IUR Derivation Based on Combined Mesothelioma and Lung-Cancer Mortality from
Exposure to Libby Amphibole Asbestos
When risks are combined, it is important to understand several concepts that are pertinent
to the evaluation and comparison of the cancer-specific mortality unit risks that will be
combined. First, there is statistical uncertainty in the potency estimate within the exposure
response model defined by each exposure metric. This within metric uncertainty is accounted
for in the confidence interval on slope. Next, there is an uncertainty in the choice of metrics for
developing IUR (cross-metric uncertainty). Finally, when unit risks corresponding to metrics are
chosen accounting for uncertainty, these are statistically combined into IUR. Details are
provided in Section 5.4.5.3.
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.
Table 6-1. Reasonable upper bound and lowest information criteria
estimates of central risks and unit risks, for mesothelioma mortality,
lung-cancer mortality, and the IUR for the combined mortality risk from
mesothelioma and lung cancer (lURs are presented in the units of excess
cancers per fibers/cc, continuous lifetime exposure)
Model
Reasonable upper bound3
Lowest information criteria13
Mesothelioma
Central
estimate
0.075
0.044
Unit risk
0.122
0.074
Lung cancer
Central
estimate
0.040
0.026
Unit risk
0.068
0.040
Combined mesothelioma
and lung cancer
Central
estimate
0.115
0.070
IUR
0.169
0.103
24
25
26
27
28
29
30
31
32
Tor mesothelioma, the selected model parameterized exposure as cumulative exposure with exponential decay
half-life of 5 years and a 15-year lag. For lung cancer, the selected model parameterized exposure as cumulative
exposure without decay and a 10-year lag.
bFor mesothelioma, the selected model parameterized exposure as cumulative exposure with exponential decay
half-life of 5 years and a 10-year lag. For lung cancer, the selected model parameterized exposure as cumulative
exposure with exponential decay half-life of 10 years and a 10-year lag.
This document is a draft for review purposes only and does not constitute Agency policy.
6-24 DRAFT—DO NOT CITE OR QUOTE
-------
1 6.2.7.1. Comparison with Other Published Studies ofLibby, MT Workers Cohort
1 Several published studies have previously evaluated risk of mesothelioma and lung
3 cancer (i.e., Larson et al., 201 Ob: Moolgavkar et al., 2010; Berman and Crump, 2008; Sullivan,
4 2007) in Libby, MT workers cohort.
5 For mesothelioma, only Moolgavkar et al. (2010) provided an exposure-response
6 relationship for absolute risk of mesothelioma mortality that would be comparable with this
7 current assessment. Based on the full cohort, with mortality data through 2001 and a
8 modification of the Peto/Nicholson exposure metric, life-table analysis would provide an upper
9 bound unit risk of approximately 0.13 per fibers/cc continuous lifetime exposure. Therefore,
10 utilization of the exposure response modeling of Moolgavkar et al. (2010), would provide an
11 IUR for excess mesothelioma mortality in close agreement with the IUR derived in this
12 assessment (see Section 5.4.5.3.1 for more details).
13 For lung cancer, all of the studies provide exposure-response relationships in terms of
14 relative risk of lung-cancer mortality and, thus, may provide risk estimates comparable number
15 to this assessment. However, inclusion criteria, length of mortality follow-up, and analytic
16 methods differ among the analyses—thus, the results are not necessarily interchangeable. For
17 comparison purposes, the lung cancer unit risk from these studies are computed from life-table
18 analyses (see Table 5-20). The lung cancer lURs calculated based on the published literature,
19 ranged from 0.010 to 0.079 per fiber/cc (based on the upper-confidence limit). This is in close
20 agreement with this current assessment where an upper-bound estimate of 0.068 per fiber/cc,
21 continuous lifetime exposure is derived (see Section 5.4.5.3.1 for more details).
22
23 6.2.8. Sensitivity Analysis
24 6.2.8.1. Sensitivity Analysis of Influence of High Exposures in Early 1960s on Model Fit
25 Although data do not exist to evaluate biological bases for model fit, other potential
26 factors can be explored where data allow. For example, because of concerns that very high
27 (>100 fibers/cc) early (1960-1963) 8-hour Libby Amphibole asbestos TWA exposures (see
28 Table 5-7) could have influenced the relative fit of the various exposure metrics, EPA conducted
29 a sensitivity analysis of the impact on the relative model fit of reducing all estimated exposure
30 intensities for 1960-1963 by 50%.
This document is a draft for review purposes only and does not constitute Agency policy.
6-25 DRAFT—DO NOT CITE OR QUOTE
-------
1 For modeling mesothelioma and lung-cancer mortality on this revised data set, there was
2 very little difference in the order of the relative fits of the exposure models as was seen for the
3 subcohort of workers hired after 1959 and included the exposures as estimated by Amandus et al.
4 (1987a) during 1960-1963 (see Tables 5-14 and 5-15). The models based on the revised data set
5 fit approximately as well for mesothelioma and for lung cancer.
6
7 6.2.8.2. Analysis of Potential Confounding of Lung Cancer Results by Smoking in the
8 Sub-cohort
9 EPA used three approaches to address the confounding issue, including restriction of the
10 cohort and two analytic evaluations of the potential for confounding by smoking including the
11 method described by Richardson (2010). Richardson (2010) describes a method to determine if
12 an identified exposure relationship with lung cancer is confounded by unmeasured smoking in an
13 occupational cohort study. EPA implemented this methodology to model the potential effects of
14 Libby Amphibole asbestos on the risk of COPD mortality on the subcohort of workers hired after
15 1959 (see Section 5.4.3.6.5). Summarizing these findings, EPA used the method described by
16 Richardson (2010) to evaluate whether exposures to Libby Amphibole asbestos predicted
17 mortality from COPD as an indication of potential confounding by smoking and found a
18 nonsignificant negative relationship, which was inconsistent with confounding by smoking.
19
20 6.2.9. Uncertainty in the Cancer Risk Values
21 It is important to consider the uncertainties in the derivation of the mesothelioma and
22 lung-cancer mortality risks in this assessment in the context of uncertainties in animal-based
23 health assessments. This assessment does not involve extrapolation from high dose in animals to
24 low dose in humans. The current assessment is based on a well-documented and well-studied
25 cohort of workers with adequate years of follow-up to evaluate mesothelioma and lung-cancer
26 mortality risks with PODs within the range of the data. The discussions in Section 5.4.6 explore
27 uncertainty in the derivation of the IUR in order to provide a comprehensive and transparent
28 context for the resulting cancer mortality risk estimates.
29 The summary below includes likely one-sided uncertainties (biases) associated with the
30 derivation of the IUR in order to provide a context for the resulting cancer risk estimates.
This document is a draft for review purposes only and does not constitute Agency policy.
6-26 DRAFT—DO NOT CITE OR QUOTE
-------
1 The sources of uncertainty that could lead to a likely underestimation of the cancer risk
2 value include the following:
5 • Use of historical phase contrast microscopy (PCM) exposure measurements. As asbestos
6 was a contaminant of vermiculite that was the primary object of production, mine and dry
7 and old wet mill ambient air may have contained material other than asbestos that could
8 have contributed to fibers counted by PCM. Therefore, it is possible that exposure
9 estimates for some or possibly a large portion of the cohort are overestimated, and,
10 therefore, the resulting IUR may be underestimated.
11 • Measurement error in exposure assessment and assignment. This current assessment
12 showed that unit risk results from analysis of the lung-cancer mortality in the full cohort
13 (see Table 5-21) compared with the sub-cohort hired after 1959 may have been
14 attenuated as much as 2-6 times (see Section 5.4.6.1.2.4). By excluding those cohort
15 members hired before 1960 for whom there was insufficient work history information to
16 estimate their exposures, the unit risk for lung cancer was less attenuated due to exposure
17 measurement error. However, exposure measurements from the 1960s are also imperfect
18 and include a lesser degree of exposure measurement error, which could have led to
19 underestimated risk—even in the sub-cohort hired after 1959.
20 • Limited length of follow-up. The IUR for mesothelioma mortality could be larger than
21 was estimated from existing data, since latency of mesothelioma can be as long as
22 60 years. The maximum length of follow-up was 46 years in this cohort. The magnitude
23 of underestimation would depend on the relationship between the number of additional
24 deaths and the increase in person-years.
25 • Use of life-tables to calculate the IUR based on cancer mortality. The life-table
26 procedure computes the extra risk of death from birth up to 85 years of age. This cut-off
27 at age 85 ignores a small additional risk of lung-cancer mortality among a small
28 percentage of people who have a higher background risk because of the increase in lung
29 cancer risk that is seen with increasing age. The lung-cancer mortality unit risk based on
30 the LECoi would be somewhat larger, on the order of 5-10%. On the other hand, the
31 additional mesothelioma mortality risk, if the life-tables were extended to account for
32 longer life spans, would be about 3%.
33 • Small number of women and ovarian cancer. While asbestos is causally associated with
34 increased risks of ovarian cancer (Straif et al., 2009), there were only 84 women in the
35 whole cohort, and there were no deaths from ovarian cancer among 24 total deaths. The
36 lack of observed ovarian cancer in this cohort may be a function of the limited number of
37 female deaths in the cohort allowing for the possibility that exposure to Libby Amphibole
38 asbestos could result in increased risk of ovarian cancer. However, it was not possible to
39 estimate the magnitude of this underestimation on the total cancer risk.
This document is a draft for review purposes only and does not constitute Agency policy.
6-27 DRAFT—DO NOT CITE OR QUOTE
-------
1 • Dependent competing risks. Competing risk of mortality from other diseases related to
2 exposure may have resulted in underestimates of the risk of mortality from either
3 mesothelioma or lung cancer. The mean length of follow-up for the Libby, MT workers
4 who died of mesothelioma was to 30.1 years, and evidence exists (Bianchi and Bianchi,
5 2009: Suzuki and Yuen, 2001) that early deaths from other exposure-related causes could
6 have precluded an individual's risks of death from mesothelioma. However, it was not
7 possible to estimate the magnitude of this effect on the total cancer risk.
8
9
10 The source of uncertainty that could lead to a likely overestimation of the cancer risk
11 value:
12
13
14 • Potential residual confounding and effect modification. The unit risk of lung-cancer
15 mortality estimated herein, and the combined mesothelioma and lung-cancer mortality
16 IUR, would over-estimate the risk in any population that had a lower prevalence of
17 smoking than that of the Libby worker cohort. Because the Libby worker cohort had a
18 large prevalence of smokers and ex-smokers and no known nonsmokers developed lung
19 cancer, it is also possible that estimated risk for lung cancer is actually risk for an
20 interaction of lung cancer and smoking, and effects of smoking and asbestos are known
21 to be between additive and multiplicative (see Section 4). However, the company
22 imposed smoking ban, and the observation that there were many ex-smokers in the
23 cohort, would tend to lessen risks that would have occurred if these individuals continued
24 smoking.
25
26
27 6.3. APPLICATION OF THE LIBBY AMPHIBOLE ASBESTOS RFC AND IUR
28 6.3.1. Sites and Materials
29 This Libby Amphibole asbestos specific assessment is based on the evaluation of worker
30 cohorts, exposed to asbestos from a single mine in Libby, MT, and it is intended to allow for
31 estimates of the risk due to exposure to the asbestos fibers from that mine, or exposures to
32 asbestos fibers that arise from the management or use of the vermiculite ore and exfoliated
33 vermiculite from this mine. Therefore, it is appropriate to apply the Libby Amphibole
34 asbestos-specific RfC and/or IUR to sites which are believed to have been contaminated by these
35 materials when assessing risk from the amphibole fibers present from this contamination. This
36 may include sites where the ore was shipped or handled, where the vermiculite was exfoliated
37 and further processed, facilities which in other ways shipped or handled the exfoliated
38 vermiculite, where products containing the raw or exfoliated vermiculite were present, the
39 consumer products themselves (e.g., vermiculite attic insulation) and any waste streams from the
This document is a draft for review purposes only and does not constitute Agency policy.
6-28 DRAFT—DO NOT CITE OR QUOTE
-------
1 above processes which contain vermiculite and the related Libby Amphibole asbestos-fibers.
2 The assessment was derived from PCM measurements taken at the Libby, MT occupational sites
3 and the mixture of minerals found in those measurements. It does not estimate the risk
4 attributable to specific subsets of those fibers whether based on size, shape, or mineral
5 composition other than the limitations on size and shape reflected in the PCM methodology and
6 counting rules. As detailed in Section 2, the amphibole asbestos present in the mine, ore and
7 expanded vermiculite, does not fit cleanly into a single category of nomenclature for amphibole
8 minerals. Most Libby Amphibole fibers are classified as winchite (84%), with lesser amounts of
9 richterite (11%) and tremolite (6%), based on the nomenclature proposed by Leake et al. (1997).
10 There are also trace amounts of magnesioriebeckite, edenite, and magnesio-arfvedsonite present
11 in Libby Amphibole asbestos (Meeker et al., 2003). Within the 30 samples taken from the mine
12 the proportion of these minerals differed between samples (Meeker et al., 2003) and the relative
13 proportions of these species may have varied over time (as ore from different locations was
14 processed). This assessment estimates the risk of exposure to the varying range of mineral fiber
15 mixtures that result from material originating from the geological deposit, recognizing there is
16 variation and uncertainty as to variations in the exposure to the underlying cohort and complex
17 variation in settings to which these estimates will be applied.
18
19 6.3.2. Exposure Units for Libby Amphibole Asbestos
20 As with the IRIS assessment for asbestos (U.S. EPA. 1988aX the RfC and IUR specific to
21 Libby Amphibole asbestos are presented here as fibers/cc exposure continuous lifetime exposure,
22 where exposure measurements are based on analysis of air filters by PCM. Early PCM analytical
23 techniques did not have the same resolution as current analytical methods, and it is understood
24 that PCM data for the majority of the exposures characterized for the Libby, MT workers and
25 Marysville, OH workers would likely have a width resolution of 0.4-0.44 |im (Amandus et al.,
26 1987a: IPCS. 1986: Kendall and Skikne. 1980). Therefore, as with the IRIS assessment for
27 asbestos (U.S. EPA, 1988a), the dimensions of the PCM fibers for the Libby Amphibole asbestos
28 unit risk are defined as fibers >5 jim in length with an aspect ratio of 3:1 or greater and a width
29 >0.4 |im.
30 Environmental air sampling for asbestos is now often analyzed by transmission electron
31 microscopes (TEM) to confirm that the fibers viewed are asbestos, and often it is used to identify
This document is a draft for review purposes only and does not constitute Agency policy.
6-29 DRAFT—DO NOT CITE OR QUOTE
-------
1 the mineralogy of the fiber. Although some historical data do exist providing TEM analysis of
2 airborne fibers from the Libby, MT mill operation (McDonald et al., 1986a: Langer etal., 1974),
3 these data are not sufficient to provide an alternative set of exposure measurements in TEM units
4 for the Libby, MT worker cohort, or provide a PCM to TEM conversion across the various work
5 environments.
6 Different sampling environments and varied site conditions may pose the potential for
7 airborne fibers from various materials. Because of that, it is expected that for many
8 environmental risk assessments conducted now and in the near future, measures of exposure may
9 be done with methods such as TEM and then adjusted through fiber-counting rules to estimate
10 the number of PCM-countable asbestos fibers. Site-specific environmental conditions should be
11 considered in determining how to best identify PCM-countable asbestos fibers in relevant air
12 samples for exposure assessments used in conjunction with this health assessment to yield
13 estimates of risk.
14
15 6.3.3. Applications to Early Lifetime and Partial Lifetime Environmental Exposure
16 Scenarios for IUR
17 The Libby Amphibole asbestos-specific unit risk derived in this assessment is a combined
18 risk of lung cancer and mesothelioma, each with its own adjustment for uncertainty in metrics.
19 The life-table analyses for Libby Amphibole asbestos do not predict greater risk from early-life
20 exposures. Thus, this assessment recommends that estimates of the risks of less-than-lifetime
21 exposures be computed by simple calculations of average lifetime exposure concentration
22 multiplied by IUR. This recommendation is consistent with standard Superfund guidance, where
23 exposures are estimated, averaged across a lifetime exposure, and the IUR simply applied to
24 calculate excess cancer risk (U.S. EPA, 2008, 2001b). The weight of evidence does not support
25 a mutagenic mode of action for Libby Amphibole asbestos carcinogenicity. Therefore,
26 according to EPA's Supplemental Guidance for Assessing Susceptibility from Early-Life
27 Exposure to Carcinogens (U.S. EPA, 2005b), the application of the age-dependent adjustment
28 factors are not recommended.
29
This document is a draft for review purposes only and does not constitute Agency policy.
6-30 DRAFT—DO NOT CITE OR QUOTE
-------
1 6.3.4. Applications to Lifetime and Partial Lifetime Environmental Exposure Scenarios for
2 RfC
3 The Libby Amphibole asbestos specific RfC should be used to derive estimates of hazard
4 from exposure to airborne materials containing Libby Amphibole asbestos as described above.
5 The Libby Amphibole asbestos RfC was derived from an evaluation of the O.M. Scott,
6 Marysville, OH worker cohort (Rohs et al., 2008; Lockey et al., 1984). Exposure-response
7 modeling of cumulative Libby Amphibole asbestosis exposure with the best-fitting model
8 (Michaelis-Menten with 10-year lagged exposure) resulted in a BMCLio of 0.1177 fibers/cc-year
9 yielding an RfC for a 70-year lifetime of 2 x 10 5 fibers/cc by calculating the average
10 concentration over a 60-year averaging period (70 years minus 10-year lag).
11 The estimate of hazard should be calculated by dividing the average daily exposure
12 concentration using an averaging period of 60 years by the reference concentration outlined in
13 Superfimd Guidance to yield a quotient representing hazard (U.S. EPA, 200 Ib). The use of the
14 reference concentration in risk assessment is further clarified in RAGs, Part F, Supplemental
15 Guidance for Inhalation risk Assessment (U.S. EPA, 2009a). The guidance provides for
16 addressing hazard for children and adults by estimating time-dependent average daily exposures.
17
This document is a draft for review purposes only and does not constitute Agency policy.
6-31 DRAFT—DO NOT CITE OR QUOTE
-------
7. REFERENCES
AGI. (American Geological Institute). (2005). Glossary of Geology. In (5th ed.). Alexandria, VA.
Amandus. HE: Wheeler. R: Jankovic. J: Tucker. J. (1987a). The morbidity and mortality of vermiculite miners and
millers exposed to tremolite-actinolite: Parti. Exposure estimates. AmJIndMed 11: 1-14.
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: Althouse. R: Morgan. WKC: Sargent EN: Jones. R. (1987b). The morbidity and mortality of
vermiculite miners and millers exposed to tremolite-actinolite: Part III. Radiographic findings. Am J Ind
Med 11: 27-37.
Amandus. HE: Wheeler. R: Armstrong. BG: McDonald. AD: McDonald. JC: Sebastien. P. (1988). Mortality of
vermiculite miners exposed to tremolite. Ann Occup Hyg 32: 459^4-65.
http://dx.doi.org/10.1093/ammvg/32.inhaledjardcles VI.459.
Anderson. HA: Lilis. R: Damn. SM: Fischbein. AS: Selikoff. IJ. (1976). Household-contact asbestos neoplastic risk.
AnnNYAcadSci 271: 311-323. http://dx.doi.0rg/10.llll/i.1749-6632.1976.tb23127.x.
Anderson. HA: Lilis. R: Damn. SM: Selikoff. IJ. (1979). Asbestosis among household contacts of asbestos factory
workers. AnnN Y Acad Sci 330: 387-399. http://dx.doi.0rg/10.llll/i.1749-6632.1979.tbl8741.x.
Andrion. A: Bosia. S: Paoletti. L: Feyles. E: Lanfranco. C: Bellis. D: Mollo. F. (1994). Malignant peritoneal
mesothelioma in a 17-year-old boy with evidence of previous exposure to chrysotile and tremolite asbestos.
Hum Pathol 25: 617-622. http://dx.doi.org/10.1016/0046-8177(94)90230-5.
Armstrong. B. (1998). Effect of measurement error on epidemiological studies of environmental and occupational
exposures. Occup Environ Med 55: 651-656. http://dx.doi.org/10.1136/oem.55.10.651.
Armstrong. BG: Mcdonald. JC: Sebastien. P: Althouse. R: Amandus. HE: Wheeler. R. (1988). Radiological changes
in vermiculite workers exposed to Tremolite. Ann Occup Hyg 32: 469-474.
http://dx.doi.org/10.1093/annhyg/32.inhaledjarticles VI.469.
Ascoli. V: Carnovale-Scalzo. C: Nardi. F: Efrati C: Menegozzo. M. (2003). A one-generation cluster of malignant
mesothelioma within a family reveals exposure to asbestos-contaminated jute bags in Naples, Italy. Eur J
Epidemiol 18: 171-174. http://dx.doi.Org/10.1023/A:1023032214444.
Asgharian. B: Menache. MG: Miller. FJ. (2004). Modeling age-related particle deposition in humans. J Aerosol Med
PulmDrugDeliv 17: 213-224. http://dx.doi.org/10.1089/iam.2004.17.213.
Athanasiou. K: Constantopoulos. SH: Rivedal E: Fitzgerald. DJ: Yamasaki. H. (1992). Metsovo-tremolite asbestos
fibres: In vitro effects on mutation, chromosome aberration, cell transformation and intercellular
communication. Mutagenesis 7: 343-347. http://dx.doi.Org/10.1093/mutage/7.5.343.
ATS. (American Thoracic Society). (2004). Diagnosis and initial management of nonmalignant diseases related to
asbestos. Am J Respir Crit Care Med 170: 691-715. http://dx.doi.org/10.1164/rccm.200310-1436ST.
ATSDR. (Agency for Toxic Substances and Disease Registry). (2000). Health consultation: Mortality from
asbestosis in Libby, Montana. Atlanta, GA. http://www.atsdr.cdc.gov/hac/pha/pha.asp?docid= 1225&pg=0.
This document is a draft for review purposes only and does not constitute Agency policy.
1-1 DRAFT—DO NOT CITE OR QUOTE
-------
ATSDR. (Agency for Toxic Substances and Disease Registry). (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: 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.
ATSDR. (Agency for Toxic Substances and Disease Registry). (2003a). Health consultation: Exposure assessment:
Western Mineral Products Site. Atlanta, GA.
http://www.health.state.mn.us/divs/eh/hazardous/sites/hennepin/western/wstrnhcl003.pdf.
ATSDR. (Agency for Toxic Substances and Disease Registry). (2003b). Report on the expert panel on health effects
of asbestos and synthetic vitreous fibers: The influence of fiber length.
ATSDR. (Agency for Toxic Substances and Disease Registry). (2004). Toxicological profile for synthetic vitreous
fibers [ATSDR Tox Profile]. Atlanta, GA: U.S. Department of Health and Human Services, Public Health
Service. http://www.atsdr.cdc.gov/ToxProfiles/tpl61.pdf.
ATSDR. (Agency for Toxic Substances and Disease Registry). (2005a). Final report of the Northeast Minneapolis
Community Vermiculite Investigation (NMCVI) and worker/household study: Cohort identification and
characterization. Atlanta, GA.
http://www.health.state.mn.us/divs/eh/hazardous/sites/hennepin/western/nmcvicohort.pdf.
ATSDR. (Agency for Toxic Substances and Disease Registry). (2005b). Health Consultation: Former O.C. Scott and
Sons Company, 14111 Scottslawn Road, Marysville, Union County, Ohio. EPA Facility ID:
OHD990834483. (PB2008-104070). Atlanta, GA.
ATSDR. (Agency for Toxic Substances and Disease Registry). (2008a). Fact sheet on the Summary report:
Exposure to asbestos-containing vermiculite from Libby, Montana, at 28 processing sites in the United
States [Fact Sheet]. Atlanta, GA.
http://www.atsdr.cdc.gov/asbestos/sites/national map/Summary Report Fact Sheet.pdf.
ATSDR. (Agency for Toxic Substances and Disease Registry). (2008b). Summary report: Exposure to asbestos-
containing vermiculite from Libby, Montana, at 28 processing sites in the United States. Atlanta, GA:
Agency for Toxic Substances & 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. J Toxicol Environ Health B Crit Rev 14: 40-75.
http://dx.doi.org/10.1080/10937404.2011.556046.
Bailey. KF: Kelse. J: Wylie. AG: Lee. RJ. (2006). The asbestiform and prismatic mineral growth habit and their
relationship to cancer studies: A pictorial presentation. Alexandria, VA: National Stone, Sand & Gravel
Association, http://www.regulations.gov/#!documentDetail:D=EPA-HQ-ORD-2003-0016-0076.
Baker. EL: Dagg. T: Greene. RE. (1985). Respiratory illness in the construction trades. I. The significance of
asbestos-associated pleural disease among sheet metal workers. J Occup Med 27: 483-489.
This document is a draft for review purposes only and does not constitute Agency policy.
7-2 DRAFT—DO NOT CITE OR QUOTE
-------
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/i. 1749-6632.1979.tb 18744.x.
Barone-Adesi F: Ferrante. D: Bertolotti. M: Todesco. A: Mirabelli D: Terracini. B: Magnanj C. (2008). Long-term
mortality from pleura! and peritoneal cancer after exposure to asbestos: Possible role of asbestos clearance.
Int J Cancer 123: 912-916. http://dx.doi.org/10.1002/iic.23609.
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: Schwartz. J. (2008). Children's response to air pollutants. J Toxicol Environ Health A 71: 238-243.
http://dx.doi.org/10.1080/15287390701598234.
Becquemin. MH: Swift PL: Bouchikhi. A: Roy. M: Teillac. A. (1991). Particle deposition and resistance in the
noses of adults and children. Eur Respir J 4: 694-702.
Bennett W: Zeman. K: Jarabek. A. (2003). Nasal contribution to breathing with exercise: effect of race and gender.
J Appl Physiol 95: 497-503. http://dx.doi.org/10.1152/iapplphysiol.00718.2002.
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.
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/iapplphvsiol.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.
Berkson. J. (1950). Are there two regressions? J Am Stat Assoc 45: 164-180.
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.
Herman. DW. (2010). Comparing milled fiber, Quebec ore, and textile factory dust: has another piece of the
asbestos puzzle fallen into place. Crit Rev Toxicol 40: 151-188.
http://dx.doi.org/10.3109/10408440903349137.
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.
This document is a draft for review purposes only and does not constitute Agency policy.
7-3 DRAFT—DO NOT CITE OR QUOTE
-------
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: 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.
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.
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. (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: Fritschj L: Olsen. N: Merler. E: Musk. A. (2004). Malignant pleural
and peritoneal mesotheliomas in former miners and millers of crocidolite at Wittenoom, Western Australia.
Occup Environ Med 61: e!4.
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.
BGS. (British Geological Survey). (2011). World mineral production. Keyworth, Nottingham, UK.
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: Bianchj 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.
Bianchj C: Bianchj T. (2009). Malignant pleural mesothelioma in Italy. Indian! Occup Environ Med 13: 80-83.
http://dx.doi.org/10.4103/0019-5278.55124.
Bignon. J: Monchaux. G: Sebastien. P: Hirsch. A: Lafuma. J. (1979). Human and experimental data on translocation
of asbestos fibers through the respiratory system. AnnN Y Acad Sci 330: 745-750.
Bignon. J: Sebastien. P: Gaudichet A. (1978) Measurement of asbestos retention in the human respiratory system
related to health effects. In CC Gravatt; PD LaFleur; KFJ Heinrich (Eds.), Proceedings of the workshop on
asbestos: definitions and measurement methods (Vol. 506, pp. 95-119). Washington, DC: National Bureau
of Standards.
Bilgrad. R. (1997). National Death Index user's 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.
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.
This document is a draft for review purposes only and does not constitute Agency policy.
7-4 DRAFT—DO NOT CITE OR QUOTE
-------
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 Minerals 6: 283-296.
Boettcher. C: Stark. H: Vanheel M. (1996). Stacked bilayer helices: A new structural organization of amphiphilic
molecules. Ultramicroscopy 62: 133-139. http://dx.doi.org/10.1016/0304-3991(95)00140-9.
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.
Broaddus. VC: Everitt JI: Black. B: Kane. AB. (2011). Non-neoplastic and neoplastic pleural endpoints following
fiber exposure. J Toxicol Environ Health B CritRev 14: 153-178.
http://dx.doi.org/10.1080/10937404.2011.556049.
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.
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.
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: 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.
Brown. JS: Wilson. WE: Grant LD. (2005). Dosimetric comparisons of particle deposition and retention in rats and
humans. Inhal Toxicol 17: 355-385.
Bunderson-Schelvan. M: Pfau. JC: Crouch. R: Holian. A. (2011). Nonpulmonary outcomes of asbestos exposure. J
Toxicol Environ Health B CritRev 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. In. New York: Springer.
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/si.bic.6603293.
Carbone. M: Albelda. SM: Broaddus. VC: Flores. RM: Hillerdal. G: Jaurand. MC: Kiaerheim. K: Pass. HI:
Robinson. B: Tsao. A. (2007). Eighth International Mesothelioma Interest Group [Review]. Oncogene 26:
6959-6967. http://dx.doi.org/10.1038/si.onc.1210515.
This document is a draft for review purposes only and does not constitute Agency policy.
7-5 DRAFT—DO NOT CITE OR QUOTE
-------
Carroll RJ: Ruppert. D: Stefanski. LA. (2006). Measurement error in nonlinear models: A modern perspective. In.
London, England: Chapman & Hall.
Case. BW: McDonald. AD: McDonald. JC. (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.
Case. BW: Abraham. JL: Meeker. G: Poolev. FD: Pinkerton. KB. (2011). Applying definitions of "asbestos" to
environmental and "low-dose" exposure levels and health effects, particularly malignant mesothelioma. J
Toxicol Environ Health B CritRev 14: 3-39. http://dx.doi.org/10.1080/10937404.2011.556045.
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.
Cazzadori. A: Malesani. F: Romeo. L. (1992). Malignant pleura! mesothelioma caused by non-occupational
childhood exposure to asbestos. Br J Ind Med 49: 599. http://dx.doi.0rg/10.1136/oem.49.8.599.
CDC. (Centers for Disease Control and Prevention). (2009). Malignant mesothelioma mortality: United States,
1999—2005. MMWR Morb Mortal Wkly Rep 58: 393-396.
CEC. (Commission of the European Communities). (1996). Guidance on risk assessment at work. Luxembourg:
Office for Official Publications of the European Communities.
http://osha.europa.eu/en/topics/riskassessment/guidance.pdf.
Cerny. FJ. (1987). Breathing pattern during exercise in young Black and Caucasian subjects. J Appl Physiol 62:
2220-2223.
Cheng. JQ: Lee. WC: Klein. MA: Cheng. GZ: Jhanwar. SC: Testa. JR. (1999). Frequent mutations of NF2 and
allelic loss from chromosome band 22ql2 in malignant mesothelioma: Evidence for a two-hit mechanism
of NF2 inactivation. Genes Chromosomes Cancer 24: 238-242. http://dx.doi.org/10.1002/(SICI)1098-
2264(199903)24:3<238::AID-GCC9>3.0.CO:2-M.
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. Chest 93: 621-628.
http://dx.doi.0rg/10.1378/chest.93.3.621.
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. (1994). Deposition and clearance of chrysotile asbestos. Ann Occup Hyg 38: 625-633.
http://dx.doi.Org/10.1093/annhyg/38.4.625.
Churg. A: Vedal S. (1994). Fiber burden and patterns of asbestos-related disease in workers with heavy mixed
amosite and chrysotile exposure. Am J Respir Crit Care Med 150: 663-669.
Claeskens. G: Hjort. NL. (2008). Model selection and model averaging. In. 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.
This document is a draft for review purposes only and does not constitute Agency policy.
7-6 DRAFT—DO NOT CITE OR QUOTE
-------
Coffin. PL: Palekar. LD: Cook. PM. (1983). Correlation of in vitro and in vivo methods by means of mass dose and
fiber distribution for amosite and fibrous ferroactinolite. Environ Health Perspect 51: 49-53.
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.
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.
Cook. PJ: Doll. R: Fellingham. SA. (1969). A mathematical model for the age distribution of cancer in man. Int J
Cancer4: 93-112. http://dx.doi.org/10.1002/iic.2910040113.
Cook. PM: Olson. GF. (1979). Ingested mineral fibers: Elimination in human urine. Science 204: 195-198.
http://dx.doi.org/10.1126/science.219478.
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.
Copes. R: Thomas. D: Becklake. MR. (1985). Temporal patterns of exposure and nonmalignant pulmonary
abnormality in Quebec chrysotile workers. Arch Environ Health 40: 80-87.
Copley. SJ: Wells. AU: Rubens. MB: Chabat F: Sheehan. RE: Musk. AW: Hansell. DM. (2001). Functional
consequences of pleura! disease evaluated with chest radiography and CT. Radiology 220: 237-243.
Cox. DR. (1972). Regression models and life-tables. J Roy Stat Soc B 34: 187-220.
Cristaudo. A: Foddis. R: Vivaldi. A: Buselli. R: Gattini. V: Guglielmi. G: Cosentino. F: Ottenga. F: Ciancia. E:
Libener. R: Filiberti. R: Neri. M: Betta. P: Tognon. M: Mutti. L: Puntoni. R. (2005). SV40 enhances the
risk of malignant mesothelioma among people exposed to asbestos: A molecular epidemiologic case-
control study. Cancer Res 65: 3049-3052. http://dx.doi.org/10.1158/0008-5472.CAN-04-2219.
Crump. KS: Howe. RB. (1985) A review of methods for calculating statistical confidence limits in low dose
extrapolation. In DB Clayson; D Krewski; I Munro (Eds.), lexicological risk assessment: Vol 1 Biological
and statistical criteria (pp. 187-203). Boca Raton, FL: CRC Press, Inc.
Cullen. MR: Barnett MJ: Balmes. JR: Cartmel B: Redlich. CA: Brodkin. CA: Barnhart. S: Rosenstock. L:
Goodman. GE: Hammar. SP: Thornquist MD: Omenn. GS. (2005). Predictors of lung cancer among
asbestos-exposed men in the beta-carotene and retinol efficacy trial. Am J Epidemiol 161: 260-270.
http://dx.doi.org/10.1093/aie/kwi034.
Cunningham HM: Pontefract RD. (1974). Placental transfer of asbestos [Letter]. Nature 249: 177-178.
http://dx.doi.org/10.1038/249177aO.
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.ll.247.
Dana. DW: Hurlbut CS: Klein. C. (1977). Manual of mineralogy. In (19th ed.). New York, NY: John Wiley and
Sons.
This document is a draft for review purposes only and does not constitute Agency policy.
7-7 DRAFT—DO NOT CITE OR QUOTE
-------
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 (Vol. 90, pp. 33-45).
Lyon, France: International Agency for Research on Cancer.
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. Ill 1/i. 1749-
6632.1991.tb24497.x.
Davis. L: Martin. T: Kligler. B. (1992). Use of death certificates for mesothelioma surveillance. Public Health Rep
107: 481-483.
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.
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: Harris. RL. (1979). Estimates of pulmonary and gastrointestinal deposition for occupational fiber
exposure. Washington, DC: National Institute of Occupational Safety and Health.
http://www.ntis.gov/search/product.aspx? ABBR=PB80149644.
Dement. JM: Welch. L: Bingham E: Cameron. B: Rice. C: Quinn. P: Ringen. K. (2003). Surveillance of respiratory
diseases among construction and trade workers at Department of Energy nuclear sites. Am J Ind Med 43:
559-573. http://dx.doi.org/10.1002/aiim. 10226.
Dianzani. I: Gibello. L: Biava. A: Giordano. M: Bertolotti. M: Betti. M: Ferrante. D: Guarrera. S: Betta. GP:
Mirabelli. D: Matullo. G: Magnani. C. (2006). Polymorphisms in DNA repair genes as risk factors for
asbestos-related malignant mesothelioma in a general population study. Mutat Res-Fundam Mol Mech
Mutagen 599: 124-134. http://dx.doi.0rg/10.1016/i.mrfmmm.2006.02.005.
Ding. M: Dong. Z: Chen. F: Pack. D: Ma. W: Ye. J: Shi. X: Castranova. V: 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: Williams. M: Corn. C: Brollo. A: Bianchj C. (1990). Asbestos content of lung tissue, lymph nodes, and
pleural plaques from former shipyard workers. Am Rev Respir Dis 142: 843-847.
Dodson. R: O'Sullivan. M: Huang. J: Holiday. D: Hammar. S. (2000a). Asbestos in extrapulmonary sites: Omentum
and mesentery. Chest 117: 486-493. http://dx.doi.org/10.1002/aiim. 10164.
Dodson. R: 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.
This document is a draft for review purposes only and does not constitute Agency policy.
7-8 DRAFT—DO NOT CITE OR QUOTE
-------
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: 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::ArD-AJiM2>3.0.CO:2-V.
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.
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.
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.
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.
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.
Duncan. K: Ohio. A: Dailev. L: Bern. A: Gibbs-Flournoy. E: Padilla-Carlin. D: Roggli. V: Devlin. R. (2010). Effect
of size fractionation on the toxicity of amosite and Libby amphibole asbestos. Toxicol Sci 118: 420-434.
http://dx.doi.org/10.1093/toxsci/kfq281.
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/gep014.
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.org/10.1093/annhyg/l 1.1.1.
Ehrlich. R: Lilis. R: Chan. E: Nicholson. WJ: Selikoff. U. (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.
Erdinc. M: Erdinc. E: Cok. G: Polatli M. (2003). Respiratory impairment due to asbestos exposure in brake-lining
workers. EnvironRes 91: 151-156. http://dx.doi.org/10.1016/S0013-9351(02)00063-4.
Ewing. WM: Hays. SM: Hatfield. R: Longo. WE: Millette. JR. (2010). Zonolite attic insulation exposure studies. Int
J Occup Environ Health 16: 279-290.
This document is a draft for review purposes only and does not constitute Agency policy.
7-9 DRAFT—DO NOT CITE OR QUOTE
-------
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/(SICI)1097-
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: Tosolinj 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.
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.470027Q111.
Gasparrini. A: Pizzo. AM: Gorini. G: Costantini. AS: Silvestri. S: Ciapini. C: Innocenti A: Berry. G. (2008).
Prediction of mesothelioma and lung cancer in a cohort of asbestos exposed workers. Eur J Epidemiol 23:
541-546. http://dx.doi.org/10.1007/sl0654-008-9257-z.
Ohio. AJ: Kennedy. TP: Whorton. AR: Crumbliss. AL: Hatch. GE: Hoidal JR. (1992). Role of surface complexed
iron in oxidant generation and lung inflammation induced by silicates. Am J Physiol 263: L511-L518.
Gibbs. AR: Stephens. M: Griffiths. DM: Blight BJ: Poolev. FD. (1991). Fibre distribution in the lungs and pleura of
subjects with asbestos related diffuse pleura! fibrosis. Br J Ind Med 48: 762-770.
http://dx.doi.org/10.1136/oem.48.ll.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: Sanchez. MS. (2009). Amphibole forensics: Using the composition of amphiboles to determine their
source, the Libby, Montana, example. American Mineralogist 94: 837-840.
http://dx.doi.org/10.2138/am.2009.3224.
Guzeliaa PS: Henry. CJ: Olin. SS. (Eds.). (1992). Similarities and differences between children and adults:
Implications for risk assessment. In PS Guzelian; CJ Henry; SS Olin (Eds.). New York: 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.
This document is a draft for review purposes only and does not constitute Agency policy.
7-10 DRAFT—DO NOT CITE OR QUOTE
-------
Hammond. EC: Selikoff IJ: Seidman. H. (1979). Asbestos exposure, cigarette smoking and death rates. Ann N Y
AcadSci 330: 473-490.
Hanahan. D: Weinberg. RA. (2011). Hallmarks of cancer: The next generation. Cell 144: 646-674.
http://dx.doi.0rg/10.1016/i.cell.2011.02.013.
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.
Hague. A: Kanz. M: Mancuso. M: Williams. G: Dodson. R. (1991). Asbestos in the lungs of children. Ann NY
AcadSci643: 419-429.
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: 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.
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: Ali. I: Vrazel DM: Uchida. T. (2001). Chrysotile asbestos fibers detected in the newborn pups
following gavage feeding of pregnant mice. J Toxicol Environ Health A 62: 23-31.
Harris. RL. Jr: Timbrell. V. (1975). The influence of fibre shape in lung deposition-mathematical estimates. Inhaled
Particles 4: 75-89.
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.
Hasanoglu. HC: Yildirim. Z: Ermis. H: Kilic. T: Koksal. N. (2006). Lung cancer and mesothelioma in towns with
environmental exposure to asbestos in Eastern Anatolia. Int Arch Occup Environ Health 79: 89-91.
http://dx.doi.org/10.1007/s00420-005-0010-6.
Hauptmann. M: Pohlabeln. H: Lubin. JH: Jockel. KH: Ahrens. W: Briiske-Hohlfeld. I: Wichmann. HE. (2002). The
exposure-time-response relationship between occupational asbestos exposure and lung cancer in two
German case-control studies. Am J Ind Med 41: 89-97.
Hawthorne. FC. (1981). Amphibole spectoscopy. Reviews in Mineralogy and Geochemistry 9: 103-139.
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.
This document is a draft for review purposes only and does not constitute Agency policy.
7-11 DRAFT—DO NOT CITE OR QUOTE
-------
Held. 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/sj.jea.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.
HEW. (U.S. Department of Health, Education and Welfare). (1979). Smoking and health: A report of the Surgeon
General. (PHS 79-50066). Washington, DC: Government Printing Office.
HHS. (U.S. Department of Health and Human Services). (1990). The health benefits of smoking cessation: Areport
of the surgeon general. Rockville, MD. http://profiles.nlm.nih. gov/ps/access/NNBBCV.pdf.
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). The human evidence: Parenchymal and pleura! changes. Ann Occup Hyg 38: 561-567.
http://dx.doi.0rg/10.1093/annhyg/38.4.561.
Hiraoka. T: Ohkura. M: Morinaga. K: Kohyama. N: Shimazu. K: Ando. M. (1998). Anthophyllite exposure and
endemic pleura! plaques in Kumamoto, Japan. Scand J Work Environ Health 24: 392-397.
Hirvonen. A: Tuimala. J: Ollikainen. T: Linnainmaa. K: Kinnula. V. (2002). Manganese superoxide dismutase
genotypes and asbestos-associated pulmonary disorders. Cancer Lett 178: 71-74.
http://dx.doi.org/10.1016/S0304-3835(01)00819-9.
Hodgson. J: McElvenny. D: Darnton. A: Price. M: Peto. J. (2005). The expected burden of mesothelioma mortality
in Great Britain from 2002 to 2050. Br J Cancer 92: 587-593. http://dx.doi.org/10.1038/sj.bjc.6602307.
Holt P. (1982). Translocation of asbestos dust through the bronchiolar wall. Environ Res 27: 255-260.
http://dx.doi.org/10.1016/0013-9351(82)90080-9.
Holt PF. (1983). Translocation of inhaled dust to the pleura. EnvironRes 31: 212-220.
http://dx.doi.org/10.1016/0013-9351(83)90077-4.
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: Lameshow. S. (2000). Applied logistic regression. In. New York, NY: John Wiley & Sons.
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.
This document is a draft for review purposes only and does not constitute Agency policy.
7-12 DRAFT—DO NOT CITE OR QUOTE
-------
ICRP. (International Commission on Radiological Protection). (1994). Human respiratory tract model for
radiological protection: A report of a task group of the International Commission on Radiological
Protection. ICRP Publication 66. Ann ICRP 24: 1-482.
ILO. (International Labour Organization). (1971). International classification of radiographs of pneumoconioses
(revised, 1968). In (Red. Ed. 1968 ed., Vol. 22). Geneva, Switzerland: International Labour Office.
ILO. (International Labour Organization). (1980). Guidelines for the use of the ILO international classification of
radiographs of pneumoconioses. In (Vol. 22). Geneva, Switzerland: International Labour Office.
ILO. (International Labour Organization). (2002). International classification of radiographs of pneumoconioses. In
(Rev. Ed. 2000 ed., Vol. 22). Geneva, Switzerland: International Labour Office.
Inase. N: Takavama. S: Nakavama. M: Miura. H: Kimula. Y. (1991). Pleura! mesothelioma after neighborhood
exposure to asbestos during childhood. Jpn J Med 30: 343-345.
http://dx.doi.org/10.2169/internalmedicinel962.30.343.
IPCS. (International Programme on Chemical Safety). (1986). Asbestos and other natural mineral fibres (Vol. 53).
Geneva, Switzerland: World Health Organization, http://www.inchem.org/documents/ehc/ehc/ehc53.htm.
Isaacs. KK: Martonen. TB. (2005). Particle deposition in children's lungs: Theory and experiment. J Aerosol Med
18: 337-353. http://dx.doi.org/10.1089/iam.2005.18.337.
Jahr. J.. (1974). Dose-response basis for setting a quartz threshold limit value. Arch Environ Health 29: 338-340.
Jakobsson. K: Stromberg. U: Albin. M: Welinder. H: Hagmar. L. (1995). Radiological changes in asbestos cement
workers. Occup Environ Med 52: 20-27.
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: Borzucki. G: Vincent JH. (1988).
Pulmonary clearance of UICC amosite fibres inhaled by rats during chronic exposure at low concentration.
BrJInd Med 45: 300-304.
Jones. JS. (2002). The pleura in health and disease. Lung 179: 397-413. http://dx.doi.org/10.1007/s004080000075.
Kamp. DW: Weitzman. SA. (1999). The molecular basis of asbestos induced lung injury [Review]. Thorax 54: 638-
652. http://dx.doi.0rg/10.1136/thx.54.7.638.
Kane. MJ: Chahinian. AP: Holland. JF. (1990). Malignant mesothelioma in young adults. Cancer 65: 1449-1455.
http://dx.doi.org/10.1002/1097-0142(19900315)65:6<1449::AID-CNCR2820650633>3.0.CO:2-0.
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.
This document is a draft for review purposes only and does not constitute Agency policy.
7-13 DRAFT—DO NOT CITE OR QUOTE
-------
Kilburn. KH: Lilis. R: Anderson. HA: Boylen. CT: Einstein. HE: Johnson. SJS: Warshaw. R. (1985). Asbestos
disease in family contacts of shipyard workers. Am J Public Health 75: 615-617.
Kilburn. KH: Warshaw. RH. (1991). Difficulties of attribution of effect in workers exposed to fiberglass and
asbestos. Am J Ind Med 20: 745-751. http://dx.doi.org/10.1002/aiim.4700200605.
Kim. CS: Kang. TC. (1997). Comparative measurement of lung deposition of inhaled fine particles in normal
subjects and patients with obstructive airway disease. Am J Respir Crit Care Med 155: 899-905.
Kim. H: Yasui Y: Burstyn. I. (2006). Attenuation in risk estimates in logistic and Cox proportional-hazards models
due to group-based exposure assessment strategy. Ann Occup Hyg 50: 623-635.
http://dx.doi.org/10.1093/annhyg/mel021.
Kimizuka. G: Ohwada. H: Havashi. Y. (1987). Co-carcinogenic effect of asbestos and benzo(a)pyrene in the lung of
hamster. ActaPathol Jpn37: 465-474. http://dx.doi.0rg/10.llll/i.1440-1827.1987.tb00380.x.
Kleinbaum. DG: Klein. M. (1996). Survival analysis: A self-learning text. In. 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: Guoying. Y: Kaminski. N: Rosas. I: Oury. TD. (2009). Oxidative stress
alters syndecan-1 distribution in lungs with pulmonary fibrosis. J Biol Chem 284: 3537-3545.
http://dx.doi.org/10.1074/ibc.M807001200.
Kohyama. N: Suzuki. Y. (1991). Analysis of asbestos fibers in lung parenchyma, pleural plaques, and mesothelioma
tissues of North American insulation workers. Ann N Y Acad Sci 643: 27-52.
http://dx.doi.0rg/10.llll/i.1749-6632.1991.tb24442.x.
Kopylev. L: Sullivan. PA: Vinikoor. LC: Bateson. TF. (2011). Monte Carlo analysis of impact of
underascertainment of Mesothelioma cases on underestimation of risk. 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.
Kouris. SP: Parker. PL: Bender. AP: Williams. AN. (1991). Effects of asbestos-related pleural disease on
pulmonary function. Scand J Work Environ Health 17: 179-183.
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. (2010). Kirk-Othmer concise encyclopedia of chemical technology. In (5th ed.). New
York: John Wiley and Sons.
Krowke. R: Bluth. V: Merker. JH: Neubert. D. (1983). Placental transfer and possible teratogenic potential of
asbestos in mice [Abstract]. Teratology 32: 26A-27A.
Landi M: Dracheva. T: Rotunno. M: Figueroa. J: Liu. H: Dasgupta. A: Mann. F: Fukuoka. J: Hames. M: Bergen. A:
Murphy. S: Yang. P: Pesatori. A: Consonni. D: Bertazzi. P: Wacholder. S: Shih. J: Caporaso. N: Jen. J.
(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.
This document is a draft for review purposes only and does not constitute Agency policy.
7-14 DRAFT—DO NOT CITE OR QUOTE
-------
Langer. AM: Mackler. AD: Poolev. FD. (1974). Electron microscopical investigation of asbestos fibers. Environ
Health Perspect 9: 63-80.
Langer. AM: Nolan. RP: Constantopoulos. SH: Moutsopoulos. HM. (1987). Association of Metsovo lung and
pleural mesothelioma with exposure to tremolite-containing whitewash. Lancet 329: 965-967.
http://dx.doi.org/10.1016/S0140-6736(87)90305-9.
Langseth. H: Johansen. BV: Nesland. JM: Kjaerheim. K. (2007). Asbestos fibers in ovarian tissue from Norwegian
pulp and paper workers. Int J Gynecol Cancer 17: 44-49. http://dx.doi.org/10. Ill 1/i. 1525-
1438.2006.00768.x.
Larsen. ES. (1942). Alkalic rocks of Iron Hill, Gunnison County, Colorado. Washington, DC: U.S. Geological
Survey.
Larson. T: Meyer. C: Kapil V: Gurnev. J: Tarver. R: Black. C: Lockev. J. (2010a). Workers with Libby amphibole
exposure: retrospective identification and progression of radiographic changes. Radiology 255: 924-933.
http://dx.doi.org/10.1148/radiol.10091447.
Larson. TC: Antao. VC: Bove. FJ. (201 Ob). Vermiculite worker mortality: Estimated effects of occupational
exposure to Libby amphibole. J Occup EnvironMed 52: 555-560.
http://dx.doi.org/10.1097/JOM.Ob013e3181dc6d45.
Leake. B.. (1978). Nomenclature of amphiboles. Mineral Mag 42: 533-563.
Leake. B.. E.: Woollev. A.. R.: Arps. C.. E. S.: Birch. W.. P.: Gilbert. M.. C.: Grice. J.. P.: Hawthorne. F.. C.:
Kato. A..: Kisch. H.. J.: Krivovichev. V.. G.: Linthout. K..: Laird. J..: Mandarino. J..: Maresch. W.. V.:
Nickel E.. H.: Rock. N.. M. S.: Schumacher. J.. C.: Smith. D.. C.: Shephenson. N.. C. N.: Ungaretti L.
.: Whittake. E.. J. W.: Youzhi. G.. (1997). Nomenclature of amphiboles: Report of the Subcommittee on
Amphiboles of the International Mineralogical Association Commission on New Minerals and Mineral
Names. Mineral Mag 61: 295-321.
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.
Li. FP: Lokich. J: Lapev. J: Neptune. WB: Wilkins. EW. (1978). Familial mesothelioma after intense asbestos
exposure at home. JAMA 240: 467. http://dx.doi.org/10.1001/jama.1978.03290050057022.
Li. FP: Dreyfus. MG: Antman. KH. (1989). Asbestos-contaminated nappies and familial mesothelioma. Lancet 333:
909-910. http://dx.doi.org/10.1016/S0140-6736(89)92916-4.
Lieben. J: Pistawka. H. (1967). Mesothelioma and asbestos exposure. Arch Environ Health 14: 559-563.
LUis. R: Miller. A: Godbold. J: Chan. E: Benkert. S: Selikoff. IJ. (1991). The effect of asbestos-induced pleural
fibrosis on pulmonary function: quantitative evaluation. In The third wave of asbestos disease: Exposure to
asbestos in place (Vol. 643, pp. 162-168). New York, NY: New York Academy of Sciences.
Lin. F: Liu. Y: Keshava. N: Li. S. (2000). Crocidolite induces cell transformation and p53 gene mutation in
BALB/C-3T3 cells. Teratog CarcinogMutagen20: 273-281. http://dx.doi.org/10.1002/1520-
6866(2000)20:5<273::AID-TCM3>3.0.CO:2-1.
Ling. S: Fried. L: Garrett E: Hirsch. R: Guralnik. J: Hochberg. M. (2000). The accuracy of self-report of physician
diagnosed rheumatoid arthritis in moderately to severely disabled older women. Women's Health and
Aging Collaborative Research Group. J Rheumatol 27: 1390-1394.
This document is a draft for review purposes only and does not constitute Agency policy.
7-15 DRAFT—DO NOT CITE OR QUOTE
-------
Lippmann. M. (1990). Effects of fiber characteristics on lung deposition, retention, and disease. Environ Health
Perspect 88: 311-317. http://dx.doi.org/10.1289/ehp.9Q88311.
Lockev. JE: Brooks. SM: Jarabek. AM: Khoury. PR: McKay. 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.
Lockev. JE. (1985) Pulmonary hazards associated with vermiculite exposure. Master's Thesis, University of
Cincinnati, Cincinnati, OH.
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.
Lund. LG: Williams. MG: Dodson. RF: Aust. AE. (1994). Iron associated with asbestos bodies is responsible for the
formation of single strand breaks in phi XI74 RFIDNA. Occup Environ Med 51: 200-204.
Luo. S: Liu. X: Mu. S: Tsai. SP: Wen. CP. (2003). Asbestos related diseases from environmental exposure to
crocidolite inDa-yao, China. I. Review of exposure and epidemiological data. Occup Environ Med 60: 35-
41; discussion41-32. http://dx.doi.0rg/10.1136/oem.60.l.35.
Magee. F: Wright. JL: Chan. N: Lawson. L: Churg. A. (1986). Malignant mesothelioma caused by childhood
exposure to long-fiber low aspect ratio tremolite. Am J Ind Med 9: 529-533.
http://dx.doi.org/10.1002/aiim.4700090604.
Magnani. C: Dalmasso. P: Biggeri. A: Ivaldi. C: Mirabelli. D: Terracini. B. (2001). Increased risk of malignant
mesothelioma of the pleura after residential or domestic exposure to asbestos: A case-control study in
Casale Monferrato, Italy. Environ Health Perspect 109: 915-919. http://dx.doi.org/10.1289/ehp.01109915.
Magnani. C: Ferrante. D: Barone-Adesi. F: Bertolotti. M: Todesco. A: Mirabelli. D: Terracini. B. (2008). Cancer
risk after cessation of asbestos exposure: A cohort study of Italian asbestos cement workers. Occup Environ
Med 65: 164-170. http://dx.doi.org/10.1136/oem.2007.032847.
Malorni. W: losi. F: Falchi. M: Donelli. G. (1990). On the mechanism of cell internalization of chrysotile fibers: An
immunocytochemical and ultrastructural study. Environ Res 52: 164-177. http://dx.doi.org/10.1016/S0013-
9351(05)80251-8.
Marsella. J: Liu. B: Vaslet C: Kane. A. (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.
Martensson. G: Larsson. S: Zettergren. L. (1984). Malignant mesothelioma in two pairs of siblings: Is there a
hereditary predisposing factor. Eur J Respir Dis 65: 179-184.
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.
Mayall. FG: Jacobsoa G: Wilkins. R. (1999). Mutations of p53 gene and SV40 sequences in asbestos associated and
non-asbestos-associated mesotheliomas. J ClinPathol 52: 291-293.
This document is a draft for review purposes only and does not constitute Agency policy.
7-16 DRAFT—DO NOT CITE OR QUOTE
-------
McConnell EE: Shefner. AM: Rust JH: Moore. JA. (1983a). Chronic effects of dietary exposure to amosite and
chrysotile asbestos in Syrian golden hamsters. Environ Health Perspect 53: 11 -25.
McConnell EE: Rutter. HA: Ulland. BM: Moore. JA. (1983b). Chronic effects of dietary exposure to amosite
asbestos andtremolite inF344 rats. Environ Health Perspect 53: 27-44.
McDonald. J: Edwards. C: Gibbs. A: Lloyd. H: Poolev. F: Ross. D: Rudd. R. (2001). Case-referent survey of young
adults with mesothelioma: II. Occupational analyses. Ann Occup Hyg 45: 519-523.
http://dx.doi.0rg/10.1093/annhyg/45.7.519.
McDonald. JC: 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.
McDonald. JC: McDonald. AD: Sebastien. P: Moy. K. (1988). Health of vermiculite miners exposed to trace
amounts of fibrous tremolite. Occup Environ Med 45: 630-634. http://dx.doi.0rg/10.1136/oem.45.9.630.
McDonald. JC: Harris. J: Armstrong. B. (2002). Cohort mortality study of vermiculite miners exposed to fibrous
tremolite: An update. Ann Occup Hyg 46: 93-93. 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.
McMillan. GH: Pethybridge. RJ: Sheers. G. (1980). Effect of smoking on attack rates of pulmonary and pleural
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. GP: Bern. AM: Brownfield. IK: Lowers. HA: Sutlev. SJ: Hoefea TM: Vance. JS. (2003). The composition
and morphology of amphiboles from the Rainy Creek Complex, near Libby, Montana. American
Mineralogist 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 pleural
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.
Miller. A: Lilis. R: Godbold. J: Chan. E: Wu. X: Selikoff. IJ. (1994). Spirometric impairments in long-term
insulators. Relationships to duration of exposure, smoking, and radiographic abnormalities. Chest 105: 175-
182.
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.
Miserocchi. G: Sancini. G: Mantegazza. F: Chiappino. G. (2008). Translocation pathways for inhaled asbestos
fibers. Environ Health Global Access Sci Source 7: 4. http://dx.doi.org/10.1186/1476-069X-7-4.
This document is a draft for review purposes only and does not constitute Agency policy.
7-17 DRAFT—DO NOT CITE OR QUOTE
-------
Moatamed. F: Lockev. JE: Parry. WT. (1986). Fiber contamination of vermiculites: A potential occupational and
environmental health hazard. Environ Res 41: 207-218. http://dx.doi.org/10.1016/80013-9351(86)80183-9.
Moolgavkar. SH: Turim. J: Alexander. DP: Lau. EC: Gushing. CA. (2010). Potency factors for risk assessment at
Libby, Montana. Risk Anal 30: 1240-1248. http://dx.doi.0rg/10.llll/i.1539-6924.2010.01411.x.
Morgan. A: Talbot RJ: Holmes. A. (1978). Significance of fibre length in the clearance of asbestos fibres from the
lung. Br J Ind Med 35: 146-153. http://dx.doi.0rg/10.1136/oem.35.2.146.
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 trachea! epithelial cells. Environ Health
Perspect 51: 331-335.
Mossman. BT: Eastman. A: Bresnick. E. (1984). Asbestos andbenzo[a]pyrene act synergistically to induce
squamous metaplasia and incorporation of [3H]thymidine in hamster tracheal epithelium. Carcinogenesis 5:
1401-1404. http://dx.doi.org/10.1093/carcin/5.11.1401.
Mossman. BT: Cameron. GS: Yotti LP. (1985). Cocarcinogenic and tumor promoting properties of asbestos and
other minerals in tracheobronchial epithelium. Carcinogenesis 8: 217-238.
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. Part Fibre Toxicol 4: 4.
http://dx.doi.org/10.1186/1743-8977-4-4.
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.
MSHA. Asbestos Exposure Limit. 30. § 56. 57. and 71. (2008).
Mukherjee. S: de Klerk. N: Palmer. LJ: Olsen. NJ: Pang. SC: William Musk. A. (2000). Chest pain in asbestos-
exposed individuals with benign pleura! and parenchyma! disease. Am J Respir Crit Care Med 162: 1807-
1811.
Muller. JG: Bohnker. BK: Philippi. AF: Litow. FK: Rudolph. G: Hernandez. JE. (2005). Trends in pleura!
radiographic findings in the Navy Asbestos Medical Surveillance Program (1990-1999). Mil Med 170:
375-380.
Muravov. O: Kave. W: Lewin. M: Berkowitz. Z: Lybarger. J: Campolucci S: Parker. J. (2005). The usefulness of
computed tomography in detecting asbestos-related pleura! abnormalities in people who had indeterminate
chest radiographs: The Libby, MT, experience. Int J Hyg Environ Health 208: 87-99.
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
Retrieved August 3, 2011, from http://seer.cancer.gov/faststats
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/si.onc.1205804.
Neri. M: Ugolini. D: Dianzani. I: Gemignani. F: Landi S: Cesario. A: Magnani. C: Mutti. L: Puntoni. R: Bonassi. S.
(2008). Genetic susceptibility to malignant pleura! mesothelioma and other asbestos-associated diseases.
MutatRes659: 126-136. http://dx.doi.0rg/10.1016/i.mrrev.2008.02.002.
This document is a draft for review purposes only and does not constitute Agency policy.
7-18 DRAFT—DO NOT CITE OR QUOTE
-------
Nerj 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 pleura! fibrosis in shipyard
workers exposed to asbestos. Am J Ind Med 30: 588-595. http://dx.doi.org/10.1002/(SICI)1097-
0274(199611')30:5<588::AID-AJIM6>3.0.CO:2-P.
Newhouse. ML: Berry. G: Wagner. JC: Turok. ME. (1972). A study of the mortality of female asbestos workers.
Occup Environ Med 29: 134-141.
Newhouse. ML: Berry. G. (1976). Predictions of mortality from mesothelial tumours in asbestos factory workers. Br
J Ind Med 33: 147-151.
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). (1994a). Method 7400: asbestos and other fibers by
PCM. In NIOSH manual of analytical methods (4 ed., pp. N/A). Cincinnati, OH: U.S.: Department of
Health and Human Services, Centers for Disease Control and Prevention, National Institute for
Occupational Safety and Health (NIOSH).
NIOSH. (National Institute for Occupational Safety and Health). (1994b). Method 7402: Asbestos by TEM. In
NIOSH manual of analytical methods (4 ed., pp. N/A). Cincinnati, OH: U.S.: Department of Health and
Human Services, Centers for Disease Control and Prevention, National Institute for Occupational Safety
and Health (NIOSH).
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). http://www.cdc.gov/niosh/docs/2011-
1597.
Noonan. C. (2006). Exposure matrix development for the Libby cohort. Inhal Toxicol 18: 963-967.
http://dx.doi.org/10.1080/08958370600835021.
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.
NRC. (National Research Council). (1983). Risk assessment in the federal government: Managing the process. In.
Washington, DC: National Academies Press.
NRC. (National Research Council). (1993). Pesticides in the diets of infants and children. In. 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) (Vol. 249). Research Triangle Park, NC.
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/nrats (feed studies) (Vol. 295). Research Triangle Park, NC.
http://www.ntis.gov/search/product.aspx? ABBR=PB86167103.
NTP. (National Toxicology Program). (1988). NTP technical report on the toxicology and carcinogenesis studies of
Crocidolite Asbestos (CAS No. 12001-28-4) InF344/NRats (Feed Studies) (Vol. 280). Research Triangle
Park, NC. http://ntp.niehs.nih.gov/ntp/htdocs/LT rpts7tr280.pdf.
This document is a draft for review purposes only and does not constitute Agency policy.
7-19 DRAFT—DO NOT CITE OR QUOTE
-------
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) (Vol. 279). Research Triangle Park,
NC. 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) (Vol. 277). Research Triangle Park, NC.
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.
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.
Oldham. P.. D. (1965). On estimating the arithmetic means of lognormally-distributed populations. Biometrics 21:
235-239.
Oliver. LC: Risen. EA: Greene. R: Sprince. NL. (1988). Asbestos-related pleura! plaques and lung function. Am J
IndMed 14: 649-656. http://dx.doi.org/10.1002/aiim.4700140604.
OSHA. (Occupational Safety & Health Administration). (1994). Occupational Safety and Health Standards: Toxic
and Hazardous Substances - Asbestos. Standard No. 1910.1001 [Standard]. Washington, DC: United States
Department of Labor.
OSHA. Asbestos. 29. § 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.
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 Global Access Sci Source 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).
Pleura! 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. In. Neenah, WI:
Big Earth Publishing.
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.
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.
This document is a draft for review purposes only and does not constitute Agency policy.
7-20 DRAFT—DO NOT CITE OR QUOTE
-------
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.
Percy. C: Stanek. E. Ill: Gloeckler. L. (1981). Accuracy of cancer death certificates and its effect on cancer
mortality statistics. Am J Public Health71: 242-250. http://dx.doi.Org/10.2105/AJPH.71.3.242.
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.
Petrovic. P: Ostojic. L: Peric. I: Mise. K: Ostojic. Z: Bradaric. A: Bota. B: Jankovic. S: Tocilj. J. (2004). Lung
function changes in pleura! asbestosis. Coll Antropol 28: 711-715.
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.
Pfau. JC: Pershouse. M: Putnam. EA. (2006). Conference summary. Directions and needs in asbestos research: new
insights. Inhal Toxicol 18: 919-923. http://dx.doi.org/10.1080/08958370600834735.
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.
Phalen. RF: Oldham. MJ: Beaucage. CB: Crocker. TT: Mortensen. JD. (1985). Postnatal enlargement of human
tracheobronchial airways and implications for particle deposition. Anat Rec 212: 368-380.
http://dx.doi.org/10.1002/ar.1092120408.
Phalen. RF: Oldham. MJ. (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: 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 Genet Toxicol Environ Mutagen 559: 19-27.
http://dx.doi.0rg/10.1016/i.mrgentox.2003.12.006.
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.
This document is a draft for review purposes only and does not constitute Agency policy.
7-21 DRAFT—DO NOT CITE OR QUOTE
-------
Putnam. EA: Smartt A: Groves. A: Schwanke. C: Brezinskj 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. 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.
Reeves. GK: Cox. PR: Darby. SC: Whitlev. E. (1998). Some aspects of measurement error in explanatory variables
for continuous and binary regression models. Stat Med 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: Fritschj L: Olsen. N: Merler. E: Musk. A.
(2007). Age and sex differences in malignant mesothelioma after residential exposure to blue asbestos
(crocidolite). Chest 131: 376-382. http://dx.doi.org/10.1378/chest.06-1690.
Reid. A: Berry. G: Hevworth. J: de Klerk. NH: Musk. AW. (2009). Predicted mortality from malignant
mesothelioma among women exposed to blue asbestos at Wittenoom, Western Australia. Occup Environ
Med 66: 169-174. http://dx.doi.org/10.1136/oem.2007.038315.
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.
Rendall R: Du Toit R. (1994). The retention and clearance of glass fibre and different varieties of asbestos by the
lung. Ann Occup Hyg 38: 757-761. http://dx.doi.org/10.1093/annhvg/38.inhaled_particles VII.757.
Rendall RE: Skikne. MI. (1980). Submicroscopic fibres in industrial atmospheres. In Biological effects of mineral
fibres (Vol. 30, pp. 837-843). Lyon, France: International Agency on the Research of Cancer.
Rev. F: Boutin. C: Steinbauer. J: Viallat JR: Alessandroni. P: Jutisz. P: Pi Giambattista. D: Billon-Galland. MA:
Hereng. P: Dumortier. P. (1993). Environmental pleura! plaques in an asbestos exposed population of
northeast Corsica. EurRespir J 6: 978-982.
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.
Rockoff SD: Chu. J: Rubin. LJ. (2002). Special Report: Asbestos-induced pleura! plaques-A disease process
associated with ventilatory impairment and respiratory symptoms. Clinical Pulmonary Medicine 9: 113-
124.
Rogaa WJ: Gladea BC: Ragan. NB: Anderson. HA. (1987). US prevalence of occupational pleura! thickening: A
look at chest X-rays from the first National Health and Nutrition Examination Survey. Am J Epidemiol
126: 893-900.
This document is a draft for review purposes only and does not constitute Agency policy.
7-22 DRAFT—DO NOT CITE OR QUOTE
-------
Rogan. WJ: Ragan. NB: Dinse. GE. (2000). X-ray evidence of increased asbestos exposure in the US population
fromNHANES I andNHANES II, 1973-1978. Cancer Causes Control 11: 441-449.
http://dx.doi.Org/10.1023/A:1008952426060.
Roggli VL: Vollmer. RT: Butnor. KJ: Sporn. TA. (2002). Tremolite and mesothelioma. Ann Occup Hyg 46: 447-
453. http://dx.doi.org/10.1093/annhyg/mef056.
Roggli. VL: Vollmer. RT. (2008). Twenty-five years of fiber analysis: What have we learned? Hum Pathol 39: 307-
315. http://dx.doi.0rg/10.1016/i.humpath.2007.07.005.
Roguin. A: Ben-Shahar. M: Ben-Dror. G: Cohen. I: Hazard. E. (1994). Malignant mesothelioma in families of
asbestos workers. Harefuah 126: 702-704, 764.
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. WN: Hammar. SP: Rusch. V: Dodson. R: Hoffman. S. (2001). Malignant mesothelioma from neighborhood
exposure to anthophyllite asbestos. Am J Ind Med 40: 211-214. http://dx.doi.org/10.1002/aiim.1089.
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: Ill-Ill, 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 (Vol. 28, pp. 361-407). Washington, DC:
Mineralogical Society of America.
Rothman. KJ. (1981). Induction and latent periods. AmJEpidemiol 114: 253-259.
Rothman. KJ. (1986). Significance questing. Ann Intern Med 105: 445-447.
Rothman. KJ: Greenland. S. (1998). Modern epidemiology. In (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.
Rushton. L. (2007a). Chronic obstructive pulmonary disease and occupational exposure to silica. Rev Environ
Health 22: 255-272.
Rushton. L. (2007b). Occupational causes of chronic obstructive pulmonary disease. Rev Environ Health 22: 195-
212.
Sahu. AP: Dogra. RK: Shanker. R: Zaidi SH. (1975). Fibrogenic response in murine lungs to asbestos.
Experimented Pathologic 11: 21-24.
This document is a draft for review purposes only and does not constitute Agency policy.
7-23 DRAFT—DO NOT CITE OR QUOTE
-------
Sakellariou. K: Malamou-Mitsi. V: Haritou. A: Koumpaniou. C: Stachouri. C: Dimoliatis. ID: Constantopoulos. SH.
(1996). Malignant pleura! mesothelioma from nonoccupational asbestos exposure in Metsovo (north-west
Greece): Slow end of an epidemic? Eur Respir J 9: 1206-1210.
http://dx.doi.org/10.1183/09031936.96.09061206.
Sanchez. MS: Gunter. ME: Dyar. MD. (2008). Characterization of historical amphibole samples from the former
vermiculite mine near Libby, Montana, U.S.A. European Journal of Mineralogy 20: 1043-1053.
http://dx.doi.org/10.1127/0935-1221/2008/0020-1866.
Scapoli L: Ramos-Nino. M: MartineHi. M: Mossman. B. (2004). Src-dependent ERK5 and Src/EGFR-dependent
ERK1/2 activation is required for cell proliferation by asbestos. Oncogene 23: 805-813.
http://dx.doi.org/10.1038/si.onc.1207163.
Schiller-Scotland. CF: Hlawa. R: Gebhart. J. (1994). Experimental data for total deposition in the respiratory tract of
children. Toxicol Lett 72: 137-144. http://dx.doi.org/10.1016/0378-4274(94)90020-5.
Schneider. J: Grossgarten. K: Woitowitz. HJ. (1995). [Fatal pleura! mesothelioma diseases caused by familial
household contacts with asbestos fiber dust]. Pneumologie 49: 55-59.
Schneider. J: Straif. K: Woitowitz. HJ. (1996a). Pleura! mesothelioma and household asbestos exposure [Review].
Rev Environ Health 11: 65-70. http://dx.doi.org/10.1515/REVEH.1996.ll.l-2.65.
Schneider. J: Rodelsperger. K: Pohlabeln. H: Woitowitz. HJ. (1996b). Umwelt-und Innenraum-Gefahrdung durch
Asbestfaserstaub als Risiko-und Einflubfaktor des diffusen malignen Pleuramesothelioms (DMM).
Zentralblatt fuer Hygiene und Umweltmedizin 199: 1-23.
Schneider. J: Rodelsperger. K: Briickel B: Kavser. K: Woitowitz. HJ. (1998). Environmental exposure to tremolite
asbestos: Pleural mesothelioma in two Turkish workers in Germany. Rev Environ Health 13: 213-220.
http://dx.doi.Org/10.1515/REVEH.1998.13.4.213.
Schneider. V: Maurer. RR. (1977). Asbestos and embryonic development. Teratology 15: 273-279.
http://dx.doi.org/10.1002/tera.1420150309.
Schwartz. DA: Galvin. JR: Yagla. SJ: Speakman. SB: Merchant JA: Hunninghake. GW. (1993). Restrictive lung
function and asbestos-induced pleural fibrosis: A quantitative approach. J Clin Invest 91: 2685-2692.
http://dx.doi.org/10.1172/JCI116507.
Sebastien. P: Billion-Galland. MA: Dufour. G: Bignon. J. (1980). Measurement of asbestos air pollution inside
buildings sprayed with asbestos. (EPA-560/13-80-026). Washington, DC: U. S. Environmental Protection
Agency. http://nepis.epa.gov/Exe/ZvPURL.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.
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.
Selevan. SG: Kimmel CA: Mendola. P. (2000). Identifying critical windows of exposure for children's health.
Environ Health Perspect 1083: 451-455.
Selikoff. IJ: Lee. DHK. (1978). Asbestos and disease. In. New York, NY: Academic Press.
Selikoff. IJ: Hammond. EC. (1979). Asbestos and smoking [Editorial]. JAMA 242: 458-459.
http://dx.doi.org/10.1001/jama.1979.03300050048029.
This document is a draft for review purposes only and does not constitute Agency policy.
7-24 DRAFT—DO NOT CITE OR QUOTE
-------
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.
Senyigit. A: Babayigit C: Gokirmak. M: Topcu. F: Asan. E: Coskunsel M: Isik. R: Ertem. M. (2000). Incidence of
malignant pleural mesothelioma due to environmental asbestos fiber exposure in the southeast of Turkey.
Respiration 67: 610-614. http://dx.doi.org/10.1159/000056289.
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: 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.
Short. RHP. (1952). Aspects of comparative lung growth. Proc Biol Sci 140: 432-441.
http://dx.doi.org/10.1098/rspb.1952.0071.
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 [Review]. Free Radic Biol Med 34: 1117-1129.
http://dx.doi.org/10.1016/80891-5849(03)00060-1.
Shukla. A: Lounsbury. KM: Barrett TF: Gell J: Rincon. M: Butnor. KJ: 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: Daskalopoulou. E: Tsarou. V: Pnevmatikos. I: Chloros. D: Vamvalis. C. (1992). Five cases of pleural
mesothelioma with endemic pleural calcifications in a rural area in Greece. Med Lav 83: 326-329.
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.
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 [Editorial]. Chest 122: 1885-1886.
http://dx.doi.0rg/10.1378/chest.122.6.1885.
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.
This document is a draft for review purposes only and does not constitute Agency policy.
7-25 DRAFT—DO NOT CITE OR QUOTE
-------
Smith. WE: Hubert. DP: Sobel HJ. (1980) Dimensions of fibres in relation to biological activity. In JC Wagner; W
Davis (Eds.), Biological effects of mineral fibres (Vol. 30, pp. 357-360). Lyon, France: World Health
Organization.
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. In. Cambridge, UK:
MRC Biostatistics Unit.
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.
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.
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.. (2009). A theoretical approach to the deposition of cancer-inducing asbestos fibers in the human
respiratory tract. The Open Lung Cancer Journal 2: 1-11.
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.
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 J Ind
Med 19: 701-704. http://dx.doi.org/10.1002/aiim.4700190603.
Suzuki. Y: Yuen. SR. (2001). Asbestos tissue burden study on human malignant mesothelioma. Ind Health 39: 150-
160. http://dx.doi.org/10.2486/indhealth.39.150.
Suzuki. Y: Yuen. SR: Ashley. R. (2005). Short, thin asbestos fibers contribute to the development of human
malignant mesothelioma: Pathological evidence. IntJHyg Environ Health 208: 201-210.
http://dx.doi.0rg/10.1016/i.iiheh.2005.01.015.
Tableman. M..: Kim. J.. S. (2004). Survival analysis using S: Analysis of time-to-event data. In. Virginia Beach,
VA: Chapman and Hall/CRC Press.
Tan. PJ: 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.
This document is a draft for review purposes only and does not constitute Agency policy.
7-26 DRAFT—DO NOT CITE OR QUOTE
-------
Topping. DC: Nettesheim. P. (1980). Two-stage carcinogenesis studies with asbestos in Fischer 344 rats. J Natl
Cancer Inst 65: 627-630.
Tossavainen. A: Karjalainen. A: Karhunen. PJ. (1994). Retention of asbestos fibers in the human body. Environ
HealthPerspect 102: 253-255.
Truhaut R: Chourourinkov. I. (1989). Effect of long-term ingestion of asbestos fibres in rats. In Non-occupational
exposure to mineral fibres (Vol. 90). Lyon, France: International Agency for Research on Cancer.
U.S. EPA. (U.S. Environmental Protection Agency). (1986a). Airborne asbestos health assessment update [EPA
Report]. (EPA/600/8-84/003F). Washington DC.
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
Report]. (EPA/630/R-98/003). Washington, DC. 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 Report]. (EPA/630/R-98/002). Washington, DC.
U.S. EPA. (U.S. Environmental Protection Agency). (1988a). IRIS summary for Asbestos (CASRN 1332-21-4).
Washington, DC. 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, Environmental Criteria and Assessment Office.
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 Report]. (EPA/600/FR-91/001). 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). (1991b). Health assessment document for vermiculite [EPA
Report]. (EPA/600/8-91/037). Research Triangle Park, NC.
http://nepis.epa.gov/Exe/ZvPURL.cgi?Dockev=30001LHV.txt.
U.S. EPA. (U.S. Environmental Protection Agency). (1994a). Interim policy for particle size and limit concentration
issues in inhalation toxicity studies [EPA Report]. Washington, DC: U.S. Environmental Protection
Agency, Office of Pesticide Products, Health Effects Division.
http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm?deid=186068.
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, Office of Research and Development, Office of Health and
Environmental Assessment, 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 Report]. (EPA/630/R-94/007). Washington, DC.
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.
This document is a draft for review purposes only and does not constitute Agency policy.
7-27 DRAFT—DO NOT CITE OR QUOTE
-------
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). Benchmark dose technical guidance document
[external review draft]. (EPA/630/R-00/001). Washington, DC: U.S. Environmental Protection Agency,
Risk Assessment Forum, http://www.epa.gov/raf/publications/benchmark-dose-doc-draft.htm.
U.S. EPA. (U.S. Environmental Protection Agency). (2000b). Sampling and analysis of consumer garden products
that contain vermiculite. (epa 744-r-00-010). Washington, DC.
http://nepis.epa. gov/Exe/ZyPURL.cgi?Dockev=P10014IJ.txt.
U.S. EPA. (U.S. Environmental Protection Agency). (2000c). Science policy council handbook: Risk
characterization. (EPA 100-B-00-002). Washington, D.C.: U.S. Environmental Protection Agency, Office
of Research and Development, Office of Science Policy, http://www.epa.gov/osa/spc/pdfs/rchandbk.pdf.
U.S. EPA. (U.S. Environmental Protection Agency). (2000d). 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). (2001a). EPA's actions concerning asbestos-contaminated
vermiculite inLibby, Montana. (2001-S-7). Washington, DC.
http://www.epa.gov/oig/reports/2001/montana.pdf.
U.S. EPA. (U.S. Environmental Protection Agency). (200Ib). Risk assessment guidance for superfund: Volume III -
part A, process for conducting probabilistic risk assessment. (EPA 540-R-02-002). Washington, DC: U.S.
Environmental Protection Agency, Office of Solid Waste and Emergency Response.
http://www.epa.gov/oswer/riskassessment/rags3adt/index.htm.
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.
http ://cfpub .epa. gov/ncea/cfm/recordisplav. cfm?deid=51717.
U.S. EPA. (U.S. Environmental Protection Agency). (2004). Dosimetry of inhaled particles in the respiratory tract.
(EPA/600/P-99/002bF). Research Triangle Park, NC.
U.S. EPA. (U.S. Environmental Protection Agency). (2005a). Guidelines for carcinogen risk assessment.
(EPA/630/P-03/00IF). Washington, DC. 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/cancerguidelines/guidelines-carcinogen-
supplementhtm.
U.S. EPA. (U.S. Environmental Protection Agency). (2006a). Aging and toxic response: Issues relevant to risk
assessment [EPA Report]. (EPA/600/P-03/004A). Washington, DC.
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.
http ://cfpub .epa. gov/ncea/cfm/recordisplay. cfm?deid= 158363.
U.S. EPA. (U.S. Environmental Protection Agency). (2006c). Phase 2 study data summary report for Libby,
Montana, environmental monitoring for asbestos, evaluation of exposure to airborne asbestos fibers during
routine and special activities. Libby, MT.
This document is a draft for review purposes only and does not constitute Agency policy.
7-28 DRAFT—DO NOT CITE OR QUOTE
-------
U.S. EPA. (U.S. Environmental Protection Agency). (2006d). Science policy council handbook: Peer review, 3rd
edition. (EPA/100/B-06/002). Washington, DC: U.S. Environmental Protection Agency, Science Policy
Council. http://www.epa.gov/OSA/spc/2peerrev.htm.
U.S. EPA. (U.S. Environmental Protection Agency). (2008). Framework for investigating asbestos-contaminated
superfund sites. Washington, DC.
http://www.epa.gov/superfund/health/contaminants/asbestos/pdfs/framework asbestos guidance.pdf.
U.S. EPA. (U.S. Environmental Protection Agency). (2009a). Risk assessment guidance for superfund volume I:
Human health evaluation manual (Part F, supplemental guidance for inhalation risk assessment): Final.
(EPA/540/-R-070/002). Washington, DC: U.S. Environmental Protection Agency, Office of Superfund
Remediation and Technology Innovation, http://www.epa.gov/oswer/riskassessment/ragsf/index.htm.
U.S. EPA. (U.S. Environmental Protection Agency). (2009b). Summary of outdoor ambient air monitoring for
asbestos at the Libby asbestos site, Libby, Montana (October 2006 to June 2008). Denver, CO.
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 ResultsSummarvReportJune2010.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. Denver, CO: U.S. Environmental
Protection Agency, Superfund Records Center.
U.S. EPA. (U.S. Environmental Protection Agency). (2010c). Public schools asbestos sampling report. Denver, CO.
U.S. GAP. (U.S. Government Accountability Office). (2007). Hazardous materials: EPA may need to reassess sites
receiving asbestos-contaminated ore from Libby, Montana, and should improve it's public notification
process. (GAO-08-71). Washington, DC. http://www.gao.gov/products/GAO-08-71.
Ugolini. D: Neri. M: Ceppi M: Cesario. A: Dianzani. I: Filiberti R: Gemignani. F: Landi. S: Magnani. C: Mutti. L:
Puntoni. R: Bonassi S. (2008). Genetic susceptibility to malignant mesothelioma and exposure to asbestos:
the influence of the familial factor. Mutat Res 658: 162-171. http://dx.doi.0rg/10.1016/i.mrrev.2007.08.001.
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.
Vaslet CA: Messier. NJ: Kane. AB. (2002). Accelerated progression of asbestos-induced mesotheliomas in
heterozygous p53+/- mice. Toxicol Sci 68: 331-338. http://dx.doi.0rg/10.1093/toxsci/68.2.331.
Versar. (Versar Inc.). (2003). EPA's pilot study to estimate asbestos exposure from vermiculite attic insulation.
Washington, DC: U.S. Environmental Protection Agency.
http://www.epa.gov/asbestos/pubs/insulationreport.pdf.
Vinikoor. LC: Larson. TC: Bateson. TF: Birnbaum. L. (2010). Exposure to asbestos-containing vermiculite ore and
respiratory symptoms among individuals who were children while the mine was active in Libby, Montana.
Environ Health Perspect 118: 1033-1038. http://dx.doi.org/10.1289/ehp.0901680.
Voisin. C: Marin. I: Brochard. P: Pairon. J. (1994). Environmental airborne tremolite asbestos pollution and pleural
plaques in Afghanistan. Chest 106: 974-976. http://dx.doi.org/10.1378/chest. 106.3.974.
Wagner. J: Chamberlaia M: Brown. R: Berry. G: Pooley. F: Davies. R: Griffiths. D. (1982). Biological effects of
tremolite. BrJ Cancer 45: 352-360.
This document is a draft for review purposes only and does not constitute Agency policy.
7-29 DRAFT—DO NOT CITE OR QUOTE
-------
Wagner. JC: Sleggs. CA: Marchand. P. (1960). Diffuse pleural mesothelioma and asbestos exposure in the north
western Cape Province. Occup Environ Med 17: 260-271. http://dx.doi.0rg/10.1136/oem.17.4.260.
Walker. C: Everitt J: Barrett JC. (1992). Possible cellular and molecular mechanisms for asbestos carcinogenicity.
AmJIndMed21: 253-273. http://dx.doi.org/10.1002/aiim.4700210214.
Wang. XR: Yano. E: Wang. M: Wang. Z: Christiani. DC. (2001). Pulmonary function in long-term asbestos workers
in China. J Occup Environ Med 43: 623 -629.
Wanner. A: Salathe. M: O'Riordan. TG. (1996). Mucociliary clearance in the airways. Am J Respir Crit Care Med
154: 1868-1902.
Wassermann. M: Wassermann. D: Steinitz. R: Katz. L: Lemesch. C. (1980). Mesothelioma in children. In Biological
effects of mineral fibres (Vol. 30). Lyon, France: International Agency for Research on Cancer.
Weill D: Dhillon. G: Frevder. L: Lefante. J: Glindmever. H. (2011). Lung function, radiologic changes and
exposure: analysis of ATSDR data from Libby, Montana. Eur Respir J 38: 376-383.
http://dx.doi.org/10.1183/09031936.00050210.
Weis. C. (200 la). Amphibole mineral fibers in source materials in residential and commercial areas of Libby pose
an imminent and substantial endangerment to public health
Weis. C. (200 Ib). Fibrous Amphibole Contamination in Soil and Dust at Multiple Locations in Libby Poses an
Imminent and Substantial Endangerment to Public Health: an Addendum to my Memorandum of May 10,
2000
Weiss. ST. (2010). Lung function and airway diseases. Nat Genet 42: 14-16. http://dx.doi.org/10.1038/ng0110-14.
Welch. LS: Haile. E: Dement. J: Michaels. D. (2007). Change in prevalence of asbestos-related disease among sheet
metal workers 1986 to 2004. Chest 131: 863-869. http://dx.doi.org/10.1378/chest.06-1155.
Wheeler. MW. (2005). Benchmark dose estimation using SAS. In Proceedings of the Thirtieth Annual SAS Users
Group International Conference. Gary, NC: SAS Institute, http://www2.sas.com/proceedings/sugi30/201-
30r
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.20Q53.
Whitehouse. A: Black. C: Heppe. M: Ruckdeschel J: Levin. S. (2008). Environmental exposure to Libby Asbestos
and mesotheliomas. Am J Ind Med 51: 877-880. http://dx.doi.org/10.1002/aiim.20620.
WHO. (World Health Organization). (1938). Manual of the international statistical classification of diseases,
injuries, and causes of death. Geneva, Switzerland.
WHO. (World Health Organization). (1948). Manual of the international statistical classification of diseases,
injuries, and causes of death. Geneva, Switzerland.
WHO. (World Health Organization). (1957). Manual of the international statistical classification of diseases,
injuries, and causes of death. Geneva, Switzerland.
WHO. (World Health Organization). (1967). Manual of the international statistical classification of diseases,
injuries, and causes of death. Geneva, Switzerland.
This document is a draft for review purposes only and does not constitute Agency policy.
7-30 DRAFT—DO NOT CITE OR QUOTE
-------
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.
Witschi. HR: Last. JA. (1996) Toxic responses of the respiratory system. In CD Klaassen (Ed.), Casarett and Doull's
toxicology (5th ed., pp. 443-462). New York, NY: McGraw Hill.
Witschi. HR: Last. JA. (2001) Toxic responses of the respiratory system. InLJ Casarett; CD Klaassen; J Doull
(Eds.), Casarett and Doull's toxicology: The basic science of poisons (6 ed., pp. 515-534). Columbus, OH:
McGraw-Hill Professional.
Wraith. D; Mengersen. K. (2007). Assessing the combined effect of asbestos exposure and smoking on lung cancer:
a Bayesian approach. Stat Med 26: 1150-1169. http://dx.doi.org/10.1002/sim.2602.
Wright. RS; Abraham. JL; Harber. P; Burnett BR; Morris. P; West P. (2002). Fatal asbestosis 50 years after brief
high intensity exposure in a vermiculite expansion plant. Am J Respir Crit Care Med 165: 1145-1149.
http://dx.doi.org/10.1164/rccm.2110034.
Wylie. AG; Skinner. HCW; Marsh. J; Snyder. H; Garzione. C; Hodkinsoa D; Winters. R; Mossmaa BT. (1997).
Mineralogical features associated with cytotoxic and proliferative effects of fibrous talc and asbestos on
rodent tracheal epithelial and pleura! mesothelial cells. Toxicol Appl Pharmacol 147: 143-150.
http://dx.doi.org/10.1006/taap.1997.8276.
Wylie. AG; Verkouteren. JR. (2000). Amphibole asbestos from Libby, Montana: Aspects of nomenclature.
American Mineralogist 85: 1540-1542.
Yano. E; Tanaka. K; Funakj M; Maeda. K; Matsunaga. C; Yamaoka. K. (1993). Effect of smoking on pleura!
thickening in asbestos workers. Br J Ind Med 50: 898-901. http://dx.doi.org/10.1136/oem.5Q. 10.898.
Yano. E; Wang. ZM; Wang. XR; Wang. MZ; Takata. A; Kohyama. N; Suzuki. Y. (2009). Mesothelioma in a worker
who spun chrysotile asbestos at home during childhood. Am J Ind Med 52: 282-287.
http://dx.doi.org/10.1002/aiim.20675.
Yazicioglu. S.. (1976). Pleura! calcification associated with exposure to chrysotile asbestos in southeast Turkey.
Chest 70: 43-47. http://dx.doi.0rg/10.1378/chest.70.l.43.
Yu. CP; Asgharian. B; Yea BM. (1986). Impaction and sedimentation deposition of fibers in airways. Am Ind Hyg
Assoc J 47: 72-77. http://dx.doi.org/10.1080/15298668691389388.
Yu. CP; Ding. YJ; Zhang. L; Oberdorster. G; Mast. RW; Maxim. LD; Utell MJ. (1997). Retention modeling of
refractory ceramic fibers (RCF) in humans. Regul Toxicol Pharmacol 25: 18-25.
http://dx.doi.org/10.1006/rtph.1996.1064.
Zanella. C; Posada. J; Tritton. T; Mossman. B. (1996). Asbestos causes stimulation of the extracellular signal-
regulated kinase 1 mitogen-activated protein kinase cascade after phosphorylation of the epidermal growth
factor receptor. Cancer Res 56: 5334-5338.
Zeger. SL; Thomas. D; Dominici F; Samet JM; Schwartz. J; Dockery. D; Cohen. A. (2000). Exposure
measurement error in time-series studies of air pollution: Concepts and consequences. Environ Health
Perspect 108: 419-426.
Zeren. EH: Gumurdulu. D: Roggli. VU Zorludemir. S: Erkisi. M: Tuncer. I. (2000). Environmental malignant
mesothelioma in southern Anatolia: A study of fifty cases. Environ Health Perspect 108: 1047-1050.
This document is a draft for review purposes only and does not constitute Agency policy.
7-31 DRAFT—DO NOT CITE OR QUOTE
-------
Zerva. LV: Constantopoulos. SH: Moutsopoulos. HM. (1989). Humoral immunity alterations after environmental
asbestos exposure. Respiration 55: 237-241. http://dx.doi.org/10.1159/000195740.
Zhao. XH: Jia. G: Liu. YQ: Liu. SW: Yan. L: Jin. Y: Liu. N. (2006). Association between polymorphisms of DNA
repair gene XRCC1 and DNA damage in asbestos-exposed workers. Biomed Environ Sci 19: 232-238.
Zhou. Y: Su. WC: Cheng. YS. (2007). Fiber deposition in the tracheobronchial region: Experimental measurements.
Inhal Toxicol 19: 1071-1078. http://dx.doi.org/10.1080/08958370701626634.
Zitting. AJ: Karjalainen. A: Impivaara. O: Kuusela. T: Maki. J: Tossavainen. A: Jarvisalo. J. (1996). Radiographic
small lung opacities and pleura! abnormalities in relation to smoking, urbanization status, and occupational
asbestos exposure in Finland. J Occup Environ Med 38: 602-609.
This document is a draft for review purposes only and does not constitute Agency policy.
7-32 DRAFT—DO NOT CITE OR QUOTE
-------
APPENDIX A. SUMMARY OF EXTERNAL PEER-REVIEW AND PUBLIC
COMMENTS AND DISPOSITION
A-l
-------
[This page intentionally left blank.]
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
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
This document is a draft for review purposes only and does not constitute Agency policy.
B-l DRAFT—DO NOT CITE OR QUOTE
-------
APPROVAL PAGE
This report, Particle Size Distribution Data for Libby Amphibole Structures Observed in Air at
the Libby Asbestos SuperfundSite, is approved for distribution.
Bonita
U.S. EPArRegion 8
Date
f
This document is a draft for review purposes only and does not constitute Agency policy.
B-2 DRAFT—DO NOT CITE OR QUOTE
-------
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.
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 of a
This document is a draft for review purposes only and does not constitute Agency policy.
B-3 DRAFT—DO NOT CITE OR QUOTE
-------
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.
This document is a draft for review purposes only and does not constitute Agency policy.
B-4 DRAFT—DO NOT CITE OR QUOTE
-------
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.
This document is a draft for review purposes only and does not constitute Agency policy.
B-5 DRAFT—DO NOT CITE OR QUOTE
-------
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.
In 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.
This document is a draft for review purposes only and does not constitute Agency policy.
B-6 DRAFT—DO NOT CITE OR QUOTE
-------
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.
This document is a draft for review purposes only and does not constitute Agency policy.
B-7 DRAFT—DO NOT CITE OR QUOTE
-------
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.
This document is a draft for review purposes only and does not constitute Agency policy.
B-8 DRAFT—DO NOT CITE OR QUOTE
-------
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
roto tilling.
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 -Nov 2001
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. EPA (2002)
U.S. EPA (2005)
U.S. EPA (2006);
(2007c)
U.S. EPAQOOTb);
(2007a)
U.S. EPA(2008a)
U.S. EPA(2009a)
U.S. EPA(2008b)
U.S. EPA(2009b)
U.S. EPA (2003)
various
Number of
LA
Structures
-------
Figure 3-1. Particle Size Distributions of LA Particles in Libby Air Samples
Length
LL
O
0.1
^s^
/
/*
^
y
/
^
j
/
..
X^
f
1 10
Length (um)
^-^
^^v
All Air Sam pies
100
L_
8
n ^ -
Width
^-S
f
A
s*
J
*
0
f
^^
J
All Air Sam pies
0.01 0.1 1
Width (urn)
10
Aspect Ratio
u_
O
n 9 -
/
/
/
/
/
/
/
f
0.1 1 10
Aspect Ratio
/
!/
s
f
s*
^
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 81%
2,366 8%
3,150 11%
54 0.2%
77?/'s document /'s a draft for review purposes only and does not constitute Agency policy.
B-10 DRAFT—DO NOT CITE OR QUOTE
-------
Figure 3-2. Particle Size Distributions of LA Particles in Libby Air Samples by Structure Type
0.9 •
0.8 •
0.7 •
0.6 •
Q 0.5 •
O
0.4 •
0.3 •
0.2 -
0.1 •
Fibers
^— Matrice
^—Bundle
0.1
s
X
/v
Sj
s'
/
/*
f
/
J
J
/
(
/
/
f
/
/,
/
/
f
s?
// /
' /
/
f
&*•
^>
S
^~
^
^^
•••
^
1 10 100
Length (urn)
1.0
0.9 •
0.8 -I
0.7
0.6
Q 0.5
O
0.4
0.3
0.2
0.1
0.0
•Fibers
-Matrices
-Bundles
0.1
Aspect Ratio
10
0.9 •
0.8 •
0.7 •
L_
O
Fibers
Matrice
Bundle
s
0.01
_.
^C-
f
^
t
t
y
~r
<*
<*
<*
s
^
.>
&
t
/
9*"*^
~/__
/
'
-•^
0.1 1 10
Width (urn)
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.
TTz/'s1 document /'s a draft for review purposes only and does not constitute Agency policy.
B-11 DRAFT—DO NOT CITE OR QUOTE
-------
Figure 3-3. Particle Size Distributions of LA Particles in Libby Air Samples by Operable Unit (OU)
100
1.0
0.9 H
0.8
0.7
0.6
S
0.4
0.3
0.2
0.1
0.0
_
•OU1
•OU2
•OU3
•OU4
•OU5
0.01
1.1 1
Width (urn)
10
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.
OU
1
2
3
4
5
N Structures
447
7,421
4,382
13,005
335
This document is a draft for review purposes only and does not constitute Agency policy.
B-12 DRAFT—DO NOT CITE OR QUOTE
-------
Figure 3-4. Particle Size Distributions of LA Particles in Libby Air Samplesby Air Type
1.0
0.9
0.8
0.7
0.6
Q 0.5
O
0.4
0.3
•Indoor ABS
•Outdoor AmbientAir
•Outdoor ABS
10
Length (urn)
100
0.9 •
0.8 •
0.7 •
n R •
L_
Q n R •
8 °-5
n A •
0\ .
n n •
Ou
— Ou
oorABS
tdoor AmbientAir
tdoor ABS
0.01
J-
^r-
f
* (
*}
J
}
f
J
^
f
J
J
/
^
J
t
i^"
/
0.1 1 10
Width (urn)
0.9 •
0.8 •
0.7 •
n R •
L_
o
n 9 •
O-i .
n n •
IndoorABS
^^"Outdoor Ambient A r
Outdoor ABS
0.1
L
r
/
j
1
j
f
A
I
/
\^
/^
/ J
f /
f
J
'
^^
^
^=
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
AmbientAir
IndoorABS
Outdoor ABS
N Structures
136
891
5,953
This document is a draft for review purposes only and does not constitute Agency policy.
B-13 DRAFT—DO NOT CITE OR QUOTE
-------
Figure 3-5. Particle Size Distributions of LA Particles in Libby Air Samples by Preparation Method
0.9 •
0.8 •
0.7 •
0.6 •
LJ_
Qn R •
O
0.4 •
0.3 •
0.2 •
0.1 •
n n .
D rect
^^"Indirect
0.1
^X^
jt
2_
/
/
/J
X
/,
/
/
/
9
f
>
y
x-
X
(
+>*'
^•*
1 10 100
Length (urn)
1 n i
0.9 •
0.8 •
n 7 •
0.6 •
u_
Q 0.5 •
O
n A •
n ^ •
n o •
n 1 •
n n •
— Direct
0.01
fr~r
^
&
>
jj
V
,4
*
X^
V
0.1 1
Width (urn)
10
0.9 •
U.o •
n 7 •
n R •
LL
Q 0.5 •
O
0.4 •
n ^ .
n "> .
n n •
^^^l H' t
0.1
/
J
y
j
z
f
/
f
_^_
/
f
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
TTz/'s1 document /'s a draft for review purposes only and does not constitute Agency policy.
B-14 DRAFT—DO NOT CITE OR QUOTE
-------
Figure 3-6. Particle Size Distributions of LA Particles in Libby Air Samples by Analysis Method
0.9 •
0.8 •
n 7 •
n R •
u_
Qn R .
O
0.4 •
^— ISO
AH ERA
0.1
j
^^
/
2
^
/
/
>
j/
>
^
'
>
/
f
S
Xx
X
T
i
^
^"
1 10 100
Length (um)
0.9 •
0.8 •
0.6 •
LL
Q n ^ .
0
n 4 .
n ^ .
n o .
n 1 .
n n .
^— ISC
Ah
0.01
ERA
^
r
{
}
$
r
j
^
^
,*•
j
^
s
^
/
J
^
^^"
0.1 1 10
Width (urn)
1 n i
n Q .
n R .
LJ_
Q 0.5 •
O
0.4 •
n ^ .
n 9 -
n 1 -
n n •
^— ISO
AH
0.1
=RA
^
^
y
^
y
/
!/
/,
//
//
//
/
//
7
*^
/
&
^
^
w
w—
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
This document is a draft for review purposes only and does not constitute Agency policy.
B-15 DRAFT—DO NOT CITE OR QUOTE
-------
5.0 REFERENCES
AHERA. 1986. Asbestos Hazardous Emergency Response Act. Title 20, Chapter 52, Sec.
4011. Public Law 99-519.
ISO. 1995. International Organization for Standardization (ISO). Ambient Air - Determination of
Asbestos Fibres - Direct-Transfer Transmission Electron Microscopy Method. ISO 10312:1995(E).
USEPA. 2000. Sampling and Quality Assurance Project Plan Revision 1 for Libby, Montana.
Environmental Monitoring for Asbestos. Baseline Monitoring for Source Area and Residential
Exposure to Tremolite-Actinolite Asbestos Fibers. Report prepared by U.S. Environmental
Protection Agency Region. January 4, 2000.
USEPA. 2001. Phase 2 Sampling and Quality Assurance Project Plan (Revision 0) for
Libby, Montana. Environmental Monitoring for Asbestos. Evaluation of Exposure to
Airborne Asbestos Fibers During Routine and Special Activities. Report prepared by U.S.
Environmental Protection Agency Region 8. March 2001.
USEPA. 2002. Final Sampling and Analysis Plan, Remedial Investigation, Contaminant Screening
Study, Libby Asbestos Site, Operable Unit 4. Report prepared by U.S. Environmental Protection
Agency Region 8, with technical support from CDM. April 30, 2002.
USEPA. 2003. Final Sampling and Analysis Plan Addendum, Post Clean-up Evaluation Sampling,
Contaminant Screening Study, Libby Asbestos Site, Operable Unit 4. Report prepared by U.S.
Environmental Protection Agency Region 8, with technical support from CDM and Syracuse
Research Corporation. December 1, 2003.
USEPA. 2005. Supplemental Remedial Investigation Quality Assurance Project Plan for Libby,
Montana. Revision 1. U.S. Environmental Protection Agency Region 8. August 5, 2005.
USEPA. 2006. Sampling and Analysis Plan for Outdoor Ambient Air Monitoring at the Libby
Asbestos Site. Revision 1. Report prepared by U.S. Environmental Protection Agency Region 8,
with technical support from CDM and Syracuse Research Corporation. December 7, 2006.
This document is a draft for review purposes only and does not constitute Agency policy.
B-16 DRAFT—DO NOT CITE OR QUOTE
-------
USEPA. 2007a. Sampling and Analysis Plan for Outdoor Ambient Air Monitoring - Operable
Units 1, 2, 5, and 6. Final Addendum prepared by U.S. Environmental Protection Agency Region 8,
with technical support from CDM and Syracuse Research Corporation. July 3, 2007
USEPA. 2007b. Sampling and Analysis Plan for Activity-Based Outdoor Air Exposures,
Operable Unit 4, Libby, Montana, Superfund Site. Final. U.S. Environmental Protection
Agency, Region 8. July 6, 2007.
USEPA. 2007c. Sampling and Analysis Plan for Activity-Based Indoor Air Exposures,
Operable Unit 4, Libby, Montana, Superfund Site. Final. U.S. Environmental Protection
Agency, Region 8. July 6, 2007.
USEPA 2008a. Request for Modification to Laboratory Activities LB-000031 A. Requested by
Lynn Woodbury of Syracuse Research Corporation. January 18, 2008.
USEPA. 2008b. Standard Operating Procedure for TEM Data Review and Data Entry
Verification. SOP No. EPA-LIBBY-09 (rev 1). Prepared by U.S. Environmental Protection
Agency, Region 8, with technical support from SRC, Inc. March 5, 2008.
USEPA. 2008c. Phase II Sampling and Analysis Plan for Operable Unit 3 Libby Asbestos
Superfund Site. Part B: Ambient Air and Groundwater. Prepared by U.S. Environmental
Protection Agency Region 8, with technical support from Syracuse Research Corporation and
NewFields Boulder LLC. July 2, 2008.
USEPA. 2008d. Final Sampling and Analysis Plan Libby Public Schools - Stationary Air
Sample Collection Libby Asbestos Site Libby, Montana. Prepared by U.S. Dept. of
Transportation and CDM Federal Programs Corp. with technical support from Syracuse
Research Corporation. December 5, 2008.
USEPA. 2009a. Remedial Investigation for Operable Unit 3 Libby Asbestos Superfund Site.
Phase HI Sampling and Analysis Plan. Prepared by U.S. Environmental Protection Agency Region
8, with technical support from Syracuse Research Corporation and NewFields Boulder LLC. May
26, 2009.
USEPA. 2009b. Final Sampling and Analysis Plan for Activity-Based Outdoor Air Exposures at
Libby Public Schools Libby Asbestos Site Libby, Montana. Prepared by U.S. Dept. of
Transportation and CDM Federal Programs Corp. with technical support from SRC, Inc. July 17,
2009.
This document is a draft for review purposes only and does not constitute Agency policy.
B-17 DRAFT—DO NOT CITE OR QUOTE
-------
1 APPENDIX C. CHARACTERIZATION OF AMPHIBOLE FIBERS FROM ORE
2 ORIGINATING FROM LIBBY, MT; LOUISA COUNTY, VA; AND PALABORA,
3 REPUBLIC OF SOUTH AFRICA
By
David L. Berry, PhD
4 U.S. EPA Region 8
5 Ecosystems Protection and Remediation
6 1595 Wynkoop Street [8EPR-PS]
7 Denver, CO 80202-1129
This document is a draft for review purposes only and does not constitute Agency policy.
C-1 DRAFT—DO NOT CITE OR QUOTE
-------
1 The O.M. Scott plant in Marysville, OH manufactured a number of products including
2 fertilizers, dyes, and pesticides that were bound to a vermiculite carrier as a delivery vehicle.
3 The plant received ore from Enoree, SC; Louisa County, VA; Libby, MT; and Palabora,
4 Republic of South Africa, which was processed in an exfoliation furnace to produce vermiculite
5 used in the manufacture of their commercial products. Only ore from South Carolina was used
6 in 1957 and 1958. From 1959 to 1971, ores from South Carolina and Libby, MT were used.
7 From 1972 to 1980, ores from Libby, MT, South Africa, and Virginia were used. No ore from
8 Libby, MT was used after 1980. Only ore from South Africa and Virginia was used after 1980
9 (see Appendix F).
10 The U.S. Environmental Protection Agency (EPA) Region 8 obtained samples of ore
11 from Libby, MT, South Africa, and Virginia from Dr. James Lockey, University of Cincinnati,
12 and analyzed the samples to determine mineralogy and particle size distribution (length, width,
13 and aspect ratio) using transmission electron microscopy (TEM) and energy dispersive
14 spectroscopy (EDS) to identify the nature of the amphibole fibers. Dr. Lockey obtained the
15 South African and Virginia ore samples from the Marysville, OH facility in 1980 and the Libby,
16 MT ore (Libby #3 ore) from an expansion plant in Salt Lake City, UT, in 1981. Region 8 was
17 unable to obtain vermiculite or ore from the Enoree, SC mine complex.
18 The ore from the Rainey Creek complex (Vermiculite Mountain Mine, Libby, MT)
19 resides in large ultramafic intrusive bodies that are rich in biotite, pyroxenite, and biotitite, a rock
20 comprised of almost pure biotite. The ultramafic intrusions are cut by deposits of syenite and
21 carbonatite, and much of the biotite has been hydrothermally altered to hydrobiotite and
22 vermiculite (Meeker et al., 2003; Frank and Edmund, 2001). The pyroxenite has been altered to
23 fibrous soda-rich amphiboles, and contacts with pyroxenite surrounding the biotitite contain the
24 vermiculite ore zone containing diopside, hydrobiotite, and apatite. Fibrous and nonfibrous
25 amphiboles are located in both veins and disseminated throughout the intrusive rock along
26 cleavage planes of pyroxene. Amphiboles from Vermiculite Mountain had been referred to as
27 soda tremolite, richterite, soda-rich tremolite, tremolite asbestos, and richterite asbestos by a
28 number of investigators. In 2000, Wylie and Verkouteren (2000) identified winchite as the
29 principal amphibole in the Vermiculite Mountain deposit based on chemical investigation
30 referencing the classification system of Leake et al. (1997) and optical properties. Meeker et al.
31 (2003) investigated amphibole types from the mine complex using electron probe microanalysis
This document is a draft for review purposes only and does not constitute Agency policy.
C-2 DRAFT—DO NOT CITE OR QUOTE
-------
1 and X-ray diffraction analysis and reported the presence of winchite, richterite, tremolite, and
2 magnesioriebeckite. Magnesio-arfvedsonite and edenite were detected in low abundance. The
3 amphibole composition of the Libby Amphiboles is roughly winchite, richterite, tremolite,
4 magnesio-riebeckite, magnesio-arfvedsonite, and edenite (84:11:6:<1 :<1 :<1). The O.M. Scott
5 facility received ore from the Vermiculite Mountain mine complex, Libby, MT from 1959
6 through 1980.
7 The Palabora Igneous Complex, located near Phalaborwa, Republic of South Africa, is
8 the location of the Palabora mine. The Palabora ore deposit shares many features with the
9 Vermiculite Mountain mine complex—including zoned deposits with ultramafic rocks
10 (pyroxenite) and intrusion by alkalic rock, primarily syenite. The primary mica at Palabora is
11 phlogopite rather than biotite, and the primary alteration product that forms vermiculite ore is
12 hydrophlogopite rather than hydrobiotite (Schoeman, 1989).
13 The Palabora ore is reported to contain little or no asbestiform fibers based on polarized
14 light microscopy by the Institute of Occupational Medicine in Edinburgh (IOM Consulting,
15 2008). Crude vermiculite from the Palabora complex was also reported to be free of asbestiform
16 fibers by polarized light microscopy (IOM Consulting, 2008). In both reports, the analysis by
17 polarized light microscopy was conducted with a detection limit of 1 ppm, and, since no
18 chrysotile or amphibole structures were detected, no further analysis by electron microscopy and
19 X-ray diffraction were conducted.
20 The ore from the Virginia Vermiculite mine in Louisa County, VA is described as mafic
21 rock intruded by a series of small pegmatites (Gooch, 1957). Meisinger (1979) classified the
22 deposits as Type 3, similar to the ores from Enoree, SC. The formations consist of potassic
23 ultramafic bodies, primarily biotite. The vermiculite ores are found primarily in hydrobiotite
24 portions of the biotite intrusions. The hydrobiotite deposits are preferentially mined because of
25 better commercial properties compared to vermiculite.
26 There is limited information on the asbestos content of the ores from the Louisa County
27 deposit. Rohl and Langer (1977) reported both chrysotile and amphibole fibers in six ore
28 samples from the Louisa County deposit. The chrysotile was reported as fibers and bundles
29 while the amphiboles fibers were classified as actinolite. Moatamed et al. (1986) analyzed a
30 Virginia ore sample collected at a processing plant in Salt Lake City, UT and reported traces of
31 fibrous amphibole asbestos identified as actionlite in the form of cleavage fragments having low
This document is a draft for review purposes only and does not constitute Agency policy.
C-3 DRAFT—DO NOT CITE OR QUOTE
-------
1 aspect ratios. Amphibole content for both unexfoliated and exfoliated ores ranged up to 1.3%
2 amphibole asbestos.
3 Ores from the Enoree, SC deposits are primarily hydrobiotite and biotite in origin.
4 Fluroapatite is a common mineral collocated with the hydrobiotite. Zircon is also widely
5 dispersed throughout the plutons along with minor accessory minerals including talc, chlorite,
6 chromite, rutile, titanite, corundum, anatase, and amphibole asbestos (Hunter, 1950). The
7 amphibole asbestos identified in the vermiculite deposit at Enoree, SC has been classified as
8 tremolite (Libby. 1975).
9 As previously noted, EPA Region 8 obtained samples of ore from Libby, MT, South
10 Africa, and Virginia from Dr. James Lockey, University of Cincinnati, and analyzed the samples
11 to determine the particle-size distribution (length, width, and aspect ratio), using TEM and EDS
12 to identify the mineral composition of the amphibole fibers. Region 8 was unable to acquire a
13 sample of ore from the South Carolina Enoree mine complex for analysis. Region 8 conducted
14 analysis of the ore and exfoliated materials to connect the exposures of workers to mineral fibers
15 in Marysville, OH, to the ore originating in Libby, MT. The connection is based on fiber
16 morphology, mineralogy, and fiber-size similarities.
17 In order to analyze the fibers from the ore and vermiculite bulk material, the fibers must
18 be loaded onto filters and prepared for analysis by TEM. Three potential methods were
19 considered for transferring the fibers from the bulk material to filters: water elutriation,
20 glove-box transfer, and the fluidized bed asbestos segregator (FBAS). Of these three methods,
21 only the glove-box and FBAS involved physical disturbance of the bulk material to elutriate
22 fibers into the air that might be similar to handling and processing of ore in the Marysville, OH
23 plant. Due to the limited quantity of test material available for analysis, Region 8 employed the
24 FBAS as an analytical instrument to load the mineral fibers onto filters for TEM analysis.
25 Briefly, samples of ore and vermiculite were prepared following the procedure outlined
26 by Bern et al. (2002). Samples were dried, ground with a Wylie mill and mortar and pestle, and
27 sieved through a 230-|im (60 mesh) sieve. Samples (exactly 2.0 g) were mixed with 18 g of
28 analytical silica sand and placed in a FBAS vessel to load 25-mm mixed cellulose ester air
29 sampling filters (0.8-|i pore size). The FBAS was run for 3 minutes to load the filter cassettes
30 with sufficient fibers for analysis by TEM. Five filters were loaded for each of the ore and
This document is a draft for review purposes only and does not constitute Agency policy.
C-4 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
vermiculite samples. After loading, the filters were prepared for TEM analysis by mounting on
copper grids, carbon coating, and subjected to TEM analysis (TEM-ISO 10312 method).
The laboratory followed fiber counting rules detailed in the Quality Assurance Project
Plan for the specific study using Libby-specific laboratory modifications. Total amphibole fibers
and Phase Contrast Microscopy equivalent (PCMe) fibers were counted for each of the
ore/vermiculite samples as described in Appendix B. A total of 1.0 mm2 area or a total of
200 asbestos structures were counted to achieve the desired analytical sensitivity (1/g; 1.5 x 104).
DS was performed on selected samples from each of the vermiculite/ore samples to provide
mineral characterization of individual fibers. Fiber counts were recorded on National Asbestos
Data Evaluation Sheet data sheets for further analysis. Only the Libby, MT vermiculite and
Libby, MT ore samples had sufficient fibers detected to construct 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, and the grid area measurement was exceeded prior to the fiber count
r\
metric (167 grid openings -1.0 mm ).
Table C-l. Fiber detected in ore and expanded product
Sample type
Virginia Ore
Virginia Expanded
South Africa Ore
South Africa Expanded
Libby #3 Ore
Libby Expanded
Grid openings
167
167
167
167
167
167
Structures counted
LA
0
1
2
0
320
100
OA
0
0
0
0
0
0
c
0
0
2
0
0
0
Concentration (s/g)
LA
0
13,008
26,403
0
1,393,873
468,213
OA
0
0
0
0
0
0
c
0
0
26,403
0
0
0
19
20
21
22
23
24
25
LA = Libby Amphibole, OA = Other amphibole, C = Chrysotile. Note: the designation of fibers as Libby Amphibole
in this instance reflects only a qualitative morphological comparison to amphiboles of the Libby, MT 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 bulk material.
This document is a draft for review purposes only and does not constitute Agency policy.
C-5 DRAFT—DO NOT CITE OR QUOTE
-------
1 Virginia expanded and South African ore contain amphibole structures represented by low fiber
2 counts. South African ore also contained chrysotile fibers as determined by morphology and
3 EDS analysis. The absence of fibers detected in the Virginia ore and the South African-
4 expanded materials probably represents actual low fiber content of the ore and is a function of
5 the detection limit for the structure analysis. The estimation of structures per gram of material
6 indicated that there were 13,000 to 26,000 fibers per gram of bulk material, which was
7 approximately 18 times lower than the Libby, MT ore samples. The decrease in fibers found in
8 the Marysville, OH facility after 1980 when only ore from Virginia, Palabora, and South
9 Carolina was used (see Appendix F) is consistent with the findings of low fiber counts for the
10 Virginia and Palabora materials. In addition, numerous nonasbestiform minerals were also
11 detected including biotite, micas, and pyroxenes in the bulk materials from Virginia and South
12 Africa.
13 Amphiboles are a complex group of minerals characterized by double chains of silicate
14 tetrahedrons and the generic chemical formula of A^iB2C5T^O22[OH\2 where A, B, C, and T
15 represent the various cations. The modern classification system of amphiboles is described in
16 Leake et al. (1997). To classify the mineral species of the amphibole, it is not sufficient to
17 determine its composition; the various cations must be assigned to the specific^, B, C, and T
18 sites. The cutoffs of the compositional ranges allowed for each amphibole mineral species are
19 based on the number of the cations in the various sites. The methodology to classify an
20 amphibole is to first determine its elemental compositions (e.g., as expressed as weight percent
21 oxide for each element or as atomic percent for each element). Then a normalized routine is
22 applied to the raw elemental measurements to calculate the number of each of the cations
23 contained in one formula unit. (This is a simple arithmetic calculation since the cation percents
24 have been measured, and the stoichiometry must balance the charges of the cations and anions.)
25 Generally, one formula unit is assumed to contain 23 oxygens. Next, the sites are filled up by
26 assigning cations to them subsequently, specifically:
27
28
This document is a draft for review purposes only and does not constitute Agency policy.
C-6 DRAFT—DO NOT CITE OR QUOTE
-------
1 T: Si4+, A13+, and Ti4+.
2 C: A13+ and Ti4+ (only after the T sites are filled first) and then Mg2+, Fe2+, Fe3+, and
3 then Mn2+.
4 B: Any remaining Mg2+, Fe2+, and Mn2+ (after the C sites are filled), all Ca2+, then
5 Na+if there is any room left.
6 A: Na+ and K+ only.
7
8
9 Once the cations are assigned to their sites, it is a simple matter to classify the minerals
10 based on the cutoffs of the composition field allowed for each mineral.
11 The Libby Amphibole asbestos1 group of minerals is a complex group of amphiboles
12 consisting of six minerals:
13
14
15 • Winchite, CaNa[Mg, Fe2+]4[Al, Fe3+]Si8O22 [OH]2
16 • Richterite, NaCaNa [Mg, Fe2+, Mn, Fe3+]5Si8O22[OH]2
17 • Tremolite, Ca2Mg5Si8O22[OH]2
18 • Magnesio-riebeckite, Na2[Mg3, Fe3+2]Si8O22[OH]2
19 • Magnesio-arfvedsonite, NaNa2[Mg4,Fe3+]Si8O22[OH]2
20 • Edenite, NaCa2Mg5Si7AlO22[OH]2
21
22
23 Libby Amphibole is characterized by a low amount of Al in the T site—and a
24 correspondingly high Si content—so, according to Leake's (1997) classification, if the Si
25 (expressed as atoms per formula unit, apfu) is at least 7.5, and Al content in the T site is <0.5, all
26 6 Libby Amphibole types can be plotted on a graph of Na content of the B site versus the
27 (Na + K) content in the A site. This approach was described by Meeker et al. (2003) for the
28 Rainy Creek complex.
29 EDS spectra (TEM/EDS) were collected from all amphibole fibers found in the South
30 Africa and Virginia samples, and six randomly selected Libby Amphibole asbestos fibers in each
lrThe term "Libby Amphibole asbestos" is used in this document to identify the mixture of amphibole mineral fibers
of varying elemental composition (e.g., winchite, richterite, tremolite, etc.), that have been identified in the Rainy
Creek complex near Libby, MT. It is further described in Section 2.2.
This document is a draft for review purposes only and does not constitute Agency policy.
C-7 DRAFT—DO NOT CITE OR QUOTE
-------
1 of the Libby, MT ore and Libby, MT expanded samples. Two bundles of asbestiform serpentine
2 (chrysotile) were found in the South African ore sample. EDS spectra were collected for one of
3 the bundles. The chemical formula of serpentine is Mg3Si2Os[OH]4. The EDS software package
4 collected and summarized each spectrum to determine the atomic percent of each element of
5 interest.
6 Several assumptions were made in the treatment of the TEM/EDS data:
7
8
9 1. Numbers of cations per formula unit are calculated on the basis of 23 oxygens. This may
10 or may not be correct because an [OH] site in the amphibole crystal can be occupied by
11 either OFT, F~, Cl~ or O2 . The calculated cation numbers will be affected if a significant
12 quantity of O2 is in the OH site.
13 2. A persistent problem with amphiboles is that they can contain both ferric [3+] and ferrous
14 [2+] iron in the same crystal. For the purposes of this report all Fe was assumed to be
15 Fe2+. A method for calculating the ratio of Fe2+ to Fe3+ is described in Leake et al.
16 (1997), but it is very complex, applies to polished sections, and was not attempted for this
17 report.
18 3. For the purposes of this report, the T sites were assumed to be filled completely full to
19 8 apfu, and the C sites were assumed to be completely full to 5 apfu. All Ca and any Mg,
20 Fe, and Mn remaining after the C site was full were then assigned to the B site. Next, Na
21 was assigned to the B site until it was full (2 apfu), then any remaining Na and all K were
22 assigned to the A site.
23
24
25 Applying these assumptions to the TEM/EDS data produces a useable graph of the Na
26 and K content of the amphibole fibers. As shown in Figure C-l, Libby #3 ore and Libby #3
27 Expanded amphiboles were characteristic of winchite and tremolite. Virginia Expanded and
28 South African ore both contained amphibole fibers characteristic of non-Libby (Na and K) in the
29 tremolite series.
30
31
This document is a draft for review purposes only and does not constitute Agency policy.
C-8 DRAFT—DO NOT CITE OR QUOTE
-------
Sodium and Potassium Content of Individual
Amphibole Fibers as Measured by TEM/EDS
4 r\
n A Site (atoms per formula
unit)
D O -
D Ul C
n
Edenite Richterite
A
A D
Tremolite-
Actinolite
Winchite
n A n A
A
I
• i
* 0.0 0.5 1.0 1
z Na in B Site (atoms per formula unit)
o Virginia Expanded
o South Africa Ore
A Libby #3 Ore
n Libby #3 Expanded
5
I
2
3
4
5
6
9
10
11
12
13
14
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^4 site as described by Meeker et al. (2003), the mineral species of the
Marysville, OH fibers can be described as:
• The single Virginia amphibole asbestos fiber is an actinolite
• Both of the South African amphibole fibers are tremolite
• 8 of the Libby Amphibole asbestos fibers from Libby, MT are winchite
• 4 of the Libby Amphibole asbestos fibers from Libby, MT are tremolite
This document is a draft for review purposes only and does not constitute Agency policy.
C-9 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
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
Length
Libby #3 Ore
Libby #3 Ore Expanded
LibbyAir
Length (um)
100
Width
Libby #3 Ore
Libby #3 Ore Expanded
Libby Air
0.1
Width (urn)
Aspect Ratio
10
Libby #3 Ore
l_ibby#3 Ore Expanded
LibbyAir
10
Aspect Ratio
100
1000
Figure C-2. Fiber-size distribution of Libby Amphibole asbestos.
This document is a draft for review purposes only and does not constitute Agency policy.
C-10 DRAFT—DO NOT CITE OR QUOTE
-------
1 Fiber-size distributions for amphibole fibers from the Libby #3 ore and Libby #3
2 expanded sources were conducted on the fibers counted during the TEM analysis of the filter
3 grids. Due to the low fiber count detected in the Virginia and South Africa sources, it was not
4 possible to develop a fiber-size distribution for these fibers. The Libby Amphibole asbestos
5 fiber-size data were plotted as a cumulative distribution frequency for fiber length, fiber width,
6 and aspect ratio. These data were compared to Libby Amphibole asbestos fibers collected in
7 Libby, MT as part of EPA's ongoing ambient air monitoring program and the Libby Asbestos
8 Superfund site (see Appendix B). The Libby, MT ore and expanded material showed an
9 increased frequency of longer and wider fibers than the fibers from the Libby, MT ambient
10 air-sampling program. Aspect ratios were nearly identical. The differences between the length
11 and width frequency were not outside of the expected range for Libby Amphibole asbestos fibers
12 and were consistent with fiber-size distributions for soil activity-based-sampling data from
13 Libby, MT.
14 Based on the TEM morphological analysis of filter grids, TEM/EDS analysis for the fiber
15 mineralogy, and the fiber-size distribution data, it can be concluded that the amphibole fibers
16 detected in the Libby # 3 ore samples from the Salt Lake Expansion facility are consistent with
17 data from authentic Libby Amphibole fibers (Meeker et al., 2003) found in Libby, MT (see also
18 Appendix B). Further, ore samples from Virginia and South Africa contained amphibole and
19 chrysotile fibers but at a much lower frequency of detection than the Libby Amphibole ore as
20 reported in Appendix F.
21
22
This document is a draft for review purposes only and does not constitute Agency policy.
C-l 1 DRAFT—DO NOT CITE OR QUOTE
-------
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 USGS Open File Report.
USGS.
Frank. D: Edmund. L. (2001). Feasibility for identifying mineralogical and geochemical traces from vermiculite ore
deposits. (EPA 910-R-01-002). Seattle, WA: U.S. EPA Region 10.
Gooch. EO. (1957). Vermiculite. 3:1-5.
Hunter. CE. (1950). Vermiculite of the southeastern states. In Symposium on mineral resources of the southeastern
United States. Knoxville, TN: University of Tennessee Press.
IOM Consulting. (2008). Sampling and analysis of crude vermiculite samples for possible asbestiform fibre and
quartz content. (609-02386). Surrey, England: Palabora Mining Co, Palabora Europe Ltd.
Leake. BE: Woollev. AR: Arps. CES: Birch. WD: Gilbert. MC: Price. 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. Mineral Mag 61: 295-321.
Libbv. SC. (1975) The origin of potassic ultramafic rocks in the Enoree "Vermiculite" District, South Carolina
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. American
Mineralogist 88: 1955-1969. http://dx.doi.org/10.1093/annnvg/38.inhaled_particles VII.639.
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. EnvironRes 41: 207-218.
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.
Schoeman. JJ. (1989). Mica and vermiculite in South Africa. J South African Institute of Mining and Mineralogy 89:
1-12.
Wylie. AG: Verkouteren. JR. (2000). Amphibole asbestos from Libby, Montana: Aspects of nomenclature.
American Mineralogist 85: 1540-1542.
This document is a draft for review purposes only and does not constitute Agency policy.
C-12 DRAFT—DO NOT CITE OR QUOTE
-------
1 APPENDIX D. ANALYSIS OF SUBCHRONIC AND CHRONIC STUDIES AND
2 CANCER BIOASSAYS IN ANIMALS AND MECHANISTIC STUDIES
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
D.I. SUBCHRONIC AND CHRONIC STUDIES AND CANCER BIOASSAYS
D.I.I. Oral
McConnell et al. (1983) describe part of a National Toxicology Program study (NTP,
1990a, b, 1988, 1985) performed to evaluate the toxicity and carcinogen!city 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 (Governeur
Talc Co., Governeur, 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-l). Citing the Stanton et al. (1981) paper, McConnell et al. (1983) stated that crystalline
tremolite can become fibrous upon grinding. Tremolite was 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-l. Fiber characteristics and distribution of 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 fim, <5 fim
1.78
61
9.85
>5 um, <10 Jim
2.87
17
3
>10 um
5.22
49
8
23
24
25
26
27
28
29
30
"Average groups, more detailed in primary paper.
Source: McConnell et al. (1983).
No significant tumor induction was observed in the animals with oral exposure to
tremolite. Although non-neoplastic lesions were observed in many of the aging rats, these were
This document is a draft for review purposes only and does not constitute Agency policy.
D-1 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
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, in general,
show decreased toxicity and carcinogenicity as compared to inhalation and
implantation/injection studies.
D.I.2. Inhalation
Davis et al. (1985) performed a chronic inhalation study examining response to tremolite
asbestos. Groups of 48 specific-pathogen-free (SPF) male Wistar rats 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 10-100-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-2 and D-3). No
pulmonary tumors were observed in the controls.
Table D-2. 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
(n = 12)
-
-
14.5 (3.8-26.9)
""Percentage of 100 squares counted in lung tissue area.
Percentage of total lung tissue area.
SD = standard deviation.
Source: Adapted from Davis et al. (1985).
This document is a draft for review purposes only and does not constitute Agency policy.
D-2 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
Table D-3. Tumors (benign and malignant) produced by tremolite exposure
4
5
6
7
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Tumor site
Control (n = 36)
Tremolite (w = 39)
Pulmonary
Adenomas
Adenocarcinomas
Squamous carcinomas
Mesotheliomas
0
0
0
0
2
8
8
2
Other organ systems
Digestive/peritoneal
Urinogenital
Endocrine
Musculoskeletal, integumentary
Reticuloendothelial/vascular
5
3
3
5
20
3
1
5
5
15
Source: Adapted from Davis et al. (1985).
Although Davis et al. (1985) did not describe the 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-2). 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-3 and appear to be unrelated to exposure.
Although a method for an injection study is described in Davis (1985), only the inhalation results
are presented. This same tremolite was used in later intraperitoneal injection experiments (Davis
et al., 1991) and might be what the authors are referring to in this article.
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-4). The long-term effects from the same exposure were
described in Bernstein et al. (2005) (6 hours per day, 5 days per week). This study describes the
full results through 1 year after cessation of tremolite exposure in Wistar rats (n = 56). The
tremolite samples were chosen to have 100 fibers/mL of fibers longer than 20 |im 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
This document is a draft for review purposes only and does not constitute Agency policy.
D-3 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
Table D-4. Chrysotile and tremolite fiber characteristics of fibers used in
inhalation exposure studies in rats
4
5
6
7
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
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
(urn) ± 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).
length >5 um, diameter <3 um, or 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 control, which
showed erythrophagocytosis (minimal severity) only at 180 days.
Table D-4 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 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-um long and
7 million fibers 5-20-um 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, which is a consistent observation with the time frame observed in
other studies (i.e., 1-year postexposure) (Smith, 1978).
This document is a draft for review purposes only and does not constitute Agency policy.
D-4 DRAFT—DO NOT CITE OR QUOTE
-------
1 D.I.3. Intratracheal Instillation
2 A recent study by Putnam et al. (2008) was designed to explore gene-environment
3 interactions in the development of asbestos-related diseases. C57B1/6 mice were exposed once
4 to either Libby Amphibole asbestos 1 (Six Mix) (100 ug via intratracheal instillation); crocidolite
5 (100 ug via intratracheal instillation); or saline (30 uL via intratracheal instillation).
6 Characteristics of fibers are described in Table D-5. Animals were sacrificed, and the lungs were
7 harvested 6 months postinstillation. The left lung was used for ribonucleic acid (RNA) isolation,
8 and the right lung was used for histology (personal communication, e-mail from E. Putnam
9 [University of Montana] to M. Gwinn [U.S. EPA] 02/26/09). Histology on mouse lungs from
10 each treatment group demonstrated an increase in fibrosis, as viewed by Gomori's trichrome
11 staining, following exposure to crocidolite and, to a lesser extent, Libby Amphibole asbestos.
12 Histologic tissue was also exposed to Lucifer Yellow stain to further analyze variability in
13 collagen following exposure. Lucifer Yellow staining revealed an increase in collagen following
14 exposure to both crocidolite and Libby Amphibole asbestos, but only crocidolite exposure led to
15 a statistically significant increase (p < 0.05). RNA was isolated from homogenized lungs and
16 purified for use in microarray analysis. Pooled RNA samples from mice in each exposure group
17 were analyzed on a OK-element mouse oligonucleotide array (MWG Biotech), and expression
18 was compared to a mouse reference standard RNA. Gene-expression results were analyzed by
19 GO Miner, and genes exhibiting at least 1.25-fold up- or down-regulation in treated lungs were
20 described. These included genes involved in membrane transport, signal transduction, epidermal
21 growth factor signaling, and calcium regulation for both crocidolite and Libby Amphibole
22 asbestos exposures, which support the increase in collagen observed above. Some limitations to
23 this study are the use of a standard reference for gene-expression comparisons (as opposed to the
24 saline controls), the practice of describing genes only if a greater than twofold difference in
25 expression is observed, and the use of pooled samples of homogenized whole lung that in some
26 cases could dilute variability between different areas of exposed lung (different lobes, fibrotic
27 versus nonfibrotic).
28
lrThe term "Libby Amphibole asbestos" is used in this document to identify the mixture of amphibole mineral fibers
of varying elemental composition (e.g., winchite, richterite, tremolite, etc.) that have been identified in the Rainy
Creek complex near Libby, MT. It is further described in Section 2.2.
This document is a draft for review purposes only and does not constitute Agency policy.
D-5 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
Table D-5. Fiber characteristics for intratracheal instillation studies in mice
Material
Libby Amphibole asbestos (Six Mix)
Crocidolite
Diameter
0.61 ± 1.22 um
0.16 ±0.09 urn
Length
7.21 ±7.01 um
4.59 ± 4.22 um
Aspect Ratio
22.52 ± 22.87
34.05 ±43.29
3
4
5
6
7
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
Source: Blake et al. (2008: 2007): Putnam et al. (2008): Smartt et al. (2010).
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 Libby
Amphibole asbestos and 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 Table D-5) 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 and
hydroxyproline assay (personal communication, e-mail from E. Putnam [University of Montana]
to M. Gwinn [U.S. EPA] 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 Libby Amphibole asbestos 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, Libby Amphibole asbestos exposure demonstrated minimal inflammation that 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, 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 altered varied.
Libby Amphibole asbestos exposure led to increased gene expression of Coll A2 at 1 week
This document is a draft for review purposes only and does not constitute Agency policy.
D-6 DRAFT—DO NOT CITE OR QUOTE
-------
1 postinstillation and Col3 Al at 1 month postexposure, while crocidolite led to no significant
2 alterations in the expression of these genes. Both crocidolite and Libby Amphibole asbestos
3 exposure led to increased CollAl gene expression as compared to saline control at 1 week and
4 1 month postexposure. Due to these differences in expression, the authors also examined the
5 collagen protein levels in the lungs to compare to the gene-expression changes. Total collagen
6 content was determined by measuring the hydroxyproline content in the caudal aspect of the left
7 lung. As compared to saline-exposed mice, a significant increase in hydroxyproline was
8 observed at 1 week and 1 month following exposure to both crocidolite and Libby Amphibole
9 asbestos; however, only lungs from crocidolite-exposed animals demonstrated a significant
10 increase at 3 months postexposure. These studies demonstrate that exposure to Libby
11 Amphibole asbestos lead to inflammation and fibrosis, although with differences in the time and
12 level of response from those of crocidolite.
13 Shannahan et al. (2011 a) exposed two rat models of human cardiovascular disease to
14 Libby Amphibole asbestos2 to determine if the preexisting cardiovascular disease in these
15 models would impact lung injury and inflammation following exposure. Healthy Wistar Kyoto
16 (WKY) rats were compared to spontaneously hypertensive (SH) and spontaneously hypertensive
17 heart failure (SHHF) rats following exposure. These rat models demonstrate pulmonary iron
18 homeostasis dysregulation (Shannahan et al., 2010). All rats (male only) were exposed to
19 0, 0.25, or 1.0 mg/rat via intratracheal instillation and were examined at 1 day, 1 week and
20 1 month postexposure. No changes were observed histopathologically, however, changes were
21 observed in markers of homeostasis, inflammation, and oxidative stress. Bronchoalveolar lavage
22 fluid (BALF) protein was significantly increased in both the SH and SHHF rat models as
23 compared to controls as early as 1 week postexposure. y-glutamyl transferase (GOT) activity was
24 increased in a concentration-dependent manner with exposure to Libby Amphibole asbestos at
25 the earliest time point measured (1 day), and was more pronounced in WKY rats as compared to
26 SH and SHHF rats. Lactate dehydrogenase (LDH) activity was also elevated in all strains but
27 was more pronounced in the SHHF rat model. Neutrophil increases were observed following
28 exposure in all strains, peaking at 1 day postexposure in all strains and persisting in the SH and
29 SHHF rats until 1 month postexposure. Macrophages showed similar results but persisted only
2Median fiber dimensions as determined by TEM: length = 3.59 urn; width = 0.23 um; aspect ratio >5:1.
This document is a draft for review purposes only and does not constitute Agency policy.
D-7 DRAFT—DO NOT CITE OR QUOTE
-------
1 in the SH rat model until 1 month postexposure. In order to determine any impact of exposure
2 on iron homeostasis, BALF ferritin and transferrin levels were measured in the lung. Increases
3 in ferritin and transferrin were observed in both SH and SHHF rats as compared to WKY
4 controls. Nonheme iron was also observed to be increased in only the SH rats at 1 day and
5 1 week postexposure. Markers of inflammation (macrophage inflammatory protein [MIP]-2) and
6 oxidative stress (heme oxygenase-1 [HO-1]) were elevated in both SH and SHHF as compared to
7 WKY rats at baseline, but limited exposure-related differences were observed. Limited changes
8 were also observed in ascorbate and glutathione levels in BALF and lung tissue. While
9 inflammation and cell injury were observed in all strains, no strain-related differences were
10 observed following exposure to Libby Amphibole asbestos (Shannahan et al., 201 la). In
11 conclusion, this study showed the potential for population variability related to cardiac disease in
12 response to exposure to Libby Amphibole asbestos, including markers of cellular injury, iron
13 homeostasis, and inflammation.
14 Shannahan et al. (20lib) tested the hypothesis that Libby Amphibole asbestos3 will bind
15 iron and increase the inflammogenic activity of fibers in vitro and acute lung injury and
16 inflammation in vivo. The authors examined the ability of Libby Amphibole asbestos to bind
17 exogeneous iron in an acellular system and evaluated iron-related alterations in the production of
18 reactive oxygen species (ROS). The authors also investigated the role of iron in the acute
19 inflammogenic response in vitro, using human bronchiolar epithelial cells, and in vivo using SH
20 rats by modulating fiber-associated iron concentrations. In a cell-free medium, Libby
21 Amphibole asbestos bound about 16 jig of iron/mg of fiber and increased ROS generation about
22 threefold. Generation of ROS was reduced by treatment with deferoxamine (DBF), an iron
23 chelator. To determine the role of iron in Libby Amphibole asbestos ROS generation and
24 inflammation, BEAS2B cells (bronchiolar epithelial cell line) were exposed to Libby Amphibole
25 asbestos (50 jig), iron-loaded Libby Amphibole asbestos, or Libby Amphibole asbestos treated
26 withDEF. No conditions altered HO-1 or ferritin mRNA expression. Libby Amphibole
27 asbestos by itself markedly increased IL-8 gene expression, which was significantly reduced by
28 iron loaded Libby Amphibole asbestos, but increased with Libby Amphibole asbestos treated
29 with DEF. To determine the role of iron in Libby Amphibole asbestos-induced lung injury in
30 vivo, spontaneously hypertensive rats were exposed intratracheally to either saline (300 jil), DEF
3Median fiber dimensions as determined by TEM: length = 3.59 urn; width = 0.23 um; aspect ratio >5:1.
This document is a draft for review purposes only and does not constitute Agency policy.
D-8 DRAFT—DO NOT CITE OR QUOTE
-------
1 (1 mg), ferric chloride (21 jig), Libby Amphibole asbestos (0.5 mg), iron loaded Libby
2 Amphibole asbestos (0.5 mg), or Libby amphibole asbestos plus DEF (0.5 mg). Neither ferric
3 chloride nor DEF increased bronchoalveolar lavage fluid (BALF) neutrophils compared to saline
4 at 24 hours after treatment. Libby Amphibole asbestos exposure led to a statistically significant
5 increase in BALF neutrophils (p < 0.05). Loading of iron on Libby Amphibole asbestos, but not
6 chelation, slightly decreased inflammation (Libby Amphibole asbestos + DEF > Libby
7 Amphibole asbestos > iron loaded Libby Amphibole asbestos). At 4 hours after exposure, Libby
8 Amphibole asbestos-exposed lung mRNA expression of MIP-2 was significantly reduced in rats
9 exposed to iron loaded Libby Amphibole asbestos, but increased by DEF (Libby Amphibole
10 asbestos + DEF > Libby Amphibole asbestos > iron loaded Libby Amphibole asbestos). Ferritin
11 mRNA expression was elevated in rats exposed to iron loaded Libby Amphibole asbestos
12 compared to the Libby Amphibole asbestos control. HO-1 expression was unchanged following
13 treatment with Libby Amphibole asbestos. The authors concluded that the acute inflammatory
14 response following exposure to Libby Amphibole asbestos might be modified by the fiber's
15 ability to complex iron, rather than redox cycling of fiber associated iron. The authors further
16 concluded that iron overload conditions may influence susceptibility to Libby Amphibole
17 asbestos-induced pulmonary disease.
18 Padilla-Carlin et al. (2011) investigated pulmonary and histopathological changes in a
19 male Fisher 344 rats following exposure to Libby Amphibole asbestos4 The rats were
20 administered a single dose of either saline, amosite (0.65 mg/rat), or Libby Amphibole asbestos
21 (0.65 or 6.5 mg/rat) by intratracheal instillation. At time from 1 day to 3 months after exposure,
22 bronchoalveolar lavage (BAL) was performed and the right and left lung was removed for
23 Rt-PCR and histopathological analysis, respectively. The results showed that amosite exposure
24 (0.65 mg/rat) resulted in a higher degree of pulmonary injury, inflammation, and fibrotic events
25 than the same mass dose of Libby Amphibole. Both amosite and Libby Amphibole resulted in
26 higher levels of cellular permeability and injury, inflammatory enzymes, and iron-binding
27 protein in both BAL fluid and lung tissue compared to saline controls. In addition
28 histopathological examination showed notable thickening of interstitial areas surrounding the
29 alveolar and terminal bronchioles in response to amosite and Libby Amphibole. However,
30 mRNA levels for some growth factors (e.g., PDGF-A and TGF-1P), which contribute to fibrosis,
4Median fiber dimensions as determined by TEM: length = 3.59 um; width = 0.23 um; aspect ratio >5.
This document is a draft for review purposes only and does not constitute Agency policy.
D-9 DRAFT—DO NOT CITE OR QUOTE
-------
1 were down regulated at several time points. The authors concluded that on a mass basis amosite
2 produced greater acute and persistent lung injury in this study.
3 In an early study, Sahu et al. (1975) described histological changes in the lungs of mice
4 exposed individually to amosite, anthophyllite, and tremolite. Fibers were described only as
5 <30-um long. Groups of 20 male albino Swiss mice were exposed to amosite, anthophyllite, and
6 tremolite at a single dose of 5 mg, and two animals from each group were sacrificed at 1, 2, 7,
7 15, 30, 60, 90, 120, and 150 days postexposure. Microscopic results following exposure to
8 tremolite showed acute inflammation of the lungs at 7 days postexposure, including macrophage
9 proliferation and phagocytosis similar to that observed with amosite and anthophyllite. Limited
10 progression of fibrotic response was observed at 60 and 90 days postexposure, with no further
11 progression of fibrotic response.
12 Blake et al. (2008) and Pfau et al. (2008) examined the role of asbestos in autoimmunity.
13 Blake et al. (2008) performed in vitro assays with Libby Amphibole asbestos, and both studies
14 performed the in vivo assays with tremolite. C57BL/6 mice were instilled intratracheally for a
15 total of two doses each of 60-ug saline and wollastonite or Korean tremolite sonicated in sterile
16 PBS, given 1 week apart in the first 2 weeks of a 7-month experiment. Detailed fiber
17 characteristics were described in Blake et al. (2007) for wollastonite and Libby Amphibole
18 asbestos, but not for Korean tremolite (see Table D-5; wollastonite and Korean tremolite not
19 shown).
20 Blake et al. (2008) described autoantibody production, monitored biweekly with blood
21 samples from saphenous vein bleeds and then by cardiac puncture following euthanization.
22 Specific autoantibodies were identified by immunoblotting with known nuclear antigens. These
23 autoantibodies were then incubated with murine macrophage cells previously exposed to Libby
24 Amphibole asbestos, wollastonite, or vehicle control (binding buffer containing 0.01 M Hepes,
25 0.14 M NaCl and 2.5 mM CaCb). Only sera from mice exposed to tremolite showed antibody
26 binding colocalized with SSA/Ro52 on the surface of apoptotic blebs (Blake et al., 2008).
27 In Pfau et al. (2008), collected serum samples, and urine were checked for protein
28 bi-weekly for 7 months. By 26 weeks, the tremolite-exposed animals had a significantly higher
29 frequency of positive antinuclear antibody tests compared to wollastinate and saline. Most of the
30 tests were positive for dsDNA and SSA/Ro52. Serum isotyping showed no major changes in
31 immunoglobulin subclasses (IgG, IgA, IgM), but serum IgG in tremolite-exposed mice decreased
This document is a draft for review purposes only and does not constitute Agency policy.
D-10 DRAFT—DO NOT CITE OR QUOTE
-------
1 overall. Further, IgG immune complex deposition in the kidneys increased, with abnormalities
2 suggestive of glomerulonephritis. No increased proteinuria was observed during the course of
3 the study. Local immunologic response was further studied on the cervical lymph nodes.
4 Although total cell numbers and lymph-node size were significantly increased following
5 exposure to tremolite, percentages of T- and B-cells did not significantly change. Because
6 tremolite is part of the makeup of Libby Amphibole asbestos (6%), using tremolite-exposed mice
7 might yield a similar response to Libby Amphibole asbestos-exposed mice. This same effect has
8 been demonstrated following exposure to ultraviolet radiation in skin cells, suggesting a similar
9 mechanism (Saegusa et al., 2002).
10
11 D.1.4. Injection/Implantation
12 LVG:LAK hamsters were intrapleurally injected with tremolite obtained from the Libby,
13 MT mine in an unpublished study by Smith (1978) prepared for W.R. Grace and Company.
14 These samples were identified as tremolite (22260p5; Sample 60) and 50% tremolite + 50%
15 vermiculite (22263p2, Sample 63). Both fiber samples were measured by optical phase
16 microscopy, and fibers were described as amorphous, irregularly shaped particles of about 5-15
17 um diameter, with Sample 60 (tremolite) also containing the occasional fiber up to 30 um long.
18 Fiber size for Sample 60 (tremolite) also was measured by scanning electron microscopy (SEM)
19 and was determined to have a geometric mean length of 2.07 um, a geometric mean diameter of
20 0.2 um, and an average aspect ratio of 10.36:1. Twenty-five milligrams of each of the two
21 samples were individually injected intraperitoneally into the pleural cavity of LVG:LAK
22 hamsters. Pathology was examined at approximately 3 months postexposure in 10 animals from
23 each group, with the remaining animals observed until death, or 600 days postexposure,
24 depending on the health of the animal. Average survivorships were 410, 445, and 421 days in
25 groups exposed to Sample 60, Sample 63, and saline, respectively (see Table D-6). Pleural
26 fibrosis was observed 3 months postexposure, and mesothelioma was observed in both treatment
27 groups between 350 and 600 days postexposure, with no mesotheliomas in control groups.
28
This document is a draft for review purposes only and does not constitute Agency policy.
D-11 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
Table D-6. Pleural adhesions and tumors following intraperitoneal injection
exposure in LVGrLAK hamsters (25 mg)
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
Endpoint
Average adhesion ratinga>b
Total tumors/animals0
Benign
Malignant
Mesothelioma
Control
0 (n = 10)
8/59
3/59
5/59
0/59
Sample 60 (tremolite)
3.3 («= 10)
8/58
2/58
6/58
5/58
Sample 63 (tremolite
and vermiculite)
3.6(»= 10)
16/61
5/61
9/61
5/61
aAs analyzed in first group sacrificed (between 41 and 92 days postexposure).
bRating for pleural 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).
The Smith et al. (1979) study 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-um
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 (1974). No further information
could be found on the characteristics of the samples used in this study.5 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, that very long fibers of the mineral talc, with narrow widths and fibrillar structure,
occur in this sample. 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
5This fiber is also analyzed in Wylie et al. (1993) and Stanton et al. (1981).
This document is a draft for review purposes only and does not constitute Agency policy.
D-12 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
prepared from a deposit of tremolitic talc from the western United States (Sample 31) and from a
specimen of asbestiform tremolite (Sample 72),6 respectively.
Each of the four samples was examined microscopically, although the data were not
reported in the paper by Smith et al. (1979). The average fibers in Sample 72 were long, thin,
crystalline fibers (>20-um long, 0.4-um 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-7. 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 (1974) or Smith et al. (1979; 1978). These data deficiencies limit the
interpretation of results from this study.
Table D-7. Fiber characteristics and numbers of resulting tumors following
intrapleural injection of 10- or 25-mg fiber samples into Syrian hamsters
Sample
FD-14
275
31
72
Average
length" (fim)
5.7
N/D
>20
>20
Average
diameter" (jim)
1.6
N/D
<0.4
<0.4
Tumors/survivors at 10 mgb
350
days
N/D
0/34
1/41
0/13
500
days
N/D
0/14
1/19
1/6
600
days
N/D
0/6
1/11
3/2
Tumors/survivors at 25 mgb
350
days
0/35
0/31
2/28
3/20
500
days
0/26
0/15
4/9
5/6
600
days
0/20
0/3
6/5
5/1
"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.
^Numerator = cumulative number of animals with tumors; denominator = number of survivors.
N/D = not described.
Source: Smith et al. (1979): Smith (1978): Smith (1974).
Although the source of this material is not reported, these studies parallel those in the unpublished studies
performed by Smith et al. (1979) for W.R. Grace that used material from Libby, MT. Whether Sample 72 is
material from Libby, MT, or another location is unknown.
This document is a draft for review purposes only and does not constitute Agency policy.
D-13 DRAFT—DO NOT CITE OR QUOTE
-------
1 Following analysis of Syrian hamsters intrapleurally injected with 10 or 25 mg of each of
2 the four samples of tremolite, Smith (1978) reported tumors at 350 days postexposure (25 mg) or
3 600 days postexposure (10 mg) for Samples 31 and 72 (see Table D-7). Although number of
4 animals was not provided by Smith et al. (1979), previous studies by these authors reported using
5 50 animals per exposure group (Smith, 1978; Smith and Hubert, 1974). The results in Table D-7
6 Present the cumulative number of tumors (numerator) at each time point analyzed over the
7 remaining survivors (denominator). The survival rate without tumor presentation was decreased
8 for animals exposed to Samples 72, 31, and 275. Smith et al. (1979) concluded that the FD-14
9 and 275 samples were noncarcinogenic, and Sample 31 was less carcinogenic than Sample 72.
10 Hamsters exposed to Sample 72 had extensive pleural fibrosis, which was observed to a lesser
11 degree in hamsters exposed to the other samples (Sample 72 > Sample 31 > Sample
12 275 = FD - 14). No statistical information was reported for these results, and because the
13 number of background tumors in control animals was not provided, no statistical analysis can be
14 performed.
15 Both studies demonstrate that intrapleural injections of Libby Amphibole asbestos7 leads
16 to an increase in pleural fibrosis and mesothelioma in hamsters compared to controls or animals
17 injected with less fibrous materials. The use of doses of equal mass for both studies makes it
18 difficult to compare potency between samples, as each sample could have vastly different fiber
19 number and total surface area. Although these studies clearly show the carcinogenic potential of
20 Libby Amphibole asbestos fibers, intrapleural injections bypass the clearance and dissolution of
21 fibers from the lung after inhalation exposures.
22 Stanton et al. (1981) also examined tremolite and describe a series of studies on various
23 forms of asbestos. Fibers, embedded in hardened gelatin, were placed against the lung pleura.
24 As an intrapleural exposure, results might not be comparable to inhalation exposures, as the
25 dynamics of fiber deposition and pulmonary clearance mechanisms are not accounted for in the
26 study design. Studies using two tremolite asbestos samples from the same lot were described as
27 being in the optimal size range for carcinogenesis; the fibers were distinctly smaller in diameter
28 than the tremolite fibers that Smith et al. (1979) used. These samples both had a high number of
29 fibers in the Stanton et al. (1981)— size range (>8-um long and <0.25-um diameter). Exposure
30 to both tremolite samples led to mesotheliomas in 21 and 22 of 28 rats exposed. The Stanton et
'Assuming Smith et al. (1979) used Libby Amphibole asbestos.
This document is a draft for review purposes only and does not constitute Agency policy.
D-14 DRAFT—DO NOT CITE OR QUOTE
-------
1 al. (1981) study also used talc that did not lead to mesothelioma production. This talc was found
2 to be the same as that used by Smith et al. (1979) and later by Wylie et al. (1993). Wylie et al.
3 (1993) stated that, although the two tremolites were consistent by size with commercial
4 amphibole asbestos, the talc used contained fibers that were much thinner and shorter, which is
5 not typical of prismatic tremolite fibers.
6 Wagner et al. (1982) examined three types of tremolite (California talc, Greenland, and
7 Korea) using SPF Sprague-Dawley (n = 48) and Wistar (n = 32) rats, then followed up with a
8 range of in vitro tests using the same fiber samples. Rats were injected intrapleurally
9 (20-mg tremolite) at 8-10 weeks of age and allowed to live out their lives. Median survival
10 times after injections were 644 days (California talc), 549 days (Greenland tremolite), and
11 557 days (Korean tremolite). Positive controls had a decreased survival time due to an infection,
12 which limits the interpretation of these data. Also, this study was performed separately using
13 different rat strains for the three tremolite samples. The authors state that, although the
14 decreased control survival time and use of different rat strains limit the usefulness of the study
15 for quantitative analysis, the results can be described qualitatively. Of the three tremolites, only
16 the Korean tremolite showed carcinogenic activity producing mesothelioma (14/47 rats, 30%).
17 Analysis of the fiber characteristics showed the Korean sample had fibers that were longer than 8
18 um and a diameter of less than 1.5 um. The California talc and Greenland tremolite had
19 little-to-no fibers in this size range (see Table D-8). Follow-up in vitro assays in the sample
20 publication (Wagner et al., 1982) confirmed the in vivo results, with the exposure to Korean
21 tremolite resulting in increased LDH and p-glucuronidase (BGL) release, cytotoxicity, and
22 giant-cell stimulation.
23 Davis et al. (1991) examined six tremolites with differing morphologies through
24 intraperitoneal injections with male SPF Wistar rats. Four of the tremolites were from
25 Jamestown, California; Korea; Wales; and Italy; and two were from Scotland. Of these, the three
26 from California, Korea, and Wales were asbestiform, and the other three were fiber bundles or
27 prismatic (see Table D-9). Rats were exposed (n = 33 or 36) with one intraperitoneal injection
28 with samples that were 10 mg/2 mL-sterile PBS. Animals were allowed to live out their full life
29 spans or until signs of debility or tumor formation developed. Although exposure was performed
30 based on sample weight, each sample was analyzed to determine the number of expected fibers
31 per milligram and, therefore, per exposure. These samples also were characterized further by
This document is a draft for review purposes only and does not constitute Agency policy.
D-15 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
Table D-8. 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
<6 um
<3 um
>8 um
Diameter
<0.8 um
<1.2 um
<1.5 um
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-um diameter
1.7
0
56.1
4
5
6
7
8
9
10
11
TEM = transmission electron microscopy.
Source: Wagner et al. (1982).
Table D-9. Fiber characteristics in a 10-mg dose (as numbers of fibers)
Sample
California
Wales
Korea
Italy
Carr Brae
Shininess
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 (x 10s)
13,430
2,104
7,791
1,293
899
383
No. of fibers
>8-um long,
<0.25-um
diameter3 (x 10s)
121
8
48
1
0
0
No. of particles
in 1-mg
injected dust
(x 10s)
18,375
4,292
13,435
20,137
9,490
5,901
Morphology
Asbestiform
Asbestiform
Asbestiform
Fiber bundles
Fiber bundles
Prismatic
12
13
14
15
16
17
18
19
20
21
22
23
24
"Stanton fibers.
Source: Davis et al. (1991).
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 diameter as measured by SEM (i.e., Stanton fibers).
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
This document is a draft for review purposes only and does not constitute Agency policy.
D-16 DRAFT—DO NOT CITE OR QUOTE
-------
1 fibers (see Table D-9). Although three samples were considered asbestiform (California,
2 Swansea, Korea), all samples had <1% of counted fibers defined as Stanton fibers. The highest
3 mesothelioma incidence was observed for the California sample, which contained the most
4 Stanton fibers (121 fibers per mg dust). The tremolite from Swansea, resulted in 97%
5 mesothelioma incidence yet contained only eight Stanton fibers per milligram (more than 90%
6 less than in the California sample). In contrast, the Italy tremolite, although containing only
7 0.08% Stanton fibers, resulted in 67% mesothelioma incidence. Little is known, however, about
8 the characteristics of particles or fibers <5-um long. This study highlights two issues associated
9 with all fiber studies: the limits of analytical techniques and the variability in response based on
10 the metric used to measure exposure. This study also supports the premise that asbestos samples
11 containing fibers that are not long and thin can be carcinogenic.
12 The Roller et al. (1996) study was designed to provide data on the dose response of
13 various fiber types in relation to their fiber dimensions (as measured by SEM). Fibers were
14 defined in this study as having an aspect ratio of >5:1 for all lengths and widths. Female Wistar
15 rats (n = 40) were given either one intraperitoneal injection of 3.3 mg or 15 mg of tremolite.
16 Rats were examined for tumors in the abdominal cavity following a lifetime (up to 30 months) of
17 observation. This paper described the fiber dimensions in depth (see Table D-10), while limited
18 discussion is focused on the exposure results. This table shows the characteristics of the fibers
19 sorted first by aspect ratio and diameter, and the fiber size distribution binned by the length and
20 diameter for those fibers with a length >5 jim. Results were described in this study in a table as
21 "positive rats" being those with histologically confirmed mesothelioma or macroscopically
22 supposed mesothelioma. No information was provided on how these determinations were made.
23 Exposure to 3.3-mg and 15-mg tremolite resulted in 9 mesotheliomas in 29 animals (64 weeks
24 postexposure) and 30 mesotheliomas in 37 animals (42 weeks postexposure), respectively. This
25 study demonstrates that intraperitoneal injection of tremolite led to mesothelioma in Wistar rats.
26 Analysis of other tissues was not described.
27
28
This document is a draft for review purposes only and does not constitute Agency policy.
D-17 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
Table D-10. Characteristics of tremolite fibers intraperitoneally injected into
Wistar rats
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
Fiber number per
ng dust and mass fraction (%)
Aspect Ratio (L/D) >5/l; D <2 um (Roller study)
Length:
>5
No.
17.4
um
Mass
32
>10
No.
6.9
jim
Mass
27
>20um
No.
1.9
Mass
18
Aspect Ratio
Diameter:
>5
No.
18.4
(L/D) <3/l; D <3 um (WHO, 1985)
um
Mass
43
>10
No.
7.0
jim
Mass
35
>20um
No.
2.0
Mass
26
Fiber-size distribution for aspect ratio (L/D) >3/l (all lengths, all diameters; SEM)
% Total
fibers
L>5 fim
22%
Length (um)
10% <
0.8
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 et al. (1996).
D.2. MECHANISTIC DATA AND OTHER STUDIES IN SUPPORT OF THE MODE OF
ACTION
D.2.1. In Vitro Studies—Libby Amphibole Asbestos
Hamilton et al. (2004) examined the potential for fibers, including Libby Amphibole
asbestos, to modify the function of antigen-presenting cells (APC). Analysis was performed at
24 hours with two forms of asbestos (crocidolite [25 or 50 |ig/mL] and Libby Amphibole
asbestos obtained from Site No. 30, Libby, MT [25 or 50 |ig/mL]) and ultrafme particulate
matter (PM2.5 [particulate matter 2.5 microns 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 Libby Amphibole
asbestos (25 or 50 |ig/mL), crocidolite (25 or 50 |ig/mL), or ultrafme 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 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
This document is a draft for review purposes only and does not constitute Agency policy.
D-18 DRAFT—DO NOT CITE OR QUOTE
-------
1 pretreatment with either asbestos or PM2.5 significantly upregulated both THI and T^ cytokines
2 (interferon gamma [IFNy]; interleukin-4 [IL-4]; and interleukin-13 [IL-13]) (p < 0.05).
3 Therefore, pre-exposure to either fibers or particles increased APC function, as reflected in
4 increased cytokine release after tetanus challenge. No significant differences, however, were
5 discernable between asbestos and PM2.s pretreatment. The authors speculated that the variability
6 in response between samples assayed—presumably due to the use of primary cells—obscures
7 statistical significance. Although this study supports a role for fibers and PM2.5 in potentiating
8 immune response, the implications of these findings to human health are unclear because many
9 agents can activate macrophages prior to antigen challenge.
10 Recent studies (Blake et al., 2008; Blake et al., 2007) compared the response of murine
11 macrophages (primary and cell line RAW264.7) to Libby Amphibole asbestos fibers and
12 crocidolite asbestos fibers. The Libby Amphibole asbestos fibers (7.21 ± 7.01-um long, 0.61 ±
13 1.22-um diameter) used in these studies were obtained from the U.S. Geological Survey and
14 were chemically representative of the Libby, MT mine (Meeker et al., 2003). The crocidolite
15 fibers (4.59 ± 4.22 um-long, 0.16 ± 0.09 um-diameter) used in these studies were provided by
16 Research Triangle Institute, NC, and the noncytotoxic control fiber (wollastonite,
17 4.46 ± 5.53 um-long, 0.75 ± 1.02 um-diameter) was provided by NYCO Minerals, NY. Cells
r\
18 were exposed for 24 hours to fiber samples measured by relative mass (5 ug/cm ), after which
19 the cells were analyzed by transmission electron microscopy (TEM) to measure internalization.
20 The results of the first study (Blake et al., 2007) indicate that Libby Amphibole asbestos fibers
21 can both attach to the plasma membrane and be internalized by macrophages, similar to the
22 crocidolite fibers. These internalized fibers were primarily less than 2-um long and were found
23 localized in the cytoplasm, in cytoplasmic vacuoles, and near the nucleus following 3-hour
9 O
24 exposure, 62.5 ug/cm . This same concentration (62.5 ug/cm ) was selected for the remaining
25 studies because cell viability was not decreased at this concentration for the Libby Amphibole
26 asbestos (92%); cell viability was decreased for crocidolite (62%), however, at this
27 concentration. As a result, the remaining assays would be expected to have decreased viability
28 following exposure to crocidolite, which may impact the levels of various responses. For
29 example, the reactive oxygen species (ROS) measurement would increase with increased cell
30 number; therefore, some of the quantitative results would be difficult to compare between fiber
31 types unless normalized to cell number.
This document is a draft for review purposes only and does not constitute Agency policy.
D-19 DRAFT—DO NOT CITE OR QUOTE
-------
1 Oxidative stress was measured by the induction of ROS and the reduction in glutathione
2 (GSH) levels. These two measurements generally complement each other, as GSH is used in
3 cells to maintain intracellular redox balance in cells in response to increased ROS levels. Both
4 Libby Amphibole asbestos and crocidolite fiber internalization generated a significant increase
5 (p< 0.05) in intracellular ROS as quantified by the oxidation of 2,7-dichlorodihydrofluorescein
6 to dichlorofluorescein with hourly readings on a fluorescent plate reader. Libby Amphibole
7 asbestos exposure significantly increased ROS in a dose-dependent manner (6.25, 32.5, and
r\
8 62.5 jig/cm ), as early as 1 hour postexposure at the highest dose (p < 0.05), as compared to a
9 no-treatment group. Only the highest concentration of crocidolite was tested. The lower
10 concentrations of Libby Amphibole asbestos were not compared to crocidolite and wollastonite,
11 but a comparison of the highest exposure concentrations (62.5 jig/cm2) of Libby Amphibole
12 asbestos, crocidolite, and wollastonite revealed greater ROS production following Libby
13 Amphibole asbestos exposure (1 hour,/? < 0.05). Blake et al. (2007) stated that similar results
14 were seen in the primary cell line but did not report the data. To differentiate the type of ROS
15 produced, dehydroergosterol fluorescence intensity levels were used, revealing that superoxide
16 anion was significantly increased following exposure to Libby Amphibole asbestos as compared
17 to controls. This observation was further confirmed with use of a free radical scavenger
18 (PEG-SOD [polyethylene glycol-superoxide dismutase]) specific to superoxide anion. This
19 coexposure of Libby Amphibole asbestos and PEG-SOD led to a significant decrease in ROS as
20 compared to cells exposed only to Libby Amphibole asbestos (p < 0.05). Total intracellular
21 superoxide dismutase (SOD) activity also was measured following exposure to Libby Amphibole
22 asbestos and showed a decrease in activity at 3 hours postexposure as compared to controls
23 (p < 0.05). Crocidolite appears to increase intracellular SOD activity at 24 hours postexposure.
24 These three assays demonstrate that Libby Amphibole asbestos exposure leads to increased
25 superoxide anion in macrophages, most likely by suppressing activity of intracellular SOD.
26 GSH levels were found to be decreased in response to Libby Amphibole asbestos and
27 crocidolite exposure in the macrophage cell line as compared to unexposed cells (p < 0.05). The
28 decreased GSH levels were more prominent following crocidolite exposure as compared to
29 Libby Amphibole asbestos. Crocidolite exposure has been shown in other studies to lead to
30 increased hydrogen peroxide but not superoxide anion (Kamp and Weitzman, 1999; Kamp et al.,
31 1992). The increased hydrogen peroxide from crocidolite exposure can then lead to increased
This document is a draft for review purposes only and does not constitute Agency policy.
D-20 DRAFT—DO NOT CITE OR QUOTE
-------
1 hydroxyl radical production (through interactions with endogenous iron), and potentially,
2 deoxyribonucleic acid (DNA) adduct formation. DNA adduct formation
3 (8-hydroxy-2'deoxyguanosine, 8-OHdG), 8-oxoguanine-DNA-glycosylase 1 (Oggl) levels, and
4 DNA damage (comet assay) also were measured. A significant increase in DNA damage in
5 exposed macrophages, as measured by increases in both 8-OHdG formation and expression of
6 Oggl, a DNA repair enzyme that excises 8-OHdG from DNA following oxidative stress, was
7 observed following exposure to crocidolite but not Libby Amphibole asbestos. Increased
8 superoxide anion following Libby Amphibole asbestos exposure does not appear to yield
9 oxidative damage similar to crocidolite. These results suggest a chemical-specific response to
10 each type of amphibole that yields varied cellular responses. Therefore, the mechanism of action
11 following response to Libby Amphibole asbestos might be different than that of crocidolite, also
12 an amphibole fiber.
13 To determine if the ROS production was related to fiber number for both Libby
14 Amphibole asbestos and crocidolite, cell-fiber interactions and fiber internalization were
15 measured following exposure to equal concentrations of crocidolite, Libby Amphibole asbestos,
r\
16 and wollastonite (62.5 ug/cm , 3 hours). With phase contrast light microscopy, the number of
17 cells interacting with one or more fibers were counted (100 cells counted for each treatment).
18 All murine macrophages bound or internalized at least one fiber from the Libby Amphibole
19 asbestos sample (mean ± SD, 4.38 ± 1.06 internalized) or the crocidolite sample (3.28 ±
20 1.58 internalized) but not the wollastonite sample (Blake et al., 2007). No significant differences
21 were observed in the responses to Libby Amphibole asbestos or crocidolite samples, suggesting
22 that the differences in measured ROS were not related to cell number. Fiber sizes varied
23 between the two samples, with the crocidolite sample containing a more homogeneous mixture
24 of long fibers (exact size not given), while the Libby Amphibole asbestos sample contained a
25 mixture of sizes and widths. These characteristics were not analyzed to determine what, if any,
26 role they might play in the varied response.
27 The second study by Blake et al. (2008) reports the effects of in vitro exposure to Libby
28 Amphibole asbestos on apoptosis by exploring autoimmune response following asbestos
29 exposure. Although Libby Amphibole asbestos was not directly used in the autoimmune studies,
30 the autoantibody (SSA/Ro52) is a known marker of apoptosis, and the in vitro studies included
31 treatment with Libby Amphibole asbestos. RAW264.7 cells exposed to Libby Amphibole
This document is a draft for review purposes only and does not constitute Agency policy.
D-21 DRAFT—DO NOT CITE OR QUOTE
-------
1 asbestos induced apoptosis over 72 hours, as measured by induction of poly (ADP-ribose)
2 polymerase cleavage and increased Annexin V staining. Redistribution of SSA/Ro52 in
3 apoptotic blebs was demonstrated in Libby Amphibole asbestos-exposed RAW264.7 cells but
4 not in the unexposed controls and wollastonite-exposed RAW264.7 murine macrophages, further
5 confirming apoptosis.
6 The role of ROS in chromosomal damage from asbestos was examined in a recent study
7 of Libby Amphibole asbestos and Union for International Cancer Control (UICC) crocidolite in
8 XRCC1-deficient human lung epithelial H460 cells (Pietruskaetal.. 2010). XRCC1 is involved
9 in the repair mechanisms for oxidative DNA damage, particularly single-strand breaks. This
10 study examined the effect of XRCC1 deficiency (induced in cells by shRNA knockdown)
11 following exposure to genotoxic (crocidolite and Libby Amphibole asbestos) and nongenotoxic
12 compounds (wollastonite, titanium dioxide) on micronucleus formation. Cells were exposed to
13 chemicals with known oxidants hydrogen peroxide (0-60 jiM) or bleomycin (0-10 |ig/ml) for 1
14 and 3 hours, or the nonoxidant paclitaxel (0-5 nM, 24 hours) to confirm the clonogenic survival
15 of the knockout cells, and as positive and negative controls. Fiber-size distribution for
16 crocidolite and Libby Amphibole asbestos is shown in Table D-l 1. Micronuclei induction was
r\
17 measured following treatment of cells by controls as described above, and by 5-|ig/cm fibers or
18 TiC>2 particles for 24 hours. Following treatment, cells were fixed, permeabilized, and blocked
19 before being exposed to anticentromere antibodies, and micronuclei were counted and scored as
20 centromere negative arising from DNA breaks (clastogenic) or centromere positive arising from
21 chromosomal loss (aneugenic). Spontaneous micronuclei induction was increased in
22 XRCC1-deficient cells as compared to control. Wollastonite and titanium dioxide did not induce
23 micronuclei in either cell type. Crocidolite and Libby Amphibole asbestos induced
24 dose-dependent increases in micronuclei formation in both cell types including an increase in the
25 proportion of micronuclei in XRCC1-deficient cells (see Table D-12). Libby Amphibole
26 asbestos exposure led to a decreased amount of micronuclei as compared to crocidolite.
27 Specifically in relation to clastogenic versus aneugenic micronuclei, crocidolite exposure led to
28 mainly clastogenic micronuclei while Libby Amphibole asbestos exposure led to a mixture of
29 aneugenic and clastogenic micronuclei. Nuclear bud formation was also observed but only with
This document is a draft for review purposes only and does not constitute Agency policy.
D-22 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
Table D-ll. Size distribution of UICC crocidolite and Libby Amphibole
asbestos used in Pietruska et al. (2010)a
4
5
6
7
8
9
10
11
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
Libby Amphibole Asbestos
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)
Table D-12. Percent clastogenic micronuclei following exposure to Libby
Amphibole asbestos or crocidolite
Libby Amphibole asbestos
(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
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
ap < 0.05 as compared to control cells.
Source: Pietruska et al. (2010).
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 Libby Amphibole
asbestos are most likely related to their physicochemical differences, particularly related to their
iron content. However, these results support a genotoxic effect of exposure to both crocidolite
and Libby Amphibole asbestos.
Mechanisms of oxidative stress following exposure to Libby Amphibole asbestos were
also studied in human mesothelial cells (Hillegass et al., 2010). Gene-expression changes were
measured with Affymetrix U133 A microarrays (analysis with GeneSifter) following exposure to
This document is a draft for review purposes only and does not constitute Agency policy.
D-23 DRAFT—DO NOT CITE OR QUOTE
-------
^99 &
1 15 x 10-|im/cm Libby Amphibole asbestos as compared to the nonpathogenic control
^99 ™^^^
2 (75 x 10 -|im /cm glass beads) in the human mesothelial cell line LP9/TERT-1 for 8 and
3 24 hours. Gene expression of only one gene (manganese superoxide dismutase [MnSOD;
4 SOD2]) was altered following exposure to Libby Amphibole asbestos for 8 hours, while
5 111 genes had an altered gene expression following exposure to Libby Amphibole asbestos for
6 24 hours (altered by at least twofold as compared to control).
7 The gene for MnSOD; SOD2 was observed to be significantly upregulated at both time
8 points (p < 0.05) as compared to nonpathogenic controls. This gene was confirmed in normal
9 human pleural mesothelial cells (HKNM-2) by quantitative RT-PCR at 24 hours following
10 exposure to the nontoxic dose of Libby Amphibole asbestos. Upregulation of three genes from
11 this and previous studies by these authors was confirmed by quantitative RT-PCR (SOD2, ATF,
12 and IL8) in HKNM-2 cells exposed to both Libby Amphibole and crocidolite asbestos. Gene
13 ontology of these results demonstrated alterations related to signal transduction, immune
14 response, apoptosis, cellular proliferation, extracellular matrix, cell adhesion and motility, and in
15 only one gene related to ROS processing. Follow-up studies at both the nontoxic dose
16 (15 x 106 |im2/cm2) and the toxic dose (75 x 106 |im2/cm2) exposure levels in LP9/TERT-1 cells
17 examined SOD protein and activity, ROS production, and glutathione (GSH) levels. At
18 24 hours, SOD2 protein levels were increased following exposure to the toxic dose of Libby
19 Amphibole asbestos (p < 0.05) but not at 8 hours. Cells exposed to all doses of Libby
20 Amphibole and crocidolite asbestos had increased copper-zinc superoxide dismutase
21 (Cu/ZnSOD; SOD1) protein at 24 hours (p < 0.05) but not at 8 hours. Although total SOD
22 activity remained unchanged, a dose-related SOD2 activity was observed following exposure to
23 both doses of Libby Amphibole asbestos for 24 hours, but this appeared to be minimal and was
24 not statistically significant (8 hours was not examined). Oxidative stress was measured by
25 dichlorodihydrofluorescein diacetate fluorescence staining detected by flow cytometry and was
26 observed as both dose- and time-dependent in cells exposed to Libby Amphibole asbestos but
27 was increased following exposure to the toxic dose of Libby Amphibole asbestos (statistical
28 analysis not possible). Oxidative stress was further supported by analysis of gene expression of
8Libby Amphibole asbestos samples for this study were characterized by analysis of chemical composition and mean
surface area (Meeker et al.. 2003). Doses were measured in surface area and described based on viability assays as
either nontoxic (15 * 106 um2/cm2) or toxic (75 * 106 um2/cm2).
This document is a draft for review purposes only and does not constitute Agency policy.
D-24 DRAFT—DO NOT CITE OR QUOTE
-------
1 heme oxygenase 1 (HO-1) following exposure to Libby Amphibole asbestos in both
2 LP9/TERT-1 and HKNM-2 cells for 8 and 24 hours. HO-1 was significantly increased
3 following exposure to the toxic dose of Libby Amphibole asbestos in both cell lines (p-value not
4 given). GSH levels were transiently depleted following 2-8 hours exposure to
5 75 x 106-|im2/cm2-levels of Libby Amphibole asbestos, with a gradual recovery up to 48 hours in
6 LP9/TERT-1 cells (HKNM-2 not analyzed). Exposure to crocidolite asbestos at the toxic dose
7 led to a significant GSH decrease at all times points up to 24 hours (p < 0.05). These studies
8 demonstrate that Libby Amphibole asbestos exposure leads to increases in oxidative stress as
9 measured by ROS production, gene expression, protein and functional changes in oxidative
10 stress proteins (SOD), and GSH-level alterations in human mesothelial cells.
11 The relative toxicity of Libby Amphibole asbestos was measured by gene-expression
12 changes of interleukin-8 (/£-#), cyclooxygenase-2 (COX-2\ heme oxygenase (HO)-l as well as
13 other stress-responsive genes as compared to amosite (Research Triangle Institute, NC) in
14 primary human airway epithelial cells (HAEC) in vitro. Comparisons were made with both
15 fractionated (aerodynamic diameter <2.5 jim) and unfractionated fiber samples (Duncan et al.,
16 2010). Crocidolite fibers (UICC) were also included in some portions of this study for
17 comparison. Fractionation was performed using the water elutriation method (Webber et al.,
18 2008) and characterized as described in Lowers and Bern (2009). Primary HAECs were
19 exposed to 0, 2.64, 13.2, and 26.4 jig/cm2 of crocidolite, amosite, AM2.5 (fractionated), Libby
20 Amphibole asbestos, or LA2.5 (fractionated) for 2 or 24 hours in cell culture. Confocal
21 microscopy was used to determine fiber content in cells exposed for 4 or 24 hours to
22 26.4-|ig/cm2 AM2.5 or LA2.5 only. At 4 hours post exposure, fibers were mainly localized on
23 the periphery of the cell with some fibers internalized. By 24 hours post exposure, most fibers
24 appeared to be internalized and localized by the nucleus. Cytotoxicity was determined by
r\
25 measurement of lactate dehydrogenase (LDH) from the maximum dose (26.4 jig/cm ) of both
26 amosite and Libby Amphibole asbestos samples, with less than 10% LDH present following
27 exposure to all four samples. Cytotoxicity was also determined for just the fractionated samples
28 of amosite and Libby Amphibole asbestos by measuring intracellular calcein fluorescence
29 emitted by live cells and showed 95% and 99% viability for AM2.5 and LA2.5, respectively.
30 These results support a limited Cytotoxicity of both amosite and Libby Amphibole asbestos under
31 these concentrations and time frames.
This document is a draft for review purposes only and does not constitute Agency policy.
D-25 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
Gene-expression changes in specific inflammatory markers (IL-8, COX-2, HO-1) were
analyzed by quantitative RT-PCR for amosite, AM2.5, Libby Amphibole asbestos, LA2.5, and
CRO at both 2 and 24 hours post exposure (all doses). Minimal increases in gene expression of
IL-8, COX-2, or HO-1 were observed at 2 hours post exposure to all five fiber types; at 24 hours
post exposure, 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-13). 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-|ig/cm2 amosite, Libby Amphibole
9
asbestos, AM2.5, or LA2.5 or to 26.4-|ig/cm LA2.5 only. The results show a pro-inflammatory
gene-expression response. Gene-expression profiles were similar between amosite and Libby
Amphibole asbestos, but differences were observed between AM2.5 and LA2.5.
Table D-13. Gene-expression changes following exposure to 26.4-ug/cm2
amphibole asbestos for 24 hours"
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
Libby
Amphibole
Asbestos
46 ±8.3
9.0 ±1.7
2.5 ±0.2
Libby
Amphibole
Asbestos,
fractionated
(LA2.5)
37 ±7.8
1.6 ±0.3
5.1 ±0.6
aAll results in fold change as compared to untreated control cells.
Source: Duncan et al. (2010).
This document is a draft for review purposes only and does not constitute Agency policy.
D-26 DRAFT—DO NOT CITE OR QUOTE
-------
1 To determine if surface iron on the fibers played a role in the inflammatory response,
2 Duncan et al. (2010) also examined surface iron concentrations by two methodologies:
3 inductively coupled plasma optical emission spectroscopy and citrate-bicarbonate-dithionite .
4 Both assays determined AM2.5 appeared to have the measured by thiobarbituric acid -reactive
5 product formation following exposure to amosite, AM2.5, Libby Amphibole asbestos, and
6 LA2.5. Both amosite samples were found to generate the greatest amount of hydroxyl radicals
7 compared to the two Libby Amphibole asbestos samples, with the fractionated AM2.5 and LA2.5
8 exhibiting small increases in ROS produced compared to the unfractionated samples.
9
10 D.2.2. In Vitro Studies—Tremolite
11 In general, all fibrous tremolite samples were shown to be carcinogenic, with those
12 containing more of the longer, thinner fibers (>10-um-length,
-------
1 samples was noted on a mass basis, data were not normalized for fiber content or size. The
2 inference is that differential results are due, at least in part, to differential fiber counts.
3 In a study to further elucidate the role of ROS following exposure to asbestos, Suzuki and
4 Hei (1996) examined the role of heme oxygenase (HO) in response to asbestos. HO is induced
5 in response to oxidative stress and functions to degrade heme; it might, therefore, prevent
6 iron-mediated hydroxyl radical production. All fibers tested led to an increase in HO, though
7 chrysotile (UICC) and crocidolite (UICC) led to a greater increase than tremolite (Metsovo,
8 Greece) and erionite (Rome, Oregon). No statistics, however, are described for these results.
9 This study focused on responses to 20 and 40 ug/mL of chrysotile and then used doses that
10 yielded 0.5 and 0.3 relative survival fractions for all other fibers (crocidolite, 20 and 40 ug/mL;
11 tremolite, 150 and 300 ug/mL; erionite, 200 and 400 ug/mL). Fibers were not characterized in
12 this paper. When normalized by survival fraction, the inductions of HO above control were
13 3.89-, 3.86-, 2.75-, and 2.78-fold above background for chrysotile, crocidolite, tremolite, and
14 erionite, respectively. Limited information is provided on the results of tremolite exposures
15 beyond an increase in HO following an 8-hour exposure. This increased HO following exposure
16 to tremolite demonstrates a response similar to that observed for crocidolite and chrysotile in this
17 study. Crocidolite is further analyzed, with exposures to the antioxidants, superoxide dismutase
18 and catalase, leading to a dose-dependent decrease in HO induction, which supports the role of
19 HO in oxidative stress.
20 Wylie et al. (1997) examined the mineralogical features associated with cytotoxic and
21 proliferative effects of asbestos in hamster tracheal epithelial (HTE) and rat pleural mesothelial
22 (RPM) cells with a colony-forming efficiency assay. HTE cells are used because they give rise
23 to tracheobronchial carcinoma, while RPM cells give rise to mesotheliomas. Cells were exposed
24 to fibers by weight, number, and surface area (see Table D-14).
25 Colony-forming efficiency assay results are expressed as the number of colonies in
26 exposed cultures divided by the control colonies multiplied by 100. Increases in colony numbers
27 indicate increased cell proliferation or survival in response to the exposure. Decreases in colony
28 numbers indicate toxicity or growth inhibition in response to the exposure. The results of the
29 analysis with fiber exposure by mass (ug/cm2) show elevated colonies in HTE cells following
30 exposures to both asbestos fibers (p < 0.05) at the lowest concentrations, while significant
This document is a draft for review purposes only and does not constitute Agency policy.
D-28 DRAFT—DO NOT CITE OR QUOTE
-------
4
5
6
7
decreases were observed for both asbestos fibers at the higher concentrations (0.5 ug/cm ,
p < 0.05) (Wvlieetal.. 1997).
Table D-14. 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
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
NIEHS = National Institute of Environmental Health Sciences.
Source: Wylie et al. (1997).
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. Comparing results of these samples when
exposure is 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,/? < 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.
Data analysis by surface area of these samples is shown in Table D-14. 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
This document is a draft for review purposes only and does not constitute Agency policy.
D-29 DRAFT—DO NOT CITE OR QUOTE
-------
1 cytotoxicity and proliferation based on the exposure measurement demonstrated differences in
2 response depending solely on how the fibers were measured (e.g., by mass, number, or surface
3 area). These results show variability in interpreting the same assay based on the defined unit of
4 exposure. Most early studies used mass as the measurement for exposure, which can impact how
5 the results are interpreted. When possible, further analysis of fiber number and surface area
6 might help elucidate the role of these metrics, particularly for in vivo studies.
7
8 D.2.2.2. Genotoxicity
9 (Athanasiou et al., 1992) performed a series of experiments to measure genotoxicity
10 following exposure to tremolite, including the Ames mutagenicity assay, micronuclei induction,
11 chromosomal aberrations, and gap-junction intercellular communication. Although a useful test
12 system for mutagenicity screening for many agents, the Ames assay is not the most effective test
13 to detect mutations induced by mineral fibers. Mineral fibers can cause mutation through
14 generation of ROS or direct disruption of the spindle apparatus during chromatid segregation.
15 Fibers do not induce ROS in the Ames system; however, and the Salmonella typhimurium strains
16 do not endocytose the fibers. Only one study was found in the published literature that used the
17 Ames assay to measure mutagenicity of tremolite. Metsovo tremolite asbestos has been shown
18 to be the causative agent of endemic pleural calcification and an increased level of malignant
19 pleural mesothelioma (see Section 4.1). To measure the mutagenicity of Metsovo tremolite,
20 S. typhimurium strains (TA98, TA100, and TA102) were exposed to 0-500 ug/plate of asbestos
21 (Athanasiou et al., 1992). This assay demonstrated that, like most asbestos fiber types tested in
22 earlier studies, Metsovo tremolite did not yield a significant increase in revertants in the Ames
23 assay, including in the TA102 Salmonella strain, which is generally sensitive to oxidative
24 damage. Although these strains can detect ROS mutations, they would not be able to produce
25 ROS from fibers alone or through necessary signaling pathways, and they do not endocytose
26 fibers. Thus, negative results in the Ames assay do not inform the cytotoxicity of Metsovo
27 tremolite.
28 Furthermore, this study demonstrated the clastogenic effects of tremolite, including
29 chromosomal aberrations and micronuclei induction. Tremolite exposure (0-3.0 ug/cm2) in
30 Syrian hamster embryo (SFIE) cells resulted in a statistically significant increase in chromosomal
31 aberrations (p < 0.02) when all treatment groups were combined and then compared to controls;
This document is a draft for review purposes only and does not constitute Agency policy.
D-3 0 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
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-15).
Table D-15. Micronuclei induction (BPNi cells) and chromosomal
aberrations (SHE cells) following exposure to tremolite for 24 hours
Asbestos dose (ng/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)
3
4
12C
9a
13C
"Significantly different from control (p < 0.05).
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 if fiber interference of the spindle
apparatus could be observed, which would have supported these results.
To determine if tremolite has some tumor promoter characteristics, Athanasiou et al.
r\
(1992) further examined intercellular communication following exposure to 0-4.0-ug/cm
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 etal., 1982). No effect on
gap-junction intercellular communication following tremolite exposure was observed.
This document is a draft for review purposes only and does not constitute Agency policy.
D-31 DRAFT—DO NOT CITE OR QUOTE
-------
1 Okayasu et al. (1999) analyzed the mutagenicity of Metsovo tremolite, erionite, and the
2 man-made ceramic (RCF-1) fiber. Whether this tremolite is the same as that used in previous
3 studies from this group is unclear. Tremolite from Metsovo, Greece, used in this study was
4 characterized as 2.4 ± 3.1-|im long and 0.175 ± 0.13-|im diameter (arithmetic mean) with the
5 number of fibers per microgram of sample equal to 1.05 x 105. Human-hamster hybrid A(L)
6 cells contain a full set of hamster chromosomes and a single copy of human chromosome 11.
7 Mutagenesis of the CD59 locus on this chromosome is quantifiable by antibody
8 complement-mediated cytotoxicity assay. The authors state that this is a highly sensitive
9 mutagenicity assay, and previous studies have demonstrated mutagenicity of both crocidolite and
10 chrysotile (Hei et al., 1992). The cytotoxicity analysis for mutagenicity was performed by
11 exposing 1 x io5 A(L) cells to a range of concentrations of fibers as measured by weight
12 (0-400 ug/mL or 0-80 ug/cm2) for 24 hours at 37°C. CD59 mutant induction showed a
13 dose-dependent increase in mutation induction for erionite and tremolite, but RCF-1 did not.
14
15 D.3. SUMMARY
16 In vitro studies have been conducted with Libby Amphibole asbestos from the Zonolite
17 Mountain mine. These studies demonstrated an effect of Libby Amphibole asbestos on
18 inflammation and immune function (Duncan et al., 2010; Blake et al., 2008; Blake et al., 2007;
19 Hamilton et al., 2004), oxidative stress (Hillegass et al., 2010), and genotoxicity (Pietruska et al.,
20 2010). These results suggest that Libby Amphibole asbestos may act through similar
21 mechanisms as other forms of asbestos, but data gaps still remain to determine specific
22 mechanisms involved in Libby Amphibole asbestos-induced disease.
23 Studies that examined cellular response to tremolite also found that fiber characteristics
24 (length and width) play a role in determining ROS production, toxicity, and mutagenicity
25 (Okayasu et al., 1999; Wagner et al., 1982). As with the in vivo studies, the definition of fibers
26 and the methods of fiber measurement vary among studies.
27
This document is a draft for review purposes only and does not constitute Agency policy.
D-32 DRAFT—DO NOT CITE OR QUOTE
-------
D.4. 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 toxicological 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: Bolin. 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.
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. Ann N 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-Carrin. 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.
Hamilton. R: Holiaa 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.
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: MacPhersoa 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.
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 [Review]. Thorax 54: 638-
652. http://dx.doi.0rg/10.1136/thx.54.7.638.
This document is a draft for review purposes only and does not constitute Agency policy.
D-3 3 DRAFT—DO NOT CITE OR QUOTE
-------
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 and tremolite 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. American
Mineralogist 88: 1955-1969.
NTP. (National Toxicology Program). (1985). NTP toxicology and carcinogenesis studies of chrysotile asbestos (cas
no. 12001-29-5) inf344/nrats (feed studies) (Vol. 295). Research Triangle Park, NC.
http://www.ntis.gov/search/product.aspx? ABBR=PB86167103.
NTP. (National Toxicology Program). (1988). NTP technical report on the toxicology and carcinogenesis studies of
Crocidolite Asbestos (CAS No. 12001-28-4) InF344/NRats (Feed Studies) (Vol. 280). Research Triangle
Park, NC. 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) (Vol. 279). Research Triangle Park,
NC. http://ntp.niehs.nih.gov/ntp/htdocs/LT rpts/tr279.pdf.
NTP. (National Toxicology Program). (1990b). NTP technical report on the toxicology and carcinogenesis studies of
tremolite (CAS no. 14567-73-8) in F344/N rats (feed studies) (Vol. 277). Research Triangle Park, NC.
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: 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.
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.
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.
This document is a draft for review purposes only and does not constitute Agency policy.
D-34 DRAFT—DO NOT CITE OR QUOTE
-------
Sahu. AP: Dogra. RK: Shanker. R: Zaidi. SH. (1975). Fibrogenic response in murine lungs to asbestos.
Experimentelle Pathologie 11: 21-24.
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: 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: 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.
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. 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. 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. Br J Cancer 45: 352-360.
Webber. JS: Blake. DJ: Ward. TJ: Pfau. JC. (2008). Separation and characterization of respirable amphibole fibers
from Libby, Montana. Inhal Toxicol 20: 733-740. http://dx.doi.org/10.1080/08958370801932544.
Wylie. AG: Bailey. KF: Kelse. JW: Lee. PJ. (1993). The importance of width in asbestos fiber carcinogenicity and
its implications for public policy. 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: Hodkinson. D: Winters. R: Mossman. BT. (1997).
Mineralogical features associated with cytotoxic and proliferative effects of fibrous talc and asbestos on
rodent trachea! epithelial and pleura! mesothelial cells. Toxicol Appl Pharmacol 147: 143-150.
http://dx.doi.org/10.1006/taap.1997.8276.
This document is a draft for review purposes only and does not constitute Agency policy.
D-3 5 DRAFT—DO NOT CITE OR QUOTE
-------
1 APPENDIX E. EVALUATION OF EXPOSURE-RESPONSE DATA FOR
2 LOCALIZED PLEURAL THICKENING IN WORKERS FROM THE
3 MARYSVILLE, OH COHORT
4 E.I. STATISTICAL ANALYSIS OF THE 2004 DATA SET FOR WORKERS HIRED IN
5 1972 AND LATER
6 All analyses were performed using SAS® statistical software v. 9.1. Benchmark dose
7 lower bound 95% confidence intervals (BMCLs) were obtained by the profile likelihood method
8 as recommended by Crump and Howe (1985) using the NLMIXED procedure in SAS (Wheeler,
9 2005). As described in Section 5.2.1.4, the critical endpoint for RfC derivation is radiographic
10 evidence of localized pleural thickening (LPT; n = 12 cases), compared with the referent group
11 with no radiographic evidence of pleural abnormality (n = 106).
12
13 E.I.I. Investigation of Explanatory Variables
14 Dichotomous statistical models describing the probability of individual response as a
15 function of cumulative exposure (CE) as measured by cumulative human equivalent exposure for
16 continuous exposure (CHEEC) in units of fiber/cc-year were used for this analysis. In order to
17 investigate the key explanatory variables for analysis, a forward-selection process was used to
18 evaluate the association of each of the potential covariates with odds of localized pleural
19 thickening, controlling for CHEEC. Covariates considered for inclusion in the model were time
20 since first exposure, age at X-ray, gender, smoking history, and body mass index (BMI). This
21 initial modeling was done using a standard logistic regression model as commonly applied in the
22 analysis of epidemiological data. The base model was a logistic regression model with CE
23 (natural log transformed) as the independent variable. This model provided an adequate fit to the
24 data (Hosmer-Lemeshow/>-value of 0.6357), and the exposure variable was statistically
25 significantly associated with the outcome (beta [standard error, SE] = 0.5676 [0.2420],
26 p-va\ue = 0.0190). Covariates were evaluated according to whether inclusion of the covariate
27 improved model fit as assessed by the Akaike Information Criterion (AIC), and statistical
28 significance of the covariate. When controlling for CE, inclusion of each of the covariates with
29 the exception of smoking increased the AIC for the model (with the exception of BMI, due to
30 missing information for some individuals), and none were associated with odds of discrete
31 pleural thickening: time since first exposure—p = 0.8879; age at X-ray—p = 0.7735;
This document is a draft for review purposes only and does not constitute Agency policy.
E-1 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
gender—•/> = 0.7660; smoking—p = 0.1669; BMI—p = 0.4095. Therefore, only exposure (i.e.,
CHEEC) was included in further analyses (see Table E-l).
Table E-l. Evaluation of covariates for the 2004 data set for workers hired
in 1972 and later
Covariate
Base model (only
ln[CHEEC])
Time since first exposure
Age at X-ray
Gender
Smoking history
BMIa
Wald p-value for
beta coefficient
corresponding to
covariate
—
0.8879
0.7735
0.7660
0.1669
0.4095
Wald p-value for
beta coefficient
corresponding to
exposure
0.0190
0.0310
0.0186
0.0195
0.0231
0.0102
AIC
75.5
77.5
77.4
77.4
75.4
56.7
9
10
11
12
13
14
15
16
17
18
19
20
21
22
"Note that only 97 observations were used, due to missing values (AIC not comparable).
E.1.2. Investigation of Candidate Models
The candidate models were logistic (with CHEEC considered as continuous and
continuous with a natural logarithm transformation), probit (with CHEEC considered as
continuous and continuous with a natural logarithm transformation), 3-parameter log-logistic,
dichotomous Hill, and dichotomous Michaelis-Menten models. These are statistical models used
to evaluate dichotomous data and were considered appropriate given the supralinear nature of the
observed relationship between Libby Amphibole asbestos1 exposure and the prevalence of
localized pleural thickening (LPT); model forms are provided in Table E-2. For each of the
candidate models, exposure lags of 0, 5, 10, 15, and 20 years were investigated. Although zero
lag exposures are not likely to be biologically relevant (i.e., some lag is expected for
development of LPT), these models were included for completeness and for comparison of
lrThe term "Libby Amphibole asbestos" is used in this document to identify the mixture of amphibole mineral fibers
of varying elemental composition (e.g., winchite, richterite, tremolite, etc.) that have been identified in the Rainy
Creek complex near Libby, MT. It is further described in Section 2.2.
This document is a draft for review purposes only and does not constitute Agency policy.
E-2 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
Table E-2. Evaluation of different model forms for the 2004 data set for
workers hired in 1972 and later
Model
Logistic
CHEEC, lag 5
CHEEC, lag 10
CHEEC, lag 15
CHEEC, lag 20
Logistic
CHEEC, lag 5
CHEEC, lag 10
CHEEC, lag 15
CHEEC, lag 20
Probit model
CHEEC, lag 5
CHEEC, lag 10
CHEEC, lag 15
CHEEC, lag 20
Probit model
CHEEC, lag 5
CHEEC, lag 10
CHEEC, lag 15
CHEEC, lag 20
3 -parameter
log-logistic
CHEEC, lag 5
CHEEC, lag 10
CHEEC, lag 15
CHEEC, lag 20
Dichotomous Hillb
CHEEC, lag 5
CHEEC, lag 10
CHEEC, lag 15
CHEEC, lag 20
Michaeris-Menten0
CHEEC, lag 5
Exposure
Metric
CHEEC
In(CHEEC)
CHEEC
In(CHEEC)
In(CHEEC)
In(CHEEC)
In(CHEEC)
Form"
P(LPT) =!-[!+ exp(-a - b x
CHEEC)]
P(LPT) =!-[!+ exp(-a - b x
In(CHEEC))]
P(LPT) =
-------
Table E-2. Evaluation of different model forms for the 2004 data set for
workers hired in 1972 and later (continued)
Model
CHEEC, lag 10d
CHEEC, lag 15
CHEEC, lag 20
Exposure
Metric
Form*
AIC
74.0
74.3
76.0
Hosmer-Lemeshow
GOF^-value
0.4163
0.5664
0.5610
BMC
0.2642
0.2097
—
BMCL
0.1177
0.0898
—
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
abkg indicates background rate, fixed at 1%.
bFor statistical modeling, the equivalent model form was used: P(LPT) = bkg + (Plateau - bkg) + [1 + exp(-a - b
ln(CHEEC))].
Tor statistical modeling, the equivalent model form was used: P(LPT) = bkg + (Plateau - bkg) + [1 + exp(-a -
ln(CHEEC))].
""Parameter estimates for the best-fitting models are as follows:
intercept = -0.1801 (SE = 1.0178), plateau = 0.5577 (SE = 0.3568, pe = 0.1207).
— = no data.
relative model fits. Similarly, although we explored models with exposure lagged by 20 years,
there were cases of LPT in the full cohort with fewer than 20 years since first exposure;
therefore, using such a long lag (which necessitates the assumption that these are background
cases) was not judged to be appropriate, and the results are not further considered.
The various model forms were compared using AIC, and general model fit was evaluated
r\
with the Hosmer-Lemeshow (2000) test (a form of the Pearson % goodness-of-fit [GOF]
statistic). This is a goodness-of-fit test that compares observed and expected events.
Observations are sorted in increasing order of estimated probability of the event occurring and
then divided into -10 groups; the test statistic is calculated as the Pearson %2 statistic of observed
and expected frequencies in these groups. The benchmark concentration (BMC) was estimated
for each candidate model using a Benchmark Response (BMR) of 10% and assuming a
background rate of 1% (see Section 5.2.3.3). BMCs and corresponding BMCLs were estimated
for each of the candidate models.
All of the candidate models had adequate fit as assessed by the Hosmer-Lemeshow test.
Models were compared using the AIC values, ranging from 74.0 to 77.8. The model with the
lowest AIC was the Michaelis-Menten model with 10-year lagged exposure (AIC = 74.0) (see
Table E-2). Note that models with exposure lagged by 0 or by 20 years, which are considered
not to be biologically relevant, are shaded grey and not included as candidate models.
This document is a draft for review purposes only and does not constitute Agency policy.
E-4 DRAFT—DO NOT CITE OR QUOTE
-------
1 There were several models that had similar model fits (within two AIC units) as the
2 best-fitting model, including the logistic and probit models with the natural log of CHEEC as the
3 exposure metric (lags of 5, 10, and 15 years), the 3-parameter log-logistic model (lags of 5, 10,
4 and 15 years), the dichotomous Hill model (lag of 10 years), and the Michaelis-Menten model
5 with exposure lagged by 5 or 15 years. All but one of these models would yield a BMCL lower
6 than that for the best-fitting model. However, the range was relatively narrow among these
7 similarly fitting models (BMCLs ranging from 0.0441 to 0.1352), with the lowest BMCL
8 -2.7 times lower than the BMCL for the Michaelis-Menten model with exposure lagged by
9 10 years.
10 The Michaelis-Menten model using the 10-year lagged exposure had ap-va\ue for fit
11 of 0.42, an AIC value of 74.0, and an estimated plateau of 0.5577 (SE = 0.3568). This model
12 yielded a BMC of 0.2642 fiber/cc-year, and corresponding BMCL of 0.1177 fiber/cc-year for a
13 10% increase in prevalence of LPT (see Table E-2 and Figure E-l).
14
15
...t
0.6 .
s i
3 ;
I °'5 '
IE i
| 0.4 i
£ i
o. I >.«* Estimated prevalence by
"S 0.3 i .*" cumulative exposure lagged by
« I .** 10 years
S | •
^ 0.2 i »V * Observed prevalence by
° I / quartiles of unlagged
a I • cumulative exposure among
cases
0
0 2 4 6 8 10
CHEEC (flbers/cc-year)
16
17 Figure E-l. Observed prevalence of localized pleural thickening and
18 estimated probability of localized pleural thickening.
19
20
This document is a draft for review purposes only and does not constitute Agency policy.
E-5 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
The potential confounding effect of covariates was re-examined in the best-fitting model
(see Table E-3). As in the initial assessment, after controlling for the effect of exposure (i.e.,
CHEEC, lagged by 10 years) there was no association between risk of LPT and time since first
exposure (p = 0.9973), age at X-ray (p = 0.8734), gender (p = 0.5544), or BMI (p = 0.3806), and
inclusion of each of these covariates increased the AIC (with the exception of BMI, due to
missing information for some individuals). The variable representing smoking history did not
meet the alpha = 0.05 criteria for statistical significance (p = 0.0841), although inclusion of this
variable decreased the AIC from 74.0 in the base model, to 72.3. Smoking was not considered
further in the derivation of the RfC due to the lack of statistical significance at the alpha = 0.05
level. However, because inclusion of the smoking variable did improve model fit, it is
investigated further as a sensitivity analysis in Section E.2.
Table E-3. Evaluation of covariates for the 2004 data set for workers hired
in 1972 and later in the best-fitting model
Covariate
Base model (only CHEEC)
Time since first exposure
Age at X-ray
Gender
Smoking history
BMI*
Wald/7-value for beta
coefficient corresponding
to covariate
—
0.9973
0.8734
0.5544
0.0841
0.3806
Plateau (SE)
0.5577 (0.3568)
0.5580 (0.3634)
0.5707 (0.3793)
0.6167 (0.4138)
0.5927 (0.3779)
0.4622 (0.2810)
AIC
74.0
76.0
76.0
75.7
72.3
55.8
17
18
19
20
21
22
23
24
25
*Note that only 97 observations used due to missing values (AIC not comparable).
To evaluate the assumption of a 1% background rate of LPT, the best-fitting model (i.e.,
Michaelis-Menten with 10-year lagged exposure) was rerun, allowing the background rate to be
estimated as a parameter rather than fixed at 1%. The resulting estimated background rate was
quite close to the assumed rate of 1%, at 3.12% (SE = 2.84%). Both the fixed and estimated
values are in the range of estimates from previous studies (see Section 5.2.3.3.), and the
This document is a draft for review purposes only and does not constitute Agency policy.
E-6 DRAFT—DO NOT CITE OR QUOTE
-------
1 difference in the BMCL when the background rate is fixed at 1% versus when it is estimated is
2 -15% (0.1177 compared to 0.1349 fiber/cc-year).
3
4 E.1.3. Derivation of the Candidate Point of Departure (POD) and Reference Concentration
5 (RfC) for Localized Pleural Thickening Using the Michaelis-Menten Model
6 The candidate point of departure (POD) is 0.1177 fiber/cc-year, the BMCLio for this data
7 set. The reference concentration (RfC) is derived from the POD using the duration of exposure
8 of 70 years, lagged by 10 years, and a total uncertainty factor of 100. See Section 5.2.4.
9 RfC = [0.1177 (fiber/cc) x (year)] x 1 -H (70 - 10) years x 1/100 = 1.96 xio"5 = 2 x 10"5
10 fibers/cc (rounded to 1 significant digit).
11
12 E.2. SENSITIVITY ANALYSIS FOR EFFECT OF SMOKING IN THE 2004 DATA SET
13 FOR WORKERS HIRED IN 1972 AND LATER
14 Due to the lack of statistical significance, smoking was not included in further analyses
15 for derivation of the RfC. However, based on the literature suggesting that smoking may play a
16 role in determining risk of LPT (see Section 5.3.6), the role of smoking was investigated further
17 for these sensitivity analyses.
18 The prevalence of any smoking history was 75.0% (n = 9) among cases, and 51.9%
19 (n = 55) among noncases. As noted above, the smoking variable was not significant at the
20 alpha = 0.05 level in the best-fitting (i.e., Michaelis-Menten) regression model controlling for
21 CHEEC lagged by 10 years (p = 0.08), but inclusion of the smoking variable did decrease the
22 AIC (AIC of 72.3 compared to 74.0 for the base model; see Table E-4). These results
23 (borderline statistical significance of the term but nontrivial improvement in model fit) may
24 indicate that smoking is associated with another variable that is associated with the outcome, or
25 that the variable is too poorly measured to accurately reflect the effect of smoking.
26 To evaluate whether smoking may modify the effect measure for the association between
27 Libby Amphibole asbestos exposure and risk of LPT, a third model was fit, which added an
28 interaction term between the exposure metric and smoking; in this model, neither the smoking
29 variable by itself nor the interaction term were significant (p = 0.2278 andp = 0.6598,
30 respectively), and the AIC increased from the base model (i.e., AIC of 74.1). Therefore, only
31 smoking (no interaction term) was retained for further sensitivity analyses.
32
This document is a draft for review purposes only and does not constitute Agency policy.
E-7 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
Table E-4. Evaluation of smoking in the best-fitting model
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
Model3
1
2
o
6
AIC
74.0
72.3
74.1
Variable
(None)
Smoke
Smoke
Ln(CHEEC, lag 10)*Smoke
Beta
—
1.8232
2.5401
0.2182
/7-value
—
0.0841
0.2278
0.6598
aThe following model forms were used for statistical analysis:
(1) P(LPT) = bkg + (Plateau - bkg) / [1 + exp(-a - ln(CHEEC, lag 10))]
(2) P(LPT) = bkg + (Plateau - bkg) / [1 + exp(-a - ln(CHEEC, lag 10) - beta*Smoke)]
(3) P(LPT) = bkg + (Plateau - bkg) / [1 + exp(-a - ln(CHEEC, lag 10) - beta*Smoke - beta2*ln(CHEEC, lag
10)*Smoke)]
The preferred model for RfC derivation (i.e., Model 1) yielded a BMC and a BMCL of
0.26 and 0.12 fiber/cc-year, respectively (see Table E-5). Model 2, which includes the smoking
variable, was used to derive estimates for smokers and nonsmokers separately. The BMC and
BMCL were derived by setting the beta coefficient for smoking to zero for nonsmokers, and to
the maximum likelihood (MLE)-estimated value (1.82) for smokers. The BMCL for nonsmokers
was about twice as high (0.25 fiber/cc-year) as that for the full cohort, while the POD for
smokers was about 1/3 that of the full cohort (0.04 fiber/cc-year).
Table E-5. Evaluation of smoking on estimated BMCs and BMCLs
Model
1
2
2
Group
All
Nonsmokers
Smokers
BMC (fiber/cc-year)
0.2642
0.9344
0.1509
BMCL
(fiber/cc-year)
0.1177
0.2463
0.0398
The lower BMCL among smokers compared to nonsmokers may indicate that smoking
increases risk for development of LPT among individuals exposed to Libby Amphibole asbestos;
another possibility is that smoking may affect the timing and progression of LPT development.
If LPT develops sooner among smokers compared to nonsmokers, this could lead to a higher
This document is a draft for review purposes only and does not constitute Agency policy.
E-8 DRAFT—DO NOT CITE OR QUOTE
-------
1 prevalence of LPT among smokers at a given observation time, and subsequently higher
2 estimated risk. The lack of detailed smoking information in this cohort (such as pack-years)
3 limits the ability to explore the effect of smoking on LPT risk among individuals exposed to
4 Libby Amphibole asbestos, but these sensitivity analyses indicate that smoking should be
5 considered when evaluating risk of respiratory health outcomes in this group.
6
7 E.3. STATISTICAL ANALYSIS OF THE FULL DATA SET
8 E.3.1. Identification of Key Explanatory Variables
9 In order to begin modeling the data, key explanatory variables were identified using
10 logistic regression to analyze the data of Rohs et al. (2008). Logistic regression was performed
11 using the R statistical software, version 2.11.1. All fitting was performed using individual data,
12 without any grouping. The dependent variable was localized pleural thickening (n = 59) noted
13 on chest X-rays of former workers in the Marysville, OH facility (n = 252) and no reported
14 history of exposure to commercial asbestos at other locations. The available potential
15 explanatory variables included CE at the time of X-ray, fiber/cc-year (equivalent to CHEEC used
16 in the University of Cincinnati report); time since first exposure (T; defined as time between first
17 exposure and date of X-ray in years); age at time of X-ray; gender; smoking status (i.e., ever,
18 never); and BMI. The BMI variable was missing for 34 individuals.
19 Initial analysis showed that CHEEC was a significant explanatory variable using both
20 CHEEC and In(CHEEC). The strategy used to determine what other explanatory variables were
21 influential consisted of including CHEEC and then adding one additional explanatory variable at
22 a time. Explanatory variables having/* >0.2 were dropped from further consideration.
23 Explanatory variables having/* <0.2 were given further consideration.
24 BMI was investigated as a potential explanatory variable because fat pads can sometimes
25 be misdiagnosed as pleural thickening. Thus, there might be a positive relation between BMI
26 and pleural thickening. Analysis of a model with CHEEC or In(CHEEC) plus BMI (n = 218)
27 showed that BMI was not a significant explanatory variable. Two subsequent models using BMI
28 cutoffs of 25 and 30 also showed that BMI was not a significant explanatory variable. Analysis
29 of a model with CHEEC or In(CHEEC) plus smoking indicated smoking was not a significant
30 explanatory variable.
This document is a draft for review purposes only and does not constitute Agency policy.
E-9 DRAFT—DO NOT CITE OR QUOTE
-------
1 Analysis of a model of CHEEC plus gender indicated gender was a potential contributing
2 explanatory variable (p = 0.18). However, it should be noted that the worker cohort was highly
3 imbalanced with 236 males and 16 females. Only three females have a cumulative human
4 equivalent exposure greater than 0.15 fiber/cc-year. These considerations indicated that the
5 potential relevance of gender as an explanatory variable should be viewed with caution.
6 Analysis of In(CHEEC) plus gender showed that gender was not a significant explanatory
7 variable. Accordingly, gender was eliminated as an explanatory variable.
8 The importance of T (time since first exposure) is clearly illustrated by comparing the
9 results of Lockey et al. (1984) with the results of Rohs et al. (2008). These two studies were
10 conducted in the same occupational cohort 24 years apart. In the initial study (Lockey et al.,
11 1984), only 2% of the individuals showed pleural changes; in the follow-up study (Rohs et al.,
12 2008), 28% of the individuals showed pleural changes. Logistic fitting of a model including
13 CHEEC or In(CHEEC) plus T showed that T was a highly significant explanatory variable with/?
14 <0.0005. This result is consistent with findings in other occupational cohorts exposed to various
15 forms of asbestos fibers that the time since first exposure is a significant explanatory variable,
16 even in the absence of continued exposure (Ehrlich et al., 1992; Jarvholm, 1992). Twas retained
17 as an explanatory variable. However, an important point of clarification is that the Invariable is
18 not the same as time of event. The LPT could have formed at any time before the X-ray was
19 taken (e.g., LPT detected in 2004 could have been present in 1990).
20 Analysis of a model of CHEEC plus age at X-ray indicated that age was a significant
21 explanatory variable with/? = 0.032. Analysis of a model of In(CHEEC) plus age at X-ray
22 showed that age at X-ray was a potentially significant explanatory variable with/? = 0.14. It
23 should be noted that this result does not mean that age is an independent risk factor for the
24 development of localized pleural thickening. In fact, there is no biological evidence that age is
25 an independent predictor of the development of localized pleural thickening without a history of
26 previous exposure to durable mineral fibers such as amphibole fibers. With a history of exposure
27 to amphibole fibers, age has been shown to be related to pleural thickening (Amandus et al.,
28 1987). However, it is quite possible that the association between age and prevalence is because
29 age at X-ray is related to rfrom first exposure, which is clearly one of the key explanatory
30 variables. Therefore, age at X-ray was not included as an explanatory variable.
31
This document is a draft for review purposes only and does not constitute Agency policy.
E-10 DRAFT—DO NOT CITE OR QUOTE
-------
1 E.3.2. Selection of Model Form
2 Figure E-2 (see Panel A) presents a plot of prevalence of LPT as a function of T,
3 stratified by CE (CHEEC). As seen, the prevalence appears to be low (i.e., close to zero) until
4 about 15-20 years after first exposure and then appears to rise in a nonlinear fashion. Figure E-2
5 (see Panel B) presents a plot of prevalence as a function of CE, stratified according to time since
6 first exposure. As seen, prevalence appears to rise rapidly with increasing CE but then tends to
7 flatten out (plateau). Based on these attributes of the base data set, the objective was to select a
8 model that included a plateau term whose value depended on T. Several alternative model forms
9 were investigated, using the dichotomous Hill model as the starting point:
10
11
12 XCHEEC) = bkg + (Plateau - bkg) - [ 1 + exp{-a - b x In(CHEEC)}]
13
14
15 In the dichotomous Hill model, the plateau term is a constant, with a value bounded
16 between background and 1.0. In order to be consistent with the data, this model was modified so
17 that the plateau term was a function of T. Several different nonlinear equations for the plateau
18 function were tested, including the following:
19
20
21 Plateau = MIN[1, bkg+ (1-bkg) x kl x 7]
22 Plateau = MIN[1, bkg + (1 - bkg) x kl x T2]
23 Plateau = MIN[1, bkg + (1 - bkg) x kl x f]
24 Plateau = bkg + (1-bkg) x
-------
Panel A. Prevalence vs. Time From First Exposure (Grouped by CHEEC)
I
100%
90% -
80% -
70% -
g 60% -
c
I 50% -
0>
£ 40% -
30% -
20% -
10% -
0% -
0
10 20 30
Time Since First Exposure (T. years)
CHEEC Bins (f/cc-yrs)
Bin No.
CHEEC 1
CHEEC 2
CHEEC 3
CHEEC 4
CHEEC 5
Min
0
0.05
0.1
0.2
1
Max
0.05
0.1
0.2
1
35
Mean
0.021
0.071
0.145
0.452
9.728
N
67
44
108
101
114
Cases
2
1
10
20
28
Prev
3.0%
2.3%
9.3%
19.8%
24.6%
Panel B. Prevalence vs. CHEEC (Grouped by Time From First Exposure)
§
I
a,
H
H £
O^
O
o
100%
T Bins (years)
Index
Tl
T2
T3
T4
T5
Min
0
10
20
30
40
Max
10
20
30
40
50
Mean
4.39
12.69
25.41
34.50
45.76
N
87
53
123
118
53
Cases
1
0
8
27
25
Prev
1.1%
0.0%
6.5%
22.9%
47.2%
Figure E-2. Raw data plots.
-------
1 AIC values when the plateau term is T, J2, T3, cumulative normal, cumulative gamma, or
2 cumulative Weibull are 293.97, 279.21, 276.12, 277.30, 277.07, and 276.98, respectively. The
3 plateau term based on T3 was not chosen because the curve reaches a plateau of 1 when Tis
4 about 50 years. Of those that have a plateau less than 1 at high T, the plateau term based on the
5 cumulative normal function was chosen because of its ease of use and familiarity.
6 Combining this equation for the plateau term with the basic probability model yields:
7
8
9 p (CHEEC,7) = bkg + (1 - bkg) x
-------
1 BMCT = exp [-a - \n{Q x ®(I]m,s) - 1}]
2
3 Where:
4 g = (l-bkg)-(BMR-bkg)
5
6
7 For a BMR of 10% extra risk, the value Q is 0.10.
8
9 E.3.4. Model-Fitting Results
10 Table E-6 provides the model-fitting results for each of the three data sets evaluated for
11 each of 5 lags of CHEEC and for each of 5 values of T. In all cases, the BMR is 10% extra risk.
12 Based on a background rate of 0.01, this BMR corresponds to a probability of LPT of 0.109.
13 Inspection of this table reveals that, for each of the three data sets evaluated, there is
14 relatively little effect of CHEEC lag over the interval 0-15 years. For the full data set and the
15 sub-cohort of workers hired in 1972 and later, the lowest AIC is achieved for a lag of 10 years.
16 It should be noted that the time since first exposure in the full cohort ranged up to 47.4 years;
17 therefore, estimates for values of T greater than 47.4 years represent extrapolation outside the
18 range of observed data, and should be interpreted with caution.
19 Figure E-3 presents a graph comparing the observed data to the predicted values from the
20 model (no lag) for the full data set. As above, this requires grouping the observed data into bins,
21 even though fitting was performed using the individual data. Because the choice of bins is
22 arbitrary, the appearance of the graphs would likely be changed somewhat if different bins were
23 chosen. Nevertheless, it seems apparent that the model predictions are in good accord with the
24 data.
25
26
This document is a draft for review purposes only and does not constitute Agency policy.
E-14 DRAFT—DO NOT CITE OR QUOTE
-------
Table E-6. Model-fitting results for the full data set
Study
1980+
2004
1980+
2004
1980+
2004
Year of Hire
All
>1972
<1972
N
434
198
236
Cases
61
13
48
CHEEC
Lag
0
5
10
15
20
0
5
10
15
20
0
5
10
15
20
MLE Parameters
m
42.38
42.44
42.58
42.86
43.28
31.41
31.58
3.5E+05
1.4E+06
2.1E+06
43.15
43.22
43.44
43.71
44.18
s
13.30
13.54
14.10
15.16
16.06
10.47
11.81
3.0E+06
5.4E+06
4.2E+06
13.33
13.46
13.88
14.83
15.84
a
1.977
2.000
2.061
2.167
2.395
-0.015
0.095
0.162
6.5E-01
1.5E+00
2.259
2.331
2.472
2.625
2.903
AIC
278.02
277.87
277.61
277.67
279.11
88.85
88.43
87.81
88.29
91.23
192.77
192.86
193.04
193.34
194.27
r=30
BMC
0.1822
0.1711
0.1477
0.1166
0.0876
0.2930
0.2623
0.2402
0.1766
0.1036
0.1689
0.1540
0.1270
0.0934
0.0642
BMCL
0.0709
0.0666
0.0580
0.0486
0.0349
0.1023
0.0956
0.0905
0.0643
0.0220
0.0227
0.0190
0.0092
-
--
r=35
BMC
0.0731
0.0707
0.0651
0.0567
0.0449
0.1900
0.1770
0.2402
0.1766
0.1036
0.0613
0.0569
0.0492
0.0406
0.0303
BMCL
0.0260
0.0253
0.0235
0.0219
0.0159
0.0399
0.0399
0.0432
0.0315
0.0059
0.0071
0.0059
0.0028
-
--
r=40
BMC
0.0421
0.0412
0.0389
0.0352
0.0286
0.1462
0.1374
0.2402
0.1766
0.1036
0.0341
0.0318
0.0279
0.0241
0.0185
BMCL
0.0138
0.0136
0.0129
0.0124
0.0091
0.0227
0.0232
0.0262
0.0185
0.0029
0.0037
0.0031
0.0015
-
--
r=so
BMC
0.0224
0.0221
0.0212
0.0197
0.0162
0.1177
0.1082
0.2402
0.1766
0.1036
0.0175
0.0164
0.0145
0.0128
0.0101
BMCL
0.0067
0.0066
0.0064
0.0062
0.0045
0.0136
0.0142
0.0162
0.0107
0.0013
0.0017
0.0014
0.0007
-
--
r=?o
BMC
0.0157
0.0154
0.0146
0.0133
0.0107
0.1128
0.1011
0.2402
0.1766
0.1036
0.0119
0.0111
0.0097
0.0084
0.0065
BMCL
0.0042
0.0042
0.0040
0.0038
0.0028
0.0109
0.0112
0.0123
0.0075
0.0007
0.0010
0.0008
0.0004
-
--
§
I
S
•3
o
§
i
a,
o!
o
a
O 3
0
H S.
w R-
o •
&
o
c
o
H
W
The BMC for any specified value of T is calculated from the model parameter estimates and the specified value of T as follows:
BMC = exp[-a - ln{(l - bkg) + (BMR - bkg) x ®(J]m,s) -I}].
The BMCL is estimated by rewriting the model so that BMC appears as an explicit term in the model, for a specified T of interest:
BMR = bkg + (1-bkg) x ®(T\m,s) + (1 + exp(-a - In(BMC)],
Solving for a yields: -a = ln[Q x (T\m,s)-l] + In(BMC),
Substituting yields: p(CREEC,T) = bkg + (1 - bkg)*O(71/w,s) + [1 + exp(z')],
Where: z' = ln(Q*d)(T\m,s) - 1) + In(BMC) - ln(CHEEC),
Simplifying yields: p(CHEEC,T) = bkg + (1 - bkg) x ®(T\m,s) + [1 + Q x d)(T\m,s) - 1) x BMC + CHEEC].
Using this equation, a trial value of the BMC is selected and treated as a constant, and the equation is refit to the data to find the MLE values of the remaining
parameters (m,s). After optimization, the value of the log-likelihood is recorded for the specified trial value of the BMC, and the process is repeated for other
trial values of the BMC. The BMCL is the trial value of the BMC where the log-likelihood decreases from the MLE log-likelihood value by an amount equal to
CHIDIST (2o, 1) + 2. For a = 0.05, the decrease is 1.3528.
-------
Panel A. Observed vs Predicted Prevalence as a Function of Time Since First Exposure (Grouped by CHEEC)
i nnn/
90% -
80% -
70% -
u 60% -
J3 50% -
> 40% -
£ 30% -
20% -
10% -
• CHEEC 1
•CHEEC 2
CHEEC 3
X CHEEC 4
X CHEEC 5
%
symbols = data /s/ X
lines = predicted JK./
V ^^^^ -~~~~~~~^
^^^ ^-^\
% & I IA. I » I • I •
0 10 20 30 40 50
Time From First Exposure (years)
CHEEC Bins (fiber/cc-
years)
Index
CHEEC 1
CHEEC 2
CHEEC 3
CHEEC 4
CHEEC 5
Min
0
0.05
0.1
0.2
1
Max
0.05
0.1
0.2
1
35
Mean
0.021
0.071
0.145
0.452
9.728
N
67
44
108
101
114
Cases
9
1
10
20
28
Prev
3.0%
2.3%
9.3%
19.8%
24.6%
Panel B. Observed vs Predicted Prevalence as Function of CHEEC (Grouped by Time Since First Exposure)
i nn% -,
90% -
80% -
70% -
u 60% "
| 50% -
1 40% -
|i 30% -
20% -
10% -
0% 1
c
1
2
• T2
T3
XT4
XT5
rx
L —
bH
symbols = data
lines = predicted
^
24681
CHEEC (fiber/cc-years)
Figure E-3. Observed versus predict
0
ed fo
TBins
Index
Tl
T2
T3
T4
T5
Min
(year)
0
10
20
30
40
Max
(year)
10
20
30
40
50
Mean
(year)
4.39
12.69
25.41
34.50
45.76
N
87
53
123
118
53
Cases
1
0
8
27
25
Prev
1.1%
0.0%
6.5%
22.9%
47.2%
4 E.3.5. Derivation of the POD and RfC for Localized Pleural Thickening Using the
5 Cumulative Normal Michaelis-Menten Model
6 For comparison with the primary analysis, a POD and RfC are derived for LPT from the
7 combined 1980 + 2004 data set as it provides the widest distribution of T-values (see
8 Section 5.2.3.2). A lag period of 5 years is used because Larson et al. (2010) showed that LPT
9 could be observed much earlier than previously thought.
10 Because the RfC is intended to provide protection for a lifetime of exposure (exposure
11 begins at birth and continues to age 70), the POD is the BMCLio with 7= 70 years of
12 0.0042 fiber/cc-year calculated with the cumulative normal Michaelis-Menten model (from
13 Table E-6). The POD is divided by duration of exposure of 70 years, lagged by 5 years, and then
This document is a draft for review purposes only and does not constitute Agency policy.
E-16 DRAFT—DO NOT CITE OR QUOTE
-------
1 divided by an uncertainty factor (see Section 5.2.4). In this case, as the model accounts for the
2 full lifetime of exposure of 70 years, the uncertainty factor of 100 is reduced to 30.
3 RfC = [0.0042 (fiber/cc) x (year)] x 1 -H (70 - 5) years x 1/30 = 2 x 10"6 fibers/cc
4 (rounded to one significant digit).
5 To provide a frame of reference, the calculation above was repeated with the data set
6 restricted to those hired in 1972 or later, when industrial hygiene data were collected in the
7 facility (from Table E-6).
8 RfC = [0.0112 (fiber/cc) x (year)] x i H- (70 - 5) years x 1/30 = 7 x io"6 fibers/cc
9 (rounded to one significant digit).
10 The reasonably good correlation in the calculated RfCs with the two different data sets
11 (2 x 10 6 versus 7 x 10 6 fiber/cc) provides some confidence in the exposure reconstruction
12 pre-1972.
13 An alternative candidate POD is the BMCLio with T= 40 years of 0.0136 fiber/cc-year
14 calculated with the Cumulative Normal Michaelis-Menten model (from Table E-6). The
15 BMCLio with T= 40 years is used because it is near the upper end of the range of T-values
16 available in the data set (rmax = 47.375 years). A lag time of 5 years and a total uncertainty
17 factor of 100 are used(see Section 5.2.5).
18 RfC = [0.0136 (fiber/cc) x (year)] x i H- (40 - 5) years x l -H 100 = 4 x io"6 fibers/cc
19 (rounded to one significant digit).
20
21 E.3.6. Sensitivity Analysis
22 The University of Cincinnati increased the exposure metric by a factor of 2 between 1972
23 and 1967 to account for conditions in the facility before engineering controls were added. For
24 the purposes of comparison, the CE was also calculated without this doubling. Plots of
25 prevalence of LPT with these two different exposure metrics are virtually identical (not shown).
26 One worker in the 1980 study was exposed only 5 months before X-ray and showed LPT.
27 Excluding this worker from the analysis did not change the calculated RfC.
28 Figure E-4 shows a plot of the PODs (fiber/cc-year) versus time since first exposure
29 (years) calculated from the Michaelis-Menten model using the 2004, data for workers hired in
30 1972 and later (see Section E.I), and from the Cumulative Normal Michaelis-Menten model
31 using the full data set (see Table E-6). Because the Michaelis-Menten model is independent of
This document is a draft for review purposes only and does not constitute Agency policy.
E-17 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
time since first exposure, the mean value of T for the data set is used. As there are few
individuals with long T (maximum of 47 years) and low CE, it is not clear whether the apparent
plateau with the Cumulative Normal Michaelis-Menten model is a reflection of the limitation of
the data or an expression of the underlying biology.
0.14
0.12
0.1
0.1177
•CumNorm MM Model All
Data
•Michaelis-Menten, 2004
Post 1972
20 40 60
Time From First Exposure (years)
7
8
9
10
11
Figure E-4. PODs (fiber/cc-year) versus time since first exposure (years).
This document is a draft for review purposes only and does not constitute Agency policy.
E-18 DRAFT—DO NOT CITE OR QUOTE
-------
1 E.4. REFERENCES
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
JIndMedl 1:27-37.
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.), Toxicological risk assessment:
Vol 1 Biological and statistical criteria (pp. 187-203). Boca Raton, FL: CRC Press, Inc.
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.
Hosmer. DW. Jr: Lameshow. S. (2000). Applied logistic regression. In. New York, NY: John Wiley &
Sons.
Jarvholm B. (1992). Pleural plaques and exposure to asbestos: a mathematical model. Int J Epidemiol 21:
1180-1184.
Larson. T: Meyer. C: Kapil V: Gurnev. J: Tarver. R: Black. C: Lockev. J. (2010). Workers withLibby
amphibole exposure: retrospective identification and progression of radiographic changes.
Radiology 255: 924-933. http://dx.doi.org/10.1148/radiol.10091447.
Lockev. JE: Brooks. SM: Jarabek. AM: Khoury. PR: McKay. 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.
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.
Wheeler. MW. (2005). Benchmark dose estimation using SAS. In Proceedings of the Thirtieth Annual SAS
Users Group International Conference. Gary, NC: SAS Institute.
http ://www2. sas .com/proceedings/sugi3 0/201-30 .pdf.
This document is a draft for review purposes only and does not constitute Agency policy.
E-19 DRAFT—DO NOT CITE OR QUOTE
-------
1 APPENDIX F. MARYSVILLE, OH WORKER OCCUPATIONAL EXPOSURE
2 RECONSTRUCTION
3 The Development of a Cumulative Human Equivalent Exposure Concentration
4
5
6
7
8 BY:
9 James E. Lockey, MD, MS
10 Carol Rice, PhD
11 EricBorton, BS
12 Timothy Hilbert, MS
13 Grace LeMasters, PhD
14
15
16
17
18
19
20 University of Cincinnati
21 Department of Environmental Health
22 3223 Eden Ave., ML 0056
23 Cincinnati, OH 45267
24
This document is a draft for review purposes only and does not constitute Agency policy.
F-1 DRAFT—DO NOT CITE OR QUOTE
-------
1 F.I. INTRODUCTION
2 This project builds on the previous work of Dr. James Lockey et al. investigating possible
3 effects of exposures to dust containing Libby Amphiboles at a plant in Marysville, OH (Rohs et
4 al., 2008; Lockey etal., 1984). The data used in the original exposure reconstruction and as
5 reported in the published manuscripts, was based on the exposures measurements available at
6 that time (Lockey et al., 1984). This exposure reconstruction is based on approximately five
7 times additional occupational fiber exposure data than was previously utilized in 1980. These
8 exposure measurements were recently obtained by the U.S. Environmental Protection Agency
9 (EPA) from the company and through trial transcripts from the United States of America vs.
10 W.R. Grace, et al., as well as the archived data used in the 1980 exposure reconstruction. Four
11 steps were undertaken to construct an exposure matrix describing exposure over each year from
12 1957 to 2000. In a final fifth step, this matrix was used to calculate an exposure metric for
13 workers.
14
15
16 1. Data searches, requests, and document selection
17 2. Document evaluation, data entry, cleaning, editing and standardization
18 3. Completeness and trends in measurements
19 4. Decisions relevant to the exposure matrix
20 5. Development of a cumulative human equivalent exposure concentration
21
22
23 F.2. DATA SEARCHES, REQUESTS, AND DOCUMENT SELECTION
24 Three sources of paper records were identified. First, sampling reports from OM Scott
25 that included measurements at the facility pre- and post-1980 were received via the EPA. These
26 reports contained both measurement results and information about the plant. OM Scott was also
27 contacted with a request for available maps of the plant layout prior to 1980. Secondly, archived
28 files from the Lockey et al. (1984) study were identified. Lastly, as a result of the recent W.R.
29 Grace trial, there was additional discovery of material relevant to the OM Scott plant. The
30 Department of Justice (DOJ) was contacted for the release of these data. There were seven
31 4" binders available for review and every page (approximately 3,150 pages) was scanned
This document is a draft for review purposes only and does not constitute Agency policy.
F-2 DRAFT—DO NOT CITE OR QUOTE
-------
1 visually to identify pages relevant to the current project. Aspects of particular interest included
2 the manufacturing process, usage and source of raw materials, engineering and design changes in
3 the plant, work practices and exposure assessment methodology. Approval was received from
4 the DOJ to utilize the relevant data for this project.
5
6 F.3. DOCUMENT EVALUATION, DATA ENTRY (QUALITATIVE AND
7 QUANTITATIVE), CLEANING, EDITING AND STANDARDIZATION
8 All of the records—both the qualitative and quantitative—were reviewed in this second
9 phase.
10
11 F.3.1. Qualitative Information
12 Written reports, letters, memos, and notes contained background information on plant
13 operations. A total of 1,489 pages were read for potentially useful and pertinent information
14 regarding OM Scott and abstracted into a data file. From these records, we obtained:
15
16
17 • Plant layout, including changes over time. This allowed us to associate the
18 descriptions used on air sampling data forms/reports with jobs or departments
19 within the plant. A limited number of aerial images were available to identify
20 major structures.
21 • Process descriptions were derived including workers per shift, workers per
22 department, sources of raw materials, and raw material volume in number of
23 railroad cars received, tonnage of railroad cars from Libby and South Carolina,
24 and tonnage of unexpanded vermiculite received.
25 • For each department a list of job titles and tasks.
26
27
28 Gaps in understanding were filled-in with information gathered from the focus groups,
29 specifically regarding:
30
31
32 • Plant lay-out and changes over time, including engineering controls,
33 • Historical pattern of job rotations within department from 1957 to 1980,
34 • Time spent in work locations at the plant site,
This document is a draft for review purposes only and does not constitute Agency policy.
F-3 DRAFT—DO NOT CITE OR QUOTE
-------
1 • Overtime associated with departments and season,
2 • Use/nonuse of respirators.
3
4
5 F.3.2. Quantitative Data
6 Air sampling reports include quantitative measurement of airborne dust and fiber
7 concentration associated with a department job. These records were computerized following the
8 data entry scheme provided on June 1, 2009 and approved. Records were double entered and
9 verified.
10 Two identical Microsoft Access databases were created for initial and duplicate entry of
11 the quantitative data. Each individual performing data entry had a unique and separate database
12 to avoid possible data entry confusion. Variables to be entered have been previously provided.
13 A random 10% check of entered data was conducted throughout the data entry process to
14 maintain quality of data, to address data entry questions and to resolve potential database issues.
15 Data entry differences were below 5% throughout the entry process.
16 Each record was assigned a document and record identification (ID) number. The
17 document ID variable was based on data source. For example, if the data were provided by the
18 EPA from OM Scott then the EPA document ID was used. Data hardcopies from the EPA,
19 Department of Justice and 1980 University of Cincinnati (UC) data were each numbered starting
20 from 1. The document ID variable states EPA, DOJ or UC followed by the document number.
21 Record IDs were generated by using a unique identifier like a sample number for each document.
22 If a unique identifier was unable to be discerned then the entry personnel was instructed to
23 consecutively number each sample per document starting from one.
24 A final verification of data entry used SAS version 9.2 PROC COMPARE to import the
25 initial and duplicate Access tables. Discrepancies were below 5% as a result of the 10% random
26 checks throughout the entry process. All discrepancies were addressed by reviewing the original
27 document. The initial and duplicate Access databases were archived. A copy of the initial
28 database was converted to Microsoft Excel format for ease of standardization and analyses.
29
30 F.3.3. Process of Standardization
31 The standardization process included categorizing entered data into appropriate variable
32 fields, spell checking, identifying duplicate record entry from duplicate documents, merging
This document is a draft for review purposes only and does not constitute Agency policy.
F-4 DRAFT—DO NOT CITE OR QUOTE
-------
1 records for the same sample or measurement, evaluating data for completeness, and categorizing
2 groups of data based on type of sample or measurement.
3 Data were reviewed and edited to ensure the information was entered into the appropriate
4 data field. A frequency of the data fields using SAS 9.2 PROC FREQ identified spelling
5 differences and patterns to ensure correct labeling of the data. Additional data variables were
6 created depending on recognized need to distinguish important pieces of data.
7 A new variable called group ID was created to identify, track, and consolidate partial
8 and/or complete duplicate data into one unique sample. Partial data were identified on a
9 combination of sample date, sample record ID, sample result, volume, sampling time and/or
10 document patterns. A document pattern would include instances where only a group of sample
11 results were available in one document and another document(s) would match the exact sequence
12 of sample results.
13 Data were further categorized based on the type of sample. Categories include dust
14 samples, bulk samples, personal and area fiber samples, limit of detection (LOD) or
15 quantification (LOQ) samples, off-site locations, and time weighted average samples. Some
16 samples were collected with a direct reading fibrous aerosol monitor, but these were not used as
17 there was no calibration information included in the records. Thus, only the fiber count data
18 collected with a sampling pump were used. In addition, group IDs lacking a sample result,
19 sample year or department were excluded.
20 Personal and area samples were plotted by year and department and found to be visually
21 similar. In addition the range, means, and standard deviations were approximately equal.
22 Therefore, personal and area sample data sets were merged and both utilized for the development
23 of the Exposure Matrix. Group IDs with only LOD or LOQ values were grouped by year and
24 categorized as trionize or background. In order to assign an estimate for the LOD or LOQ the
25 median value of each group was divided by two and assigned to all samples in that group. Given
26 the small number of LOD and LOQ samples (n = 35), it is unlikely any detectable bias was
27 introduced using this method. Time weighted average (TWA) values were not utilized when the
28 individual measurements that comprised the TWA were already available.
29 Sample analysis did not specify the type of fibers identified in the fiber counts. Counting
30 rules used included any fiber with the proper dimensions and not specifically Libby Amphibole
31 fibers. Attempts in other studies to convert from total dust to fiber count have relied on
This document is a draft for review purposes only and does not constitute Agency policy.
F-5 DRAFT—DO NOT CITE OR QUOTE
-------
1 similarities in equipment or process where side-by-side samples were collected. We did not
2 identify any 'pairs' of dust/fiber data from this plant. Moreover, fibers are a minor component of
3 the dust exposure, limiting an ability to find a relationship over time. Therefore, total dust
4 measurements were not converted to fiber counts and were not used as part of the fiber exposure
5 estimation.
6
7 F.3.4. Completeness and Trends in Measurements
8 From the paper records, we concluded that additional information would be helpful from
9 workers in order to obtain descriptions of work organization and practices. Focus groups
10 discussions were conducted with long-term OM Scott workers (n= 15) in 2010. These focus
11 groups provided valuable qualitative data in order to fill gaps regarding work plant operations,
12 especially during the earlier years.
13 As described earlier, the data used for exposure reconstruction was obtained from three
14 sources: UC archived records (reported previously by Lockey et al. (1984)), information
15 obtained by the EPA from the company, and from the DOJ documents. Table F-l shows that a
16 total of 914 IH fiber measurements were available for this analysis. Of this total, only 180
17 (19.6%) of the IH fiber measurements were available from the UC archived records. The yearly
18 number of samples collected was not uniform. As shown in Table F-2, the first fiber count
19 measurements were available in 1972 and the last in 1994. About 26% of the samples were
20 collected in 1978. Focus group participants reported working in the summer. Summer activities,
21 however, involved fewer work hours and included clean-up and repair activities in addition to
22 production. Since less than 6% of the fiber samples were collected during the summer months,
23 no seasonal trend analysis was possible.
24
25 F.4. DECISIONS RELEVANT TO THE EXPOSURE MATRIX
26 F.4.1. General Issues
27 A graphical display of fiber count results indicated that all samples in various trionizing
28 jobs generally followed the same pattern: higher in the early years of industrial hygiene
This document is a draft for review purposes only and does not constitute Agency policy.
F-6 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
Table F-l. Industrial hygiene fiber measurements by document source
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
Document source
DOJ
EPA
UC
COMBINED
Total (%)
Trionize
38
398
135
172
743(81.29)
Background
0
122
45
4
171 (18.71)
Total (%)
38(4.16)
520 (56.89)
180(19.69)
176(19.26)
914
sampling, and declining gradually over time. Further, from the focus groups, we learned that no
one, 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 gradual decline in levels of exposure rather than a dramatic step-wise drop
due to any one engineering change.
Changes in work practices such as the use of compressed air and brooms for clean-up
versus the use of wet vacuuming may result in marked decreases in exposure. We discussed
work practices in the focus groups, and no remarkable changes were documented. Participants
did note that during some years, sampling practices included leaving pumps in control rooms
during high-dust activities. High-dust activities included the use of compressed air to remove
particulate from surface areas. We did not find any documentation that high exposure work was
excluded from the sampling effort in the industrial hygiene reports. 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, no
adjustment was made for any potentially unsampled periods from 1972 through 1994 when
industrial hygiene measurements were available.
Per the focus groups, workers reported very sporadic usage of respirators due to heat and
discomfort. Because of the heat, the workers preferred paper masks, and reported reusing them
from day to day. There was no documentation of fit-testing of the paper masks. Paper masks
may 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.
This document is a draft for review purposes only and does not constitute Agency policy.
F-7 DRAFT—DO NOT CITE OR QUOTE
-------
Table F-2. Industrial hygiene fiber measurements by department and year
Dept.
Background
Trionize
Total
(Year %)
1972
o
3
9
12
(1.31)
1973
0
40
40
(4.38)
1975
2
20
22
(2.41)
1976
0
115
115
(12.58)
1977
10
68
78
(8.53)
1978
54
183
237
(25.93)
1979
2
26
28
(3.06)
1980
0
23
23
(2.52)
1981
12
38
50
(5.47)
1982
7
24
31
(3.39)
1983
3
8
11
(1.20)
1984
11
27
38
(4.16)
1985
5
14
19
(2.08)
1986
23
52
75
(8.21)
1987
13
33
46
(5.03)
1988
16
31
47
(5.14)
1993
0
3
3
(0.33)
1994
10
29
39
(4.27)
Total
(Dept.
%)
171
(18.71)
743
(81.29)
914
(100.00)
rs
s
§
I
TO
Dept. = department.
0
Grj
O
o I
2!
^
& Q
O '
c
o
-------
1 F.4.2. Vermiculite Raw Material Sources
2 Libby vermiculite usage ended in 1980 per shipping records obtained from B. Benson
3 and an Agency for Toxic Substances and Disease Registry (ATSDR) report (Benson, 2009;
4 ATSDR. 2005). Post 1980 usage included African/Virginia/South Carolina vermiculite until
5 2000. In 2000, corn cobs were introduced as an inert carrier of lawn care chemicals, and
6 vermiculite usage ended. There were two primary sources of information regarding vermiculite
7 sources:
8
9
10 • An internal UC document from the 1980 study with estimates of railroad car loads
11 delivered to the plant per year. Documents indicate railroad cars from Libby were
12 100 ton cars and from South Carolina 70 ton cars.
13 • The Chamberlain memo provides information regarding vermiculite sources for
14 1964-1972 in railroad car loads per year.
15
16
17 Per the UC document, 100% South Carolina vermiculite was estimated to be used from
18 1957-1960. Per the Chamberlain memo, Libby vermiculite began arriving in 1960. Focus
19 groups placed it earlier, in 1958 or 1959. We believe there is sufficient evidence to support a
20 1959 start date for Libby vermiculite with 1957 and 1958 assumed to be 100% South Carolina
21 vermiculite.
22 Documentation was found from the original 1980 UC documents indicating an estimated
23 Libby tonnage contribution of 32% from 1959-1963. These percentages for 1959-1963 were
24 adopted for use in this project. After adjusting for the difference in rail car sizes, the
25 Chamberlain memo indicates that Libby tonnage usage increased from 57% in 1964 to 73% in
26 1965 to 92% in 1966. Table F-3 illustrates the distribution of unexpanded vermiculite sources
27 received at the plant between 1957 and 1971. From 1959 until 1971 fiber level adjustments were
28 made based on the percent Libby versus South Carolina vermiculite tonnage received at the
29 plant. The estimates were derived from 1972 when the earliest industrial hygiene samples were
30 available and 93% of the vermiculite was Libby.
31 To develop the relationship of fiber levels between South Carolina and Libby
32 vermiculite, samples that recorded a 100% of either source for vermiculite were identified. Two
33 jobs with a higher number of samples from the same year from each source were used to
This document is a draft for review purposes only and does not constitute Agency policy.
F-9 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
Table F-3. Tonnage by year and vermiculite 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 says 1960
Chamberlain memo and 1980 chart
Chamberlain memo and 1980 chart
Chamberlain memo and 1980 chart
Chamberlain memo and 1980 chart
Chamberlain memo
Chamberlain memo
Chamberlain memo
Chamberlain memo
Chamberlain memo
Chamberlain memo
Chamberlain memo
Chamberlain memo
3
4
5
6
7
8
9
10
11
12
13
14
15
16
establish the relationship: track-unload for 1977 and expander for 1978. The samples used
included 22 Libby track-unload, 8 Libby expander, 17 South Carolina track-unload, and 7 South
Carolina expander. A weighted average of these samples generated a 10:1 fiber count ratio for
Libby: South Carolina vermiculite. This ratio was used for estimating the proportion of Libby
versus South Carolina fiber exposure levels from 1959 to 1971. From 1972 and beyond, IH
measurements were available and no adjustment in the IH data was made based on vermiculite
source. Tonnage records demonstrate that Libby was the primary source of vermiculite from
1972 until 1979, supplemented by African vermiculite, and that Libby vermiculite usage ended
in 1980.
The 100% Libby samples were compared to samples labeled as 50% Libby. The
resultant measurements were accordingly lower, demonstrating internal consistency within the
data.
This document is a draft for review purposes only and does not constitute Agency policy.
F-10 DRAFT—DO NOT CITE OR QUOTE
-------
1 Assessment of exposure in 1977 during application of the final, expanded product that
2 included a mix of South African and Libby vermiculite showed no fibers. Therefore, fiber
3 exposure estimation was restricted to jobs in the plant areas where expanding was conducted.
4
5 F.4.3. Exposure Estimates by Time Period for the Trionizing Department
6 For this project, exposures of interest were from 1957 through 2000. Exposure
7 measurements in the plant where vermiculite was used were initiated in 1972. For prior years, it
8 was necessary to estimate exposure from the measurements collected in 1972 and later and with
9 supporting qualitative information. Important changes occurred in production due to increasing
10 use of engineering controls to reduce airborne particulate. In addition, the source of vermiculite
11 changed over the years. Therefore, the exposure estimation process was divided into two efforts:
12 1972 and later when industrial hygiene measurements were available; and 1957 to 1971, when
13 no industrial hygiene measurements were available. The exposure estimation process is
14 described below, first for Trionizing where vermiculite was expanded and then for other
15 departments where either no or expanded vermiculite was used.
16
17 F.4.3.1. Trionizing Department Exposure Estimation >1972-2000
18 For the years with exposure measurements, fiber exposure level was estimated from the
19 measurement data. This was done by department.
20
21 F.4.3.1.1. Trionizing department
22 The trionizing department included jobs from the entry of vermiculite into the plant,
23 through final product. These were: track at raw material entry and production jobs of
24 screen/mill, dryer, expander, blender, resin, and clean-up, Workers rotated through the various
25 jobs within the department. Overall rotation among jobs reported in the 1980 Lockey et al. study
26 was verified by the focus groups.
27 Plots of the measurements over time were made for individual trionizing jobs. Based on
28 these plots, it was determined that all industrial hygiene sample results from the various
29 trionizing production jobs (screen/mill through clean-up) followed the same general distribution
30 and should be combined. The track job included two very different work activities: unloading
31 rail cars containing vermiculite (track unload) and general track work such as bringing in the rail
This document is a draft for review purposes only and does not constitute Agency policy.
F-l 1 DRAFT—DO NOT CITE OR QUOTE
-------
1 cars, and monitoring discharge (track other}. The two track job activities (unload and other} had
2 a substantially larger range of sampling results and were treated separately.
3 The following steps were followed:
4
5
6 1. The data were log-transformed.
7 2. For all exposure values for the combined trionizing jobs from 1972-1979, a curve
8 was drawn connecting the mean values of years having at least 40 exposure
9 measurements (1973, 1976, and 1978). This criteria was chosen to assure that
10 stable means were used to define the curve over this time period. For each year,
11 the annual exposure estimate was determined by exponentiation of the value from
12 the curve. The sharp decline seen in exposures throughout this time period
13 parallels the addition of engineering controls including dust collection, enclosing
14 vibrating conveyors, adding ventilators, erecting a wall between track and
15 trionizing, and sealing leaks in the system. As values for 1980-1994 were similar
16 and near the level of detection, the mean value for all the samples was used and
17 then extended until 2000.
18 3. The measurement results for track unload and track other were plotted and a
19 straight line produced to best fit the data points. An estimate of exposure at each
20 year was determined by exponentiation of the value on the line for that year.
21 4. For the trionizing department, it was estimated that 11% of work time was spent
22 in track and 89% in all other jobs. This is consistent with the previous weights
23 used in the 1980 Lockey study and confirmed by the focus group.
24 5. The Focus groups reported that when working track, track unload required about
25 25% of the time and track other comprised about 75% of the track job time.
26 Therefore, a weighted average for exposure at track within the trionizing
27 department was derived. This 25% time estimate for track unload is higher than
28 that previously published (Lockey et al., 1984).
29
30
31 Figure F-l illustrates on a log scale a fitted line of all usable industrial hygiene
32 measurements across all jobs (except track) within the trionizing department.
This document is a draft for review purposes only and does not constitute Agency policy.
F-l2 DRAFT—DO NOT CITE OR QUOTE
-------
^ Trionizing excluding Track
^^"Fitted Curve
4 YearlyGeometricMeans
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Log of fiber count (fibers/cc)
Figure F-l. Illustrates on a log scale a fitted line of all usable industrial
hygiene measurements across all jobs (except track) within the trionizing
department.
F.4.3.2. Trionizing Department Exposure Estimation 1957-1971
There are no industrial hygiene measurements available prior to 1972. Engineering
changes did not result in "step-function" decreases in exposures based on focus group reports.
Rather a more gradual decline in exposure occurred beginning with improvements in 1968, when
two dust collectors were added. Focus group workers report that dust exposures in trionizing
were at least two times higher in the 1960's. Track jobs, however, were outdoors and likely
unaffected by plant engineering controls. Hence, estimates for fiber exposure levels for track
duties were adjusted by type of vermiculite only.
For trionizing employees, excluding outdoor track duties, the estimate from the focus
group of "twice as high" was generated beginning from 1972 and increasing until 1967. The
year 1972 was used as the start of the "gradual" retrospective increase in exposure back to 1967
This document is a draft for review purposes only and does not constitute Agency policy.
F-13 DRAFT—DO NOT CITE OR QUOTE
-------
1 as 1972 was the first year when industrial hygiene measurements were available, and the percent
2 Libby vermiculite utilized was 93%. The year 1967 was selected as this was the year preceding
3 engineering controls. A line was drawn to connect these two points and then the adjustment was
4 made for the percent yearly Libby and South Carolina vermiculite utilized from 1967 through
5 1971. Prior to 1967, exposure was extended backward in time, assuming no change from the
6 1967 value except for a yearly adjustment for percent Libby and South Carolina usage. As
7 described above and shown in Figure F-l, after 1980 when Libby vermiculite was no longer used
8 and major environmental controls had been implemented, fiber exposure levels remained near
9 the level of detection (0.01) through the last available industrial hygiene information in 1994.
10 The levels were estimated to be the same from 1994 forward until 2000 when vermiculite was no
11 longer used.
12
13 F.4.4. Exposure Estimates for Nontrionizing Departments
14 Departments using only expanded vermiculite or no vermiculite were defined as having
15 "plant background" exposure. These included the departments of polyform, plant maintenance,
16 office, research, pilot plant, warehouse, central maintenance, and packaging. This decision was
17 based on plots of available sampling data showing similar levels, and qualitative reports
18 documenting that there were not fibers in the finished product. Plant background prior to 1972
19 was calculated using similar methodology as for trionizing. Although the background level was
20 not affected by engineering control as in trionizing, exposures would be affected by the percent
21 of Libby vermiculite used. Therefore, for the years prior to 1972, the measured plant
22 background rate in 1972 of 0.02 was adjusted by the yearly percent Libby vermiculite utilized.
23 The two years prior to Libby vermiculite usage, 1956 and 1957, were assigned level of
24 detection (0.01). This is in line with industrial hygiene measurements post Libby vermiculite
25 usage through 1994.
26 Polyform began in 1969, and no unexpanded vermiculite was used there. The
27 background exposure level was used for any time in Polyform.
28 Plant Maintenance—Although there were some differences of opinion in the focus group
29 regarding where plant maintenance spent their time, the consensus reached was to assign
30 approximately 50% of time in trionizing and 50% in areas defined as plant background for their
31 work in shop and other departments.
This document is a draft for review purposes only and does not constitute Agency policy.
F-l4 DRAFT—DO NOT CITE OR QUOTE
-------
1 • Office—Assigned plant background.
2 • Research—Assigned plant background.
3 • Pilot plant—Per the focus group participants, the pilot plant did not have its own
4 expander, and used only expanded vermiculite in test and run simulations. Plant
5 background levels were thus assigned to the pilot plant.
6 • Warehouse—Only expanded vermiculite was in this area. Although bags did
7 break, the exposure was to final product, not unexpanded vermiculite.
8 • Central Maintenance—According to the focus group, these employees worked
9 outside of trionizing for about 90% time (background) and 10% (trionizing) for
10 installation of new equipment/parts. Around 1982 central maintenance
11 department was discontinued, and the work was contracted to outside personnel.
12 • Packaging—Assigned plant background.
13
14
15 Table F-4 illustrates the fiber exposure matrix from 1957 to 2000 using this methodology.
16
17 F.4.5. Decisions Related to Break Periods and Hours Worked
18 Cumulative exposure is the product over time of the level of exposure and duration.
19 Level of exposure is derived from the exposure matrix and duration from the work history.
20 However, in this workforce, work time is complicated by breaks where exposure is at a lower
21 level and seasonal changes resulting in extra hours worked beyond the usual 40 hour week. Each
22 of these factors is described below:
23 According to the focus group data there was approximately a 30-minute break for lunch
24 and two 15-minute breaks during the day. Therefore, every worker was considered to have at
25 least one hour of background exposure daily. There was no documentation that a third ISminute
26 break was provided when working longer than 8 hours in a day.
27 Employees in some departments frequently worked extra hours each day, and weekends
28 as well, depending on the production needs and season. Decisions regarding this work
29 organization are summarized below:
30
31
This document is a draft for review purposes only and does not constitute Agency policy.
F-l 5 DRAFT—DO NOT CITE OR QUOTE
-------
Table F-4. Exposure matrix assuming doubling of fiber levels from 1972 to 1967 but with adjustment for
vermiculite source from 1957-1971
Department
Trionizing
Plant maint (50/50)
Central maint (90/10)
Background3
Trionizing
Plant maint (50/50)
Central maint (90/10)
Background3
1957
0.729
0.369
0.082
0.010
1972
3.674
1.847
0.385
0.020
1958
0.729
0.369
0.082
0.010
1973
3.007
1.513
0.319
0.020
1959
2.825
1.416
0.289
0.008
1974
2.464
1.242
0.264
0.020
1960
2.825
1.416
0.289
0.008
1975
2.019
1.020
0.220
0.020
1961
2.825
1.416
0.289
0.008
1976
1.391
0.705
0.157
0.020
1962
2.825
1.416
0.289
0.008
1977
0.150
0.090
0.030
0.020
1963
2.825
1.416
0.289
0.008
1978
0.086
0.053
0.027
0.020
1964
4.462
2.237
0.457
0.012
1979
0.077
0.044
0.017
0.010
1965
5.510
2.763
0.565
0.015
1980
0.063
0.036
0.015
0.010
1966
6.755
3.387
0.692
0.019
1981
0.063
0.036
0.015
0.010
1967
6.427
3.222
0.659
0.018
1982
0.060
0.035
0.015
0.010
1968
5.542
2.779
0.569
0.016
1983
0.060
0.035
0.015
0.010
1969
5.279
2.648
0.543
0.017
1984
0.055
0.032
0.010
1970
4.923
2.470
0.509
0.018
1985
0.055
0.032
0.010
1971
4.316
2.168
0.449
0.019
1986-2000
0.052
0.031
0.010
rs
S
I
I
s
o
a
"Background applies to Pilot Plant, Research, Polyform, Office, Packaging, Warehouse.
0 §
WX;
^
it
O '
-------
1 1. Extra hours—Were defined as hours worked in excess of 8 hours per day.
2 2. Four departments worked no extra hours—office, pilot plant, research, and central
3 maintenance. According to focus group data, the only departments that worked
4 extra hours outside of their own department were trionizing and polyform. Thus,
5 a decision was needed as to how to appropriate the amount of overtime spent
6 outside trionizing and polyform.
7 3. Extra hours for polyform workers—According to the focus groups, polyform
8 workers first worked in their own department, and went to trionizing to work
9 extra hours. According to workers, about 75% of the daily overtime was in their
10 own department. Therefore, for each 4 hours worked beyond the normal 8 hour
11 day, it is estimated that they spent 3 hours in polyform and 1 in trionizing. This
12 rule was not applied to 8-hour weekend days worked.
13 4. Extra hours for trionizing workers—As for polyform workers, above, it is
14 estimated that trionizing workers spent three hours in trionizing and one hour in
15 polyform as a daily average.
16
17
18 Schedules by season differed due to production rate:
19
20
21 • For trionizing, plant maintenance, polyform, warehouse, and packaging the spring
22 schedule was from January through May—7 days at 12 hours.
23 • For trionizing, plant maintenance, polyform, warehouse, and packaging the
24 summer schedule was from June through August—5 days at 8 hours. Due to the
25 difficulty that heat and humidity brought to the process, polyform was shut down
26 during summer. During the summer, polyform workers did outside jobs. As
27 these jobs have the same exposure level as polyform (background rate), no
28 adjustment was made for the summer polyform shutdowns. The trionizing
29 department more typically slowed down production in the summer, and this is
30 reflected in the number of hours worked from June through August.
31 • For trionizing, plant maintenance, polyform, warehouse, and packaging the fall
32 schedule was from September through December—5 days at 12 hours and
33 2 weekend days at 8 hours.
34
35
36 In light of these extra hours, exposure values by department and season were modified
37 for use in the cumulative equivalent human equivalent exposure concentration estimations.
38
This document is a draft for review purposes only and does not constitute Agency policy.
F-17 DRAFT—DO NOT CITE OR QUOTE
-------
1 F.5. DEVELOPMENT OF A CUMULATIVE HUMAN EQUIVALENT EXPOSURE
2 CONCENTRATION
3 An EPA adjustment of cumulative occupational exposure to fibers to continuous human
4 exposure to fibers (24 hours/day; 7 days/week) was provided by B. Benson. This adjustment
5 was accepted as provided for the development of a cumulative human equivalent exposure
6 concentration (CHEEC) for the Marysville, OH occupational cohort.
7
8 F.5.1. Seasonal Schedule Correction Factor
9 For this project the Correction Factor was adjusted for the specific information on work
10 schedules related to the seasonal changes to meet production demands as described above in
11 Section 4.4. UC applied these correction factors supplied by the EPA (B. Benson) to the work
12 history data obtained by UC during 1980 and updated in 2004.
13
14 F.5.2. Decision Rules to address Department Changes Occurring Within Seasons
15 Decision rules were implemented to systematically standardize each worker's
16 occupational history to a format that corresponded directly with the seasonal changes that
17 occurred at the plant. Previous decisions related to department exposure levels and seasonal
18 work resulted in six unique exposure categories: trionizing, plant maintenance, central
19 maintenance, polyform, background (office, research, pilot plant), and background with extra
20 time (warehouse, packaging). The date of any job change by a worker between these six
21 categories was adjusted so the change occurred at the starting month for the nearest season.
22
23 F.5.3. Development of CHEEC
24 In preparation for creating the CHEEC, the exposure matrix was converted to a seasonal
25 (spring, summer, fall) exposure value. This value is the estimate of the amount of exposure
26 occurring by department for each season of each year. With the worker's occupational histories
27 standardized to the same seasons, the CHEEC for each worker was then calculated as the sum of
28 exposure values for all seasons worked between 1957-2000. The correction factors used in
29 derivation of the CHEEC are outlined below.
30
31
32
This document is a draft for review purposes only and does not constitute Agency policy.
F-l 8 DRAFT—DO NOT CITE OR QUOTE
-------
1 General Procedure
2 • (Cumulative Fibers)occup x Correction Factor = (Cumulative Fibers)nEC-
3 • OCCUP = Occupational Exposure.
4 • HEC = Human Equivalent Concentration for exposure of 24 hours/day, 7
5 days/week.
6 • The Correction Factor usually used with an occupational study is
7 5 days-7 days x 10m3-20m3.
8
9
10 UC Procedure
11 CHEEC= (Exposure Est year-dept-season 1 x Correction Factor season 1
12 x Seasonal Duration Factor) + (Exposure Est year-dept-season 2
13 x Correction Factorseason 2 x Seasonal Duration Factor)
14 + ... (Exposure Est year-dept-season x x Correction Factor season
15 x Seasonal Duration Factor).
16
17 Where the Seasonal Duration Factor for the Spring is 5/12 year; the Summer is 3/12 year;
18 the Fall is 4/12 year.
19
20 F.5.3.1. Detailed Calculations Follow
21 F.5.3.1.1. Work schedule for trionizing, plant maintenance, polyform, warehouse, and
22 packaging
23 F.5.3.1.1.1. Spring
24 January 1 to May 31:7 days/week, 12 hours/day, with New Years' Day off, and
25 accounting for leap years:
26
27
28 • 151.25-1 = 150.25 days
29 • Breathing rate, working = 1.25 m3/hour x 12 hours = 15 m3
30 • Breathing rate, not working = 0.625 m3/hour x 12 hours = 7.5 m3
31 • Total breathing rate = 15 + 7.5 = 22.5 m3/day
32 • Correction Factor Spring = 150.25 + 151.25 x 15 4- 22.5 = 0.662259
33
This document is a draft for review purposes only and does not constitute Agency policy.
F-19 DRAFT—DO NOT CITE OR QUOTE
-------
1 F.5.3.1.1.2. Summer
2 June 1 to August 31:5 days/week, 8 hours/day, 2 week summer vacation:
3
4
5 • (92 - 14) x 5 -7 = 55.714286 days
6 • Breathing rate, working = 1.25 m3/hour x 8 hours = 10 m3
7 • Breathing rate, not working = 0.625 m3/hour x 16 hours = 10 m3
8 • Total breathing rate = 10 + 10 = 20 m3/day
9 • Correction Factor Summer = 55.714286 - 92 x 10 - 20 = 0.302795
10
11
12 F.5.3.1.1.3. Fall
13 September 1 to December 31:5 days/week, 12 hours/day and 2 days/week, 8 hours/day,
14 with Christmas Day off:
15
16
17 • 122-1 = 121 days
18 • Breathing rate, working, 12 hour day = 1.25 m3/hour x 12 hours = 15 m3
19 • Breathing rate, working, 8 hour day = 1.25 m3/hour x 8 hours = 10 m3
20 • Breathing rate, not working = 0.625 m3/hour x 16 hours = 10 m3
21 • Total breathing rate, 12 hour work day = 15 + 7.5 = 22.5 m3/day
22 • Total breathing rate, 8 hour work day = 10 + 10 = 20 m3/day
23 • Correction Factor Fall = 121 - 122 x (86.42857 x 15 - 22.5 + 34.57143 x
24 10-20)-121 = 0.613973
25
26
27 F.5.3.1.2. Work schedule for office, pilot plant, research, and central maintenance
28 No extra days or extra hours.
29
30 F.5.3.1.2.1. Spring
31 January 1 to May 31:5 days/week, 8 hours/day, with New Years' Day off, and
32 accounting for leap years.
33
34
This document is a draft for review purposes only and does not constitute Agency policy.
F-20 DRAFT—DO NOT CITE OR QUOTE
-------
1 • (151.25 -l)x 5 days -7 days = 107.321429
2 • Breathing rate, working = 1.25 m3/hour x 8 hours = 10 m3
3 • Breathing rate, not working = 0.625 m3/hour x 16 hours = 10 m3
4 • Total breathing rate = 10 + 10 = 20 m3/day
5 • Correction Factor Spring = 107.321429 - 151.25 x 10 - 20 = 0.354782
6
7
8 F.5.3.1.2.2. Summer
9 June 1 to August 31:5 days/week, 8 hours/day, 2 week summer vacation.
10
11
12 • (92 - 14) x 5 -7 = 55.714286 days
13 • Breathing rate, working = 1.25 m3/hour x 8 hours = 10 m3
14 • Breathing rate, not working = 0.625 m3/hour x 16 hours = 10 m3
15 • Total breathing rate = 10 + 10 = 20 m3/day
16 • Correction Factor Summer = 55.714286 - 92 x 10 - 20 = 0.302795
17
18
19 F.5.3.1.2.3. Fall
20 September 1 to December 31:5 days/week, 8 hours/day, with Christmas Day off.
21
22
23 • (122-l)x 5-7 = 86.428571 days
24 • Breathing rate, working, 8 hour day = 1.25 m3/hour x 8 hours = 10 m3
25 • Breathing rate, not working = 0.625 m3/hour x 16 hours = 10 m3
26 • Total breathing rate = 10 + 10 = 20 m3/day
27 • Correction Factor Fall = 86.428571 - 122 x 10 - 20 = 0.354215
28
29
30 F.5.4. Results of the Cumulative Human Equivalent Exposure Concentration (CHEEC)
31 To verify the accuracy of the CHEEC calculations, several quality control checks were
32 conducted. The distribution was evaluated by reviewing the mean, median, standard deviation,
33 highest 10 values, and lowest 10 values. Several workers were also randomly selected and their
34 values hand-calculated to ensure all programming was correct. Tables 5-7 provide a list of all
35 280 subjects participating in the 2004 Marysville health update (Rohs et al., 2008). These tables
This document is a draft for review purposes only and does not constitute Agency policy.
F-21 DRAFT—DO NOT CITE OR QUOTE
-------
1 describe each subject's identification number, job start and stop date, date of radiograph, age,
2 gender, body mass index, smoking history, asbestos exposures, health outcomes, and the
3 cumulative human equivalent exposure concentration (CHEEC) for all departmental exposures
4 they reported while employed at the OM Scott Marysville, OH plant.
5
6 F.6. STRENGTHS AND LIMITATIONS
7 There are major strengths in this exposure reconstruction project:
8
9
10 1. Data were gathered from court records, federal sources and archived files, totaling
11 over 3,000 pages. These data were reviewed and both qualitative and quantitative
12 data were abstracted to aid in this reconstruction.
13 2. Approximately five times more fiber measurements became available than had
14 been used in the original studies.
15 3. Two focus groups were conducted in 2010 with long term workers who provided
16 input regarding exposure and production process changes.
17 4. There were sufficient data available to examine exposure intensity over time for
18 jobs within the trionizing department as well as for other departments. These data
19 enhanced exposure estimates for all departments from 1972 to 1994.
20 5. Industrial hygiene data were available allowing for comparisons of fiber counts
21 when 100% Libby or 100% South Carolina vermiculite was used in order to
22 calculate a ratio of fibers in each.
23 6. There were data available from archived records, Scott memos, and worker
24 information that allowed for exposure estimates to be adjusted for type of
25 vermiculite used from 1957 until 1971 when no industrial hygiene data were
26 available.
27 7. Worker report data were available that provided documentation for increased
28 dustiness before industrial hygiene data were available, compared with years
29 when measurements were available.
30 8. Based on past and current data gathered in the focus group, exposures were
31 adjusted to account for seasonal work schedules by departments.
32 9. All decisions based on level of exposure by year were data driven.
33
34
35 The limitations for this project are also recognized:
This document is a draft for review purposes only and does not constitute Agency policy.
F-22 DRAFT—DO NOT CITE OR QUOTE
-------
1 1. The exposure metric used (fibers/cc) results from an analytical method that is a
2 count of fibers (defined as any viewed elongated particle in excess of 5 jim in
3 length and with a length to width ratio of 3:1) collected on a filter and viewed at
4 400x with light microscopy. The composition of the fiber is not known. Also, a
5 fiber with diameter less than a limit of resolution of 0.2 jim cannot be viewed
6 with this method.
7 2. It is unknown if other sampling results exist. If any are found in the future, these
8 can be incorporated into a future exposure assessment.
9 3. Some dusty activities may not have been sampled or rarely sampled e.g., summer
10 cleanup. We have no way of estimating the effect of these activities on overall
11 exposure estimates.
12 4. We did not reduce exposure estimates due to possible use of respiratory
13 protection. Substantially more documentation regarding enforced usage, fit
14 testing and cleaning/storage protocols would be needed for meaningful reduction
15 in exposure estimates.
16 5. By combining all individual trionizing job duties into one department exposure,
17 the nonexpander trionizing exposure estimates may have been overestimated as
18 there were more expander measurements, and these were somewhat higher than
19 for other job duties.
20 6. From 1980 forward, Libby vermiculite was not used. Thus for any individual
21 year during this period, exposure from a qualitative and quantitative perspective
22 does not reflect Libby Amphibole exposure.
23 7. Seasonal work schedule adjustments were based on recall of focus group
24 participants and may over or under estimate true durations and location of
25 additional work hours.
26
27
28 F.7. REFERENCES
29 ATSDR (Agency for Toxic Substances and Disease Registry). (2005). Health Consultation: Former O.C. Scott and
30 Sons Company, 14111 Scottslawn Road, Marysville, Union County, Ohio. EPA Facility ID:
31 OHD990834483. (PB2008-104070). Atlanta, GA.
32 Benson. R. (2009). Use of Libby Amphibole in Marysville, Ohio. Denver, CO: U.S. Environmental Protection
33 Agency.
34 Lockev. JE: Brooks. SM: Jarabek. AM: Khoury. PR: McKay. RT: Carson. A: Morrison. JA: Wiot JF: Spitz. HB.
3 5 (1984). Pulmonary changes after exposure to vermiculite contaminated with fibrous tremolite. Am Rev
36 RespirDis 129:952-958.
37 Rohs. A: Lockev. J: Dunning. K: Shukla. R: Fan. H: Hilbert. T: Borton. E: Wiot J: Meyer. C: Shipley. R:
3 8 Lemasters. G: Kapil. V. (2008). Low-level fiber-induced radiographic changes caused by Libby
39 vermiculite: A 25-year follow-up study. Am J Respir Crit Care Med 177: 630-637.
40 http://dx.doi.org/10.1164/rccm.200706-841OC.
This document is a draft for review purposes only and does not constitute Agency policy.
F-23 DRAFT—DO NOT CITE OR QUOTE
-------
1 APPENDIX G. EXTRA RISK AND UNIT RISK CALCULATION
2 G.I. MESOTHELIOMA MORTALITY
3 The increased risk of mesothelioma mortality attributable to continuous fiber exposure
4 was estimated using a life-table procedure based on the general U.S. population. The life-table
5 procedure involved the application of the estimated Libby Amphibole asbestos^specific toxicity
6 to a structured representation of the general U.S. population in such a manner as to yield
7 age-specific risk estimates for mesothelioma mortality in the absence and presence of exposure
8 to Libby Amphibole asbestos. Baseline all-cause mortality rates were included in the life-table
9 in such a way as to enable computation of the specific absolute risk of mesothelioma mortality
10 while accounting for other competing causes of mortality. For each age-interval in the life-table,
11 the effect estimates of the Poisson regression model analysis (the absolute risk) were used to
12 estimate mesothelioma mortality at a particular exposure level. These age-specific absolute risks
13 can then be summed over a lifetime. Different exposure levels are evaluated to ascertain what
14 magnitude of exposure would be expected to produce 1% absolute risk of mesothelioma
15 mortality. By this method, the exposure-response relationship determined in the Libby worker
16 cohort is used to estimate mesothelioma mortality in the general U.S. population that would be
17 expected from continuous lifetime environmental exposure to various concentrations of Libby
18 Amphibole asbestos.
19 Assuming no background risk for mesothelioma, extra risk is the same as absolute risk.
20 Absolute risk estimates were calculated using the effect estimates derived from the modeling of
21 the mesothelioma mortality risk and a life-table analysis program that accounts for competing
22 causes of death.2 The unit risk of mesothelioma is computed using the 95% upper bound to
23 estimate an upper bound for extra risk of mesothelioma due to Libby Amphibole asbestos
24 exposure. The upper bound calculation is specific to the exposure metric parameters; the effect
lrThe 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.
2This program is an adaptation of the approach previously used by the Committee on the Biological Effects of
Ionizing Radiation (BEIR. 1988). Compared to life-table methods based on full life exposures from birth, the
method used here yielded unit risk differences between full life exposure to scaled adult-only exposure between -3%
to -2% for the mesothelioma mortality unit risks for the two mesothelioma models (see Tables G-l and G-2). A
spreadsheet containing the extra risk calculation for the derivation of the LECM for mesothelioma mortality is
presented in Tables G-l and G-2.
This document is a draft for review purposes only and does not constitute Agency policy.
G-1 DRAFT—DO NOT CITE OR QUOTE
-------
1 of metric uncertainty in these values is discussed in Section 5.4.5.3. Because this human health
2 assessment derived a combined inhalation unit risk (IUR) for both mesothelioma and lung cancer
3 mortality, an interim value based on the central effect estimate (rather than the upper bound) is
4 also computed to avoid statistical concerns regarding the combination of upper bounds. Details
5 are shown in Section 5.4.5.3. This current assessment does not directly apply life-table
6 calculations to estimate partial lifetime risk scenarios; the use of the IUR for partial lifetime
7 extrapolations is discussed in Section 5.4.5.4.
8 U.S. age-specific all-cause mortality rates from the 2010 National Vital Statistics Report
9 (NVSR) for deaths in 2007 among all race and gender groups combined (Xu etal., 2010) were
10 used to specify the all-cause background mortality rates (R0) in the life-table analysis. The risk
11 with exposure (Rx) was computed up to age 85 years,3 assuming continuous environmental
12 exposure to Libby Amphibole asbestos. Conversions between occupational Libby Amphibole
13 asbestos exposures and continuous environmental asbestos exposures were made to account only
14 for differences in the amount of air inhaled per day during a higher effort occupational shift
15 (8 hours; 10 m3) compared to a standard 24-hour (20 m3) day (U.S. EPA, 1994) because results
16 were already based on a 365-day calendar year. The computation of the unit risk involved three
17 steps. The first step was to compute the unit risk for adults. This was achieved by initiating
18 exposure at age 16 years and maintaining continuous exposure throughout the remainder of life
19 while allowing for the incremental mathematical decay of previously accumulated exposure.4
20 An age of 16 years was used because it roughly matched the youngest age of a worker in the
21 subcohort and was consistent with the application of a similar life-table methodology when the
22 age-dependent adjustment factors (ADAFs) are applied; however, the application of age-
23 dependent adjustment factors was not recommended in this case (see Section 4.6.2.2). An
24 adjustment was also made in the life-table for the lag period, so that the age-specific risk
25 calculations began at 16+ (the length of the lag period) years of age. The standard assumption
26 used by the U.S. Environmental Protection Agency (EPA) is that the average lifetime spans
27 70 years. Because the adult-only-exposure unit risk excluded the first 16 years, the
28 adult-only-exposure unit risk based on 54 years was then rescaled for an entire lifetime of
3Note 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.
4Exposures in the life-tables were computed at the mid-point of each age interval and appropriately lagged.
This document is a draft for review purposes only and does not constitute Agency policy.
G-2 DRAFT—DO NOT CITE OR QUOTE
-------
1 continuous exposure by multiplying the interim value for adult-only-exposure by 70/54 to cover
2 the childhood years (<16 years) to compute the "adult-based" unit risk. After reseating, the
3 resulting "adult-based" lifetime unit risk estimate (in contrast to the unsealed
4 "adult-only-exposure" unit risk estimate obtained from the life-table calculations) may be
5 prorated for less-than-lifetime exposure scenarios in the same manner as would be used for an
6 "adult-based" unit risk estimate derived from a rodent bioassay (see Section 5.4.5.4).
7 Consistent with the Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005), the
8 same data and methodology were also used to estimate the exposure level effective concentration
9 (ECX) and the associated 95% lower confidence limit of that exposure level effective
10 concentration (LEG*) corresponding to an absolute risk of 1% (x = 0.01). A 1%-risk level is
11 commonly used for the determination of the point of departure (POD) for low-dose extrapolation
12 from epidemiological data, and the LEG value corresponding to that risk level was used as the
13 actual POD.
14 The following tables illustrate the computational details of the unit risks for
15 mesothelioma mortality (see Tables G-l and G-2). The results of Tables G-l and G-2 are shown
16 in Table 5-16 and are not adjusted for the underascertainment of mesothelioma described in
17 Section 5.4.5.1.1. The unit risks adjusted for underascertainment are shown in Table 5-17.
18
19
20 Column Definitions for Tables G-l and G-2:
21 Column A: Age interval up to age 85.
22 Column B: All-cause mortality rate for interval / (x 105/year) (Xu et al.. 2010)
23 Column C: All-cause hazard rate for interval /' (h*i) (= all-cause mortality rate x number of
24 years in age interval).
25 Column D: Probability of surviving interval /' (q,) [= exp(-/7*,)].
26 Column E: Probability of surviving up to interval /' (£•) (Si = 1; S, = Si-i x qt-i, for /' > 1).
27 Column F: Lagged exposure at mid-interval (x dose) assuming constant exposure was initiated
28 at age 16.
29 Column G: Mesothelioma mortality hazard rate in exposed people for interval. To estimate the
30 LECoi, i.e., the 95% lower bound on the continuous exposure giving an extra risk of
31 1%, the 95% upper bound on the regression coefficient is used.
This document is a draft for review purposes only and does not constitute Agency policy.
G-3 DRAFT—DO NOT CITE OR QUOTE
-------
1 Column H: All-cause hazard rate in exposed people for interval / (h*Xi) [= h*t + (hxt - hi)].
2 Column I: Probability of surviving interval / without dying from mesothelioma for exposed
3 people (qXj) [= exp(-/z*x,)].
4 Column J: Probability of surviving up to interval /' without dying from mesothelioma for
5 exposed people (Sx,) (Sx\ = 1; Sxj = Sxj-\ x qxi—\, for /' > 1).
6 Column K: Conditional probability of dying from mesothelioma in interval /' for exposed people
7 [= (hxj H- h*Xi) x Sxi x (1 - qXj)] (Rx, the lifetime probability of dying from
8 mesothelioma for exposed people = the sum of the conditional probabilities across
9 the intervals).
10
11
12 Note that the life-tables for mesothelioma mortality estimate the extra risk as the absolute
13 risk as there is no assumption of a background risk in the absence of exposure. In each of the
14 life-tables, inhalation exposure commences at age 16 years and continues at the same exposure
15 concentration for the duration of the life-table. This allows for the computation of an
16 "adult-only-exposure" occupational lifetime unit risk, which is then scaled by a ratio of 70:54 to
17 account for risk over the standard 70-year lifetime. While exposure is initiated in the life-table at
18 age 16 years, this exposure is lagged to match the corresponding exposure-response models,
19 which provide the hazard rates per unit of exposure. For example, in Table G-l, Column F
20 shows exposure lagged by 10 years so that no lagged exposure appears in the table prior to age
21 26 years (16 + 10). In Table G-2, Column F shows exposure lagged by 15 years so that no
22 lagged exposure appears in the table prior to age 3 1 years (16 + 15). Note that risks are initially
23 shown in 1-year intervals because children's risk intervals can be smaller, and there was a need
24 to be able to begin exposures at 16 years.
This document is a draft for review purposes only and does not constitute Agency policy.
G-4 DRAFT—DO NOT CITE OR QUOTE
-------
Table G-l. Mesothelioma extra risk calculation for environmental exposure to 0.1479 fibers/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 as the reasonable upper bound
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/
year)
684.5
28.6
28.6
28.6
29.9
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
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
D
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
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
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
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
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
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
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
K
Exposed
cond. prob. of
meso. 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
I
_. c
-------
Table G-l. Mesothelioma extra risk calculation for environmental exposure to 0.1479 fibers/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 as the reasonable upper bound (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
50-54
B
All-cause
mortality
(xlO5/
year)
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
509.0
C
All-cause
hazard
rate (/?*)
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
0.0255
D
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
0.9749
E
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
0.9385
F
Lagged
exp. mid. int.
(ATdose)
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.144
0.401
0.626
0.821
1.268
1.701
1.918
2.026
2.080
G
Exposed
meso.
hazard
rate (hx)
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0001
0.0002
0.0003
0.0004
0.0006
0.0009
0.0010
0.0010
0.0011
H
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
0.0011
0.0012
0.0013
0.0014
0.0062
0.0082
0.0121
0.0180
0.0265
I
Exposed
prob. of
surviving
interval
(qx)
0.9994
0.9994
0.9994
0.9990
0.9990
0.9990
0.9990
0.9990
0.9990
0.9989
0.9988
0.9987
0.9986
0.9938
0.9919
0.9880
0.9821
0.9738
J
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
0.9805
0.9793
0.9780
0.9767
0.9706
0.9628
0.9512
0.9342
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
0.0001
0.0002
0.0003
0.0004
0.0006
0.0008
0.0009
0.0010
0.0010
I
_. >J
9 s§
ON o
I
a,
H
H S
O^
O
o
-------
H
H S
O^
O
o
Table G-l. Mesothelioma extra risk calculation for environmental exposure to 0.1479 fibers/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 as the reasonable upper bound (continued)
A
Age
int.
55-59
60-64
65-69
70-74
75-79
80-84
B
All-cause
mortality
(xlO5/
year)
726.3
1,068.3
1,627.5
2,491.3
3,945.9
6,381.4
C
All-cause
hazard
rate (/?*)
0.0363
0.0534
0.0814
0.1246
0.1973
0.3191
D
Prob. of
surviving
interval
(?)
0.9643
0.9480
0.9218
0.8829
0.8209
0.7268
E
Prob. of
surviving
up to
interval
(S)
0.9149
0.8823
0.8364
0.7710
0.6807
0.5588
F
Lagged
exp. mid. int.
(ATdose)
2.107
2.121
2.127
2.131
2.132
2.133
G
Exposed
meso.
hazard
rate (hx)
0.0011
0.0011
0.0011
0.0011
0.0011
0.0011
H
Exposed
all-cause haz.
rate
(h*x)
0.0374
0.0545
0.0825
0.1256
0.1984
0.3202
I
Exposed
prob. of
surviving
interval
(qx)
0.9633
0.9470
0.9209
0.8819
0.8201
0.7260
J
Exposed
prob. of
surviving
up to int.
(Sx)
0.9098
0.8764
0.8299
0.7642
0.6740
0.5527
K
Exposed
cond. prob. of
meso. in interval
(Rx)
0.0010
0.0009
0.0009
0.0008
0.0007
0.0005
Absolute Rx = 0.0 100
I
§
I
a,
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 years); scaled occupational lifetime unit risk = 0.0876 (scaled by ratio of 70:54 to account for risk over 70-year lifetime).
-------
Table G-2. Mesothelioma extra risk calculation for environmental exposure to 0.2446 fibers/cc Libby
Amphibole asbestos using the metric of cumulative exposure with a 15-year exposure lag and a 5-year half-life of
exposure, as described in Section 5.4.5.3 as the lowest information criterion
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/ year)
684.5
28.6
28.6
28.6
29.9
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
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
D
Prob. of
surviving
interval
(q)
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
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
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
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
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
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
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
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
I
_. GS
9 s§
oo o
I
a,
H
H £
O^
O
o
-------
Table G-2. Mesothelioma extra risk calculation for environmental exposure to 0.2446 fibers/cc Libby
Amphibole asbestos using the metric of cumulative exposure with a 15-year exposure lag and a 5-year half-life
of exposure, as described in Section 5.4.5.3 as the lowest information criterion (continued)
A
Age int.
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
B
All-cause
mortality
(xl05/year)
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
110.8
110.8
110.8
110.8
C
All-cause
hazard
rate (/?*)
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
0.0055
0.0055
0.0055
0.0055
D
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
0.9945
0.9945
0.9945
0.9945
E
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
0.9777
0.9777
0.9777
0.9777
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
0.238
0.664
1.035
1.357
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
0.0001
0.0002
0.0004
0.0005
H
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
0.0011
0.0012
0.0013
0.0015
0.0016
I
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
0.9989
0.9988
0.9987
0.9985
0.9984
J
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
0.9766
0.9754
0.9741
0.9727
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
0.0001
0.0002
0.0003
0.0005
I
_. >J
9 s§
\o o
I
a,
H
H S
O^
O
o
-------
Table G-2. Mesothelioma extra risk calculation for environmental exposure to 0.2446 fibers/cc Libby
Amphibole asbestos using the metric of cumulative exposure with a 15-year exposure lag and a 5-year half-life
of exposure, as described in Section 5.4.5.3 as the lowest information criterion (continued)
A
Age int.
35-39
40-44
45-49
50-54
55-59
60-64
65-69
70-74
75-79
80-84
B
All-cause
mortality
(xl05/year)
145.8
221.6
340.0
509.0
726.3
1,068.3
1,627.5
2,491.3
3,945.9
6,381.4
C
All-cause
hazard
rate (/?*)
0.0073
0.0111
0.0170
0.0255
0.0363
0.0534
0.0814
0.1246
0.1973
0.3191
D
Prob. of
surviving
interval
(?)
0.9927
0.9890
0.9831
0.9749
0.9643
0.9480
0.9218
0.8829
0.8209
0.7268
E
Prob. of
surviving
up to
interval
(S)
0.9723
0.9652
0.9546
0.9385
0.9149
0.8823
0.8364
0.7710
0.6807
0.5588
F
Lagged
exp. mid. int.
(ATdose)
2.097
2.813
3.171
3.350
3.440
3.485
3.507
3.518
3.524
3.527
G
Exposed
meso.
hazard
rate (hx)
0.0007
0.0010
0.0011
0.0011
0.0012
0.0012
0.0012
0.0012
0.0012
0.0012
H
Exposed
all-cause
haz. rate
(h*x)
0.0080
0.0120
0.0181
0.0266
0.0375
0.0546
0.0826
0.1258
0.1985
0.3203
I
Exposed
prob. of
surviving
interval
(qx)
0.9920
0.9880
0.9821
0.9738
0.9632
0.9469
0.9207
0.8818
0.8200
0.7259
J
Exposed
prob. of
surviving
up to int.
(Sx)
0.9712
0.9634
0.9519
0.9348
0.9103
0.8768
0.8302
0.7644
0.6740
0.5527
K
Exposed
cond. prob. of
meso. in
interval
(Rx)
0.0007
0.0009
0.0010
0.0011
0.0011
0.0010
0.0010
0.0009
0.0007
0.0006
Absolute Rx = 0.0 100
I
§
I
a.
Q OQ exp. = exposure, haz. = hazard, int. = interval, meso. = mesothelioma, mid. = mid-interval, Prob. = probability.
H » Absolute risk = 0.01000; exp. level = 0.2446; Occupational lifetime unit risk = 0.01/0.2446 = 0.0409 (Based on occupational exposures beginning at age
O ^ 16 years); Scaled occupational lifetime unit risk = 0.0530 (Scaled by ratio of 70:54 to account for risk over 70-year lifetime).
O '
O
O
-------
1 G.2. LUNG CANCER MORTALITY
2 Lung cancer mortality risk computations are very similar to mesothelioma mortality
3 computations above (see Section G. 1), with one important difference that extra risk is used for
4 lung cancer. Extra risk is defined as equaling (R* - RO) + (1 - RO), where Rx is the lifetime lung
5 cancer mortality risk in the exposed population and R0 is the lifetime lung cancer mortality risk
6 in an unexposed population (i.e., the background risk). U.S. age-specific all-cause mortality
7 rates from the 20 1 0 National Vital Statistics Report (Xu et al.. 2010) for deaths in 2007 among
8 all race and gender groups combined were used to specify the all-cause background mortality
9 rates (R0) in the life-table analysis. Cause-specific background mortality rates for cancers of the
10 lung, trachea, and bronchus were obtained from a Surveillance, Epidemiology, and End Results
1 1 (SEER) report on mortality during 2003-2007 (2003-2007 Surveillance Epidemiology and End
12 Results Table 15.10, age-specific U.S. death rates).
13 The following tables show details of the computations of the unit risks for lung-cancer
14 mortality (see Tables G-3 and G-4). The results of Tables G-3 and G-4 are shown in Table 5-19.
15
16
17 Column Definitions for Tables G-3 and G-4:
18 Column A: Age interval up to age 85.
1 9 Column B : All-cause mortality rate for interval / (x 1 05/year) (Xu et al.. 2010)
20 Column C: Lung-cancer mortality rate for interval /' (x 105/year) (2003-2007 Surveillance,
21 Epidemiology and End Results Table 15.10, age-specific U.S. death rates).
22 Column D: All-cause hazard rate for interval / (//*,) (= all-cause mortality rate x number of
23 years in age interval).
24 Column E: Probability of surviving interval /' (q,) [= exp(-/7*,)].
25 Column F: Probability of surviving up to interval /' (£•) (Si = 1; Si = Si-i x qt-i, for /' > 1).
26 Column G: Lung-cancer mortality hazard rate for interval /
27 (= lung-cancer mortality rate x number of years in interval).
28 Column H: Conditional probability of dying from lung cancer in interval /'
29 [= (hi H- h*i) x Si x (1 - <7;)], i.e., conditional upon surviving up to interval / (R0, the
30 background lifetime probability of dying from lung cancer = the sum of the
3 1 conditional probabilities across the intervals).
This document is a draft for review purposes only and does not constitute Agency policy.
G- 1 1 DRAFT— DO NOT CITE OR QUOTE
-------
1 Column I: Lagged exposure at mid-interval (x dose) assuming constant exposure was initiated
2 at age 16.
3 Column!: Lung-cancer mortality hazard rate in exposed people for interval. To estimate the
4 LECoi, i.e., the 95% lower bound on the continuous exposure giving an extra risk of
5 1%, the 95% upper bound on the regression coefficient is used, i.e.,
6 Maximum Likelihood Estimate + 1.645 x standard error.
7 Column K: All-cause hazard rate in exposed people for interval /' (h*Xj) [= h*t + (hxt - ht~)].
8 Column L: Probability of surviving interval /' without dying from lung cancer for exposed
9 people (qXj) [= exp(-/7*x,)].
10 Column M: Probability of surviving up to interval / without dying from lung cancer for exposed
1 1 people (Sxj) (Sxi = 1; Sx[ = Sxt-i x ^_i, for / > 1).
12 Column N: Conditional probability of dying from lung cancer in interval /' for exposed people
13 [= (hxt + h*Xj) x Sxj x (1 - qXj)] (Rx, the lifetime probability of dying from lung
14 cancer for exposed people = the sum of the conditional probabilities across the
15 intervals).
16
17
18 In each of the life-tables, inhalation exposure commences at age 16 years and continues
19 at the same exposure concentration for the duration of the life-table. This allows for the
20 computation of an "adult-only-exposure" occupational lifetime unit risk, which is then scaled by
21 a ratio of 70:54 to account for risk over the standard 70-year lifetime. While exposure is initiated
22 at age 16 years, this exposure is lagged to match the corresponding exposure-response models,
23 which provide the hazard rates per unit of exposure. For example, in Tables G-3 and G-4,
24 Column I shows exposure lagged by 10 years so that no lagged exposure appears prior to age
25 26 years.
This document is a draft for review purposes only and does not constitute Agency policy.
G-12 DRAFT—DO NOT CITE OR QUOTE
-------
Table G-3. Lung cancer extra risk calculation for environmental exposure to 0.191 fibers/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 as the reasonable upper bound
A
Age
Int.
<1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
B
All-cause
mortality
(xlO5/
year)
684.5
28.6
28.6
28.6
29.9
13.7
13.7
13.7
13.7
13.7
18.7
18.7
18.7
18.7
18.7
61.9
C
Lung CA
mortality
(xl05/year)
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
E
Prob. of
surviving
interval
(1)
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
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
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
H
Cond. prob.
of lung CA
mortality
in interval
(Ro)
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
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
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
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
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
N
Exposed
cond.
prob.
of lung CA
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
I
§
I
a,
H
H S
O^
O
o
-------
Table G-3. Lung cancer extra risk calculation for environmental exposure to 0.191 fibers/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 as the reasonable upper bound (continued)
A
Age
Int.
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30-34
35-39
40-44
B
All-cause
mortality
(xl05/year)
61.9
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
C
Lung CA
mortality
(xl05/year)
0
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
D
All-
cause
hazard
rate
(h*)
0.0006
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
E
Prob. of
surviving
interval
(q)
0.9994
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
F
Prob. of
surviving
up to
interval
(S)
0.9898
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
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
0.0001
0.0004
H
Cond.
prob.
of lung CA
mortality
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
0.0001
0.0004
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
0.10
0.29
0.48
0.67
1.24
2.20
3.15
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
0.0001
0.0004
K
Exposed
all-cause
haz. rate
(h*x)
0.0006
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
L
Exposed
prob. of
surviving
interval
(qx)
0.9994
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
M
Exposed
prob. of
surviving
up to int.
(Sx)
0.9898
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
N
Exposed
cond.
prob.
of lung CA
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
0.0001
0.0004
I
§
I
a,
H
H £
O^
O
o
-------
Table G-3. Lung cancer extra risk calculation for environmental exposure to 0.191 fibers/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 as the reasonable upper bound (continued)
A
Age
Int.
45-49
50-54
55-59
60-64
65-69
70-74
75-79
80-84
B
All-cause
mortality
(xl05/year)
340.0
509.0
726.3
1,068.3
1,627.5
2,491.3
3,945.9
6,381.4
C
Lung CA
mortality
(xl05/year)
20.2
39.8
74.7
139.8
220.9
304.3
369.5
379.4
D
All-
cause
hazard
rate
(h*)
0.0170
0.0255
0.0363
0.0534
0.0814
0.1246
0.1973
0.3191
E
Prob. of
surviving
interval
(q)
0.9831
0.9749
0.9643
0.9480
0.9218
0.8829
0.8209
0.7268
F
Prob. of
surviving
up to
interval
(S)
0.9546
0.9385
0.9149
0.8823
0.8364
0.7710
0.6807
0.5588
G
Lung
CA
hazard
rate
(h)
0.0010
0.0020
0.0037
0.0070
0.0110
0.0152
0.0185
0.0190
H
Cond.
prob.
of lung CA
mortality
in interval
(*,)
0.0010
0.0018
0.0034
0.0060
0.0089
0.0110
0.0114
0.0091
R0 = 0.0531
I
Lagged
exp.
mid.
int.
(ATdose)
4.11
5.06
6.02
6.97
7.93
8.88
9.84
10.79
J
Exposed
lung CA
hazard
rate (hx)
0.0011
0.0022
0.0042
0.0080
0.0129
0.0181
0.0224
0.0235
K
Exposed
all-cause
haz. rate
(h*x)
0.0171
0.0257
0.0368
0.0544
0.0832
0.1275
0.2013
0.3236
L
Exposed
prob. of
surviving
interval
(qx)
0.9831
0.9747
0.9639
0.9470
0.9201
0.8803
0.8177
0.7236
M
Exposed
prob. of
surviving
up to int.
(Sx)
0.9545
0.9384
0.9146
0.8815
0.8348
0.7682
0.6762
0.5529
N
Exposed
cond.
prob.
of lung CA
in interval
(Rx)
0.0010
0.0020
0.0038
0.0069
0.0103
0.0131
0.0137
0.0111
Rx = 0.0625
I
§
I
a,
H
H £
O^
O
o
a 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 years); scaled
^ occupational lifetime unit = 0.0679 (scaled by ratio of 70:54 to account for risk over 70-year lifetime).
-------
Table G-4. Lung cancer extra risk calculation for environmental exposure to 0.333 fibers/cc Libby Amphibole
asbestos using the metric of cumulative exposure with a 10-year exposure lag and a 10-year half-life of exposure,
as described in Section 5.4.5.3 as the lowest information criterion
A
Age
int.
<1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
B
All-cause
mortality
(xl05/year)
684.5
28.6
28.6
28.6
29.9
13.7
13.7
13.7
13.7
13.7
18.7
18.7
18.7
18.7
18.7
61.9
C
Lung CA
mortality
(xl05/year)
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
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
F
Prob. of
surviving
up to
interval
OS)
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
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
H
Cond.
prob. of
lung CA
mortality
in
interval
(R0)
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
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
K
Exposed
all-cause
hazard
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
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
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
N
Exposed
cond.
prob. of
lung CA
in
interval
CR*)
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
§
I
a,
H
H £
O^
O
o
-------
Table G-4. Lung cancer extra risk calculation for environmental exposure to 0.333 fibers/cc Libby Amphibole
asbestos using the metric of cumulative exposure with a 10-year exposure lag and a 10-year half-life of exposure,
as described in Section 5.4.5.3 as the lowest information criterion (continued)
A
Age
int.
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30-34
35-39
B
All-cause
mortality
(xl05/year)
61.9
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
C
Lung CA
mortality
(xl05/year)
0
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
D
All-
cause
hazard
rate
(h*)
0.0006
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
E
Prob. of
surviving
interval
(1)
0.9994
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
F
Prob. of
surviving
up to
interval
(S)
0.9898
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
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
0.0001
H
Cond.
prob. of
lung CA
mortality
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
0.0001
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
0.16
0.48
0.77
1.04
1.74
2.64
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
0.0001
K
Exposed
all-cause
hazard
rate
(h*x)
0.0006
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
L
Exposed
prob. of
surviving
interval
(qx)
0.9994
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
M
Exposed
prob. of
surviving
up to int.
(Sx)
0.9898
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
N
Exposed
cond.
prob. of
lung CA
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
0.0001
I
§
I
a,
H
H S
O^
O
o
-------
Table G-4. Lung cancer extra risk calculation for environmental exposure to 0.333 fibers/cc Libby Amphibole
asbestos using the metric of cumulative exposure with a 10-year exposure lag and a 10-year half-life of exposure,
as described in Section 5.4.5.3 as the lowest information criterion (continued)
A
Age
int.
40-44
45-49
50-54
55-59
60-64
65-69
70-74
75-79
80—
84
B
All-cause
mortality
(xl05/year)
221.6
340.0
509.0
726.3
1,068.3
1,627.5
2,491.3
3,945.9
6,381.4
C
Lung CA
mortality
(xl05/year)
7.9
20.2
39.8
74.7
139.8
220.9
304.3
369.5
379.4
D
All-
cause
hazard
rate
(h*)
0.0111
0.0170
0.0255
0.0363
0.0534
0.0814
0.1246
0.1973
0.3191
E
Prob. of
surviving
interval
(1)
0.9890
0.9831
0.9749
0.9643
0.9480
0.9218
0.8829
0.8209
0.7268
F
Prob. of
surviving
up to
interval
(S)
0.9652
0.9546
0.9385
0.9149
0.8823
0.8364
0.7710
0.6807
0.5588
G
Lung
CA
hazard
rate
(h)
0.0004
0.0010
0.0020
0.0037
0.0070
0.0110
0.0152
0.0185
0.0190
H
Cond.
prob. of
lung CA
mortality
in interval
(*»)
0.0004
0.0010
0.0018
0.0034
0.0060
0.0089
0.0110
0.0114
0.0091
RO = 0.0531
I
Lagged
exp.
mid.
int.
(ATdose)
3.27
3.72
4.04
4.26
4.42
4.53
4.61
4.67
4.71
J
Exposed
lung CA
hazard
rate (hx)
0.0004
0.0012
0.0023
0.0044
0.0083
0.0131
0.0181
0.0220
0.0226
K
Exposed
all-cause
hazard
rate
(h*x)
0.0111
0.0172
0.0258
0.0370
0.0547
0.0834
0.1274
0.2008
0.3227
L
Exposed
prob. of
surviving
interval
(qx)
0.9889
0.9830
0.9746
0.9637
0.9468
0.9200
0.8803
0.8180
0.7242
M
Exposed
prob. of
surviving
up to int.
(Sx)
0.9652
0.9545
0.9383
0.9144
0.8812
0.8343
0.7675
0.6757
0.5527
N
Exposed
cond.
prob. of
lung CA
in
interval
(Rx)
0.0004
0.0011
0.0021
0.0039
0.0071
0.0105
0.0130
0.0135
0.0107
Rx = 0.0626
I
§
I
a.
^ ^ CA = cancer, cond. = conditional, exp. = exposure, haz. = hazard, int. = interval, mid. = mid-interval, Prob. = probability.
Q >T3 Extra risk = 0.01001; exp. level = 0.333; occupational lifetime unit risk = 0.01/0.333 = 0.0300 (based on occupational exposures beginning at age 16 years);
d §r scaled occupational lifetime unit = 0.0389 (scaled by ratio of 70:54 to account for risk over 70-year lifetime).
O^
O
O
-------
1 G.3. REFERENCES
2 BEIR (Committee on the Biological Effects of Ionizing Radiation). (1988). Health risks of radon and other
3 internally deposited alpha-emitters: BEIR IV. Washington, DC: National Academy Press.
4 http://www.rap.edu/openbook.php7isbF0309037972.
5 U.S. EPA (U.S. Environmental Protection Agency). (1994). Methods for derivation of inhalation reference
6 concentrations and application of inhalation dosimetry. (EPA/600/8-90/066F). Research Triangle Park, NC:
7 U.S. Environmental Protection Agency, Office of Research and Development, Office of Health and
8 Environmental Assessment, Environmental Criteria and Assessment Office.
9 http://cfpub.epa. gov/ncea/cfm/recordisplav.cfm?deid=71993.
10 U.S. EPA (U.S. Environmental Protection Agency). (2005). Guidelines for carcinogen risk assessment. (EPA/630/P-
11 03/001F). Washington. DC. http://www.epa.gov/cancerguidelines/.
12 Xu. JQ: Kochanek. KD: Murphy. SL: Tejada-Vera. B. (2010). Deaths: Final Data for 2007. Hyattsville, MD:
13 National Center for Health Statistics, http://www.cdc.gov/nchs/data/nvsr/nvsr58/nvsr58 19.pdf.
14
This document is a draft for review purposes only and does not constitute Agency policy.
G-19 DRAFT—DO NOT CITE OR QUOTE
-------
1 APPENDIX H. GLOSSARY OF ASBESTOS TERMINOLOGY
2 Acicular: The very long and very thin, often needle-like shape that characterizes some prismatic
3 crystals. (Prismatic crystals have one elongated dimension and two other dimensions that
4 are approximately equal). Acicular crystals or fragments do not have the strength,
5 flexibility, or other properties often associated with asbestiform fibers.
6
7 Actinolite: An amphibole mineral in the tremolite-ferroactinolite series. Actinolite can occur in
8 both asbestiform and nonasbestiform mineral habits. The asbestiform variety is often
9 referred to as actinolite asbestos.
10
11 Amosite: An amphibole mineral in the cummingtonite-grunerite series that occurs in the
12 asbestiform habit. The name amosite is a commercial term derived from the acronym for
13 "Asbestos Mines of South Africa." Amosite is sometimes referred to as "brown
14 asbestos."
15
16 Amphibole: A group of minerals composed of double-chain SiC>4 tetrahedra linked at the vertices
17 and generally containing ions of iron and/or magnesium in their structures. Amphibole
18 minerals are of either igneous or metamorphic origin. Amphiboles can occur in a variety
19 of mineral habits including asbestiform and nonasbestiform.
20
21 Anthophyllite: An amphibole mineral that can occur in both the asbestiform and nonasbestiform
22 mineral habits. The asbestiform variety is referred to as anthophyllite asbestos.
23
24 Asbestiform: A specific type of mineral fibrosity in which crystal growth is primarily in one
25 dimension, and the crystals form as long, flexible fibers. In minerals occurring in
26 asbestiform habit, fibers form in bundles that can be separated into smaller bundles and
27 ultimately into fibrils.
28
This document is a draft for review purposes only and does not constitute Agency policy.
H-1 DRAFT—DO NOT CITE OR QUOTE
-------
1 Asbestos: A generic term for silicate minerals occurring in the asbestiform habit, usually used to
2 refer to those minerals that have been commercially exploited as asbestos, including
3 chrysotile in the serpentine mineral group and tremolite asbestos, actinolite asbestos,
4 anthophyllite asbestos, cummingtonite-grunerite asbestos (amosite), and riebeckite
5 asbestos (crocidolite) in the amphibole mineral group.
6
7 Asbestos Structure: A term applied to any connected or overlapping grouping of asbestos fibers
8 or bundles, with or without other particles.
9
10 Aspect Ratio: The ratio of the length of a particle to its diameter.
11
12 Biopersistence: The ability to remain in the lung or other tissue. Biopersistence of mineral fibers
13 is a function of their fragility, solubility, and clearance.
14
15 Chrysotile: A mineral in the serpentine mineral group that occurs in the asbestiform habit.
16 Chrysotile generally occurs segregated as parallel fibers in veins or veinlets and can be
17 easily separated into individual fibers or bundles. Often referred to as "white asbestos,"
18 chrysotile is used commercially in cement or friction products and for its good
19 spinnability in the making of textile products.
20
21 Cleavage Fragment: A particle, formed by comminution (i.e., crushing, grinding, or breaking)
22 of minerals, often characterized by parallel sides. In contrast to fibers from an asbestos
23 mineral; elongate mineral particles in a population of cleavage fragments are generally
24 wider and shorter, generally have a lower aspect ratio, and do not exhibit fibrillar
25 bundling at any level of examination.
26
27 Countable Particle: A particle that meets specified dimensional criteria and is (to be) counted
28 according to an established protocol. A countable particle under the National Institute for
29 Occupational Safety and Health asbestos fiber definition is any acicular crystal,
30 asbestiform fiber, prismatic crystal, or cleavage fragment of a covered mineral that is
This document is a draft for review purposes only and does not constitute Agency policy.
H-2 DRAFT—DO NOT CITE OR QUOTE
-------
1 longer than 5 jim and has a minimum aspect ratio of 3:1 based on a microscopic analysis
2 of an airborne sample using NIOSH Method 7400 or an equivalent method.
3
4 Crocidolite: An asbestiform amphibole mineral in the glaucophane-riebeckite series.
5 Crocidolite, commonly referred to as "blue asbestos," is a varietal name for the
6 asbestiform habit of the mineral riebeckite.
7
8 Durability: The tendency of particles to resist degradation in body fluids.
9
10 Elongated Mineral Particle (EMP): Any particle or fragment of a mineral (e.g., fibril or bundle
11 of fibrils: acicular, prismatic, or cleavage fragment) with a minimum aspect ratio of 3:1,
12 based on a microscopic analysis of an airborne sample using NIOSH Method 7400 or an
13 equivalent method.
14
15 Elongated Particle (EP): A particle with a minimum aspect ratio of 3:1, based on a microscopic
16 analysis of an airborne sample using NIOSH Method 7400 or an equivalent method.
17
18 Fiber. "Fiber" can be used in a regulatory context or in a mineralogical context.
19
20 In the regulatory context, a fiber is an elongated particle equal to or longer than 5 jim
21 with a minimum aspect ratio of 3:1. The dimensional determination is made based on a
22 microscopic analysis of an air sample using NIOSH Method 7400 or an equivalent
23 method.
24
25 In the mineralogical context, a fiber is an elongated crystalline unit that resembles an
26 organic fiber and that can be separated from a bundle or appears to have grown
27 individually in that shape.
28
29 Fibril: A single fiber of asbestos that cannot be further separated longitudinally into thinner
30 components without losing its fibrous properties or appearances.
This document is a draft for review purposes only and does not constitute Agency policy.
H-3 DRAFT—DO NOT CITE OR QUOTE
-------
1 Fibrous: A descriptive characteristic of a mineral composed of parallel, radiating, or interlaced
2 aggregates of fibers, from which the fibers are sometimes separable.
O
4 Fragility: The tendency of particles to break into smaller particles.
5
6 Libby Amphibole Asbestos: The term used in this document to identify the mixture of amphibole
7 mineral fibers of varying elemental composition (e.g., winchite, richterite, tremolite, etc.)
8 that have been identified in the Rainy Creek complex near Libby, MT as described in
9 Section 2.2.
10
11 Nonasbestiform: The term used to describe fibers not having an asbestiform habit. The massive
12 nonfibrous forms of the asbestos minerals have the same chemical formula and internal
13 crystal structure as the asbestiform variety but have crystal habits where growth is more
14 equivalent in two or three dimensions instead of primarily one dimension. When milled
15 or crushed, nonasbestiform minerals generally do not break into fibers/fibrils but rather
16 into fragments resulting from cleavage along the two or three growth planes. Often,
17 cleavage fragments can appear fibrous.
18
19 Primary Structure: A fibrous structure that is a separate entity in the transmission electron
20 microscope image.
21
22 Refractory Ceramic Fiber (RCF): An amorphous, synthetic fiber produced by melting and
23 blowing or spinning calcined kaolin clay or a combination of alumina (A^Os) and silicon
24 dioxide (SiO2). Oxides (such as zirconia, ferric oxide, titanium oxide, magnesium oxide,
25 and calcium oxide) and alkalis may be added.
26
27 Solid Solution Series: A grouping of minerals that includes two or more minerals in which the
28 cations in secondary structural position are similar in chemical properties and size and
29 can be present in variable but frequently limited ratios.
30
This document is a draft for review purposes only and does not constitute Agency policy.
H-4 DRAFT—DO NOT CITE OR QUOTE
-------
1 Structure: A single fiber, fiber bundle, cluster, or matrix.
2
3 Synthetic Vitreous Fiber (SVF): Any of a number of manufactured fibers produced by the
4 melting and subsequent fiberization of kaolin clay, sand, rock, slag, etc. Fibrous glass,
5 mineral wool, ceramic fibers, and alkaline earth silicate wools are the major types of
6 SVF, also called man-made mineral fiber (MMMF) or man-made vitreous fiber
7 (MMVF).
8
9 Thoracic-size Particle: A particle with an aerodynamic equivalent diameter that enables it to be
10 deposited in the airways of the lung or the gas exchange region of the lung when inhaled.
11
12 Tremolite: An amphibole mineral in the series tremolite-ferroactinolite. Tremolite can occur in
13 both fibrous and nonfibrous mineral habits. The asbestiform variety is often referred to
14 as tremolite asbestos. Due only to changes in the International Mineralogical
15 Association's amphibole nomenclature, subsets of what was formerly referred to as
16 tremolite asbestos are now mineralogically specified as asbestiform winchite and
17 asbestiform richterite.
This document is a draft for review purposes only and does not constitute Agency policy.
H-5 DRAFT—DO NOT CITE OR QUOTE
------- |