DRAFT
Contract No. DTRS57-01-C-10044.
FINAL DRAFT
TECHNICAL SUPPORT DOCUMENT FOR A
PROTOCOL TO ASSESS ASBESTOS-RELATED RISK
(Volume 1 of 2: Text)
Prepared for:
Mark Raney
Volpe Center
U.S. Department of Transportation
55 Broadway
Kendall Square
Cambridge MA 02142
and
Chris Weis
Paul Peronard
U.S. Environmental Protection Agency
Region 8
999 18* St, Suite 500
Denver, CO 80202
Prepared by:
D. Wayne Berman
Aeolus, Inc.
751 Taft St
Albany, CA 94708
and
Kenny Crump
The KS Crump Group, Inc.
602 E. Georgia Ave
Ruston, LA 71270
September 4,2001
-------
ACKNOWLEDGMENTS
This technical support document was prepared by D. Wayne Berman of Aeolus, Inc.,
Albany, California and Kenny Crump, The KS Crump Group, Inc., Ruston, Louisiana
under contract to the Volpe Center, Department of Transportation (Contract No.
DTRS57-01-C-10044). The Technical Manager for this project is Mark Raney. The
Contracting Officer is Linda Byrne. We also wish to acknowledge the guidance and
support for this project provided by Chris Weis, Paul Peronard,.and Peter Grevatt (U.S.
EPA).
We wish to gratefully acknowledge the assistance of Nick DeKlerk, FDK Liddell,
JC McDonald, and Terri Schnorr for graciously providing the raw epidemiology data
from several key studies, without which we could not have completed our analysis.
We also wish to thank John Addison, John Dement, and Agnes Kane for their input
and assistance.
Finally, we wish to acknowledge Eric Hack and Tammie Covington for their assistance
with calculations and data management.
i
-------
DISCLAIMER
This report was prepared under contract to the U.S.
Department of Transportation for the U.S. Environmental
Protection Agency. Such support, however, does not signify
that the contents necessarily reflect the views and policies of
either the Department of Transportation or the U.S.
Environmental Protection Agency, nor does the mention of
trade or commercial products constitute endorsement of
recommendations for use.
ii
-------
TABLE OF CONTENTS
SECTION PAGE
1 EXECUTIVE SUMMARY 1-0
2 INTRODUCTION 2-0
3 OVERVIEW 3-0
4 BACKGROUND 4-0
4.1 MINERALOGY OF ASBESTOS 4-3
4.2 MORPHOLOGY OF ASBESTOS DUSTS 4-4
4.3 CAPABILITIES OF ANALYTICAL TECHNIQUES AND METHODS 4-5
USED TO MEASURE ASBESTOS
4.4 LUNG PHYSIOLOGY AND FUNCTION 4-13
5 THE ASBESTOS LITERATURE 5-0
5.1 EPIDEMIOLOGY STUDIES 5-2
5.2 HUMAN PATHOLOGY STUDIES 5-6
5.3 ANIMAL STUDIES 5-9
5.4 IN VITRO STUDIES 5-10
6 HUMAN HEALTH EFFECTS ASSOCIATED WITH OCCUPATIONAL 6-1
EXPOSURE TO ASBESTOS
6.1 APPROACH FOR EVALUATING THE EPIDEMIOLOGY 6-2
LITERATURE
6.2 LUNG CANCER 6^
6.2.1 The Adequacy of the Current EPA Model for Lung Cancer 6-5
6.2.1.1 Time dependence 6-9
6.2.1.2 Comparison with biologically based model 6-19
6.2.2 Estimating KL's from Published Epidemiology Studies 6-30
6.2.3 The Variation in Kt's Derived for Chrysotile Miners and Chrysotile 6-36
Textile Workers
iii
-------
6.2.4 Evaluating Effects Associated with the Character of Exposure 6-45
6.2.4.1 Adjusting KL's for size 6-46
6.2.4.2 Evaluating the implications of adjusting K^'s for size 6-53
6.3 Mesothelioma 6-65
6.3.1 The Adequacy of the Current EPA Model for Mesothelioma 6-66
6.3.1.1 Time dependence 6-70
6.3.1.2 Comparison with biologically based model 6-73
6.3.2 Estimating K^'s from Published Epidemiology Studies 6-79
6.3.3 Evaluating Effects Associated with the Character of Exposure 6-82
6.3.3.1 Adjusting KM's for size 6-83
6.3.3.2 Evaluating the implications of adjusting KM's for size 6-83
6.4 GENERAL CONCLUSIONS FROM HUMAN EPIDEMIOLOGY 6-86
STUDIES
7 SUPPORTING EXPERIMENTAL STUDIES THAT CHARACTERIZE 7-1
THE BIOLOGICAL ACTIVITY OF ASBESTOS
7.1 FACTORS AFFECTING RESPIRABILITY AND DEPOSITION 7-3
7.1.1 Respirability of Spherical Particles 7-4
7.1.2 Respirability of Fibrous Structures 7-7
7.1.3 Effects of Electrostatic Charage on Particle Respirability 7-12
7.1.4 General Conclusions Concerning Particle Respirablility 7-13
7.2 FACTORS AFFECTING DEGRADATION, CLEARANCE AND 7-14
TRANSLOCATION
7.2.1 Animal Retention Studies 7-18
7.2.1.1 Studies involving short-term exposure 7-19
7.2.1.2 Studies involving long-term exposure 7-32
7.2.2 Animal Histopathologicai Studies 7-41
7.2.3 Human Pathology Studies 7-47
iv
-------
7.2.4 Studies of Dissolution/Biodurability 7-53
7.2.5 Dynamic Models 7-58
7.2.6 General Conclusions Concerning Degradation, Clearance, and 7-62
Translocation
7.3 FACTORS GOVERNING CELLULAR AND TISSUE RESPONSES 7-69
7.3.1 The Current Cancer Model 7-73
7.3.2 Evidence for Transformation 7-76
7.3.3 Evidence that Asbestos Acts as a Cancer Initiator 7-77
7.3.3.1 Interference with mitosis 7-78
7.3.3.2 Generation of reactive oxygen species 7-82
7.3.3.3 Generation of reactive nitrogen species 7-92
7.3.3.4 Conclusions concerning asbestos as a cancer initiator 7-96
7.3.4 Evidence that Asbestos Acts as a Cancer Promoter 7-97
7.3.4.1 Asbestos induced proliferation 7-98
7.3.4.2 Asbestos induced cell signaling 7-105
7.3.4.3 Asbestos induced apoptosis 7-119
7.3.4.4 Asbestos induced cytotoxicity 7-121
7.3.4.5 Association between fibrosis and carcinogenicity 7-125
7.3.4.6 Synergism between asbestos and smoking 7-126
7.3.4.7 Conclusions concerning asbestos as a cancer promoter 7-127
7.3.5 Evidence that Asbestos Induces an Inflammatory Response 7-128
7.3.6 Evidence that Asbestos Induces Fibrosis 7-128
7.3.7 Evidence that Asbestos Mediates Epithelial Permeability 7-128
7.3.8 Conclusions Regarding the Mechanisms of Asbestos-Associated 7-129
Diseases
7.4 ANIMAL DOSE RESPONSE 7-131
7.4.1 Injection and Implantation Studies 7-131
v
-------
7.4.2 Inhalation Studies 7-137
7.4.3 Supplemental Inhalation Study 7-140
7.4.4 Conclusions from Animal Dose-Response Studies 7-160
7.5 CONCLUSIONS FROM AN EVALUATION OF SUPPORTING 7-151
STUDIES
8 DISCUSSION, CONCLUSIONS, AND RECOMMENDATIONS 8-1
8.1 DISCUSSION AND CONCLUSIONS 8-1
8.1.1 Addressing Issues 8-1
8.1.2 Comparison with Other, Recent Risk Reviews 8-5
8.2 RECOMMENDATIONS FOR ASSESSING ASBESTOS-RELATED 8-7
RISKS
8.3 RECOMMENDATIONS FOR FURTHER STUDY 8-15
9 REFERENCES 9-1
APPENDIX A: UPDATE OF POTENCY FACTORS FOR LUNG A-1
CANCER (KL) AND MESOTHELIOMA (KM)
APPENDIX B: BACKGROUND FOR DEVELOPMENT OF AN B-1
IMPROVED EXPOSURE INDEX FOR ASBESTOS
APPENDIX C: COMPENDIUM OF MODEL FITS TO ANIMAL C-1
INHALATION DATA IN SUPPORT OF THE BERMAN
ET AL. (1995) STUDY AND POST-STUDY WORK
APPENDIX D: DERIVATION OF LIFETIME RISKS FOR LUNG D-1
CANCER AND MESOTHELIOMA FROM MODELS
USING Kl AND K„ ESTIMATES FOR POTENCY
vi
-------
LIST OF TABLES
Table No.
Title
Page
Table 4-1: Capabilities and Limitations of Analytical Techniques 4-7
Used for Asbestos Measurements
Table 4-2: Comparison of Applicable Methods for Measuring 4-8
Asbestos in Air
Table 6-1: Fit of EPA Lung Cancer Model to Observed Mortality 6-7
Among Wittenoom Miners Categorized by Cumulative
Exposure
Table 6-2: Fit of EPA Lung Cancer Model to Observed Mortality 6-8
Among South Carolina Textile Workers Categorized by
Cumulative Exposure
Table 6-3: Fit of EPA Lung Cancer Model to Observed Mortality 6-11
Among Wittenoom Miners Categorized by Time Since
Last Exposure
Table 6-4: Fit of EPA Lung Cancer Model to Observed Mortality 6-12
Among Wittenoom Miners Categorized by Time Since
Last Exposure with Cumulative Exposure Replaced by
Internal Dose
Table 6-5: Fit of EPA Lung Cancer Model to Observed Mortality 6-15
Among South Carolina Textile Workers Categorized by
Time Since Last Exposure
Table 6-6: Fit of EPA Lung Cancer Model to Observed Mortality 6-18
Among South Carolina Textile Workers Categorized by
Time Since Last Exposure with Cumulative Exposure
Replaced by Internal Dose
Table 6*7: Results of Three Applications of the Two-Stage Model to 6-22
the Lung Cancer Data from Wittenoom (Assumed Half-
life: 20 yrs)
Table 6-8: Results of Three Applications of the Two-Stage Model to 6-23
the Lung Cancer Data from South Carolina (Assumed
Half-life 0.5 yrs)
VII
-------
Table 6-9;
Table 6-10:
Table 6-11:
Table 6-12:
Table 6-13
Table 6-14:
Table 6-15:
Table 6-16:
Table 6-17:
Table 6-18:
Table 6-19:
Table 6-20:
Table 6-21:
Fit of the Best-fitting Two-Stage Model to the Wittenoom 6-25
Lung Cancer Data Categorized by Time Since Last
Exposure {Assumed Half-life: 20 yrs)
Fit of the Best-fitting Two-Stage Model to the South 6-26
Carolina Lung Cancer Data Categorized by Time Since
Last Exposure (Assumed Half-life 0.5 yrs)
Fit of the Best-fitting Two-Stage Model to the Wittenoom 6-28
Lung Cancer Data Categorized by Time Since Last
Exposure (Assumed Half-life: 0.5 yrs)
Lung Cancer Dose-Response Coefficients (Kt's) and 6-31
Mesothelioma Dose-Response Coefficients (KM's)
Derived from Published Epidemiology Studies
Estimated Concentrations of Sized Fibers Observed in 6-42
the Lungs of Thetford Miners and South Carolina Textile
Workers
Correlation Between Published Quantitative 6-48
Epidemiology Studies and Available TEM Fiber Size
Distributions
Representative KL and K„ Values Paired with Averaged 6-51
TEM Fiber Size Distributions From Published Studies
Comparison of Original and Adjusted Kt and K* Values 6-52
Estimated Uncertainty Assigned to Adjustment for Fiber 6-55
Size
Comparison of Spread in Range of Original and Adjusted 6-56
and K„ Values for Specific Fiber Types
Results of an Analysis of the Combined Effects of Fiber 6-59
Size and Type on the Original and Adjusted KL values
Estimated Fraction of Amphiboles in Asbestos Dust 6-60
Comparison of Original and Size Adjusted Potency 6-64
Factors with Simultaneous Consideration of Fiber Type
and Uncertainty
viii
-------
Table 6-22: Fit of EPA Mesothelioma Model to Observed Mortality 6-67
Among Wittenoom Miners Categorized by Cumulative
Exposure (Based on Averaged Exposure Level and
Duration for each Category)
Table 6-23: Fit of EPA Mesothelioma Model to Observed Mortality 6-68
Among Wittenoom Miners Categorized by Cumulative
Exposure (Based on Averaged Value of Integral for Each
Category)
Table 6-24: Fit of EPA Mesothelioma Model to Observed Mortality 6-71
Among Wittenoom Miners Categorized by Time Since
Last Exposure
Table 6-25: Fit of EPA Mesothelioma Model to Observed Mortality 6-72
Among Quebec Miners in Each of Three Mining Areas
and Categorized by Time Since Last Exposure
Table 6-26: Fit of EPA Mesothelioma Model to Observed Mortality 6-76
Among South Carolina Textile Workers by Time Since
Last Exposure
Table 6-27: Results of Several Applications of the Two-Stage Model 6-78
to the Mesothelioma Data from Wiittenoom
Table 6-28: Fit of the Best-fitting Two-Stage Model to the Wittenoom 6-90
Lung Cancer Data Categorized by Time Since Last
Exposure (Assumed Infinite Half-life)
Table 6-29: Optimized Risk Coefficients 6-90
Table 6-30: Conservative Risk Coefficients
Table 7-1: Estimation of Lung Volume and Lung Surface Area 7-12
Loading Rates for Rats and Humans
Table 7-2: Clearance, Translocation, and Degradation Mechanisms 7-15
Operating in the Lung and Their Relative Rates
Table 7-3: Fraction of Fibers Retained Following Chronic Exposure 7-34
Table 7-4: Measured In-Vitro Dissolution for Various Fiber Types 7-56
Table 7-5: Putative Mechanisms by Which Asbestos May Interact 7-71
With Lung Tissue Following Inhalation to Induce Disease
ix
-------
Table 7-6: List of Chemical Species that Potentially Contribute to 7-107
Asbestos-Related Disease, Their Sources and Effects
Table 7-7: Summary Data for Animal Inhalation Experiments 7-142
Conducted by Davis and Coworkers
Table 8-1 Additional Risks per One Hundred Thousand Persons 8-10
From Lifetime, Continuous Exposure to 0.0005 TEM f/cc
Longer than 5.0 |jm and Thinner than 0.5 pm
TABLES FROM APPENDIX A GO HERE
Table B-1: Estimated Tumor Incidence and PCME Dose (with Their B-4
Respective Confidence Bounds) for IP injection Studies
of Six Tremolite Samples of Varying Asbestiform
Character
Table B-2: Estimated Tumor Incidence and Dose Defined as Interim B-8
Index Structures (with Their Respective Confidence
Bounds) for IP Injection Studies of Six Tremolite Samples
of Varying Asbestiform Character
x
-------
LIST OF FIGURES
Figure No. Title
Page
Figure 4-1: The Structure of Lung Parenchyma Showing Alveoli and 4-15
Alveolar Ducts
Figure 4-2: The Structure of Inter-Alveolar Septa 4-16
Figure 4-3: The Detailed Structure of an Alveolar Wall that is Part of 4-17
an Inter-Alveolar Septum
Figure 6-1: Plot of [(RR - a)/(CE10)] vs. Time Since Last Exposure for 6-13
Wittenoom Miners
Figure 6-2: Plot of [(RR - a)/(CE10)] vs. Time Since Last Exposure for 6-16
South Carolina Textile Workers
Figure 6-3: Plot of Estimated Kt Values and Associated Confidence 6-33
Intervals by Study Environment
Figure 6-4: Plot of Adjusted Kt Values (Kt.'s) and Associated 6-54
Confidence Intervals by Study Environment
Figure 6-5: Plot of Estimated Kw Values and Associated Confidence 6-81
Intervals by Study Environment
Figure 6-6: Plot of Adjusted K„ Values (K^'s) and Associated 6-84
Confidence Intervals by Study Environment
Figure 7-1: Fractions of Respirabie Particles Deposited in the 7-6
Various Compartments of the Human Respiratory Tract
as a Function of Aerodynamic Equivalent Diameter
Figure 7-2: Fractions of Respirabie Particles Deposited in the 7-8
Various Compartments of the Human Respiratory Tract
as a Function of the True Diameter of Asbestos Fibers
Figure 7-3: Plot of Chrysotile Lung Burden/Aerosol Concentration 7-36
vs. Time of Exposure
Figure 7-4: Schematic Representation of Fate of Asbestos Fibers in 7-63
the Lung and Surrounding Tissue
XI
-------
Figure 7-5: Fit of Model: Tumor Incidence vs. Structure
Concentration by TEM. (Length Categories: 5-40 pm, >
40 pm, Width Categories: < 0.3 pm and £ 5 Pm)
Figure 7-6: Fit of Model: Tumor Incidence vs. Structure 7-149
Concentration by TEM. (Length Categories: 5-40 pm, >
40 pm, Width: < 0.4 pm)
FIGURES FROM APPENDIX A GO HERE
Figure B-1: Fit of Tumor Incidence vs. PCME Structures for Six B-6
Tremolite Samples of Varying Asbestiform Character
Figure B-2: Fit of Tumor Incidence vs. Interim Index Structures for B-9
Six Tremolite Samples of Varying Asbestifonn Charac er
xii
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
1.0 EXECUTIVE SUMMARY
The purpose of this report is to provide a firm basis for completing a state-of-the-art-
protocol to assess potential human-health risks associated with exposure to asbestos.
Such a protocol is intended specifically for use in performing risk assessments at
Superfund sites, although it may be applicable to a broad range of situations.
The approach currently employed at the U.S. Environmental Protection Agency (U.S.
EPA) to evaluate asbestos-related risks (IRIS 1988) is based primarily on a document
completed in 1986 (U.S. EPA 1986) and has not been changed substantially in the past
15 years, despite substantial improvements in asbestos measurement techniques and
in the understanding of the manner in which asbestos exposure contributes to disease.
Therefore, this document provides an overview and evaluation of the more recent
studies and presents proposed modifications to the protocol for assessing asbestos-
related risks that can be justified based on the more recent work.
The studies relevant to developing a protocol are reviewed in this document and
combined with supporting analysis to resolve issues and identify the best candidate
procedures for assessing asbestos-related risks. Although the objective of this
evaluation was to identify the single best procedures, when current knowledge is
inadequate for distinguishing among alternatives, options are presented along with a
discussion of their relative advantages and limitations. In a few cases, limited and
focused additional research studies are recommended, which may enhance the current
state of knowledge sufficiently to resolve one or more of the important, remaining
issues.
Inhalation of asbestos dusts has been linked to several adverse health effects including
primarily asbestosis, lung cancer, and mesothelioma (U.S. EPA 1986). Asbestosis, a
chronic, degenerative lung disease, has been documented among asbestos workers
from a wide variety of industries. Although asbestosis cases have been observed at
some locations of current interest to the U.S. EPA, the disease is generally expected to
be associated only with the higher levels of exposure commonly found in workplace
settings and is not expected to contribute substantially to potential risks associated with
environmental asbestos exposure. Therefore, asbestosis is only considered in this
document to the extent required to address it's putative association with lung cancer.
Overall, the majority of evidence indicates that lung cancer and mesothelioma are the
most important sources of risk associated with exposure to low levels of asbestos.
A variety of human, animal, and tissue studies have provided insight into the nature of
the relationship between asbestos exposure and disease. Ideally, human epidemiology
1.1
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
studies are employed to determine the quantitative dose/response relationships and the
attendant risk coefficients for asbestos exposure. Risk coefficients have been
estimated for asbestos from approximately 20 epidemiology studies for which adequate
dose-response data exist. Such factors vary widely, however, and the observed
variation has not been reconciled. Among the objectives addressed in this study is to
evaluate and account for the sources of uncertainty that contribute to the variation
among the risk coefficients derived from the literature so that these estimates can be
reasonably interpreted and recommendations for their use in risk assessment
developed.
Animal and tissue studies indicate that asbestos potency is a complex function of
several characteristics of asbestos dusts including fiber size and fiber type (i.e., fiber
mineralogy). Moreover, the influence of fiber size is a complex function of both
diameter and length as critical parameters (among others). Therefore, whenever the
goal is to compare across samples with differing characteristics, it is not sufficient to
report asbestos concentrations simply as a function of mass (or any other single
parameter) and this stands in stark contrast to the treatment of chemical toxins. It has
generally been difficult to distinguish among the effects of fiber size and type in many
studies because such effects are confounded and the materials studied have not been
adequately characterized. However, several adequate studies do exist and these have
been highlighted.
The influence of such effects cannot be adequately evaluated in the existing
epidemiological studies because the analytical techniques used to monitor asbestos
exposure in these studies are not capable of resolving all of the characteristics of
asbestos dusts that other studies indicate are important. Moreover, the exposure
indices (the range of structure sizes and shapes included in an analysis) that are
employed in the existing epidemiology studies may not correspond precisely with the
characteristics of asbestos that best relate to biological activity. This hinders the ability
to compare the risk (dose-response) coefficients derived from the different studies. It
also limits the confidence with which risk coefficients derived from the existing
epidemiology studies can be applied to assess risks from asbestos exposure in other
environments. Such limitations are explored in this study, along with potential
remedies. At the same time, the existing epidemiology studies provide the most
appropriate data from which to determine the relationship between asbestos dose and
response in humans.
Briefly, the major kinds of limitations that potentially contribute to uncertainty in the
available epidemiology studies (and the effect such limitations likely produce on trends
in potency estimates) include:
1.2
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
• limitations in the manner that exposure concentrations were estimated
(likely increases random variation across studies);
• limitations in the manner that the character of exposure (i.e. the
mineralogical types of fibers and the range and distribution of fiber
dimensions) was delineated (likely increases systematic variation between
industry types and, potentially, between fiber types);
• limitations in the accuracy of mortality determinations or incompleteness in
the extent of tracing of cohort members (likely increases random variation
across studies);
• limitations in the adequacy of the match between cohort subjects and the
selected control population (likely increases random variation across
studies and may have a substantial effect on particular studies); and
• inadequate characterization of confounding factors, such as smoking
histories for individual workers (likely increases random variation across
studies and may have a substantial effect on particular studies).
Importantly, this new analysis of the epidemiology database differs from the evaluation
conducted in the 1986 Health Effects Assessment Update (U.S. EPA 1986). Not only
does it incorporate studies containing the latest available followup for the exposure
settings previously evaluated and several additional studies addressing new exposure
settings, but the manner in which the analysis was conducted incorporates important,
new features including estimation of more realistic confidence bounds for the dose-
response factors derived from each study.
Confidence bounds were adjusted to account for uncertainty contributed by the manner
that exposure was estimated, by the manner that work histories were assigned, and by
limitations in the degree of followup, in addition to the traditional practice of accounting
for the statistical uncertainty associated with the observed incidence of disease
mortality. Thus, most of the major contributors to the overall uncertainty of each study
are now addressed, at least in qualitative fashion. By better accounting for overall
uncertainty, we were better able to distinguish what can and cannot be reasonably
concluded from the existing studies.
Separately, we also completed an evaluation to consider questions concerning the
appropriateness of the range of fiber sizes characterized in the published epidemiology
studies and whether adjusted exposure indices might improve the quality of asbestos
1.3
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
risk assessments. This was motivated by conclusions from our review of the
supplemental literature, which indicate that the methods used to assess exposure in the
existing epidemiology studies do not adequately reflect the characteristics of asbestos
that determine biological activity.
Results from our evaluation of the available epidemiology studies indicate that:
(1) the results of individual epidemiology studies are uncertain, especially when one
considers all of the major potential sources of uncertainty (rather than simply
considering the statistical uncertainty associated with the number of deaths, as is
traditionally done). However, more robust conclusions can be drawn from an
analysis of the set of epidemiology studies taken as a whole than results derived
from individual studies;
(2) by adjusting for fiber size and fiber type, the existing database of studies can be
reconciled adequately to reasonably support risk assessment;
(3) the procedures recommended as a result of our analysis offer substantial
improvement over EPA's current approach to evaluating asbestos-related risks;
(4) while there is some indication that the existing EPA models for lung cancer and,
potentially, mesothelioma may not entirely reflect the time-dependence of
disease at long times following cessation of exposure, such effects appear to be
modest so that they are unlikely to adversely affect the proposed approach.
Prudence dictates, however, that limited additional study may be warranted to
adequately dismiss related concerns; and
(5) results from our review of the supplemental literature provide additional support
for the approach recommended in this document and indicate additional, specific
modifications that (if supported by limited additional study) could result in
substantial improvement even over the approach currently recommended, which
(in turn) provides substantial improvement over the current approach.
Based on our analysis, we recommend the following.
(1) Asbestos concentrations should be analyzed for structures that correspond to a
new (interim) exposure index that is defined in this document. As previously
indicated, an exposure index is a description of the sizes and shapes of fibers
(and the relative weights to be assigned to each category of size and shape) that
need to be counted in an analysis to determine asbestos concentrations.
1.4
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
(2) Because amphiboles, fiber-for-fiber, were found to be substantially more potent
than chrysotile, the individual contributions from chrysctile and the combined
amphiboles to any particular exposure need to be separately delineated and,
because amphibole exposures are the primary drivers, the analytical sensitivity
required for each particular study should be set based on amphiboles.
A set of new dose-response coefficients for chrysotile and the amphiboles are also
recommended in this document, which indicate the relative potency of each toward the
induction of lung cancer and mesothelioma, respectively.
Three options are presented in this document for incorporating the new exposure index,
the new dose-response coefficients, and the other recommendations in this document
into a general protocol for assessing asbestos-related risks:
• the first option involves performing lifetable analyses using the dose-
response models that EPA currently recommends for lung cancer and
mesothelioma. When sufficient data are available from a particular project
to support this type of analysis, it should provide the best site-specific
estimates of asbestos-related risks;
• the second option involves use of a risk table, which can be adapted to
estimate risk for most cases of interest, even ones for which exposure
estimates are crude and descriptions of the circumstances of exposure
are sparse; and
• the third option involves development.of a new unit risk factor for asbestos
(based on the latest data presented in this document), which would
replace the unit risk factor in current use.
A couple of focused, additional studies are also recommended in this document, due to
a small number of important knowledge gaps that remain to be resolved, that can be
resolved cost-effectively, and that, if resolved, can substantially increase both the
overall confidence in the proposed approach and, potentially, improve the approach as
well. Importantly, even without conducting the new studies, the approach currently
recommended was shown in this document to provide substantial improvement in the
ability to assess asbestos-related risks (in terms of reducing error) over the approach in
current use. The objectives of the proposed studies are:
1.5
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
to expand the test of the ability of the current EPA models to adequately track
the time-dependence of disease; and
to develop the supporting data needed to define adjustments for potency factors
that will allow them to be used with an exposure index that even more closely
captures the criteria that determine biological activity than the "interim" index
recommended in this document.
1.6 Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
2.0 INTRODUCTION
The purpose of this report is to provide a firm basis for completing a state-of-the-art-
protocol to assess potential human-health risks associated with exposure to asbestos.
Such a protocol is intended specifically for use in performing risk assessments at
Superfund sites, although it may be applicable to a broad range of situations.
The approach currently employed at the U.S. Environmental Protection Agency (U.S.
EPA) to evaluate asbestos-related risks (IRIS 1988) is based primarily on a document
completed in 1986 (U.S. EPA 1986) and has not been changed substantially in the past
15 years, despite substantia! improvements in asbestos measurement techniques and
in the understanding of the manner in which asbestos exposure contributes to disease.
Therefore, this document provides an overview and evaluation of the more recent
studies and presents proposed modifications to the protocol for assessing asbestos-
related risks that can be justified based on the more recent work.
In May, 2001, the U.S. EPA along with the California Environmental Protection Agency
(Cal EPA), the National Institute for Occupational Safety and Health (NIOSH), the
American Toxic Substances Disease Registry (ATSDR), and the Mine Safety and
Health Administration (MSHA) hosted an international conference on asbestos in
Oakland. The current state of knowledge (including, particularly, the status of
knowledge gaps) was reviewed during this conference by many of the leading
researchers. Consequently, several issues were raised concerning the state of
knowledge of asbestos. To the extent that such issues are relevant to the assessment
of asbestos-related risks, they are considered in this document.
Among the asbestos-related issues that have been the focus of recent attention:
• whether the dose-response models currently in use by the EPA for
describing the incidence of asbestos-related diseases adequately reflect
the time-dependence for the development of these diseases;
• whether the relative in-vivo durability of different asbestos mineral types
affect their relative potency;
• whether the set of minerals included in the current definition of asbestos
adequately covers the range of minerals that potentially contribute to
asbestos-related diseases;
2.1
Revision 1 - 6/27/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
• whether the analytical techniques and methods currently used for
determining asbestos concentrations adequately capture the biologically
relevant characteristics of asbestos (particularly with regard to the sizes of
the structures included in the various analyses) so that they can be used
to support risk assessment; and
• whether reasonable confidence can be placed in the cross-study
extrapolation of dose-response relationships that are required to assess
asbestos-related risks in new environments of interest.
These and other asbestos-related issues are described in greater detail in Chapter 3
(Overview) and Chapter 4 (Background).
The studies relevant to developing a protocol are reviewed in this document and
combined with supporting analysis to resolve issues and identify the best candidate
procedures for assessing asbestos-related risks. Although the objective of this
evaluation was to identify the single best procedures, when current knowledge is
inadequate for distinguishing among alternatives, options are presented along with a
discussion of their relative advantages and limitations. In a few cases, limited and
focused additional research studies are recommended, which may enhance the current
state of knowledge sufficiently to resolve one or more of the important, remaining
issues.
This document is an update to a previous review (Berman and Crump 1999) and
primarily incorporates information developed subsequent to the earlier review. It is also
the last in a series of documents developed as part of a multi-task project to develop a
set of mutually consistent methods for determining asbestos concentrations in a
manner useful for assessing risk and a companion protocol for conducting such risk
assessments. A method for the determination of asbestos in air (Chatfield and Berman
1990) and a companion technical background document (Berman and Chatfield 1990)
were published by the U.S. EPA in 1990. The air method has since been superceded
(improved) by the ISO Method (ISO 10312). A method for the determination of
asbestos in soils and bulk materials (Berman and Kolk 1997) was also published by the
U.S. EPA and the draft of an improved version was also recently completed (Berman
and Kolk 2000). The recommendations in this document should serve as the basis for
development of the companion risk-assessment protocol.
2.2
Revision 1 - 6/27/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
3.0 OVERVIEW
Inhalation of asbestos dusts has been linked to several adverse health effects including
primarily asbestosis, lung cancer, and mesothelioma (U.S. EPA 1986). Asbestosis, a
chronic, degenerative lung disease, has been documented among asbestos workers
from a wide variety of industries. Although asbestosis cases have been observed at
some locations of current interest to the U.S. EPA, the disease is generally expected to
be associated only with the higher levels of exposure commonly found In workplace
settings and is not expected to contribute substantially to potential risks associated with
environmental asbestos exposure. Therefore, asbestosis is only considered in this
document to the extent required to address it's putative association with lung cancer.
Overall, the majority of evidence indicates that lung cancer and mesothelioma are the
most important sources of risk associated with exposure to low levels of asbestos.
Note that the primary route of exposure of concern in association with asbestos
is inhalation. There is little evidence that ingestion of asbestos induces disease
(see, for example, U.S. EPA 1986, IRIS 1988). Therefore, this study is focused
on inhalation hazards and other routes of exposure are not addressed.
Gastrointestinal cancers and cancers of other organs (e.g. larynx, kidney, and ovaries)
have also been linked with asbestos exposures (by inhalation) in some studies.
However, such associations are not as compelling as those for the primary health
effects listed above and the potential risks from asbestos exposures associated with
these other cancers are much lower (U.S. EPA 1986). Consequently, this document is
focused on risks associated with the induction of lung cancer and mesothelioma.
A variety of human, animal, and tissue studies have provided insight into the nature of
the relationship between asbestos exposure and disease. Ideally, human epidemiology
studies are employed to determine the quantitative dose/response relationships and the
attendant risk coefficients for asbestos exposure. Risk coefficients have been
estimated for asbestos from approximately 20 epidemiology studies for which adequate
dose-response data exist. Such factors vary widely, however, and the observed
variation has not been reconciled. Among the objectives addressed in this study is to
evaluate and account for the sources of uncertainty that contribute to the variation
among the risk coefficients derived from the literature so that these estimates can be
reasonably interpreted and recommendations for their use in risk assessment
developed.
Animal and tissue studies indicate that asbestos potency is a complex function of
several characteristics of asbestos dusts including fiber size and fiber type (i.e., fiber
3.1
Revision 1 - 6/2W01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
mineralogy). Moreover, Ihe influence of fiber size is a complex function of both
diameter and length as critical parameters (among others). Therefore, whenever the
goal is to compare across samples with differing characteristics, it is not sufficient to
report asbestos concentrations simply as a function of mass (or any other single
parameter) and this stands in stark contrast to the treatment of chemical toxins. It has
generally been difficult to distinguish among the effects of fiber size and type in many
studies because such effects are confounded and the materials studied have not been
adequately characterized. However, several adequate studies do exist and these have
been highlighted.
The influence of such effects cannot be adequately evaluated in the existing
epidemiological studies because the analytical techniques used to monitor asbestos
exposure in these studies are not capable of resolving all of the characteristics of
asbestos dusts that other studies indicate are important. Moreover, the exposure
indices (the range of structure sizes and shapes included in an analysis) that are
employed in the existing epidemiology studies may not correspond precisely with the
characteristics of asbestos that best relate to biological activity. This hinders the ability
to compare the risk (dose-response) coefficients derived from the different studies. It
also limits the confidence with which risk coefficients derived from the existing
epidemiology studies can be applied to assess risks from asbestos exposure in other
environments. Such limitations are explored in this study, along with potential
remedies.
Based on the approach developed for evaluating asbestos-related cancer risk by the
U.S. EPA (1986), risk is estimated as the product of a risk coefficient and a
mathematical function that depends on the level of exposure, the duration of exposure,
and time. The risk coefficient for lung cancer is generally denoted, "KL" and the one for
mesothelioma is UKW."
A detailed description of both the lung cancer and mesothelioma models is provided in
Chapter 6. The models differ depending on whether lung cancer or mesothelioma is
being considered.
For lung cancer, the model estimates relative risk, which means that the increase in
lung cancer incidence that is attributable to asbestos exposure is proportional to the
background lung cancer incidence in the exposed population. The background cancer
incidence is the rate of lung cancer that would be expected to occur in the population in
the absence of asbestos exposure. In other words, background lung cancer incidence
is the lung cancer rate for the exposed population that is attributable to all causes other
than asbestos.
3.2
Revision 1 - 8/2B/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
The model for mesothelioma is an absolute risk model. This means that the increase in
mesothelioma attributable to asbestos is independent of the background rate of
mesothelioma, which is negligible in the general population.
Ideally, the risk coefficients derived from the existing epidemiology studies could be
combined with measurements from other exposure settings (using the corresponding
models) to estimate lung cancer and mesothelioma risks in these other exposure
settings. However, such risk estimates are only valid if both of the following conditions
are met:
(1) asbestos is measured in the exposure setting of interest in the identical
manner in which it was measured in the study from which the
corresponding risk coefficients are derived; and
(2) such measurements reflect the characteristics of asbestos exposures that
determine risk.
A growing body of evidence indicates that the way in which asbestos concentrations
were measured in the existing epidemiology studies do not reflect the characteristics of
asbestos exposure that determine risk. Therefore, measuring asbestos concentrations
in the same way in exposure settings of interest may not be sufficient to assure validity
of risk estimates derived using the published risk coefficients (and the corresponding
models). This is because the second of the above-listed conditions would not be
satisfied.
Considerations necessary to compare risk coefficients derived in different exposure
settings (or to apply a coefficient to predict risk in a setting different from the one in
which the coefficient was derived) have been elucidated clearly in a mathematical
model (Chesson et al 1989). The consequences of the model indicate that adjusting
the existing risk coefficients so that they reflect asbestos characteristics that determine
biological activity requires knowledge of the fiber size distributions of the dusts studied
in the original epidemiology studies. To the extent they exist, such data may be used to
normalize each of the published risk coefficients so that they relate to a common
exposure index reflecting asbestos characteristics that determine biological activity.
Among the goals of this evaluation is to assess the relative importance of such
limitations and to attempt to construct "fixes" as appropriate (within the confines of the
current state of knowledge). Consequently, we explored the possibility of defining an
improved exposure index (that better reflects biological activity) and of applying such an
3.3
Revision 1 - 8/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
improved index to the epidemiology data. We also evaluated improved ways of
simultaneously accounting for the effects of both fiber size and type.
Unfortunately, some of the issues that need to be resolved to support development of a
protocol for assessing asbestos-related risks cannot be entirely resolved with existing
data. We have attempted to identify such issues, to assess their relative importance,
and, when deemed appropriate, to propose limited and focused research projects
designed to provide the data required to reduce the impacts of such knowledge gaps.
3.4
Revision 1 - 8/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
4.0 BACKGROUND
Asbestos is a term used to describe the fibrous habit of a family of hydrated metal
silicate minerals. The most widely accepted definition of asbestos includes the fibrous
habits of six of these minerals (IARC 1977). The most common type of asbestos is
chrysotile, which is the fibrous habit of the mineral serpentine. The other five asbestos
minerals are all amphiboles (i.e. all partially hydrolyzed, magnesium silicates). These
are: fibrous reibeckite (crocidolite), fibrous grunerite (amosite), anthophyllite asbestos,
tremolite asbestos, and actinolite asbestos.
All six of the minerals whose fibrous habits are termed asbestos occur most commonly
in non-fibrous, massive habits. While unique names have been assigned to the
asbestiform varieties of three of the six minerals (noted parenthetically above) to
distinguish them from their massive forms, such nomenclature has not been developed
for anthophyllite, tremolite, or actinolite. Therefore, when discussing these latter three
minerals, it is important to specify whether a massive habit of the mineral or the fibrous
(asbestiform) habit is intended.
Although other minerals may also occur in a fibrous habit, they are not generally
included in the definition of asbestos either because they do not exhibit properties
typically ascribed to asbestos (e.g. high tensile strength, the ability to be woven, heat
stability, and resistence to attack by acid or alkali) or because they do not occur in
sufficient quantities to be exploited commercially.
The first four of the six asbestos minerals listed above have been exploited
commercially (IARC 1977). Of these, chrysotile alone accounts for more than 90% of
the asbestos found in commercial products.
Importantly, it is neither clear whether the term asbestos maps reasonably onto the
range of fibrous minerals that can contribute to asbestos-like health effects nor whether
individual structures of the requisite mineralogy must formally be asbestiform to
contribute to such health effects.
Regarding whether the term asbestos is a useful discriminator for health effects, it is
well established that erionite (a fibrous zeolite not related to asbestos) is a potent
inducer of mesothelioma (Baris et al. 1987), which is one of the two primary asbestos-
induced cancers (see Chapter 3). It is therefore possible that the fibrous habits of at
least some other minerals not formally included in the current definition of asbestos may
contribute to the induction of asbestos-related diseases. Therefore, what needs to be
4.1
Revision 1 - 8/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
developed, is an efficient procedure for separating potentially hazardous materials from
those that are most likely benign.
There are two issues related to the question of whether fibers must formally be
asbestiform to contribute to health effects. The first involves the relationship between
fiber structure and disease induction and the second involves measurement. Although
the evidence is overwhelming that the size and shape of a fiber affects the degree to
which it contributes to the induction of disease (see Chapter 7 and, specifically, Section
7.5), it does not appear that sizes inducing biological activity are well distinguished by
criteria that define the asbestiform habit. For example, determining the concentration of
a specific, common size range of structures was shown to adequately predict the
observed tumorigenicity of six forms of tremolite (three asbestiform varieties and three
non-asbestiform varieties) that were intrapleurally injected into rats (see Section
Appendix B). Once again, therefore, what is needed is an efficient procedure for
distinguishing between the potentially hazardous size range of fibrous structures and
those that do not contribute to disease.
Importantly, depending on the definition employed for the fibers (or fibrous structures)
that are counted during an analysis, it may or may not be necessary to distinguish
formally between asbestiform and non-asbestiform structures for the concentrations
derived from such a count to adequately reflect biological activity.
Note, the dimensions of an asbestiform fiber are determined by the manner in
which the fiber grows (Addison, J. private communication). In contrast, the
massive forms of various minerals, when cleaved, also form elongated particles
(termed "cleavage fragments") and, depending on the definition employed for
fibrous structures during an analysis, such cleavage fragments may or may not
be included along with asbestiform fibers in a count (see, Section 4.3). Although
it is clear from the manner in which they are each formed that the surface
properties of asbestiform fibers and cleavage fragments are likely to be very
different (for example, the latter will have many "unsatisified" chemical bonds),
the degree to which such differences affect potency for comparable sized
structures is not currently known.
Although it is beyond the scope of this document to present a detailed treatise on
asbestos mineralogy, the morphology of asbestos dusts, or the nature and limitations of
analytical techniques and methods used to determine asbestos concentrations, a brief
overview of these topics is presented in the following sections both to identify issues
that need to be addressed as part of the development of an appropriate protocol for
assessing asbestos risks and to provide the background required to facilitate evaluation
4.2
Revision 1 - 8/26/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
of the relevant issues. In that regard, a section on lung physiology and function is also
provided.
4.1 THE MINERALOGY OF ASBESTOS
As previously indicated, the six asbestos minerals can be divided into two general
classes.
Chrysotile is the fibrous habit of the mineral serpentine (Hodgson 1965). The smallest
fibrils of chrysotile occur as rolled sheets or hollow tubules of this magnesium silicate
mineral. The larger fibers of chrysotile form as tightly packed bundles of the unit fibrils.
Chrysotile fibrils typically range from 20 nm to approximately 300 or 400 nm (0.02 to
0.3 or 0.4 um) in diameter. Occasionally, structures up to 1000 nm (1 um) have been
observed, but these are unusual. At some point the curvature induced by the mismatch
between the magnesium and silicon layers of the fibril becomes thermodynamicalJy
unstable, so that production of thicker fibrils is prohibited (Addison, J. private
communication).
Chrysotile bundles are held together primarily by Van der Waals forces and will readily
disaggregate in aqueous solutions containing wetting agents (e.g. soap). They will also
readily disaggregate in lung surfactant (Addision, J. private communication).
The general chemical composition of serpentine is reported as Mg3(Si205)(0H)4
(Hodgson 1965). However, the exact composition in any particular sample may vary
somewhat from the general composition. For example, aluminum may occasionally
replace silicon and iron, nickel, manganese, zinc or cobalt may occasionally replace
magnesium in the crystal lattice of chrysotile (serpentine).
The five other common varieties of asbestos are all fibrous forms of amphibole minerals
(Hodgson 1965). These are ferro-magnesium silicates of the general composition:
A2.3B6(Si,AI)8022(0H)2
where:
A = Mg, Fe, Ca, Na, or K; and
B = Mg, Fe, or Al.
4.3
Revision 1 - 6/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Some of these elements may also be partially substituted by Mn, Cr, Li, Pb, Ti, or Zn.
The fibrous habits of the amphibole minerals tend to occur as extended chains of silica
tetrahedra that are interconnected by bands of cations (Hodgson 1965). Each unit cell
typically contains eight silica tetrahedra and the resulting fibers tend to be rhomboid in
cross-section. Amphibole fibrils are generally thicker than chrysotile fibrils and may
typically range from approximately 100 nm to 700 or 800 nm in diameter (Addison, J
private communication). Substantially thicker fibrils have also been observed.
4.2 MORPHOLOGY OF ASBESTOS DUSTS
Structures comprising the fibrous habits of the asbestos minerals come in a variety of
shapes and sizes. Not only do single, isolated fibers vary in length and thickness but
such fibers may be found combined with other fibers to form bundles (aggregates of
closely packed fibers arranged in parallel) or clusters (aggregates of randomly oriented
fibers) or combined with equant particles to form matrices (asbestos fibers embedded in
non-asbestos materials). Consequently, dusts (even of one mineral variety) are
complex mixtures of structures. For precise definitions of the types of fibrous structures
typically found in asbestos dusts, see Chatfield and Berman (1990).
Detailed descriptions of the characteristics of dusts typically encountered at
environmental and occupational asbestos sites have been reported in the literature and
the following summary is based on a previously published review (Berman and
Chatfield 1990). Typically, the major components of the dust observed in most
environments are non-fibrous, isometric particles. Fibrous structures consistently
represent only a minor fraction of total dust. Asbestos structures represent a subset of
the fibrous structures.
The magnitude of the fraction of total dust represented by fibers and the fraction of
fibers composed of asbestos minerals vary from site to site. However, the fraction of
asbestos in total dusts has been quantified only in a very limited number of
occupational and environmental settings (see, for example, Lynch et al 1970 or Cherrie
etal1987).
The gross features of structure size distributions appear to be similar among asbestos
dusts characterized to date (Berman and Chatfield 1990). The major asbestos fraction
of all such dusts are small structures less than 5 pm in length. Length distributions
generally exhibit a mode (maximum) between 0.8 and 1.5 pm with larger fibers
occurring with decreasing frequency. Fibrous structures longer than 5 pm constitute no
4.4
Revision 1 - 8/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
more than approximately 25% of total asbestos structures in any particular dust and
generally constitute less than 10%.
In some environments, the diameters of asbestos fibers exhibit a narrow distribution
that is largely independent of length. In other environments, diameters appear to
exhibit a narrow distribution about a mean for each specific length. In the latter case,
both the mean and the spread of the diameter distribution increases as the length of the
structures increase. The increase in diameter with length appears to be more
pronounced for chrysotile than for the amphiboles, presumably due to an increase in
the fraction of chrysotile bundles contributing to the overall distribution as length
increases.
Only a few studies have been published that indicate the number of complex structures
in asbestos size distributions. The limited data available indicate that complex
structures may constitute a substantial fraction (up to one third) of total structures, at
least for chrysotile dusts (see, for example, Sebastien et al 1984). Similar results were
also obtained during a re-analysis of dusts generated from the asbestos samples
evaluated in the animal inhalation studies conducted by Davis et al. (Berman et al.
unpublished). This is the same re-analysis used to support a study to identify asbestos
characteristics that promote biological activity (Berman et al. 1995), which is discussed
further in Section 7.4.3.
Historically, fibrous structures have been arbitrarily defined as structures exhibiting
aspect ratios (the ratio of length to width) greater than 3:1 to distinguish them from
isometric particles (Walton, 1982). However, alternate definitions for fibers have also
been proposed, which are believed to better relate to biological activity (see, for
example, Wylie et al. 1993 or Berman et al. 1995). The degree to which fibers are
combined within complex structures in a particular dust may also affect the biological
activity of the dust (Berman et al. 1995). Therefore, proper characterization of
asbestos exposure requires that the relative contributions from each of many
components of exposure be simultaneously considered. Factors that need to be
addressed include the distribution of structure sizes, shapes, and mineralogy in addition,
to the absolute concentration of structures. Such considerations are addressed further
in Chapter 7. Thus, unlike the majority of other chemicals frequently monitored at
hazardous wastes sites, asbestos exposures cannot be adequately characterized by a
single concentration parameter.
4.3 CAPABILITIES OF ANALYTICAL TECHNIQUES USED TO MONITOR
ASBESTOS
4.5
Revision 1 - 8/26/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Due to a complex history, a range of analytical techniques and methods have been
employed to measure asbestos in the various studies conducted over time (Walton
1982). Use of these various methods has affected the comparability of results across
the relevant asbestos studies (Berman and Chatfield 1990). Therefore, the relative
capabilities and limitations of the most important methods used to measure asbestos
are summarized here. Later sections of this report incorporate attempts to reconcile
effects that are attributable to the limitations of the different methods employed in the
various studies evaluated.
Analytical techniques used to measure airborne asbestos concentrations vary greatly in
their ability to fully characterize asbestos exposure. The capabilities and limitations of
four analytical techniques (midget impinger, phase contrast microscopy, scanning
electron microscopy, and transmission electron microscopy) need to be considered
here.
Midget impinger (Ml) and phase contrast microscopy (PCM) are the two analytical
techniques used to derive exposure estimates in the majority of epidemiology studies
from which the existing risk factors are derived. Scanning electron microscopy (SEM) is
an analytical technique that has been employed in several key animal studies.
Transmission electron microscopy (TEM) provokes interest because it is the only
analytical technique that is potentially capable of distinguishing all of the characteristics
of asbestos that potentially affect biological activity.
Although PCM is widely used to characterize occupational exposures, its inability to
distinguish between asbestos and non-asbestos and its lack of sensitivity limits its
usefulness in environmental settings (Berman and Chatfield 1990). In fact, PCM
analyses and TEM analyses showed no correlation among measurements collected
during the cleanup of the Oakland Hills fire (Berman, unpublished). Such lack of
correlation is expected to be observed generally whenever measurements are collected
at sites where no single, obvious source of asbestos is expected to dominate.
Consequently, TEM is the technique that has been recommended for use at Superfund
sites (Chatfield and Berman 1990 and Berman and Kolk 1997).
A general comparison of the relative capabilities and limitations of the analytical
techniques introduced above is presented in Table 4-1.
Importantly, the physical limitations of the various analytical techniques is only part of
the problem. To provide reproducible results that can be compared meaningfully to
other analyses in other studies, one must also consider the choice of procedures
4.6
Revision 1 - B/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
(methods) that address everything from sample collection and preparation to rules for
counting and quantifying asbestos structures.
Multiple methods have been published for use in conjunction with several of the
analytical techniques mentioned above (particularly TEM). Such methods differ in the
procedures incorporated for sample preparation and for the manner in which asbestos
structures are counted. The sample preparation requirements, conditions of analysis,
and structure counting rules for several of the most commonly employed methods are
presented in Table 4-2 to illustrate how the choice of method can result in substantially
different measurements (even on duplicate or split samples).
Table 4-1:
Capabilities and Limitations of Analytical Techniques Used for
Asbestos Measurements1
Parameter
Midget Phase Scanning
Impinger Contrast Electron
Microscopy Microscopy
Transmissio
n Electron
Microscopy
Range of
Magnification
Particles Counted
Minimum Diameter
(size) Visible
Resolve Internal
Structure
Distinguish
Mineralogy4
100
All
1 pm
No
No
400
Fibrous
Structures2
0.3 pm
No
No
2,000-10,000
Fibrous
Structures2
0.1 Mm
Maybe
Yes
5,000 -
20,000
Fibrous
Structures2,3
< 0.01 pm
Yes
Yes
The capabilities and limitations in this table are based primarily on the physical constraints of the
indicated instrumentation. Differences attributable to the associated procedures and practices of
methods in common use over the last 25 years are highlighted in Table 4-2.
Fibrous structures are defined here as particles exhibiting aspect ratios (the ratio of length to
width) greater than 3 (see Walton 1982).
4.7
Revision 1 - 8/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
3. TEM counts frequently resolve individual fibrous structures within larger, complex structures.
Based on internal structure, several different counting rules have been developed for handling
complex structures. See the discussion of methods presented below.
4. Most SEM and TEM instruments are equipped with the capability to record selected area electron
diffraction (SAED) spectra and perform energy dispersive X-ray analysis (EDXA), which are used
to distinguish the mineralogy of structures observed.
The second column of Table 4-2 describes the specifications of the PCM method
currently mandated by the Occupational Safety and Health Agency (OSHA)for
characterizing asbestos exposure in occupational settings. Although this method is in
common use today, several alternate methods for counting fibrous structures by PCM
4.8
Revision 1 - 8/28/01
-------
Mineralogy
Determined
Maximum
Number
Counted
Maximum
Area Scanned
100 fields .
SParsp
Adjustable 1 |*:V. -" O \x'.
• »• *-¦}» r? J.
-------
Table 2 (Page 2)
fro.:':
*"ERA
Count all structures exhibiting L > 1
prn, W < 3.0 pm, and AR > 3. Note
PCME* fraction within count.
Bundles meeting overall
dimensional criteria generally
counted es single fibers
fibers Unless ipfc 10 \
I tndvldinl Wtf^Ma&S
f can be cfisUngtished f -
Ifreproae^.S^gfe;
Cluster*
up ktaO IndrvkJualg.';:•-
fiber ends from (tip to
5) fibers that meet the.
overall dimensional
criteria. Otherwise, ;.v<
count a duster as s ;
Within a duster, count up to 3
individual fibers that meet the
overall dimensional criteria
Otherwise, dusters that contain
more than 3 fibers that meet the
overall dimensional criteria are
counted as single dusters.
Count all strudures with L > 0 5 pm
exhibiting an AR > 5. Record
individual fibers within all groupings
with fewer than 3 Intersections. Count
strudures with L > 5 pm separately
i> (PCME*).
Bundles of 3 or more fibers that meet
the overall dimensional criteria are
counted as single entities and ncrted
as bundles on the count sheet
overall dimensional
Otherwise." dusters lhat container
more th£vi> fibers that meet the;^
A collection of fibers with more than 2
Intersections where at least one
Individual projection meets the overall
dimensional criteria is counted as a
single duster.
For,e
Indude onty bufidlei longer t
Dicingufsh'dspisri^.* , ,
dusters. Count all dusters bonlalhing at
least one component fiber of bunder
satisfying apprbpriala dlmefwte»W(i~ -
criteria. Separately enumerate up to
10 component structure* teDsfylnfl ; »
appropriate dimeruAonal criteria, ¦£"
• v':*'' • "l!
-------
5
6.
7.
e.
National Institute for Occupational Safety and Health (1085). Method for Determination of Asbestos in Air Using Positive Phase Contrast Microscopy NIOSH Method 7400. NtOSH,
Cincinnati, Ohto, U.SA
National Institute for Occupational Safety and Health (1086). Method for Determination of Asbestos in Air Using Transmission Electron Microscopy. NIOSH Method 7402. NIOSH,
Cincinnati. Ohio. U.SA
Yamate. G., Agarwal, S C., and Gibbons, R.D. (1084). Methodology for the Measurement of Airborne Asbestos by Electron Microscopy. U.S. EPA Report No 68-02-3286. U.S.
Environmental Protection Agency, Washington, D.C., U.S.A.
U.S. Environmental Protection Agency (1987). Asbestos Hazard Emergency Response Ad Asbestos-Containing Materials in Schools Final Rule and Notice (Appendix A: AHERA
Method) Federal Register, 40 CFR 763, Vol 52, No 2, pp 41826-41903, October
Chatfield, E.J. (1093). Ambient Air Determination of Asbestos Fibres. Direct Transfer Transmission Electron Microscopy Procedure Submitted to the International Standards
Organization: ISO/TC 146/SC 3. Note: I need to check this number, I think It I* Incorrect
Note that the ISO Method Is a successor to the Interim Superfund Method: Chatfield, E.J. and Berman, D.W. (1090) Superfund Method for the Detennination of Asbestos in Ambient Air
Part 1: Method Interim Version. Prepared for the U.S. Environmental Protection Agency, Office of Emergency and Remedial Response, Washington, D.C. EPA/540-2-90/005a May.
Statistically balanced counting is a procedure incorporated into some asbestos methods (e.g. Ihe Superfund Methods and the ISO Methods) in which long structures
(typically longer than 3 (im) are counted separately during a lower magnification scan than used to count total structures (which are predominantly short). 1 his procedure
assures that the relatively rare longer structures are enumerated with comparable precision to that of the shorter structures.
PCME stands for phase contrast microscope equivalent and indicates the fraction of structures observed by transmission electron micrscopy that would also be visible by phase contrast
microscopy-
-------
PRELJMJNARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 4-2, Page 3
4.11
Revision 1 - 8/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
have also been used historically. Therefore, PCM measurements reported in earlier
studies (including the available epidemiology studies) may not be comparable to PCM
results collected today.
The last four columns of Table 4-2 describe TEM methods that are in current use.
Comparison across these methods indicates:
• the shortest lengths included in counts using these methods vary
between 0.06 and 1 pm. Given that structures shorter than 1 pm
represent a substantial fraction of total asbestos structures in almost any
environment (Section 4.2), this difference alone contributes substantially
to variation in measurement results across methods;
• the definitions and procedures for counting complex structures (i.e.
bundles, clusters, and matrices) vary substantially across methods, which
further contribute to variation in measurement results. For example, the
ISO Method requires that component fibers of clusters and matrices be
counted separately, if they can be readily distinguished. In contrast,
clusters are counted as single structures under the AHERA Method; and
• although all of the methods listed incorporate sample preparation by a
direct transfer process, several of the methods have also been paired with
an optional indirect transfer process. Measurements derived from split
samples that are prepared, respectively, by direct and indirect transfer,
can vary by factors as large as several 100 (Berman and Chatfield 1990).
More typically, however, counts of asbestos structures on samples
prepared by an indirect transfer procedure are greater than those derived
from directly prepared samples by factors of between 5 and 50.
Given the combined effects from the physical limitations of the various techniques
employed to analyze for asbestos and the varying attributes of the methods developed
to guide use of these techniques, the relative capabilities and limitations of asbestos
measurements derived, respectively, from paired methods and techniques in common
use can be summarized as follows:
• all four techniques are particle counting techniques;
• neither Ml nor PCM are capable of distinguishing asbestos from non-
asbestos (i.e. they are incapable of determining structure mineralogy);
4.12
Revision 1 - 8/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
• counting rules used in conjunction with Ml do not distinguish isometric
particles from fibers;
• counting rules used in conjunction with PCM limits counting to fibrous
structures longer than 5 (Jm with aspect ratios greater than 3:1;
• the range of visibility associated with PCM limits counting to fibers thicker
than approximately 0.3 pm;
• under conditions typically employed for asbestos analysis, the range of
visibility associated with SEM limits counting to fibers thicker than
approximately 0.1 pm, which is only marginally better than PCM;
• SEM is capable of distinguishing asbestos structures from non-asbestos
structures;
• TEM is capable of resolving asbestos structures over their entire size
range (down to thicknesses of 0.01 pm);
• TEM is capable of distinguishing the internal components of complex
asbestos structures; and
• TEM is capable of distinguishing asbestos structures from non-asbestos
structures.
More detailed treatments of the similarities and differences between asbestos
techniques and methods can also be found in the literature (see, for example, Berman
and Chatfield 1990).
Due to the differences indicated, measurements from a particular environment (even
from duplicate samples) that are derived using different analytical techniques and
methods can vary substantially and are not comparable. In fact, results can differ by
two or three orders of magnitude (Berman and Chatfield 1990). More importantly,
because the relative distributions of structure sizes and shapes vary from environment
to environment, measurements derived using different analytical techniques and
methods do not even remain proportional from one environment to the next.
Therefore, the results from multiple asbestos studies can only be meaningfully
compared if the effects that are attributable to use of differing analytical techniques and
methods can be quantified and reconciled. To complicate the situation, however, few of
4.13
Revision 1 - 6/28/01
-------
PRELIMINARY WORKING DRAFT- DO NOT COPY OR QUOTE
the existing studies document analytical procedures in sufficient detail to reconstruct
exactly what was done.
4.4 THE STRUCTURE AND FUNCTION OF THE HUMAN LUNG
4.4.1 Lung Structure
The lungs are the organs of the body in which gas exchange occurs to replenish the
supply of oxygen and eliminate carbon dioxide. To reach the gas exchange regions of
the lung, inhaled air (and any associated toxins) must first traverse the proximal
conducting (non-respiratory) airways of the body and the lung including the nose (or
mouth), pharynx, larynx, trachea, and the various branching bronchi of the lungs down
to the smallest (non-respiratory) bronchioles. Air then enters the distal (respiratory)
portion of the tung, where gas exchange occurs.
In humans, the respiratory portion of the lungs are composed of the respiratory or
aveolarized bronchioles, the alveolar ducts, and the alveoli (or alveolar sacs). There
are approximately 3 x 108 alveoli in human lungs (about 20 per alveolar duct) with a
cumulative volume of 3.9 x 10s cm3 (Yeh and Harkema 1993). This represents
approximately 65% of the total volume capacity of human lungs at full inspiration.
Yeh and Harkema also report that the ratio of lung volume to body weight is
approximately constant across a broad range of mammals (from a shrew - 0.007 kg to a
horse - 500 kg). Human lung volumes average a little more than 5 L.
Each human alveolus has a diameter of approximately 0.03 cm (300 |_im) and a length
of 0.025 cm (Yeh and Herkema 1993). The typical path length from the trachea to an
alveolus is approximately 25 cm. The bronchi leading to each alveolus have branched
an average of 16 times from the trachea (with a range of 9 to 22 branches).
Importantly, the mean path length, the number of branches between trachea and
alveolus, and the detailed architecture (branching pattern) of the respiratory region of
the lung vary across mammalian species. For example, rats and mice lack respiratory
bronchioles while macaque monkeys exhibit similar numbers of respiratory bronchiole
generations as humans (Nikula et al. 1997). Furthermore, branching in humans tends
to be symmetric (each daughter branch being approximately the same size and the
angle of branching for each is similar but not quite equal) while rodents tend to exhibit
4.14
Revision 1 - 8/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
monopodal branching in which smaller branches tend to come off at angles from a main
trunk (Lippmann and Schlesinger 1984).
The gas exchange regions of the lung are contained within the lung parenchyma, which
constitute approximately 82% of the total volume of the lungs (Gehr et al. 1993).
Importantly, the lung parenchyma is not a "portion" of the lung; it fills virtually the entire
volume of the organ traditionally visualized as the "whole" lung. Embedded within the
parenchyma are the larger conducting airways of the lungs and the conducting blood
vessels that transport blood to and from the capillaries that are associated with each
alveolus. In the human lung, approximately 213 ml of blood are distributed over 143 m2
of gas-exchange (alveolar) surface area (about the size of a tennis court). The gas-
exchange surface area of lungs scale linearly with body weight over most mammalian
species. The slope of the line is 0.95. Figure 4-1 is a photomicrograph showing two
views of a portion of lung parenchyma in the vicinity of a terminal bronchiole and an
avleclar duct, which are labeled. The circular spaces in the left portion of the figure and
the cavities in the right portion of the figure are alveoli. Note the thinness of the walls
(septa) separating alveoli.
Alveoli are separated from each other by alveolar septa that average only a few
micrometers in thickness (Gehr et al. 1993). Gas-exchange capillaries run within these
septa and the air-blood barrier, which averages only 0.62 pm in thickness, is composed
of three layers: the alveolar epithelium, the interstitium, and endothelium. The
epithelium is described in detail below. The interstitium is primarily composed of a
collagenous, extracellular matrix, which constitute about two-thirds of the volume with a
collection of fibroblasts, macrophages, and other cells interspersed within the matrix
(Miller et al. 1993). The cells of the interstitium constitute about one third of its volume.
The endothelial layer is composed of the smooth muscle cells and connective tissue
that constitute the walls of vascular capillaries. The amount of connective tissue in the
septa between alveoli also varies between animal species; small rodents have less and
primates more (Nikula et al. 1997).
Figures 4-2 and 4-3 show, respectively, a typical alveolar septum and a closeup of one
portion of such a septum. In Figure 4-2, one can see that the septa themselves are thin
and are filled almost entirely with capillaries. Figure 4-3 shows that the epithelial lining
of an alveolus is no more than 1 pm thick, that the underlying interstitium is no thicker,
and that the endothelium of the adjacent capillary is similarly thin. These three layers
constitute the major tissues of the air-blood barrier (the rest of the barrier includes the
limited quantity of blood plasma between the endothelial wall of a capillary and a red
blood cell and the outer membrane of the red blood cell itself).
4.15
Revision 1 - 8/28/01
-------
PRELIMINARY WORKING DRAFT- DO NOT COPY OR QUOTE
Gehr et al. (1993) also report thai interalveolar septa constitute approximately 14% of
the volume of the gas-exchange region of the lung (i.e. the lung parenchyma). Of this
tissue mass, approximately 20% is endothelium, 55% is interstitium, and the rest is
composed of cells associated with the alveolar epithelium. The remainder of the gas-
exchange region is air space. Gehr et al. also report that the interalveolar septa fold to
accommodate changes in lung volume during respiration, although the major
contribution to lung volume changes (over the range of normal inspiration) appears to
be the collapsing (folding) of alveolar ducts (Mercer and Crapo 1993).
4.16
Revision 1 - 8/28/01
-------
Figure 4-1:
The Structure of Lung Parenchyma
Showing Alveoli and Alveolar Ducts
:V5 ?s
-S
i
.iAu
* \sZ\xl
V
- (
* J
V- -
A
TB
V.
r
A
\.
•m*
f
>— (
" ' J
T
s.
V
Confidential:
iefi is> repro.^uc^ this figyggpm
i; ^ - * T. 4.^1
' "wi ¦ ^
TB
FIG. 10 lW snt $C'¦/: comparison of the cenlnacina? feaicr witn iwn different organization
and E f rp
-------
Figure 4-2:
The Structure of the Inter-Alveolar Septa
FIG. 10. Transmission electron micrograph ot an alveolus Irom a dog lung fixed by intravascular
oerlusion. The lung was inflated with air ai a pressure ot 5 cm of water on Ihe deflation limb of the
iast of three hysteresis cycles (approximately 60% TLC). It shows interalveolar septa that are
folded at this degree of inflation The surface lining layer (SLL) smoothes out every depression of
'.he alveolar surlace (arrows). A. alveoli: C. capillary. > 2.100. Inset: Hign power view of an
epithelial depression filled with a fluid surlace lining layer (SLL) • 14.600
Confidential:
Need permission to reproduce this figure
-------
Figure 4-/:
The Detailed Structure of an Alveolar Wall
that is Part of an Inter-Alveolar Septum
FIG. 19. f-actr-r- c $nu.v«. y ~' ncimsosed on alveolar capillary to measure inter-
:e:' r -ne calculation of the harmonic mean barrier thick-
ness ?;oir me tf-vT'ocv::- • ,r-.c that the barner separating blood Irom atveo-
*':_~a« .')• »v. '-if.. 'EPi. endothelium (EN), and interstitium (IN).
Confidential:
Need permission to reproduce this figure
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Figure 4-3
4.19
Revision 1 - 8/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
According to Gehr et al. (1993), human alveoli contain approximately 380 cells and the
total number reportedly increases as a power function with body mass for other
mammals. In humans, this includes approximately 12 alveolar macrophages (although
this number is larger in smokers, apparently due to chronic inflammation), 40 Type I
epithelial cells, 67 Type II epithelial cells, 106 interstitial cells, and 148 endothelial cells.
The relative number of the various types of cells associated with alveoli varies across
mammalian species (Gehr et al. 1993). In the rat, for example, Type I epithelium is
approximately 20% of the total (rather than 12% in humans) with less interstitium and
more endothelium.
According to Gehr et al. (1993), Type I epithelial cells, which constitute approximately
95% of the surface area of alveolar epithelium, are flat and platy (squamous), and
average less than 1 pm in thickness (except where cell nuclei exist, which are
approximately 7.5 pm in diameter and protrude into the alveolar space). Type II
epithelial cells, which are cuboidal, constitute no more than 5% of the epithelial surface.
Despite the small fraction of surface that they occupy, Type II cells serve to maintain
the integrity of the overall epithelial lining so that, for example, they limit the tissue's
permeability and control/prevent transport of macromolecules from the interstitium to
the alveolar space, or the reverse (Leikauf and Driscoll 1993). Type II cells also secrete
lung surfactant. The basement membrane of alveolar epithelium is a collagenous
structure.
In contrast, the epithelial cells lining the trachea and bronchi (including the respiratory
bronchioles) are ciliated and columnar and averages between 10 and 15 pm in
thickness (based on photomicrographs presented in St. George et al. 1993).
Tracheobronchial epithelium reportedly contains at least 8 distinct cell types (St.
George et al. 1993): ciliated epithelium, basal cells (small flat cells situated along the
basal lamina and not reaching the luminal surface), mucous goblet cells, serous cells,
nonciliated bronchiolar (clara) cells, small mucous granule (SMG) cells, brush cells, and
neuroendocrine cells. The relative abundance of the various cells varies across
mammalian species as well as across the various airway generations (branches) and
even the opposing sides of specific airways. The number of cells per length of basal
lamina also varies across mammalian species.
4.4.2 The Structure of the Mesothelium
The mesothelium is a double layered membrane and each layer is a single-cell thick.
The two layers of the mesothelium are separated by a space (e.g. the pleural space),
which contains extracellular fluid and free macrophages (Kane and McDonald 1993).
4.20
Revision 1 - 8/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
The pleural space is drained at fixed locations by lymphatic ducts. Each layer of the
mesothelium overlies a collagenous basement membrane containing dispersed spindle
cells. Depending on the location of the mesothelium, the basement membrane may
overlie the skeletal muscle of the diaphragm or the rib cage (in the case of the parietal
pleura, which is the outer layer). For the inner layer or visceral pleura, the basement
membrane overlies visceral organs (including the lungs) within the rib cage. Healthy
mesothelium is quiescent, meaning that cells are not actively dividing.
The relative size and thickness of mesothelial tissue varies across mammalian species
(Nikula et al. 1997). For example, rats have relatively thin pleura with limited lymphatic
ducts. In contrast, nonhuman primates have thicker pleura with greater lymphatic
drainage than rats and humans have even thicker pleura and relatively abundant
lymphatic ducts.
4.4.3 Cytology
Alveolar Epithelium. In the respiratory region of the lung, Type II epithelial cells are
progenitor cells for Type I epithelium (Leikauf and Driscoll 1993). Type I epithelial cells
are not proliferation competent (Nehls et al. 1997). After injury to Type I cells, Type II
cells proliferate and reestablish the continuous epithelial surface. Type II cells also
secrete surfactant. It appears that the identity and location of the progenitor cells for
Type II epithelial cells are not currently known.
Injury or alteration of Type II cell function are associated with several diseases included
idiopathic pulmonary fibrosis (Leikauf and Driscoll 1993). Also, crystalline silica and
other toxic agents have been shown to directly modify Type II cellular activity. For
example, crystalline silica (at sub-cytotoxic levels, <100 pg/ml) stimulates Type II
proliferation in tissue culture. In contrast, neither titanium dioxide nor aluminum coated
silica induce proliferation at corresponding concentrations.
In culture, Type II epithelial cells transform slowly into Type I cells and thus have limited
population doubling capacity (Leikauf and Driscoll 1993). Rarely can the number of
Type II cells expand past three to 10 passages (20-30 doublings). During this time,
cells terminally differentiate, develop cross-linked envelopes, and appear squamous,
enlarged, and multinucleated. The process is noted to be accelerated by the presence
of transforming growith factor beta (TGF-p).
Macrophages. Alveolar macrophages (the largest population of macrophages in the
lung) are mobile, avidly phagocytic, present antigens, and release cytokines that trigger
various other immune responses (Leikauf and Driscoll 1993). They also initiate
4.21
Revision 1 - 8/2B/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
inflammatory responses and other repair mechanisms that are designed to restore
tissue homeostasis.
The next largest population of macrophages in the lung are the interstitial macrophages
(Leikauf and Driscoll 1993). These are localized in the peribronchial and perivascular
spaces, the interstitial spaces of the lung parenchyma, the lymphatic channels, and the
visceral pleura. The various populations of macrophages in the lung express different
surface proteins, show different proliferative capacity, and show differences In
metabolism.
The sizes of alveolar macrophages varies substantially across mammalian species
(Krombach et al. 1997). Krombach and co-workers provide a table summary of the
relative sizes across several species of interest:
Animal Mean diameter (um) Mean Volume (um3)
Rats 13.1 +/-0.2 1166+/-42
Syrian Golden Hamsters 13.6+/-0.4 1328+/-123
Cyanomolgus Monkeys 15.3+/-0.5
Healthy Humans 21.2 +/-0.3 4990+/-174
As indicated later (Section 7.2), the relative size of various macrophages has direct
implications regarding the dependence of clearance mechanisms on fiber size.
Mesothelium. Importantly, mesothelial cells are proliferation competent and may be
their own progenitor cells (Kane and McDonald 1993). It is also possible, however, that
as yet to be identified progenitor cells are located along opposite walls of pleura or at
other locations within the pleural space.
Tracheo-bronchiolar epithelium. As previously indicated, tracheo-bronchiolar (i.e.
non-respiratory) epithelium is composed primarily of ciliated, columnar cells that are 10
to 15 pm thick. Although some report that Clara cells serve as progenitor cells for
tracheo-bronchiolar epithelium (Finkelstein et al. 1997), others report that both Clara
cells and bronchiolar epithelium are proliferation competent (Nehls et al. 1997). It
appears that the identity and location of the progenitor cells for Clara cells are not
currently known.
4.4.4 Implications
The potential implications of the above observations concerning lung structure arid
cytology are:
4.22 Revision 1 - B/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
• that the thicknesses of Type I epithelial cells, endothelial cells, and the
interstitium in the alveolar septa are all very small relative to the lengths of
the putative asbestos fibers that contribute to disease;
• that an entire alveolus is only 300 pm across and a typical Type I cell is
46 pm in radius by < 1 pm thick;
• that distances across alveolar septa are only on the order of a few pm
and such septa contain both the endothelium and the interstitum. Thus,
the distances that have to be traversed to get to these structures are also
small relative even to the length of a fiber;
• that the alveolar septa and the walls of the alveolar ducts fold during
respiration, which may provide mechanical forces that facilitate movement
of fibers into and through the alveolar epithelium;
• that Type I epithelium do not proliferate so they cannot be the cells that
lead to cancer. It is the Type II epithelial cells (and potentially
macrophages, basal cells, or endothelial cells) that contribute to cancer in
the pulmonary portion of the lung. Type II cells eventually terminally
differentiate to Type I cells;
• that other cells in the lung that have variously been reported to be
proliferation competent (so that they potentially serve as target cells for
the induction of cancer) include Clara cells and bronchiolar epithelial cells;
• that the distance from the most distal airways and aiveoli to the pleura is
small relative to the lengths of a fiber; and
• that mesothelial cells are proliferation competent and thus serve as
potential targets for the induction of cancer;
4.23
Revision 1 - 8/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
5.0 THE ASBESTOS LITERATURE
This is a description of the common types of studies in the asbestos literature and an
overview of the sources of potential uncertainty typically associated with each. Such
limitations must be considered when drawing conclusions from these studies and, more
importantly, when comparing inferences across studies. Throughout this document, we
have endeavored to identify the major sources of uncertainty in the studies we
examined and have endeavored to account for such uncertainties during our evaluation
and interpretation of study results.
The types of studies available for examining relationships between risk and asbestos
exposure include human epidemiology studies, human pathology studies, a broad
variety of animal studies, and a broad variety of in-vitro studies in both tissue cultures
and cell-free systems. To properly compare and contrast the results from such studies:
• the method(s) employed for asbestos characterization in each study need
to be reconciled;
• the procedures employed for evaluating study end points need to be
compared and contrasted;
• the relationship between the route of exposure employed in each type of
study and the exposure route of interest (i.e. human inhalation) needs to
be examined; and
• other major, study-specific sources of uncertainty need to identified and
addressed.
Among study conditions and procedures that must be considered before evaluating
study conclusions, it is particularly important to address the analytical methodologies
employed to characterize the nature of exposures (or doses) in each study and such
considerations are common to virtually all of the various types of studies of interest.
As indicated in Section 4.3, the only instrument capable of completely delineating
asbestos structure size distributions is TEM (or TEM combined with other techniques).
Thus, for example, conclusions regarding variations in biological effects due to
differences in such things as fiber size must be viewed with caution when fiber sizes are
characterized using only PCM, SEM, or other, cruder analytical techniques. Similarly,
the ability to adequately determine fiber mineralogy (fiber type), particularly of what may
be minor constituents of various dusts or bulk materials, also depends strongly on the
5.1
Revision 1 - 8/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
instrumentation employed for analysis as well as the strategy for sampling and for
conducting the actual structure count. All of these factors must be considered.
Not only does the specific instrumentation (analytical technique) that is employed in an
asbestos measurement affect the outcome of that measurement, but the particular
method employed to guide the measurement affects the outcome. As previously
indicated (Section 4.3), details concerning the definition of structures to be included in a
count, the strategy for counting, and the minimum number of specific types of structures
to be included in a count (all features that vary across analytical methods) affect the
precision with which fiber concentrations (particularly of longer and thinner fibers) are
delineated. It is not uncommon, for example, that asbestos concentrations measured in
the same sample may vary by several orders of magnitude, due simply to a difference
in the analytical method employed for the analysis (even when the same analytical
technique is employed, see Section 4.3).
Other important sources of uncertainty tend to be study-type specific and are thus
addressed separately below.
5.1 HUMAN EPIDEMIOLOGY STUDIES
A good overview of the kinds of limitations that contribute to uncertainty in the available
epidemiology studies was presented in the Health Effects Assessment Update (U.S.
EPA 1986). As described in Appendix A of this document, while evaluating dose-
response factors derived from the human epidemiology studies, we tried to address
most of the major sources of uncertainty commonly associated with such studies, which
are described briefly below.
Epidemiology studies, which track the incidence of disease (or mortality) within a
defined group (cohort) sharing comparable exposures, have been performed on cohorts
of workers exposed to asbestos and other mineral fibers in a variety of occupational
and environmental settings. Among these, studies that include quantification of
exposures are particularly useful for evaluating dose/response relationships and
deriving risk factors.
Generally, the most severe limitations in an epidemiology study involve the exposure
data. Both estimates of the level of exposure and determination of the character of
exposure are affected by such limitations. Regarding the character of exposure,
because exposure measurements from most of the available quantitative epidemiology
studies are based on midget impinger measurements or PCM measurements, detailed
characterization of the size distribution or the mineral type of fibrous structures
5.2
Revision 1 - 8/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
(particularly of minor constituents) that contributed to exposure in such studies is
generally lacking (Section 4.3). This is particularly important because of the evidence
that neither midget impinger or PCM are capable of providing measurements that
remain proportional (across study environments) to the biologically-relevant
characteristics of an asbestos dust (see Section 7,5 and Appendix B). This limits the
ability both to compare results across the existing studies and to extrapolate such
results to new environments for which risks need to be estimated. At the same time,
effects on the ability to observe dose-response trends within a single study are not
typically impaired.
Samples collected prior to the mid-1960s were often analyzed by measuring total dust
in units of millions of particles per cubic foot (MPCF) using impingers or thermal
precipitators. A description of the relative strengths and weaknesses of these
techniques is provided in Section 4.3. The fibrous portion of the dust was not
monitored. Impinger measurements are sometimes related to fiber counts (based on
PCM) using side-by-side measurements of total dust and fiber counts collected during a
relatively brief period of time (e.g., McDonald et al, 1980b, Dement et al, 1983a).
However, the correlation between fiber counts and total dust is sometimes poor within a
plant and generally poor between plants (see, for example, U.S. EPA 1986). Thus,
conversions based on limited sets of paired measurements are of questionable validity.
In some studies (e.g., McDonald et al, 1983b) the only available measurements are
impinger measurements in MPCF and these have been related to f/ml by PCM using
conversion factors derived in other plants, which raises further questions concerning
validity.
Even conversion to PCM may not be adequate for assessing risk in a manner that
allows extrapolation across exposure environments or studies. As indicated in
Section 7.5, comparing dose-response factors derived in different exposure
environments (or extrapolating to new environments to predict risk) requires that
asbestos measurements reflect the characteristics of asbestos structures (size, shape,
mineralogy) that determine biological activity. If surrogate measures are employed
(e.g. measures of asbestos structures displaying characteristics other than those that
determine biological activity), there is no guarantee that concentrations of such
surrogate measures and the true biologically active structures will remain proportional
from one environment to the next. As a consequence, the relationship between
exposure (measured by surrogate) and risk may not remain constant from one
environment to the next. Importantly, several studies suggest that PCM may, at best,
be no more than a surrogate measure (see, for example, Berman et al. 1995).
Moreover, the technique was adapted to asbestos in an ad hoc fashion with only limited
thought given to biological relevance (Walton 1982).
5.3
Revision 1 - 8/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Use of surrogate measures of asbestos exposure may be less of a problem within a
single exposure environment (where airborne asbestos structures likely have been
generated in a similar manner from similar source material). Thus, surrogate measures
of asbestos exposure may remain approximately proportional to the true biologically
active structures, which suggests why monotonically increasing dose-response
relationships have likely been observed with PCM-measured concentrations in single
exposure environments. In different exposure environments, however, airborne
asbestos structures likely are generated in different processes from different source
material. There is thus little reason to expect surrogate measures of exposure to
remain proportional across such environments.
None of the published epidemiology studies incorporate TEM measurements of
asbestos and such measurements are not widely available in occupational settings
(Section 6.2.4). However, TEM is the method currently used (and recommended) to
assess exposure in environmental settings (due both to questions concerning biological
relevance (Section 7.5) and to problems with measuring environmental asbestos
concentrations by PCM (Section 4.3 and Appendix B).
In some cases, the limited exposure characterization presented in specific epidemiology
studies can be augmented by pairing such studies with published TEM
characterizations of dusts from the same or similar exposure settings, to the extent the
appropriate supplemental studies are available. In fact, this is the procedure adopted in
this document to adjust the existing risk factors to exposure indices that are thought to
better relate to biological activity (Sections 6.2.4 and 6.3.3). Such an approach is
limited, however, to the extent that the published asbestos characterizations actually
represent exposure conditions in the corresponding epidemiology studies. To the
extent reasonable, the limitations of this approach have been addressed in this study by
assigning and incorporating additional factors into the calculation of the confidence
intervals associated with the adjusted potency factors (Sections 6.2.4 and 6.3.3).
Regarding levels of exposure in the epidemiology studies, in most cases, air
measurements were collected only at limited points in time and measurements may be
entirely lacking from the earliest time periods, when exposures may have been
heaviest. In such cases, exposures are typically estimated either by extrapolation from
periods when measurements are available or by expert judgement based on personal
accounts and records of changes in plant operations, industrial hygiene procedures, air
standards, etc. Moreover, the majority of exposure measurements used in these
studies are based on area (ambient) rather than personal samples. Typically, only a few
areas of a plant have been sampled so that levels in other areas must be approximated
using expert judgement by persons familiar with operations at the plant.
5.4
Revision 1 - 8/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
It is difficult to judge the degree that available asbestos concentration measurements
are representative of actual exposures in the existing studies. In some cases, it seems
likely that operations were shut down or otherwise modified in preparation for sampling.
Likewise, in some operations there are brief episodes of very intense exposure and it is
questionable whether such episodes are adequately represented in the available data.
Most of the asbestos measurements used in the published epidemiology studies were
collected for insurance or compliance purposes. They were not intended to provide
representative estimates of the direct level of exposure to workers. Some of the
published epidemiology studies lack any direct exposure data. For example, exposures
were estimated for the cohort studied by Seidman (1984) based on conditions
simulated many years later in a similar plant to the one from which Seidman studied the
original cohort. In fact, the equipment in the plant from which Seidman obtained
exposure estimates came originally from the plant where Seidman studied the workers;
it was purchased and moved. Recently, an epidemiology study was also completed for
a cohort working at that new plant (Levin et al. 1998).
In addition to problems with the actual analysis of asbestos concentrations, individual
exposures are generally estimated in the existing epidemiology studies by relating
ambient asbestos measurements to job descriptions and integrating the duration of
exposure over the recorded time that each worker spent in each job category.
However, sometimes there are no records of specific areas in which an employee
worked, so that work areas must be assumed based on job title. Some types of
workers (e.g., maintenance workers) may have spent time in many different areas of a
plant so their exposure varies from what might otherwise be assumed.
Although the greatest problems with the data in existing epidemiology studies likely lies
within the estimates of exposure, problems with disease-response data also exist.
Mesothelioma is rare and this disease may have been under-reported as a cause of
death in older studies. This is probably less of a problem in more recent studies, since
the association of mesothelioma with asbestos exposure is now well known. In fact, the
opposite tendency (over-reporting) may now be occurring because an asbestos worker
with mesothelioma is probably eligible for compensation. Some studies have
re-diagnosed causes of death from all of the available data (e.g., Selikoff et al, 1979);
however, this creates the problem of lack of comparability to control populations (for
which such re-diagnosis is not generally performed).
The choice of an appropriate control population is also an important consideration.
Local cancer rates may differ substantially from regional or national rates and the
choice of an appropriate control is not always clear. A related problem is the lack of
5.5
Revision 1 - 8/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
smoking data in many of the studies. Because of the interrelation between smoking
and asbestos in lung cancer, errors could occur in lung cancer risk estimates if the
smoking patterns of the cohort are substantially different from those of the control
population.
In some of the studies, a substantial portion of the population is lost to follow-up (e.g.,
Armstrong 1988), and this adds additional uncertainty to the analysis. Also, the effect
of exposure may be inaccurately evaluated if the follow-up of the population is too brief.
This may be a limitation, for example, of the Levin et al. (1998) study.
Another problem frequently associated with these studies is that available data are not
reported in a form that is well-suited to risk assessment. The EPA lung cancer model,
for example, requires that exposure be estimated as cumulative exposure in f/ml-years
excluding the most recent 10 years (Section 6.2); generally the data are not published
in this form. The data are also frequently not available in a form that permits study of
the shape of the lung cancer dose/response curve, so it is not possible to determine
how well the EPA model describes the data. The form of the data for mesothelioma is
generally even less appropriate for risk assessment. Ideally, what is needed is the
incidence of mesothelioma subdivided according to exposure level, age at beginning of
exposure, and duration of exposure (Section 6.3). Such data are almost never
available in published studies and crude approximations must be made to account for
this lack.
It is important to appreciate the type and magnitude of effect that each of these sources
of uncertainties are likely to have on the distribution of potency estimates derived from
the set of available studies for lung cancer and mesothelioma, respectively. Some of
the above-described limitations likely introduce random errors that simply decrease the
overall precision of a potency estimate. However, other types of limitations may cause
systematic errors in particular studies, which potentially biases the potency estimate
either high or low. Some of the limitations may only affect between-study differences
and some may introduce a systematic bias between either industry types or fiber types.
Examples of some of these types of variation are provided in Section 6.1.
We also note that, although individual estimates of potency factors from individual
studies may be highly uncertain, by combining results across multiple studies while
properly addressing such uncertainties, it may be possible to draw conclusions with
greater precision than reasonable for any individual study. This is the essential
advantage of the type of meta analysis discussed in Section 6.2.4 for lung cancer and
6.3.3 for mesothelioma.
5.6
Revision 1 - 8/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
5.2 HUMAN PATHOLOGY STUDIES
Human pathology studies provide a characterization of disease morphology and
correlations between causes of death and the types of asbestos fibers retained in the
lungs and other bodily tissues. These studies generally involve microscopic
examination of tissue samples for indications of morphologic changes characteristic of
disease and/or microscopic examination of digested tissue specimens to characterize
the mineral fibers extracted from such tissue.
The results of human pathology studies need to be evaluated carefully by addressing
effects that are attributable to:
• the way that tissue samples are fixed for preservation;
• the way that tissue samples are prepared for analysis (e.g. ashing, bleach
digestion, digestion in alkali, or some combination);
• the choice of methods employed for characterization of asbestos; and
• the choice of locations within tissues from which samples are collected for
analysis.
Because tissue samples obtained from deceased individuals are typically stored for
long periods of time before they may be analyzed as part of a human pathology study,
such samples are commonly fixed by treatment with chemical preservatives prior to
storage. However, Law et al. (1991) studied the effects of two common fixatives
(Karnovsky's fixative and formalin fixative) on asbestos fibers and concluded that such
fixatives degrade and dissolve asbestos fibers (including both chrysotile and crocidolite)
at measurable rates (Section 7.2.4). Therefore, particularly for samples that are stored
"wet" (as opposed, for example, to storage in paraffin blocks), the concentrations and
character of the tissue burden of asbestos may be altered during storage. Even for
studies in which relative (as opposed) to absolute concentrations are being compared,
alterations associated with preservation may limit the ability to make such comparisons,
particularly among samples stored for widely disparate periods of time or stored using
widely disparate procedures.
Fiber concentrations in tissue samples have also been shown to vary as a function of
the method employed for preparing such samples for analysis. Historically, samples
that are digested in bleach or alkali have tended to exhibit lower recovery of asbestos
fibers than samples that are ashed (REF). However, more recent studies suggest that
5.7
Revision 1 - 8/26/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
improving technique has narrowed these differences so that this is no longer a major
consideration (REF). Thus, when comparing results across studies, due consideration
needs to be given for the time frame during which such studies were conducted and the
comparability/differences in the techniques employed for tissue sample preparation.
Once prepared, both the character and the concentration of the tissue burden
measured in a tissue sample will also depend heavily on the particular analytical
method employed to characterize asbestos and differences attributable to such
techniques must be reconciled before measurements across samples or conclusions
across studies can be reasonably compared. A more detailed description of the effects
attributable to asbestos measurement was presented in the previous section on human
epidemiology studies (Section 5.1) and the same issues obtain for human pathology
studies. Unless measurements are made using comparable instrumentation with
comparable methodology, comparisons across such measurements can be very
misleading.
Perhaps the biggest limitation hindering the kinds of evaluation that can be conducted
based on human pathology studies is that due to the strong dependence of asbestos
concentrations on the specific location within a tissue from which a sample is obtained.
Numerous authors have reported that asbestos is non-uniformly distributed in lung
parenchyma and other tissues following exposure (see, for example: Davis et al. 1986a,
Bignon et al. 1979, or Pooley 1982). The incidence of lesions and other pathological
effects attributed to asbestos exposure correspondingly exhibit a non-uniform
distribution.
For lung tissue samples (which tend to be among the primary interests in human
pathology studies) the relationship between sample location and asbestos
concentration is particularly important. To sample deep lung tissue reproducibly, it has
been shown necessary to select a specific section of lung parenchyma from a defined
portion of the bronchio-alveolar tree. Pinkerton et al. (1986) showed that the deposition
of asbestos in the lungs is an inverse function both of the pathlength and the number of
bifurcations between the trachea and the site. Thus, analyses of samples from different
animals of the same species can only be compared meaningfully if the samples are
collected from identical locations in the bronchio-alveolar tree. Similar, nonuniform
depositional patterns have also been observed in humans (Raabe 1984). Futhermore,
due to the complex branching and folding pattern of the lung, adjacent sections of lung
parenchyma frequently represent disparate portions of the bronchio-alveolar tree (REF).
Consequently, lung burdens derived even from adjacent samples of lung parenchyma
can show broadly varying concentrations (dffering by orders of magnitude).
5.8
Revision 1 - 8/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Unfortunately, however, tissue samples that are available for analysis in support of a
human pathology study are typically "opportunistic" samples, which means that they
were selected and stored for an entirely different purpose, usually with little or no
thought given to the specific location from which they were collected (except in the most
general way). It is therefore seldom possible to address the effects of sample location
directly. Consequently, comparisons of tissue burden concentrations across samples
from different individuals in a human pathology study are at best qualitative and may
only be useful when averaged over large numbers of individuals and only when large
differences in concentrations (several orders of magnitude) are being distinguished.
Moreover, because the parts of a tissue that undergoes morphologic changes induced
by asbestos typically corresponds to the parts of a tissue where asbestos burdens are
the highest, even comparison of morphologic effects across tissue samples requires
proper consideration of the effects of the locations from which such tissue samples
were derived.
5.3 ANIMAL STUDIES
In animal studies, various species are exposed to measured doses of size-selected
mineral fibers and the resultant biological responses are monitored. Animals may be
dosed either by inhalation, intratracheal installation, implantation, or injection (U.S. EPA
1986). Such studies are conducted for several purposes. As with human pathology
studies, animal pathology studies are those in which the transport of asbestos
structures is tracked through the various organs and tissues of the animal and the
attendent cellular and molecular changes are characterized. In parallel with
quantitative epidemiology studies, animal dose/response studies track the incidence of
disease among a population that has been exposed in a controlled manner. One of the
advantages of animal dose/response studies over epidemiology studies is that
exposures are controlled and can be well characterized. The major disadvantage is
that there are many uncertainties introduced when extrapolating the results of animal
data to predict effects in humans. Therefore, attempts to adapt such things as animal
derived dose-response factors to humans are not generally recommended.
As with human epidemiology and pathology studies, the validity of conclusions drawn
from animal studies depends strongly on the techniques and methods used to
characterize and quantify asbestos structures either in the delivered dose or in the
tissues of the dosed animals (see Section 5.1). The ability to reconcile conclusions
derived from many animal studies with the rest of the asbestos literature is limited
because SEM was commonly employed to measure asbestos in animal studies but not
other kinds of studies. Even many of those studies in which TEM was employed for
asbestos analysis suffer from use of non-standard methods that cannot be easily
5.9
Revision 1 - B/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
reconciled with the more traditional TEM methods, particularly because such
specialized methods are seldom adequately documented to allow comparison.
As with human pathology studies, the location of a tissue sample excised for analysis is
a critical factor that also governs the quality of an animal pathology study (Section 5.2).
However, one potential advantage frequently available in animal pathology studies over
human studies is the ability to carefully identify and select the precise tissue samples to
be analyzed. The extent that a particular animal pathology study exploits this capability
can affect the overall utility of the study. Thus, such issues need to be addressed
carefully when evaluating and comparing the results across animal pathology studies.
The route of exposure employed in a particular animal study is also important to
consider. Each of the routes of exposure commonly employed in these studies
(inhalation, intratracheal installation, and injection or implantation) delivers different size
fractions of asbestos to a target tissue with varying efficiencies. For example, injection
or implantation studies deliver 100% of all size categories of structures to the target
tissue. However, the efficiency that each size category is delivered by inhalation is a
function of the aerodynamic properties of the asbestos structures and the air flow
characteristics of the lungs (Section 7.1). Thus, the relationship between dose and
exposure depends upon the route of exposure employed. Importantly, because the
ultimate goal is to understand the effects that inhaled fibers have on humans,
differences between the character of the delivered dose in an animal study and the
character that such a dose would have, had it originally been inhaled, typically need to
be addressed.
Regarding the measurement of health effects, many of the results in animal studies
suffer from a lack of statistical significance because of small numbers of observed
tumors. Consequently, trends cannot be established conclusively. Interestingly,
several studies draw conclusions regarding the relative potency of different sample
types when variations within the dose/response trend of a single sample type is larger
than observed differences between types (REFERENCE).
For animal inhalation studies, meaningful comparison of the relative deposition of
asbestos dusts across species is not direct. To extrapolate results across species, the
detailed differences in the physiology of the respiratory tracts between the species need
to be addressed (Sections 4.4 and 7.1). However, if measurements are available for
both species, differences in physiology are addressed, and the manner in which tissue
burdens are analyzed is considered, it may be possible to compare relative tissue
doses (mass of asbestos per mass of tissue) across species.
5.10
Revision 1 - 8/26/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
5.4 IN-VITRO STUDIES
A broad range of in-vitro studies provide useful insight on the effects of asbestos.
These include, for example, studies in cell-free systems (which have been used to
evaluate such things as asbestos dissolution rates or the kinetics of free radical
formation on the surface of asbestos fibers) and studies of the effects of asbestos on
cultures of a broad variety of cell types and tissues.
As with other studies, the potential limitations and sources of uncertainty associated
with in-vitro studies need to be considered when evaluating the validity of study results
or comparing such results to those of other studies, particular studies of varying type.
Also, as with other studies, among the primary sources of uncertainty that need to be
addressed for in-vitro studies is the manner in which asbestos doses are characterized
and quantified (Section 5.1). For in-vitro studies (as with animal studies dosed by
routes other than inhalation), this also extends to the need to consider the relationship
between the character of the asbestos dose applied in vitro and the character that a
similar exposure might possess following inhalation exposure in vivo (Section 5.3). This
can be particularly problematic for studies in tissue cultures because it is not clear how
the application of a suspension of fibers (with known concentration) to a dish containing
cultured cells can be related to doses that reach corresponding tissues following
administration to whole animals.
In-vitro studies, of necessity, represent isolated components of living systems observed
under conditions that may vary radically from those under which such components
operate in vivo. Consequently, the behavior of such components may also vary
radically from the behavior of the same components in vivo. Therefore, additional
study-specific considerations (concerning the design of a study and the conditions
under which a study is conducted) also need to be addressed before evaluating the
validity or relevance of the results from an in-vitro study to what might otherwise be
observed in a whole animals. Examples of such considerations include:
- for cell-free systems:
• whether conditions under which the study is conducted are sufficiently
similar to conditions in vivo to expect that the observed effect is likely to
occur in vivo; and
• if the observed variables describing the nature or magnitude of the effect
are also likely to reflect what may occur in vivo;
5.11
Revision 1 - 6/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
- for tissue cultures:
• whether conditions under which the study is conducted are sufficiently
similar to conditions in vivo to expect that the observed effect is likely to
occur in vivo;
• whether responses by specific tissues or cells in culture are likely to
behave similarly in vivo where their behavior may be suppressed,
enhanced, or modified in some other manner due to additional stimuli
provided by responses of other tissues and cells that are components of a
complete organism but that may be lacking in culture; and
• whether the conditions required to establish and maintain a tissue culture
for experimentation sufficiently alters the characteristics and behavior of
the cells being studied to minimize the relevance of results from such a
study to conditions in vivo.
Among the most important examples of the last consideration relates to the general
need to create immortalized cells to maintain tissue cultures. Thus, questions must
always be raised concerning whether the alterations required to create immortalized
cells for culture (from what are normally mortal cells in vivo) also alter the nature of the
responses being studied.
Note, many studies in the current literature also incorporate combined aspects of
several of the four general study types described in this chapter. For these studies, a
corresponding combination of the considerations described must therefore be
addressed when evaluating such studies and comparing their results with inferences
derived from the rest of the literature.
5.12
Revision 1 - 8/28/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
6.0 EPIDEMIOLOGY STUDIES
The existing epidemiology studies provide the most appropriate data from which to
determine the relationship between asbestos dose and response in humans. As
previously indicated, however, due to a variety of methodological limitations
(Section 5.1), the ability to compare and contrast results across studies needs to be
evaluated to determine the confidence with which risk may be predicted by
extrapolating from the "reference" epidemiology studies to new environments where risk
needs to be assessed. This requires both that the uncertainties contributed by such
methodological limitations and that several ancillary issues (identified in Chapter 2) be
adequately addressed.
A detailed discussion of the methodological limitations inherent to the available
epidemiology studies was provided in the Health Effects Assessment Update (U.S. EPA
1986) and additional perspectives are provided in this document (Section 5.1). The
manner in which the uncertainties associated with these limitations are addressed in
this document are described in Appendix A. As previously indicated (Chapter 2), the
ancillary issues that need to be addressed include;
• whether the models currently employed to assess asbestos-related risk
adequately predict the time-dependence of disease;
• whether the relative in-vivo durability of different asbestos mineral types
affect their relative potency;
• whether the set of minerals included in the current definition for asbestos
adequately covers the range of minerals that potentially contribute to
asbestos-related diseases; and
• whether the analytical techniques and methods used to characterize
exposures in the available epidemiology studies adequately capture the
characteristics of exposure that affect biological activity.
All but the third of the above issues are addressed in this Chapter. Currently, the third
issue can best be addressed by evaluating inferences from the broader literature (see
Chapter 7). The remaining issues are addressed separately for lung cancer and
mesothelioma following a brief overview of the approach adopted for evaluating the
epidemiology literature.
6.1
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
6.1 APPROACH FOR EVALUATING THE EPIDEMIOLOGY LITERATURE
To develop dose/response relationships (and corresponding risk coefficients) for use in
risk assessment from epidemiological data, two basic types of information are
necessary: information on disease mortality in the study population (cohort) and
information on the asbestos exposure experienced by each member of the cohort. So
that disease mortality attributable to asbestos can be distinguished from other
(background) causes of death, it is also necessary to have knowledge of the rates of
mortality that would be expected in the study population, absent exposure. Normally,
such information must be determined based on a "reference" or "control" population.
Ideally, one would like to have complete knowledge of exposure at any period of time
for each individual in the cohort and complete access to the data in order to fit different
types of dose/response models to the data so that the approach for evaluating the
relationship between dose and response can be optimized. In most instances,
unfortunately, the data suffer from multiple limitations (see Section 5.1 and Appendix A)
and the epidemiologist is further constrained by less than complete access to the data.
Briefly, the major kinds of limitations that potentially contribute to uncertainty in the
available epidemiology studies (and the effect such limitations likely produce on trends
in potency estimates) include:
• limitations in the manner that exposure concentrations were estimated
(likely increases random variation across studies);
• limitations in the manner that the character of exposure (i.e. the
mineralogical types of fibers and the range and distribution of fiber
dimensions) was delineated (likely increases systematic variation between
industry types and, potentially, between fiber types);
• limitations in the accuracy of mortality determinations or incompleteness in
the extent of tracing of cohort members (likely increases random variation
across studies);
• limitations in the adequacy of the match between cohort subjects and the
selected control population (likely increases random variation across
studies and may have a substantial effect on particular studies); and
6.2
Revision 1 - 0/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
• inadequate characterization of confounding factors, such as smoking
histories for individual workers (likely increases random variation across
studies and may have a substantial effect on particular studies).
More detailed discussion of the above limitations is provided in Section 5.1. The
manner in which these limitations are being addressed in this evaluation are described
briefly below and in more detail in Appendix A.
The existing asbestos epidemiology database consists of approximately 150 studies of
which approximately 35 contain exposure data sufficient to derive quantitative
dose/response relationships. A detailed evaluation of 20 of the most recent of these
studies, which includes the most recent followup for all of the cohorts evaluated in the
35 studies, based on the considerations presented in this overview, is provided in
Appendix A.
Importantly, this new analysis of the epidemiology database differs from the evaluation
conducted in the 1986 Health Effects Assessment Update (U.S. EPA 1986). Not only
does it incorporate studies containing the latest available followup for the exposure
settings previously evaluated and several additional studies addressing new exposure
settings, but the manner in which the analysis was conducted incorporates important,
new features. These new features include:
• estimation of more realistic confidence bounds for the dose-response
factors derived from each study; and
• for the lung cancer model, introducing a parameter, "a", which represents
the ratio of the background rate of lung cancer in the cohort relative to the
control population.
Confidence bounds were adjusted to account for uncertainty contributed by the manner
that exposure was estimated, by the manner that work histories were assigned, and by
limitations in the degree of followup, in addition to the traditional practice of accounting
for the statistical uncertainty associated with the observed incidence of disease
mortality. Thus, most of the major contributors to the overall uncertainty of each study
are now addressed, at least in qualitative fashion. A detailed description of how the
new confidence bounds were constructed is provided in Appendix A.
Note, in the text of this Chapter, we occasionally refer to significant differences between
certain measured or estimated values. When such differences are determined
informally (for example, by comparing confidence intervals) they are simply referred to
6.3
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
as "significant." When they are determined formally (by statistical test), they are
denoted as "statistically significant.*
In this analysis, dose-response coefficients were estimated for each cohort both by
requiring that a =1 (the traditional approach) and by allowing a to vary while optimizing
the fit of the data. Allowing a to vary addresses such potential problems as differences
in the relative fraction of smokers among cohort and controls, respectively, or
differences in background cancer rates due to lack of the appropriateness of the
selected control population for the cohort. As indicated in Appendix A, such
adjustments have a substantial effect on the fit of the EPA model to the data for several
specific cohorts and a corresponding effect on the estimates of the lung cancer dose-
response coefficients for those cohorts.
To better evaluate the sources of variation in estimated dose-response coefficients
observed across the available studies and potentially reconcile such variation so that
the precision of asbestos risk assessments might be improved (Sections 6.2 and 6.3),
we were also able to obtain the original, raw data for selected cohorts from a limited
number of some of the most important of the published epidemiology studies. This
allowed us to more formally evaluate the appropriateness of the existing U.S. EPA
models for lung cancer and mesothelioma, respectively, and to evaluate the effects of
such things as fiber biodurability. The models were evaluated by determining the fit of
the models to various representations of the time-dependence of mortality observed in
these raw data sets. A biologically based model, the two-stage model developed by
Moolgavkar et al. (Moolgavkar and Luebeck 1990) was also fit to the observed time-
dependence and contrasted with results from the U.S. EPA models.
Separately, we also completed an evaluation to consider questions concerning the
appropriateness of the range of fiber sizes characterized in the published epidemiology
studies and whether adjusted exposure indices might improve the quality of asbestos
risk assessments.
6.2 LUNG CANCER
The Airborne Health Effects Assessment Update (U.S. EPA 1986) utilizes a model for
lung cancer in which the asbestos-related age-specific incidence of lung cancer, "t"
years from onset of exposure, is proportional to cumulative asbestos exposure at time t-
10 years, multiplied by the age and calendar year incidence of lung cancer in the
absence of asbestos exposure. A linear relationship between cumulative dose and
response was assumed based on the ten epidemiology studies identified (in the 1986
6.4
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
EPA document) as containing sufficient information to establish a dose/response cuive
for asbestos induced lung cancer:
where:
"lL" is the overall incidence of lung cancer (expected new cancers per year
per person) adjusted for age and calendar year;
"lE" is the corresponding cancer incidence in a population not exposed to
T is the concentration of asbestos (expressed as PCM fibers longer than
5 pm in the existing evaluations of the published epidemiology data);
"t" is the time since onset of exposure in years;
"d(,.10)" is the duration of exposure excluding the most recent 10 years; and
"Kt" is the proportionality constant between dose and response. This is the
risk coefficient that represents the potency of asbestos.
The above model is a relative risk model in that it assumes that the excess incidence of
lung cancer from asbestos is proportional to the incidence in an unexposed population.
Since smokers have a much higher incidence of lung cancer, If smoking-specific
incidence rates are applied, the model predicts a higher excess incidence of asbestos-
related lung cancer in smokers than in non-smokers. This is consistent with the
multiplicative relationship between smoking and asbestos that has been observed in
epidemiological studies (see, for example, Hammond et al. 1979). Note that the KL in
the model pertains to an occupational pattern of exposure (e.g., 8 hours per day, 240
days per year) and must be modified before application to environmental exposure
patterns.
Equation 6.1 is generally written as follows to indicate that is a proportionality
constant between dose and response:
iL= yi+Kt+d(t.10)]
(6.1)
asbestos;
Kt(f)(d') = RR-1
(6.2)
where:
6.5
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
"RR" is the relative risk and is equal to (lL/lE) and d' is now the duration
excluding the final 10 years before observation of total lung cancer
deaths.
6.2.1 The Adequacy of the Current EPA Model for Lung Cancer
Access to the raw epidemiology data from a small number of key studies allowed us to
evaluate the adequacy of the EPA model for describing the time-dependence for lung
cancer in asbestos exposed cohorts. For this analysis, the raw data for the cohort of
crocidolite miners in Whittenoom, Australia was graciously provided by Nick DeKlerk
(unpublished) and the raw data for the cohort of chrysotile textile workers (described by
Dement et al. 1994) was graciously provided by Terri Schnorr of NIOSH. The
Whittenoom cohort was originally described by Armstrong et al. (1988), but the data
provided by DeKlerk included additional followup through 1999.
The EPA lung cancer model was fit to the raw data from both South Carolina and
Wittenoom. In these analyses, each person-year of followup was categorized by
cumulative exposure defined using a lag of 10 years, in accordance with the EPA lung
cancer model. The data were then grouped into a set of cumulative exposure
categories and the observed and expected numbers of lung cancers were computed for
each category. For South Carolina, expected numbers were based on sex-race-age-
and calendar-year-specific U.S. rates and separate analyses were conducted for white
males, black males, and white females, as well as for the combined group. For
Wittenoom expected rates were based on age- and calendar-year-specific rates for
Australian men. These data were also categorized by cumulative exposure lagged 10
years. The resulting fit of the EPA model to the data from Wittenoom and South
Carolina are presented in Tables 6-1 and 6-2, respectively.
The data in these tables were fit assuming that the observed cancers in each
cumulative exposure category had a Poisson distribution with mean response equal to
the expected response in the comparison population times a (1 + Kt*CE10), where the
linear slope, K^, is the "lung cancer potency factor" in the EPA lung cancer model, and
CE10 is the mean cumulative exposure for a category computed using a lag of 10 years.
For the Wittenoom data (Table 6-1), the fit with a = 1 is poor (p < 0.0001), as the model
overpredicts the number of cancers in the highest exposure category and greatly
underpredicts at lower exposures. By contrast the fit of the model with a variable is
adequate (p = 0.1). This model predicts a high background of lung cancer in this cohort
(a = 2.1) and a fairly shallow slope, with the relative risk increasing only from 2.1 at
background to 3.6 in the highest exposure category. A test of the hypothesis that a = 1
6.6
Revision 1 - 6/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
for this cohort is rejected (P < 0.01) and the model fit to the data (with a variable)
predicts KL= 0.0047.
The fit of the EPA model to the South Carolina lung cancer data categorized by
cumulative exposure is shown in Table 6-2. The model with a = 1 cannot be rejected
both when the model is applied to white males only (p = 0.54) or with all data combined
(p = 0.92). Since the values a and Kt estimated from white males only are similar to
those estimated using the complete cohort (black and white males and white females),
the fit to the complete data is emphasized in this analysis. This model fit to the data
predicts a = 1.2 and KL = 0.021 (f/rnl-y)'1. A test of the hypothesis that a =1 for this
cohort cannot be rejected (P = 0.21).
6.7
Revision 1 • 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 6-1
6.8
Revision 1 - 9/2/01
-------
Table 6-1:
Fit of EPA Lung Cancer Model to Observed Mortality Among Wittenoom Miners
Categorized by Cumulative Exposure
Mean
Cumulative
Exposure
Lagged Cases Predicted Log
10years (f- Observed Expected (a<1) (a variable) Likelihood
vrVcc
0
5
4.6
4.6
9.8
-3.8
0.193
27
7.9
8.0
17.0
-4.1
0.692
11
8.2
8.3
17.6
-4.4
1.59
22
11.6
12.1
24.9
-3.1
3.27
28
12.9
14.0
27.9
-2.7
6.19
38
14.3
16.7
31.4
-3.0
11.8
31
13.2
17.4
29.8
-2.6
21.5
21
9.2
14.5
21.6
-2.5
41.1
25
11.6
24.5
29.6
-2.7
142
43
11.6
56.5
41.6
-6.4
Totals
251
105.1
176.6
251.0
-35.1
Goodness of Fit P-value 1.3316E-13 0.10
Test of Ho: a=1
p < 0.01
Estimates of KL ((f-yr)/cc)"1
( ° * 1 )
Kl = 0.027
90% CI = ( 0.020, 0.035 )
(a variable)
Kl = 0.0047
90% CI = ( 0.0017, 0.0087 )
-------
Table 6-2:
Fit of EPA Lung Cancer Model to Observed Mortality Among South Carolina Textile
Workers Categorized by Cumulative Exposure
Cumulative
Exposure Lagged 10
years (f-yr)/cc Cases Predicted Log
Mean Rang* Observed
Expected
(ae1)
(a variable)
Likelihood
0.00 0
0
0.88
0.88
1.06
-1.06
0.35 [0.0,0.5)
4
2.64
2.67
3.23
-1.72
0.75 [0.5,1.0)
7
5.51
5.63
6.78
-1.91
1.66 f 1.0, 2.5 )
10
9.66
10.12
12.11
-2.27
3.37 [2.5,4.5)
13
7.89
8.65
10.23
-2.55
6.23 [ 4.5, 8.5 )
11
7.69
9.05
10.53
-2.14
11.80 [8.5,16)
11
7.20
9.61
10.86
-2.13
21.28 [16,28)
8
5.26
8.43
9.20
-2.05
41.49 [ 28,60)
14
6.34
13.80
14.33
-2.25
69.17 [ 60,80)
9
3.03
8.98
8.97
-2.03
93.28 [80, 110)
10
2.95
10.76
10.53
-2.09
173.36 (>=110)
25
4.32
25.56
24.17
-2.55
Totals
Goodness of Fit P-value
122
63.36
114.14
0.92
122.00
0.96
Test of Ho: a=1
p = 0.21
Estimates of Ki ((f-yrj/cc)"1
(a«1)
Kl = 0.028
90% CI = (0.021, 0.037 )
(a variable)
= 0.021
90% CI = ( 0.012, 0.034 )
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 6-2
6.9
Revision 1
9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
6-2.1.1 Time Dependence
The EPA model was next evaluated to determine whether it adequately describes the
time-dependence of the mortality observed in the Wittenoom and South Carolina
cohorts. The model predicts that relative risk will remain constant following 10 years
from the time of last exposure.
Table 6-3, shows the fit of the EPA model to the Wittenoom data categorized by time
since last exposure. In each category of time since last exposure, the number of lung
cancer deaths are predicted by the EPA model using the parameter values calculated
in Table 6-1, based on the expected cancers and the average cumulative exposure for
each category. This categorization allows investigation of the assumption inherent in
the EPA model that the relative risk remains constant following 10 years from last
exposure. The Wittenoom data appear to be consistent with this assumption, as the
excess relative risk per f/ml-y of cumulative exposure varies only between 0.059 and
0.073 for categories representing more than ten years from last exposure.
The conclusion that relative risk remains constant following 10 years from last exposure
among Wittenoom miners is further illustrated in Figure 6-1. Figure 6-1 is a plot of the
quantity: (relative risk - a)/(cumulative exposure) versus time since last exposure.
Based on the modified form of the EPA model (in which a is allowed to vary), this
quantity, (RR-a)/(CE10), should be constant and equal to a*Kt (beyond 10 years from
last exposure). Figure 6-1 demonstrates that this is exactly the behavior observed
among Wittenoom miners, where the value of the quantity plotted on the Y-axis rises
rapidly until 10 years after exposure ends and then remains constant for all longer
times. Moreover, a test of the hypothesis that the slope of the data depicted in
Figure 6-1 (beyond 10 years after the end of exposure) differs from zero cannot be
rejected
Because it has been reported that asbestos fibers (especially chrysotile but including
amphiboles) dissolve in biological fluids and that animal studies suggest that the
tumorigenicity of fibers is proportional to in-vivo durability (Section 7.2.4), the
Wittenoom data categorized by time since last exposure was also fit to a modification of
the EPA lung cancer model in which the cumulative exposure variable is replaced by an
estimate of internal dose derived assuming first-order clearance of fibers. Results are
presented in Table 6-4
In developing Table 6-4, a preliminary investigation, based upon an "exact" likelihood
that did not involve categorization of exposures, indicated that this model provided a
best fit using a half-life of 11 years and a lag of 12 years. As this table indicates,
6.10
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
despite optimizing the model with respect to the lag and half-life, it did not fit the data as
well as the standard EPA model based upon cumulative number of inhaled fibers.
Thus, there is no evidence from the Wittenoom data that crocidolite fibers degrade (or
6.11
Revision 1 • 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 6-3
6.12
Revision 1 • 9/2/01
-------
Table 6-3:
Fit of EPA Lung Cancer Model to Observed Mortality Among Wittenoom Miners
Categorized by Time Since Last Exposure
Cumulative Excess
Exposure relative
Time Since Last Exposure Lagged Predicted risk per
Mean
Range
10 years
Cases
Expected
( a = 1 ) (a variable )
f/ml*y
(Yr»)
fYrs)
ff-Yry/cc
0.267
f0- 1]
1.18
0
0.64
0.658
1.368
-0.8505
3.01
[1-5]
1.23
1
1.60
1.652
3.432
-0.3042
7.49
[5- 10]
7.68
8
3.93
4.750
8.698
0.1347
14.9
[10-20]
24.43
45
18.39
30.557
43.792
0.0592
24.7
[20 - 30]
22.48
96
36.41
58.584
86.002
0.0728
34.0
[30- 40]
21.12
81
34.20
53.766
80.306
0.0648
43.6
[40-1000]
16.62
20
9.91
14.376
22.828
0.0612
Totals
251
105.08
164.342
246.427
Goodness of Fit
Chi sq
43.2811
6.7165
P value
3.24E-08
0.152
Using Australian Mortality Rates
-------
Table 6-4:
Fit of (EPA Lung Cancer Model to Obeerved Mortality Among Wittenoom Miners Categorized by
Time Since Last Exposure with Cumulative Exposure Replace by Internal Dose
Time Since Last
Internal
Cases
EPA model
Log
Exposure (Yrs)
Dose
Obssrved
Expected
Predicted
(a variable)
Likelihood
0
15.8
0
0.5
1.2
1.4
-1.16
(0-5)
18.4
0
1.5
3.8
3.4
-3.82
[5-10)
19.9
5
3.6
9.7
8.7
-3.13
[10-20)
9.1
42
17.5
41.3
43.8
-2.80
[20 - 30)
6.4
91
35.7
81.6
86.0
-3.69
[30 - 40)
4.2
88
35.4
79.0
80.3
-3.65
(40-100)
1.7
25
11.0
23.7
22.8
-2.57
Totals
251
105.1
240.3
-20.81
Goodness of Fit
Chi sq
14.83
672
P value
0.01
0.15
-------
LU
O
03
¦
tc
a:
0.5
0
-0.5
-1
-1.5
-2
-2.5
Figure 6-1:
Plot of [(RR - alpha)/(CE10)] vs. Time Since Last
Exposure for Wittenoom Miners
0
10 20 30 40
Time Since Last Exposure (Years)
50
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
that the effects of crocidolite fibers are diminished) in vivo over time scales that are
short relative to a human life time and this is consistent with the data presented in
Section 7.2.4 that suggest crocidolite fibers likely last many decades in human lungs, at
a minimum. It is also consistent with the appearance of Figure 6-1, which shows no
decrease in the effect of crocidolite following cessation of exposure.
Table 6-5 shows the fit of the EPA model to the South Carolina data categorized by
time since last exposure. This table was constructed in the same manner as Table 6-3.
In contrast to the data from Wittenoom, the time dependence of the South Carolina data
cannot be adequately fit with the EPA model, whether a is held equal to one
(P < 0.0001) or allowed to vary (P < 0.0007). A plot of (RR-a)/CE10 versus time (Figure
6-2) indicates why.
In Figure 6-2 (as with the Wittenoom data), the quantity plotted on the V-axis increases
sharply to a maximum at approximately 10 years following the end of exposure. In
contrast to what is observed for Wittenoom, however, the value of this quantity
decreases with time beyond the first 10 years following the end of exposure. This
decrease is clear from the figure, despite some scatter in the data. Thus, in contrast to
the assumption inherent to the EPA model, relative risk decreases precipitously in this
cohort after 10 years following the cessation of exposure. This suggests that the
effects of chrysotile are somehow diminished after long periods of time.
There are several reasonable hypotheses that might explain the decrease in relative
risk with time within the South Carolina cohort. Among those that have attracted
interest is the possibility that risk is a function of biodurability (for example, see
Section 7.2.4). We therefore decided to explore this possibility.
Given that the expected lifetime for respirable chrysotile fibers in biological fluids is
anticipated to be only on the order of a year (see Section 7.2.4), further analysis of the
decreasing trend observed in Figure 6-2 is warranted. Therefore, the South Carolina
data were also fit using the modified form of the EPA model that accounts for internal
dose (rather than cumulative exposure). In this model, as indicated above, the internal
dose is derived from cumulative exposure, except it is assumed that, from the moment
that a fiber is deposited in the lung, it dissolves by a first order process. Actually, the
best analysis of the dissolution process indicates that it is likely zero order, but the first
order decay (see Section 7.2.4) is assumed as an approximation because this provides
an easier description of the lifetime of the material.
6.15
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 6-5
6.16
Revision 1 - 9/2/01
-------
Table 6-5:
Fit of EPA Lung Cancer Model to Observed Mortality Among South Carolina Textile Workers Categorized by Time Since Last
Exposure
Cumulative
Excess
Time Since Last
Exposure
relative
Exposure
Lagged
Predicted
Log Likelihood
risk per
Mean Range
10 Years (f
Cases
Expected
(o-1)
(a variable)
(o=1) (avariable)
(f-yr)/cc
-------
Figure 6-2:
Plot of [(RR-alpha)/(CE10)] vs. Time Since Last
Exposure for South Carolina Textile Workers
0.16
0.14
o 0.12
0.1
in
o
¦f 0.08
Q.
7 0.06
£ 0.04
0.02
0
0 20 40 60
Time Since Last Exposure (Years)
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Figure 6-2
6.17
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 6-6 shows the fit to the South Carolina data to the EPA model using internal
dose. This Table was derived from the South Carolina data in the same manner that
Table 6-4 was derived from the Wittenoom data. Since data appeared to be about
equally consistent with a range of lags and half-lives, for consistency, the same values
were used as with the Wittenoom data. This table indicates that this model provides an
adequate fit to the South Carolina data (P = 0.16).
Assuming a first order decay process, the best fit trend line to the data beyond 10 years
following cessation of exposure in Figure 6-2 suggests a half-life for chrysotile of
approximately 20 years. A similar analysis of the Whittenoom data suggests a half-life
for crocidolite that is > 700 years, which is no different than infinity.
Although this is in stark contrast to the dissolution half-life for chrysotile fibers estimated
based on in-vitro data, these results may not be inconsistent. There are at least two
reasonable explanations for this apparent discrepancy. First, the data in Figure 6-2
track the risk attributable to the presence of chrysotile, not the presence of chrysotile
itself. It is therefore reasonable that, even if the chrysotile fibers disappear readily, the
risk from their initial presence may decay only much more slowly; the biological
changes induced by the initial presence of the fibers may persist.
Second, as indicated in the in-vitro studies (see Section 7.2.4), the dissolution rate of
chrysotile is limited by the rate at which fluid flows past the material. The maximum
dissolution rate (i.e. the rate reported in the in-vitro studies) is observed only when flow
is adequate. Otherwise, partial saturation of the solution surrounding the chrysotile
fibers will slow the dissolution rate until it is only a fraction of what is observed in the
optimally designed studies. Therefore, that a slower dissolution rate is suggested by
the South Carolina data may simply indicate that the various biological compartments in
which the biologically active fibers are located do not allow for adequate flow of an
adequate reservoir of biological fluids to maintain the dissolution rate at its theoretical
maximum.
An alternate hypothesis is that the fibers are not dissolving at all but are simply being
sequestered in some manner that reduces their biological activity with time. Such a
hypothesis, however, does not appear consistent with the animal data that suggests
that tumorigenicity is indeed a function of the in-vivo durability of fibrous materials (see
Section 7.2.4). However, other hypotheses may also reasonably explain the data.
Therefore, although relative risk clearly decreases with time after cessation of exposure
among the South Carolina workers and such observations are not inconsistent with the
possibility that this is an effect of biodurability, it is premature to conclude such a link
without further study.
6.18
Revision 1 - 9/2/01
-------
Table 6-6:
Fit of EPA Lung Cancer Model to Observed Mortality Among South Carolina Textile Workers
Categorized by Time Since Last Exposure with Cumulative Exposure Replace by Internal Dose
Time Since Last
Exposure
Internal
Cases
EPA model
Log
Mean Range
Doss
Observed
Expected
Predicted
(a variable)
Likelihood
1Y»») (YibI
0 0
16.4
12
3.8
8.6
7.2
-2.77
3(0-5 )
18.9
10
2.1
5.1
3 5
-3.95
7 [5 -10 )
20.6
11
3.4
8.4
6.0
-2.46
14 [10-20)
15.1
13
7.5
16.4
12.1
-2.59
25 [ 20 - 30)
10.8
25
11.9
23.1
18.2
-2.61
35 [30-40)
5.9
27
19.9
33.5
28.8
-3.24
45 [ 40 - 100)
4.7
23
14.8
23.9
19.3
-2.51
Totals
121
63.5
119.0
95.1
-20.14
Goodness of Fit
Chi sq
7.94
21.5
P value
016
0.0007
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Importantly, the time-dependent behavior described above for crocidolite and chrysotile
are each based on analysis of data from only a single study. Thus, although such
trends are also consistent with additional evidence from animal studies and other
implications from the literature, it would be useful to confirm these observations by
analyzing additional raw epidemiology data sets. If they can be acquired, the raw data
from Libby may serve as a second example for amphiboles and the lung cancer data
from Quebec may serve as a second example for chrysotile.
6.2.1.2 Comparison with bioloaicallv-based model
To further evaluate the behavior of the EPA model, results were compared with those of
a biologically based (two-stage) model.
Definition of the two-stage model.- The two-stage model of cancer implemented in
this report (Moolgavkar and Luebeck, 1990) assumes that there is a population of
normal cells under homeostatic control whose size remains fixed at X. A normal cell
mutates at a given rate to an intermediate cell. (Mathematically, intermediate cells are
generated according to a non-homogeneous Poisson process with intensity v, which
implies that the per-cell mutation rate is v/X. For a given value of v the hazard function
does not depend upon X and, without loss of generality, X was fixed at 107 cells.) An
intermediate cell divides into two intermediate cells at a rate a, dies at a rate P, and
divides into one intermediate and one malignant cell at a rate p. (Mathematically a
clone of intermediate cells is a non-homogeneous birth, death and mutation process )
Importantly, the "a" in this model is in no way related to the "a" variable used to adjust
for background mortality in the EPA lung cancer model and the two should not be
confused.
The mortality hazard function (rate of dying of cancer at age t among persons who
survive to that age) is assumed to be the hazard for the occurrence of the first
malignant cell, with a delay of L years. That is, the mortality hazard at t years of age is
assumed to be h(t-L), where h is the hazard for the occurrence of the first malignant
cell. Thus, L represents the delay between the occurrence of the first malignant cell
and the subsequent death from the resulting cancer. Following the assumption of Doll
and Peto (1978) and Moolgavkar et a/. (1993), a lag of L = 4 years was generally
assumed. We note that the lag in this model is conceptually different from the lag of ten
years used in the EPA models. Whereas the lags in the latter model were empirically
derived based on goodness of fit, the lag in the two-stage model has a mechanistic
interpretation and, consequently, should not necessarily have the same value.
6.21
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
To introduce the effect of exposure to asbestos into the model, one or more of the
parameters v, a, P or |J were assumed to be linear functions of the instantaneous
whole-body fiber burden, d(t), where t indicates age:
v(t) = v0 + v, * d(t),
a(t) = a0 + a, * d(t),
P(t) = P0 + Pi* d(t),
M(t) = Mo+ Mi * d(t).
The hazard function, h(t) and the corresponding survival function (probability of
surviving to age t without death from lung cancer in the absence of competing causes
of death), S(t) = exp(-J0l h(s)ds, predicted by this model, although mathematically
complex, can be solved for explicitly whenever the parameters are piecewise constant
(Moolgavkar and Luebeck, 1990; Heidenreich et a!., 1997).
In application of this model to lung cancer (and mesothelioma) data from Wittenoom,
South Carolina, and Quebec: the fiber body burden, d(t), was computed (up to a
multiplicative constant) from a subject's exposure pattern, assuming a first-order
elimination process with an elimination rate of r (half-life of-Ln(.5)/r). In these
calculations, an average exposure rate during work hours was computed for each year
of work. Based on the first-order assumption, the body burden at time t resulting from
exposure to an average of f f/ml during the year between s and s+1 is
d(t) = (f/r)[1 - exp(-r(t-s)] for t ss+1
= (f/r)[1 - exp(-r)]exp[-r(t-s-1)] for t > s+1.
This expression does not include the multiplicative constant needed to convert from the
air concentration of asbestos to internal deposition rate, but without loss of generality is
assumed to be included in the parameters a1t J31t and Vv
The average body burden between t and t+1 years due to exposure between s and s+1
years was computed by integrating this expression from t to t+1. The average body
burden during each year was then computed by summing the contributions to the
average from each year (by summing over sst). The body burden d(t) was
approximated by the step function formed by these yearly average body burdens With
this approximation the two-stage parameters v(t), a(t), p(t) and p(t) are piecewise
constant and the exact solution to the two-stage model for this case (Heidenreich et al.,
6.22
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
1997) was used to calculate the hazard, h(t) and survival function, S(t). This calculation
was made using a FORTRAN subroutine kindly provided by Drs. Suresh Moolgavkar
and Georg Luebeck.
This model was fit to the raw lung cancer and mesothelioma data from Wittenoom and
South Carolina, and to the raw mesothelioma data from Quebec using the method of
maximum likelihood. The contribution to the likelihood for a subject who began work at
age to and died of cancer at age t, was h(t1)*S(t1)/S(t0), and the corresponding
expression for a subject whose followup ended at t, without resulting in death from
cancer was S(t1)/S(t0). The complete likelihood was computed as the product of these
individual contributions. The software program, MINUIT, developed by CERN in
Geneva, was used to maximize the log-likelihood.
The full two-stage model described above has eight parameters (nine if the elimination
rate, r, is included), which is too many to be estimated reliably using the epidemiologic
data available. Moreover, the general model is not fully identifiable, as different
combinations of parameters can lead to the same hazard (Heidenreich et at., 1997).
Accordingly, in applications of this model, not all of the parameters were estimated
independently. For example, in most cases only one of the two-stage parameters (V, a,
P, p) was allowed to be dose-related. Also, to reduce the number of estimated
parameters, in most cases either a or 3 was fixed at zero, because, when the mortality
rate is low, h(t) is practically a function of the difference between these two parameters.
Also, the elimination rate, r, was not estimated from the epidemiologic data, but was
determined from evidence on the durability of different types of asbestos fibers (see
Section 7.2.4).
Although the parameters in the two-stage model at least nominally represent
physiological phenomena, it must be realized that the model is, at best, likely to be a
highly idealized representation of a very complex process (see Section 7.3.1).
Accordingly, the results from this model are probably best used to investigate the
relative contributions to risk from exposures at different ages (e.g., if the model predicts
that v is dose-related, then earlier exposures are more important in determining risk,
whereas if the model predicts that p is dose-related, then more recent exposures are
more important). Moreover, if the model predicts that v is dose-related, the offending
toxic substance can be considered to be acting as a classical initiator; if a (or, more
precisely, a - 3) is dose-related, the toxic substance is more likely a promoter. If p is
dose related, the toxin is likely a late-stage actor, which also would present different
behavior than an initiator. An important advantage of the two-stage model is that, since
it models risk as a function of internal dose, it can naturally incorporate consideration of
fiber clearance.
Applying the two-stage model to lung cancer. Table 6-7 summarizes the results of
three applications of the two-stage model to the lung cancer data from Wittenoom, and
6.23
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 6-8 summarizes a similar application of the model to the South Carolina lung
cancer data. With each set of data, the model is fit allowing either V, a, or p to be a
linear function of the fiber body burden. In keeping with the half-life estimates based on
the durability of chrysotile and amphibole fibers in Section 7.2.4, for the South
6.24
Revision 1 - 9/2/01
-------
Table 6-7:
Results of Three Applications of the Two-Stage Model to the Lung Cancer Data from
Wittenoom (Assumed Half-life: 20 yrs)
Application 1 Application 2 Application 3
Parameter
Fitted
Value
Standard
Error
Fitted Value
Standard
Error
Fitted
Value
Standard
Error
a0
0.106
0.00535
0.345
0.0071
0.338
0.00702
al
0
fixed
0.000754
0.000301
0
fixed
po
0
fixed
0
fixed
0
fixed
Pi
0
fixed
0
fixed
0
fixed
V©
2 03
0.304
0.00669
0.000682
0.00744
0.000774
vl
0.474
0.223
0
fixed
0
fixed
pO
2.03E-07
3.O4E-0B
6.69E-10.
6.82E-11
7.44E-10
7.74E-11
pl
0
fixed
0
fixed
2.57E-11
1.7BE-11
X
1.00E+07
fixed
1.00E+Q7
fixed
1.00E+07
fixed
Log
Likelihood
Predicted
Cancers
-1669.29
235.6
-1642.19
240.9
-1642.33
240.7
-------
Table 6-8:
Results of Three Applications of the Two-Stage Model to the Lung Cancer Data from
South Carolina (Assumed Half-life: 0.5 yre)
Application 1 Application 2 Application 3
Parameter Fitted Standard Fitted Value Standard Fitted Value Standard
Value Error Error Error
aO
0.100
0.00815
0.0964
0.00725
0.107
0.00795
al
0
fixed
0.00610
0.00115
0
fixed
po
0
fixed
0
fixed
0
fixed
PI
0
fixed
0
fixed
0
fixed
vO
1.63
0.385
2.07
0.418
1.61
0.375
vl
20.3
5.77
0
fixed
0
fixed
no
1.63E-07
3.85E-08
2.07E-07
4.18E-08
1.81E-07
3.75E-08
0
fixed
0
fixed
8.13E-08
3.05E-08
X
1.000E+07
fixed
1.000E+07
fixed
1.000E+07
fixed
Likelihood -825.4 -B30.9 -832.9
Predicted
Cancers 114.6 113.9 115.3
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Carolina data, the half-life of internally deposited fibers was assumed to be 0.5 years (r
= -ln(0.5)/20 = 1.39/year) and, for the Wittenoom data, the half-life was assumed to be
20 years (r = 0.035/year).
Note that these analyses were completed before the analysis of the time-dependence
of the data using the EPA model, so that we did not have the opportunity to go back
and run this biological model using half-lives that might better represent what has been
observed with the time-dependence of the Wittenoom and South Carolina data
(Section 6.2.1.1). However, now that we have developed an approach that allows
evaluation of the time-dependence of lung cancer mortality based on estimates of
internal dose using the EPA model, further analysis using the two-stage model may not
be warranted (see Section 6.3.1).
When applied to the Wittenoom data the model provided comparable descriptions of
the data (i.e., similar likelihoods) when either the proliferation rate of intermediate cells,
a, or the mutation rate of intermediate cells, [J, was assumed to be dose related.
These models provided a much larger likelihood than the model in which the mutation
rate of normal cells, V, was assumed to be dose-related. Note that the "likelihoods"
reported in the table are "negative likelihoods" so that the smaller the number in the
table, the larger the actual likelihood.
Table 6-9 shows the fit of the best-fitting 2-stage model to the Wittenoom data
categorized by time since last exposure, compared to the fit of the EPA lung cancer
model. As this Table shows, the simpler EPA model fits these data better than the 2-
stage model, although the fit of both models is adequate (lack of fit p-value s 0.12 for 2-
stage model and 0.35 for EPA model).
The above indicates that the best-fitting model runs for Wittenoom adequately describe
the time-dependence for lung cancer and further indicate that asbestos primarily affects
a (and potentially }J). This suggests that crocidolite may be acting primarily as a
promoter (or at least a late-stage actor) rather than an initiator toward the induction of
lung cancer, which is consistent with inferences from several recent in-vitro studies
(Section 7.3.3.4 and 7.3.4.7) and a recent retrospective cohort study of retired workers
(Sanden et al. 1992).
Table 6-10 shows the fit of the 2-stage model (in which v, the mutation rate from normal
to intermediate cells is dose-related, i.e. the best-fitting model) to the totality of the
South Carolina data categorized by time since last exposure, compared to the fit of the
EPA model. Although the 2-stage model fits the data only marginally well (p s 0.05), it
fits the data much better than the EPA model (p < 0.001). The EPA model
underpredicts the number of cases for short times since end of exposure.
6.27
Revision 1 • 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
g
° Revision 1 - 8/2/01
-------
Table 6-9:
Fit of the Best-fitting Two-Stage Model to the Wlttenoom Lung Cancer Data Categorized by Time
Since Last Exposure (Assumed Half-life: 20 yrs)
Time Since Last
Exposure Hazard Cases EPA Log EPA Log
Mean Range Observed Predicted (available) Likelihood Likelihood
Cfo)
0.310-1 )
0.00013
0
1.3
1.4
-1.35
-1.37
3.011-5 )
0.00018
1
3.3
3.4
-2.13
-2.20
7.515-10 )
0.00035
8
8.8
8.7
-2.01
-2.00
14.9110-20)
0.00092
45
41.7
43.8
-2.95
-2.84
24.7 1 20 - 30)
0.0024
96
88.5
88.0
-3.51
-3.76
34.0 [30-40)
0.0047
81
91.4
80.3
-3.73
-3.12
43.6 ( 40 -100)
0.0064
20
21.6
22.8
-2.48
-2.80
Totals
251
256.7
246.9
-18.18
-17.89
Goodness of Fit
Chi sq
7.26
6.72
P value
0.12
0.35
-------
Table 6-10:
Fit of the Best-fitting Two-Stage Model to the South Carolina Lung Cancer Data Categorized by
Tim* Since Last Exposure (Assumed Half-life: 0.5 yrs)
Time Since Last
Exposure
Cases
EPA
Log
EPA Log
Mean Range
Hazard
Observed
Predicted
(a variable)
Likelihood
Likelihood
lYra) (Yrs)
0.06(0-1 )
0.00055
15
12.9
7.2
-2.43
-5.43
3.011-5 )
0.00060
7
6.1
3.5
-1.97
-3.25
7.515-10 )
0.00080
14
9.7
6.0
-3.06
-6.17
14 8 [10-20)
0.00089
11
19.3
12.1
-4.23
-2.18
24.8 [ 20 - 30)
0.0012
28
21.4
18.2
-3.51
-4.83
34.6 [30-40)
0.0024
28
30.7
28.8
-2.71
-2.60
43.5 [ 40 - 100)
0.0044
19
24.0
19.3
-2.96
-2.40
Totals
122
124.2
95.1
-20.88
-26.86
Goodness of Fit
Chlsq
P value
9.52
0.05
21.47
0.0015
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
The results with the South Carolina data provide a stark contrast to the Wittenoom
results. With the South Carolina data, the model with V dose-related provided the best
fit. Thus, the results from South Carolina suggest that chrysotile may be acting
primarily as an initiator. Such an interpretation must be tempered, however, by several
inconsistencies and potential problems with the South Carolina fits that may be
attributed, among other possibilities, to the short half-lives assumed for chrysotile.
The two fits (for South Carolina and Wittenoom) also provide very different estimates of
the background parameters (v0, a0 and p0). The (J0 and v0 estimated from the South
Carolina data both are about 250 times larger than those estimated from the Wittenoom
data. Although these rates could differ somewhat in these cohorts, owing to different
lifestyles and genetic traits in the two populations, these differences appear too large to
be attributable to such causes.
Due to the apparently conflicting results from the analyses of the lung cancer data from
Wittenoom and South Carolina, we conducted several additional analyses to attempt to
identify the source of the apparent conflict. We initially explored the possibility that the
assumed half-life for crocidolite was too long and completed additional model runs
assuming a 0.5 yr half-life (identical to what was assumed for South Carolina) and
these results are described below. Subsequently, we discovered during our analysis of
time-dependence using the EPA model (see Section 6.2.1.1), that the half-lives
measured in vitro for these fibers are likely much too short, which suggests a solution to
the apparent conflict and this is also described below. Unfortunately, however, this
revelation came too late to allow additional, confirmatory runs with the two-stage model
to be completed in time for this report. Therefore, the apparent conflict between the
Wittenoom and South Carolina results (at least with regard to the two-stage model)
remains to be definitively resolved.
The results of the two-stage model are based on the assumption that one of the two-
stage parameters is a linear function of the instantaneous body burden predicted by a
first order retention model that relies on dissolution lifetimes for the asbestos fibers that
are observed in vitro (Section 7.2.4). Alternately, it could be the case that an inhaled
fiber is soon neutralized in some manner (e.g. by some type of sequestration
mechanism) so that any contribution to cancer induction must occur soon after
exposure. To explore this possibility, the two-stage model was also fit to the Wittenoom
data assuming a half-life of only 0.5 year (the same value used for the South Carolina
cohort) and results are indicated in Table 6-11.
When the Wittenoom data are evaluated assuming a fiber half-life of 0.5 yrs, as was the
case using a 20-year half life, the best-fitting of the three submodels considered was
the one with a dose-related. This fit using this submodel was not appreciably worse
than that obtained assuming a 20-year half life (difference in likelihoods of 1.16).
6.31
Revision 1 - 9^2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 6-11
6.32
Revision 1 • 9/2/01
-------
Table 6-11:
Results of Three Applications of the Two-Stage Model to the Lung Cancer
Data from Wittenoom (Assumed Half-life: 0.5 yrs)
Application 1 Application 2 Application 3
Parameter
Fitted
Standard
Fitted
Standard
Fitted
Standard
Value
Error
Value
Error
Value
Error
ctO
0.106
0.00533
0.342
0.00739
0.349
0.00730
ai
0
fixed
0.00856
0.00439
0
fixed
po
0
fixed
0
fixed
0
fixed
PI
0
fixed
0
fixed
0
fixed
vO
2.01
0.302
0.00717
0.000731
0.00692
0.000707
vl
4.66
2.21
0
fixed
0
fixed
no
2.01 E-07
3.02E-08
7.17E-10
7.31E-11
6.92E-10
7.07E-11
0
fixed
0
fixed
0.00E+00
1.49E-11
X
1.00E+07
fixed
1.00E+07
fixed
1.00E+07
fixed
Likelihood -1669.40 -1643.36 -1647.27
Predicted
Cancers 235.6 240.9 241.1
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
However, with a 0.5-year half-life, the fit of the model that assumed p was dose-related
was appreciably worse than the comparable model assuming a 20-year half life
(change in likelihoods of 5). As this analysis demonstrates, choice of an appropriate
internal dose can be confounded with selection of the appropriate two-stage parameter
that is affected by that dose. The analysis also implies that, due to the number of
variables that can be optimized, the two-stage model does not appear to be a good
discriminator for evaluating the biodurability of fibers.
These results do not suggest that asbestos acts at short times and is then sequestered.
However, they do suggest that favo'red dose-dependent variables may be strong
functions of other constraints placed on the model (such as the assumed half-life of the
fibers). This further implies that the South Carolina results reported above might be
anomalous, due to the assumption of a half-life that is short relative to the actual rate at
which risk is observed to decrease for this cohort (based on the results from the
analysis using the EPA model, Section 6.2.1.1).
It is unfortunate that it now appears that we may have evaluated these data using the
wrong estimates of dissolution half-lives (which were, nevertheless, based on the best
estimates from the literature). Our analysis of the time-dependence of the data using
the EPA model (Section 6.2.1.1) suggests that both crocidolite and chyrsotile dissolve
in vivo much more slowly than in vitro studies suggest and that this should not be
surprising given that such materials only dissolve at their optimum rates when the fluid
they are dissolving in is rapidly and continually renewed. Fluid flow rates in most
compartments of the lung may simply not be adequate to facilitate dissolution of
asbestos at the maximum rate.
It would be interesting to rerun the Moolgavkar model using half-lives based on what
was observed with the analyses using the EPA model (approximately 20 years for
chrysotile and > 700 yrs or virtually infinite for crocidolite). The goal would be to
determine whether the fitting the South Carolina data might then show that chrysotile
also acts primarily as a promoter (in agreement both with what is observed for
Wittenoom and with what is indicated by the most current mechanism studies, see
Sections 7.3.3.4 and 7.3.4.7).
That the two-stage model with v dose-dependent only marginally fits the time-
dependence of the South Carolina data, may further reflect the inadequacy of this
model (because the assumed half-life for asbestos is much too short). Incorporating a
longer half life may improve this fit. In contrast, model fits for Wittenoom in which
longer half-lives were assumed (although still short based on observations reported
from our analysis using the EPA model), do adequately describe the time-dependence
for lung cancer mortality observed in that cohort. As previously indicated, however, the
marginal benefit derived from the two-stage model in this application may not be
6.33
Revision 1 • 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
justified compared to the cost and effort that may be required to obtain the new results.
Especially given that we can now address questions concerning bio-durability with the
adaptation we developed for the EPA lung cancer model (Section 6.2.1.1).
6.2.2 Estimating Kt's from the Published Epidemiology Studies
The EPA model for lung cancer (described in Section 6.2.1.1) was applied to each of
the available epidemiology data sets to obtain study-specific estimates for the lung
cancer risk coefficient, Kt. Based on the results presented in Section 6.2.1.1, there is
no indication that the current model does not provide an adequate description of lung
cancer mortality, at least in association with exposure to amphiboles. There is some
indication that the current model may not provide an adequate description of the time-
dependence of mortality following exposure to chrysotile. However, it is believed that
the effect that such deviations might have on estimates of KL derived using the model
are likely to be small, except in rare cases when studies incorporate long periods of
followup after cessation of exposure. It will be prudent to evaluate the observed
deviations further (if additional raw data sets can be obtained) and to adapt the current
model to account for the observed effect, should it be confirmed in other studies.
However, at this point in time, we see no reason not to proceed with the analysis of
published studies using the existing model.
The set of KL values derived from available epidemiology studies are presented in Table
6-12. Note that this table is a reproduction of Table A-1 in Appendix A.
In Table 6-12, Column 1 lists the fiber types for the various studies, Column 2 lists the
exposure settings (industry type), and Column 3 indicates the specific locations studied.
Column 4 presents the best estimate of each value derived for studies, as reported
in the original 1986 Health Effects Update and Column 5 presents the reference for
each respective study. Columns 6, 7, and 8 present, respectively: the best-estimates
for the Kl values derived for all of the studies currently available (including studies
corresponding to those in the original Health Effects Update); the estimated lower and
upper bounds for each Kt value (derived as described in Appendix A); and the
reference for the respective study from which the data were derived. To assure
comparability across studies, values for all studies (even those that have not been
updated since their inclusion in the 1986 Health Effects Update) were re-derived using
the modified procedures described in Appendix A. The remaining columns of the table
present information relevant to risk coefficients for mesothelioma and are discussed in
Section 6.3.2.
The Ki values derived in this study and the corresponding values derived in the original
1986 Health Effects Update generally agree; most vary be no more than a factor of 2, a
couple vary by a factor of 3, and one varies by a factor of about 5. In no case,
6.34
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
however, are the differences significant (based on visual comparison of confidence
intervals).
6.35
Revision 1 • 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 6-12
6.36
Revision 1 - 9/2/01
-------
I 12
Lung Csnc+r Oom-Nmpodw CMfflclmls |Kt» and Mnolhtlloma Doa^nponM Co«fncl*nt» (K^'•) D*rtv»d from PubNshtd Epidemiology StutflM
that Tjpa Op.nmn
Cohort
EPA (IMC)
K,i100 Rail alio
TMa u|Klala
K^ilOO Rang* IManno
EPA (1994)
K^i10* Ra(M*nc*
TMa
K^ulo* Range
update
Reference
CtvyMM* Mnki|niMMng
OubKntainf
into
006 McDonald d d.. (1900)
0029(0 0085.011) Uddeldd (1997)
017 MdntMn « d. (1979)
Asbestos, Ousbac
THedtadMkies.
Qiatac
0 013(0 003.0049)
0 021 (0 0065.0065)
Udddldd (1997)(rawdda)
Udddldd (1997) (raw dda)
Frtcfcon Product!
Cdaanl nMnufadura
Kalian mne and m*
Cormedctf plant
Mew (Mem ptants
South CvoCntpM
0081 PWdtodd (1990)
001 McDondd d d. (1964)
2.8 Damant eld., (1983)
0035(0. 1.1)
0 (0. 22)
04 (0.16)
2 4 (0 81. 5.6)
Piddtodd. (1990)
McDonald el d. (1984)
Hughes aid (1987)
Dement d d. (1994 raw dda)
0 (0.0 9)
02(0028.14)
011 (0 037.0 25)
McOendddd (1984)
Hughes d d (1987)
per camm Dement (2001)
25 McDonald el 4., (1983a)
1 15(022.4 9)
McOonddatd (1983a)
0 088(0 002. 18)
ft
ft
l
CnCkMta IMngandMKng
WMsnoan, Aualrafa
047 (0.08.8.9)
DdOark. unpiMshed data
7 9 (3 5,18)
DeKlarii. unpubN»h«iJ(Wa
1
i
PaBervn. MJ fedary
Tytar. Tub factory
4 3SaMman(1984)
28 (0 17.27)
013(0.68)
Seidman (1986)
letrfneld (19B8)
32 Seidmai(1B84)
39(0 74.20)
S«idman (100S)
Tin Mill Wamlcitfe mines Ink
Ltfcy. Moitone
081 10 04.5 3)
O H (0 025.4 7)
Amandus and Wheeiar (1987)
McDondd et d. (1988)
!>¦) FnckonProdudt
Camanl manufacture
BrtKsMadory
OnUnotacWy
0-0S8 Batty and Mw^wun (1083)
48 Finkdstain (1983)
0 058 (0.24)
12(0.29)
Beny and Newhouse (1983)
Frtehlwn (1984)
12 Finkefctein (1983)
18(2 1,149)
Fintobtein (1WM)
New Ortarra plants
0 S3 Wai el al. (1979). We* (1984)
04 (0. 16)
Hughaadd (1087)
03(0062. 1 4)
Hughes el ri (1967)
SnadWipM
007(0.259)
AJbiidd (1990)
Factory mftasrt
Batgun factory
U.S. uliw
049 Henderson and Enlerfne (1879)
0 (0. DM)
015(0011, 1)
Laquetdd (1980)
Enterfnedd (1906)
MauMon appicdion
Asbestos, Quebac
US. ImMien wortm
Ptnmylvnt*mt
RnMA England
pw
0 75 SeBtaflel d. (1979)
1.4 McOondddd. (1983b)
11 Pels (1980)
0 4(0012. 51)
18(0 073.185)
0 9 (0 18. 5)
S&Mt and Seidman (19911
McDondd d d (1983b)
Palo (1985)
SdkoHeld
15(1979)
Prfo(1960) Pdo
1 el d (1082)
0 092(0 02. 0 35)
1 3 (0 25. 6 5)
1 1 (0 13 8 8)
1 31 (0 78 5 9)
Lktteiatri (1997)(imwcMn)
Stfcdf and Setirnan(19B1)
McDonaMtf M (1M3b)
PWoftW)
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Perhaps the most interesting of the changes between the 1986 value estimates and
the current value estimates involves the friction products plant in Connecticut
(McDonald et al. 1984). Although a relatively small, positive dose-response was
estimated from this study in the 1986 Health Effects Update, the best current estimate
is that this is essentially a negative study (no excess risk attributable to asbestos). The
difference derives primarily from allowing a to vary in the current analysis. The
exposure groups in this cohort do not exhibit a monotonically increasing dose-response
relationship and, in fact, the highest response is observed among the group with the
lowest overall exposure. Thus, the most likely explanation for the observed relative risk
in this cohort is that there is poor agreement in the background cancer rates between
the study cohort and the control population. As indicated in Appendix A, lack of a
monotonically increasing dose-response relationship is a problem observed in several
of the studies evaluated.
Among the Kt values derived in the current study, the lowest and highest of the best-
estimate values differ by a factor of 90 (excluding the two negative studies, which would
make the spread even larger) and several of the pair-wise comparisons across this
spectrum are significant (based on a comparison of their confidence intervals). For
example, the value derived for the chrysotile miners in Quebec is significantly
smaller than the KL values derived for chrysotile textile workers (in either of the two
studies for South Carolina, which are of highly redundant cohorts in the same plant), for
textile workers in the Roachdale, UK plant, and amosite insulation manufacturers (in the
Seidman study).
The Kl values and the associated confidence bounds derived in the current study are
plotted in Figure 6-3. Each exposure environment is plotted along the X-axis of the
figure and is labeled with a four-digit code that indicates fiber type (chrysotile, mixed,
crocidolite, or tremolite), industry (mining, friction products, asbestos-cement pipe,
textiles, insulation manufacturing, or insulation application); and a two-digit numerical
value indicating the study from which the data were derived. A key is also provided. In
the Figure, the chrysotile studies are grouped on the left, amphibole studies are
grouped on the right, and mixed studies are in the middle. Note also that studies
conducted in the same facility (generally among highly overlapping cohorts), such as
the Dement et al. (1994) and the McDonald et al. (1983a) studies of the same South
Carolina textile facility, are combined (averaged) and presented as a single entry in the
Figure. The two Libby Vermiculite studies were similarly combined.
6.37
RevWon 1-9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Figure 6-3
6.38
Revision 1 • 9/2/01
-------
Figure 6-3:
Plot of Estimated KL Values and Associated Confidence Intervals by Study
Environment
Ol
O
A
-J
¥
1000
100 -
10 -
1
0.1 - T
0.01
0.001
-------
Key for Figure 6-3 Code
Fiber Types
(First Digit of Code)
Study Environment
(Second Digit of Code)
A s amosite
C = chrysotile
M = mixed fibers
R = crocidolite
T = tremolite (in vermiculite)
A = insulation application
F = friction products manufacturing
I = insulation manufacturing
M = mining
P = ac pipe manufacturing
T = textile manufacturing
X = misc. products manufacturing
Study Cohorts
(Last 2 digits)
1 = Quebec miners (Lidded et al. 1997)
2 = Quebec miners (Liddell et al. 1997, raw data)
3 = Italian miners
4 = Connecticut friction product workers (McDonald et al. 1984)
5 = New Orleans ac pipe manufacturers (Hughs et al. 1987)
6 = South Carolina textile manufacturers (Dement etal. 1994,raw data)
7 = British friction product manufacturers (Berry and Newhouse 1983)
8 = Ontario ac pipe manufacturers (Finkelstein 1984)
9 = New Orleans ac pipe manufacturers (Hughes et al. 1987)
10 = Swedish ac pipe manufacturers (Albin et al. 1990)
11 = Belgium ac pipe manufacturers (Laque et al. 1980)
13 * Retired factory workers (Enterline et al. 1986)
14 = Factory workers (Liddell et al. 1997)
15 = Insulation appliers (Selikoff and Seidman 1991)
16 = Pennsylvania textile workers (Mcdonald et al. 1983b)
17 = British textile workers (Peto 1985)
18 « Australian crocidolite miners (DeKlerk, unpublished, raw data)
19 « New Jersey insulation manufacturers (Seidman 1986)
20 = Texas insulation manufacturers (Levine et ai. 1998)
21 = Libby vermiculite miners (Amandus and Wheeler 1987)
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Key for Figure 6-3
6.38
Revision 1 • 8/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Comparisons of KL values across the available studies are instructive. Within chrysotile
studies alone, lowest and highest values vary by approximately a factor of 80 (again,
excluding the negative friction products study), which is little different than the range
observed across all studies. Moreover, as indicated above, the differences between
the lowest (non-zero) value (for Quebec miners) and the highest value (textile workers)
is significant. Differences between the negative friction product study and the other
estimates for chrysotile are not significant, primarily due to the wide confidence interval
associated with the negative study.
Among the apparent variations, differences in lung cancer potency observed among
Quebec miners versus that observed among South Carolina textile workers has been
the subject of much discussion and evaluation, which is worthy of review (Section
6.2.3). In fact, the inability to reconcile these differences, appears to be among the
biggest obstacles to reliably estimating an overall chyrsotile potency factor for lung
cancer.
Among "pure" amphibole studies, the lowest and highest of the best-estimate KL values
vary by a factor of approximately 20 and, interestingly, the two extremes both derive
from within the same industry (amosite insulation). However, these two estimates for
amosite insulation are not significantly different (based on a comparison of their
confidence intervals) Taking the arithmetic mean of the values for the two amosite
insulation studies, results indicate that KL values estimated, respectively, for crocidolite
mining, amosite insulation manufacture, and mining of vermiculite contaminated with
tremolite vary by less than a factor of 3, which constitutes stellar agreement (given the
magnitude of the associated confidence intervals, as shown in Figure 6-3), if this is
indeed the appropriate interpretation. However, alternate interpretations cannot be
excluded.
As indicated in Section 6.2.3, for example, it is possible that mining studies tend to
exhibit low Ki values relative to studies of asbestos products industries. As indicated in
that section, this may be due to the presence of large numbers of cleavage fragments
in the dusts, which may not contribute to biological activity (because the majority of
these may not exhibit the requisite size to be biologically active) but which,
nevertheless, are included in estimates of asbestos concentrations in the original
epidemiology studies. Under this interpretation, the KL estimates for Libby and for
Wittenoom might both be raised substantially, if they could be adjusted for similar
effects, so that they are closer in value to that obtained from the Seidman study. Then
the "outlier" Levin study might be re-evaluated to see if inadequate consideration of lag,
the relatively young age of the cohort studied, or other factors might contribute to a low
Kt estimate for this study. As indicated in Section 6.2.3, such considerations have
implications both for identifying appropriate size ranges of structures to be included In
the analysis of asbestos and for evaluating the relative potency of chrysotile and the
6.40
Revision 1 • 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
amphiboles. Of course, because the difference between the Kt values from the Levin
et al. study and the Seidman study do not appear significant, such differences may
simply be due to unavoidable statistical variation.
It is also instructive to compare variation within and between industries. Within
industries (especially for a single fiber type), the data are limited. The studies from the
two chrysotile mines (Quebec and Italy) show remarkably close agreement, varying by
less than about 20%. However, the studies of the two amosite insulation plants
(Paterson, New Jersey and Tyler, Texas) show values that vary by a factor of 20;
due primarily to the uncertainty of the Kt value for the Tyler facility; these differences
are not significant. As previously mentioned, the same equipment (literally) was used at
both of these insulation plants and the sources of asbestos were also apparently
similar, although the cohorts at the two plants are entirely independent.
The differences between the Kt values from the New Jersey and Texas facilities may
represent the practical limit to agreement within industries that can be expected.
However, it is also possible that an "explanation" for the difference may eventually
present itself.
Across all asbestos types (including mixed), the asbestos-cement pipe industry shows
the greatest variation, including a negative study (best estimate KL - 0) and three
positive studies with values that range amongst themselves by a factor of 17. The
friction products industry includes one negative and one positive study. Better
agreement is observed among textiles. The two mixed textile plants show Kt values
that vary by no more than a factor of 2.5 from each other and from the pure chrysotile
textile plant value. However, given the variation observed within other industry groups,
such agreement may be fortuitous.
Kl estimates within the mining industry (but between fiber types) range over a factor of
20, from a low of 0.029 for Quebec miners to a high of 0.62 for tremolite in vermiculite
miners. Based on inspection of confidence intervals, however, none of the pairwise
differences within this industry appear significant.
6.2.3 The Variation in K^'s Derived for Chrysotile Miners and Chrysotile
Textile Workers
The difference between the observed risk of lung cancer for comparable levels of
chrysotile exposure among Quebec miners (most recent followup: Liddell et al. 1997)
and South Carolina textile workers (McDonald et al. 1983a and Dement et al. 1994) has
been the focus of much attention. Although the discrepancy in lung cancer risk
between Connecticut friction products workers (McDonald et al. 1984) and South
Carolina textile workers is potentially even greater (Table 6-12), the large uncertainties .
6.41
Revision 1 • 8/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
associated with the risk estimate in the friction products study make comparisons with
friction products more difficult to evaluate (Section 6.2.2). Moreover, reasonably good
agreement between results from the Quebec studies and another study of chrysotile
miners in Italy (Piolatto et al. 1990) coupled with reasonably good agreement between
results from the South Carolina Plant and results from two other textile plants: one in
Mannheim, Pennsylvania (McDonald et al. 1983b)and one in Roachdale, England (Peto
1980), suggest that the difference between Quebec and South Carolina may reflect a
general difference between the two industries, rather than specific locations (see Table
6-12 and Section 6.2.2). This appears true despite the fact, for example, that cohorts at
the two other textile plants may have been exposed to mixed fiber types (amphiboles as
well as chrysotile), see Appendix A and Section 6.2.2. Importantly, this "discrepancy"
has been recognized as an issue for lung cancer only, the differences in mesothelioma
rates observed in the Quebec mines and the South Carolina textile plant are not widely
disparate (see Table 6-12).
Three main hypotheses have been advanced to explain the difference in the risk per
unit exposure observed among miners and textile workers (see, for example, Sebastien
et al. 1989). These are:
(1) the low reliability of exposure estimates in the various studies;
(2) differences in fiber size distributions in the two industries (with textile-related
exposures presumably involving greater fractions of longer fibers); or
(3) simultaneous exposure to a co-carcinogen (oil that may have been sprayed on
the asbestos fibers) in the textile industry.
It has also been proposed that differences in the concentration of long tremolite
(amphibole) fibers in dusts from each of the two industries might represent an
explanatory factor (see, for example, McDonald 1998b). However, this would also
require a large relative difference between the potencies of tremolite (amphiboles) and
chrysotile toward the induction of lung cancer. This latter issue is addressed further in
Section 6.2.4. McDonald (1998b) also presents an overview of the current status of
each of the hypotheses described above.
In an attempt to distinguish among the above-listed hypotheses, Sebastien et al. (1989)
conducted a study to determine lung fiber concentrations in tissue samples from
deceased members of the cohorts studied from both the Quebec mines (specifically,
from the Thetford mine) and the South Carolina textile plant. These researchers
ultimately analyzed tissue samples from 72 members of the South Carolina cohort and
89 members of the Thetford (Quebec) cohort. Because the tissue samples came from
cohort members, they could be matched with estimates of the exposure experienced by
6.42
Revision 1 • 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
each of the individuals as well as details concerning the age at first employment, th©
age at death, the years of employment, and the number of years following employment
until death.
In the Sebastien et al. study, tissue samples were obtained in formalin-fixed or paraffin
blocks, which were then digested in bleach, filtered, and analyzed by TEM. Tissue
samples were apparently "opportunistic." Only fibers longer than 5 pm with an aspect
ratio greater than 3:1 were included in the count. For consideration of the limitations
associated with such preparations, see Section 5.2.
Results from matching of tissue samples with the histories of corresponding cohort
members indicate that tissue samples obtained from each cohort covered a broad
range of exposure levels, duration of exposure, and years since the end of exposure.
They also indicate that South Carolina cohort members included in the Sebastien et al.
study experienced, on average, 13.5 years of exposure with 18.1 years between the
end of exposure and death. In contrast, Thetford workers included in this study
experienced an average of 32.6 years of exposure with only 11.6 years between the
end of exposure and death. Corresponding to differences in exposure levels observed
across the two cohorts in the original epidemiology studies, mean exposure levels
experienced by Thetford cohort members included in this study were about 10 times
mean exposure levels experienced by South Carolina workers (19.5 mpcf vs. 1.9 mpcf).
Because Sebastien and coworkers recognized the general lack of a good model
describing the retention and clearance of asbestos fibers in the lungs at the time their
study was conducted, they performed most of their analyses either on pairs of members
(one from each cohort) matched for duration of exposure and time since end of
exposure or on groups of members from each cohort similarly stratified by duration of
exposure and time since end of exposure.
Results from their study indicate that, overall, lung burdens observed among Thetford
cohort members are substantially higher than those observed among South Carolina
cohort members. Geometric mean lung chrysotile concentrations are reported to be 5.3
and 0.63 fibers/pg dry lung tissue in Thetford workers and South Carolina workers,
respectively. Furthermore, despite tremolite representing only a minor contaminant in
the chrysotile from Quebec and the dusts to which the miners were exposed (Sebastien
et al. 1986), the majority of fibers observed in the lungs of Thetford miners were in fact
tremolite (mean concentration 18.4 f/pg dry lung). Since the raw material used in the
South Carolina plant came largely from Quebec, tremolite was also expected to be a
minor contaminant in the dusts to which textile workers were exposed. Yet among
these workers also, tremolite represented a substantial fraction of the lung fibers
observed (mean concentration 0.36 f/pg). Thus, the ratio of tremolite concentrations
6.43
Revision 1 • 0/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
observed among Thetford miners and that observed among South Carolina workers
(18.4:0.36 or 51) is even more extreme than the ratio observed for chrysotile (8.4).
To evaluate the first of the above-listed hypotheses, it is instructive to compare the
ratios of chrysotile or tremolite fibers observed in the lungs of deceased workers from
Thetford and South Carolina, respectively, with the overall exposures that each
received. A rough estimate of cumulative exposure for each set of workers in the
Sebastien et al. (1989) study representing each cohort can be derived as the product of
the mean duration of exposure and the mean intensity of exposure. Thus, for example,
mean cumulative exposure in Thetford was 32.6 years x 19.5 mpcf or 635.7 mpcf-yrs.
Similarly, for South Carolina, mean cumulative exposure was 25.65 mpcf-yrs, which
gives a Thetford/South Carolina ratio of 24.8. This presumably represents the relative
cumulative exposure to chrysotile. For tremolite, Sebastien et al. report that, based on
a regression analysis, air concentrations of tremolite were likely only 0.4 times as much
in South Carolina than in Thetford (where they likely averaged 1% of total fibers).
Therefore, the ratio of cumulative exposures to tremolite for the sets of cohort members
studied by Sebastien is likely 62.
Comparing the ratio of Thetford:South Carolina lung burden estimates with the ratios of
the corresponding cumulative exposures, it appears that the chrysotile lung burden ratio
(8.4) is only a third of the ratio predicted based on cumulative exposure (24.8).
However, the ratio of lung tremolite concentrations (51) is much closer to the
corresponding cumulative exposure ratio (62). It thus appears that, although, airborne
concentrations may not closely track the exposures that led to the observed lung
burdens, the overall trend in exposures predicted by airborne measurements is
approximately correct. It is therefore likely that overall exposure concentrations in
Thetford were in fact substantially higher than in South Carolina (in agreement with
airborne measurements). Thus, we concur with Sebastien et al. that the unreliability of
exposure estimates in these two cohorts is unlikely to explain the observed difference in
the risk per unit of exposure observed for each cohort.
Importantly, although the general trend in relative overall exposure levels predicted by
airborne measurements between Thetford and South Carolina appear to have been
confirmed by mean lung fiber concentrations in the Sebastien et al. (1989) study,
estimated exposures appear to do a very poor job of predicting lung burdens for any
particular individual. To demonstrate this, we analyzed the Thetford: South Carolina
ratios of lung chrysotile concentrations and, separately, lung tremolite concentrations
reported by Sebastien et al. for their set of 32 matched pairs of cohort workers to
determine whether trends in these ratios adequately matched trends in the
corresponding estimated airborne exposure level ratios for the same matched pairs. To
do this, we subjected the ratios presented in Table 7 of the Sebastien et al. (1989)
study to a Rank Von Neuman test (Gilbert 1987). Results indicate that trends in
6.44
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
neither lung chrysotile concentration ratios nor lung tremolite concentration ratios can
be predicted by the observed trend in the estimated airborne concentration ratios
among these 32 matched pairs.
There are numerous sources of potential uncertainty that may mask the relationship
between airborne exposure estimates and resulting lung burdens (Section 5.2).
Potentially the largest of these is the variation expected among lung burden estimates
derived from use of "opportunistic" tissue samples, which are not controlled for the
portion of the respiratory tree represented by the sample. Even for samples collected
from adjacent locations in lung parenchyma, observed fiber concentrations may vary
substantially and such variation is magnified between samples taken from different
individuals at locations in the lung that may not in any way correspond to their relative
position in the respiratory tree.
Other potentially important sources of variation that may mask the relationship between
airborne exposure concentrations and resulting lung burden estimates may primarily
involve limitations in the degree to which the airborne estimates from an epidemiology
study represent actual exposures to the individual members of a study cohort (Section
5.1). Thus, such things as: potential differences between individual exposures versus
area concentrations (which are what is typically measured), the adequacy of
extrapolation to the earliest exposures in a cohort (when measurements were generally
not available), or the adequacy of estimating job x time matrices for individual workers
that can then be integrated with work area exposure estimates to derive individual
exposure estimates may all contribute to the uncertainty of exposure estimates.
The second of the above-listed hypotheses, involves potential differences in the size of
structures that may have been present in the airborne concentrations in Thetford and
South Carolina, which may not have been adequately represented by the exposure
measurements. More generally, this is a question of the degree to which measured
exposures in the two environments adequately reflect potential differences in the
character of exposure that relate to biological activity.
Sebastien et al. (1989) considered this second hypothesis by generating and
comparing size distributions for the fibers observed in the lungs of workers from
Thetford and, separately, South Carolina. Importantly, the size distributions for each
cohort were generated by including the first five fibers observed from every member of
that cohort, without regard to the duration of exposure, level of exposure, or time since
exposure experienced by each cohort member. Therefore, the size distributions
obtained are "averaged" over very different time frames during which differing degrees
of fiber retention and clearance will have taken place, each of which potentially alters
the distributions of fiber sizes (Section 7.2). Thus, the two distributions generated are
each actually collections of samples from multiple/varied size distributions (rather than
6.45
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
single distributions) and this likely masks distinctions between the two work
environments. It is therefore not surprising that the authors found relatively little
differences in the two size distributions.
The portion of the generated size distributions that are least likely to have been affected
by the limitations due to the manner in which they are generated (as Sebastien et al.
suggest) is the fraction of tremolite (amphibole) fibers longer than 20 pm. This is
because (1) tremolite fibers (unlike chrysotile) are biodurable and (2) biodurable fibers
longer than approximately 20 pm have been shown to clear from the lung only very
slowly, if at all (Section 7.2). Thus, the Thetford:South Carolina ratio of long tremolite
fibers may provide the best indication of the relative size distributions in the two
environments.
Table 6-13 presents a table of the estimated, relative concentrations of specific lengths
of fibers observed in lung tissue among Thetford miners and South Carolina workers,
respectively. The length category for various fibers is presented in the last column of
the table. The estimated concentrations, presented in Columns 2 (for Thetford) and 3
(for South Carolina) of this table were derived as follows. For the first length category
(L > 5 pm), concentrations are taken directly from Table 5 of the Sebastien et al. paper
(the geometric means are presented). Concentrations for the remaining length
categories were estimated by multiplying the concentrations for this first length category
by the fraction of the size distribution represented by each succeeding length category
(as provided in Table 4 of the Sebastien et al. paper). So that the relative precision of
these concentration estimates can be evaluated, an estimate of the numbers of fibers
included in each length category (from the total used to derive the size distribution in
Table 4 of Sebastien et al.) are provided in Columns 6 (for Thetford) and 7 (for South
Carolina)! respectively. The Thetford:South Carolina ratios of the concentrations of
fibers in each length category (for each fiber type) are provided in Column 5 of the
table.
It is instructive to compare the ratios presented in Table 6-13 to the Thetford:South
Carolina ratios of mean cumulative exposures estimated above for chrysotile and
tremolite among the cohort members included in the Sebastien et al. study (24.8 and
62, respectively). As indicated in Table 6-13, for chrysotile, the ratio remains
approximately constant at about 9 (varying only between 8.4 and 10) for all of the size
ranges reported except the longest. For the longest category (L > 20), however, the
ratio drops to 5. Because fibers longer than 20 pm are expected to be the most
persistent in the body (Section 7.2), it may be that the ratio of 5 best represents the
relative concentration of long chrysotile structures among the two sets of cohort
members.
6.46
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Because this ratio (for the long fibers found in the lung) is only approximately one fifth
of the estimated ratio for the cumulative exposure to chrysotile (24.8), this suggests that
the South Carolina cohort may indeed have been exposed to dusts enriched in long
fibers relative to dusts experienced at Thetford. Because the estimate of this ratio is
based on counts of at least 11 fibers from Thetford and South Carolina, respectively, it
is unlikely that this ratio will vary by more than a factor of two or three (the 95%
confidence interval around 11 fibers, based on a Poisson distribution is 6 to 19).
6.47
Revision 1 • 8/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 6-13
6.48
Revision 1 - 0/2/01
-------
Table 6-13:
Intimated Concentrations of Sized Fibers Observed in the Lungs of Thetford
Miners and South Carolina Textile Workers1
MEAN LUNG CONCENTRATION NUMBER OF FIBERS
titer
South
Ratio:
South
Size Range
Thetford
Carolina
Units
Th/SC
Thetford
Carolina
of Fibers
"Chrys
5.3
0.63
f/pg lung
8.41
371
226
Length > 5 pm
lr»m
18.4
0.38
f/pg lung
48.42
405
175
Chrys
1.73
0.17
f/pg lung
10.00
121
62
Length > 8 pm
Tr»m
3.90
0.091
f/pg lung
42.95
86
42
Chrys
0.59
0.070
f/pg lung
8.41
41
25
Length > 13pm
Trcm
0.72
0.024
f/pg lung
30.46
16
11
Chrys
0.16
0.031
f/pg lung
5.15
11
11
Length > 20 pm
Trent
0.037
0.008
f/pg lung
4.40
1
4
1
Derived from data presented in Tables A and S of Sebastien et al. 1989
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
The trend with tremolite is even more striking. Moreover, as previously indicated,
because tremolite fibers are biodurable, it is the tremolite fibers longer than 20 pm that
may best represent the ratio of long fibers to which these two groups of cohort
members were exposed. The ratios observed among tremolite fibers steadily decrease
from approximately 50 for fibers longer than 5 pm to 4.4 for fibers longer than 20 pm,
although this last value is uncertain (due to it being based on only 1 fiber observed
among Thetford-derived lungs and only 4 fibers among South Carolina-derived lungs).
In fact these data are statistically consistent even with a ratio considerably less than
one, (i.e., with a considerably higher concentration of long tremolite fibers in South
Carolina than in Quebec). Given that the ratio of the original cumulative exposures for
tremolite was estimated to be 62, that the ratio of long tremolite fibers is only 4.4
suggests that dusts in South Carolina may have been highly enriched in long fibers.
Observations that the fibers to which textile workers were exposed were longer and
thinner than those found in mining are further supported by various published size
distributions of fibers determined in air samples collected in these environments (see,
for example, Gibbs and Hwang 1975,1980). Also, as noted in Curmp et al. (1985), the
raw fiber purchased by textile plants was commonly described as the longest grade of
product (see Table 22 of Crump et al.) Size issues are addressed further in Section
6.2.4.
At this point it is worth mentioning some of the potential differences in the
characteristics of mining dusts and textile mill dusts that may affect biological activity,
but that may not be adequately delineated when measuring exposures by PCM (in f/ml)
and almost certainly not delineated when exposures are measured by midget impinger
(in mpcf), see Section 4.3. During the mining of asbestos, only a small fraction of the
rock (generally no more than 10%) that is mined is typically composed of the fibers of
interest.
While the host rock in a mine may be of similar chemical composition, it generally
represents an entirely different crystalline habit. Nevertheless, a large fraction of the
dust that is created during mining is likely composed of fragments from the host rock
and many of these fragments will be of a size that wouJd be included in the particles
counted by midget impinger. Furthermore, at least some fraction of the fragments
created by the crushing and cutting of the host rock will be elongated "cleavage"
fragments (Section 4.0) so that at least some fraction of these may be included even in
PCM counts, despite many of them being either too thick to be respirable or too short or
thick to be biologically active (see Section 7.2). Note, although Sebastien et al.
employed TEM to characterize fibers in their 1989 study, they apparently employed a
fiber definition that was sufficiently broad that they too would have counted large
numbers of structures that may not contribute to biological activity.
6.49
Revision 1 - 0/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
In comparison, the dusts created in a textile factory are likely composed almost
exclusively of true asbestos fibers. The raw material received by the factory will already
have been milled and beneficiated to remove the vast majority of non-fibrous material.
It is therefore, much less likely that extraneous fragments (even cleavage fragments)
exist that might be counted either by midget impinger or PCM. We make this point
because, if this represents the true situation, it would be expected that risk per unit
exposure estimates (i.e. dose-response factors) derived from any mining site, may be
smaller than estimates derived for the same fiber type in occupational environments
where only finished fiber is used. Thus, another interpretation of the variation observed
among estimated KL and KM values for amphiboles (reported in Section 6.2.2 and
Section 6.3.2, respectively) is that the mining values are somewhat low. The
implications of this possibility are discussed further in each respective section.
Note, although the Sebastien et al. paper suggests that (mpcf) exposure estimates from
Thetford and South Carolina grossly suggest the relative range of lung burdens
observed, there is too much scatter in the data to determine how closely the air ratios
track the lung burden ratios. For example, ratios derived from arithmetic means (rather
than the geometric means) for the Sebastien et al. data are substantially different.
Moreover, as indicated above, there may be substantially different size distributions in
the two environments, which might at least in part be explained by the inclusion of large
numbers of cleavage fragments (with dimensions inappropriate for biological activity) in
the mining environment.
Although the third of the above-listed hypotheses was not addressed by Sebastien and
coworkers, the question of whether a co-carcinogen contributes to the overall observed
lung cancer rate among textile workers has been considered by several other
researchers. To test the hypothesis of whether oils potentially contributed to disease in
South Carolina, Dement and Brown (1994) performed a nested case-control study
among a subset of the cohort members previously studied by Dement et al. (most
recent update, 1994). In this analysis, Dement and Brown qualitatively assessed the
probability of mineral oil exposure for cases and controls based on knowledge of
historic descriptions of mineral oil use. The extent of such exposure was then further
categorized into three strata: none or little, moderate, or heavy, based on where each
worker was longest employed. Cases and controls were then further categorized based
on years at risk and level of asbestos exposure. Results from this nested analysis
indicated no significant change in the estimated exposure-response slope for asbestos
after adjusting for mineral oil exposure.
Additional, albeit qualitative, evidence that oils may not represent an adequate
explanation for the relative lung cancer risks observed in mining and textiles is provided
by McDonald (1998b). McDonald suggests that oils were not used in the Roachdale
plant until 1974. Therefore, due to latency, it is unlikely that the use of such oils would
6.50
Revision 1 • 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
have had a substantial impact on the observed lung cancer cases at the point in time
that the study was conducted (Peto 1980).
Taken as a whole, the evidence presented in this Section suggests that the relative
distribution of fiber sizes found in dusts in the textile industry and the mining industry,
respectively, may be the leading hypothesis for explaining the observed differences in
lung cancer risk per unit of exposure between these two industries.
€.2.4 Evaluating Effects Associated with the Character of Exposure
As indicated in Chapter 3(Overview), valid extrapolation of risk coefficients derived from
a reference site for use at a new site (to predict risk) requires that both of the following
two conditions be satisfied:
(1) that asbestos be measured in both environments in an identical manner,
and
(2) that such measurements reflect (or at least remain proportional to) the
characteristics of asbestos exposure that determine biological activity.
Because there is mounting evidence that the manner in which asbestos was measured
in the available epidemiology studies may not adequately reflect the characteristics that
relate to biological activity (Section 7.4 and 7.5), the second of the above two criteria
may not be satisfied when dose-response factors (i.e.Kj/s and KM's) derived from these
studies are used to predict risk in new environments, without first applying some type of
adjustment. Therefore, we explored the possibility that an improved index of exposure
could be identified and that dose-response factors from the available human
epidemiology studies could be adjusted so that they could be matched with
measurements based on an improved exposure index to predict risk.
Based on an extensive evaluation of the broader literature and the results from a series
of supplemental studies (Chapter 7), the primary features required of an improved index
of exposure were identified and an "optimum" exposure index was defined. As
indicated below, however, compromises were required to adapt the risk coefficients
derived from human epidemiology studies to match measurements using the improved
index. Due to the limitations of the data available for adjusting the existing Kt's and
Kw's, an improved, but less than optimum exposure index was defined, which
nevertheless represents the best of adjustments that can be justified, based on existing
data.
6.51
Revision 1 • 8/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Following a description of the procedures employed to adjust the existing K^'s and KM's,
the effects of such adjustments and their appurtenant implications for risk assessment
are explored.
6.2.4.1 Adjusting K,'s for size
As previously indicated, considerations necessary to compare risk coefficients derived
in different exposure settings (or to apply a coefficient to predict risk in a setting
different from the one in which the coefficient was derived) have been elucidated clearly
in a mathematical model (Chesson et al 1989). The consequences of the model
indicate that adjusting the existing risk coefficients so that they reflect asbestos
characteristics that determine biological activity requires knowledge of the fiber size
distributions of the dusts studied in the original epidemiology studies. To the extent
they exist, such data may be used to normalize each of the published risk coefficients
so that they relate to a common exposure index reflecting asbestos characteristics that
determine biological activity.
Adjustment of the existing risk coefficients is necessary because they have all been
derived based primarily on PCM and Ml measurements and neither PCM nor Ml
measurements relate adequately to biological activity (Section 7.5). Thus, analytical
techniques employed at the time these studies were conducted were not capable of
providing asbestos measurements that directly reflect biological activity.
Ideally, the existing risk coefficients should be adjusted so that they are all normalized
to an exposure index such as that defined by Equation 7.12, which has been shown to
adequately reflect the characteristics of asbestos that determine biological activity
(Section 7.5). However, as indicated below, the database available for describing
structure-size distributions for exposures in the original epidemiology studies do not
contain sufficient information to individually characterize frequencies for structures in
separate length categories longer than approximately 10 pm, while Equation 7.12
requires that fibers longer than 40 pm be separately enumerated. Therefore, an ad hoc
exposure index is presented, which represents a compromise between theoretical
needs defined in Chapter 7 (Section 7.5) and the limitations of the available data.
Importantly, the conclusions that can be drawn from a review of the general asbestos
literature (Section 7.5) is that the asbestos fibers that drive risk are almost certainly
longer than 10 pm with potency likely increasing for longer structures, at least up to a
length of 20 pm. Therefore, due to the fact that structures longer than 20 pm have not
been adequately characterized in any of the epidemiologically studied environments, It
is not possible to test optimal exposure indices such as the one defined in the earlier
analysis of animal inhalation studies (Berman et al. 1995). At the same time, it is
apparent that an exposure index focusing on structures longer than 10 pm as the
6.52
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
primary contributors to risk should be superior to the arbitrary index originally defined
for PCM analysis and employed in the original epidemiology studies. Moreover, the
new index proposed here is also limited to structures that are clearly respirable and
includes counting of even the thinnest of respirable structures while the index defined
for PCM does neither (Appendix B).
In the remainder of this section, the K/s derived in Section 6.2.2 (and summarized in
Table 6-12) are normalized (using a series of published size distributions derived from
TEM measurements) to an exposure index that is expected to better reflect biological
activity. The distributions are defined for exposure settings that correspond to several
of the settings evaluated in the epidemiology studies from which the published risk
coefficients were derived. The coefficients are then combined with the asbestos risk
models presented previously (U.S. EPA 1986) and evaluated above (Section 6.2.1) to
derive a set of adjusted KL values (Kj/s), which are appropriately matched for use with
the proposed, new exposure index. The proposed, new index is also defined below.
For the lung cancer model and the associated risk coefficient, "KL," to be applicable
generally, the asbestos concentration, "f," needs to be expressed in an exposure index
that incorporates the characteristics of asbestos that determine risk. Correspondingly,
the risk coefficient, "KL" needs to be normalized to a model in which T is properly
expressed.
As indicated previously, published structure size distributions derived from TEM
measurements are used to adjust the existing KL's by normalizing them to an exposure
index that is expected to better relate to biological activity than the measurements in the
epidemiology studies from which the KL's were obtained.
Assuming that the available TEM size distributions are representative of dust
characteristics for the exposure settings (industries) studied, these were paired with
corresponding epidemiology studies. The TEM size distributions are then used to
convert the exposure measurements used in the epidemiology studies to the new
exposure index that is potentially more characteristic of biological activity. Studies were
paired as indicated in Table 6-14
Only a subset of the TEM size distributions listed in Table 6-14 were actually employed
in the effort to normalize values. To minimize uncertainty introduced by between-
study variability, it was decided to employ distributions from common studies, to the
extent that this could be accomplished without reducing the number of "size-distribution
- K" pairs available for inclusion in the analysis. Also, studies containing the best
documented procedures were favored over those in which the limitations of the
distributions were less clear. Ultimately, the size distributions selected for use (with one
exception) came from only two studies, which were reported in three publications:
6.53
Revision 1 • JW2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Dement and Harris (1979), Gibbs and Hwang (1980), and Hwang and Gibbs (1981). In
the case of the one exception, size distributions for Libby (tremolite asbestos in
vermiculite) were derived from TEM data recently acquired directly from the site.
6.54
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 6-14
€.55
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 6-15 presents bivariate fiber size distributions derived from the published TEM
data that are paired with representative KL values from the corresponding epidemiology
studies. These data were used to adjust KL values in the manner described below.
The procedure for adjusting risk coefficients is straightforward. First, Equation 6.2 is re-
written to reflect the fact that the fiber concentrations, "f," in the original epidemiology
studies were typically measured by PCM:
where:
"CPCM" is the concentration of asbestos structures that are typically included in a
PCM measurement (Section 4.3); and
all other parameters have been previously defined.
The left side of Equation 6.3 is then multiplied by one (expressed as a product of
reciprocal ratios):
where:
^pcme" 's the fraction of asbestos structures in the published size distribution that
are typically included in PCM measurements;
"f0Dl" is the fraction of asbestos structures in the published size distribution that •
represent the optimal exposure index (i.e. the index of exposure that
determines biological activity); and
"Ctot" is the concentration of total structures from which the fractions derive.
Remembering that the product of any fraction and the total concentration of a
distribution is equal to the concentration of that fraction and that CPCME is supposed to
equal CPCW (Section 4.3) and canceling terms accordingly, one is left with a new
equation for risk in which asbestos concentrations are expressed in terms of the optimal
exposure index:
Kt(cPCM)(d/) " RR'1
6.3
(CpCM)(d') - RR- 1 6.4
KL.(coPtXd/) - RR-1
6.5
6.56
Revision 1 - 9/2/01
-------
TABLE 6-14:
CORRELATION BETWEEN PUBLISHED QUANTITATIVE EPIDEMIOLOGY
STUDIES AND AVAILABLE TEM FIBER SIZE DISTRIBUTIONS
Fiber Type Exposure Setting Distribution Reference Epidemiology Reference
Chrysotile
Textiles
Dement and Harris 1979
Cherrieetaf. 1979
Dement etal 1994,1963b
McOonald et af. 1983a. b
Peto 1960 and 1965
Friction Products
Dement and Harris 1979
Marconi et at. 1984
Winer and Coesett 1979
Roberts and Zumwaide 1982
Rood and Scott 1989
Berry and Newhouse 1983
McDonald etal. 1984
Mining and Milling
Gtotos and Hwang 1980
Winer and Cossett 1979
Liddeiltfal. 1997
McDonald et al. 1980
Nicholson ettf. 1979
Pidattoat* 1990
Asbestos Cement Manu.
Dement and Harris 1978
Snyder et al. 1967
Winer and C06sett 1979
Hughs et* 1987
Chrysolite and
Crociddite
Asbestos Cement Manu.
Hwang and Gibbs 1981
Finkeistein 1984
Fmkeistein 1983
Hughs etal. 1987
Weietal. 1979
Wei# etal. 1984
Attn etal. 1990
Crocidolite Mining and Wiling
Gibbs and Hwang 1980
Hwang and Gibbs 1981
Armstrong etal. 1988
DeKleric, unpublished
Amcsite
Insulation Manu.
Dement and Harris 1979
Levineatal. 1990
Seidman 1986
Seidman 1984
Insulation Application
Insulation Ciearwice
Snyder etal. 1987
Chenieetal. 1979
Selikaretal. 1979
Tremoiite Vemnculite Mining
EPA, unpublished
McDonald etal. 1986
Amandus and Wheeler 1987
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
where:
"K^* is now the adjusted risk coefficient, defined (and derived) as described in
Equation 6.6:
Ki. ¦ K,
PCME
Oot
6.6
where all terms have been previously defined.
The adjusted risk coefficients, the "KL.s", derived from the data provided in Table 6-15 in
the manner described above are presented in Table 6-16. Adjusted risk coefficients for
mesothelioma, "K^-'s", are also provided. K^-'s and K^.'s are derived in precisely the
same manner. The model adopted for mesothelioma risk is described in the next
section.
Ideally, would be set equal to the concentration of asbestos structures defined as
described in Equation 7.12. However, as previously described, published size
distributions cannot be used to adjust risk coefficients to be consistent with the
exposure index described by Equation 7.12. This is because the longest size category
individually represented in the published distributions is for all structures longer than 10
pm and the exposure index described by Equation 7.12 requires that structures longer
than 40 pm be individually enumerated. Therefore, Equation 7.12 was modified in an
ad hoc fashion to generate the exposure index that is evaluated in this document:
« 0.003CS + 0.997Cl 6.7
w
where:
"CMb" is the concentration of asbestos to be used to estimate risk;
"Cs" is the concentration of asbestos structures between 5 and 10 pm in length
that are also thinner than 0.5 |Jm; and
"CL" is the concentration of asbestos structures longer than 10 |Jm that are
also thinner than 0.5 |Jm.
Although not optimized, as previously indicated, this index is nevertheless expected to
represent a substantial improvement over the PCM-related index in current use,
6.57
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
because it reflects more of the general criteria for biological activity defined in Section
7.5.
6.58
Revision 1 • 9/2/01
-------
Table 6-15:
Reprasntative KL and KM Values Paired with Averaged TEM Fiber Size Distributions From Published Papers
C«*
ML
MM
5b« Dtetrl
button*:
Rtlirmci
PC ME
• <•1
L<99 9.4
L<»S
-------
Table 6-16:
Comparison of Original and Adjusted KL and KM Values
Study
Environment
Code
KL Reference
KM reference
KL
KM KL*
KM*
Quebec mines and mills
CM1
Liddell et al. (1997)
0.029
0.11
Quebec mines
Italian mine and mill
CM1
CM2
Piolatto et al. (1990)
Liddell et a I. (1997) (raw data)
0.035
0.017
0.13
0.062
Connecticut plant
CF3
McDonald et al. (1984)
McDonald et al. (1984)
0
0
0
0
New Orleans plants
CP4
Hughes et al. (1987)
Hughes et al. (1987)
0.4
0.2
0.86
0.43
South Carolina plant
CT5
Dement et al. (1994 raw de per. comm. Dement (2001)
2.4
0.25
6.4
0.66
Wittenoom. Australia
RM14
DeKlerk, unpublished data DeKlerk, unpublished data
0.47
7.9
0.87
14
Patterson, NJ factory
Tyler, Texas factory
AI15
AI16
Seidman (1986)
Levin et al. (1998)
Seidman (1986)
2.6
0.13
3.9
17.4
0.87
26
Libby, Montana
British factory
TM17
MF5.5
Amandus and Wheeler (1987)
Berry and Newhouse (1983)
0.61
0.058
2.4
0.13
Ontario factory
New Orleans plants
Swedish plant
Belgium factory
MPS
MP7
MB
M9
Finkelstein (1984)
Hughes et al. (1987)
Albin et al. (1990)
Laquet et al. (1980)
Finkelstein (1984)
Hughes et al. (1987)
1.2
0.4
0.07
0
18
0.3
3.2
1.1
0.19
49
0.81
U.S. retirees
Asbestos, Quebec
M10
MF
Enterllne et al. (1886)
Liddell et al, (1997)(raw data)
0.15
0.092
U.S. insulation workers
MI11
Selikoff and Seidman (199
Selikoff and Seidman (1991)
0.4
1.3
7.4
24
Pennsylvania plant MT12
Rochedale, England plant MT13
McDonald et al. (1983b)
Peto (1985)
McDonald et al. (1983b)
Peto (1985)
1.8
0.9
1.1
1.31
4.8
2.4
2.9
3.5
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
The justification for selecting the above exposure index and the potential limitations
associated with its use are described in Appendix B. Until such time that a study is
completed in which size distributions are generated that allows use of an exposure
index incorporating consideration of longer structures (such as that defined by
Equation 7.12), it is recommended that asbestos concentrations be defined in terms of
the exposure index defined by Equation 6.7 when evaluating asbestos risks. We
therefore recommend that this index be incorporated into an interim protocol for
conducting asbestos risk assessments.
fi 2.4.2 Evaluating the implications of adjusting K, .'s for size
To compare the effects of adjusting Kt values for fiber size, the adjusted values (the
KL.'s), are plotted (along with their associated confidence intervals) in Figure 6-4, which
exhibits a format identical to that of Figure 6-3, to facilitate direct comparison. The "key"
provided in Figure 6-3 is also directly transferable to Figure 6-4.
At first glance, the two figures look quite similar. However, closer inspection indicates
that: (1) two of the points plotted in Figure 6-3 (for MP11 and MX13) are missing in
Figure 6-4; (2) the confidence intervals are larger in Figure 6-4 than in Figure 6-3; and
(3) the relative location of the best estimate values (central points within each
confidence interval) have changed slightly relative to one another. The implications of
these observations are discussed below.
Regarding Figure 6-4, the two points missing (for MP11 and MX13) were omitted
because it was felt that none of the available size distributions used for adjusting the KL
values were suitably applicable for these study sites, so no conversions were possible.
The confidence intervals are larger in Figure 6-4 because, as previously indicated, we
have attempted throughout our analysis of the epidemiology data to account for major
sources of uncertainty. Thus, the confidence intervals depicted in Figure 6-4 are
modified from those depicted in Figure 6-3 to account for the uncertainty of making the
adjustment for fiber size using paired data from published size distributions. The
intervals were adjusted by multiplying the upper bound and dividing the lower bound by
a factor thought to represent the relative contribution to uncertainty contributed by the
need to match data from separate studies to perform the conversions. The factors
employed are provided in Table 6-17.
The visual impressions from a comparison between Figures 6-3 and 6-4 regarding
changes in the relative locations of the central values (best estimate) Kt and adjusted
Kl values are subtle. Thus, numerical representations are provided in Table 6-18.
6.61
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Figure 6-4
6.62
Revision 1 - 9/2/01
-------
1000
Figure 6-4:
Plot of Adjusted KL Values (KL*'s) and Associated Confidence Intervals by
Study Environment
CM
o
"O
B
V)
3
-o
<
100 -
10 -
1 -
2. 0.1 -
0.01 -
0.001
c
M
1
c
M
3
C
F
4
C
P
5
C
T
8
M
F
7
M
P
8
M
P
9
M M
P P
1 1
0 1
M
X
1
3
M M
A T
1 1
5 6
M
T
1
7
R A
M I
1 1
8 9
A
1
2
0
T
M
2
1
-------
Table 6-17:
Estimated Uncertainty Assigned to Adjustment for Fiber Size
Study
8tudy Location
Coda
Quebec mines and mMe
CM1
Quebec mines
CM2
Italian mine end mW
CM3
Connecticut piert
CF4
New Orleans plants
CPS
South Carolna plant
CT6
British factory
MF7
Ontario factory
MPS
New Orleans plants
MPS
Swedish plant
MP10
Belgium factory
MF11
U.S. retiree*
MF13
AftbMtoft, Quebec
MF14
U.S. insulation workers
MI15
Pennsylvania plant
MT16
Rochedale, England plant
MT17
Whtanoom, Australia
RM18
Patterson, MJ factory
AI10
Tyler. Texas factory
AI20
Ubby, Montana
TM21
Estimated
Uncertainty
Factor Explanation
1 Location common to epidemiology study and size study
1 Location common to epidemiology study and sizs study
1 75 Soma industry, separata locations for epidemiology and size studies
1.25 Epidemiology location one of several combined for size study
1.25 Epidemiology location one of several combined for size study
125 Epidemiology location one of se\«ral combined for size study
1 5 Same Industry, separate locations fa epidemiology and size studies
15 Epidemiology location probably one of several combined for size study
2 Same industry, separata locations, mixed exposures
2 Same industry, separate locations, mixed exposures
2 Generally similar industries studied for epidemiology and size
2 Same Industry, separata locations, mixed e^MSUtes
2 Same Industry, separate locations, mixed exposures
1.75 Same industry, separata locations for epidemiology and size studies
1.25 Epidemiology location one of several combined for size study
1.25 Epidemiology location one of several combined for size study
1.75 Extrapolated from limited, marginally associated air data
KL Reference
Uddeaetal. (1987)
Pioiatto eld. (1990)
McDonald et al. (1984)
Hughes etal. (1987)
Dement et al. (1894 raw data)
Berry and Newhouse (1983)
Finkeistein (1984)
Hughes et at. (1987)
AMnetal. (1990)
Laquetetal. (1980)
Enteriineetal. (1966)
SeWtoff and Seldman (1991)
McDonald etal. (1983b)
Peto(1965)
DeKlerk, unpublished data
Seldman (1986)
Levin etal. (1998)
Amandus and Wheeler (1987)
KM reference
UddeH et el. (1997) (raw data Loc.
McDonald etal. (1984)
Hughes et al. (1987)
per. comm. Dement (2001)
Finkeistein (1984)
Hughes et al. (1987)
Lidded et al. (1997)(raw dale)
SelikoW and Seldman (1991)
McDonald et al. (1983b)
Peto (1985)
DeKlerfc, unpublished data
Seldman (1986)
-------
TABLE 6-18:
COMPARISON OF SPREAD IN RANGE OF ORIGINAL AND ADJUSTED KL AND
K„ VALUES FOR SPECIFIC FIBER TYPES
Fiber Type Ranges of Values
Number Kt. Number K* K..
in Ki, in K*,
Kt. Sets
K,*. Sets
All Tiber types combined
15
90
160
10
1089
789
Chrysotile only (excluding mixed)
4
83
59
3
15
11
Chrysotile and mixed settings
11
83
68
8
1089
789
Amphibole and mixed settings
11
90
160
7
1089
789
Amphiboles only (excluding mixed)
4
20
20
2
2
2
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 6-18 presents the magnitude of the spread in the range of original and adjusted
Kl values (and also original and adjusted Kw values) estimated as the quotient of the
maximum and minimum values of each range. Note that, of necessity, such an analysis
requires that the zero values obtained for the Connecticut friction products plant (CF4)
be omitted. Note further that the data sets evaluated in Table 6-18 for the original and
adjusted values are identical (i.e., the two studies for which no suitable conversion
factor could be found were excluded).
In Table 6-18, it is apparent that the spread in values among "pure" fiber types (i.e.
chrysotile only or amphibole only) are both smaller than those for mixed data sets
(containing both fiber types). This suggests that there may be differences in the
potencies of each fiber type toward each respective disease (lung cancer and
mesothelioma). In fact, the effect appears to be even more extreme among
mesothelioma values than among lung cancer values (discussed further in
Section 6.3.3.2). Moreover, that the spread in adjusted values for the data sets
containing only the pure fiber types are smaller than for the unadjusted values, but (at
least for lung cancer) larger than the unadjusted values for mixed data sets suggests
that the effects from adjusting the KL (and KM) values for fiber size may be confounded
with differences in the potency of fiber types. Therefore, we endeavored to develop a
procedure for evaluating these data that would simultaneously account for both fiber
size and type.
To address these issues, we considered that the mixed exposure environments
represent situations in which there would be combined contributions to risk from both
amphiboles and chrysotile. It is also known that chrysotile from the Quebec mines
contains small amounts of tremolite (amphibole) asbestos (Sebastien et al. 1986). We
therefore set up .the following formalism to separate the effects of chrysotile asbestos
from the effects of the amphiboles. Note, based on results of a previous study (Berman
et al. 1995) and lack of compelling evidence elsewhere in the literature (assuming that
size effects are adequately addressed, see Chapter 7), we assume in this formalism
that all similarly-sized amphibole fibers are equipotent.
We start by defining potency factors for truly pure fiber types: for pure amphiboles
and Kte for pure chrysotile. Next we define the relative potency of chrysotile, "RPC" as
the ratio of these factors. Thus:
KU=RPC*KU 6.8
If one then knows the fraction of amphibole asbestos, in an exposure setting, the
relationship between the observed, overall potency factors (e.g. the Kt's) and these
fiber-specific factors is given by:
6.65
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Kt = + RPC*Kt/(1 .fBmph) 6.9
Note that, if RPC = 1 then Equation 6.9 reduces to the identity KL = and
Equation 6.8 reduces to = K^, meaning that all fibers are equipotent and there is no
distinction for fiber type. Also, if RPC = 0, then chrysotile is non-potent and
Equation 6.9 reduces to Kt = K^'f^. Moreover, because famph is always less than
one, this latter equation means that the true potency of the amphibole fibers is greater
than would be indicated by the overall in a study (whenever RPC = 0).
Finally, solving Equation 6.9 for yields:
K,
K,
IW * RPC, (1 - f8mph)]
6.10
and we further define the denominator on the right side of Equation 6.10 as the
function "Q".
Because the value for should be constant across all study environments, if we also
know the fraction of exposure in each environment contributed by amphiboles, we can
apply Equation 6.10 to each of the available K/s and search for the value of RPC that
minimizes the spread in the resulting Ku values. Furthermore, by applying this same
procedure both to the original KL values and to the size adjusted Kt values (the Kt.'s)
allows simultaneous consideration of fiber size and type. Results of applying this
analysis are presented in Table 6-19.
The results provided in Table 6-19 are based on the estimates of the fraction of
exposures in each environment that is represented by amphiboles presented in
Table 6-20. In Table 6-20, the best estimate of the fraction of amphiboles is provided in
Column 3 along with an estimate of the range (in Column 4), which indicates the degree
of uncertainty that should be associated with each estimate of the fraction of
amphiboles. The source of information from which each estimate was derived is
provided in Column 5. Note that, depending on the particular study environment, such
fractions either represent a time/operations average (e.g. commercial amphiboles were
employed in a certain fraction of the plant for a certain period of time) or a composition
average (e.g. tremolite constitutes some defined fraction of the chrysotile used).
6.66
Revision 1 • 8/2/01
-------
Table 6-19:
Results of an Analysis of the Combined Effects of Fiber Size and Type on the
Original and Adjusted KL values1
Studies
Range
Representing
Optimum
Max/Min of K.
Ratio KJK..
Ratio
Max.Min K,
Number of
Value for
Values at
at Optimum
at
Values at
Data Sets
Studies
RPC
Optimum RPC
RPC
RPC = 1
Optimum RPC
Among Values
Among Values
Among KHa Values
Among KWa. Values
15
15
10
10
0.5
0.35
< 0.00001
< 0.00001
83
59
36
26
0.98
0.72
0.72
0.72
1.8
0.72
CT6:CM1
CT6:CM1
MP8:CM1
MP8:CM1
Excluding "zero" value for Connecticut friction products (CF4)
-------
Tab* t-3*
Estimated Fraction of Amphlbola* m AiUHw (Mi
Fraction of Amphlbolas
Study
Best
Estimated
Study Location
Code
Estimate
Range
Source of Eetfmete
KL Reference
KM reference
(%)
(%)
Quebec mines and mHIa
CM1
1
0-4
Sebastlen el ai. (1988). Extrapolated from air data
UddeMetai. (1997)
Quebec mines
CM2
1
0-4
Sebastlen el al (1986), Extrapolated from air data
llddeii et al. (1997) (raw data L
Itaiian mine and mHi
CM3
0.3
0.1-0 5
CHECK THIS
Piolattoetai. (1990)
ConnecMcut plant
CM
0.5
0-2
McDonald et al. (1984), Extrapolated from planty history
McDonald et el. (1984)
McDonald et al. (1984)
New Orleans plants
CP5
t
0-2
Hughs at ai. (1987), Extrapolated from plant history.
Hughes etal (1987)
Hughes etai. (1987)
South Carolina plant
CT6
0.5
0-2
Extrapolated from Quebec source material:
Dement et al. (1994 raw data)
per. comm. Dement (2001)
Based on Sebastlen et at. (1686)
BrWeh factory
MF7
0.5
0-2
Berry end Newhouse, Extrapolated from plant history
Berry and Newhouse (1983)
Ontario factory
MPS
30
10-50
Flnkelsteln (1984), Extrapolated from plant history
Flnkelstetn (1984)
FinketsMn (1984)
New Orleans plants
MPS
S
2-15
Hughe* et el. (1687), Extrapolated from plant history
Hughes el ai. (1967)
Hughes etal. (1987)
Swedish plant
MP10
3
o-e
Albin et al. (1990), Extrapolated from plant history
Albin etal. (1990)
Belgium factory
MF11
Laquet at al (1980)
U.S. redress
MF13
Entertlneetal. (1988)
Asbestos, Quebec
MF14
UddeR et at (1997)(raw data)
U.S. twsullfcm workers
MI1S
50
25-75
Guessttmste for broed Industry
Seiikolf and Seidmsn (1991)
SeHkoffand Setdman(199t)
Pemeytvanla plant
MT19
8
3-15
McOoneld et ai. (1983b), Extrapolated from plant history
McDonald el al. (1983b)
McDonald etal (1983b)
Rochedale, England plant
MT17
S
2.5 -15
Peto (1985), Extrapolated trom plant history
Peto (1985)
Peto (1985)
WNtenoom. Australia
RM16
67
95-100
General Estimate*
OeKlerk, unpublished data
DeKlerk, unpublished date
Patterson, N J factory
AltO
97
95-100
General Estimate*
Seidmsn (1988)
Seidman (1988)
Tyler, Texee factory
AI20
87
95-100
General Estimate*
Levin etal. (1998)
Llbby. Montana
TM21
95
90-100
General Estimate'
Amendus and Wheeler (1987)
* Allows for the possibility ol some foreign material Practical effect lor the analysis Is n*. Might just as weH assume 100%.
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
In Table 6-19, the different data sets evaluated (i.e. the Ku's, K^'s. K^'s, and KMa.*s)
are identified in Column 1. The number of studies included in each data set is listed in
Column 2. Column 3 presents the values of the optimized RPC's (the RPC's for which
the spread in Kx< values across each data set is smallest). Column 4 presents the
minimized (optimum) range in the corresponding Kxa values for each data set
(expressed as the ratio of the maximum/minimum values). Column 5 presents the
corresponding ratio of the original and adjusted K,, values. Column 6 presents this
same ratio of original and adjusted K„ values with RPC held fixed at one, and
Column 7 indicates which study environments represent the maximum and minimum
values for the optimized range of each data set.
Several conclusions can be drawn from the data presented in Table 6-19. First, it
appears that, for both lung cancer and mesothelioma, the best estimate is that there is
a difference in potency between chrysotile and amphiboles. For lung cancer, chrysotile
may only be one third to one half as potent as the amphiboles, depending on whether
one also adjusts risk factors for fiber size. For mesothelioma, no matter whether risk
factors are adjusted for fiber size or not, it appears that chrysotile is much less potent
than amphiboles and, in fact, may be non-potent. However, because improvement in
the reconciliation of potency estimates across studies becomes marginal for RPC's
smaller than about 1/1000 (and changes are no longer noticeable for RPC's smaller
than about 1/10,000), a conservative estimate suggests that, at most, chrysotile is no
more than about 1/1000 as potent as amphiboles toward the induction of
mesothelioma. Other inferences concerning mesothelioma are addressed further in
Section 6.3.3.2.
Although the improvements in reconciliation across potency estimates for the different
studies are modest when one adjusts for fiber size (the ranges are decreased by about
30%), these reductions are based on spreads over the entire, available data sets (15
studies for lung cancer and 10 studies for mesothelioma). Moreover, such reductions
are achieved using the relatively crude procedure of matching published size
distributions to published epidemiology studies (rather than by obtaining true study-
specific size distributions). Thus, combined with the strong inferences from the rest of
the literature (see Chapter 7) that such size adjustments better reflect the
characteristics of biological activity, adjusting values for fiber size is well justified and
will improve the confidence with which risks can be extrapolated to new sites.
In this analysis, the range in lung cancer potency estimates is reduced from about a
factor of 90 (when data are neither adjusted for fiber size or type, Table 6-18) to 59
(when data are adjusted for both fiber size and type). Interestingly, the extremes of this
range are contributed by South Carolina textiles (CT6 - the maximum value) and
Quebec mining (CM1 - the minimum value). Once adjusted for fiber size and type, the
lung cancer potency estimates from all other studies (including those both for mixed
6.69
Revision 1 • 8/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
exposures and for nominally pure amphibole exposures) falls between these two
extremes. As previously indicated (Section 6.2.3) the differences in potency estimates
in these two environments has been the focus of much attention. Based on the
analysis presented in Section 6.2.3, the best of candidate explanations for the
differences between mining and textile production is the range of sizes of fibers in these
two environments. Further evidence that this is the case is apparent from Table 6-19.
As indicated from a comparison of the range in KL and Kr values in the table, adjusting
values for fiber size (even given the crude approach currently available for making such
adjustments) already reduces the difference in the potency estimates from these two
environments by about 30%.
Although useful, the above analysis of ranges is potentially limited by the need to
exclude the nominally zero value from the Connecticut friction products plant (CF4) and
the lack of formal consideration of uncertainty. Therefore, an additional analysis was
conducted in which uncertainty is addressed formally and the value from the
Connecticut study was included.
In this analysis, we assumed that the log of the measured KL values (or K^, Kw, KM.
values) were normally distributed with mean Ln{Ku*[fart)Ph + rpc*(1-f8mptl)}, where Ku,
RPC, and famptl were all as previously defined. The variance of KL was assumed to be
composed of two or three independent components. The first component, aiP was
assumed to be equal to one-half of the difference between the log transforms of the
upper limit of the confidence interval and the point estimate of (see Table 6-12).
This variance was assumed to represent statistical uncertainty in response and
exposure measurements (i.e, uncertainty that the confidence intervals were designed to
represent, see Appendix A). Note that these values represent the irreducible fraction of
uncertainty in the potency estimates.
A second component, o, of the standard deviation was assumed to be constant for all
studies, and may be thought of as representing the uncertainty in the estimate resulting
from differences in exposure conditions, such as fiber size distributions, which are not
represented in the c(. The overall standard deviation of the Kt from study i was
assumed to be (O,2 + a2)14. As an alternate approach, the uncertainty in f8mpW was
incorporated by adding a third independent component of exposure, given by one-
half of the difference between the log transforms of the upper limit of the confidence
interval for famphi and the best estimate of f8(T)phi (see Table 6-20). Thus, when the
uncertainty in fBmpf, was accounted for, the overafl standard deviation of the KL from
study i was (a* + o2+ a,2)5*.
Using this model, a likelihood test was applied to estimate the parameters, Ln (Ku),
RPC, and o, and to conduct likelihood tests of the hypotheses that chrysotile was non-
potent (RPC=0) or equally potent (RPC = 1) with amphibole. The same test was
6.70
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
applied to the Kr, Kw, and K„. values. When applied to the adjusted (starred)
variables, the uncertainty in the fiber size distribution data was not accounted for so that
the relative effect of adjusting for fiber size could be evaluated.
In this approach, the overall likelihood is estimated as the product of a series of terms
that each represent the likelihood of obtaining a KL value equal to that obtained from
one of each of the available studies, given that they are derived from a common normal
distribution with mean, and standard deviation (of + a2)*, as indicated above. The
optimized (best estimate values) for and 0 are those for which the overall likelihood
is maximized. Results from the analysis with the two parameter standard deviation are
summarized in Table 6-21.
In Table 6-21, the first and fifth columns identify the specific data set being considered
and the size of each data set is also presented. Columns 2 and 6 indicate the specific
variables estimated. Columns 3 and 7 indicate the values for the estimated parameters
with the value for RPC optimized. Columns 4 and 8 indicate the values for the
estimated parameters with the value for RPC fixed at 1 and Columns 5 and 9 indicate
the values for the estimated parameters with the values for RPC fixed at 0.
The results presented in Table 6-21 closely parallel the results obtained from the
analysis of the range of values presented in Table 6-19. The best estimated value of
relative potency for lung cancer indicates that chrysotile is only between a fifth and a
half as potent as the amphiboles toward the induction of lung cancer, depending on
whether potency factors are adjusted for fiber size. The results of a hypothesis test
indicate that this is not significantly different from the hypothesis that chrysotile and
amphiboles are equipotent (in either the original or the size-adjusted data sets).
However, the hypothesis that chrysotile is non-potent toward the induction of lung
cancer is clearly rejected. Size-adjusting the potency values provides a small
improvement (reduction) in the overall variance of the data set of about 3% (although
substantially larger improvements are noted for mesothelioma). Assuming the optimal
difference in potency for the different fiber types, best estimate KL and Kt. values for
chrysotile and amphiboles are also presented in Rows 8 and 9, respectively, of the
table. Although the data are not shown, the analysis using the three parameter
estimate of standard deviation gave entirely comparable results.
Discussion of the results from the analysis of the mesothelioma values (also presented
in the table) is provided in Section 6.3.3.2. Based on this analysis and the rest of the
evaluation described in this chapter, a recommended set of lung cancer (and
mesothelioma) potency factors has been developed and is presented in Section €.4.
6.71
Revision 1 • 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 6-21
6.72
Revision 1 - B/2/01
-------
Table 6-21:
Comparison of Original and Size Adjusted Potency Factors with Simultaneous
Consideration of Fiber Type and Uncertainty
[Standard Deviation = (0* + o2)y,J
Data
Optimized
Data
Optimized
Set
Variable
Values
RPC *1
RPC = 0
Set
Variable
Values
RPC =1
RPC = 0
Kc
Kf
Z
ii
3
RPC
0.407
1
0
(N = 16)
RPC
0.179
1
0
Ln(K,.)
-0.206
-0.836
1.806
LntK,,.)
1.549
0.378
3.087
a
1.076
1.085
1.806
O
1.041
1.106
1.619
LL
-17.6182
-17.9464
-22.2693
LL
-17.516
-18.6264
-21.1145
P value
0.42
0.002
P value
0.14
0.007
100%,
0.81
100%,.
4.71
100*Ku
0.33
100%c.
0.84
km
K„.
(N = 11)
RPC
0.0040
1
0
(N= 11)
RPC
0.0017
1
0
Ln(K^)
2.084
-0.234
2.234
LnflW)
3.358
1.004
3.431
a
0.8123
1.817
0.846
a
0.6561
1.925
0.672
LL
-10.1285
-15.9751
-10.3145
LL
-8.854
-16.638
-8.916
P value
0.0006
0.54
P value
< 0.0001
0.72
10^
8.04
10**K(h,.
28.7
10-K^
0.032
tf'Kuc-
0.050
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
6.3 MESOTHELIOMA
The EPA model used to describe the incidence of mesothelioma in relation to asbestos
exposure is the model proposed in the Airborne Health Assessment Update (U.S. EPA
1986). This model assumes that the incidence of asbestos induced mesothelioma is
independent of age at first exposure and increases according to a power of time from
onset of exposure, as described in the following relationship:
'm = KMf {(T-10)3 - (T-10-d)3] forT > 10+d (6.11)
= Kwf (T-10)3 for 10+d > T > 10
= 0 for 10 > T
where:
"lM" is the mesothelioma mortality observed at "T" years from onset of
exposure to asbestos for duration "d" and concentration T;
"K*/ is the proportionality constant between dose and mesothelioma
response and represents the potency of asbestos: and
all other factors have been previously defined.
This is an absolute risk model, which means that the incidence of mesothelioma
predicted by the model is a direct function of asbestos exposure that does not depend
on the background incidence of the disease. Background mesothelioma cases are rare
in the general population in any case.
The EPA model for mesothelioma is the solution to the following integral equation,
assuming that exposure remains constant:
'm = 3^ - 10- x,2dx (6-12)
This expression is more general than Equation 6.11, as it applies io both constant and
variable exposures (see Appendix A).
6.73
Revision 1 • 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
6.3.1 The Adequacy of the Current EPA Model for Mesothelioma
Access to the raw epidemiology data from a small number of key studies allowed us to
evaluate the adequacy of the EPA model for describing the time-dependence for
mesothelioma in asbestos exposed cohorts. For this analysis, the raw data for the
cohorts for the chrysotile miners in Quebec was graciously provided by Drs. Liddell and
McDonald (described in Liddell et al. 1997), the raw data for the cohort of crocidolite
miners in Whittenoom, Australia was graciously provided by Nick DeKlerk (unpublished)
and the raw data for the cohort of chrysotile textile workers (described by Dement et al.
1994) was graciously provided by Terri Schnorr of NIOSH. The Whittenoom cohort was
originally described by Armstrong et al. (1988), but the data provided by DeKlerk
included additional followup through 1999.
To identify potential effects due to varying procedures, different methods for fitting the
EPA mesothelioma model to epidemiological data were evaluated. In this evaluation,
three methods were used to fit the EPA mesothelioma model to data from Wittenoom.
In the first approach the data were categorized in a manner often available in published
form, so this method mimics the method generally used when raw data are not
available. The observed mesotheliomas and person years of observation were
categorized by time since first exposure, and the mean exposure level and duration of
exposure were calculated for each such category. The EPA model is then applied to
such data using the approach for the typical situation (as described in Appendix A) and
results for the Wittenoom cohort are presented in Table 6-22. The KM value estimated
for Wittenoom using this approach is 7.15E-8 (90% CI: 6.27, 8.11). The fit of this
model to data categorized by time since first exposure is good (p = 0.65).
Note that, for most of the published epidemiology data sets, the average level and
duration of exposure are not available so that these have to be estimated from cruder
data, such as cohort-wide averages.
A second approach to fitting the EPA model to epidemiology data, exploits the fact that
the mesothelioma model (Equation 6-12) expresses the mesothelioma mortality rate as
the product of Kw and an integral involving the exposure pattern, the time of
observation, and the ten-year time lag, but not any variable parameters. The value of
this integral was calculated for each year of followup of each subject. Person-years of
followup and mesothelioma deaths were then categorized according to the values of
the integral, and the average value of the integral determined for each category.
Results are presented in Table 6-23.
6.74
Revision 1 - 9/2/01
-------
Table 6-22:
Fit of EPA Mesothelioma Model to Observed Mortality Among Wittenoom Miners
Time Since First Exposure Duration Concentration Cases Predicted Log
Range (Yrs) Mean (Yre) Likelihood
10.5)
2.5
0.6
30.2
0
0
0.00
15.10)
7.4
0.9
30.2
0
0
0.00
{10. 15 )
124
1
30
1
0.6
-1.14
[ 15. 20)
17.4
1
29.7
5
6.6
-1 95
120, 25 )
224
1
29.5
20
17.6
-2.58
125.30)
274
1
29.3
25
30.1
-2.99
I 30. 40 )
34.1
1
29
90
78.1
-4.04
[ 40. 100)
43.6
1.1
28.1
23
31.1
-3.65
Totals
164
164
-16.35
Goodness of Fit
Chi sq
5.95
P value
0.65
Km x 10'
7.15
90% Confidence Interval
(6.27. 8.11 )
-------
Table 6-23:
Fit of EPA Mesothelioma Model to Observed Mortality Among
Wittenoom Miners Categorized by Cumulative Exposure (Based on
Averaged Value of Integral for Each Category)
Average Value
of Integral
Cases
Predicted
Log
Likelihood
0
0
0
0.00
28.4
0
0
-0.03
206.3
1
0.3
-1.56
732.3
4
1
-4.15
2038.1
10
2.9
-727
5153.7
23
6.8
-14.20
12385.8
32
13.8
-11.40
30639.7
27
25.8
-2.60
144801
67
113.3
-14.14
Totals
Goodness of Fit
ChiSq
P Value
164
164
79.8
1.5E-14
-55.35
K« x 10*
90% Confidence
Interval
9.00
(7.89,10.2)
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Km was estimated by maximum likelihood from Table 6-23 assuming that the observed
numbers of cancer deaths were independently Poisson distributed with a mean equal to
the mean value of the integral for that category times KM. The value of KM obtained in
this fashion was Kv = 9.00E-8 (90% CI: 7.89,10.2). The fit of the model to data
categorized by the integral is not good (P < 0.00001), as the model predicts too few
mesotheliomas for small values of the integral and too many mesotheliomas for large
values of the integral. However, the estimates for KM obtained from applying the model
in this manner are still very similar to those obtained using the exact method (described
below).
A third method of estimating KM (termed the "exact method") employs a likelihood that
does not involve any categorization of data. With this method, the hazard function, h(t)
= lM(t) and the corresponding survival function (probability of surviving to age t without
death from mesothelioma in the absence of competing causes of death),
S(t) = exp(-J0' h(s)ds, are computed for time at the end of followup. The contribution to
the likelihood of a subject who died of mesothelioma t, years after beginning of followup
is h(t,)*S(t1), and the contribution of a subject whose followup was not terminated by
death from mesothelioma is S(t,). The complete likelihood is the product of such terms
over all members of the cohort. The estimate of KM obtained by maximizing the
logarithm of this likelihood was 7.95E-8 (90% CI: 7.0, 9.0).
We consider the exact method of computing Kw to be the most accurate and results
from this method are reported in the summary tables (see, for example, Table 6-12).
The Quebec and South Carolina data sets were thus evaluated using the exact method.
However, it is noteworthy that the two other methods described above, one of which is
often applied to published data, give similar estimates of Kv, at least for this data set.
The Quebec cohort was subdivided into three subcohorts, believed to correspond to
differing amounts of amphibole exposure due to tremolite contamination within the
chrysotile (Liddell et al. 1997). Location I consisted of workers at the mine at Asbestos
where the ore reportedly had less tremolite contamination. Locations 3 and 4 consisted
of workers at the large central mine and at smaller mines, respectively, near Thedford,
where the ore was more heavily contaminated with tremolite. Location 2 consisted of
workers at an asbestos products factory at Asbestos, which processed some
commercial amphibole fibers in addition to chrysotile. The exact method of calculating
Kw produced the following estimates: Location 1 (8 cases): Kw = 1.3E-10 (90% CI:
0.3E-10, 4.9E-10); Location 2 (5 cases): 9.2E-10, (90% CI: 2.0E-10,35E-10); Locations
3 and 4 (22 cases): KM = 2.1E-10 (95% CI: 0.65E-10, 6.5E-10. The relative magnitudes
of these estimates track with the relative amounts of amphibole exposure estimated for
these locations (Liddell et al. 1997), which is consistent with the hypothesis that the
mesothelioma risk in this cohort is due, at least in large measure, to exposure to
amphiboles.
6.77
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
There were only two confirmed mesothelioma deaths in the South Carolina cohort and
four additional suspected deaths. These were too few to permit detailed analysis.
Based on both confirmed and suspected mesothelioma deaths, the exact method of
analysis gave an estimate for KM of Kw = 0.43x10"8, 90% CI: (0.20x10"®, 0.79x10"8).
Using only the two confirmed mesotheliomas, the same analysis yielded Kw =
0.14x10"®, 90% CI: (0.034x10"®, 0.38x10"®). The same estimates were obtained by
estimating KM from data categorized by time since first exposure and fitting a linear
model to the categorized value of the integral in the definition of the EPA model. Thus,
for this cohort (in parallel with the cohort from Wittenoom) comparable Kw values are
estimated no matter which of the three methods described above are used for fitting the
EPA mesothelioma model to the epidemiology data.
6.3.1.1 Time dependence
The EPA mesothelioma model was next evaluated to determine whether it adequately
describes the time-dependence of mesothelioma mortality following cessation of
exposure in the Wittenoom and Quebec cohorts. The small number of mesotheliomas
observed among the South Carolina cohort precluded our completing this analysis for
that cohort. The model predicts that risk rises as the difference between the third
power of time since the start of exposure and the third power of time since the
cessation of exposure (both lagged by 10 years).
Table 6-24 shows the fit of the EPA mesothelioma model to Wittenoom data
characterized by time since last exposure, based on the KM estimated from the exact
analysis. Although this fit appears adequate (P = 0.21), there appears to be a slight
tendency for the EPA model to under-predict the mesothelioma rate at long times
following cessation of exposure. However, given that the EPA model does not appear
to under-predict the mesothelioma rate at long times for the other two cohorts evaluated
(see below), it is not clear that this is a real limitation in the model. At the same time,
the number of mesotheliomas in the other two cohorts are relatively small. Moreover,
both of these other cohorts nominally involve chrysotile exposure (as opposed to
crocidolite). Therefore, it may prove useful to obtain additional cohorts (preferably with
relatively large numbers of mesotheliomas and preferably involving amphibole
exposure) before concluding definitively whether the current EPA mesothelioma model
adequately describes the time dependence for mesothelioma.
Table 6-25 shows the observed and expected mesotheliomas in the three separate
locations at Quebec, categorized by time since last exposure, and computed using the
Ku values obtained using the exact fitting method. Although the small numbers of
mesotheliomas make it difficult to draw definite conclusions about the adequacy of the
model, there is little evidence that the model under or over-predicts the numbers of
mesotheliomas at long periods after the end of exposure.
6.78
Revision 1 • 9/2/01
-------
Table 6-26:
Fit of EPA Mesothelioma Model to Observed Mortality Among Quebec Miners In Each of Three Mining Areas and
Time Since
Averaged
Last Exposure Predicted
Cases
Value of
Relative Rate
Log
K„
Goodness of Fit
Range (YrsJ
the Integral
Risk
Likelihood
ChlSq P-value
Location 1
10.10)
3.1
3
94267.5
0.98 3.90E-05
-1.50
1.34E-10
4.83 0.44
[10,20)
1.4
2
119190
1.47 7.30E-05
-1.44
(20,30)
1.2
0
119971
0 O.OOE+OO
-1.15
130,40)
1.4
1
181470
0.72 5.50E-05
-1.06
140.50)
1.4
2
278118
1.44 1.70E-04
-1.43
50+
1.0
0
339160
0 0.00E+00
-0.95
30+
3.7
3
0.80429
Totals
9.3
8
0.860215
-7.52
Location 2
10.10)
2.6
0
56377.5
0 0.00E+00
-2.55
9.55E-10
12.33 0.03
110.20)
0.8
2
56445.7
2.44 4.10E-04
-1.91
(20,30)
0.7
0
58307.7
0 0.00E+00
-0.70
(30.40)
0.7
2
78685.8
2.71 6.40E-04
-2.04
(40.50)
0.5
0
88541.2
0 0.00E+00
-0.49
50+
0.1
1
59590.4
7.03 1.30E-03
-2.09
30+
1.4
3
2 189781
Totals
5.4
5
0.919118
-9.78
Location 3&4
(0.10)
12.7
13
194913
1.02 1.40E-04
-2.21
2.18E-10
13.34 0.02
(10.20)
4.3
5
262898
1.16 2.10E-04
-1.79
(20,30)
2.3
0
179052
0 0.00E+00
-2.30
(30.40)
2.5
0
251766
0 O.OOE+OO
-2.47
(40,50)
2.1
1
348264
0.48 1.10E-04
-1.38
50+
0.9
3
298743
3.29 6.70E-04
-2.98
30+
5.5
4
0.729927
Totals
24.8
22
0.886739
-13.12
-------
Table 6-24:
Fit of EPA Mesothelioma Model to Observed Mortality Among Wittenoom
Miners Categorized by Time Since Last Exposure
Time Since Last
Average
Exposure
Predicted
Cases
Value
Relative
Log
Range
Mean
of Integral
Risk
Likelihood
lYrsi
lYisl
(0.1)
03
0.0
0
35
0
-0.03
11.5)
3.0
0.1
0
72
0
-0.12
15. 10)
7.5
0.6
0
327
0
-0.64
110. 20)
14.8
14.5
10
4025
0.7
¦2.85
120, 30 )
24.7
52.0
58
18058
1.1
-3.19
130.40)
34.0
81.3
70
39802
1.3
-5.44
40+
43.6
15.5
17
57952
1.1
-2.41
30+
76.6
96.0
1.3
Totals
144.9
164
-14.68
Goodness of Fit
ChiSq
8.41
P value 0.21
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 6-26 similarly shows the fit of the EPA model to the South Carolina mesothelioma
data (including all four suspected mesotheliomas as real). As with the Quebec cohort,
there is little evidence from this cohort that the model under or over-predicts the number
of mesotheliomas at long times following the end of exposure. Particularly with this
cohort however, conclusions must be drawn with caution due to the very small number
of mesotheliomas observed.
Based on the analysis performed, we see marginal evidence that the current EPA
model for mesothelioma may not provide an adequate description of the long term risk
of mesothelioma in cohorts exposed to asbestos (even for extended periods of time
following cessation of exposure); the fit of the model to the Wittenoom cohort is
adequate, despite the small apparent discrepancy at long times. There also appears to
be little evidence that there is a difference in the time development of mesothelioma
between the Wittenoom and Quebec cohorts, although it is acknowledged that the data
from the Quebec cohorts were divided into sets containing only small numbers of
mesotheliomas. This latter observation contrasts with what was observed for the Jung
cancer model, where there is evidence that the effect that chrysotile has on relative risk
appears to decrease with time following cessation of exposure while the effects of
crocidolite appear to remain constant (Section 6.2.1.1).
If the observed difference in the time dependence for lung cancer risk observed for
chrysotile (among South Carolina textile workers and Quebec miners) and crocidolite
(among Wittenoom miners) is indeed due to fiber type, that a similar effect is not
observed for mesothelioma suggests, among other possibilities, that similar fiber types
may be primarily responsible for driving mesothelioma risk. This provides further, albeit
circumstantial, support for the "amphibole hypothesis" (that amphiboles are primarily
responsible for the induction of mesothelioma, even when present as minor
contaminants in chrysotile).
Given the importance of these two issues: (1) the relative potencies of chrysotile and
the amphiboles and (2) the adequacy of EPA models for predicting the time
dependence of disease, limited analysis of raw data from a small number of additional
cohorts is warranted. Such recommendations are discussed further in the conclusions
to this Chapter (Section 6.4).
6.3,1.2 Comparison with a biologically based model
Definition of the two-stage model. The two-stage model of cancer implemented in
this report (Moolgavkar and Luebeck, 1990) was previously described (Section 6.2.1.2).
6.81
Revision 1 • 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 6-26
6.82
Revision 1 - 9/2/01
-------
Table 6-28:
Fit of EPA Mesothelioma Model to Observed Mortality Among South Carolina Textile Workers
Time Since
Last Exposure
Predicted
Cases
Averaged
Value of
Relative
Rate
Log
Range (Yrs)
Mean (Yrs)
the Integral
Risk
Likelihood
[0.1)
0 06
11.9
1
118165
0.08
0.00004
-9.40
(1.5)
3.00
2.1
2
47292
0.97
0.0002
-1.31
(5, 10)
7.50
2.4
0
46052
0
0
-242
[10.20)
14.84
3.7
2
40029
0.54
0.00009
-1.78
I 20. 30)
24.80
2.7
1
35825
0.37
0.00006
-1.70
[ 30.40)
34.65
1.9
0
35224
0
0
-1.94
40+
43.49
0.8
0
33443.5
0
0
-0.79
30+
2.7
0
0
0
Totals
25.5
6
-19.33
Goodness of Fit ChlSq 29.44
P value 5.0E-05
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Applying the two-stage model to mesothelioma. The 2-stage model was fit to the
mesothelioma data using the same general approach applied to the lung cancer data
and described earlier (Section 6.2.1.2). Since the background risk of mesothelioma is
very small, the first stage mutation rate, v0, was fixed at zero in these analyses. Also,
the death rate for intermediate cells, (S0, was likewise fixed at zero. (This is not
expected to affect the range of hazards predicted by the model, since so long as the
hazard is small, it is approximately a function of the difference a - p.). Four situations
were modeled:
(1) assuming only the mutation rate of normal cells into intermediate cells is
dose-related [v = v,d(t), a = a0, |J = pj;
(2) assuming both the mutation rate of normal cells and the proliferation rate
of intermediate cells are dose-related [v = v^t), Ot = cc0 * oc^t), p = pj;
(3) assuming both mutation rates are dose-related [V = v,d(t), a = a0, p s p„
+ Pid(t)]; and
(4) assuming that both mutation rates are dose-related and equal on a per-
cell basis {v = pX = v,d(t), a = a0. where X = 107 is the assumed number
of normal cells].
The two-stage model was fit to the mesothelioma data from both Wittenoom and
Quebec in a variety of situations. Table 6-27 shows results of fitting the two-stage
model to the Wittenoom mesothelioma data assuming either an infinite half-life of
internally deposited fibers (r = 0) or a ten-year half-life (r = 0.069). This Table shows
the results of fitting allowing each one of the four parameters a, J3, v, and p individually
to be related to internal exposure (Model 1-4) and allowing all four simultaneously to be
exposure-related (Models 5,6). The best fit (largest likelihood) using a single exposure-
related parameter came from allowing v to depend upon exposure. This was true when
either an infinite half-life or a ten-year half-life was assumed. The likelihoods from the
two assumed half-lives were similar. The estimated parameters were also similar,
except for v.,, which was larger using to ten-year half life in order to compensate for the
smaller estimated exposures in this case.
The fit with an infinite half-life allowing all four parameters to be exposure-related
(Model 5) did not improve the fit, even though parameter estimates differed
considerably. This suggests that the likelihood is quite flat and diverse parameter
estimates can provide comparable fits. However, the differences a0 - P0 are nearly
constant, which is consistent with the observation by Moolgavkar et al. (198B) that in
many cases the model depends upon these parameters mainly through their difference.
6.63
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 6-27
6.84
Revision 1 - 9/2/01
-------
Table 6-27:
Reaulta of Several Application* of the Two-Slag* Modal to the Mesothelioma Data from Wlttanoom
MODEL 1
MODEL 2
MOOEL 3
MODEL 4
MODEL 5
MODEL 6
Parameter
Fitted Valua
Standard
Flttad Valua
Standard
Flttad Valua
Standard
Flttad Valua
Standard
Flttad Valua
Standard
Flttad Value
Standard
Error
Error
Error
Error
Error
Error
aO
1.42E+01
2.78E-01
1.42E+01
4.57E-02
1.42E+01
3.00E-02
2.26E-01
2.51 E-02
4.98E-01
1.18E-01
5.50E-01
1.01E-01
al
1.31E-03
3 47E-04
0
fixed
0
fixed
0
fixed
1.49E-01
8.19E-02
1.28E-01
2.45E-03
III
P0
1.40E+01
2.79E-01
1.39E+01
4.56E-02
1.41E+01
2.99E-02
1.44E-09
9.13E-01
1.13E-02
1.81E-01
5.94E-02
9 92E-02
IL
J
PI
0
fixed
-1 27E43
3.29E-04
0
fixed
0
fixed
1.50E-G1
6.19E-02
1.27E-01
2.44E-03
S
vO
2 66E-01
8.12E-02
2.88E-01
5.88E-02
1.05E+00
2.83E-01
5.05E-03
7.52E-04
7.30E-O5
7.24E-05
0
fixed
<
x
vl
0
fixed
0
fixed
2.95E+00
8 33E-01
0
fixed
2.78E-03
6.08E-04
2.24E-03
4.23E-04
S!
p0
206E-O8
6.12E-09
2 66E-0B
588E-09
1.05E-07
2 63E-08
5.05E-10
7 52E-11
7.38E-12
724E-12
0
fixed
E
- X
m
Loq
0
fixed
0
fixed
0
fixed
1.29E-07
1.22E-07
1.G8E-07
9.76E-08
2.86E-07
1.80E-07
\L
z
Likelihood
-11496
-1149.9
-1131.7
-1144.8
-1098.4
-1101.8
q0
147E+01
4.16E-02
1.47E+01
1.39E+00
1.47E+01
297E-02
1.48E+01
4.36E-02
5.74E-01
1.49E-01
5.97E-01
1.48E-01
al
1 79E-03
4.39E-04
0
fixed
0
fixed
0
fixed
3.73E-01
1.13E-01
2.75E-01
1.01E-01
ui
fio
1.45E+01
4.10E-O2
1.45E+01
1.39E+00
1 46E+01
2.88E-02
1.45E+01
4.37E-02
7.08E-O2
147E-01
9 98E-02
1.48E-01
5
Pi
0
fixed
3.44E-13
2 90E-05
0
fixed
0
fixed
3.76E-01
1.13E-01
2 78E-01
1.01E-01
!i
vO
344E-01
7.72E-02
2.77E-01
8.30E-02
9.88E-01
242E-01
2.09E-O1
8.52E-02
4.58E-05
4 50E-05
0
fixed
<
X
vl
0
fixed
0
fixed
460E+00
1.01 E+00
0
fixed
7.24E-03
1 85E-03
5.13E 03
1.50E-03
a
MO
3.44E-08
7.72E-09
2.77E-08
6 30E-09
9.68E-08
2.42E-08
2.89E-08
6 52E-09
4.58E-12
4.50E-12
0
fixed
ui
*¦
Ml
Loa
0
fixed
0
fixed
0
fixed
291E-10
808E-10
3.13E-07
2.41 E-07
4.84E-07
3.09E-07
° Likelihood
-1157.3
-1171.7
-1132.1
-1171.8
-1091.7
-irv" »
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Allowing each of the parameters to be exposure-related did significantly improve the fit
of the model assuming a ten-year half life over that obtained by assuming any one of
the parameters to be exposure-related, based on the improvement in the log-likelihood.
The model assuming an infinite half-life and all four two-stage parameters dose-related
predicts that the background hazard increases sharply between age 50 and 60 and
reaches a plateau of about 7x10"5 per year from age 60 onward. This seems
somewhat implausible given the low mortality rate for mesothelioma in persons without
asbestos exposure. Consequently, this model was refit with the background
mesothelioma risk of zero, by assuming both v0 and jj0 were fixed at zero (Model 6).
This model, although being perhaps more realistic, gave a poorer fit (p - 0.03 for infinite
half-life and p=0.08 for ten-year half-life, based on likelihood ratio test) than the model
with the high background.
That the highest likelihoods for the Wittenoom data arise when the mutation rate from
normal to intermediate cells (v) is dose related, suggests that asbestos is acting as an
initiator for mesothelioma. That the fit improves when either the proliferation rate or the
mutation rate of intermediate cells is also allowed to vary with dose suggests that
asbestos may also be serving as either a promoter or a late-stage actor for
mesothelioma. This contrasts with the two-stage model results for lung cancer, in
which having v dose dependent did not improve the fits. Given the apparent limitations
of the two-stage model, the interpretation of such results needs to be taken with great
caution. However, the implication that asbestos serve primarily as a promoter (or late
stage actor) toward lung cancer but also acts as an initiator toward mesothelioma is
consistent with inferences from the broader literature (see Section 7.3).
Table 6-28 shows the fit of the two-stage model with an an infinite half-life and without
background (v0 = Mo = 0) to the Wittnoom data categorized as to time since end of
exposure. The fit of these data to the two-stage model is poor (P = 0.03). For
comparison purposes the fit of the EPA mesothelioma model to the same data is also
provided. The EPA model provides an adequate fit (p = 0.21), which is the same as
previously indicated (Table 6-24).
Based on this limited exploration of the application of the two-stage model to the
Wittenoom data, as well as the Quebec data (not shown), the decision has been
reached that this model is probably not suitable for routine use in asbestos risk
assessment. There are several reasons for this conclusion. The two-stage model was
originally considered because there was some evidence from the literature that the
mesothelioma risk did not continue to rise after the end of exposure, as predicted by the
EPA model. The hazard function from the two-stage model predicts that risk will
eventually fall back to baseline after exposure ceases, and consequently can predict
6.85
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 6-28
6.86
Revision 1 • 8/2/01
-------
Table 6-28:
Fit of the Best-fitting Two-Stage Model to the Wittenoom Mesothelioma Data
Time Since
EPA
Log
EPA Log
Last Exposure
Cases
Predicted
Prediction
Likelihood
Likelihood
Range
Mean
(Yrs)
(Yrs)
10. 1)
0.27
0
0.1
0.0
-0.05
-0.03
M.5)
3.01
0
02
0.1
-0.15
-0.12
[5,10)
7.49
0
0.7
0.0
-0.65
-0.64
110,20)
14.89
10
10.9
14.5
-2.12
-2.05
120, 30)
24.74
58
68.2
52.9
-3.76
-3.19
(30, 40)
33.95
79
74.0
61.3
-3.27
-5.44
>=40
43.55
17
13.1
15.5
-2.06
-2.41
Total
164
167.1
144.9
-12.05
-14.68
Two-Stage
EPA
Goodness of Fit
ChlSq
4.75
8.41
P-value
0.03
0.21
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
this behavior. However, a detailed analysis of the mesothelioma data from both
Wittnoom and Quebec indicates that neither demonstrates such behavior
(Section 6.3.1.1). In fact, if anything, the EPA model may tend to under-predict the risk
at the longest time periods following cessation of exposure (based on the Wittenoom
data), although the results with the other two cohorts may not support this observation
(see Tables 6-24, 6-25, and 6-26).
Although the two-stage model has been shown to be a useful research tool, some of its
features make it less suitable for routine use in risk assessment. The model is complex
and would be more difficult to comprehend by stakeholders in an asbestos risk
assessment. Because of the numerous parameters in the model, numerical
calculations are time-consuming and the results are sometimes problematic. Data are
sometimes compatible with a range of possible outcomes under this model (e.g.,
exposure affecting either the first mutation rate or the cell death rate); currently the
mechanism of mesothelioma causation is not well enough understood to permit choice
among these possibilities with any degree of confidence, although our findings are not
inconsistent with inferences from the literature (see Section 7.3). Moreover, the
absence of a single "potency parameter" in the two-stage model would complicate
simple comparisons of potency across studies or environments. For these reasons, the
decision was reached to not pursue the two-stage model further in this project, but
instead to consider simple modifications to the existing EPA model for cases in which
this model was not adequate (Section 6.2.1.1).
An advantage of the two-stage model is that, since it models risk as a function of
internal dose, it can naturally incorporate effects of fiber clearance.' However, we found
that, due to the complexity of the model, it was neither particularly useful at predicting
fiber half-lives nor distinguishing among assumed half-lives. We also found it possible
to develop simpler adaptations to the EPA model that could reasonably address this
consideration (Section 6.2.1.1) and found this latter approach more fruitful.
6.3.2 Estimating KM's from Published Epidemiology Studies
At the time that the Health Effects Update was published (U.S. EPA 1986), four studies
were found to provide suitable quantitative data for estimating a value for KM and six
additional studies provide corroborative support for the mesothelioma model applied.
Currently, there are 14 published studies with adequate data for deriving an estimate of
Km (including updates to all four of the quantitative studies evaluated in 1986).
The EPA mesothelioma model described at the beginning of Section 6.3 was applied to
each of these data sets to obtain study-specific estimates for the mesothelioma risk
coefficient, Ku. The resulting set of KM values are presented in Table 6-12 (which has
been previously described, Section 6.2.2. In Table 6-12, Columns 9 and 10 list,
6.87
Revision 1 > 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
respectively, the best estimate of each Ku value derived for studies reported in the
original 1986 Health Effects Update and the reference for each respective study.
Columns 11, 12, and 13 present, respectively: the best-estimates for the Kw values
derived for all of the studies currently available (including studies corresponding to
those in the original Health Effects Update); the estimated lower and upper bounds for
each Km value (derived as described in Appendix A); and the reference for the
respective study from which the data were obtained. To assure comparability across
studies, values for all studies (even those that have not been updated since their
inclusion in the 1986 Health Effects Update) were re-derived using the modified
procedures described in Appendix A.
The Kw values derived in this study and the corresponding values derived in the original
1986 Health Effects Update are in close agreement, none vary by more than a factor of
2 and none of the differences appear significant (based on a comparison of their
relative confidence intervals).
Among the Kw values derived in the current study, the lowest and highest of the best-
estimate values differ by a factor of approximately 1400 (excluding the one negative
study) and many of the pair-wise comparisons across this spectrum appear significant
(because their confidence intervals do not overlap). For example, none of the
confidence intervals for the Ku values derived for any of the environments involving
exposure to chrysotile overlap the confidence intervals associated with the KM values
derived for either crocidolite mining or asbestos-cement manufacture using mxed fibers
at the Ontario plant (Finkelstein 1984). Furthermore, neither of the confidence intervals
for the Km values derived for each of the two sets of chrysotile mines in Quebec
(Asbestos and Thetford) overlap the intervals around Kw values for any of the
amphibole environments or any of the mixed environments, except the Quebec factory
that is associated with the Asbestos mine (Liddell et al. 1997).
The Km values and the associated confidence bounds derived in the current study are
plotted in Figure 6-5. Each exposure environment is plotted along the X-axis of the
figure and is labeled with a four-digit code that indicates fiber type (chrysotile, mixed,
crocidolite, or tremolite), industry (mining, friction products, asbestos-cement pipe,
textiles, insulation manufacturing, or insulation application); and a two-digit numerical
value indicating the study from which the data were derived. The same key provided
for Figure 6-3 also applies to this figure. In the Figure, the chrysotile studies are
grouped on the left, amphibole studies are grouped on the right, and mixed studies are
in the middle. Note also that studies conducted in the same facility (generally among
highly overlapping cohorts), such as the Dement et al. (1994) and the McDonald et al.
(1983a) studies of the same South Carolina textile facility, are combined as a single
entry in the Figure. The cohorts from Asbestos Quebec and Thetford are also
combined in the Figure.
6.88
Revision 1 - 9/2/01
-------
Figure 6-5:
Plot of Estimated KM Values and Associated Confidence Intervals by
Study Environment
1000
100 -
10 -
0.1 -
0.01 -
0.001
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Figure 6-5
6.69
Revision 1 • 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
As with the KL values, comparison of Ku values across the available studies are
instructive. Within chrysotile studies alone, lowest and highest Ku values vary by
approximately a factor of 15 (again, excluding the negative friction products study),
which is minimal variation relative to the spread across the values from all of the studies
in the table. Based on the overlap of their confidence intervals, none of these
differences appear to be significant.
The Kw values for the two "pure" amphibole studies (crocidolite mining and amosite
manufacturing) agree to within a factor of 2 and, with the exception of the value from
the Ontario asbestos-cement plant (Study No. 8 in the Figure) and the factory plant
associated with the Asbestos, Quebec mines (Study No. 14), the mixed exposures fall
in between the high values for the "pure" amphibole exposures and the lower values
reported for all of the chrysotile sites. Although the Ku value for the asbestos-cement
plant (Study No. 8) appears high, it does not appear to be significantly different than the
values from either of the "pure" amphibole studies or values for many of the other mixed
exposures (based on the degree of overlap of their respective confidence intervals).
The Kw value reported for the factory associated with the Asbestos, Quebec mine
(Study No. 14) is the lowest of those reported for mixed exposures, but (again based on
the overlap of confidence intervals) it does not appear to differ significantly from values
for several of the other mixed exposures nor the highest values among the chrysotile
studies.
As with Kl values, the asbestos-cement pipe industry shows the greatest variation in KM
values across all asbestos types, with a range that varies over a factor of 90. Moreover,
the difference between the highest and lowest values in this industry may be significant
(due to the lack of overlap of their intervals). Within the textile industry, the Kw value for
the South Carolina chrysotile plant is only a tenth of the values reported for the two
textile plants using mixed fibers, but these differences do not appear to be pairwise
significant. The potential sources of variation across all of the KM values are likely
attributable to the same sources of variation identified for the KL values and previously
described in detail (Section 6.22).
6.3.3 Evaluating Effects Associated with the Character of Exposure
To evaluate the effects of fiber size on the KM values listed in Table 6-12 (in parallel
with the manner in which the KL values were evaluated) KM values were also,adjusted to
match them to the improved exposure index defined in Equation 6.7 and the
implications of these adjustments were evaluated (in a manner entirely analogous to
that described in Section 6.2.4.2). The justification for the development of the improved
exposure index is provided in Appendix B.
6.90
Revision 1 • 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
6.3.3.1 Adjusting K.,'s for size
The Km values were adjusted for size in the same manner described in Section 6.2.4.1
for the Kt values. To accomplish this, the same, published fiber size distributions were
matched to the same epidemiology studies (Table 6-14) and the conversions performed
as described by Equations 6.4 to 6.6 using the data presented in Table 6-15. The
resulting adjusted values, the KM.'s are listed in Table 6-16.
fi.3,3.2 Evaluating the implications of adiustino K.,'s for size
To compare the effects of adjusting values for fiber size, the adjusted values (the
Kw.'s), are plotted (along with their associated confidence intervals) in Figure 6-6, which
exhibits a format identical to that of Figure 6-5. In fact, the formats of Figures 6-3
(unadjusted K/s), 6-4 (the adjusted, Kr's), 6-5 (unadjusted KM's), and Figure 6-6
(adjusted, KM.'s) are also designed for direct comparison. The "key" provided for
Figure 6-3 is also directly transferable to all of the figures, including Figure 6-6.
Although Figures 6-5 and 6-6 may look similar at first glance, there are several
differences. First, the value for study MF14 had to be omitted from Figure 6-6 because
no suitable size distribution is available to allow the potency factor from this study to be
adjusted. The confidence intervals presented in Figure 6-6 are also larger than those
depicted in Figure 6-5. This is because the one's depicted in Figure 6-6 incorporate an
additional factor to account for the uncertainty of adjusting for fiber size using paired
data from published size distributions. Table 6-17 lists the factors by which the upper
bounds are multiplied and lower bounds are divided to derive the confidence intervals
depicted in Figure 6-6.
The most important differences between Figures 6-5 and 6-6 are the changes in the
relative positions of the central value estimates for the potency factors depicted.
Therefore, numerical representations of these changes are provided in Table 6-18.
Table 6-18 presents the magnitude of the spread in the range of original and adjusted
Kw values (original and adjusted values are also presented, see Section 6.2.4.2).
Such spreads are estimated as the quotient of the maximum and minimum values of
each range. As previously indicated, this type of analysis (of necessity) requires that
the zero value obtained for the Connecticut friction products plant study (CF4) be
omitted. The one study for which no suitable conversion factor could be found for the
corresponding KM, (Study No. MF14) was also omitted.
6.91
Revision 1 - 8/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Figure 6-6
6.92
Revision 1 • 9/2/01
-------
00
O
1000
100 -
10 -
1 -
Table 6-6:
Plot of Adjiusted KM Values (KM*fs) and Associated Confidence
Intervals by Study Environment
T3
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
The trends apparent in Table 6-18 that were described for a comparison between KL
and Kr values are similar but exaggerated between the Kw and Kw. values. Briefly, the
spread in values among "pure'fiber types (i.e. chrysotile or amphiboles only) are much
smaller than the spread observed over all values (or over data sets that include mixed
exposures) and this is true for both the adjusted and unadjusted KM values. This
suggests that chrysotile and the amphiboles exhibit different potencies toward the
induction of mesothelioma. Also, the spreads for all of the data sets presented in
Table 6-18 (except the one for pure amphiboles, which is already minuscule) are
reduced when values are adjusted for fiber size. However, even for adjusted values,
the spreads in data sets including mixed exposures are still rather large. As for lung
cancer, this implies that the effects of fiber type and fiber size are also confounded for
mesothelioma. Therefore, an evaluation similar to that conducted for lung cancer
potency factors was also conducted for mesothelioma potency factors to simultaneously
address the effects of fiber size and fiber type. The manner in which the evaluation
was conducted is described in Section 6.2.4.2.
Table 6-19 presents the results of the analysis in which the Kv estimates are adjusted
both for fiber size and for the fraction of the exposures in each environment contributed
by amphiboles. As previously described, values are the estimated potencies of
pure amphiboles, K^.'s are the estimated potency of pure amphiboles adjusted for fiber
size, and RPC is the ratio of the potency of pure chrysotile to pure amphibole toward
the induction of mesothelioma, in this case (see Section 6.2.4.2). In the analysis, the
spread of the range in values for each data set was minimized by optimizing the value
for RPC.
That the optimum values for RPC (presented in Column 3 of Table 6-19) are
substantially smaller than one for both the unadjusted KW4 values and the values
(which are adjusted for fiber size) suggests that chrysotile is substantially less potent
toward the induction of mesothelioma than amphiboles and may be non-potent.
However, because Improvement (reduction) in the spread in values for these data sets
becomes marginal for values of RPC less than about 1/1000, a conservative estimate
suggests that, at most, chrysotile is no more than 1/1000 as potent as amphiboles
toward the induction of mesothelioma.
Based on a comparison of the magnitude of the spreads in range for each data set
(listed in Column 4), adjusting mesothelioma potency factors for fiber size results in a
30% improvement in agreement across estimated values. Moreover, simultaneously
accounting for the effects of fiber type and fiber size results in very effective
reconciliation of potency estimates for mesothelioma. Once adjusted in ths manner, the
spread in potency estimates across all 10 of the studies available for this analysis is
only a factor of 26 (only a little larger than an order of magnitude). This should provide
6.93
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
reasonable confidence that a representative value for amphibole potency that might be
selected from this range can be applied to new environments to estimate risk.
As with lung cancer, a more sophisticated analysis of mesothelioma potency values
was also completed, which incorporates formal consideration of uncertainty and (due to
the manner in which the analysis is performed) allows us to include the nominally zero
value from the Connecticut friction products plant (Study No. CF4). The details of the
analysis are provided in Section 6.2.4.2. Results are provided in Table 6-21.
As with lung cancer, the results of this more formal analysis of mesothelioma potency
factors closely parallels the results described in the above analysis. The best estimated
values of the relative potency of chrysotile toward mesothelioma lies between 1/250th
(0.004) and 1/570th (0.002) of that for the amphiboles and these values are not
statistically different from an estimate of zero potency for chrysotile (P = 0.54 for the
unadjusted data set and P = 0.72 for the size adjusted data set).
Adjusting the mesothelioma potency factors for fiber size results in an about a 23%
improvement in the overall agreement among values (determined by comparing the
relative magnitudes of the optimized "a" estimated for the size adjusted data set
(0.6561) and the unadjusted data sets (0.8123), respectively. Thus, even using the
relatively crude procedure of adjusting KM values based on published (rather than
study-specific) size distributions, results in moderate improvement in the agreement in
values across studies.
6.4 GENERAL CONCLUSIONS FROM HUMAN EPIDEMIOLOGY STUDIES
Results from our evaluation of the available epidemiology studies indicate that:
(1) the results of individual epidemiology studies are uncertain, especially when one
considers all of the major potential sources of uncertainty (rather than simply
considering the statistical uncertainty associated with the number of deaths, as is
traditionally done). However, more robust conclusions can be drawn from an
analysis of the set of epidemiology studies taken as a whole than results derived
from individual studies;
(2) by adjusting for fiber size and fiber type, the existing database of studies can be
reconciled adequately to reasonably support risk assessment;
(3) the procedures recommended as a result of our analysis offer substantial
improvement over EPA's current approach to evaluating asbestos-related risks;
6.94
Revision 1 • 8/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
(4) while there is some indication that the existing EPA models for lung cancer and,
potentially, mesothelioma may not entirely reflect the time-dependence of
disease at long times following cessation of exposure, such effects appear to be
modest so that they are unlikely to adversely affect the proposed approach.
Prudence dictates, however, that limited additional study may be warranted to
adequately dismiss related concerns; and
(5) results from our review of the supplemental literature provide additional support
for the approach recommended in this chapter and indicate additional, specific
modifications that (if supported by limited additional study) could result in
substantial improvement even over the approach currently recommended, which
(in turn) provides substantial improvement over the current approach.
More detailed conclusions and recommendations are described below.
Conclusions and Recommendations For Lung Cancer
Without adjustments, the KL values estimated from the 18 existing studies vary by a
factor of 90 and pairwise differences between several individual studies are likely
significant (based on a comparison of their associated confidence intervals). By
simultaneously adjusting these values for fiber size and type, variation across the 16
studies (for which data are available to make the required adjustments) is reduced to a
factor of about 60.
Especially given the crude manner available for performing size adjustments coupled
with the very consistent picture that such adjustments improve matters and that there is
a strong theoretical basis for expecting that such adjustments should improve cross-
study comparison or extrapolation, we believe that such adjustments increase overall
confidence in cross-study comparison/extrapolation. We therefore recommend their
use.
The maximum and minimum values driving the factor of 60 come from cohorts of South
Carolina textile workers and Quebec miners, respectively, and the difference in the lung
cancer potency estimated between these studies has long been the subject of much
attention. A detailed evaluation of the studies addressing this difference, the results of
our analysis of the overall epidemiology literature, and implications from the broader
literature, indicate that the most likely cause of the difference between these studies is
the relative distribution of fiber sizes in the two study environments. It is therefore likely
that the variation between these studies can be further reduced by implementing one or
a combination of the following:
6.95
Revision 1 - 0/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
• deriving improved characterizations of the exposures experienced by cohort
members in each, respective environment based on study-specific size
distributions (instead of relying on the cruder use of published distributions, as
currently required);
• deriving improved characterization of the exposures experienced by cohort
members in each, respective environment based on improved analysis of
relevant material (instead of relying on the existing, published distributions, which
are insufficiently detailed to capture all of the biologically-relevant aspects of size
distributions);
• from a selected subset of key studies, developing the improved characterization
data required to allow the existing lung cancer potency factors to be adjusted to
an exposure index that is even closer to an optimized index than the one
recommended in this document;
• procedures for accomplishing some or all of the these objectives are described
below.
Importantly, the exposure index recommended in this document (and used throughout
this analysis), while not perfect, when coupled with the recommended adjustments for
fiber type, already provides substantial improvement over EPA's current approach to
evaluating asbestos-related risks. This is evident, for example, if one compares certain
values of the variance (o) provided in Table 6-21. Because, under the current
approach, chrysotile and amphiboles are considered equipotent and lung cancer and
mesothelioma risks are combined, the overall variance estimated for the unadjusted Ku
values with RPC =1 (1.817) should be compared with the variance estimated for
adjusted Kw. values with RPC optimized (0.6561), which represents the recommended
approach. Thus, the new approach reduces variance by approximately a factor of three
over the current approach. Actually, the improvement is even larger because the KM
values employed in our analysis include a greater number of studies than available to
the U.S.' EPA in 1986. Therefore, despite our recommendations below to conduct
limited, additional study that may lead to further improvements, there is no reason not
to implement the recommendations from our current analysis now.
Although the EPA lung cancer model appears to adequately describe the time
dependence of disease for amphiboles, it may not adequately describe risks for
chrysotile at long times following cessation of exposure. In contrast to predictions from
the model (potentially due to the observed, lower biodurability of chrysotile), lung cancer
risks from chrysotile decrease with time following cessation of exposure. Thus, if the
model is applied to any new site to predict risks, if anything, risk estimates will be
conservative. However, particularly for studies with long followup times after cessation
6.96
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
of exposure, the K^'s estimated from existing studies (using the EPA model) may be
underestimated. Importantly, these conclusions are based on raw observations from
only one chrysotile and one amphibole study. Thus, it may be prudent to evaluate raw
data from a minimum of one additional chrysotile study and one additional amphibole
study to confirm that such observations are indeed general. However, it is likely that
these time-dependent effects are relatively small.
Conclusions and Recommendations For Mesothelioma
Without adjustments, the Kw values estimated from the 12 existing studies vary by more
than a factor of 1000 and pairwise differences between several individual studies are
likely significant (based on a comparison of their associated confidence intervals). By
simultaneously adjusting these values for fiber size and type, variation across the 11
studies (for which data are available to make the required adjustments) is reduced to a
factor of about 26. Given that this is little larger than an order of magnitude, it appears
that we have essentially reconciled the mesothelioma studies.
Although the EPA mesothelioma model appears to adequately describe the time
dependence of disease in general, there was some indication that it may under
estimate risks at long times following cessation of exposure for amphiboles. Therefore,
as part of the analysis recommended for further testing the time dependence of the lung
cancer model, it may also be prudent to further evaluate the behavior of the
mesothelioma model.
Conclusions and Recommendations for Risk Assessment
Based on our analysis, we recommend the following.
(1) Asbestos concentrations should be analyzed for structures that correspond to
the exposure index defined in Equation 6.7. Thus, count only structures longer
than 5 pm and thinner than 0.5 pm with structures longer than 10 pm weighted
as indicated in Equation 6.7. Further, count both parent fibers and bundles and
fibers and bundles that are components of more complex structures, but that
individually satisfy the above-defined dimensional criteria. Importantly, analysis
needs to be performed by TEM following preparation by a direct-transfer
procedure.
(2) When mixed, the individual contributions from chrysotile and the combined
amphiboles to any particular exposure need to be separately delineated and,
because amphibole exposures are the primary drivers, the analytical sensitivity
required for each particular study should be set based on amphiboles.
6.97
Revision 1 - 8/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
(3) Combine such counts with the matched potency factors for lung cancer and
mesothelioma described in Tables 6-29 or 6-30 below using the appropriate EPA
model for each disease.
Based on our analysis, the optimized potency coefficients that are adjusted for size (so
that they are appropriately matched to the exposure index defined above) are obtained
from Table 6-21 and reproduced here:
Table 6-29:
Optimized Risk Coefficients
for Pure Fiber Types
Fiber Type KL.x100 K^xlO®
Chrysotile 0.85 0.05
Amphiboles 4.5 30
In keeping with the tradition of incorporating conservative estimates to minimize the
chance of under-estimating risks, conservative estimates for the potency factors listed
in Table 6-29 are derived by up-adjusting the amphibole potency values for both lung
cancer and mesothelioma so that they match the largest value from the set of available
studies and adjusting the chrysotile values proportionally. Results from this calculation
are presented in Table 6-30.
Table 6-30:
Conservative Risk Coefficients
for Pure Fiber Types
Fiber Type Kt. x 100 K*. x 10s
Chrysotile 3.0 0.1
Amphiboles 15 50
We leave it as an option for deciding which values to employ. Interestingly, however,
due to the great degree of reconciliation among the disparate values for the potency
estimates already achieved in this document, the ratio of the conservative and optimum
K^. estimates is only a little larger than a factor of 3 while the conservative and optimum
Km. values vary by less than a factor of 2.
6.98
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Recommendations for Limited, Further Study
The two major objectives identified above for further study are:
(1) to expand the test of the ability of the current EPA models to adequately track
the time-dependence of disease; and
(2) to develop the supporting data_needed to defiQ& adjustments for potency factors
that will allow them to be used with an exposure index that even more closely
captures the criteria that determine biological activity (see Section 7.5). Among
other things, this may ultimately provide the data allowing complete reconciliation
of the potency factors derived for Quebec miners and South Carolina textile
workers, one of the major outstanding issues identified among asbestos
researchers.
The first of the above objectives requires that we gain access to raw data from a small
number of selected, additional epidemiology studies. The best candidate studies
include: (for chrysotile) the lung cancer data from Quebec (best) or, potentially, from the
New Orleans asbestos-cement pipe plant studied by Hughes et al. For amphiboles,
the best candidate studies include: the lung cancer and newest mesothelioma data
from Libby (best), or, potentially the lung cancer and mesothelioma data from the
Paterson, New Jersey insulation manufacturing plant studied by Seidman. The
possibility of obtaining some or all of these data sets needs to be further explored.
The second of the above objectives requires more detailed size characterization data
for the environments of interest. Although archived air samples do not appear to be
available from any of the study locations of interest, we believe that suitable data can
be developed from appropriate bulk samples. Thus, for example, it would be useful to
obtain samples of the raw ore from Libby, Wittenoom, and Quebec and the textile,
asbestos-cement pipe, and friction-product grade products from Quebec. We have
already approached individuals who can provide access to these materials and they
have all expressed interest in collaborating.
Results from our review of the supporting literature suggest that the optimum cutoff for
increased potency occurs at a length that is closer to 20 pm thanIO pm, (the latter of
which is the cutoff in the exposure index provided in this study). Data do not currently
exist to improve on this latter cutoff. However, provided that study-specific size
distribution data could be obtained as indicated above, with the appropriate analyses, it
will be possible to develop the size distributions necessary to evaluate a range of
considerations including:
€.99
Revision 1 - 9/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
delineation of size fractions among individual length categories out to
lengths as long as 30 or 40 pm;
determination of the relative presence and importance of cleavage
fragments (of non-biologically relevant sizes) in mine ores vs. finished
fibers; and
the relative fraction of fibrous material vs. non-fibrous particles in the
various exposure dusts of interest.
6.100
Revision 1 • 8/2/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
7.0 SUPPORTING EXPERIMENTAL STUDIES
To evaluate the plausibility of cancer risk models for asbestos it is useful to examine
what is currently known (and what is not known) about (1) the mechanisms that
facilitate the transport of asbestos to the various target tissues of interest (in the lungs
and mesothelium) and (2) the mechanisms that contribute to the development of cancer
in these target tissues. Accordingly, a detailed review of the relevant literature is
provided in this Chapter.
Although much progress has been made over the last decade toward elucidating the
fiber/particle mechanisms that contribute to transport and subsequent cancer induction,
at least two critical data gaps remain:
• no one has yet been able to track a specific lesion induced by asbestos in a
specific cell through to the development of a specific tumor. There have been
experiments that show altered DNA and other types of cellular and tissue
damage that are produced in association with exposure to asbestos and other
studies demonstrating that various tumors of the kinds that result from asbestos
exposure exhibit specific patterns of DNA alteration or other damage. There are
also studies that show that exposure to asbestos can lead ultimately to
development of tumors. However, these types of studies have yet to be linked;
and
• although the target tissues in which asbestos-induced cancers develop have
generally been identified, the specific target cells that serve as precursors to
these tumors are not known with certainty.
Because of the first of the above limitations, researchers have tended to report on a
broad range of tissue and cellular effects induced by asbestos that may lead generally
to various kinds of cellular damage or injury. Direct cytotoxicity, for example, is one of
the end points typically tracked as a marker for asbestos-induced injury. However, not
all of these effects necessarily contribute (either directly or indirectly) to the
development of cancer. Therefore, one of the goals of the following discussion is to
distinguish (within the limitations of the current state of knowledge) among effects that
likely contribute to the development of cancer from those that are less likely or unlikely
to contribute.
In addition, because the relative effects of fiber size, shape, and mineralogy need to be
elucidated to better indicate how asbestos concentrations should be characterized to
support risk assessment, studies that address these topics are highlighted. Of
7.1
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
particular interest are studies that contrast the effects of different sized fibers or the
effects of fibers from the effects of non-fibrous particles of similar mineralogy and
studies that contrast the effects of fibers of comparable morphology (size and shape)
but differing mineralogy.
The types of studies that have contributed to the state of knowledge of the effects of
asbestos (in addition to the human epidemiology studies previously reviewed, see
Chapter 6) include:
• whole animal inhalation studies;
• whole animal instillation studies;
• whole animal injection, implantation studies;
• human pathological studies;
• in-vitro studies in cell cultures; and
• in-vitro studies in ceh-free systems.
Note, depending on the outcome(s) monitored, the animal studies may alternately be
categorized as retention studies, histopathology studies, or dose-response studies.
Each type of study possesses certain advantages and exhibits certain limitations, which
have previously been described (Chapter 5) along with descriptions of the nature of
each of these study types. In addition to the advantages and limitations that are
attributable to the type of study, the quality of the characterization of asbestos (or other
particulate matter) determines the utility of the study for addressing issues associated
with fiber morphology and mineralogy. Unfortunately, for many published studies, both
the characterization of the asbestos (or other particulate matter) and descriptions of the
manner in which such materials were handled are insufficient to establish the detailed
morphology or mineralogy. Thus, such limitations need to be considered when
comparing across study results or evaluating the validity of study conclusions.
As previously indicated (Chapter 3), the primary route of asbestos exposure of concern
for humans is inhalation. Moreover, the biological activity of inhaled asbestos depends
on the following factors:
• the extent that asbestos structures are respirable and the pattern of
deposition of Inhaled structures;
• the extent that deposited structures are subsequently cleared or
degraded;
7.2
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
• the extent that deposited structures are transported or migrate to the
various target tissues; and
• the extent that retained structures induce a biological response in each
target tissue.
7.1 FACTORS AFFECTING RESPIRABILITY AND DEPOSITION
Discounting systemic affects resulting from other forms of exposure, factors affecting
respirability are common to all of the toxic end points associated with asbestos
exposure considered in this study (asbestosis, pulmonary carcinomas, and
mesothelioma). Moreover, respirability is common to the factors affecting the toxicity of
inhaled, insoluble particles in general. To be respirable, an inhaled particle must pass
the blocking hairs and tortuous passageways of the nose and throat and be deposited
in the lungs. Particles deposited in the naso-pharyngeal portion of the respiratory tract
are not considered respirable.
Not all of the inhaled particles that reach the lungs will be deposited. Small particles
may not impact lung surfaces during inhalation and are subsequently exhaled. Once a
particle impacts on a surface, however, it is likely to remain because the surfaces of the
lungs are wetted with a surfactant (Raabe 1984).
Adverse health effects potentially result when particles are deposited in the lungs and
remain in contact with tissue within the lungs for a sufficient period of time to provoke a
biological response. To affect the mesothelium, an offending particle may also need to
migrate or be transported from the lung to this surrounding tissue. However, due to the
proximity of the mesothelium to peripheral portions of the lung parenchyma (which
include locations where particles are typically deposited), it is also possible that
diffusable molecules produced in lung tissue in response to deposited particles can
have an adverse effect on the mesothelium (see, for example, Adamson 1997). Such
effects are considered among the mechanisms of disease induction addressed in the
discussion of biological responses (Section 7.3).
It is long-term retention of particles in the respiratory tract that potentially leads to
adverse health effects. Consequently, the interplay between deposition and removal
(clearance) is an important determinant of biological activity and separating the impact
of these two processes is difficult. The term "retention" is used here to represent the
fraction of particles remaining in the lungs beyond the time frame over which only the
7.3
Revision 1 • 0/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
most rapid removal processes are active (i.e. muco-ciliary clearance). The factors
affecting retention are addressed further in Section 7.2.
Published inhalation studies divide the respiratory tract into three units (see, for
example, Raabe 1984). The naso-pharyngeal portion of the respiratory tract extends
from the nares in the nose through the entrance to the trachea. The tracheo-bronchial
portion of the respiratory tract includes the trachea and all of the branching bronchi
down to the terminal bronchioles. The respiratory bronchioles and the alveoli, which
are collectively referred to as the "deep lung", are the bronchio-alveolar (or pulmonary)
portion of the respiratory tract. For a more detailed description of the features of the
respiratory tract, see Section 4.4.
The dimensional requirements for respirability have been studied and reviewed by
several authors (see, for example, Raabe or 1984 or U.S. EPA 1986). A more recent
review is also presented by Stober et al. (1993). Much of the data reviewed by these
authors is based on earlier studies in which researchers exposed animals or human
volunteers to a series of monodisperse spherical particles (although Stober etal. also
reviews fiber experiments). In this manner, the impact of the diameter of spherical
particles on respirability was elucidated. The respirability of fibrous materials (such as
asbestos) tends to be described in terms closely associated with those employed for
spherical particles, but with adjustments for density and shape, importantly, because
respirability is a mechanical process: the size, shape, and density of a particle (or fiber)
determine its respirability along with the morphometry of the airways through which the
particle passes (Stober et al. 1993). Other than affecting the particle's density or the
distribution of fiber shapes, the chemical composition (mineralogy) of a particle (or fiber)
does not influence respirability.
The respirability of particles and fibers by humans and a variety of other mammals of
experimental interest has also been the subject of increasingly sophisticated modeling
efforts (Stober et al. 1993). The latest refinements of such models predict particle
deposition with a degree of accuracy that is beyond what can be validated with existing,
experimental data. The application of several of these models to asbestos (and other
fibrous materials) are considered throughout this chapter. However, a detailed
overview of the state of the art of such modeling is beyond the scope of this document.
Such an overview is presented by Stober et al. (1993).
7.1.1 Respirability of Spherical Particles
Spherical particles larger than 10 pm in diameter are considered non-respirable
because virtually all particles in this size range are trapped in naso-pharyngeal
7.4
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
passageways and blocked from entering the lungs. As the diameter of the particles fall,
an increasing fraction traverses the nose and throat and may be deposited in the lungs.
About half of particles 5 pm in diameter are blocked before entering the lungs. Virtually
all particles smaller than 1 pm enter the lungs, although other factors determine
whether they are in fact deposited or simply exhaled. Figure 7-1 (Source: Raabe 1984)
is a representation of the relative deposition in the various compartments of the
respiratory tract as a function of particle diameter.
Within the lungs (Figure 7-1), the greatest fraction of respirable particles (over the entire
range of diameters down to less than 0.01 pm) are deposited in the deep lung (the
broncho-alveolar portion of the respiratory tract), primarily at alveolar duct bifurcations
(see, for example, Brody et al. 1981, Davis et al. 1987, Johnson 1987, and Sussman et
al. 1991a). These studies also indicate that biological responses appear to be initiated
where deposition is heaviest. Generally, the fraction of particles deposited in the deep
lung increases regularly with decreasing diameter until a maximum of 60% deposition in
the deep lung is reached at about 0.1 pm diameter.
As indicated in Figure 7-1, a transition occurs at particle diameters between 0.5 and
1 pm. For particles in this range and smaller, deposition in the deep lung competes
primarily with deposition in the tracheo-bronchial tree and with exhalation; smaller
particles have an increasing probability of being exhaled without ever impacting the
surface of an air passageway. For particles larger than this transition range, broncho-
alveolar deposition is limited chiefly by the fraction of particles that are removed from
the air stream prior to reaching the deep lung (either by deposition in the naso-
pharyngeal or the tracheo-bronchial portions of the respiratory tract).
The transition between naso-pharyngeal competition with deep-lung deposition and
competition from other removal processes is important because, during mouth
breathing, a process that bypasses the tortuous pathways of the nose and throat, it has
been observed that larger particles (up to several micrometers in diameter) may be
deposited in the deep lung (Raabe 1984). Studies of the effects of mouth breathing are
also reviewed by Stober et al. (1993). Because most people spend at least small
amounts of time mouth breathing, especially during exertion or during snoring, this
mechanism for allowing larger particles to settle in the deep lung should not be ignored.
A diameter of 0.5 pm also happens to represent the transition between the regime
where inertial flow and the regime where diffusional flow becomes the major factor
controlling deposition in the lungs. Below the 0.5 pm transition, the diffusional diameter
becomes more important in determining deposition than the aerodynamic equivalent
diameter (defined below).
7.5
Revision 1 >8/3/01
-------
PRELIMINARY WORKING DRAFT- DO NOT COPY OR QUOTE
Figure 7-1. OLD Figure 5-1
NEED TO GET PERMISSION TO REPRODUCE FIGURE AND NEED TO DEVELOP
FIGURE.
7.6
Revision 1 • 9/3/01
-------
Figure 7-1:
FRACTIONS OF RESPIRABLE PARTICLES DEPOSITED IN THE VARIOUS
COMPARTMENTS OF THE HUMAN RESPIRATORY TRACT AS A FUNCTION
OF AERODYNAMIC EQUIVALENT DIAMETER"
i i 11111Pi'—i i n i tilf
too
I I .< M Mil
O.OI
i i'i'i iTih" "ill i>ihi
0.1 l.o 10.0
Diameter of Spherical Particle with Density Equal to One gfcm'fum) c
m
Source: Raabe 1984.
b. 3
Assumes a typical tidal value of 1450 cm and a rate ot 15 breaths per minute.
• C
Aerodynamic equivalent diameter.
Confidential:
Need permission to reproduce this figure
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
7.1.2 Respirability of Fibrous Structures
Several authors have investigated the effect of the shape of non-spherical particles
(including fibers) on respirability and deposition (see, for example, Harris and Timbrell
1977, Sussman et al. 1991a and b, Strom and Yu 1994, or Yu et al. 1995a andb). It
has been found that the behavior of non-spherical particles can be related to the
behavior of spherical particles by introducing a concept known as the aerodynamic
equivalent diameter. The aerodynamic equivalent diameter is the diameter of a
hypothetical spherical particle of unit density that would exhibit the same settling
velocities and aerodynamic behavior as the real, non-spherical particle of interest.
Factors that affect the aerodynamic equivalent diameter are density, true diameter, true
length (for elongaled particles such as fibers) and the regularity of the particle shape.
Harris and Timbrell Finriingc
Because fibrous particles tend to align primarily along the axis of travel under the flow
conditions found in the lungs, respirability is predominantly a function of the diameter of
a fiber and the effect of length is secondary (Harris and Timbrell 1977). Fibrous
structures (of unit density) with aspect ratios greater than 3:1 behave like spherical
part'des with diameters up to 3 times larger {the aerodynamic equivalent diameter) and
exhibit only a very weak dependence on length. This is demonstrated in Fioure 7-2
where the true diameter of a fiber is graphed on the top horizontal axis aqainst
spherical (aerodynamic equivalent) diameters on the bottom horizontal axis Fiaure 7-2
is an overlay of Figure 7-1. Note that, to adjust for the density of asbestos, Ihe lue
diameters listed >n the figure have been shifted to the right of where they would appear
rf the relationship was exactly one third of the aerodynamic equivalent diameter.
Two vertical dashed lines in Figure 7-2 represent effective limits to the range of
respirable asbestos. The left line in the figure represents the limiting diameter of the
smallest chrysot.le fibril (about 0.02 pm true diameter) and thus represents a lower limit
to the deposition chart that is of concern when considering asbestos The right vertical
line represents the cutoff where deposition in the deep lung becomes unimportant
to removal of such particles by the nasopharyngeal passageways. This latter cutoff
corresponds to a true fiber diameter of 2.0 Mm, which theoretically represents the upper
limit to the size of asbestos that is respirable. As indicated in the figure, however
deposition in the deep lung drops precipitously for fibers thicker than about 0 7 urn so
acL"h the deep lung06"' °' aSbeE'°S fiberilhicker ,han ¦PP™"™**1 Mm
7.7
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
7.6
Revision 1 - 8/3/01
-------
FtguraT^.
FRACTIONS OF RESPIRABLE PARTICLES DEPOSITED IN THE VARIOUS
COMPARTMENTS OF THE HUMAN RESPIRATORY TRACT AS A FUNCTION
OF THE TRUE DIAMETER OF ASBESTOS FIBERS"**
The Diameter of Asbestos Fiber
Source of original: Raabe 1984. ^
b Assumes a typical tidal value of 1450 cm and a rale of 15 brealhs per minute.
" The relationship between true diameters and aerodynamic equivalent
diameters derived from Harris & Umbrell (1977). Diameters
adjusted for shape and density of asbestos fibers.
4 Aerodynamic equivalent diameter.
Confidential:
Need permission to reproduce this figure
-------
Figure 7-2 (old Figure 5-2)
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
7.9
Revision 1 - 9/3AJ1
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Harris and Timbrell (1977) also evaluated the relationship between the overall shape of
a particle and the extent of deposition. Over the range of diameters that potentially
represent the range of asbestos fibers likely to be encountered, pulmonary deposition
decreases with increasing complexity of shape beyond simple cylinders (such as
clusters and matrices) at the expense of increasing naso-pharyngeal or tracheo-
bronchial deposition. This change also becomes increasingly important as the length of
the structure increases.
For structures less than 25 pm in length, the difference in deposition between simple
fibers and complex clusters or matrices may vary by up to a factor of 2 with the complex
structures being more likely to be removed in the naso-pharyngeal portion of the
respiratory tract and the fibers more likely to be deposited in the deep lung. At 100 pm
lengths, the fraction of complex structures that survive passage through the nose and
throat in comparison with simple fibers may vary by a factor of five. This means that
large structures become relatively less respirable as their complexity increases.
However, during mouth breathing large clusters and matrices may enter the deep lung.
When all of the factors that Harris and Timbrell (1977) addressed are considered, the
efficiency of the deposition of asbestos structures in the deep lung is maximal for short,
thin, single fibers (less than 10 pm in length with a true diameter less than 0.7 pm).
The efficiency decreases slowly with increasing length (up to an effective limit of 200
pm), moderately with increasing complexity of shape, and rapidly with increasing
diameter (up to an effective limit of 2.0 pm, true diameter). Thinner fibers, down to the
lower limit of the range for asbestos fibers (0.02 pm, true diameter), are deposited with
roughly the same efficiency. Approximately 20 to 25% of the fibers between 0.7 and
0.02 pm in diameter (and less than 10 pm in length) are deposited in the deep lung.
Sussman et al. Findings
Based on a series of experiments on human tracheal bronchial casts, Sussman et al.
(1991a and b), also developed models of fiber deposition in the human lung. Such
experiments are in fact illustrative of several research groups who have developed
deposition models based on results from experiments on airway casts (for a review, see
Stober et al. 1993).
The results reported by Sussman et al. (1991a and b) appear to be generally consistent
with the results reported by Harris and Timbrell (above) and Yu and coworkers (below),
although the manner in which their results are reported make them somewhat less
directly comparable. Briefly, Sussman et al. report that deposition increases along most
generations of the bronchial tree with increasing fiber length and increasing airflow rate
7.10
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
for any fixed aerodynamic diameter. This increased deposition efficiency is
demonstrated for airway generations at least through the ninth bifurcation and is implied
to continue through to airway generations that would be representative of the
respiratory (pulmonary) portion of the lung (i.e. airway bifurcations greater than
approximately 16 to 22). For definitions and a description of airway generations, see
Section 4.4.
Findings of Yu and Coworkers
In a series of studies, Yu and coworkers combined an improved model of human lung
physiology (Asgharian and Yu 1988) with a series of more rigorous equations to
describe fiber mobility (Chen 1992) and used these to evaluate the deposition of
various types of fibrous materials in the lung. The trends indicated in their studies show
general agreement with those reported by Harris and Timbrell, but with several notable
refinements.
In a study of refractory ceramic fibers (Yu et al. 1995a), a maximum deposition
efficiency of 15% is reported for fibers that are approximately 6 pm long and
approximately 1 pm in diameter. This is close to the fiber size at which maximal
deposition is reported by Harris and Timbrell (1977). As with Harris and Timbrell, Yu et
al. also report that deposition efficiency decreases precipitously for thicker structures
and more slowly for thinner structures. For thinner structures, deposition efficiency
increases with both decreasing width and length. As fibers get longer, optimum
deposition occurs with decreasing thickness. Thus, for example, a maximum deposition
rate of 10% occurs for fibers that are 20 pm long at a thickness of 0.8 pm.
In a study of silicon-carbide whiskers (Strom and Yu 1994), the deposition model is
extended to fiber widths as narrow as 0.01 pm. Results from this study indicate that
fibers between 0.01 and 0.1 pm in thickness are deposited with a minimum efficiency of
5% up to lengths of approximately 40 pm before efficiency drops below 5%. For thin
fibers (thinner than 0.5 pm), shorter fibers tend to be deposited in the deep lung much
more efficiently than longer fibers. More than 25% of thin fibers shorter than 1 pm are
deposited in the deep lung following inhalation. Strom and Yu also report that the
efficiency of deposition in the deep lung of long structures increases substantially
during mouth breathing.
Comparing the results reported for refractory ceramic fibers (density = 2.7 g/cm3) and
silicon-carbide whiskers (density = 3.2 g/cm3), it also appears that the efficiency of
deep-lung deposition increases for thinner and for longer structures as the density of
the structures increases. Given the observed density effect, longer and thinner
7.11
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
chrysotile fibers would be deposited in the deep lung less efficiently than (denser)
amphibole fibers of the same size. However, shorter and thicker chrysotile structures
would be deposited somewhat more efficiently than similarly sized amphiboles. This
suggests that a greater fraction of the mass of chrysotile that gets deposited in the deep
lung will be composed of very short fibers and somewhat longer bundles than the mass
fraction of short fibers or longer bundles in the air breathed. Also, to the extent that
chrysotile fibers are curved, these would be deposited somewhat less efficiently than
straighter (amphibole) fibers of comparable size.
Based on the deposition efficiencies predicted by Yu and coworkers, fibrous structures
that reach the deep lung in humans are effectively limited to those thinner than
approximately 1 pm. It is also apparent that aspect ratio constraints affect only the
shortest fibers (i.e., those shorter than approximately 3 pm). The thickness constraint
for all longer structures is best described as a maximum width (rather than an aspect
ratio).
Rats versus Humans
Yu and coworkers also modified their models to evaluate the rates that fibrous materials
are deposited in rat lungs and compared these with results for humans. Such
comparisons have implications for the manner in which results from animal inhalation
studies are extrapolated to humans.
Results from Yu et al. (1994) suggest that pulmonary deposition of all fibrous structures
with lengths between about 1 and 100 pm and thinner than approximately 1 pm occurs
at much higher rates in rats than in humans. Fibers as long as 90 pm are deposited in
rat lungs at efficiencies exceeding 20% while fewer than 5% of structures this long are
deposited in the pulmonary region of human lungs. In fact, it is only structures
between 1 and about 20 pm within a very narrow range of thicknesses (centered
around 1 pm) that are deposited more efficiently in the deep lungs of humans than In
rats.
Yu et al. (1995a) also indicate that, even when deposition efficiencies are comparable
in rats and humans, due to differences in the total lung mass and breathing dynamics
across species, the resulting lung burdens (i.e. the mass or number of structures per
mass of lung tissue) are five to 10 times higher in the rat than in humans for any given
exposure. Lung burden per lung surface area are also higher in the rat than in humans.
To illustrate, assume rats and humans are similarly exposed to a concentration of
0.1 f/cm3 (100 f/L)of some fibrous material with a length at which both species retain
7.12
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
approximately 10% of the fibers inhaled. The Following Table 7.1 then indicates the
calculations required to determine the relative rates at which the lung (volume and
surface area) burdens in each species would develop.
Table 7-1:
ESTIMATION OF LUNG VOLUME AND LUNG SURFACE AREA
LOADING RATES FOR RATS AND HUMANS
Species
Body
Weight
(*g)
Lung
Volume
(L)
Lung
Surface Area
(m2)
Rest Breaths
per Minute
(bpm)
Tidal Lung
Volume
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
7.1.3 The Effects of Electrostatic Charge on Particle Respirability
Electrostatic charge has been shown to affect the retention of particles within the lungs
(see, for example, Vincent 1985). Since processes that generate airborne particles
generally involve some form of abrasion, airborne dust particles frequently exhibit
varying degrees of electrostatic charge. Although this potentially leads to variation in
the efficiency of particle retention in the lungs as a function of the source of the dust, a
detailed relationship between surface charge and retention was not described in this
paper. A more detailed and quantitative treatment was developed by Chen and Yu
(1993) and the implications of the Chen and Yu model are described below (following
discussion of results reported by Davis et al.).
Davis et al (1988) report that animals exposed to dusts containing fibrous chrysotile,
whose surface charge is reduced with a beta minus source, retain significantly less
chrysotile than animals dosed with dusts containing particles whose surface charge has
not been reduced. However, the magnitude of the difference in the mass of fibers
retained is less than a factor of two, implying that the absolute variation due to this
effect may be small. Further research in this area is needed.
Chen and Yu (1993) report that, based on modeling of lung deposition, overall
deposition increases with increasing charge density on the particles inhaled. However,
due to the pre-filtering by the naso-pharyngeal and and tracheo-bronchial portions of
the respiratory tract, the effects of electrostatic charge on deep lung deposition appear
to be only slight to modest.
Given the results of the above studies, the overall effects of electrostatic charge on
particle deposition in the deep lung appear to be relatively minor. Therefore, such
effects do not need to be considered explicitly when evaluating the health
consequences of asbestos.
7.1.4 General Conclusions Concerning Particle Respirability
Based on the information provided in the last several sections, it is apparent that in
humans:
• deposition of asbestos fibers in the pulmonary portion of the lung occurs
primarily at alveolar duct bifurcations;
• electrostatic effects on pulmonary deposition are likely minor
7.14
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
• fibers that are deposited in the pulmonary portion of the lung are largely thinner
than approximately 0.7 pm and virtually all are thinner than 1 pm (except during
mouth breathing, when thicker and more complex structures may be respired);
• the length of a fiber has limited impact on respirability up to a length of
approximately 20 pm, but the efficiency of deposition of longer fibers decrease
slowly with increasing length for longer fibers;
• as the length of the fibers that are inhaled increases, the thinner fibers are
deposited with greater efficiency. Thus, the longer the fibers inhaled, the thinner
the fibers retained;
• due to differences in density, shorter and thicker chrysotile structures will be
deposited more efficiently in the pulmonary portion of the lung than
corresponding amphibole structures and longer and thinner amphibole structures
will be deposited more efficiently than corresponding chrysotile structures;
• curly chrysotile structures are less likely to reach the pulmonary portion of the
lung than straight amphibole (or chrysotile) structures;
• except for a very narrow range of fiber sizes (centered around 6 pm in length
and 1 pm in diameter), virtually all size fibers are deposited with greater
efficiency in rat lungs than human lungs;.
• due to body morphology and the dynamics of breathing, rats exposed to similar
air concentrations will accumulate fiber burdens both per mass (volume) of lung
tissue and per lung surface area at a rate that is several times the rate such
burdens accumulate in humans; and
• the dynamics of fiber lung deposition can now be accurately predicted in great
detail using currently available models.
7.2 FACTORS AFFECTING DEGRADATION, TRANSLOCATION, AND CLEARANCE
Degradation and clearance mechanisms compete with deposition to determine the
fraction of asbestos that is retained in the lungs. Other (translocation) mechanisms
mediate the movement of asbestos from sites of initial deposition to various target
tissues within the lung and mesothelium. These factors affect all of the toxic end points
of interest. Studies indicating the dependence of the various contributing mechanisms
7.15
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
on fiber size and mineralogy are highlighted as well as studies indicating differences
between mechanisms in humans and laboratory animals.
The three units of the respiratory tract defined in the last section (naso-pharyngeal,
tracheo-bronchial, and bronchio-alveolar units) differ primarily by the types of clearance
(and translocation) mechanisms operating in each unit (Raabe, 1984). These are
summarized in Table 7-2 along with rough estimates of the time frames over which
each mechanism may operate.
7.16
Revision 1 - B/3AJ1
-------
TABLE 7-2
RELATIVE RATES. HALF-LIVES FOR PARTICLES CLEARED BY THE VAROUS OPERATING MECHANISMS OF A HEALTHY LUNO
(Thli labia needs lo ba completed)
TIMUBLUNO REOtME
Msrtisnlaws
(Componanl Machanlama)
SPECIES
FIBER
TYPE
PARTICLE
HALF-LIFE*
(days)
KINETIC
ORDER
MNERALOOICAL EFFECTS
SIZE EFFECTS
REFERENCE COMMENTS
NASAL-PMARVNOEAL
Evaporation and Swadowlng
Muco-dMry Transport
Human
Human
Part Idas
minimal
0 0028
Zero
NoElfed
NoEfled
Reaba 1984
TRACHEO-MONCHML
Mwco-dtery Trenaport
Human
Pallidas
0.021-0.21
Zero
No Eltad
NoElfed
Raaba 1984
PULMONARY (¦RONCMO-ALVEOLAR)
AM Phagocytoeis. Traniport lo MC EaeaMor
Ral
Rat
Particles
Short Chr
49
14
Ps - First
NoERed
Inhibited by Length/Cone
Flbars < 4 um
Stober et al. (1990) In Stobar et si. (1993)
Yu at al. (1B90)
Dissolution In Eidracalular Fluid
ii
Chr
Cre
ISO
11000
Zsro
Affects Rata
Diameter determines lifetime Hume and Rhnstldt 1992
Zollus at el. 1907
Tranaport totha IntantWum
(Phagocytoata and ajyuiaioii by apWialal calla)
(AM phagocytoala, tranaport through aplthalum)
(OMualve transport through tha apllhatum)
(Forcad mechenicd tranaport through tha apKhatum)
Sequestration
(Phagocidoali and IntemetesHon by apWtaM cala)
(AM ptiseocytoaie.lmmoMteaUon due to overload)
Ral
Pari Idas
2.3
???
77?
777
Stobar at al. (1990) In Stober el el (1893)
PULMONARY (INTERSTITIAL)
IM Phalogyctoala, Transport lo Lymphatics
CHECK THIS
Ral
Dog
Partlclss
Ams
2300
2200
Unspecified Negative E(ed
Inhibited by Length/Cone
Stober el al. (1990) In Stober el al. (1993)
Oberdorater et el. (1988)
DMuaiva Fluid Transport lo Lymph 7??
2200
check • should be none
chsck
Churg 1994
Transport to Endottwluai, Ptaura
OM phagocytosis, transport through MaretMum)
(DMuslve tranaport through lha MaratWum)
(Forcad mechanical tranaport through tha InteriUUum)
SequaatraUon
(Encapsulation In granulomatous tlssus)
(htsmabsflon by tntantM/andothaU cafa)
THE PLEURA
PM Phetogyctoaie. Tranaport to Lymphatic Stomala
Dissolution In mtaosMar Md
Sequestration
(EncapauMkn by granulomatous tissue)
(Phagocytoeis by maaathalal cadi)
(Sama as Above)
For zero order mechanisms, haHlvai raportad am half of tha tlnta required tar complate clearance tor lha process thai la con slant with lima
For Iral order machantama, lha tura tislflvos (I.e. lha lima nqukad tor half of lha Initial papulation lo diseppear) la raportad.
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
TABLE 7-2
7.17
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Briefly, the structures of the nose and throat are bathed in a continual flow of mucous,
which is ultimately swallowed or expectorated. The mucous traps deposited particles
and carries them out of the respiratory tract. The air channels of the tracheo-bronchial
section of the respiratory tract are lined with cilia and mucous secreting cells. As in the
nose and throat, the mucous traps particles deposited in these air pathways and the
ciliary escalator transports the mucous up to the throat where It may be swallowed or
expectorated.
Neither the alveolar ducts nor the alveoli of the pulmonary compartment of the lung are
ciliated (inferred from St. George et al. 1993). Therefore, particles deposited in this
section of the respiratory tract can only be cleared by the following mechanisms:
• if the deposited particles are soluble, they may dissolve and be
transported away from the lungs in blood or lymph; or
• if they are sufficiently compact, they may be taken up by alveolar
macrophages and transported outward to the muco-ciliary escalator of the
tracheo-bronchial portion of the respiratory tract.
Due to a combination of chemical and physical stresses in the environment of the lung,
deposited asbestos structures may degrade by splitting. Longitudinal splitting, primarily
of bundles, produces thinner structures and transverse splitting produces shorter
structures. In both cases, the number of structures produced may be larger than the
number of structures initially deposited.
By changing the size and number of structures that were initially deposited in the lungs,
splitting may affect the rates and efficiency with which the various other degradation
and clearance mechanisms operate.
Particles and fibers that are deposited in the pulmonary portion of the lung may also be
transported by a variety of mechanisms into and through the epithelium lining the
alveolar ducts and alveoli to the underlying interstitium and endothelium that are
located within the interalveolar septa (See Section 4.4). In those portions of the lung
parenchyma that lie proximal to the pleura, such mechanisms may also facilitate
transport to the mesothelium. Putative mechanisms by which such transport may
occur include:
• if particles are sufficiently compact to be phagocytized by alveolar
macrophages, they may be transported within macrophage "hosts"
through the epithelium to the interstitium;
7.18
Revision 1 • 6/3/01
-------
TABLE 7-7:
SUMMARY DATA FOR ANIMAL INHALATION EXPERIMENTS CONDUCTED BY DAVIS AND COWORKERS"
Fiber
Type
Description
Abbreviations
Mas*
Concentration
(mg/m'|
PCM
f/ml
Number of
Animals
Number of
Benign
Pulmonary
Tumors
Number of
Malignant
Pulmonary
Tumors
Total Number
ot Pulmonary
Tumors
Mesotheliomas
Reference
Chryaotile
UICC-A
UC
2
390
42
6
2
e
1
Davis et al. 1978
Chrysolite
mcc-A
UC
10
1.950
40
7
8
15
0
Davis et al. 1978
CtvysotMe
Long
LC
10
5.510
40
8
12
20
3
Davis et al. 1968a
Ctaysotte
Short
SC
10
1.170
40
1
6
7
1
Davis et al. 1988a
Chrysotfle
UCC-A
UC
99
2.560
36
6
8
14
0
Davis et al. 1988b
CtaysoMe
UICC-A (Discharged)'
DC
99
2.670
39
4
6
10
1
Davis etal. 1968b
CtvysoOe
WDCYarW
WC
3.6
679
41
5
13
18
0
Davis etal. 1986b
Amosits
UICC
UA
10
550
43
2
0
2
0
Davis et al. 1978
Annate
Long
LA
10
2,060
40
3
8
11
3
Davis etal. 19B6a
Annate
Short
SA
10
70
42
0
0
0
1
Davis et al. 1666a
Ooddolite
UICC
UR
4.9
430
43
2
0
2
1
Davis et al 1978
- -»¦«-*«» -
UfOCKJOHIQ
UICC
UR
10
860
40
1
0
1
0
Davis et al. 1978
Tremotlte
Korean
KT
10
1,600
39
2
16
18
2
Davis etal. 1B85
Nona
Control
C
0
20
0
0
0
0
Davis et al 1078
Nona
Control
C
0
36
0
0
0
0
Davis et al 1S65
None
Control
C
0
61
1
1
2
0
Davis et al. 1986a
None
Control
C
0
64
1
1
2
0
Davis et al 1966b
None
Control
C
0
47
1
1
2
0
Davis el al. 16B8a
t. Source: Berman et al 1995
2. Exposure occurred for 7 hours per day, 5 days per week for 1 year.
3. UICG-A chrysotlle in this experiment was treated with mixed polarity air (produced with a source of beta radiation) following generation to reduce the surface charge on individual particles
within the dud.
4. Chryaotile samples used for dust generation in this experiment were obtained from material treated by a commercial wet dispersion source
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
• if particles are sufficiently compact to be phagocytized by the epithelial
cells lining the air passageways of the deep lung, they may be transported
into cell interiors or transported through to the basement membrane, the
interstitium, the endothelium, and (eventually) the pleura;
• particularly when associated biological effects cause changes in the
morphology of epithelial cells (Section 7.3.6), particles may diffuse
between the cells of the epithelium to underlying tissues; and/or
• particles may be transported through respiratory epithelium mechanically
due to physical stresses associated with respiration within the lung.
Although the transport of fibers and particles from airway lumena to the interstitium is
apparent in many studies (see below), the precise mechanisms by which such transport
actually occurs has yet to be delineated with certainty.
Particles deposited in the interstitium can also be cleared and the processes by which
these particles are ultimately cleared are similar to, but may be substantially slower
than, the mechanisms by which particles deposited in airway spaces can be cleared.
Such mechanisms include:
• if the deposited particles are soluble, they may dissolve and be
transported away from the lungs in blood or lymph; or
• if the particles of the interstitium are sufficiently compact to be
phagocytized by interstitial macrophages, they may be taken up and
transported to the lymphatic system for removal.
The mechanisms by which particles that reach the pleura and mesothelium may be
cleared are also similar to those operating in the interstitium:
• if the deposited particles are soluble, they may dissolve and be
transported away from the lungs in blood or lymph; or
• if the particles that reach the pleura are sufficiently compact to be
phagocytized by pleural macrophages, they may be taken up and
transported to the lymphatic system for removal.
7.19
Revision 1 • 6/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Note, particles cleared from the pleura by macrophages appear subsequently to be
deposited at sites of lymphatic drainage along the pleura (i.e. at lymphatic ducts) from
where they are ultimately cleared in lymph (Kane and McDonald 1993).
The various degradation, clearance, and transport mechanisms that affect the retention
of asbestos in the lung and other target tissues (identified above) exhibit disparate
kinetics that may be further altered by the size, shape, mineralogy, and concentration of
the particles affected. Therefore, the kinetics of these mechanisms are considered
below. The mechanisms evaluated include:
• dissolution;
• muco-ciliary transport;
• macrophage phagocytosis and transport; and
• diffusional transport.
Evidence for the existence of these mechanisms and inferences concerning their
kinetics derive primarily from retention studies, which may include both studies of
retained structures in animals following either short-term or chronic exposure or human
pathology studies in which the lung burdens of deceased individuals are correlated with
their exposure history. Other information also comes from in-vitro studies. Various,
increasingly sophisticated models have also been developed to predict the individual
and combined effects of these mechanisms.
7.2.1 Animal Retention Studies
Retention studies track the time-dependence of the lung burden of asbestos or other
particulate matter (i.e. the concentration of particles in the lung) during or following
exposure. Thus, such studies are designed to indicate the degree to which inhaled
structures are retained. Depending on the time frame evaluated, however, effects due
to deposition and those due to clearance may not easily be distinguished in such
studies. Moreover, due to the near impossibility of isolating the various compartments
of the lung when preparing for quantitative anlaysis of tissue burden (e.g. the pure
respiratory components vs. the larger airways or the tissues directly associated with
airway lumena vs. the underlying interstitium or endothelium), it is nearly impossible to
separate the effects of the various clearance mechanisms, which typically operate over
vastly different time scales (Table 7-2). This is why modeling has proven so important
to distinguishing effects attributable to individual mechanisms.
Results from retention studies must be evaluated carefully. In addition to the limitations
highlighted above, the lung burden estimates from such studies may be affected by the
7.20
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
manner in which asbestos is isolated from lung tissue for measurement and the manner
in which the concentration of asbestos is quantified (Chapter 5). For example, lung
burden estimates may vary substantially depending on what portions of lung
parenchyma are sampled or whether whole lungs are homogenized. Results may also
vary depending on whether lung tissue is ashed or dissolved in bleach during sample
preparation. More importantly, because several clearance mechanisms are affected by
the size and even the mineralogy of the structures being cleared, studies (particularly
older studies) that track lung burden by mass or by total fiber number may not
adequately capture such distinctions.
7.2.1.1 Studies involving short-term exposures
The latest retention studies tend to focus on the fate of long fibers (typically those
longer than 20 |Jm) in support of the generally emerging recognition that these are the
fibers that cannot be readily cleared from the pulmonary compartment of the lung and
that, not coincidentally, contribute most to disease (Section 7.4).
Hesterberg et al. (1998a), for example, tracked the time-dependent retention in rats of
two fiber categories: (1) WHO fibers1 and (2) WHO fibers longer than 20 pm for a range
of manmade vitreous fibers (MMVF's), a refractory ceramic fiber (RCF1a), and amosite
following a 5-day (6 hr/day), nose-only exposure. Rats were sacrificed at intervals up to
a year following exposure. The amosite was size selected to contain a high proportion
of fibers longer than 20 pm. Aerosol concentrations were also adjusted to maintain
target concentrations of 150 f/cm® for long fibers for each sample tested. Airborne
mass concentrations varied between 17 mg/m3 for amosite to as much as 60 mg/m3 for
the other fiber types. Lungs (without trachea or main bronchi) were weighed and stored
frozen. For analysis, each lung was dried to constant weight, minced and a portion was
ashed. The ashed portion was further washed with filtered, household bleach, then
filtered and applied to an SEM stub. Fiber numbers and dimensions (in both aerosols
and tissue) were determined by SEM with a minimum of 200 fibers counted. In
addition, analysis continued until a minimum of 30 fibers longer than 20 jjm were
counted.
Hesterberg et al. tracked the ratios of retained fiber concentrations with time to the
concentration retained 1-day following cessation of exposure. The observed time-
dependent decay in these ratios were then fit to one-pool (single first order decay) or
two-pool (weighted sum of two first order decays) models. With zero time assumed to
WHO fibers are those longer than 5 (Jm, thinner than 3 pm with an aspect (length to
width) ratio greater than 3 (WH01985).
7.21
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
be the time immediately following cessation of exposure. The authors recognize that at
least some clearance likely takes place during the five days of exposure so they
expected the assumption that retained concentrations at the end of exposure on day 5
to be equal to deposited concentrations would cause their analysis to slightly
underestimate clearance rates. They also recognized that waiting 24-hours after
cessation of exposure to measure retention allows some short-term clearance of upper
airways, so that they expected their analysis would better focus on slower clearance
from deeper in the lung.
Results reported by Hesterberg et al. indicate that the dimensions and concentrations
of fibers in aerosols from the five synthetic fibrous materials were all similar but that the
amosite aerosol contained a substantially greater number of fibers (including the
longest fibers) but that the fibers, on average, were somewhat shorter and substantially
thinner than the other aerosols. Of the fibers initially deposited in the lung (based on
measurements made 1-day following cessation of exposure, comparable fiber numbers
of long (> 20 fjm) fibers were retained across all six fiber types. Deposited
concentrations of fibers 5-20 |jm in length were more variable, but values within one
standard deviation still overlapped. About 6 times as many short amosite fibers (< 5
pm) were initially deposited than for any of the other fiber types. The authors also
indicate that the dimensions of retained fibers were generally shorter and thinner than
the original aerosol and were much more similar across retained fiber types than the
original aerosols.
Clearance of long fibers (>20 pm) for all six fiber types could best be described using a
two-pool model. The first pool cleared relatively rapidly (within the first 90 days) and
represented a minimum of 65% of the lung burden observed 1-day following exposure.
The second pool cleared much more slowly. For amosite fibers in the second pool,
during the approximately 275 days of clearance, retention was only reduced to 80% of
the 90-day value. In contrast, all five of the synthetic fibers were reduced to less than
30% of their 90-day value during this period. For amosite, the first pool decayed with a
half-life of 20 days (90%CL: 13-27) and all of the other fibers with half-lives of 5-7 days
(with varying confidence bounds). For the slower pool, amosite fibers exhibited a half-
life of 1160 days (90% CL: 420-®) with the other fibers showing half lives varying
between 24 and 179 days. The combined, weighted half-life for amosite was 418 days
(90%CL: 0-1060). The authors also note that data reanalyzed from an earlier study
(Hesterberg et al. 1996) indicate a corresponding weighted half life for crocidolite of 817
days (246-oo) and indicate that this was best fit using a single exponential (a one-pool
model).
Hesterberg et al. indicate that in this and previous studies approximately 20 to 60% of
long fibers typically clear from the lung within 2-weeks post exposure. They further
7.22
Revision 1 • B/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
suggest that this rapid clearance may be attributable to muco-ciliary clearance from the
upper respiratory tract. They further report from the present study that short amosite
fibers cleared much more rapidly than long fibers. Fibers < 5 pm in length were reduced
by 90% in the first 90 days (in comparison to 65% for long fibers). However, from 90 to
365 days, little or no clearance was observed for amosite fibers of any length.
For four of the synthetic fibers, long fibers cleared at the same rate as short fibers (all
more rapidly than amosite) and the authors report that the data suggest transverse
breakage for these fibers. Moreover, they attribute the more rapid clearance of long
fibers among the MMVF's to dissolution, since these fibers exhibit in-vitro dissolution
rates that are rapid relative to the time scale of macrophage clearance. One synthetic
fiber MMVF34, which is a stonewool) disappeared much more rapidly than any other
fiber and the long fibers disappeared more rapidly than the short fibers. MMVF34
shows the greatest in-vitro dissolution rate at neutral pH for any of the fibers tested in
ths study and dissolves particularly rapidly at pH 4.5 (the pH found in the phagosomes
of macrophages). The authors postulate that clearance of all of the synthetic fibers are
enhanced over amosite by dissolution and breakage.
In summary, Hesterberg et al. observed that:
• multiple clearance mechanisms (operating over multiple time scales)
contribute to clearance;
• for sufficiently soluble fibers, long fibers clear more rapidly than short
fibers;
• for insoluble fibers, a subset of long fibers clears rapidly within the first few
months following exposure and the remaining long fibers clear only
extremely slowly, if at all;
• short fibers of all types are cleared at approximately the same rate (much
more rapidly than long, insoluble fibers);
• a small, residual concentration of short fibers may not always clear and
may remain in the lungs (sequestered in alveolar macrophages) for
extended periods; and
• in this study, there is some suggestion that short amosite fibers clear
somewhat more slowly than short fibers of the other, non-asbestos
mineral types studied.
72.2
Revision 1 • 6/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Regarding the last observation, whether this is attributable to differences in fiber
thicknesses among the various mineral types, due to partial contributions (even among
short structures) to dissolution, or due to a unique, toxic effect of amosite is unclear.
However, the likeliest of these candidate hypotheses is that the effect is due to partial
dissolution.
This general pattern of observations are consistent with the findings of most, recent
retention studies following short-term exposure.
In an earlier study of similar design, Bernstein et al. (1996) evaluated the deposition
and clearance of a series of 9 glass and rock wools. These authors similarly found that
clearance could be modeled using a double exponential for all length fibers (in similar
length categories of < 5, 5-20, and > 20 pm) and that for soluble fibers, long fibers clear
more rapidly than short fibers (with the intermediate length fibers in between).
For the Bernstein et al. study, if one assumes that the pool of longer-lived fibers is
representative of macrophage clearance, this suggests that the efficiency of clearance
by macrophages decreases with increasing fiber length and that the longest structures
are not phagocytized at all, so that they remain exposed to the extracellular medium
where dissolution occurs. Lending further support to this interpretation, the authors also
report that the clearance rate for long fibers correlate with measured in-vitro dissolution
rates at neutral pH while the clearance rates for short fibers neither correlate with in-
vitro dissolution rates at neutral pH or at pH 4.5. Although the latter pH corresponds to
the pH found in the phagosomes of macrophages, there is likely too little fluid available
in such organelles to support efficient dissolution. The authors also indicate that a
sufficient number of fibers were counted during the study to suggest that breakage is
not playing a role in clearance (except at very early times) and that the clearance rate
for short fibers appears to be the same or slower than that observed for nuisance dusts.
In another, earlier study of similar design Eastes and Hadley (1995) evaluated four
types of manmade mineral fibers (MMMF's) and crocidolite. All of the samples
(including crocidolite) had been size-selected to assure a large fraction of fibers longer
than 5 pm. Unfortunately, due to differences in reporting, it is not possible to compare
the initial loading of crocidolite fibers to those reported for amosite in the Hesterberg et
al. (1998a) study. However, results from this study further support the physical
interpretation of clearance suggested in the studies discussed above. In fact, the
authors report that the time-dependent size distribution of retained fibers observed in
this study agree well with a computer simulation of fiber clearance. The simulation
assumes that long fibers dissolve at the rate measured for such fibers in vitro and that
7.24
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
short fibers of every type are removed at the same rate as short fiber crocidolite (which
is practically insoluble). This is strong evidence that short fibers are cleared by
macrophage phagocytosis and that long fibers cannot be cleared by macrophages, but
may dissolve in extracellular fluid provided that they are sufficiently soluble.
Regarding crocidolite, the data from the Eastes and Hadley study suggest that short
crocidolite fibers appear to clear at a rate that is somewhat slower than observed for
any of the short MMVF fibers. Importantly, however, the interpretation of short fiber
clearance in this paper is somewhat confounded because, unlike the studies discussed
above, short fibers in this paper are defined as all fibers < 20 pm so there may be some
confounding with MMVF fibers that are dissolving. As previously indicated, the
Hesterberg et al. (1998) work also suggests that short asbestos (amosite) fibers may
clear more slowly than short fibers of differing mineralogy and Hesterberg et al only
includes fibers < 5 pm in their definition. Nevertheless, it is still possible that some
effects due to dissolution may still be affecting the clearance of these shorter fibers.
Surprisingly, a visual inspection of the data presented in their table suggests a lack of
any long-term clearance for long fiber crocidolite (> 20 pm). Vet, the authors model
long fiber crocidolite clearance using a single exponential (suggesting no rapidly
clearing compartment). The long-term half-life reported for crocidolite in this study is
approximately 220 days (with estimated confidence limits of 165 to 566 days). This
overlaps with the long term clearance half-life reported by Hesterberg et al. (1996) for
crocidolite of approximately 820 days (246-°°).
Equally surprising, Hesterberg et al. also modeled crocidolite clearance as a single
exponential, which might suggest better penetration to the deep lung by crocidolite, less
clearance by muco-ciliary transport or alveolar macrophage transport, or better
penetration to the interstitium than other fibers. More likely, however, it may simply
indicate that the two-pool model does not represent a statistically significant
improvement in model fit over the one pool model. However, relative size distributions
would need to be evaluated carefully before drawing any such conclusions. Eastes and
Hadley also report clearance of short fiber crocidolite is modeled as a double
exponential with short and long half-lives of: 25 and 112 days, respectively. Since this
fiber category contains fibers up to 20 pm in length (in this study only), this does
suggest at least some contribution from muco-ciliary and alveolar macrophage
mediated clearance for crocidolite.
In two studies, Coin et al. (1992 and 1994) evaluated the fate of chrysotile fibers in rats
exposed for 3 hrs to 10 mg/m3 (reportedly containing > 5000 fibers longer than
5 Mm/cm3). For lung analysis, the left lung was separated into peripheral and central
7.25
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
regions under a dissecting microscope. Slices of peripheral and central portions were
separately weighed and minced. Tissue was digested in sodium hypochlorite and then
filtered. A quality control test indicated that the digestion process caused a slight
(-10%) decrease in fiber number and slight decreases in fiber diameter and fiber
length. Fiber size distributions were evaluated by SEM. A stratified counting procedure
was employed to assure equal precision for each length category of interest.
Measurements for each category were then converted to mass equivalents.
Results from the Coin et al. studies indicates no difference between deposition in
central or peripheral regions of the lung. They also confirm that chrysotile splits
longitudinally in the lungs with a half life that is competitive with the clearance rates
measured in this study. Clearance was found to be very length dependent so that
rates decrease from a half life of about 10 days for fibers about 4 pm in length through
30 days for fibers 8 pm to 112 days (which is no different from zero) for fibers longer
than 16 pm (all after adjusting for longitudinal splitting). Importantly, the brief foliowup
period (30 days) is too short to provide an adequate evaluation of the longer term
clearance pools observed in other studies and certainly too short to evaluate any
effects potentially associated with chrysotile dissolution. Also, that the decay curves for
clearance were limited to four points, makes evaluation of the slopes for these curves
highly uncertain.
Coin et al. also report that the mass of chrysotile deposited during these short
exposures (i.e. no more than 20 pg) is very small compared to levels at which overload
has been reported to occur (approximately 1 mg, see, for example, Yu and Yoon 1991)
and that the volume of the 16 pm fibers, which have an average diameter of 0.2 pm
and therefore a mean volume of 0.5 pm3, is small relative to the volume at which
macrophage clearance of non-fibrous particles is reported to be hindered (Morrow
1988). Thus, the authors conclude that fiber length presents an additional constraint
on macrophage clearance, independent of any other overload. They also indicate that
inhibition of clearance due to fiber length is independent of fibrosis.
Coin et al. also discuss the effect of fibrosis on clearance. They indicate that, although
increased concentrations of short fibers are observed in focal areas of fibrosis, it is
more likely that such fibers accumulate because clearance is hindered by fibrosis in
these areas than the hypothesis that the short fibers are causing fibrosis. This is
because, as they point out, there are too many studies demonstrating the lack of ability
of short fibers to induce fibrosis.
Evidence in the Coin et al. studies suggests that no translocation from central to
peripheral regions of the lung were detected. An upper bound rate that is about 20% of
7.26
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
clearance is reported. However, the short followup time in this study would have
precluded slower processes from being detected. Despite the lack of evidence of
translocation, the authors report that duct bifurcations in peripheral regions of the lung
where fibers are deposited are no more than 1-2 mm from the visceral pleura. In fact,
in the 1994 study, the authors show that 50% of the primary duct bifurcations in the
peripheral portion of the rat lung occur within 1 mm of the visceral pleura and some
occur as close as 220 |Jm. Deposited fibers may also affect the pleura by inducing
generation of diffusable, inflammatory agents.
A short term inhalation study by Warheit et al. (1997) evaluated retention of chrysotile
and aramid fibers. In this study, rats (and hamsters) were exposed nose-only for
6 hrs/day, 5 days/wk for 2 weeks by inhalation to UICC chrysotile and p-aramid fibers
(each at two doses of 460 or 780 fibers/ml, although the size range of these fibers is not
stated nor is the manner in which they were analyzed). Fixed lungs were digested in
chlorox during preparation for asbestos analysis. Animals were followed for up to a
year post exposure.
As in studies described above, results from the Warheit et al. study indicate rapid
clearance of short chrysotile fibers but slow to non-existent clearance of fibers longer
than 20 |Jm. In contrast, aramid fibers apparently degrade and are subsequently
cleared fairly rapidly in vivo. Based on the data provided in figures, although the
reported concentrations of chrysotile and aramid fibers to which animals were exposed
were equivalent, at both the lower and higher concentrations, it appears that rats
initially retain three to five times as many aramid fibers as chrysotile fibers (at least for
the size range counted, which was not reported). For both fiber types, clearance
appears rapid for an initial period of approximately 90 days post exposure. During this
time, the mean length of chrysotile fibers also appears to increase steadily, which
suggests rapid, preferential clearance of short structures. After the initial period, it
appears that (as the authors suggest), a residual concentration of longer fibers are
cleared only very slowly, if at all.
Oberdorster et al. (1988) instilled a 3 ml suspension of irradiated amosite into the
bronchio-alveolar space of the right diaphragmatic lobe of the lungs of dogs to evaluate
clearance and transport. The amosite used was modified by sedimentation from UICC
amosite to contain only fibers shorter than 20 pm. One dog also had unmodified UICC
amosite instilled directly into a lymph node in the thigh. The dogs had been cannulated
to allow collection of lymph from the right lymph duct-RLD and the thoracic duct- TD
(both in the neck).
7.27
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Results from Orberdorster et al. indicate that within 4 hours following instillation in the
lung, low activity was noted in postnodal lung lymph but not in either the RLD or TD.
Within 24 hours, however, activity and fibers (determined by SEM) were observed in
both the RLD and the TD. The median length of fibers observed in the lymph were
significantly longer than the instilled material, although there appeared to be a cutoff
length of 16 pm in fibers observed at nodes and 9 pm in fibers observed directly in
lymph. Fibers recovered from lymph were also significantly thinner and appeared to
exhibit an absolute cutoff at a maximum width of 0.5 Mm. Fibers recovered from the
TD and RLD in the dog that had unmodified UICC amosite instilled directly into leg
lymph were all short (with a maximum length of 6 pm). Since collection times were all
short, the authors indicate that it is unknown whether longer fibers would have been
observed at later times. The authors also note the almost total absence of fibers
shorter than 1 pm in lymph, .which they assume are cleared rapidly and efficiently by
alveolar macrophages.
Oberdosrster et al. also report that a rough calculation, based on the fraction of the
material originally instilled that was recovered in the first 24 hours, it would take
approximately 6 years to clear all of the instilled asbestos (assuming no other clearance
mechanisms were active).
Everitt et al. (1997) performed a short-term inhalation study that is interesting
particularly because it focused on pleural (as opposed to lung) fiber burden. The
authors exposed rats and hamsters to one type of refractory ceramic fiber (RCF-1) by
nose-only inhalation for periods of 0, 4 and 12 weeks and animals were held for
observation for up to an additional 12 weeks post exposure. Exposures were
conducted for 4 hrs/day, 5 days/wk, at 45.6 +/-10 mg/m3 Groups of 6 animals were
held for 0,4,12, and 24 weeks to determine pleural fiber burden. An agarose casting
method was reportedly used to recover fibers from the pleura. Analysis was by electron
microscopy. Fibers were observed in the pleura at each time point examined (including
samples from rats sacrificed immediately following the last day of a five-day exposure).
Fibers were all reported to be short and thin (geometric mean length: 1.6 pm with GSD:
1.8, geometric mean diameter: 0.1 pm with GSD: 1.5). Concentrations averaged
approximately 40,000 fibers (per whole pleura??? - units not reported). The authors
indicate that such fibers would not typically be visible by optical microscopy. They also
indicate that use of casts may be a more efficient method of recovering fibers from the
pleura.
Everitt et al. indicate that observation of rapid translocation of short, thin fibers to the
pleura has also been observed in studies of chrysotile so these results are not unique.
Although it is stated that the mechanisms facilitating translocation are currently
7.28
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
unknown, the authors indicate that their finding of site-specific mesothelial proliferation
supports observations by Boutin et al. (1996) that asbestos fibers accumulate in the
parietal pleura of humans at sites associated with lymphatic drainage. Kane and
McDonald (1993) have suggested that fibers are transported to these locations by
pleural macrophages. However, the mechanisms by which fibers are transported from
the lung to the pleura are still unconfirmed.
Older Retention Studies
Although the older retention studies generally support the results of newer studies
(such as those cited above), older studies are sometimes limited by such things as the
tracking of lung burden in terms of fiber mass or use of analytical techniques such as
infrared spectroscopy for detection of asbestos, which are neither capable of
distinguishing individual fibers nor provide any information on their sizes. Tracking of
lung burdens in terms of mass may not reflect the fate of long, thin fibers, which (by
increasing concurrence) appear to be the legitimate focus of studies evaluating
biological hazards attributable to asbestos.
In two studies (Roggli and Brody 1984 and Roggli et al. 1987), Roggli and coworkers
tracked the behavior of chrysotile (not UICC) and UICC crocidolite in rats following 1 hr
exposure by inhalation to 3.5-4.5 mg/m3 dusts. The authors indicate that this results in
deposition of approximately 21 pg of dust. Portions of the lower lung lobes of selected
rats were collected and digested for asbestos analysis using a scheme that was shown
to be representative. To evaluate size distributions, more than 400 fibers from each
sample were characterized by SEM. Fiber dimensions were then used to estimate total
fiber mass.
Based on their study, Roggli and coworkers indicate that similar fractions of inhaled
chrysotile and crocidolite dust are deposited in the lung during inhalation (23% and
19%, respectively). The authors therefore concluded that respirability and deposition
do not depend on fiber type. Importantly, however, the manner in which this study was
conducted does not facilitate distinguishing deposition in the deep lung from deposition
in the upper respiratory tract.
Roggli and coworkers further indicate that clearance rates for the two fiber types appear
comparable. Of the chrysotile initially deposited, they report that 81% of this material is
cleared after four weeks. Similarly, 75% of the crocidolite is cleared. Importantly,
because this is based on total mass (estimated by summing volume contributions from
observed fibers), it may not reflect the specific behavior of long, thin structures.
Therefore, it is difficult to compare such results with those of more recent studies.
7.29
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
However, the authors do report that short structures are cleared more readily than long
structures and that chrysotile. is observed to split longitudinally in vivo (based on
observation that the total number of chrysotile structures initially increases and the
mean length increases). The authors further conclude that clearance rates appear to
be independent of fiber type.
In another short-term study, Kauffer et al. (1987) report that the average length of
retained chrysotile structures increases in rat lungs following 5-hrs inhalation of
chrysotile dust. Based on their results, the authors report that fibers shorter than
approximately 8 pm are preferentially cleared. Kauffer and coworkers also confirm that
chrysotile fibrs split longitudinally in the lung. In fact, several other studies (Le Bouffant
et al. 1978, Kimizuka et al. 1987) also provide supporting observations that chrysotile
fibers (or bundles) split longitudinally in the lung.
In two studies (Morgan et al. 1978 and 1980), Morgan and coworkers report on the fate
of fibers following short term inhalation of radio-labeled fibers by rats. In the first study
(Morgan et al. 1978), rats inhaled UICC anthophyllite at 35 mg/m3 for a total of 8.4 hrs
spread over three days. The authors report that the rats retained approximately 190 pg
of dust at the end of exposure, mostly in the alveolar region; the authors assumed that
conducting airway clearance is sufficiently rapid to clear this portion of the lung within a
few days. Beginning about 7 days following exposure, the rats were then sacrificed
serially for a period up to 205 days following exposure. Because anthophyllite fibers
are relatively thick, fibers were analyzed by optical microscopy. Fibers were
determined both in free cells (mostly macrophages) recovered in bronchopulmonary
lavage and in lung tissue. Tissue samples and cells were digested in KOH and peroxide
in preparation for fiber analysis.
Based on this first study, Morgan et aJ. report that anthophyllite lung content declined
steadily by a process that could be described as a simple first order decay with a half
life of approximately 76 days. Free macrophages recovered by lavage, initially
contained about 8 pg and this too declined steadily with a half life of about 49 days.
The authors further indicate that, if the number of macrophages remains constant with
time (i.e. they are replaced at the same rate they are cleared), then the decay of the
load in the macrophages should match what is observed in the rest of the lung. They
suggest that the discrepancy may be due either to an influx of an increasing number of
macrophages in response to injury with time and/or to transfer of some fibers through
the alveolar wall. They also cite unpublished work indicating that uptake of fibers by
alveolar macrophages is essentially complete within hours after cessation of exposure.
7.30
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
The authors also report that, initially, the lengths of fibers recovered in lung lavage was
greater than in the original aerosol but that the prevalence of the longest fibers
decreased after the first seven days. In lung tissue, however, the fraction of longer
fibers (among total fibers) steadily increased with time. This suggests rapid clearance
of naked fibers by muco-ciliary transport (which is a length independent process) with
later times dominated by slower clearance in the deep lung by alveolar macrophages,
which is a length-dependent process.
Morgan and coworkers also radiometrically determined the fraction of fibers in rat
faeces (prior to sacrifice). They assumed that after 14 days, this would represent the
fraction of asbestos cleared primarily from the alveolar region of the lung. Initially 1.4%
of lung anthophyllite content was excreted daily but this fell to 0.5% after 120 days.
The authors indicate that this suggests that the elimination from asbestos in lung tissue
cannot be described by a single exponential because multiple processes are involved
and that, over longer periods of time, the slower processes become increasingly
important.
In the second study, Morgan et al. track the fate of several size-selected radiolabeled
glasses in rats, again following short term inhalation. From an analysis of the size
dependence of deposited fibers in this study, the authors suggest that alveolar
deposition in the rat is limited to structures with aerodynamic equivalent diameters less
than about 6 pm and that deposition in this region of the lung falls precipitously for
fibers with thicknesses between about 2 and 3 pm (aerodynamic equivalent diameter).
For fibers that are the density of asbestos, this represents an upper bound limit to
alveolar deposition for the absolute thickness of a fiber of approximately 1.5 pm with
fibers deposition of fibers thicker than approximately 0.7 pm being drastically reduced.
This is in concordance with conclusions concerning deposition provided in
Section 7.1.4. Alveolar deposition efficiency is also shown to decrease with increasing
fiber length, at least for fibers longer than approximately 8 pm, also in concordance with
findings presented in Section 7.1.4.
Intratracheal Instillation
The fate of fibers following intratracheal instillation into the lungs has also proven
informative in some studies. For example, Wright and Kushner (1975) intratracheally
instilled paired samples each of several types of glass fibers, fluoramphibole, and
crocidolite into guinea pigs. For each mineral tested, a sample with predominantly short
structures (25 mg total dose for crocidolite, reportedly 99% < 5 pm) and another with
predominantly long structures (4 mg total dose for crocidolite, reportedly 80% >10 pm)
were evaluated. Unfortunately, the authors do not report how fibrous structures were
7.31
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
characterized. Results in the cited paper report observations only after 2-years
following the last injection.
Wright and Kuschner report that long structures uniformly caused fibrosis (primarily
involving the respiratory bronchioles and alveoli and abutting the terminal bronchioles)
while the short structures were uniformly phagocytized and generally removed to
thoracic lymph nodes. Among other things, clearance to lymph suggests that fibers
reached the interstitium (Section 7.2.5). It is interesting that, even after 2 years of
recovery, the authors observe elevated levels of macrophages in the alveoli of animals
dosed with short structures. Based on the relative size distributions of the samples
analyzed, the authors report that structures up to 10 }jm in length appear to be
efficiently scavenged by macrophages. Based on the observation of larger numbers of
short structures than expected in comparison with their fractions in the original samples,
the authors further conclude that glass structures underwent biodegradation so that
longer structures broke down into shorter structures that could be phagocytized.
Wright and Kuschner also report that long fibers are occasionally visible within the
fibrotic interstitium of dosed animals. The long-fiber dosed animals also show
macrophages in hilar lymph nodes containing fibers that are too small to resolve and all
of them are short. In short-fiber dosed animals, some fibers are seen to remain in the
lung within aggregates of macrophages, both in alveoli and the interstitium. Short-fiber
dosed animals also show many more macrophages within the hilar lymph nodes than
long-fiber dosed animals.
In two reports of the same study (Bellmann et al. 1986 and 1987), Bellmann and
coworkers followed the fate of UICC chrysotile, UICC crocidolite, several fibrous
glasses, and other manmade mineral fibers following a single intratracheal instillation of
0.3 ml of fibrous material in rats. Groups of rats were then sacrificed at 1, 6,12,18,
and 24 months following instillation. Lungs were low temperature ashed and the
resulting, filtered suspension analyzed by transmission electron microscopy. Some of
the fiber types were also acid treated with 0.1 M oxalic acid for 24 hrs prior to
instillation.
Bellmann and coworkers report that short fibers (< 5 pm) from all of the fiber types were
shown to be cleared from the lungs with half-lives of approximately 100 days, with the
asbestos varieties tending to exhibit slightly longer half lives than the other fibers.
Short crocidolite fibers exhibited a half life of 160 days The half-life for clearance of
short chrysotile was reported to be 196 days (the longest of all). However, this was
attributed to positive contributions from breakage of longer fibers.
7.32
Revision 1 • 0/3/01
-------
PRELIMINARY WORKING DRAFT- DO NOT COPY OR QUOTE
Bellmann and coworkers report that the behavior of the different long fibers (> 5 pm) for
the different fiber types was radically different. The authors report no observed net
decline in long crocidolite fibers over the 2 yrs of followup. The also report no
observable changes in width of these fibers with time. In contrast, long chrysotile fibers
increased in number with time throughout the 18 month followup period and this was
attributed to longitudinal splitting. The width of these fibers reportedly decreased with
time.
Bellmann et al. also report that a more detailed examination of the time dependence of
the width of chrystoile fibers indicates a rapid increase in the number of thin fibrils (<
0.05 pm in width) and thin bundles (< 0.1 pm in width) within 100 days (at the expense
of thicker bundles). The authors suggest that this would result in rapid decrease in the
number of chrysotile structures visible by optical microscopy and, possibly, increased
clearance of the thinnest fibrils by dissolution, but this study shows no increased rate of
clearance for thinner chrysotile structures compared to thicker structures (when viewed
by electron microscopy). In contrast, long chrysotile fibers that were acid-leached prior
to instillation reportedly disappeared with a half life of 2 days.
Generally, the rate of clearance of the long fractions of the other fibers reported in the
Bellmann et al. papers varies as a function of solubility and overall thickness.
Importantly, all half lives are reported to have high standard errors in this study, due to
the small number of animals included for examination.
In summation, virtually all short-term retention studies indicate that:
• fibers retained in the lung tend to be shorter and thinner than the aerosols
from which they derive and the size distributions of retained structures
tend to be more similar overall than the size distributions observed in the
original aerosols;
• chrysotile asbestos undergoes rapid, longitudinal splitting in the lung while
amphiboles do not;
• by mass, chrysotile and amphibole asbestos are deposited in the lung
with comparable efficiencies, although it is not clear whether chrysotile
dusts tend to contain sufficient numbers of curly fibers to limit deposition
in the deep lung;
• multiple clearance processes operate over different time frames and some
of these processes are strongly length-dependent. Fibers shorter than
7.33
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
approximately 10 pm appear to be cleared rapidly relative to longer fibers
and those longer than approximately 20 pm are not cleared efficiently at
all (if the fibers are insoluble). The Bellmann et al. studies appear to
contrast with other studies in this regard in that they suggest fibers longer
than 5 pm do not readily clear;
• the quickest clearance process (presumably muco-ciliary clearance) is not
dependent on length; and
• the effects of fiber diameter on clearance have not been well delineated
overall, although fibers that reach the deep lung appear to be largely
limited to those thinner than approximately 0.7 pm.
These findings are in addition to those mentioned previously from the newer studies:
• multiple clearance mechanisms (operating over multiple time scales)
contribute to clearance;
• for sufficiently soluble fibers, long fibers clear more rapidly than short
fibers;
• for insoluble fibers, a subset of long fibers clears rapidly while the
remaining long fibers clear only extremely slowly, if at all;
• short fibers of all types are cleared at approximately the same rate (much
more rapidly than long, insoluble fibers);
• a small fraction of short fibers may be retained for long periods under
certain circumstances (sequestered in alveolar macrophages) despite
overall rapid clearance of these structures; and
• there is some suggestion that short asbestos fibers clear somewhat more
slowly than short fibers of the other, non-asbestos mineral types studied.
Regarding specifically the clearance of long fibers, it appears that a component of all
such fibers clears rapidly within the first two weeks and this likely represents muco-
ciliary clearance. A second component (representing as much as 60% of the fibers)
clears within 90 days and this likely represents clearance by alveolar macrophages.
The remaining long fibers are cleared only very slowly, if at all, and this likely represents
fibers that are sequestered in granulomas or that escape into the interstitium.
7.34
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
7.2.1.2 Studies involving chronic or sub-chronic exposures
Although the results of older retention studies following longer term (sub-chronic or
chronic) exposure were difficult to reconcile with the results following shorter-term
exposures, newer studies suggest greater consistency and a clearer picture of the fate
of fibers in the lung. Moreover, although there are further suggestions of mineralogy-
(fiber type) dependent effects with some clearance mechanisms, it is important that size
effects be considered simultaneously, if the dynamics of these processes are to be
understood.
In some of the latest studies, for example, Heterberg et al. (1993,1995, and 1998b)
exposed rats (nose-only) by inhalation to a series of manmade vitreous fibers (including
a variety of fibrous glasses, rock wools, and refractory ceramic fibers) and two kinds of
asbestos: chrysotile (intermediate length NIEHS fiber) and crocidolite (size selected).
Animals were dosed for 6 hrs/day, 5 days/wk for up to two years at target
concentrations of 10 - 60 mg/m3. The target concentration for chrysotile and crocidolite
was 10 mg/m3. Animals were periodically sacrificed during the exposure regimen to
determine the character of the retained fibers. Vitreous fiber aerosols were
characterized by PCM, SEM or, for chrysotile, by TEM. The right accessory lung lobe
of sacrificed animals was tied off, frozen, and stored for lung burden analysis.
For analysis, lung lobes were dried to constant weight, ashed, the residue suspended in
distilled water, and then filtered on Millipore filters (for examination by optical
microscopy) or Nuclepore filters (for analysis by SEM or TEM for chrysotile).
Approximately 100 fibers were reportedly characterized to establish fiber size
distributions. However, this is problematic for this study because chrysotile asbestos
concentrations in the aerosols to which the animals were exposed contained
approximately 100 times as many fibers as the other aerosols. Thus, although no fibers
longer than 20 pm were observed during characterization of the chrysotile, the
concentration of such long fibers could still have been larger in this aerosol than the
other aerosols and it would not necessarily have been observed. This is also true of
lung burden analyses especially because indirect preparation tends to magnify the
number of short chrysotile structures observed in a sample.
A comparison of the retention patterns of chrysotile and RCF-1 from the Hesterberg et
al. studies is particularly instructive. First, it should be noted that, in contrast to the
values reported by the authors of this study, chrysotile and RCF-1 in fact appear to
exhibit comparable in-vitro dissolution rates (12.7 vs. 8 ng/cm2-hr, respectively) when
rates are measured using comparable techniques (see discussion in Section 7.2.4).
7.35
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
The dissolution rates quoted in the Hesterberg study are not derived in comparable
studies.
Although a full set of time-dependent analyses are not apparently available for
chrysotile, it is reported that approximately 14% of those chrysotile WHO fibers
observed to be retained after 104 days of exposure continue to be retained after 23
days of recovery. Under the same conditions, it is reported that 43% of RCF-1 fibers
are retained, which suggests more rapid clearance for chrysotile. Even adjusted for the
fraction of RCF-1 structures that are longer than 20 pm (and that are presumably
cleared even more slowly), approximately 37% of the RCF-1 WHO fibers (< 20 pm) are
apparently retained over this period, which is still more than twice the rate reported for
chrysotile. Still, more detailed characterization of the size distributions of these two
fiber types would need to be evaluated before it could be concluded with confidence
that chrysotile is cleared more rapidly than RCF-1 or that dissolution plays a role. In
fact, dissolution would tend to cause more rapid clearance only of the longest fibers
(i..e. the ones that cannot be cleared by macrophages - see Section 7.2.1.1), which
would further reduce the apparent retention rate of the shorter RCF fibers, making it
even more comparable to the chrysotile number.
Based on these studies, Hesterberg et al. report that fibers deposited and retained in
the lung tend to be shorter and thinner on average than the sizes found in the original
aerosol. It is also apparent from their data that long RCF-1 fibers clear more rapidly
than short RCF-1 fibers (although a small fraction of long structures are retained at all
time points following a recovery period after cessation of exposure), which is consistent
with observations in other studies for fibers that dissolve at reasonable rates. In the
short term study performed by the same laboratory, Hesterberg et al. 1998a, long fiber
(> 20 pm) RCF-1 a appears to clear at approximately the same rate as the shorter
structures (< 5 pm), although the scatter in the data (and an unexplained initial rise in
long-fiber RCF) prevent a more careful comparison. Similar results are also apparent in
the data presented for MMVF21. It should be noted that the dissolution rates for RCF-1
and MMVF21 bracket the estimated dissolution rate for chrysotile asbestos (when the
three are derived from comparable studies, see Section 7.2.4).
The data presented in Table 3 of Hesterberg et al. (1998b), which are reproduced in the
following Table 7-3, can also be used to evaluate the time-trend of retention during
chronic exposure.
Table 7-3:
Fraction of Fibers Retained Following Chronic Exposure1
7.36
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Chrysotile
RCF-1
Exposure
Lung/Aerosol
Lung/Aerosol
Long
Lung/Aerosol
Period
WHO Fibers
Ratio
WHO Fibers
Ratio
WHO fibers
Ratio
(wks)
f/lunq x 10A6
f/lung x 10A6
f/lung x 10A6
0.0357
0.009
4.81 E-05
0.002
1.98E-05
13
250
0.024
39
0.209
3
0.030
26
180
0.017
56
0.299
6
0.059
52
1020
0.096
119
0.636
20
0.198
78
853
0.080
173
0.925
21
0.208
104
1600
0.151
143
0.765
25
0.248
Aerosol Concentration:
(f/ml)
10600
187
101
1. Source: Hesterberg et al. (1998b)
The values presented in Columns 2, 4, and 6 of the above Table present, respectively,
measurements of the lung burden for chrysotile WHO fibers, RCF-1 WHO fibers, and
RCF-1 long WHO fibers (> 20 (Jm) in animals sacrificed immediately following cessation
of exposure for the time period indicated in Column 1. Unlike results reported in some
earlier chronic studies based on mass (see below), there is no evidence from this table
(based on fiber number) that chrysotile lung burdens reach a plateau. Rather chrysotile
lung burdens (as well as RCF-1 lung burdens) continue to increase with increasing
exposure.
The data presented in Table 7-3 can also be used to gauge the relative efficiency with
which the chrysotile and RCF fibers are retained. Considering that the number of fibers
inhaled (Ninh) over the period of exposure would be equal to the product of the aerosol
concentration (C^ in f/cm3), the breathing rate of the exposed animal (Rg in cm3/wk),
and time (in weeks):
N^CVRe't 7.1
and the efficiency of retention is simply equal to the quotient of the number of fibers
retained (N,ung) and the total number inhaled: N^N^, then the efficiency of retention is
estimated by the following simple relationship:
Efficiency of retention = N)unfl/(C./RB*t) (7.2)
By rearranaging Equation 7.2, one obtains:
7.37
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Efficiency of retentions = (N(uns/(Cair)*(1/Re) (7.3)
Because the breathing rate for the rats in the Hesterberg et al. study can be considered
a constant for all experiments, Equation 7.3 indicates that the slope of a plot of
versus time should yield estimates of the relative efficiency of retention for each of the
fiber types evaluated. The plot for chrysotile is presented in Figure 7-3. Results from
this plot and similar plots for RCF-1 WHO fibers and long WHO fibers (data not shown),
result in the following estimates of the relative efficiencies of retention (along with the
corresponding Rz value for the fit of the linear trend line):
chrysotile WHO fibers: 0.0014, R2 = 0.856
RCF-1 WHO fibers: 0.0095, R2 = 0.694
RCF-1 long, WHO fibers 0.0026, R2 = 0.880
Thus, it appears that chrysotile WHO fibers are retained somewhat less efficiently than
either RCF-1 WHO fibers or RCF-1 long WHO fibers. However, whether this is due to
less efficient deposition or more efficient clearance cannot be determined from this
analysis. It is also not possible to determine whether such differences are due to the
effects of differences in size distributions among the various fiber types. Interestingly,
based on the data presented by Hesterberg et al., which indicates that long RCF-1
WHO fibers clear more rapidly than regular RCF-1 WHO fibers, the differences in the
relative retention of these two length categories of fibers is due primarily to relative
efficiency of clearance.
7.38
Revision 1 • 8/3/01
-------
o
10
g
<5
= o"
® c
^ o
3 O
CQ
o>
c
3
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
Figure 7.3:
[Chrysotile Lung Burden/Aerosol Concentration]
vs. Time of Exposure
0
~
~
y = 0.0014x
R2 = 0.8556
~
I I I I 1
20
40 60 80
Time of Exposure (weeks)
100
120
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Figure 7-3
7.39
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
In a similar study involving chronic exposure to Syrian golden hamsters (Hesterberg et
al. 1997), fiber retention and biological effects associated with exposure to amosite and
a series of MMVFs were evaluated. The amosite was size selected and hamsters were
exposed to one of three levels (0.8+/-0.2, 3.7+/-0.6, and 7.3+/-1.0 mg/m3). Amosite
lung burdens were shown to increase regularly with dose and time of exposure. The
time dependence for accumulation of some of the MMVF's was more complicated.
None of the animals were apparently followed for any recovery periods following
cessation of exposure. The authors also indicate that the severity of the effects
observed (inflammation, cellular proliferation, fibrosis, and eventually several
mesotheliomas), appear to correlate well with the concentration of fibers longer than
20 fjm.
Earlier Studies
Earlier studies, in which asbestos concentrations tend to be monitored as total mass
tend commonly to show that chrysotile asbestos is neither deposited as efficiently as
various amphibole asbestos types nor is it retained as long (i.e. it is cleared much more
rapidly from the lungs). In fact, several such studies tend to show that chrysotile
asbestos concentrations eventually reach a plateau despite continuing exposure, which
suggests that clearance and deposition come into balance and a steady state is
reached. In contrast, amphibole asbestos concentrations continue to rise with
increasing exposure, even at the lowest exposure levels at which experimental animals
have been dosed.
Such observations do not appear to be entirely consistent with those reported in newer
studies (see above) that track fiber number concentrations (in specific size categories).
In these newer studies, chrysotile retention is not observed to level off, but continues to
increase in a manner paralleling amosite or other fibers. As indicated below, however,
the limitations associated with these older studies suggest that, although it may not be
easy to reconcile them quantitatively with the newer studies, results from these studies
are not necessarily inconsistent with those of the newer studies. Moreover, the trends
observed in the newer studies are likely more directly relevant to issues associated with
the induction of asbestos related disease.
The problems with the older studies are that:
• the trends seen in the older studies (based on mass) may mask the more
important trends associated with deposition and retention of long, thin
fibers. Thus, results from such studies may not be directly relevant to
considerations of risk; and
7.40
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
• the observed differences between chrysotile and the amphiboles may be
attributed to differences in size distribution (among other possibilities).
Thus, lacking detailed information on size distributions, it is difficult to
reconcile the results from the older studies with results from the newer
studies, which explicitly track specific size ranges of fibers.
Given these limitations, the earlier studies are only mentioned briefly.
Middleton et al. (1979) tracked the fate of asbestos (as mass measured by infrared
spectroscopy) in rats following inhalation of several asbestos aerosols (UICC chrysotile
A, UICC amosite, and UICC crocidolite) at multiple concentrations (reported at 1, 5 or
10 mg/m3). To account for possible differences in the nocturnal (vs. daytime) activity
level of rats, several groups of rats were also exposed in a "reversed daylight" regimen
(in which cages were darkened during the real day and bathed in light during the real
night). Acclimatized rats in these groups were thus dosed at times corresponding to
their night. Exposure continued for 7 hrs per day, 5 days per week, for six weeks.
Results from the Middleton et al. study were fit to a three compartment model (originally
proposed by Morgan et al. 1977) and the authors concluded that clearance was
independent of fiber type, but that the initial deposition of fibers was very dependent on
fiber type. This was indicated by a "K-factor" representing the efficiency of initial
deposition. Chrysotile showed K factors that range between 0.17 and 0.36 and vary
inversely with the initial exposure concentration. In contrast, amosite exhibits a K factor
of 0.69 and crocidolite a K factor of 1.0 and both are independent of exposure level.
Although the design of this experiment precluded fitting of the shortest two
compartments of the model (with half lives of 0.33 and 8 days, from Morgan et al.
1977), they did optimize the half life of the longest compartment. Fibers in this
compartment were cleared with a half life of 170 days.
In a series of studies, Davis and coworkers (1978,1980,1988a, and 1988b) report that
retention of asbestos (measured in terms of mass) appears to be a function of fiber type
and surface charge in addition to fiber size. With regard to fiber type, for example,
Davis et al (1978) report that substantially more amphibole (amosite) asbestos appears
to be deposited and retained in the lungs of exposed rats than chrysotile. Chrysotile is
also apparently cleared more readily than amosite. However, mineralogical effects
should only be judged after adjusting for fiber size.
Rats in the Davis et al. studies were dosed at 6hrs/day, 5 days/wk for up to one year at
dust concentrations of 2, 5 or 10 mg/m3 (depending on the specific experiment). Right
7.41
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
lungs (used for determining lung burden) were ashed and the residue was washed in
distilled water and filtered. The residue was formed into a potassium bromide disc and
asbestos (mass) content was determined by infrared spectroscopy.
Jones et al. (1988) report that the lung-tissue concentration of amosite increases
continually with exposure (at 7 hrs/day, 5 days/wk for up to 18 months) and the rate of
increase is proportional to the level of exposure. A leveling off of amphibole
concentrations in lung-tissue was not observed in this study as long as exposure
continued, even for the lowest level of exposure (0.1 mg/m3) studied. The lowest
exposure concentration evaluated in this study is only 1 % of the concentration at which
chrysotile lung burdens were shown to reach equilibrium in other retention studies (see
below). Importantly, however, these are only the older studies in which fiber burden is
tracked by mass. The newer studies don't show this effect.
The authors also report lack of any apparent change in size distribution with time
among the fibers recovered from the animal's lungs, which suggests lack of substantial
clearance even of short fibers. However, the longest recovery period following the
cessation of exposure evaluated in this study is only 38 days, which may be too short to
allow evidence for differential clearance as a function of size to become apparent (at
least in a chronic study; the time dependence in chronic studies such as this are more
complicated than for short-term studies). Moreover, the apparent inclusion of lymph
nodes as part of the lung homogenate may have caused short fibers initially cleared
from the lung to be added back in. In this study, lungs were recovered intact including
the associated mediastinal and hilar lymph nodes, which were ashed in toto. Ash
residue was washed in acid and water, ultrasonicated and filtered for electron
microscope analysis. Note that such a procedure would include any fibers cleared to
local lymph nodes.
In a widely cited study, Wagner et al. (1974), report that amphibole lung burdens
increase continually as long as exposure to amphiboles continues and that amphibole
concentrations in lung tissue decrease only slowly following cessation of exposure. In
contrast, chrysotile lung burdens reach a plateau despite continued exposure
Importantly, asbestos content was estimated by determining total lung silica content
and adjusting for similar analysis on filtered samples of the original aerosols. Thus, in
addition to suffering from the limitations associated with tracking fiber burden by mass,
there are questions concerning the validity of using total silica to represent asbestos
content. Therefore, for these reasons and the additional reason of the lack of
controlling for fiber size, the ability to interpret this study and reconcile its conclusions
with those of newer studies is severely limited.
7.42
Revision 1 - 6/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Chronic Inhalation of Non-fibrous Particulate Matter
A recent study involving chronic inhalation of non-fibrous materials is helpful at
elucidating the relative localization of particles in rats and primates. Nikula et al. (1997)
studied lung tissue from a two-year bioassay originally conducted by Lewis et al.
(1989), in which cyanomolgus monkeys and F344 rats were exposed to filtered,
ambient air or air containing one of three particulate materials: diesel exhaust
(2 mg/m3), coal dust (2 mg/m3, particles < 7 um in diameter), or a 50-50 mix of diesel
exhaust and coal dust (combined concentration: 2 mg/m3).
Results from Nikula et al. (1997) indicate that responses to all three particulate
materials were similar. The particles tended to localize in different compartments of the
lung in a species-specific manner:
• 73% of particles remain in the alveolar lumen of rats but only 43% in
monkeys. The remainder can be found in the interstitium;
• in both the alveolar lumen and in the interstitium, virtually all of the
particles are observed to be isolated within macrophages; and
• the particles in the interstitium reside in macrophages within the alveolar
septa, the interstitium of respiratory bronchioles, the adventitia and
lymphatic capillaries surrounding arterioles and veins of pulmonary
parenchyma, or in the pleura.
It is not known whether free particles penetrate the epithelial lining of the airway lumena
and escape into the interstitium or whether such particles are first engulfed by
macrophages and then transported in their macrophage "hosts" into the interstitium.
Importantly, even after two years of exposure, the particles in the interstitium do not
appear to have elicited a tissue response. Also, the aggregates of particle-laden
macrophages observed in alveolar lumena elicited significantly less of a tissue
response im monkeys than in rats. Such responses included: alveolar epithelial
hyperplasia, inflammation, and focal septal fibrosis.
The authors further indicate that "epithelial hyperplasia concomitant with aggregation of
particle-laden macrophages in alveolar lumen is a characteristic response to. many
poorly soluble particles in the rat lung, both at exposure concentrations that result in
lung tumors and at concentrations below those resulting in tumors. Such a response,
however, was not characteristic of what was observed in monkeys. Among other
things, these differences in responses suggest that rats may not represent a good
7.43
Revision 1 • 0/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
model for human responses to inhalation of poorly soluble particulate matter. It would
also have been interesting had they tested a "benign" dust such as TiOz.
In summation:
• results of (newer) sub-chronic and chronic retention studies are generally
consistent with those of retention studies that track lung burden following
short-term exposure (Section 7.2.1.1);
• there is some indication in these sub-chronic and chronic studies that
chrysotile asbestos may not be retained as efficiently as amphibole
asbestos. It is likely, however, that such distinctions are due more to fiber
size than fiber type so that definitive conclusions concerning such effects
cannot be reached until better studies that properly account for both size
and type are conducted;
• although earlier studies that track mass instead of fiber number suggest
otherwise, chrysotile and amphibole asbestos concentrations (when
measured by fiber number) continue to increase with time as long as
exposure continues. Due to a lack in the ability to distinguish among size-
dependent effects when lung burdens are tracked by mass, the results of
the earlier studies are not necessarily inconsistent with the results of the
later studies;
7.2.2 Animal Histopathological Studies
Studies in which the lungs of dosed animals are examined to determine the fate and
effects of inhaled asbestos are helpful for understanding the movement and distribution
of retained particles within the lung and surrounding tissue. Both the newest retention
studies and the older retention studies tend to include at least some of this type of
information. They also tend to indicate a consistent picture of the fate and effects of
asbestos. While such studies tend to confirm that translocation in fact occurs, they are
less helpful for elucidating the specific mechanisms by which translocation occurs.
Newer Studies
llgren and Chatfield (1998) studied the biopersistence of three types of chrysotile
("short chrysotile" from Coalinga in California, "long* Jeffrey fiber from the Jeffrey mine
in Asbestos, Quebec, and UICC - B (??) Canadian Chrysotile, a blend from several
mines in Quebec). Both the Coalinga fiber and the Jeffrey fiber were subjected to
7.44
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
further milling prior to use. In this study, rats were exposed via inhalation for 7 hrs/day,
5 days/wk for up to two years. Concentrations were: 7.78 +/-1.46 mg/m3 for Coalinga
fiber, 11.36 +/- 2.18 mg/m3 for Jeffrey fiber, and 10.99 +/- 2.11 mg/m3 for the UICC-B
fiber. An additional group of rats was also dosed for a single 24-hour period with
Jeffrey fiber at a concentration of 5,000 f/ml > 5 um. Estimates of lung content of
chrysotile were based on measurements of total silica content. The character of the
three chrysotile types evaluated in this study was previously reported (Campbell et al.
1980, Pinkerton et al. 1983). The animal studies were conducted previously with the
overall approach reported by McConnell et al. (1983a and b ) and Pinkerton et al.
(1984). Based on the characterization presented in these papers, Coalinga fiber is
short, but not as extremely short as suggested by the authors:
Such calculations also suggest that the single, "high" dose in this experiment was
equivalent only to a concentration that is approximately 10 times the other
concentrations studied, so that it is equivalent only to a 10-day exposure and small
relative to the longer term (up to 2 yrs) exposures considered in this study.
Results reported by llgren and Chatfield indicate that the lung burden for short fiber
chrysotile initially increases with exposure, reaches a steady state, and then decreases
steadily following cessation of exposure. Approximately 95% of this material is cleared
within 2 years. Thus, short fibers appear to exhibit the trend suggested by the older
chronic retention studies that tracked burden by mass and is consistent with the newer
studies indicating rapid clearance of short structures (Section 7.2.1.2).
In the studies reported by llgren and Chatfield, most short fibers are found initially within
alveolar macrophages and, while concentrations in Type I epithelium, interstitial cells,
and interstitial matrix increase with time, little appears to be taken up by Type II
epithelium. Small amounts of short fiber chrysotile are also observed to be taken up by
endothelial cells. There is also little sign of inflammation or fibrosis following exposure
to the Calidria chrysotile.
Jeffrey chrysotile also initially appears to be taken up primarily by alveolar
macrophages but later becomes most prevalent in the interstitial matrix and, to a lesser
extent, in interstitial cells. Substantial numbers of fibers are also taken up by Type I
epithelium and small amounts by endothelial cells. Similar, but slightly delayed effects
Ratio of fibers longer than: 5 (Jm: 10 pm: 20 pm
Coalinga fiber:
Jeffrey fiber:
200: 78: 0.98
591: 220: 78
7.45
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
were seen for UICC-B chrysotile (which has a smaller fraction of long fibers and
therefore, the authors suggest, takes longer to accumulate). Note that, by 12 months,
the majority of long fibers from both these types were found in interstitial matrix while
the majority of Calidria material was still found in alveolar macrophages This is
consistent with observations concerning behavior between short and long fibers
reported by Wright and Kushner (1975), Section 7.2.1.1, At this point, the Jeffrey
material also caused substantial thickening of the basement membrane and most fibers
in the interstitium were trapped within the collagenous matrix. The authors note that
some of the most severe interstitial changes occurred adjacent to areas of bronchiolar
metaplasia. Such effects were not seen with Calidria exposure.
Thickened basement membranes, calcium deposits, metaplastic changes, and
structural abnormalities were all observed with long fiber exposure but not with calidria
exposure. Interstitial macrophages also showed morphological changes following
phagocytosis of fibers. While calidria material was about evenly distributed between
interstitial matrix and cells, the vast majority of long fiber material was found in the
matrix. Movement into the matrix was also observed to increase even after exposure
ceased. With time, the number of long fibers in interstitial cells declined modestly but
declined precipitously for Calidria material
Jeffrey fibers accumulated in Type I epithelial cells during exposure and then levels
decreased slowly after exposure ceased. UICC-B fibers accumulated more slowly,
never reaching the same levels as for Jeffrey and decreased more rapidly.
Concentrations of Calidria fibers in Type I epithelial cells was low at all time points.
Type II epithelial cells accumulated very few fibers of any type (although they took up
slightly more Jeffrey fiber than the others). All three fiber types caused substantial
increases in interstitial cells (mostly macrophages) at three months and this increase
persisted for the Jeffrey fiber but decreased to background after 24 months for the other
two fiber types. Fibroblast numbers also increased with the long fiber types but not
Calidria chrysotile.
Type II epithelial cells showed decreases in volume and number that persisted until
exposure to long fiber ceased and these cells displayed dramatic structural aberrations
despite absence of a fiber load. One possible explanation for the observed changes In
Type II cells, especially the reduction in their number, is that they were undergoing
terminal differentiation to Type I cells (see Section 4.4) In fact, the apparent absence of
fibers observed within Type II cells might be explained by such cells taking up fibers but
being induced to terminally differentiate, once fibers are accumulated. Overall, Type II
cells displayed greater cellular response than Type I cells (which might also suggests a
7.46
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
role for cytokines). All effects were observed to be fiber length dependent and all were
exaggerated following exposure to long fiber material.
Rats exposed to long fiber had numerous accumulations of dust-laden interstitial
macrophages and/or small focal accumulations of dust within the interstitium at the end
of the lifetime study, but such changes were not observed for Calidria exposed animals
The lung burden for rats exposed to the single, "high" Jeffrey fiber exposure (based on
total silica) at 12 months (i.e. 12 months post exposure) was not different from controls.
Therefore, the authors conclude that short-term, "high" exposures are rapidly cleared
(even exposures containing substantial quantities of long fibers). Other changes
induced by the single, short-term high exposure of Jeffrey fiber that was followed for
24 months also showed reversion to close to background status. Importantly, these
observations are not based on quantitation of fiber burden in lung tissue. Rather they
are inferred by observing the effects caused by the presence of fibers. This may
suggest, for example, that more than 10 day's worth of exposure would be required at
this level of exposure before irreversible lesions develop.
In the study by Hesterberg et al. (1997), which was previously discussed (Section
7.2.1.2), the authors note (among dosed hamsters) that the magnitude of cellular
effects appeared to differ among the fibrous glasses as a function of their relative
biodurability. For animals dosed with the least biopersistent glass, only transient effects
(influx of macrophages, development of microgranulomas) were observed and these
did not progress further. For the more persistent glasses, injury progressed through
more intense inflammation, interstitial fibrosis, pleural collagen deposition, mesothelial
hypertrophy and hyperplasia and eventually, mesothelioma. The authors also suggest
that amosite appears to be more potent than chrysotile, even when aerosols contain
comparable numbers of fibers longer than 20 pm.
Choe et al. (1997) exposed rats to chrysotile and crocidolite (both NIEHS samples) by
inhalation for 6 h/d, 5 d/wk, for 2 weeks. The rats were then sacrificed and their pleural
cavities lavaged. Results indicate that significantly more pleural macrophages were
recovered in plueral lavage fluid at one and six weeks following exposure than sham
exposed rats. The centrifuged pellet from pleural lavage fluid from one of four rats also
exhibited long (> 8 pm), thin (< 0.5 pm) crocidolite fibers (one week following
exposure). The concentration of fibers in this pellet suggested approximately 1 f per
4000 cells in the pleura. Note that chrysotile rats were not examined for fiber content.
Older Studies
7.47
Revision 1.0/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
In the series of studies by Davis et al. (1978,1980,1985, and 1986a), the authors
generally report similar histopathological observations that emphasizes a marked
distinction between effects from long and short fibers. From the 1986 study, for
example, Davis et al. report that at the end of 12 months of exposure, rats exposed to
long fibers (amosite in this case) exhibited deposits of granulation tissue around
terminal and respiratory bronchioles. They further indicate that the granulation tissue
consists primarily of macrophages and fibroblasts with occasional foreign body giant
cells.
As the animals aged, there was increased evidence of collagen deposits in these
lesions and the oldest lesions consist mainly of acellular, fibrous tissue. The alveolar
septa in these older animals showed progressive thickening. Initially, this was
apparently due primarily to hyperplasia of Type II epithelial cells but with time was
increasingly due, first, to reticulin and, later, to collagenous deposits in the septal walls.
Asbestos dust was frequently visible in these deposits. Epithelial cells lining alveoli
adjacent to the oldest lesions also tended to become cuboidal in shape. As the animals
aged, these areas of interstitial fibrosis became more extensive.
In contrast, animals exposed to short fibers (also amosite in this case) showed no such
lesions (peribronchial fibrosis) at any point in time. At the end of exposure, the lungs of
these animals contained large numbers of pulmonary macrophages packed with fibers,
but these cells remained free in the avleolar spaces. The authors report that large
numbers of laden macrophages sometimes aggregated in alveoli close to respiratory
bronchioles, but that there would be no formation of graunulation tissue or thickening of
alveolar septa at these locations. Thus, with the exception of the presence of dust-
laden macrophages, the structure of the lung of these rats was not altered.
In the Davis et al. 1987 study of chrysotile, a slight variation of the above scenario is
worth noting. In this study, the development of peribronchial fibrosis was reported for
animals dosed both with the long fiber material and with the short fiber material.
However, the authors also report that the short fiber chrysotile in this study in fact
contains a sizable fraction of longer structures and this finding was corroborated by
more formal size characterization (Berman et al. unpublished) later conducted in
support of a study to evaluate the effects of size (Berman et al. 1995).
In this study, Davis et al. also reported observations on the morphological changes
observed in the mesothelium during these studies. The authors indicate that the older
animals in these studies exhibit "areas of vesicular pleural metaplasia consisting of
loose, fibrous tissue containing large vesicular spaces lined with flattened cells." The
7.48
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
authors also report that examination in previous studies indicates that these cells are of
a mesothelial type.
Davis et ai. report that, "...occasionally the walls between vesicular spaces were so thin
that they consisted of two closely opposed layers of extended and flattened cells with
no basement membrane in between them. Where cells were supported by areas of
fibrous tissue, a basement membrane was present. While no method for the direct
quantification of this pleural metaplasia has been developed, its occurrence is closely
related to the presence of advanced interstitial fibrosis or adenomatosis in the lung
tissue and it is particularly common where patches of this type of parenchymal lesion
have reached the surface. It is not known whether such lesions are precursors to
mesothelioma." Davis et al. also note that neither of the two mesotheliomas observed
in this study showed histological patterns consistent with the observed vesicular
hyperplasia.
Brody et al (1981) tracked the distribution of chrysotile following inhalation by rats.
Asbestos was initially deposited almost exclusively at alveolar duct bifurcations. In
agreement with Pinkerton et al (1986), the degree of deposition appeared to be an
inverse function of the pathlength and bifurcation number for each alveolar duct.
Uptake by macrophages and type 1 epithelial cells were observed following deposition.
Asbestos was observed both in lipid vesicles and free in the cytoplasm of type 1 cells.
After 8 days, alveolar duct bifurcations became thickened with an influx of
macrophages. Asbestos was also observed in basement membrane below the
epithelium. Apparently, structures had been transported through type 1 cells to the
basement membrane. Once in the basement membrane, asbestos may enter the
interstitium. Predominantly short structures were monitored in this study. Long
structures were not readily observed (but this is likely a counting problem; under such
circumstances, short structures may serve as surrogates for the presence of other
structures).
Intratracheal instillation
Bignon et al (1979) studied the rate of translocation of various materials in rats.
Chrysotile, crocidolite, and glass fibers were intrapleural^ injected into rats and their
concentration was monitored as a function of time in lung parenchyma and other
tissues removed from the pleura. Within one day following injection, asbestos was
detectable in lung parenchyma. After 90 days, asbestos was found in all of the tissues
analyzed. Based on the rate of translocation to the lung, crocidolite migrates about 10
times more rapidly than chrysotile (on a mass basis). The rate of migration of glass is
in between the two asbestos types. Structures initially found in the lung were
7.49
Revision 1 -9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
significantly shorter than the average size of structures injected. After seven months,
however, the average lengths of structures in all tissues monitored were longer than the
average length of structures originally injected. Thus, short structures migrate more
rapidly than longer structures (possibly by a different mechanism) but long structures
eventually translocate as well. Within a target tissue, preferential clearance of short
structures also contributes to observed increases in the average length of the structures
with time.
Studies of Non-Fibrous Particulate Matter
Studies of the fate and effects of respirable, non-fibrous particulate matter provide
evidence for at least one mechanism by which particles (and fibers) may be transported
to the interstitium.
Li et al. (1997) evaluated the effect of urban PM 10, carbon black, and ultrafine carbon
black on rats following intratracheal instillation (0.2 ml volume instilled containing
between 50 and 125 pg of particles). After 6 hrs, there was a noted influx of
neutrophils (up to 15% of total cells observed in bronchioalveolar lavage - BAL fluid)
and increases in epithelial permeability was surmised based on increased total protein
(including increases in levels of lactate dehydrogenase, which is a marker for cell
membrane damage) in BAL fluid.
Conclusions
Overall, observations among both the newer and the older studies (including the study
by Wright and Kuschner 1975, see Section 7.2.1.2) tend to be highly consistent,
particularly with regard to the distinction between the effects of short and of long fibers.
Typically, long fibers initially produce substantial inflammation characterized by an influx
of macrophages and other inflammatory cells. Ultimately, exposure to such fibers
cause thickening of alveolar septa (particularly near avleolar duct bifurcations) due to a
combination of epithelial hyperplasia and deposition of reticulin and, later, collagen
resulting in interstitial fibrosis. In contrast, short fibers cause an initial influx of
macrophages and, long term, show persistent accumulations of fiber-laden
macrophages both in alveolar lumena and in pulmonary lymph nodes, but otherwise no
structural changes are observed in the lung tissue of these animals.
These studies also provide ready evidence of the effects of fiber translocation, but
generally offer only limited evidence for elucidating the mechanisms by which such
translocation occurs. There is evidence that Type I (and possibly Type II) epithelium
phagocytize particles and fibers and it is possible that such fibers may be passed
7.50
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
through to the basement membrane and the interstitium. For Type I cells, the distance
between the alveolar lumen and the basement membrane averages less than 1 pm in
any case (Section 4.4). Certainly fibers are also observed in the interstitium.
Particulate studies also indicate that oxidative stress induced by particulate matter and
fibers may cause morphological changes in Type II cells with consequent loss of
integrity of the epithelium, which increases it's permeability overall and may also allow
diffusional passage of particules and fibers.
7.2.3 Human Pathology Studies
Human pathology studies provide additional information concerning the nature of
asbestos deposition, clearance, and retention. These are the studies in which lung
burdens are measured in samples of lung tissue and correlated with the exposures
received by the individuals from which the lung samples derive.
Among the advantages of human pathology studies is that they provide direct insight
into the behavior of asbestos in humans. They are also limited, however, by the lack of
ability to obtain time-dependent estimates of lung burden (because samples are derived
from deceased individuals), by the manner in which lung tissue is stored (several of the
fixatives employed to store tissue samples have been shown to enhance dissolution of
asbestos, Law et al. 1990 and 1991), by the manner in which samples are prepared for
asbestos analysis, by the manner in which asbestos is analyzed, and by the limited
ability to re-construct the uncontrolled exposures experienced by study subjects
(Section 5.2).
Perhaps most importantly, the ability to construct anything but the coarsest quantitative
comparisons across subjects is also typically limited by use of "opportunistic" tissue
samples (i.e. use of samples that happened to have been collected and stored during
autopsy or necropsy) because such samples are not controlled for location on the
respiratory tree (i.e. the linear distance and branch number from the trachea) that is
represented by the sample. Because it has previously been shown that deposition is a
strong function of such location (see, for example, Yu et al. 1991), comparisons across
lung samples not controlled for these variables are problematic. In fact, it has also
been shown that samples collected from adjacent locations in lung parenchyma can in
fact exhibit strikingly different fiber concentrations due specifically to thepdifferences in
the location of the respiratory tree represented by the alveoli and respiratory
bronchioles in the spatially adjacent samples (Pinkerton et al. 1986, Brody et al. 1981,
Section 5.2).
7.51
RtvWon 1-80/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Despite the above indicated cautions, when interpreted carefully, human pathology
studies can provide useful evidence regarding fiber deposition, clearance, and retention
in the human lung.
Mpwpr Studies
Among the most recent studies, Finkelstein and Dufresne (1999) evaluated trends in
the relationship between lung burdens for different fiber types and different size ranges
as a function of historical exposure, the duration of such exposure, the time since last
exposure, and other variables. The analyses were performed among 72 cases from
which tissue samples could be obtained (including 36 asbestosis cases, 25 lung cancer
with asbestosis cases, and 11 mesothelioma cases).
Due to the excessive scatter in the data, most of the analyses presented depend on
"Lowess Scatterplot Smoothers." NEED TO COMMENT ON THIS Moreover, although
not stated, it is likely that the tissue samples obtained were "opportunistic" In that they
were not matched or controlled for relative position in the respiratory tree.
The authors employed the multi-compartment model developed by Vincent et al. (1985)
to evaluate trends in their data. The features of this model include:
• a compartment representing conducting airways that are cleared within
minutes to hours by muco-ciliary transport;
• a compartment representing the subset of fibers reaching the pulmonary
portion of the lung that are cleared by alveolar macrophages and
transported to the muco-ciliary escalator. This type of clearance is also
considered relatively rapid with half lives of no more than several days to
several weeks. Macrophage clearance is also considered size-dependent
and long fibers are cleared less efficiently than short fibers;
• when sufficient dust is inhaled (or dust is sufficiently cytotoxic) to impair
the motility of macrophages (either by volumetric overload or by toxicity), a
sequestration compartment forms that consists of laden but immobile
macrophages. Although this compartment may ultimately be cleared to
lymphatic drainage, such clearance is assumed to be slow and size
dependent (with half lives of 2 or 3 years for short fibers and 8 years for
fibers longer than 10 pm); and
7.52
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
• once the macrophage system is overloaded, fibers may cross the alveolar
epithelium and reach the interstitium and this compartment must be
cleared by transport to lymphatic drainage, which is assumed to be an
extremely slow process.
Finkelstein and Dufresne indicate that chrysotile splits both longitudinally and
transversely in the lung and that chrysotile lung burdens decrease significantly with time
since last exposure (with short fibers clearing even faster than long fibers) while
tremolite burdens do not appear to decrease with time since last exposure. They also
suggest that smoking does not appear to affect clearance rates.
Finkelstein and Dufresne indicate that these type of studies are not useful for examining
the behavior in rapidly clearing compartments of the lung but they may provide insight
concerning the more slowly clearing compartments. Based on their modeling, they
suggest that tremolite is transferred to the sequestration compartment at rates that are
6 to 20 times that of chrysotile (which they indicate is comparable to what was found for
crocidolite by DeKlerk). The authors suggest that retained chrysotile concentrations
tend to plateau after accumulation of about 35 years of exposure while tremolite
concentrations continue to increase. They also, suggest, however, that chrysotile
concentrations may begin to increase again after 40 years (suggesting the overload is
eventually reached for chrysotile as well). Reported half-lives from the long-term
compartment are:
chrysotile
fibers shorter than 5 Mm
fibers 5-10 pm jn length
fibers Jonger than 10 pm
tremolite
fibers shorter than 5 pm
fibers 5-10 pm in length
fibers longer than 10 pm
In a case-control study, Albin et al.
workers from the asbestos-cement plant previously studied for mortality (Albin et al.
1990). In this study, details of the procedures used to prepare lung tissue for analysis
were not provided. It is also assumed that available tissue samples were
"opportunistic" in that they were not matched or controlled for relative position in the
respiratory tree.
3.8 yrs
5.7 yrs
7.9 yrs
14.3 yrs (not different from °°)
15.8 yrs (not different from «)
150 yrs (not different from «)
(1994) examined the lung burdens of deceased
7.53
Revision 1 -8/3*01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Results from Albin et al. (1994) are consistent with, but do not necessarily demonstrate
that, chrysotiie is cleared more readily from a long-term sequestration compartment
than amphiboles. The authors also report that chrysotiie fibers observed in this study
are much shorter than the amphibole fibers observed, so that differences in clearance
rates might be attributable to size differences. The authors also suggest that clearance
is impaired by fibrosis.
Studies of Quebec Miners
Several authors also studied the lung content of various groups of deceased chrysotiie
miners in Quebec and found, despite overwhelming exposure to chrysotiie from the ore
(which contains only trace quantities of tremolite, see Section 6.2.3 and Sebastien et al.
1986), a substantial number of fibers (in some cases the majority of fibers) observed in
the lungs of deceased miners from this area are tremolite. Thus, for example:
• in a study of lung burdens in 6 mesothelioma victims, Churg et al (1984)
showed that amphiboles structures were 5 to 15 times as plentiful as
chrysotiie despite the predominantly chrysotiie exposure;
• in a study of lung tissue from 20 asbestosis cases, Pooley (1976) found
substantial concentrations of tremolite in the lungs of deceased Quebec
chrysotiie workers; and
• in a much larger study comparing lung burdens of Quebec workers with
those from the South Carolina textile mill, which has also been extensively
studied for asbestos-related mortality (Section 6.2.3), Sebastien et al.
(1989) examined 161 lung tissue samples (89 from the Quebec mines).
Results from this study indicate that geometric mean tremolite fiber
concentrations were more than three times mean chrysotiie fiber
concentrations (18.4 vs. 5.3 f/pg dry lung tissue) among the deceased
Quebec miners evaluated. It was also found that, despite these
differences, the overall size distributions of tremolite and chrysotiie fibers
observed in lung tissue were approximately the same. A more detailed
discussion of the results of this study is provided in Section 6.2.3.
• McDonald, Gibbs and Rowlands (1985) suggest that lung burden data
from quebec indicate little evidence of decreasing chrysotiie concentration
with time since last exposure. Rather they suggest simply that tremolite is
initially deposited in the deep lung more efficiently. These authors also
7.54
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
indicate that tremolite fibers are mostly optical while chrysotlie are mostly
"Stanton" or thinner.
These authors report good correlation of both tremolite and chrysotile with
estimated past exposures. Reported geometric mean measurements for
the various fibers in lung are: tremolite: 1x 10" -18.2 x 10®, chrysotile : 1.5
x 106 -15.7 x 10®, respectively, when exposure varied from < 30 mpcf to
>300 mpcf. Note that, if these are PCM measurements, this may not be
telling the whole story. The authors also report that 66% of those who
died 10 yrs since first exposure and half of those who died 30 yrs since
first exposure showed high chrysotile concentrations in their lungs.
These observations provide evidence that either amphibole (tremolite) asbestos is
deposited more efficiently in the compartments of the lung where clearance is slow or
chrysotile asbestos is cleared more rapidly and efficiently from even the slowest
clearing compartments of the lung (or both). Moreover, this conclusion appears to
apply similarly to both short and long fibers.
Some of the pathology studies that have been published suggest that at least some
clearance mechanisms show a dependence on fiber size, which is consistent with what
is observed in animal studies (Section 7.2.1). Notably, for example, Timbrell (1982)
studied deceased workers and relatives from the Paakkila anthophyllite mine in Finland.
He found that structures shorter than 4 pm and less than 0.6 pm in diameter are
completely cleared from healthy lungs. The efficiency of clearance decreases slowly
with increasing size. Structures longer than 17 pm and thicker than 0.8 pm in diameter
are not significantly cleared. The study is based on a comparison of structure size
distributions in lungs compared to the structure size of the material in the original dust
exposure. Timbrell also noted that asbestosis suppresses the removal process.
When considering the dependence of clearance on size (particularly via mechanisms
involving phagocytosis), it is necessary to address differences in human and animal
physiology. Due to differences in the morphology, for example, human macrophages
have been shown capable of phagocytizing larger particles and longer fibers than
macrophages found in mice and rats (Krombach et ai. 1996, lor details, see Section
4.4). Thus, the range of fibrous structures that are efficiently cleared from human Kings
is expected to include longer fibers than the range efficiently cleared in mice or rats.
Unfortunately, given the limited precision of the available data, the size ranges that are
reported to be cleared efficiently in rats and humans, respectively, cannot be easily
distinguished.
7.55
RevWon 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Several human pathology studies also support observations from animal studies
indicating that clearance may be inhibited by the development of fibrosis (Albln et al.
1994, Churg et al. 1990, Morgan and Holmes 1980) or by heavy smoking. However,
other studies do not indicate such hindered clearance either with smoking (Finkelstein
and Dufresne 1998) or with fibrosis.
Older Studies
Morgan and Holmes (1980) examined tissue samples from 21 patients in England (10
who died of mesothelioma, 3 who died of lung cancer, and 8 who died of other causes).
In this study, formalin-fixed tissue samples were digested with hypochlorite. The
residue was then rinsed, diluted, and an aliquot filtered. The filter was mounted on a
microscope slide and clarified for analysis by phase contrast optical microscopy.
Importantly, the authors note that chrysotile fibers were ignored In this study because
they would not generally have been detected by this technique. Portions of the filters
were also carbon coated and prepared for TEM analysis. Based on the observation
that only 19% of the fibers observed in this study were between 2.5 and 5 pm, when
the authors expect airborne distributions to contain closer to 90% of the fibere'within
this size range, the authors conclude that short fibers are preferentially cleared from the
lung. They also conclude, based on one subject with asbestosis whose lung tissue
exhibited 72% short fibers, that asbestosis hinders clearance. The authors also note
that fewer than 1 % of ferrugenous bodies (iron-coated asbestos bodies) are less than
10 pm in length, which indicates (in agreement with previously published work) that
such bodies seldom form on short fibers. They also suggest that virtually all fibers
longer than 20 fJm tend to be coated in the distributions they observe.
Le Bouffant (1980) studied the concentrations, mineralogy, and size distributions of
asbestos fibers found in the lungs and pleura of deceased asbestos workers. Based on
the analysis, he found that the average ratio of chrysotile fiber concentrations found in
the lung versus the pleura is 1.8 while for amosite the ratio is 34, This indicates that
chrysotile migrates from the lung to the pleura more rapidly than amphiboles resultina in
a higher fraction of total fibers in the pleura being composed of chrysotile (3% in the
lungs versus 30% in the pleura). With regard to size, the researchers found the size
distribution of amosite is virtually identical in the lung and pleura while chrysolite fibers
found in the pleura are much shorter than chrysotile fibers found in lung tissue This
suggests that the movement of chrysotile is a result of a combination of translocation
and degradation to shorter fibers (or that tissue samples have been contaminated with
environmentally ubiquitous short, chrysotile structures). The authors indicate that
chrysotile fibers apparently degrade to shorter fibers more rapidly than amosite and
translocate to the pleura more rapidly than amosite. Thus, a greater fraction of
7.56
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
chrysotile fibers (albeit short fibers) reach the pleura than amosite fibers over fixed time
intervals. However, the results of this study also confirm that the longer amosite fibers
do eventually translocate, although on a much more extended time scale than the
translocation of chrysotile.
Importantly, the results of this study need to be evaluated carefully. Boutin et ai. (1996)
showed that the majority of asbestos fibers in the pleura (particularly the long fibers) are
aggregated in localized "black spots" (which surround the sites of lymphatic drainage).
Thus, if the tissue samples analyzed by Le Bouffant do not contain representative sets
of such spots, the conclusions drawn by Le Bouffant may be subject to question.
In summation, human pathology studies tend generally to support the findings of other
studies regarding the size effects of asbestos (i.e. short fibers tend to clear more rapidly
than long fibers, which can be retained in pulmonary tissues for extended periods).
They also appear to highlight drastically different behavior between chrysotile and the
amphibole asbestos types (particularly tremolite) regarding the heavily favored retention
of the latter, which has also been indicated in animal studies. Unfortunately, the ability
to draw quantitative conclusions from human pathology studies is hampered by the
severe limitations of these studies (Section 5.2).
7.2.4 Studies of Dissolution/BioDurability
Although asbestos minerals are relatively insoluble in vivo in comparison, for example,
to various fibrous glasses or other manmade mineral fibers (see, for example,
Hesterberg et al. 1998a or Estes and Hadley 1996), they do eventually dissolve in the
body. Therefore, this pathway may contribute importantly to the overall biological
clearance of asbestos. Moreover, it has been suggested by several researchers (see,
for example, McDonald 1998 and other references cited below) that differences in
biodurability between chrysotile and the amphiboles may at least partially explain the
disparate potencies observed for these fiber types toward the Induction of
mesothelioma and, potentially, lung cancer (see Sections 6.2.4.2 and 6.3.3.2).
Note that the term "biodurability* is used here to indicate the persistence of a
particle or fiber attributable specifically to solubility (in the absence of other
clearance or degradation mechanisms). In contrast, the term "biopersistence" Is
used to indicate the overall persistence of a particle or fiber in the body
attributable to the combined effect of all mechanisms by which it might be
removed. Thus, for example, while biopersistence can be evaluated in vivo,
biodurability can best be inferred from in vitro dissolution studies so that effects
from other clearance mechanisms can be eliminated.
7.57
R«vMon 1 - warn
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Several studies further indicate that both the in-vivo biopersistence and the bio-activity
(including carcinogenicity) of various fiber types may be linked to their observed, in-vitro
dissolution rates (Eastes and Hadley 1995, 1996, Bernstein et al. 1996, Hesterberg et
al. 1998a, 1998b). Such studies, however, typically involve fiber types with dissolution
rates that are rapid relative to the rates of clearance by other mechanisms (Sections
7.2.1.1 and 7.2.1.2). In such studies, moreover, the various types of asbestos are
typically employed as negative (insoluble) controls. In fact, most of these studies are
based on experiments with rats and the 2-yr lifetime of a rat is comparable to the
anticipated lifetime of chrysotile asbestos in the body and short compared to the
anticipated lifetimes for the amphiboles (see below). Therefore, such studies are not
particularly sensitive to differences in the relative biodurability of the different asbestos
types. In fact, in the majority of these studies, the dissolution rates reported for
asbestos were derived indirectly by analogy with other minerals or are quoted from
other studies that derive rates similarly and may therefore be somewhat unreliable.
Nevertheless, a review of a subset of these studies is instructive.
• Eastes and Hadley (1995 and 1996) report a simple model that
reasonably predicts the relative fibrogenicity and tumorigenicity for a
range of synthetic fibers based on the dissolution rates of the fibers
measured in vitro. The authors found that they could explain observations
by assuming that the effects of the various fibers are a function of an
adjusted dose that accounts for biodurability. Thus,
F = f(ax)
where "F" is the observed incidence of the end point, T is the dose-
response function proposed for the effect, "x" is the measured dose^ and
"a" is an adjustment factor that accounts for durability.
In the model, "a" is determined simply as "VV where t^ is the time that a
fiber of diameter, "D" remains in the lung and ^ is the lifetime of the
exposed animal (e.g. 2 years for rats). This simple model reasonably
reconciles the results observed in animal inhalation and injection studies
of manmade mineral fibers (MMVF's), refractory ceramic fibers (RCF's),
and asbestos for end points including lung tumors, degree of fibrosis, and
(for intrapleural injection studies) mesothelioma. Based on a chi-square
test, the simple model is shown to adequately fit the data to a number of
databases reviewed. In contrast, the unadjusted doses do not.
Importantly, the dissolution rates used for the various asbestos minerals in
7.58
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
this study were estimated by analogy with similar minerals and therefore
may be unreliable.
• In studies comparing in-vivo biopersistence with dissolution rates
measured in-vitro, Bernstein et al. (1996) and Hesterberg et al. (1998a),
indicate that it is necessary to consider only long fibers (typically longer
than 20 pm), because shorter structures are typically cleared by other
mechanisms. They also indicate, at least for this type of study in which
whole lungs were homogenized and dissolved prior to preparation for
asbestos analysis, that clearance is initially rapid. This reflects muco-
ciliary clearance from the upper respiratory tract. Therefore, it is
clearance of long fibers from a longer term pool that tracts in-vitro
dissolution rates.
These authors "also report that long, soluble fibers (longer than 20 pm) are
actually cleared more rapidly than short fibers in these studies. They
indicate that this is likely due to long fibers being too long to be effectively
phagocytized by macrophages so that they are left to dissolve in the
extracellular fluid at neutral pH. Shorter fibers are effectively taken up by
macrophages so that dissolution is hindered by the more acidic
environment of the phagosomes (pH 4.5) and by the limited volume of
fluid within which to dissolve.
• Law et al. (1991) studied the dissolution of a range of fibers in solutions
used as common fixatives for biological samples. The authors report that
chrysotile and crocidolite, as well as many other fibers, dissolve at
measureable rates in the fixatives studied (Karnovsky's fixative and
formalin fixative). They therefore recommend that fiber concentrations
and size distributions obtained from tissue samples stored in such
fixatives should be evaluated carefully to account for the possible effects
of the fixatives.
• Although Coin et al. (1994) reported seeing no effective reduction in long
fiber (>16 (Jm) chrysotile (nor other evidence of dissolution) In their study
of fiber biopersistence, the limited time frame of this study {30-days) may
have been too short to allow detectable changes to accumulate.
The most consistent data for the comparative biodurability of chrysotile and the
amphiboles (specifically crocidolite) is found in two in-vitro studies of the dissolution
rates of fibers that were conducted under comparable conditions. In the first of these
7-5® Revision 1 >0/3*01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
studies* Hume and Rimstidt (1992) measured the dissolution rate for chrysotile
asbestos at neutral pH under conditions analogous to biological systems. The
dissolution rate that they report for chrysotile converts to: = 12.7 ng/cm2-hr and this
is reportedly independent of pH. In a comparable study Zoitus et al. (1997) report the
following dissolution rate for crocidolite: = 0.3 ng/cm2-hr, which is 40 times slower
than for chrysotile. Dissolution rates for several MMVFs and RCF-1 are also reported
in the latter paper, which are listed from fastest dissolving to slowest in Table 7-4.
Table 7-4:
Measured In-Vitro Dissolution Rates for Various Fibers1
Fiber Type H- (na/cm2-hr)
MMVF 10 259
MMVF11 142
MMVF 22 119
MMVF 21 23
Chrysotile 12.72
RCF1 8
Crocidolite 0.3
1 Source: Zoitus et al. (1997)
2 Source: Hume and Rimstidt (1992)
Note that dissolution rates for other amphiboles, such as amosite are probably
no more than a factor of two or three different than that reported above for.
crocidolite (see, for example, Hesterberg et al. 1998a,b).
To compare the effect of biodurability on the in-vivo biopersistence of asbestos and
other fiber types, both the detailed kinetics of dissolution and the distribution of fiber
sizes must be considered.
As reported by Zoitus et al. (1997), at a sufficiently high rate of fluid flow, the rate of
mass loss from a fiber is proportional to its surface area, A. Thus:
-dM/dt = -kA. (7.4)
This means that for a uniform mass fiber dissolving congruently:
= 2kt/Dep (7.5)
7.60
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
where:
M is the mass at time t;
M0 is the initial mass at time t« 0;
D0 is the initial diameter of the fiber; and
p is the density of the fiber.
Substituting the equation relating the mass and the diameter of a fiber (M « pnd2h/4)
into the above equation, cancelling terms, and rearranging indicates that (during
dissolution) the diameter of a fiber decreases linearly with time:
D = D0-2kt/p (7.6)
where:
D is the diameter at time t; and
ail other terms have been previously defined.
Furthermore, the rate of reduction in radius is given by: k/p. Based on the dissolution
rates given above for chrysotile and crocidolite, the radius reduction rates (v^) for
these fiber types are determined to be: 1.26 x 10"® pm/sec and 2.6 x 1O"t0 pm/sec,
respectively. Thus, the dissolution of each fiber is a zero order process
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Importantly, fibers in vivo will only dissolve at the rates predicted by the above
equations if the fluid in which they are dissolving flows past the fibers sufficiently rapidly
to prevent saturation from limiting the rate (Mattson 1992). Especially for slow
dissolving materials like asbestos, however, it is expected that the observed dissolution
rate in vivo will generally be slower than the rates predicted based on in vitro
measurements. Even for more rapidly dissolving fibers like most fibrous glasses and
manmade mineral fibers, dissolution is hindered in compartments of the body in which
the volume of available solute is limited.
In summary, dissolution is a zero-order (i.e. constant with time, independent of
concentration) clearance mechanism that is dependent on fiber mineralogy, that the
effect it has on fiber populations (concentrations) is a function of the distribution of fiber
diameters within the population, and that the theoretical rate of dissolution may not be
achieved in all tissues in all compartments of the lung or mesothelium due to limits in
the rate of in-vivo solute flow.
7.2.5 Dynamic Models
Unlike particle deposition in the lungs, which is an entirely mechanical process,
clearance, transport, and degradation mechanisms tend to be complex biochemical
processes. Due to the incomplete understanding of such processes, state-of-the-art of
degradation and clearance modeling is not as advanced as that for deposition. Even
the most sophisticated of these models remain semi-empirical. Although, current
models in this area show general agreement with the sparse, available data, there are
clearly areas of weakness that require additional research. Nevertheless, the models
provide a good indication of the kinds of processes that are important in the body and
their overall constraints. It is also noted that models for degradation and clearance in
humans tend to be better developed than those for animals primarily due to
confounding uncertainties associated with animal ventilation rates. An overview of the
state of the art, which was current as of the date of publication, can be found in Stober
et al. (1993).
According to Stober et al. (1993), the general conclusions that can be drawn from the
current models are that:
• clearance from ciliated airways is rapid, independent of particle/fiber type,
apparently independent of particle size, and can be described as the sum of two,
weighted exponentials (i.e. an assumed combination of two first-order decay
processes), although the process may in fact be zero order (i.e. the reduction in
concentration with time is constant and independent of concentration) with rates
7.62
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
that differ primarily by the distance that a particle must traverse to return to the
trachea.
Based on studies of particle clearance reported by Raabe (1984), muco-ciliary
transport in the nose and throat generally exhibits a half life for clearance of 4
minutes. Clearance of the tracheo-bronchial section of the respiratory tract is a
function of the distance from the trachea and generally varies from a half life of
30 minutes for the largest bronchi to approximately 5 hours for the smallest and
most remote bronchi. In healthy humans, material deposited in this region is
generally cleared within 24 hours. In contrast, the clearance mechanisms
operating in the deep lung, beyond the muco-ciliary escalator, operate over time
frames of many days to years (see Table 7-2);
clearance of insoluble particles from the pulmonary portion of the lung occurs
primarily by macrophage transport and such transport has several components.
One component represents the population of "free" macrophages located within
the alveoli that engulf particles and transports them to the muco-ciliary escalator.
Macrophages are also renewed at some rate of recruitment that may be
dependent on particle concentrations. In fact, numerous studies have
demonstrated that macrophage recruitment is induced by the deposition of
asbestos and other particles in the lung (Section 7.3.5). Particles also migrate
into the interstitium where another population of macrophages clears these
particles to lymph. This second component (interstitial clearance) is much
slower than the first (see Table 7-2);
each macrophage can carry a maximum load and the mobility of each
macrophage decreases with increasing load. At sufficient loading, macrophages
become immobile and aggregates of overloaded macrophages in the alveoli may
then sequester particles for some period of time as this clearance mechanism is
shut down. In the interstitium, masses of immobile macrophages may trigger
development of granulomas that sequester particles for extended periods of time
by effectively preventing clearance of the particles within such tissue, at least
until or unless the granulomas resolve. Thus, these models incorporate overload
mechanisms and the incorporation of such overload mechanisms are required to
explain observed trends in experimental results; and
in various published studies, overload (immobilization of laden macrophages)
has been modeled as dependent on the total volume or mass of phagocytized
material (for compact particles) and (additionally) on the length of phagocytized
material (for fibers). It is also possible that the motility of macrophages and the
7.63
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
consequent overall rate of this clearance process is additionally a function of
fiber diameter and/or particle toxicity (the latter for special cases).
Interestingly, while it is reported that large particles are not readily cleared by this
mechanism, the range of sizes over which clearance becomes hindered
corresponds reasonably well to the limits of overall respirability. In contrast,
fibers that are clearly too long to be cleared by macrophages, if they are
sufficiently thin, are quite respirable. Thus fibrous materials present a unique
challenge to the respiratory tract based solely on the dimensions of these
materials.
Stober et al. (1993) also notes that many models incorporate the assumption that most
clearance processes are first order (i.e. that the rate of reduction of mass or fiber
number is proportional to the remaining mass or fiber number, respectively, and
independent of other factors) and that the combined effects of multiple clearance
processes can therefore be expressed as a weighted sum of exponentials, has been
fairly successful. This means, however, that the half-lives "t1/2's" attributed to the
various first order decays are empirical and do not necessarily correspond to any
specific physiological or biochemical features of the processes being modeled.
Depending on the specific process, clearance rates may be zero order or may be a
more complicated function of multiple variables than can be described by a first order
decay. Nevertheless, models incorporating these simplifying assumptions have shown
good success at adequately describing observed effects.
Note that half lives for first order decay processes represent the time required for half of
the initial mass to decay (or be transported or whatever) and can be estimated as: t,a =
(In2)/k with k being the first order rate coefficient or proportionality constant between
rate and mass. This is why so many of the retention studies cited above provide
estimates of a series of decay constants or half-lives that are assumed to correspond
approximately to the major clearance processes contributing to the observed, overall
reduction in concentration.
Due to the complexity of the processes involved, only a small number of dynamic
models for fiber retention have been developed. Interpretation of the results of these
models requires that the meaning of the term "retention" first be reconciled across
studies.
• Dement and Harris (1979) report that, based on a mathematical model,
the fraction of structures retained in the deep lung is unlikely to vary by
more than a factor of 2 for different asbestos mineral types. In this study,
7.64
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
however, the term retention appears to refer primarily to a very short time
period that primarily includes consideration of deposition but not clearance
processes;
Using a definition for retention that reflects long-term residence in the
lung, Vu et al. (1990a) developed a model of chrysotile retention that
explicitly incorporates longitudinal splitting, dissolution, and size-
dependent clearance. Time-dependent lung burden estimates derived
using the model were shown to compare reasonably well with published
data (Abraham et al. 1988, as cited by Yu et al.) both in terms of fiber
concentrations and fiber size-distributions.
In a later modification of their retention model for chrysotile, Yu et al.
(1991) also considered the effect of airway asymmetry on fiber retention.
In this version of the model, Yu and coworkers incorporated information
concerning the geometry of the bronchio-alveolar tree (including mean
distance and the mean number of airway bifurcations between the trachea
and the alveoli in each section of the lung) and studied the effects of such
considerations. The modified model predicts a non-uniform distribution of
the asbestos that is retained in the lung and the predictions reasonably
reproduce the distributions observed by various researchers and
measured formally by Pinkerton et al. (as cited by Yu et al.).
Yu et al. (1990b) also modeled the long term retention of amosite in rat
lungs. In contrast to the models employed for chrysotile, the model
presented for amosite incorporates a term for the clearance rate that is
not a constant but, rather, is a function of the lung concentration of
asbestos, which was adapted from an earlier model for diesef soot. This
modification was incorporated to adequately mimic the suppression of
clearance with increasing lung burden that has been observed by several
research groups (e.g. Wagner et al. 1974 or Davis et al. 1978) for
amphiboles. Conditions under which elevated asbestos (or dust)
concentrations are observed to reduce clearance are referred to as
"overload" conditions. Model predictions were shown to reasonably
reproduce the time-dependence of amosite lung burdens (in terms of
mass) in several studies.
Importantly, the overload conditions addressed in the Yu et al. (1990b)
model were primarily observed among older retention studies where lung
burden was tracked as total asbestos mass (Section 7.2.1). Such studies
7.65
Revision 1 • 9/3/91
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
tend to suggest a difference in the behavior of chrysotile and the
amphiboles. As indicated in Section 7.2.1, however, later retention
studies, which track lung burden as a function of fibers number (in specific
size categories) tend to show this effect is a function of fiber length more
than fiber type and newer models may need to incorporate such factors
that indicate reduced macrophage motility as a function of fiber length.
Moreover, it is important to consider the major, confounding effects, if the
goal is to develop a model that not only reproduces the time-dependence
of clearance but also captures relevant physical phenomena. Thus, for
example, Yu et al. (1994) were able to reproduce the time-dependence of
the retention in rats of inhaled RCF-1 (as a function of fiber size) using a
model in which macrophage motility was limited only be total lung burden
and not dependent on fiber size. However, these authors also failed to
consider that long RCF-1 fibers in fact dissolve at rates competitive with
the clearance of short fibers (see Section 7.2.1), which is probably why
they did not find a dependence on length; the two effects cancelled out.
7.2.6 General Conclusions Regarding Deposition, Translocation, and
Clearance
The current literature on deposition, translocation, and clearance paint a consistent
picture of the fate of fibrous structures in the lung. The ultimate fate of biodurable fibers
depend overwhelmingly on their size. Although there may be additional effects due to
mineralogy (addressed further in Section 7.3) and for rare, special cases this may be
important: generally such effects appear to be minor.
The primary effect attributable to mineralogy that is important to consider in relation to
clearance is that associated with biodurability. Fibers that dissolve in the lung at rates
that are competitive with the other clearance mechanisms described below may be
cleared sufficiently rapidly to preclude adverse effects, even when such fibers are too
long to be cleared efficiently by macrophages (see Section 7.2.1).
A schematic representation of the complex set of mechanisms that contribute to the
translocation and clearance of fibrous structures that have been deposited in the deep
lung is presented in Figure 7-4. This description was developed based on the complete
spectrum of observations reported in each of the previous sections of this Chapter
including, primarily, the descriptions of the most sophisticated of the models reviewed
by Stober et al. (1993).
7.66
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
As shown in Figure 7-4, briefly, the first reaction to the introduction of fibers (or other
particulate matter) into the alveolar lumen is scavenging by alveolar macrophages. It
has been reported that the initial uptake by macrophages is a rapid process that is
essentially complete within hours after initial deposition. Rates for several of the
mechanisms depicted in Figure 7-4 have been estimated in the literature and are
summarized in Table 7-2.
FIGURE 7-4
7.67
Revision 1 • 6/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
FIGURE 7-4, PAGE 2
7.68
Revision 1 • 8/3/01
-------
Lwk«V|
j T * H'Vj* IN Qtt^ ^ "^«l^;,
i I ; ¦ ^
I ! • ^
! ! ! »
z-^-^L. I" " 'WtM
_C : OWm* '
• r^tyri«Ww\
I • •
40* ^
;—l-=H
CjXXMAJomr^
Sttjjv
I
-------
Figure 7-4 (cont, page 2 of 4)
Key for Putative Mechanisms for Clearance and Translocation of Fibers
in the Lung
K, = rate constant for phagocytosis of fibers by alveolar macrophages. This
mechanism is an inverse function of fiber length and, likely, diameter. This
mechanism is likely pseudo first order (assuming sufficient numbers of
macrophages, the rate will be proportional to the number of fibers);
K2 = rate constant for phagocytosis of fibers by Type I epithelial cells. This
mechanism is an inverse function of fiber length and, likely, diameter. This
mechanism is likely pseudo first order;
K3 = rate constant for diffusion of fibers from the alveolar air space (lumen) through
the epithelial lining to the underlying interstitium. This mechanism is diffusion
limited and likely independent fiber size or type. This mechanism likely parallels
the behavior of diffusion in through a finite, column of fixed diameter CHECK
THIS;
«4 = rate constant for phagocytosis of fibers by Type II epithelial cells. This
mechanism is an inverse function of fiber length and, likely, diameter. This
mechanism is likely pseudo first order;
Ks = rate constant for transport by macrophage to the muco-ciliary escalator. This
mechanism is likely an inverse function of fiber length and the volume (or mass)
of the fibers phagocytized. Macrophages that become immobilized tend to
aggregate in alveolar lumena. For macrophages with fixed loads, this
mechanism may be first order or zero order CHECK THIS;
Kg = rate constant for putative discharge to interstitium of phagocytized fibers by Type
I epithelial cells. There has been no direct verification of this mechanism;
K7 = rate constant for phagocytosis of fibers by endothelium. This mechanism is an
inverse function of fiber length and, likely, diameter. This mechanism is likely
pseudo first order;
KB = rate constant for putative mechanism in which fibers internalized by
macrophages are transported through the epithelial lining of the alveolar space
to the underlying interstitium. There has been no direct verification of this
mechanism;
Kg = rate constant for diffusion of fibers from the interstitium through the endothelial
lining to the enclosed, capillary lumen. This mechanism is diffusion limited and
-------
Figure 7-4 (cont, page 3 of 4)
Key for Putative Mechanisms for Clearance and Translocation of Fibers
in the Lung
likely independent fiber size or type. There has been no independent
verification of this mechanism;
K10 = rate constant for phagocytosis of fibers by interstitial macrophages. This
mechanism is an inverse function of fiber length and, likely, diameter. This
mechanism is likely pseudo first order (assuming sufficient numbers of
macrophages, the rate will be proportional to the number of fibers);
Kn = rate constant for transport by macrophage from the interstitium to the lymphatic
system. This mechanism is likely an inverse function of fiber length and the
volume (or mass) of the fibers phagocytized. Macrophages that become
immobilized tend to induce formation of interstitial granuloma. For macrophages
with fixed loads, this mechanism may be first order or zero order CHECK THIS;
K12 = rate constant for putative discharge to capillary lumena of phagocytized fibers by
endothelial cells. There has been no direct verification of this mechanism;
K,3 = rate constant for phagocytosis of fibers by mesothelial cells of fibers transported
to the mesothelium. This mechanism is an inverse function of fiber length and,
likely, diameter. This mechanism is likely pseudo first order;
k M = rate constant for putative mechanism in which fibers internalized by
macrophages are transported from the interstitium through intervening tissue
and the mesothelium to the pleural space. There has been no direct verification
of this mechanism;
K|;, = rate constant for phagocytosis of fibers by pleural macrophages. This
mechanism is an inverse function of fiber length and, likely, diameter. This
mechanism is likely pseudo first order (assuming sufficient numbers of
macrophages, the rate will be proportional to the number of fibers);
K„ = rate constant for transport by macrophage to sites of lymphatic drainage
(lymphatic ducts) along the pleura. This mechanism is likely an inverse function
of fiber length and the volume (or mass) of the fibers phagocytized.For
macrophages with fixed loads, this mechanism may be first order or zero order
CHECK THIS;
Not shown: apparently fibers that are too large to pass through lymphatic ducts
attract accumulation of macrophages at sites of lymphatic drainage (Kane and
McDonald (1993).
-------
Figure 7-4 (cont, page 4 of 4)
Key for Putative Mechanisms for Clearance and Translocation of Fibers
in the Lung
R1 = rate constant for the renewal of the alveolar macrophage population. While the
there is likely a steady state rate for background renewal, given that the average
life of an alveolar macrophage is reported to be on the order of seven days
(REF), this rate is also stimulated in response to insult by foreign substances in
the lung;
R2 = rate constant for the renewal of the interstitial macrophage population. While
the there is likely a steady state rate for background renewal, this rate is also
expected to be stimulated in response to insult by foreign substances in the
lung;
R2 = rate constant for the renewal of the interstitial macrophage population. While
the there is likely a steady state rate for background renewal, this rate is also
expected to be stimulated in response to insult by foreign substances in the
lung;
R3 = rate constant for the renewal of the interstitial macrophage population. While
the there is likely a steady state rate for background renewal, this rate is also
expected to be stimulated in response to insult by foreign substances in the
lung;
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
The rate of removal (to the muco-ciliary escalator) by alveolar macrophages is then
determined by a variety of effects. Macrophage motility is a size dependent-process so
that only fibers that are sufficiently compact (< -20 pm) can be removed from the lung.
The rate of removal by this process may also be suppressed both for fibers of
intermediate lengths (10-20 pm, which are short enough to be phagocytized but long
enough to suppress macrophage motility) and by the overall mass/volume of particles
deposited (and, proportionally, taken up by each macrophage). Note that the
dimensions provided are the ones that are apparently appropriate for humans. For rats,
the corresponding dimensions may be somewhat smaller.
Likely competing with scavenging by macrophages are (1) phagocytosis by the
epithelial cells lining the alveolus and (2) diffusive transport to the intersititium. Both
Type I and Type II epithelial cells appear to phagocytize fibers. Although relatively few
fibers are observed to be taken up by Type II cells, as previously discussed (Section
7.2.2), one possible explanation for the limited observation of fibers in Type il cells is
that uptake of fibers induces terminal differentiation to Type I cells. It is expected that
phagyctosis by epithelial cells is a size-dependent process.
Especially when the presence of fibers (or other particulate matter) induces
morphological changes in Type II cells that increase the overall permeability of the
epithelial lining (Section 7.3.7), fibers can apparently diffuse into the interstitium. This
process, potentially supplemented with expulsion of phagocytized fibers by epithelial
cells and/or transport of fiber laden macrophages through the epithelial lining,
represents the set of putative mechanisms by which fibers may reach the interstitium. It
is expected that these processes are somewhat slower than uptake by alveolar
macrophages (see Table 7-2). Therefore, if this latter mechanism is operating at peak
efficiency, relatively few fibers may reach the interstitium. Note that diffusive transport
is likely independent of fiber length, but may be dependent on fiber width with thinner
fibers more rapidly diffusing to the interstitium.
Fibers reaching the interstitium are likely cleared primarily by interstitial macrophages,
which phagocytize the fibers and transport them to the lymphatic system. Both the
efficiency of phagocytosis and motility of macrophage transport in the interstitium likely
depend on fiber size and the total volume/mass of fibers in the same way described
above for transport by alveolar macrophages. However, all such mechanisms are
substantially slower in the interstitium than in the alveolar lumen (see Table 7-2).
Macrophages that have been immobilized (due to fiber size or volume/mass) in the
interstitium tend to aggregate and induce formation of granulomas, which may
sequester the fibers in these cells. Although there is less evidence for this, fibers free
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
in the interstitial matrix might also trigger such a process. Such fibers would typically be
too large to have been effectively phagocytized by any of the cells of the interstitium.
Fibers may also reach the endothelium and be taken up by endothelial cells lining the
capillaries of the deep lung. Because fibers have also been observed in capillary
lumena, mechanisms similar to those described for transport through the alveolar
epithelium to the interstitium may be operating to transport fibers into capillary lumena.
While it is expected that such mechanisms will also show size dependence similar to
that previously described, little is known about the details or the rates of such
processes.
Also by mechanisms similar to some or all of the putative mechanisms described for
translocation of fibers from the alveolar lumen to the interstitium, fibers may reach the
pleura. Whether fibers can also reach the pleura via transport in blood or lymph has
not been definitively determined. Fibers reaching the pleura may be phagocytized by
mesothelial cells or may pass through such cells to the pleural cavity. Fibers reaching
the pleural cavity are apparently phagocytized by pleural macrophages (probably
showing a similar size or volume/mass effect as described above for similar
mechanisms) and are apparently transported to (and deposited at) sites of lymphatic
drainage along the pleura. If such fibers are then too large to pass through the
lymphatic ducts, they may trigger accumulation of additional macrophages and other
inflammatory cells.
Overall, the effects of size appear to be:
• few fibers thicker than approximately 0.7 pm and virtually none thicker
than approximately 1.5 pm appear to reach the deep lung. Of these, the
longer the fiber, the thinner the fiber. Importantly, the distribution of sizes
of structures deposited in the deep lung tend to be much more similar
across studies than the distributions in the aerosols originally inhaled.
Thus, deposition is a very effective filtering process; and
• short fibers (or compact particles) that are shorter than somewhere
between 10 and 20 pm tend to be taken up almost entirely by
macrophages and are either cleared via the muco-ciliary escalator,
isolated in immobilized macrophages that remain within alveolar lumena,
or transported to lymphatics (presumably after first reaching the
interstitum). These processes appear to be efficient for the shortest of the
fibers in this range (and all shorter fibers) so that no further effects are
manifest. In contrast, longer fibers, which are not efficiently cleared or
7.72
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
isolated by macrophages either in the alveolar lumen or the interstitium
appear to trigger a range of additional responses, some of which appear
to lead to disease.
Therefore, based on deposition, translocation, and clearance, It is fibers thinner than
approximately 0.7 pm and longer than a minimum of approximately 10 pm
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
and other chemical signals produced by asbestos exposed cells in lung tissue
immediately proximal to the mesothelium may also play a role (see Adamson 1997, as
described in Section 7.3.4.1).
That the specific biochemical triggers for asbestos-related diseases have not been
definitively delineated as of yet is not surprising. Despite great progress in elucidating
candidate mechanisms, the number of candidate mechanisms is large and confounded
by "cross-talk" between mechanisms. Moreover, similar toxic end points may result
from entirely independent mechanisms that exhibit disparate dose-response
characteristics but, nevertheless, may be triggered by the same or similar toxins. In
such cases, however, the relative contributions from each mechanism to a particular
end point may vary substantially. Unfortunately, the ability to compare results across
studies of different mechanisms is currently limited due to the inability to reconcile the
quantitative effects of dose and response across dissimilar studies.
Table 7-5 illustrates the range and complexity of the biological responses triggered by
asbestos in lung tissue. The table was developed based on the information gleaned
from the studies described in this Section. Importantly, not all of the mechanisms listed
contribute equally to the toxic end points that are attributable to asbestos, but their
relative importance has yet to be entirely delineated. The toxic end points of potential
interest to which each of the listed mechanisms potentially contribute are indicated in
bold italics.
As indicated previously (Section 7.0), although much has been learned about specific
components of the underlying mechanisms by which asbestos causes the above-listed
diseases, substantial knowledge gaps remain. Moreover, because of these gaps,
multiple candidate effects have been explored as potential contributors to
carcinogenicity (or fibrogenicity) and one of the goals of this document is to distinguish
among those effects that are likely to contribute to the induction of cancer from those
that are less likely or unlikely to contribute (given the current state of knowledge).
Accordingly, an overview of recent studies is presented below following a brief
description of the current model of the general mechanism for cancer. Note that, due to
the availability of several recent reviews (including, for example, Floyd 1990, Kamp et
al. 1992, Kane and McDonald 1993, Economou et al. 1994, Oberdorster 1994,
Mossman et al. 1996, Mossman and Churg 1998 and Robledo and Mossman 1999),
only the most recent primary articles are included in the following review.
Also, as indicated below, many of the biological responses provoked by retained
asbestos are in fact dependent on fiber size and type. Therefore, studies that
distinguish among effects induced by different size fibers (or fibers and non-fibrous
7.74
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
particles) of the same mineralogy and studies that distinguish among effects induced by
comparably sized fibers (or non-fibrous particles) but differing mineralogy are
highlighted. However, due to the limits with which fibrous materials have tended to be
characterized in many of these studies, the database from which to distinguish among
the effects of size and mineralogy are limited and conclusions from differing studies
must be compared with caution.
7.75
Revision 1 - 9/3/D1
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 7-5, Page 1
7.76
Revision 1 • 9/3/01
-------
Table 7-5:
Putative Mechanisms by Which Asbestos May Interact with Lung Tissue to Induce Disease
Following Inhalation
Asbestos (In vivo):
generates reactive oxygen species (ROS)
may affect neighboring cells and tissues
Interacts with macrophages
- interacts with lung epithelium
Asbestos Interacts with macrophages:
Induces generation and release of reactive oxygen species/reactive nitrogen species (ROS/RNS)
may affect neighboring cells and tissues
- may induce inflammatory response (which may promote cancer)
may Induce fibrogenesis (which may promote cancer)
induces signaling cascades In macrophages and neighboring tissues
mediates apoptosis (which may regulate cancer)
mediates proliferation (which may promote cancer)
causes genotoxlc effects in neighboring tissues
may cauae cancer Initiation
- may induce arrest of cell cycle
Induces signaling cascades
mediates apoptosis (which may regulate cancer)
mediates proliferation (which may promote cancer)
- depletes reserves of antioxidants In macrophages and neighboring tissues
may induce cytotoxicity (which may promote cancer, by Inducing proliferation)
may increase susceptibility to insult by other toxic agents (which may promote cancer)
Induces release of various cytokines
affects neighboring cells and tissues
- mediates Inflammatory response (which may promote cancer)
- mediates fibrogenesis (which may promote cancel)
- Induces signaling cascades in macrophages and neighboring tissues
mediates apoptosis (which may regulate cancer)
- mediates proliferation (which may promote cancer)
mediates proliferation In neighboring tissues (which may promote cancer)
- Induces signaling cascades and mediates apoptosis In macrophages (which mey regulate cancer)
at high enough concentrations, promotes cytotoxic cell death (which may promote cancer by Inducing proliferation)
-------
Table 7-5 (Page 2 of 2)
Asbestos interacts with Lung Epithelium:
- induces release of reactive oxygen species/rreactlve nitrogen species (ROS/RNS)
- may affect neighboring cells and tissues
- may induce Inflammatory response (which may promote cancer)
- may induce fibrogenesis (which may promote cancer)
- induces signaling cascades In epithelium and neighboring tissues
- mediates apoptosis (which may regulate cancer)
- mediates proliferation (which may promote cancer)
- causes genotoxic effects in epithelium and neighboring tissues
- may cause cancer Initiation
- may induce arrest of cell cycle
- induces signaling cascades
- mediates apoptosis (which may regulate cancer)
- mediates proliferation (which may promote cancer)
depletes reserves of antioxidants in epithelium and neighboring tissues
- may induce cytotoxicity (which may promote cancer by Inducing proliferation)
may increase susceptibility to insult by other toxic agents (which may promote cancer)
induces release of various cytokines
- affects neighboring cells and tissues
- mediates inflammatory response (which may promote cancer)
- mediates fibrogenesis (which may promote cancer)
- induces signaling cascades in epithelium and neighboring tissues
mediates apoptosis (which may regulate cancer)
mediates proliferation (which may promote cancer)
mediates proliferation In epithelium and neighboring tissues (which may promote cancer)
- causes genotoxic effects
- may cause cancer Initiation
may induce arrest of cell cycle
- induces signaling cascades
mediates apoptosis (which may regulate cancer)
mediates proliferation (which may promote cancer)
- induces signaling cascades
mediates apoptosis (which may regulate cancer)
mediates proliferation (which may promote cancer)
- increases epithelial permeability
encourages fibrosis (which may promote cancer)
facilitates translocation
at high enough concentrations, promotes cytotoxic cell death (which may promote cancer by Inducing proliferation)
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 7-5, Page 2
7.77
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
A large body of evidence supports the conclusion that it is primarily (if not exclusively)
long fibers (those longer than a minimum of 5 to 10 pm) that contribute to disease (see
Sections...7.2.1, 7.2.2 and 7.4). Much evidence also indicates that the potency of long
fibers increases with length at least up to a length of approximately 20 pm. Therefore,
because short and long fibers are both respirable (for fibers that are sufficiently thin,
less than approximately 0.7 pm in diameter), differences in the ultimate response to
short and long fibers must be attributable to differences in tissue and cellular responses
to the retained fibers in each size range.
At least at a histopathological level, clear differences have been observed in the
responses evoked by short and long structures (see Sections 7.2.1. and 7.2.2). It is a
goal of this section to determine whether the biochemical triggers that mediate the
disparate responses to short and long fibers can also be identified. Unfortunately, while
a large body of knowledge has been amassed, definitive conclusions are not yet
possible. This is because the specific mechanisms by which asbestos acts have still
not been definitively determined, although many candidate mechanisms have been
elucidated (see above). However, important inferences can still be gleaned from the
available studies.
Evidence for the relationship between fiber diameter and disease is somewhat less
clear. Although there appears to be a fairly sharp cutoff in the diameter of fibers that
are respirable (see Section 7.1.4), several studies suggest that the most potent fibers
are substantially thinner than the sharp cutoff in respirability (see Section 7.4). If these
latter observations are valid then, as with length, differences in the ultimate response to
thick and thin fibers must be attributable to differences in the tissue and cellular
responses elicited by retained fibers of each width. As indicated with length, however,
definitive conclusions regarding biochemical triggers and the effect of size on such
triggers are not yet possible because the specific triggers that lead to specific asbestos-
related diseases have not been definitively identified. Still, useful inferences can be
developed.
7.3.1 The Current Cancer Model
The following description of the current model for cancer is derived from the ideas
presented in Moolgavkar et al. 1988, Mossman 1993, and Economou et al. 1994.
In the current model of cancer, normal cells must accumulate specific, multiple
mutations before a tumorigenic cell is created that can lead to the development of
cancer. Each successive mutation produces an initiated cell (a cell that is transformed
from normal cells because it incorporates one or more of the requisite mutations, but
7.78
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
that has not yet acquired all of the changes needed to produce cancer). Each initiated
cell may then proliferate to generate a population of similarly initiated cells, which
increases the probability that other events will lead to further mutation in at (east one of
these cells. Individual mutations may occur spontaneously or may be induced by
exposure to mutagens.
Generally, the minimum, heritable changes required before a normal cell is transformed
into a metastatic tumor cell include, but may not be limited to (see, for example, Hei et
al. 1997 or Kravchenko et al. 1998):
• escape from terminal differentiation or programmed cell death (especially in
response to DNA damage);
• escape from anchorage/neighbor dependent growth inhibition;
• development of self-promoting growth and proliferation; and
• active expression of cytokines needed to promote angiogenesis and allow tissue
invasion.
It is not clear whether the mutations associated with these charges need to occur in a
particular order, although the first of the above-listed changes would facilitate
accumulation of all later changes. It is also unclear which of the above changes
contribute to the time dependence of tumor development and this may vary among
tumor types. It is likely that only a subset of the required mutations determine the
ultimate time development of the associated cancer. For example, once a cell begins
self-promoting growth, later mutations (even relatively rare ones) may become
incorporated rather quickly. Also, some mutations may be rare or may require
intervention by a toxic agent, while other mutations may occur spontaneously and may
thus occur frequently, once a sufficient number of initiated precursor cells are
generated. This is one of the reasons that models with as few as one or two stages
have proven successful at predicting the time course of many types of cancer (see, for
example, Moolgavkar et al. 1988).
To produce cancer, the mutations that occur must also be "heritable" meaning that they
must be preserved and passed on to daughter cells during mitosis. Thus, it is not only
necessary to cause alteration in the DNA of a cell (genetic damage), but the
mechanisms by which the cell subsequently repairs such changes or prevents cell
division (e.g. arrest of the cell cycle, programmed cell death, terminal differentiation) in
the presence of such changes must also be defeated. Generally, if a cell proceeds
7.79
Revision 1 - B/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
through DNA synthesis (in preparation for mitosis) before accumulated alterations to
DNA are repaired, the sites of such alterations can lead to errors in replication during
synthesis, which in turn result in permanent, heritable mutations in one or both of the
daughter cells that are created from mitosis.
Toxic agents that (directly or indirectly) cause DNA alterations may contribute to the
development of cancer by inducing one or more of the set of requisite mutations
required for cancer to develop. In traditional parlance, such agents are termed
"initiators". In addition, any toxic agent that enhances proliferation also facilitates the
development of cancer both by increasing the probability of creating spontaneous
mutations (due to errors that infrequently but unavoidably occur whenever DNA is
synthesized in support of mitosis) and by increasing the numbers of any initiated cells
that may be present, which may then serve as additional targets for initiators or may
incorporate additional spontaneous mutations. Agents that facilitate the development of
cancer by inducing proliferation are traditionally termed "promoters."
Multiple mechanisms have been identified by which both initiators and promoters may
act. Initiators, for example, may react directly with DNA to cause genetic damage, or
may induce generation of other reactive species (such as reactive oxygen species or
reactive nitrogen species) that, in turn, react with DNA to cause genetic damage. In
addition, fibrous materials may uniquely damage chromosomes by interfering with
mitosis causing aneuploidy (incorporation of an incorrect number of chromosome
copies in cells) and/or various clastogenic changes (alterations in the organization and
structure of specific chromosomes).
Promoters may also induce proliferation via a variety of mechanisms. Cytotoxic agents,
for example, may induce proliferation in a tissue by damaging or killing cells and
thereby induce stem cells to proliferate to replace the damaged cells. Promoters may
also induce release of various growth factors that, in turn, induce proliferation in
targeted tissues. This may occur, for example, as part of the inflammatory response to
tissue insult.
Promoters that are biopersistent (such as long asbestos fibers) or promoters that are
continually reintroduced (through chronic, external exposure) may also chronically up
regulate (or down regulate) certain cell signaling cascades that may contribute to
cancer development in a variety of ways including: (1) activation of genes that mediate
proliferation or production of various growth factors (including any of various
oncogenes) or (2) suppression of genes that inhibit proliferation or growth (including
any of various tumor-suppressor genes).
7.80
Revision 1 -S/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
There is growing evidence that all varieties of asbestos fibers {and certain other fibrous
materials) can act both as cancer initiators and promoters. However, the biological
responses to these materials appear to vary in different tissues so that it may be
important to separately evaluate the behavior of asbestos in specific tissues.
Biological responses to varying fiber types also appear to vary, particularly in relation to
a fiber's biopersistence (Sections 7.2.1 and 7.2.4). Accordingly, an overview of the
generic evidence that specific types of asbestos may act as an initiator and, separately,
as a promoter is reviewed below, followed by consideration of the more limited data
suggesting tissue-specific variation in biological responses.
7.3.2 Evidence for Transformation
Several recent studies provide evidence that specific types of asbestos can induce
transformation of cells in specific tissues (both lung epithelium and mesotheiium) to
create tumor cells. This provides further, confirmatory support for the whole animal
studies in which cancer is induced by exposure to asbestos (see Sections 7.4).
• An immortalized but non-tumorigenic cell line of human bronchial epithelial cells
(BEP2D cells) was transformed by a single exposure to 4 ug/cm2 UICC
chrysotile B (Rhodesian). Surviving cells (the treatment caused 18% cell death)
went through several transformations including: altered growth kinetics,
resistence to serum induced terminal differentiation, and loss of anchorage
dependent growth, before becoming tumorigenic (Hei et al. 1997).
Tumorigenicity developed in the various exposed cell lines over a period of
several to 11 weeks following exposure. When injected into nude mice,
secondary tumors developed with a latency of 8 to 10 weeks.
The authors indicate that there were no mutations in these cells at either codon
12 or 13 or the ras gene (mutations that have sometimes been observed in
asbestos-induced lung cancers. Also, because this cell line already contains
alterations in genes for p53 (a protein that plays a role in cell-cycle control,
among other things, see Table 7-6) and Rb (Table 7-6), the authors speculate
that such changes are not rate controlling for transformation to cancer.
It should also be noted that cultures of Type II cells have been particularly
difficult to maintain due to the tendency of these cells to undergo terminal
differentiation (to Type I cells) once they are removed from their natural
environment in the epithelial lining of lung alveoli (see Leikoff and Driscoli 1993,
as described in Section 4.4).
7.81
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
• Kravchenko et al. (1998) indicates that, unlike cultures of lung epithelial cells, in-
vitro cultures of rat mesothelial cells tend to transform spontaneously to
tumorigenic cells. The major changes that occur with time include: altered
response to epithelial growth factor (from growth-proliferation inhibition by this
factor to growth-proliferation stimulation), morphological changes (from polygonal
epithelial-like cells to elongated fibroblast-like cells and, eventually,
polynucleated cells exhibiting broad polymorphism), and multi-layered growth
and an ability to grow as colonies and masses in semi-solid agar. Eventually,
these cells become immortal and induce cancer when harvested and injected
into other rats. The authors indicate that asbestos-induced transformations in
such cultures proceed through identical stages but that they occur much more
rapidly. For example, incorporation of the stimulating response to EGF occurs
spontaneously at passages 9 or 10 but at passages 6 or 7 when induced by
asbestos. Asbestos was applied at 5 ug/cm2, which was noted to be sub-lethal
(95% cell survival was noted at this rate of application).
• In a study in which p53 deficient mice were intrapleural^ injected with UICC
crocidolite (200 pg/week), Marsella et al. (1997) showed that p53 deficient mice
exhibited substantially increased susceptibility to development of mesothelioma.
In this study, 12.5% of homozygous mice (p53 deficient) developed
mesothelioma and the rest died of lymphomas or hemangiosarcomas that
develop spontaneously in these mice; 76% of heterozygous mice died of
mesothelioma, and only 32% of wild-type mice (p53 competent) died of
mesothelioma. The authors suggest that p53 deficient mice are susceptible to
excess proliferation induced by crocidolite due to loss of control at the G1/S
check point that is normally mediated by p53.
In further confirmation of this hypothesis, Marsella et al. report in the same study
that 7.5 pg/cm2 of crocidolite applied to wild murine mesothelial cells in culture
induced substantial apoptosis while p53-deficient cells were resistant to
apoptosis. The authors also note that most of the p53-deficient cells are
tetraploid (suggesting a loss of a spindle check point) while the wild type cells
are all diploid.
Note that, although these studies suggest that asbestos alone can induce complete
transformation of both lung epithelial cells and mesothelial cells, such studies must be
evaluated with caution. In the case of the Hei et al. study, for example, the effects of
the (asbestos-independent) mutations required to initially establish the immortal line of
lung epithelial cells are not entirely known. Therefore, the response to asbestos of
epithelial cells in vivo may be substantially different.
7.82
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
In the case of the Kravchenko et a|. study, it is clear that mesothelial cells in vitro do not
behave in the same manner as those in vivo; in vitro, they spontaneously transform to
tumorigenic cells. This suggests that one or more growth inhibitory signals exist in vivo
(which are absent in vitro) that are critical to maintaining the health of the mesothelium.
7.3.3 Evidence that Asbestos Acts As a Cancer Initiator
As indicated in a review by Jaurand (1997), historically, the status of asbestos as a
mutagen was questioned, due primarily to the failure to produce detectable gene
mutations in short-term assays. However, mote recent studies provide clear evidence
that asbestos (and other fibrous materials) can produce mutations. Moreover, fibrous
materials may induce mutations by multiple mechanisms including, for example:
• direct interference with mitosis;
• production and release of reactive oxygen species (ROS); or
• production and release of reactive nitrogen species (RNS).
Of these, the most consistent, positive evidence that asbestos can act as a cancer
initiator (that is, a genotoxin or mutagen) has been from assays designed to detect the
kinds of genetic damage that result from interference with mitosis (Jaurand 1997).
Although generation of ROS and RNS plays a clear role in mediating asbestos-induced
disease, the direct link to a role of asbestos as an initiator is somewhat more tenuous
(see Sections 7.3.3.2 and 7.3.3.3).
7.3.3.1 Interference with mitosis. Apparently, the physical presence of asbestos
fibers can interfere with proper spindle formation and the function of other structural
units required for mitosis (Jaurand 1997). This typically leads to aneuploidy (an
incorrect number of copies of the chromosomes contained within a cell), development
of micronuclei (fragments of chromosomes enclosed in a membrane that are isolated
from the main nucleus of the cell), and has also been shown to lead to clastogenic
effects (changes in the organization and structure of the chromosomes). Assays for
these kinds of genetic alterations have consistently shown asbestos capable of
inducing these effects.
Jaurand (1997) also indicates that:
• fibers must be phagocytized by the target cells before they can interfere
with mitosis;
7.83
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
• once phagocytized, all asbestos types are observed to interfere with
mitotic activity; and
• samples enriched in long, thin fibers enhance these effects. In contrast,
short fibers do not appear to contribute to these effects.
Jaurand notes, however, that results related to fiber size have not been entirely
consistent, primarily because not all studies have rigorously controlled for or even
adequately characterized the sizes of the fibrous structures in test materials. Jaurand
also notes that this mechanism is not dependent on the formation of reactive oxygen
species or any other reactive free radicals.
Among studies that suggest that surface chemistry (or presumably fiber type) may not
be an important factor in determining the degree to which asbestos (or other fibrous
materials) interfere with mitosis, Keane et al. (1999), exposed cultured V79 cells
(hamster lung fibroblasts) to untreated and HCI-treated chrysotile asbestos. The acid
treatment substantially reduces the magnesium content on the surface of the chrysotile.
The authors also noted a "small" effect on fiber size due to treatment (treated fibers
showed a 20% excess of short fibers). Cells were exposed to doses ranging between
0.4 and 12.7 jjg/cm2 (each dose left in place for 24 hrs and then rinsed off). Cells were
harvested after an additional 24 hrs and evaluated for cytotoxicity and the presence of
micronuclei.
Results from the Keane et al. study indicate that untreated fibers were shown to be
slightly more cytotoxic than treated fibers but both treated and untreated fibers were
observed to increase the abundance of micronuclei in a similar, dose-dependent
fashion. The induction of micronuclei appeared to saturate at approximately 35/1000
cells observed at an applied asbestos concentration of 40 |jg/cm2. In contrast,
substantial cytotoxicity was only observed at the highest doses employed. The authors
thus concluded that surface chemistry (at least in terms of magnesium content) does
not appear to have a major affect on induction of micronuclei and that the observed
genetic damage and cytotoxicity appear to occur through entirely independent
mechanisms.
In other studies, cultured cells were assayed for a variety of genotoxic effects following
exposure to a range of fibrous materials.
• For example, Dopp and Shiffmann (1998) dosed human amniotic fluid (HAF)
cells or Syrian Hamster Embryo (SHE) cells with UICC amosite, Rhodesian
7.84
Revision ^ • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
chrysotile, crocidolite, or ceramic fibers at concentrations of 0.5,1.0, 5.0, and
10.0 pg/cm2 and assayed the cultures for formation of micronuclei and a variety
of clastogenic effects.
Based on their study, Dopp and Shiffmann report that all asbestos types induced
formation of micronuclei in SHE celts in a dose-dependent fashion at rates that
were significantly elevated over control animals. The effect appeared to
saturate at doses between 1 and 5 pg/cm2 and rates appeared to peak at
between 48 and 66 hrs post exposure. Ceramic fibers, which were noted to be
longer but thicker than the asbestos fibers tested, also showed significantly
increased induction of micronuclei, but at rates less pronounced than for
asbestos. However, it is difficult to judge whether this is due to differences in
fiber type because the data in the paper are not adequate to distinguish effects
of fiber size and number from effects of fiber type. Similar results were obtained
with HAF cells, but overall rates were about one third those observed in SHE
cells.
The authors also note observing various disturbance of chromatin structure
during interphase. They report observing formation of chromatin bridges and
chromosome displacements in meta and anaphases and impaired chromatin
separation in mitosis. They also report that cytokinesis was frequently blocked.
In addition to the effects that they observed that are attributable to fiber
interference with mitosis, Dopp and Shiffmann report observing a variety of
clastogenic effects. In all cases where authors labeled specific regions of
specific chromosomes, fiber-exposed cells showed significantly greater
frequencies of DNA breaks over controls or gypsum-exposed animals. Different
regions of various chromosomes also showed significantly different frequencies
of breakage with patterns that were specific to the different fiber types. The
authors hypothesize that the observed clastogenic effects may be due to
production of reactive oxygen species, to formation of some type of clastogenic
factor, or to the direct interactions between fibers and chromosomes during
mitosis that are the apparent cause of the disturbances discussed in the
previous paragraphs. However, It was not possible to distinguish among
hypothetical causes of the observed clastogenic effects in this study.
• Kodama et al. (1993) exposed cultures of human bronchiolar epithelial (HBE)
cells to asbestos (chrysotile at 0 to 4 pg/cm2 and crocidolite at 0 to 300 (jg/cma).
They then examined cells at 24,48, 72, and 96 hrs following exposure for
cytotoxic effects and cytogenetic effects. Results indicate that both fiber types
7.85
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
induced concentration-dependent inhibition of cell proliferation and colony-
forming ability but chrysotile was 100 to 300 times more toxic. The authors
report this translates to a 40 fold increase in toxicity on a fiber for fiber basis
(although the range of sizes included in this count is not indicated).
Kodama et al. also report that, at 72 hrs, chrysotile (4 pg/cm2) caused a 2.7 fold
increase in binuclei and a 1.6 fold increase in micronuclei. Over the same time
interval, at 300 pg/cm2, crocidolite caused a 1.9 fold increase in binuclei but did
not cause micronuclei. They also report that chrysotile failed to produce
significant numerical chromosome changes in HBE cells and increased structural
aberrations only at the 24 hr time point. The frequency of neither aneuploidy or
polyploidy was increased at any time point following exposure to asbestos in this
study. The authors indicate that this contrasts with observations of relatively
high incidences of asbestos-induced chromosome changes observed in some
rodent cell cultures and clastogenic effects observed in human mesothelial cell
lines. They further speculate that phagocytic cells with high mitogenetic activity
are likely most susceptible to the effects of asbestos, which primarily interferes
with mitosis. However, they suggest that epithelial cells that are exposed to
fibers may undergo terminal differentiation (from Type II to Type I cells) and thus
cease mitosis. This would effectively prevent adverse genetic effects from
asbestos exposure. Such pathways are not available to mesothelial cells.
• Hart et al. (1994) studied the effects of a range of fiber types (with varying size
distributions) on Chinese hamster ovary (CHO) cells. The authors indicate that
such cells are very different from cells in pulmonary tissues in that they are
immortalized, aneuploid, undifferentiated, and preneoplasic. They also note that
the responses observed in these cells differs from responses observed in cells
from pulmonary tissues. Nevertheless, the implications concerning fiber size are
instructive. Fibers evaluated included long, medium, short, and UICC crocidolite
and chrysotile along with a range of manmade vitreous fibers (MMVF's),
refractory ceramic fibers (RCF's), and fibrous glasses. Exposure concentrations
ranged between 10 and 225 pg/cm2.
Results from the Hart et al. study indicate that all of the fibers caused .
qualitatively similar toxic effects: concentration-dependent reduction in cell
numbers and an increase in the incidence of abnormal nuclei with little or no loss
in viability. Fiber-induced cell death in CHO cells appears to be minor, even at
relatively high exposure concentrations. Based on mean dimensions (which is
problematic), the diameter dependence on the observed toxic effects, particularly
on the formation of aberrant nuclei, was slight or absent. However, the effect
7.86
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
with length was striking. For lengths up to at least 20 pm, potency toward both
cytotoxicity and the induction of aberrant nuclei increased dramatically with
increasing length. The authors also note that the lack of an observed fiber
composition associated effect on the toxicity of CHO cells does not correlate with
findings from recent rodent inhalation studies using the same test fibers. The
authors therefore speculate that CHO cells may not represent an appropriate in-
vitro model for fiber effects. However, it may also be that effects in vitro occur
over time scales that are rapid relative to those that occur in vivo so that
biodurabifity is not important in vitro.
• In a study by Takeuchi et al. (1999) cultured human mesothelioma cells
(MST0211H) and, separately, cultured humanpromyelocyte leukemia cells
(HL60), which are not phagocytic, were dosed with crocidolite between 0.6 and
6.6 pg/cm2 (no size range indicated). Studies with latex beads confirmed that
the mesothelioma cells are actively phagocytic but that the leukemia cells are
not. The authors indicate that dosed mesothelioma cells showed significantly
increased numbers of polynucleated cells, tetraploid cells, and cells with variable
DNA content at the G0VG1 transition in the cell cycle and that the extent of
effects was dose-dependent. Leukemia cells showed no such effects.
The authors further indicate that, when the mesothelioma cells were sorted by
fiber content, those with the highest fiber content showed the greatest effects.
The authors also indicate that cells are stimulated neither to release superoxide
nor NO at the concentrations of crocidolite studied. However, they hypothesize
that intracellular ROS may have been generated because they report finding
increased levels of 8-OH-dGua (and oxidized form of one of the DNA bases)
following crocidolite exposure. Nevertheless, the authors conclude that the
mechanisms by which crocidolite induce cytotoxicity and, potentially,
carcinogenicity is related to phagocytosis. Importantly, the effects described in
the Takeuchi et al. study are entirely consistent with effects attributable to
interference with mitosis, despite any speculation by the authors.
7.3.3.2 Generation of reactive oxygen species (ROS). ROS have been implicated
as mediators in a variety of toxic effects (including cancer initiation) associated with a
broad range of toxins (see, for example, Floyd 1990). Moreover, substantial evidence
indicates that asbestos can induce generation of ROS by several mechanisms and that
asbestos-induced ROS plays a role in several of the toxic effects attributable to
asbestos (see below). However, whether ROS plays an important role in asbestos-
associated cancer initiation is less clear and needs to be evaluated carefully.
Therefore, evidence that asbestos induces the production of ROS and that ROS
7.87
Revision 1-S/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
contributes to the adverse health affects attributable to asbestos are reviewed below
with particular attention to effects that may contribute to the initiation of cancer.
Contributions by ROS to other asbestos-related toxic effects are also evaluated in later
sections of this chapter.
Note, that generation of certain reactive nitrogen species (specifically the peroxinitrite
ion) is closely associated with ROS generation so that evidence for the generation of
certain reactive nitrogen species (Section 7.3.3.3) constitutes additional evidence for
the generation of ROS.
Asbestos-Induced Generation of ROS
Asbestos has been shown capable of generating a variety of reactive oxygen species
(ROS) including hydrogen peroxide (H202), superoxide (02*), and hydroxyl radical (OH-)
via several mechanisms (see, for example, Kamp et al. 1992, Jaurand 1997, Fubini
1997) including:
• catalytic production of superoxide from oxygen in aqueous solution;
• catalytic production hydrogen peroxide from oxygen in aqueous solution;
• catalytic production of hydroxyl radical by the Fenton reaction
(degradation of hydrogen peroxide catalyzed by iron on the surface of
fibers or mobilized from the surface of fibers);
• catalytic production of several ROS by redox cycling of iron on the surface
of fibers or mobilized from the surface of fibers (Haber-Weiss type
reactions);
• catalytic production of ROS by release of heme and heme protein from
various cellular components;
• by inducing release of various ROS species from phagocytes during
"frustrated" phagocytosis; and
• by binding to cell receptors, integrins for such receptors, or other features
of surface membranes that trigger signaling cascades mediating the
production and release of various ROS (and RNS).
For a general review of the chemistry involved in these processes, see Floyd (1990).
7.88
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Fertton, Haber-Weiss, and related reactions. Most of the evidence for Fenton and
Haber-Weiss reactions (and related free radical generating reactions) that take place on
the surface of asbestos fibers comes from experiments in cell-free systems (see below)
Therefore, their relevance to the conditions found in vivo may be limited. Moreover, the
size of the fibers (especially in terms of their cumulative surface area) and the history of
their surfaces (in terms of metal contaminants or coatings that might be present) may
substantially alter the effects of such experiments (Fubini 1997). Unfortunately,
however, few of the available studies report characterization of fiber sizes or surface
conditions in sufficient detail to judge the importance of these effects.
For fibers:
• Zalma et al. (1987) evaluated a range of fibers (UICC crocidolite, UICC amosite,
UICC Canadian and Zimbobwean chrysotile, industrial chrysotile, and magnetite)
for their ability to produce free radicals by the direct reduction of oxygen in
aqueous solution. In some cases, hydrogen peroxide was also added to the
solution. Results indicate that all of the fibers tested were able to generate
hydroxyl radicals (even in the absence of hydrogen peroxide) but that the
efficiency of production was a strong function of the activation (by grinding) or
pacification (by coating with benzene or other agents) of the fiber surface.
Chrysotile was reported to be the most efficient at generating radicals and the
authors assumed that this is due to iron contamination on the surface (since iron
is not a component of the "ideal" chrysotile fiber). However, such conclusions
are difficult to evaluate in the absence of simultaneous consideration of fiber
size.
• Governa et al. (1998) evaluated the ability of wollastonite fibers to generate ROS
in both a cell free system and a suspension of polymorphonucleocytes (PMN's).
The fibers were observed to produce ROS in both systems and that ground
wollastonite produced substantially more ROS than unground. The efficiency of
ROS generation in PMN suspension is also reported to be greater for
wollastonite than for either chrysotile or crocidolite (tested in previous studies).
However, no size information is given. Based on additional work with various
inhibitors added to the system, the authors also indicate that only a fraction of
the ROS generated was composed of hydroxyl radicals.
• Brown et al. 1998 subjected several different fiber types (amosite, silicon carbide
whiskers, RCF-1, and various fibrous glasses) to two standard chemical assays
for free-radical production (in cell-free systems). The authors indicate that, of the
fiber types tested, only amosite showed free radical activity significantly above
7.89
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
controls in both assays and only RCF-1 additionally showed significantly
elevated free radical activity in one assay. However, there is not enough
information provided in this study to determine whether the observed differences
are due to differences in fiber sizes, sample preparation (i.e. surface condition),
or fiber type. In apparent contrast, for example, Gold et al. (1997) report that
amosite and crocidolite produce few free radicals in cell free systems, unless
they are ground.
• Weitzman and Graceffa (1984) indicate that chrysotile, crocidolite, and amosite
are all capable of catalyzing the generation of hydroxyl radicals and superoxide
from hydrogen peroxide in vitro and that, based on experiments with various iron
chelators, these reactions are iron dependent. The authors further indicate that
hydrogen peroxide is produced in large quantities as a normal bi-product of
tissue metabolism, but that it is effectively scavenged by various enzymes. The
authors speculate that, by physically damaging cell membranes, asbestos may
allow release of the precursor hydrogen peroxide before it can be scavenged.
For particles:
• Silica, residual fly ash, and ambient air dusts also can create ROS in vitro and
the efficiency of production correlates with ionizable concentrations of various
transition metals complexed on the dust (Martin et al. 1997). In vivo, such metal
containing particles also cause release of ROS from macrophages (Martin et al.
1997). Additionally, binding of silica to plasma membranes of airway lung cells
and phagosomes provokes generation of ROS.
• Castronova et al. (1997) further indicate that it is the concentration of
contaminating iron on the surface of freshly fractured quartz that enhances free
radical production in aqueous solution (in cell free systems). These authors also
showed that high iron-containing (430 ppm) quartz dust inhaled by rats (at 20
mg/m3 for 5 hrs/day for 10 days) induced five times more leukocyte recruitment,
two times more lavageable red blood cells, 30 to 90% increase in macrophage
production of ROS, 71% increase in nitric oxide production by macrophages, and
38% increase in lipid peroxidation of lung tissue than observed in rats exposed
to low iron-containing (56 ppm) quartz.
Although iron is required for the reactions considered here, studies indicate that the iron
content of the fiber itself is not a good indicator of reactivity (Gold et al. 1997). Studies
also indicate that the iron that participates in these reactions need not originate with the
fiber (Jaurand 1997, Fubini 1997) and biological systems contain abundant sources of
7.90
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
iron. Therefore, both iron-containing fibers and iron-free fibers have been shown to
participate in these reactions in vivo.
Release of heme and heme protein. At least one research group (Rahman et al.
1997) indicates that heme and heme protein cause extensive DNA damage in the
presence of asbestos in vitro and, based on previous studies, that this may involve
heme catalyzed production of ROS following asbestos-induced release of heme from
cytochrome P-450, from prostaglandin H synthetase (or perhaps from other heme
containing proteins). Importantly, the authors indicate that such observations relate to a
nuclear pool of heme, which suggests that ROS generation via this mechanism may
occur in the immediate vicinity of DNA. The work by this group suggests at least one
additional pathway by which asbestos may induce the production of ROS and by which
ROS-mediated damage to DNA might occur.
Frustrated phagocytosis. Numerous studies indicate that long asbestos fibers (longer
than somewhere between 10 and 20 |Jm) cannot be efficiently phagoycitized by
macrophages (see, for example, Sections 4.4 and 7.2) and that macrophages that are
damaged by such "frustrated" phagocytosis release ROS (see, for example, Mossman
and Marsh 1991 or Kamp et al. 1992). Due to the differences in the size of
macrophages across species (see discussion of Krombach et al. 1997 in Section 4.4)
The minimum length beyond which phagocytosis may become frustrated may differ in
different animals. However, it is clearly longer fibers that contribute to this mechanism
for generating ROS. Shorter fibers (<10 ym) are unlikely to cause frustrated
phagocytosis in any of the mammalian species of potential interest.
• Lim et al. (1997) showed that alveolar macorphages in culture (after stimulation
with lippopolysaccharide) generated free radicals (ROS) when subsequently
exposed to chrysotile, crocidolite, or amosite (all UICC). They found chrysotile to
be the most potent inducer of free radical activity (which is not surprising given
that UICC chrysotile contains the highest fraction of long fibers of the UICC
samples tested (Berman unpublished). Based on tests with various inhibitors,
the authors indicated that the free radicals generated by the alveolar
macrophages occurred through a pathway mediated by protein tyrosine kinase,
phospholipase C, and protein kinase C and that the effects are dose-related.
• Kostyuk et al. (1998) cultured peritoneal macrophages and showed that
treatment with chrysotile asbestos (1 pg, no size data given) resulted in
production of frustrated phagocytosis and cell injury (the latter as evidenced by
release of lipid dehydrogenase - LDH, a marker for membrane damage). By
working with various chelators and flavonoids (natural plant products, some of
7.91
Revision 1 -8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
which quench superoxide and some of which chelate iron), the authors indicate
that cell injury was likely induced by superoxide and that the superoxide was
likely produced by an iron-dependent mechanism. Note that this contrasts with
the above studies that suggests production of radicals by frustrated phagocytosis
in culture is an iron-independent mechanism.
• At least for one kind of phagocyte: polymorphonucleocytes (PMN's), a study by
Ishizaki et al. (1997) suggests that crocidolite and erionite may induce production
of ROS from PMN's by each of two mechanisms. The first requires phagocytosis
and may represent the traditional, "frustrated" phagocytosis pathway indicated
above. The second pathway is triggered by an interaction between the fiber and
the cell surface and is mediated by NADPH. The authors also cite evidence that
chrysotile may similarly act through both of these pathways.
• Afaq et al. (1998) cultured alveolar macrophages and peripheral red blood cells
(RBC's) that were harvested from rats 30-days following a single 5 mg
intratracheal instillation of UICC crocidolite, UICC chrysotile, or ultrafine titanium
dioxide. The authors indicate that populations of alveolar macrophages were
significantly increased (over sham-exposed rats) for all three particle types and
that acid phosphatase and lipid dehydrogenase (LDH), which are markers of cell
membrane damage, were observed in cell-free lung lavage from animals
exposed to all three particle types. Both alveolar macrophages taken from
asbestos-exposed animals (both types) showed significantly elevated lipid
peroxidation and hydrogen peroxide production over titanium dioxide exposed
animals. However, the latter also showed elevated peroxidation and peroxide
production that were significantly elevated over sham-exposed animals. Similar
results were observed among RBC's from asbestos-exposed animals but not
from titanium dioxide-exposed animals. Note, it is possible that ROS production
induced by Ti02 occurs by a different mechanism (or set of mechanisms) than
that for asbestos (see, for example, Palekar et al. 1979, Section 7.3.4.4)
In-vivo evidence for asbestos-generated ROS. Several studies involving whole
animals also indicate that asbestos exposure induces the generation of ROS.
Importantly, in such studies, evidence for generation of ROS is generally determined
based on observation of the putative effects of ROS, rather than ROS directly.
• Ghio et al. (1998) intratracheal^ instilled 500 pg of crocidolite (NIEHS) into rats.
This was observed to induce a neutrophilic inflammatory response within 24
hours (in contrast to saline-exposed rats). The authors collected chloroform
extracts from exposed lungs and subjected them to electron spin residence
7.92
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
(ESR) spectrometry. Results indicate the presence of a carbon-centered radical
adduct that has a structure consistent with products of lipid-peroxidation. The
radical signal was only observed in asbestos-exposed animals and persisted
even after one-month following exposure. The authors also report that depletion
of neutrophils did not affect the signal and that dextrin-induced inflammation did
not produce the signal.
• Yamaguchi et al. (1999) studied effects in rat lung tissue at 1, 3, 5, 7, and 9 days
following a single intratracheal instillation of 2rrig of either glass fiber or UICC
crocidolite. The authors indicate significantly increased levels of B-OH-Guanine
(an oxidized form of the DNA base Guanine) one day after crocidolite instillation
and increasing repair activity for this oxidized form of guanine with time that
became significant at days 7 and 9 following instillation. Glass fibers (noted to
be non-fibrogenic and non-tumorigenic) did not produce either increases in 8-
OH-Gua or its repair activity. The effects associated with crocidolite were all
noted to be dose related.
Several of these studies also suggest distinctions in ROS generation (either the
absolute generation of ROS or generation of specific ROS species by specific tissues)
due to differences in fiber (or particle) size or type :
• Nehls et al. (1997) intratracheal^ instillled rats either with quartz (2.5 mg) or
corundum (2.5 mg). The latter mineral is reportedly non-tumorigenic. Results
indicate that lung epithelial cells in quartz exposed rats exhibited increased 8-
oxo-Guanine levels (a DNA adduct generated by reaction with ROS, see above)
that persisted for up to 90 days post exposure. Elevated levels of the DNA
adduct appeared in all cell types in all areas of the lung. The authors suggest
that the observed persistence of the elevated levels of 8-oxo-Guanine suggests
that it was produced at a rate in excess of the lung's capacity for repair. The
authors also report enhanced and persistent inflammation, cell proliferation, and
an increase in neutrophil population in bronchio-alveolar lavage (BAL) fluid and
an increase in in tumor necrosis factor alpha (TNF-a) in BAL fluid in the quartz
exposed animals. TNF-a is a cytokine linked to a variety of effects including the
recruitment of inflammatory cells (see Table 7-6) In contrast, exposure to
corundum produced none of the effects observed with quartz.
• Timblin et al. (1998b) report that ROS induced responses by rat lung epithelial
cells vary depending on whether exposure is to crocidolite, hydrogen peroxide,
or cadmium chloride (CdCl2). In response to ROS generation induced by the first
two agents increase levels of cJun protein (a protooncogene, Table 7S) and
7.93
Revision 1 -9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
crocidolite, but not hydrogen peroxide, causes elevation in the levels of
manganese-containing superoxide (MnSOD) dismutase (an enzyme that
catalyzes dismutation of superoxide, Table 7-6), Neither of these agents affect
levels of either of two common stress proteins (Table 7-6): GRP78 or HSP72/73,
nor do they affect cellular glutathione levels. In contrast, cadmium chloride does
not alter MnSOD levels but increases levels of GRP78 and HSP72/73 in addition
to cJun protein. Therefore, ROS-related mechanisms may be complex and may
be toxicant-specific. Thus, it may not be correct to assume that all fibers and
particles act through common ROS-related pathways.
In contrast to the results of the above study (which showed no affect on
glutathione levels), Golladay et al. (1997) showed that human lung epithelial cells
(cultured A549 cells) exposed to crocidolite (NIEHS sample) showed substantial
reduction in intracellular levels of glutathione (without increases in the oxidized
forms of glutathione). Rather, an associated increase in extracellular, reduced
glutathione was observed, suggesting that crocidolite induces release of
glutathione from the interior of these cells to the environment. The authors also
indicate that, given that the half-life for reduced glutathione outside of cells is on
the order of an hour, while extracellular reduced glutathione levels remained
elevated for more than 24 hrs following exposure, cells must have be releasing
reduced glutathione continuously. Because no concomitant release of lipid
dehygrogenase (LDH) or labeled adenine was observed (despite loading of cells
with labeled adenine prior to the experiment), the authors conclude that release
of glutathione is not associated with membrane disruption or apoptosis (which is
induced to some degree by exposure to crocidolite). Also, all of the effects
described above were similarly associated with exposure to de-ironized
crocidolite. Thus, the iron content of the fibers does not play a role in this
process.
Kaiglova et al. (1999) intrapleural^ injected rats with 10 mg of long amosite and
collected bronchio-alveolar lavage (BAL) fluid 24 hours following exposure and at
later time intervals. They indicate that total protein and alkaline phophatase (AP)
were both elevated in BAL 24 hrs after exposure and that AP remained elevated
for at least 3 mo following exposure. They also noted increased levels of lipid
peroxides in BAL at 24 hrs but not 3 mo following exposure. The authors
indicate that antioxidants were significantly decreased following exposure:
glutathione was significantly decreased in lung tissue at both 24 hrs and 3 mo
following exposure but was normal in BAL fluid at all time points; a-tokopherol
and retinol were significantly decreased at 3 mo in lung tissue; and ascorbic acid
was significantly decreased in both lung tissue and BAL at 24 hrs and remained
7.94
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
low at 3 mo. The authors indicate that decreases in antioxidants implies a role
for ROS (or RNS) in lung tissue injury. It is also possible that the varied
responses of specific antioxidants may suggest a role for toxin- and injury-
specific ROS/RNS.
Note that the apparent difference in the reported effect of crocidolite exposure on
intracellular glutathione levels in the above two studies might be due (individually or in
combination) to differences in the cell-types studied, differences in the size distribution
of the crocidolite employed, differences in study design, or other factors. Insufficient
information is available to distinguish among these possibilities.
Conclusions concerning generation of ROS. Except for frustrated phagocytosis
(which is unique to long fibers), ROS generation by the mechanisms discussed above
are generally considered a common response to respiration of particles in general (see,
for example, Martin et al. 1997 who suggest that ROS "...may be a global signaling
mechanism mediating response to particulate insult mostly by activation of kinases and
transcription factors common to many response genes." They further indicate that if the
load of ROS generated is to great, or the airway in which it is generated has been
previously impaired, "...these same mechanisms can result in deleterious respiratory
lesions and outright pathology"). However, not all non-fibrous particles are similarly
capable of inducing production of ROS. As indicated above, for example, while
crystalline silica is a potent inducer of ROS production, carundum is not (Nehls et al.
1997). Moreover, the spectrum of ROS species that are induced by particular toxicants
are generally specific to the offending toxicant (Timblin et al. 1998b). Therefore, the
generic grouping of ROS mediated pathways by particles and, especially, by particles
and fibers, does not appear justified. These mechanisms are more complex and
individualized than such generic grouping suggests.
ROS can be generated by multiple pathways that are variously dependent on particle
size, whether a particle is a fiber, fiber size, and particle or fiber type (i.e. chemical
composition). The primary mechanism(s) through which is ROS generated in response
to one type of particle or fiber may be very different than that through which ROS
generation is induced by another and the resulting suite of ROS may also differ.
Importantly, the relationship between dose and response for each mechanism may also
differ (see, for example, Palekar et al. 1979, Section 7.3.4.4).
Given the above, comparing among the ability of fibrous materials and non-fibrous
analogs to induce generation of ROS requires that such analogs be properly matched
before valid conclusions can be drawn. Thus, for example, the appropriate non-fibrous
analog for crocidolite is the massive habit of reibeckite and the appropriate analog for
7.95
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
chrysotile is the massive habit of antigorite or lizardite. Due to differences in both
chemistry and crystal structure, crystalline silica is not an appropriate non-fibrous
analog for any of the asbestos types. Moreover, because ROS can be generated by
different mechanisms, critical comparison across analogs requires more than the simple
confirmation that ROS is generated or even whether the relative efficiency with which
'ROS is generated is comparable. It is also necessary to contrast the specific
complexion of ROS generated and the specific tissue/cellular environments (i.e.
locations within a cell) in which they are generated in response to each analog.
It is also clear that ROS is generated by both iron-dependent and iron-independent
pathways. Even for the iron-dependent pathways, however, the source of iron need not
derive from the fibers or particles themselves. Therefore, since iron is abundant in vivo
(and the environment), both iron-containing and iron-free fibers (or particles) can
potentially participate in both the iron-independent and the iron-dependent pathways.
Effects Mediated by ROS
ROS have been implicated as mediators in a variety of toxic effects (including cancer
initiation) associated with a broad range of toxins (see, for example, Floyd 1990, Martin
et al. 1997). Cellular and tissue effects that have been associated with the effects of
ROS include:
• enhancement of overall uptake of particles by epithelium;
• stimulation of inflammatory responses;
• stimulation of various signaling cascades and production of cytokines;
• inducement of apoptosis;
• cytotoxicity;
• mediation of cell proliferation;
• formation of oxidized macromolecule (including DNA) adducts; and
• induction of DNA strand breaks.
However, only the last two of the above list of ROS effects are potentially relevant to
the initiation of cancer (the topic of this section). The other effects in the above list
likely contribute to the induction of other asbestos-related diseases and may even
promote (but not initiate) asbestos-related cancer. Thus, they are addressed further in
later sections of this chapter (see below).
Although, ROS generation is associated with exposure to various particles and fibers
(including all forms of asbestos), generation of ROS does not necessarily imply
carcinogenesis. For example, Zhu et al. (1998) indicate that ROS generation is
7.96
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
induced in response to exposure to asbestos, crystalline silica, and coal mine dust.
However, based on an extensive record of human exposure, the latter (coal mine dust)
is not carcinogenic in humans. Therefore, evidence related to the last two of ROS-
associated effects listed above, are examined in more detail below.
Several studies indicate that, once generated by exposure to asbestos, ROS can
interact with DNA in vitro and in vivo to produce oxygenated adducts, primarily 8-oxo-
Guanine. Of the studies reviewed above, for example, Yamaguchi et al. (1999)
observed that crocidolite but not (non-tumorigenic) glass produced dose-dependent
increases in 8-oxo-Guanine in rat lung tissue following intratracheal instillation. Also,
Leanderson et al. (1988), Park and Aust (1998), Keane et al. (1999) indicate that
asbestos induces formation of oxidized DNA adducts in in-vitro assays. Brown et al.
(1998) indicate that asbestos induces ROS-mediated DNA strand breaks.
However, not all DNA strand breaks attributable to asbestos occur through pathways
involving ROS. Ollikainen et al. (1999) exposed cultures of human mesothelial cells
(MeT-5A, transfected with SV-40 but nontumorigenic) to hyrdrogen peroxide (100 pM)
or crocidolite (2 - 4 pg/cm2), either alone or in combination with TNF-a and performed
assays for DNA strand breaks. The authors note that the concentrations of asbestos
evaluated are well below those that have been associated with cytotoxic effects.
Crocidolite alone was shown to produce DNA strand breaks at the concentrations
tested. The authors also note that, at lower concentrations, only reversible effects were
observed (presumably indicating DNA repair). Co-exposure to crocidolite and TNF-a
increased the observed incidence of DNA damage, but the effect was less than
additive. The authors indicate that additional studies with antioxidants indicate that the
DNA damage induced by crocidolite in this study occurs through a mechanism that
does not involve ROS. In fact, a potentially much more substantial mechanism by
which asbestos may induce DNA breaks and various clastogenic alterations involves its
ability to interfere with mitosis (Section 7.3.3.1).
There is also evidence that different tissues may respond to asbestos-induced ROS
generation differently. For example, Zhu et al. (1998) indicate that MnSOD is found in
the mitochondria of Type II epithelial cells of rats exposed to crocidolite. The authors
further indicate that, because fibroblasts, alveolar macrphages, or endothelial cells do
not display this protein when stimulated by exposure, this suggests a difference in the
susceptibility of epithelial cells to certain types of asbestos-induced injury.
Importantly, although these studies provide evidence that indicates asbestos is capable
of producing oxidized DNA adducts (or strand breaks) through ROS mediated
processes, they do not address the question of whether such adducts can lead to
7.97
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
heritable mutations in DNA in vivo nor do they indicate whether such adducts lead to
tumor production. Therefore, such studies should only be construed to suggest a
potential that asbestos can act as a cancer initiator through ROS-mediated pathways.
As previously indicated, not all mechanisms involving the generation or effects of ROS
are similarly fiber-size fiber-type dependent and those that are do not necessarily
depend on these variables in the same way. At this time, it is not possible to distinguish
among the relative importance of the different mechanisms so that it is difficult to judge
the importance of the relative effects. However, the overall general implication (with the
few exceptions noted) is that ROS more likely contributes to other asbestos-related
diseases and cancer promotion than cancer initiation.
One final note concerning the effects of ROS that specifically involves the behavior of
the hydroxyl radical is also warranted. The hydroxyl radical, is an extremely reactive
species. Whether in vitro or in vivo, this species will react with virtually 100% efficiency
with every organic molecule it encounters. Therefore the effects attributable to the
hydroxyl radical are limited to those involving reactions in the immediate vicinity of the
location at which it is generated. Thus, unless asbestos-induced generation of this
radical occurs within the nucleus and in the immediate vicinity of susceptible strands of
DNA, it is unlikely that these radicals are the direct cause of DNA damage.
Rather, hydroxyl radicals tend to react with cell membranes and other cellular
components to produce further intermediate radicals (primarily lipid peroxides), which
are much more stable than the hydroxyl radical and may migrate substantial distances
before having an effect. It is likely that these intermediate radicals are ultimately
responsible for any ROS-mediated DNA damage that may be attributed to asbestos.
7.3.3.3 Generation of reactive nitorgen species (RNS). Nitric oxide (NO) is
produced ubiquitously in biological systems and serves many functions (REF). it is
highly reactive and, therefore, generally a short-lived in vivo. However, nitric oxide has
been shown to react with superoxide (02 ) to form the peroxynitrite ion (ONOO") at near
diffusion-limited rates (see, for example, Zhu et al. 1998). This RNS may represent the
primary species responsible for the effects attributed to RNS.
Asbestos-induced Generation of RNS
Alveolar macrophages, lung endothelial and epithelial cells, and alveolar epithelium (in
both rats and humans), when stimulated by inflammatory agents generate both
superoxide and over-produce nitric oxide that then combine to produce the perxoynitrite
ion (Zhu et al. 1998, Martin et al. 1997). These cells up-regulate production of NO
7.98
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
when stimulated by various cytokines, lipopolysaccharide, and interferon Because
the peroxynitrite ion is a strong oxidant and nitrating agent and is extremely reactive,
evidence for its production is generally indicated in most studies by the presence of
nitrotyrosine, the stable product of tyrosine nitration, (Zhu et al. 1998). Evidence for
production of NO is frequently indicated by observation of nitrite. Numerous studies
also provide evidence of nitric oxide and peroxynitrite ion production specifically in
response to exposure to asbestos in various tissues.
• Both chrysotile and crocidolite up-regulate the production of nitric oxide by
alveolar macrophages in the presence of interferon-y and the interaction
between asbestos and interferon is synergistic. Non-fbrogenic carbonyl iron did
not induce nitric oxide formation (Zhu et al. 1998). These authors also cite a
study in which intratracheal instillation of silica and coal mine dust caused more
inflammation and nitric oxide formation than Ti02 or carbonyl iron (on an equal
particle basis) This suggests both the geometry and chemical composition of
particles determine their ability to up-regulate nitric oxide.
• Inhalation of chrysotile or crocdolite induce secretion of both TNF-a and nitric
oxide by pleural macrophages (Tanaka et al. 1998) Tanaka et al. (1998) studied
the effects of RNS in rats exposed by inhalation to 6 to 8 mg/m3 crocidolite or
chrysotile (both NIEHS samples) for 6 hrs/day, 5 days/wk for 2 wks. Rats were
sacrificed at 1 and 6 weeks following exposure. The authors indicate that
asbestos induces formation of stable products of nitric oxide in cells obtained by
lung lavage 1 wk after asbestos exposure. Nitrotyrosine (a marker for ONOO"
formation) was also observed. Also, a greater number of alveolar macrophages
and pleural macrophages were shown to express iNOS protein (the inducible
form of nitric oxide synthase, Table 7-6) than sham exposed animals. Exposed
rats showed significantly elevated immuno-staining for nitrotyrosine in the region
of thickened duct bifurcations as well as within bronchiolar epithelium, alveolar
macrophages, and mesothelial cells of both the visceral and parietal pleura.
Nitrotyrosine staining was persistent, being observed at both 1 and 6 wks
following exposure.
• Quinlin et al. (1998) studied the production of nitric oxide in rats exposed to
crocidolite or chrysotile asbestos (both NIESH reference samples). Rats were
exposed by inhalation at 6 hrs/day, 5 days/wk and lavaged at 3,9, and 20 days.
Lavaged macrophages showed significantly increased nitrite/nitrate (indicating
production of nitric oxide) and this was suppressed with inhibitors to iNOS.
Thus, nitric oxide is produced via an iNOS pathway. The authors also note that
nitric oxide production correlated temporally with neutrophil influx in the lavage
7.99
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
fluid. They also indicate that asbestos exposed animals showed a 3 to 4 fold
increase in iNOS positive macrophages in their lungs.
Quinlin et al. also exposed cultured murine alveolar macrophages (RAW 264.7
cells) to crocidolite, riebeckite, and crstobalite silica in vitro. These cells showed
increased iNOS mRNA following exposure to asbestos and even greater
increases if the cells were also stimulated with lipopolysaccharide (LPS). Both
crocidolite and riebeckite (but not cristobalite silica) stimulated increased iNOS
promoter activity when applied in combination with LPS. Thus, in this case, their
appears to be a mechanism that is sensitive to particle composition but not size.
• Park and Aust (1998) treated cultures of human lung epithelial (A549) cells with
crocidolite and observed induction of iNOS and reduction of intracellular
glutathione (GSH) levels. Based on studies with inhibitors, the authors further
indicate that iron mobilized from crocidolite was required both for formation of
nitric oxide and to generate 2'-deoxy-7-hydro-8-oxoguanosine but not for the
observed decrease in intracellular glutathione. The authors note that
approximately five times as much chrysotile (containing approximately 3% iron)
as crocidolite was required to produce the same level of nitric oxide formation.
Importantly, these experiments were conducted in vitro in serum-free medium, so
that no extra-biological source of iron was present.
• Choe et al. (1998) dosed cultured rat pleural mesothelial cells with either
chrysotile or crocidolite (both NIEHS samples) at concentrations between 1.05
and 8.4 pg/cm2 with or without co-stimulation with 50 ng/ml of interleukin-1{3 (IL-
1(3). The authors report that mRNA for iNOS in asbestos and IL-10 dosed cells
increased progressively from 2 to 12 hours following exposure. Both types of
asbestos also stimulated production of nitric oxide (measured as nitrite) in IL-13
stimulated cells in a dose- and time- dependent fashion. Both types of asbestos
also induced expression of iNOS protein and formation of nitrotyrosine (based on
nitrate detection) in IL stimulated cells. In contrast, carbonyl iron particles did not
induce any of the effects observed for asbestos in IL stimulated cells. Thus,
formation of RNS appears to be either fiber size dependent (not induced by
particles) or mineralogy-dependent (or both).
As with the production of ROS, RNS production is apparently a function of multiple,
complex mechanisms. Also, as with production of ROS, the dose-response
characteristics of the various mechanisms differ. Several of the mechanisms show a
strong dependence on fiber size and some dependence on fiber type. However, there
are other mechanisms that are dependent primarily on fiber (or particle) type, but may
7.100
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
not be dependent on size (or at least not dependent on fiber size). At this point in time,
it is not possible to gauge the relative importance of the various mechanisms by which
RNS may be generated, except that it is likely that the importance of the various
mechanisms likely differ in different cells and tissues and likely differ as a function of the
specific toxin whose presence is inducing RNS production.
There are also indications that production of RNS may be species specific. For
example, Jesch et al. (1997) report that alveolar macrophages harvested from rats
expressed iNOS when stimulated with either LPS (lippopolysaccharide) or interferon-y.
In contrast, iNOS expression could not be induced in hamster, monkey or human
macrophages.
Effects Mediated by RNS
Zhu et al. (1998), over-production of nitric oxide can:
inactivate critical enzymes;
cause DNA strand breaks that result in activation of poly-ADP-ribosyl
transferase (PARS);
inhibit both DNA and protein synthesis; and
form peroxynitrite by reaction with superoxide.
In turn, peroxynitrite ion may:
• initiate iron-dependent lipid peroxidation;
• oxidize thiols;
• damage the mitochondrial electron transport chain; and
• nitrate phenolics (including tyrosine).
Also, some of the damage to alveolar epithelium and pulmonary surfactant system
previously attributed to reactive oxygen species may actually be caused through ROS
generation of peroxynitrite. For example, Chao et al. (1996) report that crocidolite
treatment of human lung epithelial cells (A549 cells) results in formation of 8-hydroxy-2'-
deoxyguanosine (8-OHdG) in DNA and synthesis of mRNA for iNOS. An iNOS inhibitor
reduces intracellular nitrite and eliminates production of 8-OHdG. Addition of
independent NO donor, recovers production of 8-OhdG. Thus, production of the
oxygenated DNA adduct in this case appears to be generated by reaction with RNS.
As with ROS, it is primarily the potential for RNS to contribute to DNA damage {e.g.
strand breaks or generation of oxygenated adducts) that represent the primary
pathways by which RNS might participate in cancer initiation (as opposed to cancer
promotion or other asbestos-related diseases). It appears that RNS-mediated DNA
Based on
7.101
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
damage is closely associated with ROS generation and mediation of DNA damage (see
Section 7.3.3.2). Thus, there is little to add here.
7.3.3.4 Conclusions concerning asbestos as a cancer initiator.
The strongest, most consistent evidence that asbestos can act as a cancer initiator
relates to the tendency of asbestos to interfere with mitosis. Although there is evidence
that asbestos may induce production of DNA adducts and DNA strand breaks (through
ROS and RNS mediated pathways), whether such adducts or breaks ultimately lead to
permanent, heritable changes to DNA remain to be demonstrated. The relative
importance of the ROS/RNS mediated pathways compared to the pathway involving
interference with mitosis also remains to be determined. As indicated in later sections,
however, ROS/RNS mediated pathways may play substantial roles in cancer promotion
and induction of other asbestos-related diseases.
Regarding the primary mechanism by which asbestos may initiate cancer (interference
with mitosis), the pathway is length dependent (short fibers do not appear to contribute
to the effect). Further, although there may also be a dependence on fiber type
(chemical composition), it is apparent that all types of asbestos can act through this
pathway. The dependence of this pathway on fiber diameter is less clear. Thus, other
than to suggest that fibers must be respirable (and therefore thinner than approximately
0.7 pm to have an opportunity to act, Section 7.1), whether further diameter constraints
are associated specifically with the mechanism of cancer initiation is not known.
Importantly, although crystalline silica may also act to produce some of the same
effects as asbestos (including potentially carcinogenicity), there is substantial evidence
that this material does not act through the same pathways and that the characteristics
of its respective dose/response relationships may differ. Thus, for example, while
asbestos likely initiates cancer through a mechanism that favors long (and potentially
thin) fibers, silica more likely acts through a mechanism that is dependent on total
surface area, with freshly and finely ground material likely being the most potent. In
contrast, grinding asbestos fibers tends to lesson its carcinogenicity overall. Due to
differences in chemistry and crystallinity (reinforced by studies indicating a lack of
correspondence in behavior) crystalline silica does not appear to be an appropriate
analog for any of the asbestos types. Rather, for example, the appropriate non-fibrous
analog for crocidolite is riebeckite and the appropriate non-fibrous analog for chrysotile
is antigorite or lizardite.
There is also some evidence that the relative importance of asbestos as a cancer
initiator may differ in differing tissues. Thus, the results of Zhu et al. (1998) suggests
7.102
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
that Type II epithelial cells (but not alveolar macrophages, fibroblasts, or endothelial
cells) express MnSOD (manganese-dependent superoxide dismutase) in mitochondria
in response to exposure to asbestos, which may serve to mitigate effects associated
with ROS mediated mechanisms.
More importantly, the results reported by Kodama et al. (1993) indicate that human
bronchiolar epithelial cells are relatively resistant to the overall ability of asbestos to
induce genotoxic effects (including those associated with interference with mitosis) in
comparison, for example, to human meosthelial cells or various animal cells lines that
have been studied by others.
Further, Sanden et al. (1992) indicates from a prospective cohort study of shipyard
workers that the risk of lung cancer decreases with time since last exposure and the
authors argue that the time-dependence of disease is more consistent with asbestos as
a promoter than an initiator.
Taken as a whole, the above studies suggest that asbestos may serve primarily as a
promoter of lung cancer, rather than an initiator. In contrast, it appears that asbestos
may serve as both an initiator and a promoter of mesothelioma. Among other things,
this may suggest a reason for the observed difference in the time-development of these
two diseases among asbestos exposed individuals.
7.3.4 Evidence that Asbestos Acts As a Cancer Promoter
Primarily, asbestos may promote cancer by facilitating tissue proliferation. However,
additional mechanisms associated with the observed synergism between asbestos
exposure and smoking (see, for example, Hammond et al. 1979) also need to be
considered.
Substantial evidence exists indicating that asbestos induces proliferation in target
tissues associated with lung cancer and mesothelioma and this is summarized below
followed by an overview of studies that suggest the various mechanisms by which
asbestos may facilitate such proliferation. Evidence suggestinjg the various
mechanisms related to the synergism between smoking and asbestos exposure are
also briefly reviewed.
There are numerous mechanisms by which asbestos may facilitate proliferation
including:
• direct cell signaling to Induce proliferation. This may occur by:
7.103
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
- direct interactions between fibers and receptors on the cell surface;
- interaction with integrins for such receptors;
- interactions between phagocytized fibers and intracellular
components of signaling cascades;
- or interactions between cells and intermediate species (e.g. ROS
or RNS) whose generation and release has been induced by
asbestos; or
• response to induced cell death in target tissues, which then stimulates
stem cells to proliferate to replace killed cells. Cell death may be induced
either through:
- inducing apoptosis (programmed cell death); or
- direct cytotoxic effects, which leads to necrosis.
These pathways are summarized in Table 7-5, which provides a perspective on the
complexity of the interactions between asbestos and the cells and tissues of the body.
Regarding asbestos-induced cell death, asbestos-induced apoptosis (and all of the
other effects described above) typically occurs at exposure concentrations that are
much lower than required to induce frank cytotoxic effects (Sections 7.3.4.3 and
7.3.4.4). Therefore, it is primarily the former that is of potential interest in terms of
implications for asbestos-induced diseases in humans. Due to the high exposure
concentrations typically required, the importance of contributions from frank cytotoxicity
to human disease is unclear (Section 7.3.4.4).
7.3.4.1 Asbestos-induced proliferation. Numerous in-vivo studies indicate that all
types of asbestos induce proliferation in target tissues relevant to lung cancer and
mesothelioma. Such proliferation is also suggested by the animal histopathological
observations previously described (Section 7.2.2). Moreover, many of these studies
suggest that the underlying mechanisms may be fiber size- and fiber-type specific.
Responses may also be species-specific.
Importantly, there are some studies (primarily in-vitro studies) that suggest asbestos
acts to inhibit (rather than induce) proliferation. Although it is likely that the contrasting
observations found in such studies are due to differences in conditions, timing, dose, or
the type of asbestos employed, the underlying reason for the contrasting observations
is not always apparent.
The evidence for proliferation is summarized by tissue below.
7.104
Revision 1 - 0/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
In lung epithelium:
• Brody et al. (1997) showed that rats and mice exposed for a brief (5 hr) period to
chrysotile asbestos at 1000 fibers/cm3 (no size indicated) exhibited focal scarring
at bronchio-alveolar duct junctions that are identical to those seen in asbestos-
exposed humans. After 3-consecutive exposures, the lesions persisted for 6
months. In regions where fibers are deposited, macrophages are observed to
accumulate, epithelium is injured, and proliferation is observed to occur. In this
study, the authors also showed by immunohybridization staining that the four
genes required to express three peptide growth factors (TGF-a, TGF-fl, and the
A and B chains of PDGF) and the proteins themselves are expressed in
bronchio-alveolar tissue within 24 hrs of exposure. PDGF is expressed almost
immediately and expression remains elevated for 2-weeks post exposure, but
only in regions where fibers are deposited.
The authors report that PDGF is a potent growth factor for mesenchymal cells,
TGF-a is a potent mitogen for epithelial cells, and TGF-3 inhibits fibroblast
proliferation but stimulates synthesis of extra-cellular matrix (Table 7-6). The
authors also report additional experiments with knockout mice indicate that TNF-
a is required to induce the early stages of proliferation. The authors also
indicate that Type II cells produce TGF-01 and TGF-02 (two of three isoforms of
this protein) and that they are stimulated to do so when co-cultured with
macrophages.
• Adamson (1997) intratracheal^ instilled size-separated (by sedimentation) long
and short crocidolite fibers into rats (0.1 mg in a single dose) and noted that the
long fibers damaged the bronchiolar epithelium and that fibers were incorporated
into the resulting connective tissue; granulomas formed with giant cells
containing fibers). The long fibers also appeared to escape into the interstitium.
Labeled thymidine uptake (which indicates DNA synthesis and suggests
proliferation) following long fiber exposure was seen in lung epithelial cells,
fibroblasts, and pleural mesothelial cells. Such labeling peaked at 2% in
mesothelial cells and 3% in epithelial cells within one week following exposure.
Proliferation appeared to end shortly beyond one week. Short fibers were
observed to have been efficiently phagocytized by alveolar macrophages and a
small increase in macrophage population appeared to have been induced.
Otherwise, none of the other effects attributable to long fibers (described above)
were observed with short fibers. Note that such observations are entirely
consistent with those reported for the range of studies described in
Section 7.2.2.
7.105
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
• McGavran et al. (1989) exposed both normal (C5+) mice and compliment
deficient (C5-) mice to asbestos (at 4 mg/m3) by inhalation and observed
proliferation of bronchio-alveolar epithelium and interstitial cells at alveolar duct
bifurcations (based on incorporation of tritiated thymidine) between 19 and 72
hrs after a single, 5-hr exposure. Sham exposed rats showed fewer than 1% of
epithelial and interstitial cells at alveolar duct bifurcations incorporate labeled
thymidine. In contrast, thymidine uptake in asbestos exposed animals is
significantly elevated for the first few days, begins to decrease at 8 days, and
returns to normal by one month following exposure. Both C5+ and C5- mice
show similar increases in volume density of epithelial and interstitial cells at 48
hrs post exposure. However, one month following exposure C5+ mice
developed fibrotic lesions while C5- mice were no different than controls. The
authors conclude that the depressed macrophage response in C5- mice does
not appear to change the early mitogenic (and proliferative) response to
asbestos but apparently attenuates later fibrogenesis.
• Chang et al. (1989) describes the morphometric changes observed in rats
following 1 hr inhalation exposure to chrysotile (at 13 mg/m3). Within 48 hrs
following exposure: the volume of the epithelium increased by 78% and the
interstitium by 28% at alveolar duct bifurcations relative to sham-exposed
animals. Alveolar macrophages increased 10 fold and interstital macrophages 3
fold. Numbers of Type I and Type II epithelial cells increased by 82% and 29%,
respectively. At one month following exposure, the numbers of Type I and Type
II pneumocytes were still elevated, but not significantly. However, the volume of
the interstitium had increased by 67% accompanied by persistently high
numbers of interstitial macrophages, accumulation of myofibroblasts, smooth
muscle cells, and an increased volume of interstitial matrix.
In lung endothelium:
In addition to the general evidence for proliferation of endothelial cells provided by
Adamson (1997), McGavran et al. (1989), and Chang et al. (1989), as cited above, a
more detailed description of the nature of asbestos-induced proliferation of endothelial
cells is also available.
• Proliferation of endothelial cells and smooth muscle cells of arterioles and
venules near alveolar duct bifurcations is induced in rats inhaling chrysotile (no
size information given) for 5 hrs at 4 mg/m3 (McGavran et al. 1990). This is
based on the observed uptake of labeled thymidine (which indicates DNA
7.106
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
synthesis and suggests proliferation) that is significantly increased over controls
between 19 and 72 hrs following exposure. 28% of vessels near bifurcations
exhibited labeled cells 31 hrs after exposure. One month following exposure, the
thickness of the smooth muscle layers around these blood vessels is significantly
increased (doubled). In contrast, labeling of these same endothelial and smooth
muscle cells in sham exposed rats is zero. The authors indicate that endothelial
cells and smooth muscles associated with pulmonary blood vessels are normally
quiescent with turnover rates on the order of years.
In mesothelium:
• In the second part of the study addressing epithelial proliferation, Adamson
(1997) reports that rats instilled with 0.5 mg of unmodified, UJCC crocidolite were
sacrificed at one week and six weeks following exposure and subjected to
bronchiolar lavage and pleural cavity lavage. Lavaged alveolar macrophages
were observed to contain fibers but pleural macrophages did not. At one week,
collected pleural macrophages were shown to induce proliferation of fresh
mesothelial cells in culture and pleural lavage fluid showed an even greater
effect. No effects were observed at six weeks. Further work with anti-bodies to
various cytokines indicated that early, transient proliferation of mesothelial cells
was dependent on kertinocyte growth factor (KGF) but not on PDGF, FGF, or
TNF-a (Table 7-6). This suggests that early, transient proliferation is induced by
diffusing cytokines rather than direct fiber exposure. Adamson further reports
that KGF is a fibroblast-derived cytokine that acts on epithelial cells so that it's
up-regulation likely results from epithelial injury with penetration of asbestos to
the interstitium (where fibroblasts are found, Section 4.4). A similar transient
proliferative response in mesothelial cells was also observed following exposure
to chrysotile and in response to crystalline silica exposure. Thus, it appears that
the mesenchymal proliferative response may be mineralogy specific for particles
and is size-specific for fibers.
• Everitt et al. (1997) exposed rats and hamsters by inhalation to RCF-1 (45 mg/m3
- 650 f/cm3) for 12 weeks and then let them recover for up to an additional 12
weeks prior to sacrifice. The authors indicate that both rats and hamsters
showed qualitatively similar levels of inflammation at time examined (4 wks and
12 wks). They also indicate the mesothelial cell proliferation was observed in
both animal types but was more pronounced in hamsters at all time points
examined. The greatest proliferation in both species was in the parietal pleura
lining the diaphragm. The authors also report that fibers (primarily short and
thin) were also observed in the pleural cavities of both species at all time points.
7.107
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Several in vitro studies provide evidence that asbestos either induces or inhibits
proliferation in lung tissues of interest and that at least some of the mechanisms
involved are fiber size and mineralogy dependent. As previously indicated, the specific
reasons for the apparent contrast between results observed in vivo (where asbestos
consistently promotes proliferation) and the inhibition sometimes observed in in-vitro
studies is not always apparent. However, it must be due to the special conditions that
must be created to conduct in-vitro studies, which may not support certain mechanisms
that are important in vivo.
• Timblin et al. (1998a) completed a study in which rat pleural mesothelioma
(RPM) cells in culture were dosed with crocidolite or various cation-substituted
erionites. The expression of several gene and gene products were then tracked.
Cultures in this study were exposed to 1, 5 or 10 ug/cm2 of the various fibrous
materials. Analysis of the fibrous materials indicated that crocidolite contained
many more fibers per gram of material (probably because they are thinner) and
that the preparation contains somewhat longer fibers than the erionites
evaluated. In crocidolite, for example, 88% of the fibers are longer than 5 um,
68% longer than 10 um, and 37.5% longer than 20 um. All of the cation
substituted erionites showed approximately the same size distribution: 50%
longer than 5 um, 10-20% longer than 10 um, and 1-5% longer than 20 um.
Results indicate that the various cation substituted erionites behave differently
and that the Na substituted erionite shows the largest overall potency, at least for
some end points but not others. Fe, Na, and Ca substituted erionite all appear to
induce c-fos expression in a dose-dependent fashion (increasing regularly
among the 1, 5, and 10 ug applications). K-erionite may also show the same
pattern, but the changes were not indicated as significant over controls
(apparently due to greater variability). Only Na-erionite showed significantly
increased expression of c-jun (at 1 and 5 ug applications but not significantly at
10 |jg, apparently due to greater variability. However, the mean result for 10 ug
shows a consistent trend with the lower concentration application results).
Na-erionite induces c-fos at the same or greater rates as crocidolite asbestos for
the same mass application (but not the same fiber number). Crocidolite also
appears to show a dose-response trend for c-jun expression, but only the result
for the highest application (10 ug) is significantly different from controls. In
contrast, Na-erionite appears to show greater induction of c-jun expression at
lower dose than crocidolite, but the increase with increasing dose is much lower
for Na-erionite. Crocidolite also appears to induce substantial apoptosis (even at
7.108
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
the low dose of 5 ug) and that the induction is dose-dependent. In contrast, non-
fibrous riebeckite does not appear to induce apoptosis. Comparison between
crocidolite dosed cultures and Na-erionite dosed cultures indicate that crocidolite
induces substantial apoptosis at all time periods following application but that
Na-erionite induces little apoptosis even at higher mass dose and longer time
periods than crocidolite. The authors also indicate that crocidolite and Na-
erionite appear to stimulate DNA-synthesis, which appears to be a compensating
mechanism to fiber cell toxicity. The authors indicate that chemistry is important
in fiber toxicity as Na-erionite was a strong inducer of c-jun, even at relatively low
concentrations, but several of the other cation substituted erionites (including Fe-
erionite) were not. They further suggest that, given the difference in the fiber
lengths of the crocidolite samples and Na-erionite samples, that fiber length may
be a less important consideration than fiber surface chemistry. However, despite
the author's assertion, considering that non-fibrous riebeckite does not induce
any of the effects observed for crocidolite, there appears to be a clear size effect.
It may simply require that a defined, minimum length is necessary to induce the
effect.
The authors also indicate that balance between proliferation and apoptosis is
required to maintain homeostatis in healthy tissue. They further indicate that
other studies suggest that c-Jun expression is linked to proliferation and
induction of cancer, while c-fos expression is linked to apoptosis. Thus,
suppression of c-fos may be linked to carcinogenesis by allowing establishment
or maintenance of a transformed cellular phenotype. This is in fact an early step
in carcinogensis. Many environmental agents stimulate both apoptosis and
proliferation and, depending on the degree, may cause imbalances that lead to
disease. Relative stimulation of c-fos and c-jun may reflect some of these
pathways. Since crocidolite induces both c-fos and c-jun in this study, the
implication is that it mediates both apoptosis and proliferation.
• Wylie et al. (1997) dosed hamster tracheal epithelial (HTE) cells and rat pleural
mesothelial (RPM) cells with various asbestos and talc samples and evaluated
proliferation based on a colony-forming efficiency (CFE) assay. The samples
were NIEHS crocidolite and chrysotile and three different talcs. Samples were
characterized in the paper by mineralogical composition, surface area, and size
distributions.
The authors indicate that both asbestos samples increased colony formation of
HTE cells (suggesting induction of proliferation) but talc samples did not. RPM
cells, in contrast, showed only dose-dependent decreases in colony forming
7.109
Revision 1 ¦ 8/3101
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
efficiency for all samples, which the authors indicate is a sign of cytotoxicity. The
authors report that all samples show corresponding effects when concentrations
are expressed as fibers longer than 5 pm or by total surface area. They also
suggest that the "unique" proliferative response by HTE cells could not be
explained by either fiber dimension or surface area and suggests a mineralogical
effect. NOTE: MY OBSERVATIONS FROM RESULTS OF THIS STUDY ARE
THAT ABESTOS AFFECTS BOTH HTE AND RPM CELLS, BUT THAT TALC
ONLY APPEARS TO AFFECT RPM CELLS. EXPRESSING EXPOSURE AS
LONG FIBER NUMBER CONC OR SURFACE AREA TENDS TO PARTIALLY
BUT NOT COMPLETELY NORMALIZE DOSE-RESPONSE ACROSS SAMPLE
TYPES FOR RPM CELLS, BUT NOT FOR HTE CELLS. THE TWO
ASBESTOS SAMPLE TYPES BEHAVE SIMILARLY TWARD BOTH CELL
TYPES WHEN EXPOSURE IS EXPRESSED EITHER AS LONG FIBERS OR
SURFACE AREA (NEED TO GO BACK AND DEMONSTRATE THIS WITH
TABLES FROM THE PAPER.)
• Barchowsky et al. (1997) dosed cultured (low passage) endothelial cells to
NIEHS chrysotile, crocidolite, or RCF-1. After 1 to 3 hrs exposure to 5 pg/cm2
(non-lethal concentrations), asbestos (but not RCF-1) causes changes in cell
morphology (cells elongate), increases in cell motility, and increases in gene
expression. Further work by the authors indicate that these effects are mediated
by interaction between asbestos and the receptor for urokinase-type
pasminogen activator (uPAR). The authors also suggest that attachment of
asbestos to cell membranes, internalization of asbestos fibers by the cells, and
the morphological changes induced by asbestos are each mediated by different
proteins.
Examples of in vitro studies that indicate asbestos (and other fibrous materials) may
inhibit proliferation in culture include the following.
• In contrast to the above, Levresse et al. (1997) found that chrysotile and
crocidolite act to inhibit proliferation in cultured rat pleural mesothelioma (RPM)
cells. In this study, RPM cells (diploid, no more than 25 passages) were dosed
with either UICC crocidolite or NIEHS Zimbabwean Chrysotile at concentrations
varying between 0.5 and 20 pg/cm2. The authors also note that the chrysotile
sample contains approximately four times the number of fibers as the crocidolite
samples. Cells were then examined at 4, 24 and 48 hrs following treatment. In
untreated cultures, the number of cells in replicative phase decrease with time,
which indicates that such cells are headed for confluence (completion of a
monolayer on the culture medium). At 48 hrs, for example, 10% of untreated
7.110
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
control cells were observed to be in replicative stage. Chrysotile (but not
crocidolite) decreased the fraction of cells in replicative phase in a time- and
dose-dependent manner. At 48 hrs, for example, cells treated with 10 pg/cm2
chrysotile showed only 1.5% in replicative phase. Further tests confirmed that
this was due to blockage of cells at the G1/S boundary of the cell cycle. Both
chrysotile and crocidolite appears to induce a time-dependent increase in the
number of cells at G2/M in the cell cycle, although this effect was not observed to
be dose-dependent for chrysotile. Even on a fiber-number basis, chrysotile
appears to be elicit a greater response (arrest a greater percentage of cells) than
crocidolite.
The authors also indicate that chrysotile caused nuclear-localized, time-
dependent increases in p53 concentrations. Crocidolite produced much lower
levels that were not detectable in the nucleus. Chrysotile was also observed to
produce blockage at the G0/G1 transition of the cell cycle, but crocidolite did not
They also note that p53 is known to mediate arrest at this stage in the cycle so
observing that chrysotile induces arrest at this transition in the cell cycle may be
consistent with the observed increased expression of p53. The authors also
note that chrysotile triggers apoptosis in this study and that crocidolite shows a
smaller, but detectable effect. Spontaneous apoptosis in untreated cultures ran
between 0.5 and 1% at 24 to 48 hours whereas chrysotile induced 4% apoptosis,
peaking at 72 hrs following exposure to 10 pg/cm2. The authors indicate that the
lower level of effects observed with crocidolite could be due to the substantially
smaller number of long fibers in UICC crocidolite compared to NIEHS chrysotile.
• The previously reviewed (Section 7.3.3.1) study by Hart et al. (1994) also
suggests that long, medium, short, and UICC crocidolite and chrysotile along
with a range of MMVF's show a dose-dependent inhibition on proliferation of
cultured CHO cells and that potency toward the effect is a direct function of fiber
length.
Several of the above-described studies, in addition to providing evidence that asbestos
induces proliferation in various lung tissues, also suggests certain mechanisms.
Asbestos may induce proliferation, for example, by inducing production of specific
cytokine growth factors (Brody et al. 1997, Adamson 1997), or by inducing certain
signaling cascades (Barchowsky et al. 1997, Timblin et al. 1998b). It is also possible
that the two effects may be related (i.e. that stimulation of a particular signaling cascade
may result in production of certain growth-stimulating cytokines). Other mechanisms
may also be important (Table 7-5).
7.111
Revision 1 • 8/3/D1
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
7.3.4.2 Asbestos induced cell signaling. Asbestos has been shown to induce a
variety of cell signaling cascades in a variety of target cell and tissue types. Such
signaling may then trigger effects in the stimulated cells that may include:
• proliferation;
• morphological changes;
• generation and release of various cytokines, enzymes, or extracellular
matrix; or
• programmed cell death.
Note, due to the large number of chemical species that need to be considered in this
discussion, Table 7-6 provides a summary of the sources of such species (Table 7-6A)
and the effects attributable to such species (Table 7-6B).
In specific cases, asbestos may initiate cell signaling by interacting directly with
receptors on the cell surface, by interacting with integrins for such receptors, by causing
generation and release of intermediate species (e.g. ROS or RNS) that trigger cell
signaling, or (for phagocytized fibers) by interacting with intracellular components of a
particular signaling cascade. The specific responses to cell signaling induced by
asbestos are frequently cell- or tissue- type specific. Moreover, depending on the
specific mechanism, cell signaling by asbestos may be dependent on fiber size and/or
type.
• Barchowsky et al. (1998) showed in a set of studies that long chrysotile and long
crocidolite but not RCF-1 fibers (at concentrations between 1 and 10 pg/cm2),
which is reportedly below levels that typically induce cytotoxic effects) induced
up-regulation of urokinase-type plasminogen activator (uPA) and its receptor
(uPAR) in both lung endothelial cells (vascular cells) and lung epithelial cells.
They also showed that the increased pericellular protolytic activity (requiring
cleavage of plasminogen to plasmin) that is induced by asbestos in these cells is
mediated by uPA.
In prior studies, chrysotile has been shown to cause endothelial cells to elongate and
increase expression of adhesion molecules for phagocytes. They also show enhanced
proteolytic activity and matrix interactions. Chrysotile has also been shown to stimulate
fibrinolytic activity in lung epithelial cells and extravasating macrophages. All such
stimulation appears to result from up-regulation of uPA. Thus, this mechanism may
explain the observed asbestos-induced changes in lung endothelial and epithelial cells
including vascular remodeling, development of vascularized granular tissue, increased
matrix turnover, and leukocyte extravasation (which in turn may be caused by cell
7.112
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
activation and elaboration of proteases and adhesion molecules). The authors suggest
that asbestos-induced up-regulation of uPA and uPAR expression may represent a
global mechanism for pulmonary toxicity and fibrosis induced by crystalline fibers.
Importantly, chrysotile was shown to Induce uPA and uPAR expression in the absence
of serum, so the effect is apparently due either to direct binding of fibers to cell-surface
receptors or integrins for such receptors.
• Mossman et al. (1997) observed that concentrations of 1.25 to 5 pg/cm2 of
crodidolite (sample from TIMA) caused expression of c-jun and AP-1 in both
cultured hamster tracheal epithelial (HTE) cells and cultured rat pleural
mesothelial (RPM) cells. Crocidolite was also observed to trigger the EGFR-
regulated kinase (ERK) and mitogen activated protein kinase (MAPK) pathways
in RPM cells. The authors indicate that these pathways are also stimulated by
hydrogen peroxide and that NF-k0 induction stimulated by crocidolite is also
stimulated by crystalline silica (although silica stimulation of the MAPK pathway
was not investigated). They also note that the non-fibrous analog to crocidolite,
riebeckite does not elicit these activities.
7.113
Revision 1 - B/3A)1
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 7-6A, page 1
7.114
Revision 1 - 8/3/01
-------
Chemical Species
Signal Transmitters
Activator Protein-1
Table 7-6A:
Sources of Various Cytokines and Other Chemical Transmitters
Abbrevlatl Sources
Fifth Component of Compliment
Cyclln Dependent Kinases
CycHn Dependent Kinase Inhibitor!
Cytokine-induced Neutrophil Chamoattractant
Epithelial Growth Factor
AP-1
BAX
Bd-2
C5
COK-a
CDI'a
CINC
EGF
induced fay high cones of p53
Reference
Lechnarata. 1997
Epithelial Growth Factor Receptor
EGFR regulated kinase
Granutocy te-macroph age colony Stimulating Factor
human neutrophil eiastase
Intercellular adhesion molecule-1
lnterleukln-1
lnterteukln-6
lnterleukln-8
Inducible Nitric Oxtd Synthase
Keritonizing Growth F actor
Mitogen Activated Protein Kinase
EGFR
ERK
GM-CSF
GRP78
HNE
HSP72/73
ICAM-1
IL-1
IL-6
IL-8
INOS
K.GF
MAPK
released from asbestos activated PMN
Kamp at al. 1903
released by king epithelial and endothelial oals following exposure lo silica but not Ti02 Nario and Hubbard 1S97
enhanced expression by RPM oals expowd to ctvrys or etc Choeet al. 1997
induced by ROS in Type II epithelial oeNs
expressed by activated macrophagea, but not in primates
derived from fibroblasts???
Luster and Skneonova 1998
Jeachetal. 1997
Adamaon 1997
-------
MVMafeha
Macrophage Inflammatory Pratein-2 MIP-2
Matrix MetaNoprotelnases
Nuclear Factor-KB
Nitric Oxide Synthase
Ornithine Decarboxylase
Platelet-Derived Growth Factor
MMPa
Neu
NF-KB
NOS
ode
PDGF
relative expression affected by dgaret smoke and fiber exposure
Morlmotoetel. 1997
Activated macrophages
Elevated expression in mesothelioma oells
asbestos and iron treated AM and Interstitial macrophages release PDGF
Baumanetal. 1990
Lechneretal. 1997
Osomto-Vargaa at al 19B3
poty(AOP-ribosyl)polymera8e
????
Tissue inhibitors of MMP's
Tumor Necrosis Factor Alpha
pS3
Rb
TIMP's
Induced by DNA damage
Inhibited by 3amlnobenzamide
relative expression affecte Morimoto et al. 1997
Lechneretal. 1997
Broaddua et al. 1997
TNF-alpha some particles Induce activated macrophages to release this (shown both In vitro and in viv DriacoH et al 1997
regulated by oxldant-sensitlve transcription of NF-KB Drlscoll et el. 1997
macrophages in BAl from Babesiosis and Idiopathic Interstitial fibrosis patients release TNI Zhang et el. 1993
Transforming Growth Factor Alpha
Transforming Growth Factor Beta
TGF-alpha expressed in BA epithelium exposed to asbestos
Brody et al. 1997
TGF-Beta expressed in BA epithelium exposed to asbestos Brody et al. 1997
macrphage stimulated Type II epi cells produce both TGF-beta1 and TGF-beta2 Brody et al. 1997
all three Isoforms expressed in the flbrotic lesions (by hyperplaUc type II cells) of asbestoslt Jaglirdlr et al. 1997
mesothelial tumor cells expressed only TGF-beta1 and 2 Jaglirdlr et al. 1997
also type II oells of silicosis patients express al three forms Jagilrdlr et al. 1997
asbestos and Iron treated AM and Interstitial macrophages release TGF-toeta Osornto-Vargas et al. 1903
»duniw type Plasminogen Activator
uPA
-------
Vascular Cell Adhesion Molecule VCAM-1 enhanced expression by RPM cells exposed to chrys Of etc ChoeeOL 1997
WT1
Enzymes
Mn-dependent Superoxide Olsmutaae Mn-SOO
Gene Products
c-fos
c-jun
mdm2
c-myc
H-RAS
K-RAS
-------
Chemical Species
Signal Transmitters
Activator Protein-1
Table 7-6B:
Effects of Various Cytokines and Other Chemical Transmiffers
Abbrevlat Effects
Fifth Component of Compliment
Cyclln Dependent Kinases
Cydln Dependent Kinase Inhibitors
Cytoklne-lnduoed Neutrophil Chemoattradant
Epithelial Growth Factor
Epithelial Growth Factor Receptor
EGFR regulated kinase
Granulocyte-macrophage colony Stimulating Factor
human neutrophil elastase
knferoelltilar adhesion molecule-1
AP-1
BAX
Bd-2
C5
CDICs
CDrs
CINC
EGF
EGFR
ERK
GM-CSF
GRP78
HNE
HSP72/73
ICAM-1
a transcription factor:
binds the DNA promoter region of genes governing Inflammation, proliferate, and apoptosls
Induces apoptosls
Inhibits apoptosls
mediates asbeetoslnduoed llbrosla
mediates cell cyde
Inhibits advance of cell cyde
Inhibits proliferation of mesothellal oells
stimulates proliferation of mesothelioma tumor oells
mediates MAPK signaling
mediates apoptosls
By Itself. Increases pulmonary epithelial cell detachment In culture
with asbestos, Increases pulmonary epithelial cell lysis
fadlitalea migration at leukocytes out of blood to sites of Injury
Reference
Mossman and Churn 19S8 (Citing
Lechner et al. 1097
Lechneret al. 1997
MoGavran et al. 1989
Lechneret al. 1997
Lechneret al. 1997
Kravchenko et al. 1996
Goodgllck et al. 1697
Mossman et al. 1997
Mossman et al. 1997
Kamp et al. 1993
Kampet al. 1993
Narto and Hubbard 1997
kiterieukln-1
MerieuWfvfl
MerieukirvO
Inducible NHric Otdd Synthase
KerWonblng Growth Factor
Mitogen Activated Protein Kinase
Macrophage Mammatory Protein-1 alpha
Macrophage Iniammetory PraMn-2
IL-1
1-6
H.-8
INOS
KGF
MAPK
Is a plelotroplc cytokine
Is a neutrophil chemoattradant
Induces transient proliferation of mesrthellal cells
mediates transcription of AP-1
mediates ERK signaling
MtP-1 alpha
Mf>-2 a chemoattradant far PMNs. but may not be Involved with Inflammatory response
Luster and Slmeonova 1998
Luster and Slmeonova 1998
Adamson 1997
Mossman el al. 1997
Mossman et al. 1997
Osier et a! 1997
-------
Matrix Metalloproteinases
MAPs
Nudear Fador-KB
Nitric Cbdde Synthase
Ornithine Decarboxylase
Platelet-Derived Growth Factor
Neu
NF-KB
NOS
ode
POGF
a transcription factor that regulates genes Involved with the Inflammatory response, cell adhesion, and growth c Barchowsky et al. 1098
promoted by AP-1
encodes a key enzyme for tie biosynthesis of pdyarrines
Induoes proliferation of lung fibroblasts
Induce proliferation of mesothdlal cells
Is a chemotactlc attradant for rat lung fibroblasts
Mossman and Chuig 1998
Mossman and Chug 1998
Baumanetal. 1990
Brady etal. 1997
Osomlo-Vargas et al. 1993
polyCADP-rlbosyi polymerase
????
Tissue Inhibitors of MMPs
Tumor Neaosls Factor Alpha
Transforming Growth Factor Alpha
Transforming Growth Factor Beta
pS3
Rb
TINTS
TNF-alpha
at low oonc. arrests cell cyde at G1/S
at higher oonc, Induoes BAX protein, which induoes apoptosls
????
Untried etal 1997
777?
Induces JNK aim of MAPK cascade Mossman etal. 1997
Is Involved In the recruitment of Inflammatory cells Drtscolt et al. 1997
stimulates maaphages, epithelial oells.endctfieilal cells, fibroblasts to secrete chemoWnes and adhesion motec DtlsooH et al. 1997
can Induce apoptosls among neutrophils Leigh et al. 1997
can Induce ROS production In leukocytes Kalglova etal. 1997
Intradermal InJ stimulates focal accumulation of fibroblasts and collagen Zhang el at. 1993
In vitro, stimulates fibroblast ONA synthesis and proliferation Zhang et al. 1993
TGF-alpha Is a mitogen for epithelial cells
TGF-Beta Inhibits fibroblast proliferation but stimulates formation of ertraoellular matrix
can Induos apoptosls In different types of cells
does not appear to Induce chemolaxia of rat lung fibroblasts
Induces chemctaids of rat mononuclear leukocytes
Brodyetal. 1997
Brady el al. 1997
Leigh el al. 1997
Osomlo-Vargas et al. 1993
Osomlo-Vargas et al. 1993
Urokinase-type Plasminogen Activator
Vascular Cell Adhesion Molecule
uPA
VCAM-1
WT1
assodated with Increased pericellular protydytic activity In endothelltf tissue
forms a heteradimer w p53 and alters behavior
Barchowsky et al. 1996
Untried et al 1997
Mn-dependent Superoxide Dismutase
Mn-SOD
-------
Gene Products
c-fos linked to apoptosta TtmbllnWal. 1898
transcription factor that activates the TGF-Mai promotor Jaglrdar at si. 1697
prOMr\» of c-fae dtmerize wKth c-Jun to create activator protein-1 (AP-1) Mossmao and Chutg 1089
c-)un HnVed 1o proliferation Timbtln et el. 199B
transcription factor that activates the TGF-betal promotor Jaglrder at at. 1097
transcription facta Rut activates the TGF-betal promotor Moavran and Churn 199B
mdm2 a protococoncooene that Inhibits p53 Lechneretal. 1997
omyc
H-RAS
K-RAS
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 7-6B, Page 3
7.119
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
The authors indicate that induction of the NF-Kp cascade was inhibited by
excess glutathione, which is stimulated by N-acetylcysteine (NAC), suggesting
that this pathway is induced by asbestos-caused oxidative stress (perhaps
through ROS or RNS intermediates). Application of NAC also diminished
crocidolite induced c-fos and c-jun RNA levels and inhibited activation of the
ERK-MAPK cascade. Further work suggests that asbestos triggers the MAPK
pathway by interaction with the Epithelial Growth Factor Receptor (EGFR), either
directly or by phosphorylation of this receptor by ROS. At the concentrations
examined, crocidolite induces substantial apoptosis (apparently through
activation of the ERK-MAPK cascade). In contrast, the authors note that TNF-a
induces the JNK arm of the ERK-MAPK cascade, which leads to proliferation.
Asbestos does not elevate JNK over the time-period of the study. The authors
note that it has been shown in some studies that inhibition of ERK in some cells,
also inhibits asbestos-induced apoptosis. Importantly, given that these
processes are induced by crocidolite but not by its non-fibrous analog, riebeckite,
induction of the ERK-MAPK cascade appears to be a fiber-size dependent
process.
• In a related study to that conducted by Mossman et al. (1997), Zanella et al.
(1999) report that the (TIMA) crocidolite (at concentrations of 2.5 to 10 ug/cm2,
but not its non-fibrous analog, riebeckite, eliminated binding of EGF to its
receptor EGFR. Because EGF does not bind to crocidolite in the absence of
membrane, this is not simply a case of crocidolite tying up ligand. Crocidolite
also induces a greater than twofold increase in steady-state message and
protetin levels of EGFR.
The authors also note that the tyrphostin, AG-1478 (which specifically inhibits the
tyrosine kinase activity of EGFR), significantly mitigated asbstos-induced
increases in mRNA levels of c-fos but not c-jun and that the asbestos action was
not blocked by a non-specific tyrphostin, AG-10. Moreover, pretreatment of RPM
cells with AG1478 significantly reduced asbestos-induced apoptosis. Therefore,
the authors concluded that asbestos -induced binding to EGFR initiates signaling
pathways responsible for increased expression of the protooncogene c-fos and
the development of apoptosis. This apparently occurs through the EGFR-
extracellular signaling regulated kinase (ERK). It is hypothesized that asbestos
may induce dimerization and activation (phosphorylation) of EGFR, which also
prevents binding of EGF. Asbestos apparently serves the same role as EGF in
that it promotes aggregation of EGFR, which in turn promotes binding to the
extracellular domain of tyrosine kinase receptors and the activation of their
intracellular kinases. The authors also indicate that other work suggests that
7.120
Revision 1 • 8/3fl)1
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
crocidolite fiber exposure leads to aggregation and accumulation of EGFR at
sites of fiber contact and that asbestos also stimulates biosynthesis of the EGFR
and activates ERK in an EGFR-dependent manner.
The authors speculate that asbestos binding may not be ligand-site specific but
may be charge related or may induce EGFR phophorylation by local production
of ROS, which has previously been demonstrated to cause EGFR activation. As
previously indicated, that this effect is driven by crocidolite exposure but not by
exposure to riebeckite indicates that this mechanism is fiber-size specific.
Johnson et al. (1997) showed that UICC crocidolite, but not JM-100 glass,
applied to a culture of immortalized human Type II epithelial (A549) cells at non-
cytotoxic concentrations (for 20 hours) results in increased expression of p53,
Cip1, and GADD153 in a dose- and time-dependent fashion. Expression was
observed to be maximum at 18 hours. The crocidolite treatment was also shown
to cause an increase in the number of cells arrested in Stage G2 of the cell cycle
(with a persistent decrease in the number of cells in G1). This was considered
surprising because both p53 and Cip1 are known to mediate arrest in Stage G1.
The authors suggest that these findings indicate a strong dependence on both
fiber type and fiber size (JM-100 glass contains substantially more long fibers
than crocidolite). However, it is not possible to separate the effects of fiber type
versus fiber size in this study.
Luster and Simeonova (1998) indicate that at high concentrations, ROS may
induce frank cytotoxicity. At low or moderate levels, ROS is more likely to induce
cell signaling cascades that may, in turn, contribute to asbestos-related disease.
The authors dosed cultures of immortalized human Type II epithelial (A549)
cells, originally derived from a lung carcinoma, and normal hufnan
bronchioepithelial (NHBE) cells with long (Certain-Teed supplied) crocidolite
(reported mean length: 19 pm) at concentrations ranging between 0 and
24 jjg/ml. Results indicate that secretion of both lnterluken-8 (IL-8) and IL-6 was
stimulated by crocidolite exposure in a dose-dependent manner. In contrast,
increases in LDH levels (which indicate cell damage) was only detected at the
highest exposure concentrations tested. Further work indicates that stimulation
of IL-6 and IL-8 secretion occurs through ROS that is generated in an iron-
dependent process (that may also include NF-k(3 induction). Note that the trend
of cell signal induction at low and moderate levels of asbestos exposure with
evidence for cytotoxicity observed only at the highest exposure concentrations is
common to many of these kinds of studies.
7.121
RevWon 1-8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Choe et al. (1999) conducted a combined in-vivo/in-vitro study of the effects of
low level exposure to chrysotile or crocidolite at inducing leukocyte attachment to
rat pleural mesothelial cells. The authors note that similar populations of rat
pleural leukocytes (74% macrophages, 2% neutrophils, 10% mast cells, and
10% eosinohils) were observed in both asbestos-exposed and unexposed rats.
In the second part of the study, cultured RPM cells were exposed to either
crocidolite or chrysotile (both NIEHS samples) at concentrations ranging
between 1.25 and 10 pg/cm2, which was noted to be below levels at which
substantial cytotoxicity is observed. Attachment of rat pleural leukocytes to RPM
cells was then observed to increase with increasing dose of asbestos to the RPM
cells. In contrast, carbonyl iron (a non-fibrous particle) also induced enhanced
attachment but at much lower levels and the effect was not dose-dependent.
Further analysis indicated that asbestos-induced adhesion is mediated by up-
regulation of IL-1(3 (but not dependent on TNF-a or nitric oxide production,
although it is noted that TNF-a independently increases attachment). Asbestos
also induces increased expression of vacular cell adhesion molecule (VCAM-1).
The authors also note that rat pleural leukocytes harvested from asbestos-
exposed rats also showed increased adhesion to in-vitro RPM cells over
leukocytes harvested from sham exposed rats. Thus, asbestos appears to
trigger alterations in these cells as well.
In a recent paper, Driscoll et al. (1997) reviewed the role of tumor necrosis factor
alpha (TNF-a) in mediating the inflammatory response to lung insult by
particulate matter. The authors indicate that quartz, coal dust, crocidolite, and
chrysotile are all potent inducers of TNF-a production. Titanium dioxide (Ti02),
corundum (aluminum oxide: Al203), and latex beads are not. Pulmonary
macrophages have been shown to secrete TNF-a in vivo in response to
exposure to some of the dusts listed above (including asbestos). This is ailso
observed among macrophages from asbestosis patients.
The authors indicate that rats immunized against TNF-a show reduced
recruitment of neutrophils, which demonstrates that TNF-a is involved In the
recruitment of inflammatory cells. TNF-a stimulates macrophages, epithelial
cells, endothelial cells, and fibroblasts to release chemokines that include
adhesion molecules (Eselectin, ICAM, VCAM). Inflammatory cells then interact
with such molecules and migrate along gradients from vascular structures in the
lung to the lung interstitium and even the lung air spaces. It has also been
shown that release of TNF-a by macrophages is apparently dependent on
oxygen stress (i.e. exposure to ROS/RNS) that is induced in pathways that
7.122
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
require iron. Production of TNF-a and other compounds that mediate the
inflammatory response are regulated by the oxidant-sensitive transcription factor
NF-Kp. This factor exists as a heterodimer in the cytoplasm in an inactive form
because it is bound to the inhibitory 1-kappaBalpha (1-KBa), which masks a
nuclear translocation signal. Appropriate stimulation of a cell induces
phosphorylation of 1-KBa, which then marks it for proteolytic degradation. Then,
NF-kP translocates to the nucleus and induces transcription. The process is
stimulated by oxidants and inhibited by antioxidants.
In another review, Finkelstein et al. (1997) indicated that Type II epithelial cells
and Clara cells (non-ciliated bronchiolar epithelial cells) respond to and produce
specific cytokines during the inflammatory process. Early responses to particle
challenge include increases in mRNA and protein for IL-1{3, IL-6, and TNF-
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
the Quebec chrysotile. The authors also indicate that the hyperplastic epithelium
of silicosis patients also show elevated expression of all three isoforms. They
further indicate from previous studies that mesothelioma tumor cells frequently
express TGF-02 while the cells in the stroma of such tumors frequently express
TGF-31. It is also noted in this study that jun and fos are both transcription
factors that activate the TGF-P1 promoter.
• Zhang et al. (1993) indicate that macrophages obtained in BAL fluid from
idiopathic pulmonary fibrosis (IPF) patients and asbestosis patients show
significantly increased secretion of TNF-a and asbestosis patients also showed
significantly increased secretion of IL-1 p. Macrophages and monocytes
obtained from both kinds of patients also show elevated expression of mRNAfor
these cytokines. In an in-vitro part of this study, Zhang and coworkers, showed
that chrysotile, crocidolite, amosite, and crystalline silica all stimulated IL-13 and
TNF-a release and up-regulated their respective mRNA in both macrophages
and monocytes. The authors also report that these two cytokines have been
shown to up-regulate collagen Types I and III and fibronectin gene expression in
human diploid lung fibroblasts after short term, serum free exposure in vitro.
• Holian et al. (1997) exposed cultures of normal human alveolar macrophages
(AM) (obtained by lavage) to varying concentrations (up to 25 pg/ml) of short
chrysotile, UICC crocidolite, ground silica, wollastonite, and titanium dioxide to
determine whether these materials cause a phenotypic shift in macrophage
populations by inducing selective apoptosis. The authors indicate that normal
lungs contain: 40-50% RFD1+7+ suppressor AM, and 5-10% RFD1+ immune
activator. In this study, the fibrogenic subset of the particles tested (not
wollastonite or titanium dioxide) increased the ratio of activator/suppressor AM
by a factor of 4 within a few hours and the effect was seen to increase with time.
The authors also note that fibrogenic particles decrease the abundance of
RFD7+ AM (phagocytic), but the consequences of this phenotypic shift are
unclear.
The authors indicate that AM taken from fibrotic patients release a variety of
proinflammatory mediators capable of stimulating fibroblast proliferation and
collagen synthesis. Even in the absence of evidence of fibrosis, workers who
have been heavily exposed to asbestos yield similarly activated AM. In contrast,
they also note that, in vitro studies in which AM are stimulated with fibrogenic
particles, while such AM are activated to release inflammatory cytokines, such
releases are orders of magnitude less than that seen from AM derived from
7.124
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
fibrotic patients. The authors indicate that the apoptosis-driven phenotypic shift
in AM that is indicated by this study may explain the apparent discrepancy.
• In a previously described study, Timblin et al. (1998a), see Section 7.3.4.1,
showed that crocidolite asbestos and several cation substituted erionites all
stimulate c-jun and c-fos in rat pleural mesothelial cells, but to varying degrees
depending on fiber chemistry. The effect also appears to be size dependent as
crocidolite but not its non-fibrous analog riebeckite induces the effect.
In general, the kinds of signaling cascades that are potentially stimulated by exposure
to asbestos are important due to their potential to contribute to the promotion of cancer.
Such pathways, for example, may mediate proliferation or may suppress apoptosis.
Alternately, they may mediate an inflammatory response that in turn may lead to
proliferation or to production and release of other, mutagenic agents (e.g. ROS or
RNS). Pathways that facilitate development of fibrosis may also contribute to cancer
promotion, given the apparent link between fibrosis and the development of lung
cancer, which may relate (among other possibilities) to inhibition of fiber clearance
(Section 7.3.4.5).
The ERK-MAPK signaling pathway evaluated in multiple studies by Mossman and
coworkers (Mossman et al. 1997, Timblin et al. 1998a and b, Zanella et al. 1999) in rat
pleural mesothelial (RPM) cells is a case in point (see above). These studies suggest
that crocidolite stimulates the ERK-MAPK cascade through interaction with the EGF
receptor. This ultimately leads to transcription of mRNA for c-fos; Crocidolite has also
been shown to induce c-jun (apparently through a separate mechanism) and the
balance between c-jun and c-fos has been implicated in guiding a cell toward either
proliferation or apoptosis. Although the direct connection between c-jun/c-fos and
apoptosis has not been established, it is observed that crocidolite induces substantial
apoptosis in RPM cells at the same concentrations at which it induces substantial
expression of c-fos and c-jun. The link is also implied because inhibition of the ERK
pathway has been shown in some studies to inhibit asbestos-induced apoptosis. Na-
erionite has also been shown to induce c-fos at levels comparable or higher than
crocidolite for comparable exposures (at least on a mass basis) and induce c-jun at
higher levels. However, it is not known whether Na-erionite and crocidolite act via the
same pathways. Potentially due to the increased, relative expression of c-jun induced
by Na-erioinite, increased apoptosis is not observed in association with exposure to Na-
erionite. However, the link between increased c-jun expression and inhibition of
apoptosis has not been demonstrated explicitly.
7.125
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Both crocidolite and Na-erionite were also shown by Mossman and coworkers to induce
uptake of bromodeoxyuridine (BrdU) by RPM cells in these same experiments. Uptake
of BrdU is an indicator of DNA synthesis. Since it has also been shown that crocidolite
is capable of damaging DNA via ROS and other pathways (Sections 7.3.3) and both
crocidolite and erionite are known to induce mesotheliomas in any case, the balance
between proliferation and apoptosis in this cell population that is struck by exposure to
these toxins may very well determine whether development of cancers are promoted or
prevented. Of course, both proliferation and apoptosis may also be mediated by other
pathways independent of the ones described here.
The problem is that the range of responses that are induced by asbestos in the lung are
varied and complex (see Table 7-5) so that it has not yet been possible to definitively
identify the biochemical triggers that lead to lung cancer or mesothelioma. It is even
likely, for example, that different mechanisms (or combinations of mechanisms)
predominate under different exposure conditions or in association with differing fiber
types or particle sizes. Still, examination of the dependence of candidate mechanisms
on fiber type and particle size can be instructive, especially to the degree that such
indications are consistent with observations in whole animal studies (see, for example,
Section 7.2.2). For the signaling cascade described above by Mossman and
coworkers, for example, the effects attributable to crocidolite are clearly dependent on
fiber size because the non-fibrous analog to crocidolite, riebeckite does not induce any
of the effects. It also appears that the chemistry of the fibers is important, given the
observed differences in responses among the various, substituted erionites.
7.3.4.3 Asbestos-induced apoptosis. Apoptosis (programmed cell death) is generally
triggered when a cell accumulates certain types of genetic damage, when cell signaling
cascades are triggered by external stimuli that may occur, for example, as part of the
need to maintain tissue homeostasis or to cause a phenotypic shift in response to toxic
challenge (see, for example, Holian et al. 1993, Section 7.3.4.2), or when a cell has
completed a pre-programmed number of divisions. Asbestos can induce apoptosis in a
variety of cells by several mechanisms including primarily:
• by causing sufficient genetic damage to trigger apoptosis; or
• by triggering a signal cascade that leads to apoptosis.
As previously indicated, asbestos may trigger signaling cascades by interacting directly
with receptors on the cell surface, by reacting with integrins for such receptors, or by
inducing production of intermediate species (such as ROS or RNS) that may in turn
induce cell signaling.
7.126
Revision 1 • B/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Some of the mechanisms by which asbestos may act to induce apoptosis may be fiber
size- or type-dependent. Also, responses may vary in different target tissues.
Fibers:
• As indicated in Section 7.4.3.2, crocidolite (but not the non-fibrous analog
riebeckite) induces apoptosis in hamster tracheal epithelial cells and rat pleural
mesothelial cells when applied at non-cytotoxic concentrations (Mossman et al.
1997). Results from the study also indicate that apoptosis is triggered in this
case by inducing an ERK-MAPK signaling cascade as a consequence of
interaction with EGF receptors on the cell surface. The interaction may be direct
or may be caused by asbestos-induced ROS. In a related study (also previously
summarized, Section 7.3.4.1), Timblin et al. (1998a) indicate that the asbestos-
induced apoptosis reported in the Mossman et al. work is fiber-type specific and
the pathway involved appears to stimulate expression of c-fos.
• In a study previously reported in greater detail (Section 7.3.4.1), Levresse et al.
(1997) indicate that chrysotile induces apoptosis in cultured rat pleural
mesothelioma cells with the effect peaking at 4% at 72 hours following exposure
to 10 pg/cm2. Although the authors observed a much smaller effect with
crocidolite, they indicate that the difference is likely due to the much smaller
number of long fibers in the particular crocidolite sample evaluated.
• Broaddus et al. (1997) indicates that crocidolite (not UICC) but not wollastonite,
glass beads, or non-fibrous riebeckite cause substantial apoptosis in rabbit
pleural mesothelioma cells in culture in a dose-dependent fashion. The extent of
apoptosis induced was inhibited by treatment with catalase and by
3-minobenzamide (an inhibitor of poly(ADP-ribosyl) polymerase. The former
indicates a role for ROS mediation and the latter indicates that this enzyme,
which mediates DNA repair, also mediates asbestos-induced apoptosis (perhaps
triggered by asbestos-induced DNA damage). Asbestos induced apoptosis was
also inhibited by treatment with desfeoxamine, but effects were restored by
adding iron to the medium. The authors note that in other studies, crocidolite
has been shown to induce DNA strand breaks within 2 hrs after exposure and
induces unscheduled DNA synthesis within 24 hrs following exposure. Asbestos
also induces production of poly(ADP-ribosyl) polymerase
Non-fibrous particles:
7.127
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
• Leigh et a\. (1997) intratracheal!/ instilled rats with silica (a non-fibrous particle)
at doses varying between 2 and 22 mg. They then collected cells by bronchio-
alveolar lavage (BAL) 10 days after instillation. The authors observed large
numbers of apoptotic cells in BAL fluid and that the number of such cells was
dose-dependent. The dead cells were primarily neutrophils (so that this might
represent some type of mechanism to restore homeostasis). Engulfment of
apoptotic cells by macrophages was also observed. The authors report that, 56
days after instillation, apoptotic cells were observed in granulomatous tissue
within the lungs of rats exposed to silica. This suggests that apoptosis may also
occur in response to chronic inflammation. The authors conclude that silica
induces apoptosis among granulomatous cells and alveolar cells and that such
apoptosis and the subsequent engulfment of apoptotic cells by macrophages
may play a role in the evolution of silica-related disease. The authors also note
that granuloma formation is a hyperplasia-related event.
At least some of the mechanisms suggested above for asbestos-induced apoptosis are
dependent on fiber size (the non-fibrous analog of crocidolite does not induce the
effect) and dependent on the chemistry of the fibers involved (various, cation-
substituted analogs of erionite exhibit disparate ability to induce the effect). Although
non-fibrous particles (such as crystalline silica) may also induce apoptosis, as
previously suggested, this may be through separate mechanisms from those
responsible for asbestos-related effects, even if the same end point results.
7.3.4.4 Asbestos-induced cytotoxicity. While there is ample evidence from various
in-vitro studies that asbestos is cytotoxic, such effects are observed almost exclusively
at the highest concentrations evaluated in an experiment (for example, Luster and
Simeonova 1998, Section 7.3.4.2 and Choe et al. 1999, Section 7.3.4.2). Many
experiments are conducted at concentrations below those for which cytotoxicity is
important because the other toxic effects attributable to asbestos occur at substantially
lower exposure levels and researchers prefer to study such effects in the absence of
potentially confounding cytotoxicity. For in-vitro studies, for example, non-cyotoxic
effects are typically studied at concentrations less than approximately 10 pg/cm2 (or 20
pg/ml) while substantial cytotoxicity is not typically observed until exposure
concentrations are several times higher.
Because most of the other effects attributable to asbestos occur at concentrations that
are substantially lower, this begs the question as to whether frank cytotoxicity is an
effect that is relevant to human exposures. There is also substantial evidence that the
mechanisms associated with asbestos-induced cytotoxicity are separate from the
mechanisms that mediate most of the other asbestos-related effects of interest.
7.126
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Kamp et al. (1993) dosed cultured pulmonary epithelial (PE) cells with UICC
amosite asbestos. In some studies, polymorphonuclear leukocytes (PMN) were
also added to the culture. Typical doses in this experiment were on the order of
250 pg/cmJ, which is quite high for these types of studies. For example,
compare this level with the levels reported for studies of ROS/RNS generation
(Section 7.3.3), cell signaling (Section 7.3.4.2), proliferation (Section 7.3.4.1) or
apoptosis (Section 7.3.4.3).
Kamp and coworkers indicate that the effect of amosite exposure on cultured PE
cells (at the concentrations studied) was to induce substantial cell lysis
(cytotoxicity) and little cell detachment (from the culture medium), which would
indicate increased cell motility. Addition of PMN to the culture resulted in both
increased cell lysis and cell detachment for comparable exposures to amosite.
The observed cell detachment was mitigated in a dose-dependent fashion by
adding protease inhibitors. Further work indicated that asbestos induces release
of human neutrophil elastase (HNE), which may mediate the combined effects
with PMN. PE cell exposure to HNE alone causes increases in cell detachment
in a dose-dependent fashion. However, when combined with asbestos exposure,
cell lysis increases at the expense of cell detachment. The authors suggest that
HNE becomes bound to asbestos, which also becomes bound to PE cells and
this facilitates augmented cytotoxicity by proteases that are secreted by PMN's.
Blake et al. (1998) studied the effect of fiber size on the cytotoxicity of alveolar
macrophages in vitro. Cultured cells were dosed with concentrations varying
between 0 and 500 pg/ml of each of 5 different length preparations of JM 100
glass fiber. Cytotoxicity was monitored by assays for extracellular LDH and by
chemiluminescence following zymosan addition. The latter assy is intended to
show macrophage stimulation. Results indicate that all samples showed dose-
dependent increases in toxicity (i.e. increasing LDH and decreasing
chemiluminescence). Comparing across samples, relatively long fibers
(mean = 17 pm) showed the greatest toxicity. The authors further indicate that
microscopic examination suggests that frustrated phagocytosis plays a role in
cytotoxicity.
Goodglick and Kane (1990) studied the effect of three different length
preparations of crocidolite (long, short, and UICC) on elicited macrophages
(stimulated initially with thioglycolate) in vitro and in vivo. The long and short
samples were reportedly prepared from the UICC sample by repeated
centrification. Goodglick and Kane report that all three types of crocidolite
7.129
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
stimulated release of ROS from macrophages. At sufficiently high
concentrations, all three also caused substantial cytotoxicity, although apparently
due to the longer time required to settle in culture, the full effects from short
fibers take longer to develop. They suggest that, on a total fiber number or
surface area basis, long and short crocidolite appear to exhibit approximately
equal potency toward the cytotoxicity of macrophages. Further work with
various inhibitors indicates that cytotoxicity is mediated by production of ROS
and that ROS is produced via an iron-dependent pathway. They also indicate
that, among the effects of crocidolite exposure is that macrophage mitochondrial
membranes are depolarized
Goodglick and Kane also evaluated the effects of long and short crocidolite in
vivo. This was done by evaluating the effects of intraperitoneal injection of the
various samples (long, short, or mixed crocidolite or titanium dioxide particles) in
C57B1/6 mice. Results indicate that a single injection of long crocidolite (480
jjg) induced an intense inflammatory response, leakage of albumin, and fibers
observed scattered across the diaphragm. In contrast, a single injection of short
crocidolite (600 pg) induced only a relatively mild inflammatory response and
only limited clusters of fibers observed on the surface of the diaphragm.
To test whether short fibers would show a greater response, if they were not
cleared more readily than long fibers, Goodglick and Kane also subjected mice
to 5 consecutive, daily injections of 120 pg of short crocidolite and noted more
substantial aggregations of fiber clusters along the diaphragm as well as a more
pronounced inflammatory response. Cell injury was also assessed by Trypan
blue staining (which indicates cell death). All mice singly or multiply injected with
mixed or long crocidolite showed marked Trypan blue staining. Single injections
of short structures showed only limited Trypan blue staining. However, following
5 daily injections of short fibers, multiple Trypan blue stained cells were observed
on the diaphragm in the vicinity of the locations were clusters of fibers were also
observed. The authors also indicate (in contrast) that neither single injections of
160 or 800 pg nor 5 consecutive (160 pg) injections of titanium dioxide produced
any Trypan blue staining.
The authors conclude from this study that both short and long crocidolite fibers
appear to be cytotoxic to macrophages while titanium dioxide particles are not
(suggesting that not only fiber length, but fiber type is important to cytotoxicity).
They further suggest that, while short fibers tend to be cleared rapidly in vivo,
when such clearance mechanisms are overwhelmed (such as by repeated insult
through repeated, daily injections in this study), then the toxic effects of short
7.130
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
structures becomes apparent. As indicated in other studies, however, there may
be multiple mechanisms working to produce similar responses, that such
mechanisms may exhibit varying dose-response characteristics, and that
cytotoxicity may not generally be directly related to mechanisms that contribute
to carcinogenesis. Moreover, there almost certainly are at least some longer
fibers in the short fiber preparation and extended analysis to determine their
relative concentration with adequate precision would be helpful to see if the
relative magnitude of the observed effects correlate.
• Palekar et al. (1979) studied the ability of four different samples of
commingtonite-grunerite, each also subjected to varying degrees of grinding, to
induce hemolysis of mammalian erythrocytes and cytotoxicity to Chinese
hamster ovary (CHO) cells. The samples studied include: UfCC amosite (4.13
m2/g surface area/mass), which is denoted as "asbestiform grunerite; "semi-
asbestiform" commingtonite (3.88 m2/g); acicular commingtonite (ground to three
particle sizes: 3.76, 2.45, and 0.82 m2/g), and acicular grunerite (2.82 m2/g).
Results from this study indicate that amosite induced the greatest hemolysis of
erythrocytes by far (approximately 50%) while acicular, unground grunerite
caused no hemolysis. However, grinding the acicular grunerite to increasingly
smaller particle sizes and greater surface area ultimately results in some
hemolysis. Both semi-asbestiform and acicular, unground commingtonite show
hemolytic activity between amosite and unground, acicular grunerite and grinding
acicular commingtonite also increased its hemolytic activity.
Similar results were also observed for cytotoxicity. Amosite was by far the most
cytotoxic and the effect was dose-dependent. A dose of 0.05 mg/ml caused
approximately 75% cell death for CHO cells. At 0.2 mg/ml, only 1% of cells
survived. Acicular grunerite was nontoxic even at 0.5 mg/ml. With grinding,
acicular grunerite cytotoxicity increased, albeit only slowly. The most heavily
ground sample killed fewer than 25% of cells at 0.2 mg/ml and killed only 65% at
0.5 mg/ml. The cytotoxicity of semi-asbestiform commingtonite was substantially
less than amosite but greater even than ground, acicular grunerite. For this
material, 0.2 mg/ml killed approximately 65% of cells and 0.5 mg/ml killed
approximately 90%. Interestingly, approximately the same dose-response curve
for cytotoxicity was observed for the 3.88 and 1.61 specific surface area samples
of this material. The 1.21 samples was somewhat less cytotoxic. Acicular
commingtonite was somewhat less cytotoxic (for corresponding doses) than
semiasbestiform commingtonite at the highest specific surface area (3.76) and
its toxicity decreased with decreasing surface area.
7.131
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
The authors also indicate that neither surface charges on crystal particles nor
Magnesium ion content appear to correlate with biological activity. The authors
conclude that the degree of "asbestiform" character of a mineral has a dominant
effect on biological activity. Moreover, although non-fibrous particles may also
be biologically active (and their activity increases with increasing specific surface
area), the effects of particles and fibers lie along entirely separate dose-response
curves. The biological activity of fibrous materials does not appear to depend
directly on specific surface area.
Importantly, the results of the Palekar et al. (1979) study are also consistent with the
possibility that fibrous structures within a specific range of sizes and shapes contribute
strongly to biological activity while largely non-fibrous particles act through a separate
mechanism that depends primarily on total surface area but that, particle-for-particle,
elicits a substantially lower overall response than the mechanism by which fibers act.
Such a scenario is supported by several studies. Jaurand (1997), for example, indicate
that ROS is implicated in the cytotoxicity of long but not short fibers on tracheal
epithelial cells. Although the evidence for distinct mechanisms for fibers and particles
discussed here is specific to cytotoxic and hemolytic effects, evidence in other studies
suggest similar scenarios for other toxic end points (potentially including end points that
contribute to carcinogenicity).
Comparisons of the rate and extent of effects observed in epidemiology studies, whole
animal dose-response studies, and in vitro studies suggests that cytotoxicity may not be
important to human exposures. Unfortunately, however, there is currently insufficient
information to compare doses and exposures across these studies in a more
quantitative fashion. Therefore, the importance of cytotoxicity to human asbestos
exposure cannot be definitively determined at this time.
7.3.4.5 Association between fibrosis and carcinogenicity. The hypothesis that lung
tumor induction is associated with the fibrosis has been examined by several authors.
There appears to be a debate as to whether fibrosis is a necessary precursor for
development of lung cancer (associated with exposure to fibers), whether the presence
of fibrosis is an additional factor contributing to increased risk for lung cancer, or
whether the two diseases are largely unrelated. This is an important consideration
because the characteristics of the dose-response relationship between asbestos and
lung cancer or asbestosis (fibrosis) apparently differ (see Chapter 6 and Sections 7.3.6
and 7.4).
Based on animal studies, Davis and Cowie (1990) found that rats that developed
pulmonary tumors during inhalation experiments exhibited a significantly greater clinical
7.132
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
degree of fibrosis than rats that did not develop tumors. Furthermore, Davis and Cowie
reported suggestive evidence that the pulmonary tumors that did develop in the dosed
rats tended to develop within portions of the rat's lungs that were already scarred by
fibrosis.
As part of a review, Mossman and Churg (1998) indicate that fibrosis of any cause
(including diffuse idiopathic fibrosis) appears to be associated with an increased risk of
lung cancer and that this is observed in both human and animal studies. They also
indicate that only those strains of mice and hamsters that develop fibrotic lesions
following exposure to crystalline silica show an increased risk of developing cancer.
They also report that, in parallel to what is reported for asbestos by Davis and Cowie
(1990), lung tumors that develop following exposure to crystalline silica tend to occur
primarily (if not exclusively) in those portions of the lung where fibrotic lesions
predominate.
In contrast, Case and Dufresne (1997) from their study of lung burdens among Quebec
miners and millers indicate that there is high overlap in the range of concentrations that
lead to both lung cancer and asbestosis and those that lead to lung cancer alone, which
the authors suggest show a lack of relationship between the two diseases (one is not
predictive of the other). The authors indicate that, based on regression of the 111
cases they examined, the only indicator that reasonably tracks lung cancer is severity
of smoking and they indicate that this is true despite the level of fiber content in the
lung.
Although a definitive determination concerning the relationship between fibrosis
(asbestosis) and lung cancer cannot be developed at this time, it does appear that
there is some association between the two diseases. Most likely, fibrosis is an
additional risk factor for lung cancer and thus represents an additional set of
mechanisms that may contribute to the overall risk of developing lung cancer in
association with exposure to asbestos. However, based on the evidence as a whole,
including the evidence that the character of the dose-response relationship for lung
cancer and asbestosis differ, it is not clear that development of fibrosis is an absolute
precursor that is required before asbestos-related lung tumors can develop.
Interestingly, based on the studies reviewed by Mossman and Churg (1998), the
relationship between fibrosis and lung cancer induced by silica may be substantially
stronger than that between fibrosis and lung cancer induced by asbestos.
7.3.4.6 Synergism between asbestos and smoking. Numerous studies have
confirmed the synergistic relationship between smoking and asbestos exposure toward
the induction of lung cancer (see, for example, Hammond et al. 1979, Kamp et al. 1992,
7.133
Revision 1 • 6/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Kamp et al. 1998, Mossman et al. 1996). Smoking is also suspected to facilitate
development of asbestos-induced fibrosis (Kamp et al. 1992).
Putative mechanisms that may contribute to synergism between asbestos exposure
and smoking include:
• facilitated transport of carcinogenic components of smoke that may be
adsorbed on the surface of asbestos fibers, which may then serve as
vehicles to transport these materials through cell membranes to cell
interiors and even to locations adjacent to or within the nucleus (see, for
example, Fubini 1997, Mossman et al. 1996);
• asbestos-catalyzed production of more highly mutagenic metabolites of
the various components of smoke, including benzo(a)pyrene (see, for
example, Mossman et al. 1996);
• smoke-product induced inhibition of clearance of asbestos fibers and/or
asbestos induced inhibition of clearance of smoke products (see, for
example, Mossman et al. 1996); and
• smoke-product induced facilitation of uptake of asbestos by lung
epithelium (see, for example, Mossman et al. 1996).
Although much progress has been made at elucidating the nature of these mechanisms
and other candidate mechanisms, at this point in time it is possible to indicate
definitively neither what mechanisms are important to the observed synergism between
smoking and asbestos exposure nor to indicate the relative magnitude of the
contributions from such mechanisms. Moreover, a detailed review of such
mechanisms is beyond the scope of this document.
7.3.4.7 Conclusions concerning asbestos as a promoter. There is strong evidence
that asbestos acts as a promoter for cancer. While this may primarily involve
mechanisms the contribute to the induction of proliferation, mechanisms associated
with the well-established synergism between smoking and exposure to asbestos (for
the induction of lung cancer; smoking does not appear to affect mesothelioma) are also
important. There also appears to be an association between the development of
fibrosis and an increased risk for lung cancer.
Evidence indicates that multiple mechanisms may be involved with asbestos-induced
cancer promotion and that such mechanisms may be complex and interacting. The
7.134
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
different mechanisms also appear to exhibit dose-response relationships with differing
characteristics. While there are indications that the most important among these
mechanisms may be strong functions of fiber size (with long fibers contributing most to
the induction of disease), mechanisms that depend primarily on surface area or total
fiber (particle) number (for any size range) may also contribute to overall cancer
promotion. Importantly, these latter mechanisms also appear to be strongly associated
with the composition of fibers (particles) and may therefore contribute more
substantially to the disease induction of agents that have been shown to be particularly
toxic (such as crystalline silica), as opposed to particles, fibers, or asbestos in general.
At this point in time, the available data may not be sufficient to distinguish among the
relative contributions from the various mechanisms to the overall promotion of cancer,
at least in terms of the mechanistic data itself. Importantly, however, the mechanistic
data should not be considered to be inconsistent with the results from whole animal
studies, where there are clearer indications that fiber size plays a major role in
carcinogenicity and fiber (particle) type is also important (see Sections 7.2 and 7.4).
Such studies indicate, for asbestos (and other biodurable fibers) that:
• short fibers (less than somewhere between 5 and 10 pm) do not appear
to contribute to disease;
• . potency likely increases regularly for fibers between 10 pm and a
minimum of 20 pm (and, perhaps, continues to increase up to lengths of
at least 40 pm); and
• fiber type may be important primarily in determining biodurability.
They further indicate that particularly (or uniquely) toxic particles (such as crystalline
silica) may act through a different set of mechanisms that are not dependent on fiber
length but that induce toxic end points paralleling those observed for asbestos.
Importantly, the mechanisms by which asbestos may act as a promoter appear to occur
in cell lines that may contribute both to the induction of lung cancer and mesothelioma.
7.3.5 Evidence that Asbestos Induces an Inflammatory Response
There is ample evidence that asbestos induces an inflammatory response in pulmonary
tissues and the pleura (see, for example, Sections 7.2.2 and 7.3). Moreover, there
appears to be multiple biochemical triggers that mediate this response and various
mechanisms may be fiber size- and/or fiber type-specific (Table 7-5). Because the role
7.135
Revision 1 - 6/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
that inflammation plays in the induction of cancer has been addressed elsewhere
(Sections 7.3.3 and 7.3.4), it is beyond the scope of this document to provide a detailed
review of the mechanisms that lead specifically to inflammation.
7.3.6 Evidence that Asbestos Induces Fibrosis
There is ample evidence that asbestos induces fibrosis in pulmonary tissues (see, for
example, Sections 7.2.2 and 7.3). Moreover, there appears to be multiple biochemical
triggers that mediate this response and various mechanisms may be fiber size- and/or
fiber type-specific (Table 7-5). Because the role that fibrosis plays in the induction of
cancer has been addressed elsewhere (Sections 7.3.3 and 7.3.4), it is beyond the
scope of this document to provide a detailed review of the mechanisms that lead
specifically to fibrosis. Such mechanisms have also been the subject of recent reviews
(see, for example, Mossman and Churg 1998 and Robledo and Mossman 1999).
7.3.7 Evidence that Asbestos Mediates Changes in Epithelial Permeability
As previously indicated (Section 4.4), maintaining the overall integrity of the epithelial
surface of the lung is among the various functions of Type II epithelial cells (Leikauf and
Driscoll 1993). It has been shown that asbestos induces changes in the morphology of
Type II epithelial cells (see, for example, llgren and Chatfield 1998), which has the
effect (among others) of increasing the overall permeability of lung epithelial tissue to
various macromolecules and, potentially to asbestos fibers themselves. The former
plays a role in asbestos-induced fibrosis (by allowing cytokines that stimulate fibroblast
proliferation or stimulate fibroblasts to generate extracellular matrix to pass through the
epithelium and reach the underlying fibroblasts, Section 7.3.6). The latter may be
important to facilitating transport of asbestos from the alveolar lumen to the interstitium
(see, for example, Lippmann 1994).
Changes in epithelial permeability may be triggered by cytokines released from other
cells or by the action of asbestos fibers on epithelial cells directly. Moreover, some of
the mechanisms that mediate this response may be sensitive to fiber size and/or fiber
type. For example, Gross et al. (1994) showed that monolayers of human bronchial
epithelial cells cultured over a porous medium and exposed to cryogenically ground
chrysotile (average length: 1 pm, average aspect ratio:14 at 15 pg/culture plate)
became permeable to fibrin breakdown products (FBP's). The cultures were grown
over human serum with labeled fibrinogen. This was based on observed increased
concentrations of FBP's (double in 24 hrs) in the ablumenal chambers of exposed cells
compared to cells in control cultures. Because the epithelium showed greater
permeability to all concentrations, the increased concentrations were not due to
7.136
Revision Y - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
increased breakdown. The observed FDP flux was not vectoral, not saturable, and
required neither proteolytic processing nor active transport. Thus, asbestos increases
the paracellular flux of intact FDP across airway epithelium.
7.3.8 Conclusions Regarding the Biochemical Mechanisms of Asbestos-
Related Diseases
That the specific biochemical triggers for asbestos-related diseases (particularly, the
asbestos-related cancers) have not been definitively delineated as of yet is not
surprising. The detailed interactions between fibers (and particles) and the cells and
tissues of the lung are complex and there are complex, multiple, interacting
mechanisms by which such interactions may contribute to disease. Despite great
progress in elucidating candidate mechanisms, the number of candidate mechanisms is
large and distinguishing among their relative contributions has been difficult. This is
because, among other things, the ability to compare results across studies of different
mechanisms is currently limited due to the inability to reconcile the quantitative effects
of dose and response across dissimilar studies.
Nevertheless, a number of important implications can be gleaned from the available
literature. First, it appears that asbestos can function both as a cancer initiator and a
promoter. It also appears that both the initiation and promotion of cancer may occur
through more than one mechanism.
Regarding cancer initiation, asbestos likely acts primarily through a mechanism
involving interference with mitosis. By this mechanism, asbestos fibers are
phagocytized by target cells, migrate to perinuclear locations, and interact with the
spindle apparatus and other cell assemblages required to complete mitosis. This tends
to result in aneuploidy and may cause various clastogenic effects. This mechanism is
driven by long fibers; short fibers do not appear to contribute to the effect. It also
appears that all asbestos fiber types (and potentially other durable fibers with sufficient
dimensions) cause genetic damage via this mechanisms. If there are effects due to
fiber type, they appear only to play a secondary role.
Although there is also evidence that asbestos may induce production of DNA adducts
and DNA strand breaks (through ROS and RNS mediated pathways), whether such
adducts or breaks ultimately lead to permanent, heritable changes to DNA remain to be
demonstrated. The relative importance of ROS/RNS mediated pathways for initiating
cancer, compared to the pathway involving interference with mitosis, also remains to be
determined.
7.137
-Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
There is also some evidence that the relative importance of asbestos as a cancer
initiator may differ in different tissues. Lung epithelial cells, for example, appear to be
relatively resistant to the mechanisms by which asbestos may initiate cancer.
Mesothelial cells are not. Among several possibilities, this may be due to the ability of
proliferation-competent lung epithelial cells (Type II cells) to undergo terminal
differentiation when challenged with certain toxins and this is a pathway not available to
mesothelial cells.
The mechanisms by which asbestos may promote cancer primarily involve mechanisms
that contribute to the induction of proliferation, although mechanisms associated with
the well-established synergism between smoking and exposure to asbestos to induce
lung cancer are also important. There also appears to be an association between the
development of fibrosis (including asbestosis) and an increased risk of lung cancer.
Evidence indicates that multiple mechanisms may be involved with asbestos-induced
cancer promotion and that such mechanisms may be complex and interacting. The
different mechanisms also appear to exhibit dose-response relationships with differing
characteristics. While there are indications that the most important of these may be
strong functions of fiber size (with long fibers contributing the most to carcinogenicity),
mechanisms that depend primarily on surface area or total fiber number (for any size
range) may also contribute to overall cancer promotion. These latter mechanisms also
appear to be strongly associated with the composition of fibers and may therefore
contribute more substantially to the disease induction of agents that have been shown
to be particularly toxic, as opposed to particles, fibers, or asbestos in general.
Although crystalline silica may act to produce some of the same effects as asbestos
(including carcinogenicity), there is substantial evidence that this family of materials do
not act through the same pathways and that the characteristics of their respective dose-
response relationships may differ. Thus, for example, while asbestos likely induces
cancer through mechanisms that favor long (and potentially thin) fibers, silica more
likely acts through a mechanism that is dependent on total surface area, with freshly
and finely ground material likely being the most potent. In contrast, grinding asbestos
fibers tends to lesson its carcinogenicity overall. Due to differences in chemistry and
crystallinity (reinforced by studies indicating a lack of correspondence in behavior),
crystalline silica does not appear to be an appropriate analog for any of the asbestos
fiber types. Rather, for example, the appropriate non-fibrous analog for crocidolite is
riebeckite and the appropriate non-fibrous analog for chrysotile is antigorite or lizardite.
7.4 ANIMAL DOSE RESPONSE STUDIES
7.138
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Ideally, human epidemiology studies (reviewed in Chapter 6) provide the best data
from which to judge the effects of asbestos in humans and they certainly provide the
only reliable information from which to derive dose-response factors for humans.
However, animal dose-response studies have proven useful for elucidating certain
features of the relationship between asbestos dose and response that cannot be
adequately explored in the human studies, primarily due to limitations in the manner
that exposures were characterized in the human studies -(see Chapter 5).
Unlike human epidemiology studies, exposures in animal studies are controlled and
better quantified. Frequently, the characteristics (in terms of fiber size, shape, and
type) of such exposures have also been better quantified and this has allowed
exploration of the effects that such characteristics (fiber size, shape, type) have on
disease response. Accordingly, an overview of animal dose-response studies is
provided in this section. Both injection-implantation studies and inhalation studies are
reviewed. Particular attention is also focused on a "supplemental" animal inhalation
study that we conducted with the specific aim of identifying the characteristics of
asbestos that best relate to risk. The strengths and limitations of these kinds of studies
are described in Chapter 5.
7.4.1 Injection-Implantation Studies
Because the fibrous materials in injection and implantation studies are placed
immediately against the target tissue, the effects of processes associated with
inhalation, retention, and translocation are avoided. The only active mechanisms that
need to be considered in these studies are those that occur directly in the target tissue
(including degradation, clearance, and biological responses of the types described in
the previous sections of this chapter) Fibrous materials placed against the tissue
surface are subject to dissolution, phagocytosis by macrophages, and phagocytosis by
the cells of the target tissue. These mechanisms are described in greater detail in
Section 7.2. A range of biologic responses have also been observed (described in
Section 7.3).
Numerous researchers have performed these types of studies.
The work of Stanton et al. In a series of studies, Stanton et al. (1972,1977, and
1981) implanted fibrous materials and induced mesotheliomas in rats. In the studies, a
pledgette composed of coarse glass is loaded with hardened gelatin containing sample
material and is surgically implanted immediately against the left pleura of the rats.
Control studies demonstrate that the coarse glass of the pledgette does not induce
significant tumors in the absence of other tumorigenic agents in the gelatin.
7.139
Revision 1 • 9/3/D1
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Although the mass dose of material implanted was the same for all experiments
(40 mg), the observed incidence of mesothelioma varied among samples. By
characterizing the dimensions of fibrous structures in the samples using a microscope,
the researchers were able to explore the relationship between fiber size and the
incidence of mesothelioma. By studying a wide range of fibrousrmaterials, Stanton and
his coworkers concluded that the induction of mesothelioma is determined primarily by
the physical dimensions of fibers and that mineral composition is secondary. Further,
potency appears to increase with the length and decrease with the diameter of fibrous
structures. The researchers also concluded that the incidence of malignant tumors
correlates with the degree of fibrosis induced by the presence of the fibrous materials.
This does not necessarily imply, however, that fibrosis is a necessary step in the
induction of asbestos-induced tumors (see Section 7.3.4.5).
Conclusions from the Stanton et al. studies indicating that mineralogy is not a factor in
biological response conflicts with evidence provided in Chapter 6 and implications
gleaned from mechanism studies presented in Section 7.3. However, the Stanton et al.
studies have been shown to suffer from certain methodological limitations (Berman et
al. 1995) so that results from these studies should be considered more qualitative than
quantitative.
Due to limitations in the ability to produce samples composed of uniform fibers,
quantitative relationships between size and potency were explored by Stanton et al.
(1972, 1977, and 1981) using a regression analysis. Structures longer than 8 pm with
diameters less than 0.25 pm or longer structures with diameters less than 1.5 pm were
found to represent the range of sizes that best correlate with carcinogenicity. It was
further stated that such correlations did not eliminate the possibility that other size
ranges also contribute to potency, only that the two size ranges identified appear to
correlate best. Samples that varied significantly from the reported correlations were
attributed to errors in the characterization of structure size distributions in those
samples. However, other methodological limitations might also have contributed to the
observed deviations or such "outliers" may also suggest evidence for a mineralogical
effect that is similar to what is reported in other studies (see Section 7.3 and Berman et
al. 1995).
The precision of estimates for the ranges of sizes that contribute to biological activity
that are derived from the Stanton et al. studies is limited so that such estimates should
also be considered qualitative. Size distributions were determined by characterizing
200 to 1,000 structures using TEM and there is no indication that statistically balanced
counting rules were employed (Section 4.3). Under such conditions, counts of
structures longer than 8 pm are likely small and subject to large uncertainties for most
7.140
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
of the samples characterized. Confidence intervals are not provided for any of the
counts presented in these studies!
Potentially larger errors in the Stanton et al. studies could have been introduced by the
method employed to relate fiber counts to sample mass. As indicated in Chapter 5,
estimating contributions to mass by sizing total particles and assuming that this is
proportional to total sample mass is subject to error from the limit to the precision of
characterizing structure dimensions (particularly diameter) and by not accounting for
nonasbestos (and possibly nonfibrous) material in the samples.
Thus, for example, there is no discussion of the precision with which the cut point of
8 pm was determined in these studies. Based on the manner in which size distribution
data are presented, it is unlikely that cut points that vary by up to a factor of 2 would
lead to substantial difference in the quality of the correlation. For example, a cut point
of 5 pm may not be distinguishable from 8 pm within the database examined.
Re-analysis and extension of the Stanton studies. Several researchers have re-
evaluated data from the implantation studies to test additional hypotheses. Using the
Stanton data, Bertrand and Pezerat (1980) examined the relationship between
mesothelioma incidence and several characteristics not evaluated by Stanton et al.
including: average fiber length, average fiber diameter, average fiber aspect ratio, total
fiber surface area and total fiber volume. Results from the regression analysis indicate
that potency varies directly with average length and inversely with average diameter but
that neither parameter is a good indicator alone. Combining the effects of length and
diameter, average aspect ratio is highly correlated with potency. Biological activity does
not correlate highly with structure count, surface area or volume except when fiber sizes
are restricted to the long, thin structures that Stanton defined. Results of this study are
not inconsistent with those originally presented by Stanton except ttiat they emphasize
a set of characteristics that relate parametrically to biological activity rather than
expressing exposure as a single restricted size range of structures.
It is important to note that Bertrand and Pezerat were able to find good correlations
between response and specific "average" characteristics of the samples that are not
proportional to the quantity of the material present in the sample ("intensive"
characteristics). Such intensive characteristics as average aspect ratio, average length,
or average diameter are properties that are independent of the mass of material in a
sample. Since response must be a function of the quantity of sample present, intensive
characteristics should have to be multiplied by characteristics that are proportional to
the mass of a sample (e.g. fiber number, sample mass, or sample volume) in order to
7.141
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
relate them to response. Properties that vary with the mass of a sample are termed
"extensive" properties.
The correlations between intensive properties and response reported by Bertrand and
Pezerat likely succeed within the Stanton database because a constant sample mass
(40 mg) was employed for all of the implantation experiments. However, to apply
dose/response relationships that are dependent only on intensive characteristics
beyond the Stanton data (where mass dose will not be constant), it is necessary to pair
intensive characteristics with extensive characteristics (such as mass or number of
fibers per sample). Therefore, it is unclear how the conclusions from this paper may be
generalized to other data sets.'
In a similar study, Bonneau, et al. (1986) also examined parametric relationships
between structure characteristics and mesothelioma induction. The paper examined
specifically correlations between carcinogenicity and dose in terms of two specific
relationships: dose expressed as fibers longer than 8 |Jm that are thinner than 0.25 pm
("Stanton" fibers) and dose expressed as mean aspect ratio. The researchers conclude
that mean aspect ratio provides an excellent indication of carcinogenicity for individual
fiber types but that each fiber type must be treated separately. Poorer correlations are
found for the relationship between the concentration of "Stanton" fibers and
mesothelioma, even when fiber types are considered independently. Although these
results appear to be consistent with findings reported from mechanistic studies (Chapter
6, Sections 7.2 and 7.3) in that they posit a role for fiber mineralogy, the relationships
evaluated by Bonneau et al. also suffer from the limitation of expressing dose only in
terms of intensive quantities, as discussed above. Direct comparison with other studies
is therefore difficult.
Following up on the reported problems characterizing crocidolite in Stanton's work,
Wylie et al (1987) reanalyzed seven crocidolite samples originally studied by Stanton.
She and coworkers then used the new size distributions to reevaluate the "Stanton
hypothesis" (that the concentration of Stanton fibers in a sample correlates with
carcinogenicity). Wylie and coworkers note that substantial deviations from the Stanton
hypothesis occur for specific samples. They conclude that a specific structure size
range alone is not sufficient to characterize biological activity and that a parametric
relationship with other structure characteristics (potentially including mineral type) may
be necessary to sufficiently describe biological activity.
Conclusions from the Wylie et al. paper must be interpreted carefully because the
researchers evaluated only the relationship between carcinogenicity and the single
specific size range indicated ("Stanton" fibers). Thus, the possibility that improved
7.142
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
correlations exist between biological activity and different size ranges or a combination
of size ranges cannot be ruled out. Qualitatively, conclusions presented in this paper
are not inconsistent with the conclusions reported by Stanton et al. regarding the
general relationship between response and fiber dimensions.
The Wylie et al. study appears to suffer from several methodological problems. These
relate to the manner in which the sample reanalysis was performed. The drop method
for preparing electron microscopy grids (used in this study) is not satisfactory for
preparing grids. In fact, as reported in the study itself, grids prepared as duplicates by
this method were shown to be non-uniform at the 95% confidence interval using a chi
squared test. In addition, only 100 to 300 fibers were counted for each sample. Since
there is no indication that statistically balanced counting was performed, the uncertainty
associated with counts of Stanton fibers may be substantial. Such errors would be
further multiplied by uncertainty introduced during the sizing of total particles to
determine the number of fibers per unit mass.
In a later study, Wylie et al. (1993) examined the effect of width on fiber potency. In this
latter study, results from animal injection and implantation studies were pooled and
subjected to regression analyses to identify correlations between exposure and tumor
incidence. The animal studies selected for inclusion in this analysis were performed on
a variety of tremolite samples exhibiting a range of morphological and dimensional
characteristics.
In their regression analyses, Wylie et al. evaluated a range of exposure indices that
emphasize different morphological or size characteristics to help elucidate the
characteristics of asbestos that induce a biological response. Because all of the animal
studies included in their analyses involved tremolite, mineralogy was not an issue.
Results from the Wylie et al. study suggest that fibers longer than 5 |Jm and thinner
than 1 Mm best correlate with tumor incidence among the animal injection and
implantation studies examined. Further, they suggest that a width limit, rather than a
limit on aspect ratio, better reflects the bounds of the asbestos characteristics that
determine biological activity. They also suggest that complex structures (bundles and
clusters) need to be evaluated as part of the determination of exposure because such
structures can breakdown and contribute to the population of thinner fibers.
Although the results of the Wylie et al. study are interesting and tend to support the
general conclusions in this document related to width, if not to length (see Section 7.5
and Appendix B), as the authors themselves indicate, such results should be
7.143
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
considered qualitative due to the limitations imposed on their study by the methodology
employed. Their study was conducted by:
- combining results from multiple studies without careful consideration of
variation introduced by methodological differences across the studies;
- employing asbestos concentrations determined by SEM and without
careful consideration of differences in the counting methodologies
employed by differing research groups across studies; and
- considering injection and implantation studies, which (as opposed to
inhalation studies) do not account for all of the mechanisms that affect the
dose-response relationship in humans.
The limitations imposed by the above constraints are highlighted in Chapters 4 and 5 of
this report.
Other injection studies. A series of injection studies were conducted by several
research groups. In these studies, fibrous materials were suspended in saline and
injected into rats immediately adjacent either to the pleura or peritoneum. A large
number of fibrous materials have now been studied by this process, as reported by Pott
et al. (1974, 1976, 1978, 1982, and 1987), Muhle et al. (1987), Bolton et al. (1982,
1984, and 1986), Davis etal. (1985, 1986a, 1986d, 1987, and 1988a), and Wagner et
al. (1976, 1980, 1982, 1984, and 1985). Newer studies are also discussed in
Section 7.2. Results confirm that it is the fibrous nature of the materials that is the
primary factor leading to the induction of tumors and that potency appears to depend
directly on length and inversely on diameter.
The authors of these studies tend to indicate that, except where fibers are not
persistent in vivo due to solubility or other degradation processes, the mineralogy of the
fibers appears to play only a secondary role in determining disease incidence.
Researchers conducting injection experiments also tended to report a correlation
between tumor incidence and the degree of fibrosis induced by the sample. These
observations are consistent with the ideas originally articulated by Stanton.
Pott developed Stanton's ideas further by suggesting that carcinogenicity is a
continuous function of fiber dimensions, which decreases rapidly for lengths less than
10 pm and also decreases with increasing diameter. The possibility was also raised
that the apparent inverse dependence on diameter may be an artifact due to the limited
number of thick fibers that can be injected in a sample of fixed mass.
7.144
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Although the published injection studies indicate that potency decreases with
decreasing length, researchers in these earlier studies were reluctant to identify a
length below which contributions to carcinogenicity can be considered inconsequential.
This may be due in part to the skewed distribution of fiber sizes typical of asbestos
dusts. Thus, for example, even if structures less than 5 pm are only 1% as potent as
structures longer than 5 pm, they may be as much as 100 times as plentiful in some
asbestos dusts, so that the total contribution to potency would be equal for both size
fractions.
Reasonable dose/response curves have been generated using various sample masses
of a single material in some of these studies. This has been demonstrated for UICC
crocidolite and UICC chrysotile "A" (Bolton 1984). Results indicate that the relationship
between tumor incidence and the log of the dose may be linear and there is no effective
threshold. A consistent difference between the two dusts is apparent; the points lie
along separate curves and chrysotile appears to be more potent per unit sample mass.
In general, details of the analytical techniques used for quantifying size distributions in
these studies are not fully documented. To the extent that they are, it appears that
similar approaches were adopted to those described for the implantation studies above.
Consequently, similar limitations apply to the interpretation of results. Briefly, large
uncertainties are likely associated with counts of long fibers and estimates of the
number of fibers per unit sample mass. Counts in several of the studies also suffer
from limitations in the ability of SEM or PCM to detect thin fibers (Section 4.3);
whenever SEM or PCM was employed, the thinner fibers were likely under-represented
in reported fiber size distributions.
Because samples are placed against mesenchyme in the published implantation and
injection studies, results of these studies most directly represent processes associated
with the induction of mesothelioma. Assuming, however, that clearance and
degradation processes are similar in the deep lung, once a fiber reaches a target
tissue, results from the implantation and injection studies may also provide a model of
biological response in lung tissue and the factors that lead to the induction of pulmonary
tumors. Such a model must be considered qualitative at best, however, because it has
been shown that the mechanisms of tissue response to the presence of asbestos in
lung parenchyma and in the mesenchyme differ in detail (Section 7.3). The time
periods over which the various clearance mechanisms operate in the deep lung and the
mesenchyme also differ (Section 7.2), although, it is apparent that the general nature of
the clearance and degradation processes in the two tissue types are generally similar.
7.4.2 Animal Inhalation Studies
7.145
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Animal inhalation studies measure response to exposure in controlled systems that
model all of the relevant variables associated with asbestos disease mechanisms in
humans (including respirability, retention, degradation, clearance, translocation, and
tissue-specific response). Thus, the available inhalation studies are the best database
from which to evaluate the integrated effects that lead to the development of asbestos-
related disease. Such studies can be used both to identify the characteristics of
asbestos that determine biological activity and to qualitatively elucidate the nature of
the corresponding relationship between exposure (via inhalation) and the induction of
disease.
In this section, the nature and results of the existing animal inhalation studies are
described. In the following section, a project undertaken to overcome the limitations of
the existing animaJ inhalation studies is described (the supplemental inhalation study).
Because this latter project was specifically designed to support the risk protocol
presented in this document, the nature and results of this latter project are described in
detail.
The existing animal inhalation database consists of approximately 30 studies of which
approximately 20 contain dose/response information based on lifetime monitoring of
exposed animals, including the work of Davis et al. (1978, 1980, 1985, 1986a, 1986d,
1988a, and 1988b), Wagner et al. (1974, 1982, 1985, and 1987), Bellman etal. (1987,
1995), Bolton et al. (1982), Le Bouffant et al. (1987), Lee et al. (1981), McConnel et al.
(1982), Muhle et al. (1987), Platek et al. (1985), Smith et al. (1987), and Goldstein et al.
(1983). The studies are similar in overall design, although detailed differences
potentially affect the comparability of results from separate studies.
in the inhalation studies, plugs are formed from bulk samples of fibrous asbestos and
related materials, which are placed in a dust generator to be aerosolized. The
generators (Beckett 1975), usually a modified version of the apparatus originally
designed by Timbrell (Timbrell et al. 1968), consist of a rotating brush that sweeps over
an advancing plug of bulk material liberating fibers that are entrained in the controlled
air flow passing through the device. The airborne dust is then passed either into a
delivery system for nose-only exposure or into an exposure chamber where animals are
kept for fixed periods of time (usually 7 hours per day) on a weekly routine (typically 5
days per week). The exposure routine is continued for as long as 2 years in some of
the studies. In some, but not all of the studies, fiber-containing air is passed through a
cyclone or elutriator prior to the exposure chamber so that exposure consists primarily
of particles sized within the respirable range.
7.146
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Asbestos concentrations in the animal inhalation experiments are monitored by a
combination of techniques. The concentration of total dust in the chamber is generally
monitored gravimetrically. Simultaneously, membrane filter samples are collected and
fibers counted by PCM. The quotient of these two measurements yields the number of
(PCM) fibers per unit mass of dust (Section 4.3). The distribution of fiber sizes within
the dusts introduced into the animal exposure chambers may also be determined in
these studies by any of a variety of methods. As indicated previously (Section 4.3),
however, the utility of such measurements depends on the precise manner in which
they are derived.
To derive fiber size distributions, dust samples from these studies have generally been
collected on polycarbonate filters for analysis by SEM. However, such distributions
suffer both from the limitations of SEM (Section 4.3) and from the manner in which they
are tied to the inhalation experiments (see Chapter 5).
Theoretically, the dose of any fiber size fraction can be estimated in a two-step process.
The procedure incorporates consideration of a size fraction termed the PCM-equivalent
fraction (PCME), which is the fraction of structures measured by SEM-(or TEM) that
correspond to the size range of structures known to be visible and therefore -countable
by PCM. First, the concentration of the PCM-equivalent fraction of the fiber size
distribution (measured by SEM) is normalized by dividing its value by the PCM-
measured concentration per unit dust mass observed in the inhalation experiment.
This ratio is then multiplied by the fractional concentration of any specified size range of
interest within the distribution (measured by SEM) to determine the exposure level for
that size fraction. However, because bivariate (length by diameter) size distributions
have not typically been developed in the available studies and because the number of
total fibers longer than 5 pm observed by SEM (without adjustment for width) does not
correspond to the number of total fibers longer than 5 pm observed by PCM, it is not
possible to derive a true PCME fraction from the SEM data. Therefore, the theoretical
approach described above for estimating exposure to specific size fractions cannot
generally be applied in the existing studies.
As indicated above, the data within the published animal inhalation studies are further
constrained by the limitations of the analytical methods employed to generate the data
(Section 4.3). Comparison of data between studies is also hindered by the lack of
sufficient documentation to indicate the specific methods and procedures employed in
each study. Frequently, for example, it is unclear whether respirable dusts or total
dusts have been monitored. Also, several studies fail to report one or both of two
critical pieces of information: fiber-number-to-mass conversion factors and fiber size
7.147
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
distributions. In addition, few studies indicate the precise counting rules employed for
generating size distributions.
When structure-number-to-mass conversion factors are provided, unless the conversion
factor is derived by counting fibers in a specific size range in a known mass of sample
and fiber concentrations in other size ranges are normalized to this count, several types
of error may be introduced. For example, if total sample mass is assumed proportional
to calculated mass derived from volume characterizations of the particles counted,
unless isometric particles are sized along with fibers and both asbestos and
nonasbestos particles are included in the count, a bias will be introduced in the
conversion factor because total sample mass will have been under represented to the
extent that such particles are ignored in the estimation of fiber mass. Even if such
particles are included, significant uncertainty may result from estimating the volumes of
irregular particles and the limited precision associated with the count of the largest
particles (due to their limited number). The uncertainty in the measurement of a fiber's
diameter is squared in contributing to the uncertainty associated with a mass estimate.
Among reported variations in study design, differences in the detailed design and
operation of the aerosolization chamber and the frequency and duration of exposure
also potentially contribute to variation in results between studies. Also, use of differing
animal strains and species across the various studies suggest the possibility that
physiological differences may contribute to the observed variation in study results.
Such differences are discussed further in Chapter 5.
A small subset of the asbestos dusts evaluated in the animal inhalation studies have
been analyzed by TEM. However, even the published fiber size distributions from these
TEM studies are subject to variation from differences in procedures used for sample
preparation, from differences in counting rules, and from precision limitations due to the
limited number of fibers actually characterized (Section 4.3). This latter limitation
particularly affects the precision with which longer fibers are counted.
Although fiber size distributions are primarily based on SEM analyses rather than TEM
analyses in the existing animal inhalation studies, results generally echo the results of
the injection and implantation studies. Thus, longer fibrous structures are observed to
contribute most to asbestos biological activity, at least qualitatively. For example, dusts
containing predominantly long amosite or long chrysotile fibers induce far more
pulmonary tumors than samples containing predominantly short structures (Davis et al
1986a and 1988b). However, dusts evaluated in the existing inhalation experiments
have not been characterized sufficiently to distinguish the dependence of biological
activity on fiber diameter. Neither are the existing studies sufficient to evaluate the
7.148
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
importance of mineralogy (or other potentially important asbestos characteristics) in
determining risk.
7.4.3 Supplemental Inhalation Study
Given the problems with the existing animal inhalation studies, a project was
undertaken to overcome some of the attendant limitations. To control for effects from
variation in study design and execution (including choice of animal strain, animal
handling procedures, equipment design, sample handling procedures, dosing regimen,
and pathology protocols), the project focused on a set of studies generated from a
single laboratory (i.e., the studies published by Davis et al.). Ultimately, the results from
six studies covering nine different asbestos samples (including four types of asbestos
with samples exhibiting multiple size distributions for two asbestos types) and a total of
13 separate experiments (some samples were studied at multiple exposure levels or in
duplicate runs) were pooled for analysis. The database of experiments employed in the
project is described in Table 7-7.
To overcome the limitations in the Davis et al. studies associated with the
characterization of asbestos itself, the dusts studied in the thirteen experiments listed in
Table 7-7 were regenerated by the same group who performed the original studies,
from the same starting materials, using the same equipment, and reproducing the same
conditions under which the original studies were conducted. Samples of the
regenerated dusts were then collected and analyzed by TEM using a modified version
of the Superfund air method (Chatfield and Berman 1990) to generate bi-variate size
distributions that also include detailed characterization of the shapes and complexity of
fibrous structures observed.
The total mass concentration of the regenerated dusts and fiber measurements by
PCM were also collected to provide the data required to link size distributions in the
regenerated dusts to absolute structure concentrations in the original inhalation
experiments. The manner in which such calculations are performed has been
published (Berman et al. 1995).
The concentration estimates (for asbestos structures exhibiting a range of
characteristics of interest) that were derived from the TEM analyses of the regenerated
dusts were then combined with the tumor response data from the set of inhalation
experiments Fisted in Table 7-7 and a statistical analysis was completed to determine if
a measure of asbestos exposure could be identified that satisfactorily predicts the lung
tumor incidence observed. A more limited analysis was also performed to address
mesothelioma; the small number of mesotheliomas observed in the Davis et al. data
7.149
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
constrained the types of analyses that could be completed for this disease. The
detailed procedures employed in this analysis and the results from the first part of the
study have been published (Berman et al. 1995). These are summarized below along
with results from the parts of the study that remain to be published.
7.150
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 7-7
7.1S1
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
In the statistical analysis performed in this study, the individual dose/response profiles
from each of the two data sets were fit to a linear dose/response model:
P, = 1 - EXP( - Q0 <7J)
where:
"P," is the probability of inducing pulmonary tumors observed in the "ith" study.
A probability of one is equivalent to 100% incidence among the animals
dosed in study i;
"Q0" is the correction factor for background derived from the background
incidence of pulmonary tumors observed among the control populations
(pooled from all studies):
"Xj" is the concentration of the "jth" size fraction of fibers in the "ith" study;
"a" is the coefficient of potency for the "jth" size fraction of fibers in the "ith"
study; and
"b" is a coefficient representing a normalization factor accounting for
differences between the dose/response factors found for the summed
contributions of the "j" exposure indices in the "ith" study.
The "aj"s in this analysis are constrained to be positive because it is assumed that no
fiber prevents cancer. The "a^s are also constrained to sum to one so that
contributions to overall potency from individual size fractions are normalized to the
overall concentration of asbestos. Correspondingly, size fractions evaluated represent
disjoint (mutually exclusive) sets.
The dose/response model was set up as indicated in Equation 7.7 to provide flexibility.
The model allows separate potency coefficients to be assigned to individual size
fractions in a dose/response relationship that depends on multiple size fractions.
Simultaneously, the "b" coefficients allows separate potency coefficients to be assigned
to different fiber types or to results from different studies performed under different
experimental conditions.
7.152
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Several investigators (Stanton et al. 1977; Bertrand and Pezerat 1980; Bonneau et al.
1986; and Wylie, et al. 1987) have used a logit curve to investigate the dose/response
relating various measures of asbestos exposure to tumor response. The logit formula
specifies that the tumor probabilities satisfy the relation:
where log x is some measure of asbestos exposure, such as log of concentration of
fibers in some size range. In some instances, the logit model was expanded by
replacing b-log x with a term representing a linear combination of exposure indices so
that multiple exposure indices could be explored simultaneously. The models were fit
using standard linear regression based on normal theory.
An equivalent form for the logit model is:
Written in this form, it is clear that this model does not permit a background response
(i.e. P = 0, whenever x = 0). This is not a serious limitation when there are no tumors in
control animals, such as was the case in Stanton et al. (1977). However, the model
will not adequately fit data in which tumors are found in control animals. This was one
reason for adopting the linear model (Equation 7.7) used in the investigation of the
animal data reported in the study described here.
There is no evidence from this study that the linear model is inadequate. For cases in
this study in which the fit between exposure and response is shown to be inadequate,
the lack of fit is typically observed to be due to an inconsistent (non-monotonic)
dose/response curve so that there is no indication that a non-linear model, such as the
logit, would provide a better fit.
The linear model (Equation 7.7) used in this study was fit using a maximum likelihood
(Cox and Lindley 1974) approach that utilizes the actual underlying binomial
probabilities. This is a more efficient estimation method model than use of regression
methods based on normal theory, which was the fitting method used in the earlier
studies (described above). The regression procedure indicates only whether the
exposure measures that were studied are significantly correlated with tumor response.
In contrast, the statistical analyses performed in this study indicate whether exposures
that are described by a particular characteristic (or combination of characteristics)
log[P/(1-P)] = a + b-log x
(7.8)
P = eV/(1 + eV)
(7.9)
7.153
Revision 1 •9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
satisfactorily predict the observed tumor incidence. To illustrate, it is apparent from
Text Figure 2 of Stanton et al. (1981) that the exposure measure they identify as being
most highly correlated with tumor incidence (fibers longer than 8 pm and thinner than
0.25 pm) does not provide an acceptable fit to the observed tumor incidence. Similarly,
although all of the univariate exposure measures listed in Table 2 of Berman et al.
(1995) are highly correlated with tumor incidence, none of them adequately describe
(fit) lung tumor incidence.
To test for the goodness of fit in this study, each relationship was subjected to the
standard "P" test where the model was rejected if P was significant at the 5% level
indicating that the true model would provide a worse fit only 5% of the time. Among
models that were not rejected based on a P test, several hypotheses concerning the
relative merit of the various models were also examined according to the method of
maximum likelihoods (Cox and Lindley 1974).
An example of an adequate fit to the tumor response data is provided in Figure 7-5
(Figure 3 of Berman et al. 1995). Note that Figures 2 and 3 of the original paper were
inadvertently switched during publication; the correct Figure 3 is reproduced here. The
exposure index plotted in Figure 7-5 is the sum:
Exposure = 0.0017C, + 0.853C2 + 0.145C3 (7.10)
where:
"C," is the concentration of structures between 5 and 40 pm in length that are
thinner than 0.3 pm;
"C2" is the concentration of structures longer than 40 pm that are thinner than
0.3 pm; and
"C3" is the concentration of structures longer than 40 pm that are thicker than 5
pm.
This index of exposure represents one of the optimum indices reported in Berman et al.
1995.
As is clear from the figure, when exposure is expressed in the manner described above,
the tumor responses observed in the 13 separate experiments that were evaluated
increase monotonically with increasing exposure. It is also apparent that the data
points representing each study fall reasonably close to the line representing the
7.154
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
optimized model for this exposure index. Thus, exposure adequately predicts
response.
Results obtained from completing more than 200 statistical analyses to determine
whether various measures of asbestos exposure adequately predict lung tumor
response (Berman et al. 1995) indicate that:
• neither total dust mass nor fiber concentrations determined by PCM
adequately predict lung tumor incidence;
• no univariate measure of exposure (i.e. exposure represented by the
concentration of a single size category of structures as measured by
TEM) was found to adequately predict lung tumor incidence. Of the
univariate measures of exposure examined, the concentration of total
structures longer than 20 JJm provides the best fit-(although still
inadequate); and
• lung tumor incidence can be adequately predicted with measures of
exposure representing a weighted sum of size categories in which longer
structures are assigned greater potency than shorter structures.
The set of analyses completed In support of this work are summarized in Appendix C.
One example of an exposure measure that adequately describes lung tumor incidence
is presented in Figure 7-5. Another exposure measure shown to provide an adequate
fit is:
Exposure = 0.0024C. + 0.9976Cb (7.11)
where:
"C," is the concentration of structures between 5 and 40 pm in length that are
thinner than 0.4 (Jm; and
"Cb" is the concentration of structures longer than 40 pm that are thinner than
0.4 |Jm.
The fit of this index is depicted in Figure7-6.
7.155
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Figure 7-5
7.156
Revision 1 • 9/3/01
-------
Figure 7-5:
Fit of Model.
Tumor Incidence vs. Structure Concentration by TEM
(Length Categories: 5-40 um, > 40 um, Width Categories: < 0.3 um and > 5 um)
80
40
30
20
to
I *
IE
I J
\fatkai bare 90% oorrttimncm HwmIm.
AbbrwMona lor data are axptalnad In T*bto 1.
Lm tompto* duatara and matrtcaa raplaoad by oomponanta.
Original ooncanMlona aHmated by muMplylnfl oonoanMkma In raconatmctad duali by ratto d PCM oonoanMtona.
r * * 1 ' i
o 1
i 1 1 1 1 1 1 1 1 1 1 1 1 T r
2
"T
rill 1 r 7 r 1 1 1 1 1 1 1 1 1 r—i 1 1—
4 5 0 7
WaHjMad Akborm Concartraton
10
Length Categories: 5/
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
In addition to the above, a series of hypotheses tests were also conducted to test such
questions as: whether fiber type affects potency or whether the component fibers in
complex clusters and matrices should be counted individually. Questions concerning
whether mesothelioma incidence can be adequately described by the same measure{s)
of exposure that describe lung tumor incidence were also addressed. Taken as a
whole, the results presented in Berman et al. (1995) support the following, general
conclusions:
• structures contributing to lung tumor incidence are thin (< 0.5 pm) and
long (> 5 pm) with structures longer than 20 pm being the most potent;
• the best estimate is that short structures (< 5 pm) are non-potent. There
is no evidence from this study that these structures contribute anything to
risk;
• among long structures, those shorter than 40 pm appear individually to
contribute no more than a few percent of the potency of the structures
longer than 40 pm;
• lung tumor incidence is best predicted by measurements in which the
component fibers and bundles of complex structures are individually
counted;
• at least for lung tumor induction in rats, the best estimate is that chrysotile
and the amphiboles are equipotent;
• for equivalent size and shape structures, amphiboles are more potent
toward the induction of mesothelioma than chrysotile; and
• after adjusting for the relative potencies of fiber type, the size categories
that contribute to lung tumor incidence appear also to adequately describe
mesothelioma incidence.
A number of supplemental analyses were also conducted, primarily to identify optimal
procedures for performing asbestos analysis and for estimating concentrations. These
analyses have not yet been published, but they are included in the summaries in
Appendix C. The most important results of the supplemental analyses are that:
• tumor incidence can only be adequately fit by data derived from TEM
analysis of samples prepared by a dir-ect transfer procedure.
7.157
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Measurements derived from indirectly prepared samples could not be fit to
lung tumor incidence in any coherent fashion; and
7.158
Revision 1 - 9/3/01
-------
I il vji ivivuvi.
Tumor Incidence vs. Structure Concentration by TEM
(Length Categories: 5-40 um, > 40 urn, Width Categories: < 0.4 um)
Weighted Airborne Concentration
Length Categories: 5pm-40 pm, £ 40 pm
Width Categories: < 0.4 pm
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Figure 7-6
7.159
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
• it was not possible to identify an exposure measure in which potency is
expressed in terms of a single, continuous function of structure length.
Regarding the last point, although we were not able to identify a continuous function of
length that provides an adequate fit to the tumor incidence data, the general results
from the above analysis are not inconsistent with the hypothesis that potency is a
continuous function of length (i.e. the Pott hypothesis, Pott 1982). The Pott hypothesis
suggests that relative potency is low for short fibers, rises rapidly over an intermediate
range of length, and approaches a constant for the longest fibers.
7.4.4 Conclusions Concerning Animal Dose-Response Studies
Results from our evaluation of the animal dose-response data for asbestos (including
the existing injection/implantation studies, the existing inhalation studies, and our
supplemental study) indicate that:
• short structures (less than somewhere between 5 and 10 |jm in length) do
not appear to contribute to cancer risk;
• beyond a fixed, minimum length, potency increases with increasing
length, at least up to a length of 20 pm (and possibly up to a length of as
much as 40 pm);
• the majority of structures that contribute to cancer risk are thin with
diameters less than 0.5 pm and the most potent structures may be even
thinner. In fact, it appears that the structures that are most potent are
substantially thinner than the upper limit defined by respirability;
• identifiable components (fibers and bundles) of complex structures
(clusters and matrices) that exhibit the requisite size range may contribute
to overall cancer risk because such structures likely disaggregate in the
lung. Therefore, such structures should be individually enumerated when
analyzing to determine the concentration of asbestos;
• for asbestos analyses to adequately represent biological activity, samples
need to be prepared by a direct-transfer procedure; and
• based on animal dose-response studies alone, fiber type (i.e. fiber
mineralogy) appears to impart only a modest effect on cancer risk (at least
among the various asbestos types).
7.160
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Regarding the last of the above bullets, that only a modest effect of fiber mineralogy
was observed in the available animal dose-response studies (when large effects are
observed among human studies, Chapter 6), may be due at least in part to the limited
lifetime of the rat relative to the biodurability of the asbestos fiber types evaluated in
these studies, although it is also possible that different mechanisms drive the effects
observed in the animal studies than those that dominate for asbestos-induced cancers
in humans and that such mechanisms depend more strongly on mineralogy. Other
explanations are also possible. Issues relating to both fiber mineralogy and fiber size
are addressed further in Section 7.5 and Chapter 8.
7.5 CONCLUSIONS FROM AN EVALUATION OF SUPPORTING STUDIES
Although gaps in knowledge remain, a review of the literature addressing the health-
related effects of asbestos-(and related materials) provides a generally consistent
picture of the relationship between asbestos exposure and the induction of disease
(lung cancer and mesothelioma). Therefore, the general characteristics of asbestos
exposure that drive the induction of cancer can be inferred from the existing studies and
can be applied to define appropriate procedures for evaluating asbestos-related risk.
Furthermore, although it would be helpful to definitively identify the underlying
biochemical triggers and associated mechanisms that drive asbestos-induced cancer,
this is not an absolute prerequisite for the development of a technically sound protocol
for assessing asbestos-related risk.
As previously indicated (Sections 7.3.8), the biochemical mechanisms that potentially
contribute to the induction of asbestos-induced cancer are complex and varied.
Moreover, different mechanisms appear to exhibit differing dose-response
characteristics (i.e. the various mechanisms do not all show the same kind of
dependence on fiber size or fiber type). Some mechanisms, for example, suggest that
fiber length is important and that only structures that are sufficiently long induce a
response. In contrast, other mechanisms suggest that fibers (and even non-fibrous
particles) may all contribute to response and that the magnitude of the response is a
function of the total surface area of the offending fibers (or particles). Among these
mechanisms, additionally, some suggest that fiber type (i.e. mineralogy) is not an
important determinant of potency while other mechanisms indicate that fiber type is the
predominant determinant of potency.
Unfortunately, the existing studies are not currently adequate to support definitive
identification of the specific mechanisms that drive the induction of asbestos-related
cancer (versus other mechanisms that may contribute only modestly or not at all).
7.161
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
However, whatever mechanisms in fact contribute to the induction of disease, they must
be consistent with the gross characteristics of exposure that are observed to predict
response in the available whole-animal dose-response studies and human
epidemiology studies. Therefore, the implications from these latter studies regarding the
dependence of asbestos-induced cancers on fiber size and type are reviewed here in
some detail. Further, in Chapter 8, they are used to support development of a protocol
for evaluating asbestos-associated risks.
Fiber Length
Fibers less than a minimum length between 5 and 10 pm do not appear to contribute to
risk. This is supported both by the results of our re-analysis of the animal inhalation
studies conducted by Davis et al. (Section 7.4.3 and Berman et al. 1995), in which this
hypothesis was tested formally, and by inferences from the broader literature. As long
as fiber size is adequately characterized, the animal inhalation studies (Section 7.4.2)
and injection/implantation (Section 7.4.1) studies consistently indicate lack of ability of
short structures to contribute to the induction of cancer. Furthermore, animal retention
studies (Section 7.2.1) and histopathology studies (Section 7.2.2) provide strong
mechanistic evidence that explains the lack of potency for short structures; they are
readily cleared from the respiratory tract. Even when sequestered in large numbers in
macrophages within the lung, there is little indication that such structures induce the
kinds of tissue damage, and related mechanisms that appear to be closely associated
with the induction of cancer.
Although there are mechanism studies that may suggest a role for short fibers in the
induction of asbestos-related disease (see, for example, Goodglick and Kane 1990,
Section 7.3.4.4), such studies do not track cancer as an end point. Therefore, the
relationship between the toxic end point observed and the induction of cancer needs to
be adequately addressed before it can be concluded definitively that short structures
can contribute to cancer.
Beyond the minimum length below which structures may be non-potent, potency
appears to increase with increasing length, at least up to a length of 20 pm and
potentially up to a length of 40 pm. The latter limit is suggested by our re-analysis of
the Davis et al. studies (Section 7.4.3) in which it was also found that structures longer
than 40 pm may be as much as 500 times as potent as those between 5 and 40 pm in
length. The former limit is suggested by broader inferences from the literature that
suggest the cutoff in the length of structures that are at least partially cleared by
macrophages from the lung may lie close to 20 pm and that the efficiency of clearance
likely decreases rapidly for structures between 10 and 20 pm in length (Section 7.2).
7.162
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Such inferences are further reinforced by measurements of the overall dimensions of
macrophages in various mammals by Krombach et al. (1997), as reported in
Section 4.4.
Importantly, the inferences that potency increases for structures longer than 10 pm {up
to some limiting length) from these various studies are strongly reinforcing, even though
the upper limits to the points at which potency stops increasing do not precisely
correspond. Furthermore, that the longest structures are substantially more potent
than shorter structures (and that the shortest structures are likely non-potent) dictates
that asbestos analyses performed in support of risk assessment need to provide
adequate sensitivity and precision for counts of the longest structures.
Fiber Diameter
Because fibers that contribute to the induction of cancer must be respirable, they must
also be thin. The studies reviewed in Section 7.1 indicate that respirable fibers are
thinner than 1.5 pm and the vast majority of such structures are thinner than 0.7 pm. In
fact, the results of injection, implantation, and inhalation studies reviewed in
Sections 7.4.1 and 7.4.2 and the results of our supplemental re-analysis of the Davis et
al. studies (Section 7.4.3) indicate that the fibers that contribute most to the induction of
asbestos-related cancers are substantially thinner than the limit suggested by
respirability alone.
Importantly, the results of all of the studies cited above indicate that it is a cutoff in
absolute width that defines the bounds of biological activity rather than a cutoff in
aspect ratio (the ratio of length to width) that has been used to define fibrous structures
heretofore. That is why the exposure index recommended based on our review of the
Davis et al. studies incorporates a maximum width as a cutoff, rather than a minimum
aspect ratio.
Fiber Complexity
In our supplemental evaluation of the Davis et al. studies, the tumor incidence data
from the animal inhalation studies were best fit (predicted) by exposure indices in which
the component fibers and bundles of complex structures (clusters and matrices) were
separately enumerated and included in the exposure index used to represent
concentration (Section 7.4.3). The appropriateness of such an approach is further
supported by the observation that loosely bound structures (including, for example,
chrysotile bundles) readily disaggregate in vivo (Section 7.2). Therefore, it is
recommended in this report that those components of complex structures that
7.163
Revision 1-9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
individually exhibit the required dimensional criteria be individually enumerated and
included as part of the count during analyses to determine the concentration of
asbestos in support of risk assessment.
Fiber Type (Mineralogy)
Two separate but interrelated issues need to be addressed with regard to the effects of
fiber mineralogy. The first is that mineralogy appears to be an important determinant of
cancer risk. The second is that the magnitude of the mineralogical effect appears to be
at best modest in animal dose-response studies but appears to be substantial in human
epidemiology studies. It must also be emphasized that, due to confounding, the effects
of fiber size and fiber mineralogy need to be addressed simultaneously, if one is
interested in drawing useful conclusions concerning fiber mineralogy.
Results from some (but not all) of the animal injection, implantation, and inhalation
studies previously reviewed (Sections7.4.1 and 7.4.2) suggest that mineralogy plays an
important role in determining biological activity. However, the nature of the effects of
mineralogy are not easily separated from size effects, due to the methodological
limitations of the studies cited. Therefore, the evidence from these studies can be
considered ambiguous. Formal hypothesis testing during our re-analysis of animal
inhalation studies (Section 7.4.3) indicates that, when size effects are addressed,
chrysotile and the amphiboles exhibit comparable potency toward the induction of lung
cancer. In contrast, amphiboles appear to be approximately three times more potent
than chrysotile toward the induction of mesothelioma, once fiber size effects are
addressed.
Several of the human pathology studies cited previously (Section 7.2.3) suggest that
mineralogy is an important factor in determining cancer risk, but these studies similarly
suffer from methodological difficulties that introduce ambiguity into the inferences
drawn. However, it is clear from the human epidemiology data (Chapter 6) that
mineralogy plays a substantial role in the determination of risk for human cancer
(primarily, mesothelioma).
The underlying cause(s) for the observed difference in potency between chrysotile and
the amphiboles may relate to differences in fiber durability (Section 7.2), to size/shape
related differences in fibers that are a function of mineralogy and that cause differences
in deposition, retention, or translocation (Sections 7.1 and 7.2), and/or to the
dependence on mineralogy of the specific mechanisms underlying the biological
7.164
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
responses of specific tissues (Section 7.3). The relative magnitudes of such effects on
animal and human pathology also need to be considered, if the observed differences in
potency among animal and human studies, respectively, is to be reconciled. Such
considerations are addressed further in Chapter 8.
Importantly, whether the observed differences in the role of mineralogy toward animal
and human pathology can be reconciled, the effects of mineralogy can be adequately
addressed when assessing asbestos-related cancer risk for humans by incorporating
dose-response coefficients explicitly derived from the human epidemiology data.
Because this is the approach proposed in this document, effects due to mineralogy are
properly addressed.
An Appropriate Exposure Index for Risk Assessment
The optimum exposure index defined based on our re-analysis of the Davis et al.
animal inhalation studies is a weighted sum of two fiber size categories:
= 0.0024C. + 0.9976Cb (7.12)
where:
"Cop," is the concentration of asbestos expressed in terms of the optimum
exposure index (i.e. as the weighted sum of two size categories);
"Ca" is the concentration of structures between 5 and 40 pm in length that are
thinner than 0.4 pm; and
"Cb" is the concentration of structures longer than 40 pm that are thinner than
0.4 |Jm.
This index was shown to adequately fit (predict) the tumor incidence data across the 13
separate animal inhalation experiments evaluated = 0.09). Whether the exposure
index defined in Equation 7.12 is also optimal for capturing the relevant characteristics
of fibers that contribute to the induction of human cancer is an open question. Because
it captures the major characteristics (concerning length and diameter) identified above
that are indicated to be important for human exposures, it represents a promising
candidate. Unfortunately, however, the data required to match this index to a set of
human-derived dose response coefficients does not currently exist (Section 6.2.4).
Therefore, until such time that the data required to adapt a set of human-derived
coefficients to the above index become available, the following interim index (defined
7.165
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
originally in Equation 6.7 and reproduced here), should be used to evaluate asbestos-
related cancer risks for humans:
Casb = 0.003CS + 0.997Cl (7.13)
where:
"Casb" is the concentration of asbestos to be used to estimate risk;
"Cs" is the concentration of asbestos structures between 5 and 10 pm in length
that are also thinner than 0.5 pm; and
"CL" is the concentration of asbestos structures longer than 10 pm that are
also thinner than 0.5 pm.
Although not optimized (so that, for example, it does not provide an adequate fit to the
animal inhalation studies), this index nevertheless captures substantially more of the
characteristics (concerning length and diameter) that are indicated to be important for
human exposures than the index (traditionally defined for phase contrast microscopy) in
current use by EPA (which does not provide an adequate fit to the animal inhalation
studies either). The index currently recommended by EPA is defined as "CpcM" and is
the concentration of asbestos structures longer than 5 pm, thicker than 0.25 pm, with
an aspect ratio greater than 3. A comparison between this index and the "interim" index
defined by Equation 7.13 is instructive.
First, we will ignore the limitations in PCM instrumentation (Section 4.3), which among
other things has been shown incapable of providing reliable measurements of asbestos
concentrations in outdoor settings that are removed from obvious, dominating sources
of asbestos (Berman unpublished). Therefore, we will begin this comparison by
assuming that measurements for the index currently in use by EPA are actually derived
using TEM (so that they can be considered to be "PCM-equivalent concentrations" or
Cpcme)-
Comparing the dimensions of structures incorporated in CPCME and C^, respectively, it
is clear that:
• the interim index, recommended in this document, is weighted toward
longer structures (> 10 pm vs. > 5 pm) than the current EPA approach
and this better captures the expected trend with length;
7.166
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
• the interim index is focused on the thinnest structures, which also tracks
the expected trends in biological activity. In contrast, the current index
excludes the thinnest structures (which are likely more potent and more
numerous than thicker structures in most environments of interest); and
• even more important, the interim index includes only structures that are
respirable while the current index potentially includes counting of a large
fraction of structures that are not respirable (which clearly cannot
contribute to risk).
As indicated above, the size criteria for the interim index proposed in this document
provides substantially better agreement with the characteristics expected of biologically
active fibers (described above) relative to the index currently recommended by EPA
(which is also the index to which the existing epidemiology studies are currently
normalized). It is likely for this reason that human dose-response coefficients adjusted
to match the interim index show improved cross-study agreement relative to unadjusted
coefficients (Sections 6.2.4.2 and 6.3.3.2). Additional advantages of the interim index
are also discussed in Appendix B, where, for example, the interim index is shown to
adequately fit (predict) the relative tumorigenicity of six different tremolite samples that
vary in the degree of their asbestiform character. In contrast, the index currently
recommended by EPA does not adequately fit the tremolite data.
Although the interim index appears to provide substantial improvement in performance
for use in risk assessment over the index in current use, conclusions from the literature
reviewed in this Chapter indicate that additional, substantial improvement may be
gained by adopting an index that more heavily weights structures even longer
thanIO jjm. Although the ideal may be to incorporate separate weights for structures
as long as 40 pm, even incorporating separate weights for structures longer than 20
pm would likely achieve most of the desired improvement. However, this will require
completion of limited additional study to develop the data needed to adjust the existing
human dose-response coefficients to match the improved index. Importantly, it is also
likely that, additional improvement over the interim index can be achieved by
conducting the additional study in a manner that would provide size data appropriate for
the specific facilities evaluated in the existing epidemiology studies (rather than
generically similar sites, matched by industry). The adjustments used in this document
to match the existing dose-response coefficients to the interim index are industry
generic (Section 6.2.4). It is therefore interesting that, even given this cruder approach
for adjusting coefficients, improved agreement across the adjusted coefficients is still
observed (Sections 6.2.4.2 and 6.3.3.2).
7.167
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Interpreting Exposure Indices
It is instructive to examine the relationship between the true (but unknowable) function
that represents biological activity and the functions represented, respectively, by each
of the exposure indices considered in this Section. All of the indices considered above
(the current index, the interim index, and the proposed, optimum index) all represent
step functions with precise dimensional cutoffs. The existing index is a single-
parameter function with cutoffs for minimum length, minimum width, and minimum
aspect ratio. Both the interim index and the proposed, optimal index are two-parameter
indices with cutoffs for minimum lengths, an interim length (at which potency
incrementally increases), and maximum width.
In contrast, the true relationship between fiber dimension and biological activity likely
represents a continuous function without distinct, hard boundaries (although relatively
steep declines in potency for minimum length and maximum width are likely
approximated reasonably well by a hard boundary). Unfortunately, due to the
limitations of the studies conducted to date, it has not been possible heretofore to
define a continuous function that adequately predicts risk. However, as long as the
sensitivity of the associated analysis is set sufficiently low so that a minimum of several
structures (say four or five) would be counted to estimate concentrations the relate to
levels of risk of potential concern, exposure indices that are step functions
(incorporating hard, dimensional boundaries) will provide a reasonably good
approximation to results that would be obtained, should the true, continuous function be
known. Moreover, agreement between results from a step function and the true,
continuous function will improve as the number of structures counted increases (i.e. as
the estimated concentration and associated risk increases), due to the averaging of
potency over the multiple fibers that are observed.
7.168
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
8.0 DISCUSSION, CONCLUSIONS, AND RECOMMENDATIONS
Although gaps in knowledge remain, a review of the literature addressing the health-
related effects of asbestos (and related materials) provides a generally consistent
picture of the relationship between asbestos exposure and the induction of disease
(lung cancer and mesothelioma). Therefore, the general characteristics of asbestos
exposure that drive the induction of cancer can be inferred from the existing studies and
can be applied to define appropriate procedures for evaluating asbestos-related risk.
Moreover, such procedures provide substantial improvement in the confidence that can
be placed in predicting risks in exposure environments of interest compare to
predictions derived based on procedures in current use.
Following a general discussions of the findings of this study, specific recommendations
(including options) for finalizing a protocol for assessing asbestos-related risks using
the procedures identified in this document are provided in Section 8.2.
Recommendations are described in Section 8.3 for limited, focused, additional studies
to:
(1) settle a small number of outstanding issues -{concerning whether the
current dose-response models adequately track the time-dependence of
lung cancer and mesothelioma) and
(2) provide the data required to fully optimize the approach recommended in
this document.
8.1 DISCUSSION AND CONCLUSIONS
8.2.1 Addressing Issues
The issues identified in the introduction (Chapter 2) as part of the focus of this study
can now be addressed. These are:
• whether the dose-response models currently in use by the EPA for
describing the incidence of asbestos-related diseases adequately reflect
the time-dependence for the development of these diseases;
• whether the relative in-vivo durability of different asbestos mineral types
affect their relative potency;
8.1
Revision 1-9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
• whether the set of minerals included in the current definition of asbestos
adequately covers the range of minerals that potentially contribute to
asbestos-related diseases;
• whether the analytical techniques and methods currently used for
determining asbestos concentrations adequately capture the biologically
relevant characteristics of asbestos (particularly with regard to the sizes of
the structures included in the various analyses) so that they can be used
to support risk assessment; and
• whether reasonable confidence can be placed in the cross-study
extrapolation of dose-response relationships that are required to assess
asbestos-related risks in new environments of interest.
Time dependence
Regarding the first of the above-listed issues, the lung cancer model appears to
adequately reflect long-term risk associated with exposure to amphiboles (for which
relative risk remains constant beyond 10 yrs after the end of exposure) but may over-
predict long-term risk for chrysotile (Section 6.2.1). Relative risk for chrysotile-exposed
individuals appears to decrease substantially with time following 10 yrs after the end of
exposure, rather than remain constant, as the model predicts. However, these findings
are based on evaluation of only a single chrysotile cohort and a single amphibole cohort
so that further evaluation of a minimum of two additional cohorts (one each of
amphibole and chrysotile), to determine whether these trends are general, may be
warranted (Section 8.3).
Although there are hints from our evaluation that the mesothelioma model may
somewhat under-predict long term mesothelioma risk following exposure to amphiboles
(Section 6.3.1), the effect appears small and may not be important. Still, because this
finding is based on evaluation of only a single amphibole study, evaluation of at least
one additional cohort may prove prudent. There was no indication that the
mesothelioma model under-predicts risk for chrysotile-exposed individuals, although
only small numbers of mesothejiomas were observed in the data sets evaluated for
chrysotile.
Biodurabilitv
Because the in-vitro dissolution rate for chrysotile in biological fluids is substantially less
than for crocidolite and, likely, other amphiboles (Section 7.2.4), effects potentially
8.2
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
attributable to differences in the relative biodurabiliy of these asbestos types are
addressed in several sections of this document. Among other things, this difference
may explain the reduction in relative risk for lung cancer following cessation of exposure
observed for chrysotile-exposed workers that was not apparent for amphibole-exposed
workers (Section 6.2.1). Both of these affects may also explain the small difference in
potency toward the induction of lung cancer between chrysotile and the amphiboles
observed in this study (Section 6.2.4). Importantly, however, the observed difference is
based only on best estimates of relative potency; formally, the difference was not found
to be significant.
Although the relative biodurability of chrysotile and the amphiboles may potentially
underlie the substantial difference in potency observed between these fiber types
toward the induction of human mesothelioma, which is significant, (Sections 6.2.4.2 and
6.3.3.2), direct evidence for the specific effects contributing to this difference are not
apparent. Thus, for example, the current mesothelioma model appears to adequately
describe the time-dependence for mesothelioma following exposure to both chrysotile
and amphiboles (Section 6.3.1), although the chrysotile data sets examined were fairly
small, which limits the power of this analysis. Moreover, there is evidence suggesting
that, even among the chrysotile-exposed cohorts, the observed mesotheliomas may be
attributable to the small concentration of tremolite (amphibole) asbestos that was found
as a contaminant in the material to which these cohorts were exposed (Section 6.3.3).
That tremolite fibers represent a substantial (sometimes predominant) fraction of the
fibers found in the lungs of deceased workers from these cohorts (Sections 6.2.3 and
7.2.3), despite its trace-level presence in the material handled, suggests either that
chrysotile is not deposited as efficiently in the deep lung (which would not be affected
by biodurability) or that it is cleared much more rapidly (which may imply a role for
biodurability, although size effects may also dominate). Animal retention studies
addressing these effects are somewhat ambiguous (Section 7.2.1 and 7.2.2), although
they too support the general impression that chrysotile (on a mass basis) is either not
deposited as efficiently in the deep lung or cleared more readily from the deep lung (or
both). Still, the fact that long fibers of chrysotile (which are expected to contribute most
to potency) are observed to be retained for extended periods of time, clouds the
potential role for biodurability.
Overall, the substantially larger potency toward the induction of mesothelioma observed
in humans may have several causes, but direct evidence for such causes is not
definitive either among the supporting animals studies or the in-vitro studies. Thus,
while this effect can be adequately addressed by the procedures recommended in this
8.3
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
document for risk assessment (because the recommended risk coefficients are based
on human data), it cannot be entirely explained.
Minerals of concern
Regarding the range of fibrous minerals that potentially contribute to lung cancer and
mesothelioma, available evidence (Sections 7.2 and 7.4) suggests that:
• several minerals and the most biodurable among synthetic fibers (such as
erionite or refractory ceramic fibers) in addition to those included strictly
within the definition for asbestos have been shown capable of inducing
lung cancer and/or mesothelioma (as long as the corresponding fibers fall
within the appropriate size range);
• fibrous minerals that differ radically in chemical composition or crystal
structure (such as erionite, chrysotile, and the amphibole asbestos types)
appear to exhibit substantially different potencies (even after adjusting for
size), which may be largely, but not necessarily entirely, explained by
relative biodurability; and
• fibrous minerals that exhibit closely related chemical compositions and
crystal structures (such as the family of amphibole asbestos types) appear
to exhibit relatively similar potencies (once effects are adjusted for size).
Note, to illustrate that biodurability may not entirely describe relative potencies across
the different fiber types, the biodurability-based model by Eastes and Hadley (1995 and
1996), which is described in Section 7.2.4, would (properly) predict different potencies
for chrysotile and amphiboles for both lung cancer and mesothelioma. However,
because the dose-response models for both lung cancer and mesothelioma are linear
in fiber concentration, the Eastes and Hadley model would incorrectly predict that the
ratio of relative potencies of chrysotile and the amphiboles would be the same in
humans for these two diseases, which is definitely not what is observed (Section 6.4).
Given the above-described observations, it is clear that fibrous minerals beyond those,
included in the definition for asbestos can contribute to lung cancer and mesothelioma.
8.4
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
It is also likely that potencies for minerals that exhibit similar chemistry and crystal
structure (and which therefore likely exhibit similar physical-chemical properties) also
exhibit corresponding potency (for similarly sized fibers). However, the carcinogenicity
of fibers exhibiting radically different chemical compositions and crystal structures than
those already identified as carcinogenic, should likely be evaluated on a case-by-case
basis. Two additional considerations may be useful for focusing such evaluations.
First, the size distribution for fibers composed of the mineral of concern should be
shown to include substantial numbers within the range of structures that potentially
contribute to biological activity (i.e. that fall within the size range defined by the interim
index, Equation 7.13). Second, such fibers should also be shown to be relatively
biodurable (i.e. that they exhibit dissolution rates less than approximately 100 ng/cm2-
hr, Lippmann 2000). Note that, as long as the interim index of exposure is used to
define fiber sizes, whether a fiber is strictly asbestiform appears to be irrelevant
(Appendix B)
Analytical methods
Given the need to detect the thinnest fibers and the need for reliable measurements in
outdoor settings, the only analytical technique that appears to be capable of providing
quantitative data useful for supporting risk assessment is TEM (Sections 4.3 and 7.5).
Further, given the specific size range of structures that need to be evaluated and the
specific manner in which they need to be counted (to assure both cross-study
comparability and compatibility with the recommended dose-response coefficients),
analyses should be performed using the specific analytical methods recommended in
this document. These are ISO 10312 (modified to focus on interim index structures) for
air and the Modified Elutriator method (Berman and Kolk 2000) for soils or bulk
materials. Note that, on a study-specific basis, other methods may be shown to provide
comparable results so that they can also be used as part of a properly integrated
investigation.
Confidence in risk assessment
Given the degree of reconciliation achieved across the various epidemiology studies
(with the adjustments proposed in this document, Section 6.4), it appears that
asbestos-related risks estimated using the recommended procedures may be
considered good to within one to two orders of magnitude. Further, if proper
procedures are employed to derive exposure estimates that are then properly matched
with the recommended dose-response coefficients, the chance of severely
underestimating risk can be minimized without the need to introduce overly
conservative assumptions.
8.5
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
8.1.2 Comparison with Other, Recent Risk Reviews
Although several other reviews have also recently been published that nominally
address risk-related issues for asbestos (including questions concerning the
identification of an appropriate exposure index and the relative potency of varying fiber
types), these studies are either qualitative or involve analysis of data in a manner that
does not allow formal evaluation of nature of the specific dose-response relationships
for the various diseases. Therefore, they are not well suited to support development of
a protocol for conducting formal assessment of asbestos-related risks. Nevertheless,
the general conclusions from these reviews are not inconsistent with our findings.
Hodoson and Darnton (2000)
Hodgson and Darnton (2000) conducted a comprehensive quantitative review of
potency of asbestos for causing lung cancer and mesothelioma in relation to fiber type.
They concluded that amosite and crocidolite were, respectively, on the order of 100 and
500 times more potent for causing mesothelioma than chrysotile. They regarded the
evidence for lung cancer to be less clear cut, but concluded nevertheless that
amphiboles (amosite and crocidolite) were between 10 and 50 times more potent for
causing lung cancer than chrysotile. In reaching this latter conclusion they discounted
the high estimate of chrysotile potency obtained from the South Carolina cohort.
Hodgson and Darnton concluded that inter-study comparisons for amphibole fibers
suggested non-linear dose response relationships for lung cancer and mesothelioma,
although a linear relationship was possible for pleural mesothelioma and lung tumors,
but not for peritoneal mesothelioma.
The Hodgson and Darnton study was based on 17 cohorts, 14 of which were among
the 20 included in the present evaluation. This study had different goals from the
present evaluation and used different methods of analysis. Hodgson and Darnton did
not use the dose response information within a study. Instead, lung cancer potency
was expressed as a cohort-wide excess mortality divided by the cohort mean exposure.
Likewise, mesothelioma potency was expressed as the number of mesothelioma
deaths divided by the expected total number of deaths, normalized to an age of first
exposure of 30, and by the mean exposure for the cohort. These measures have the
advantage of being generally calculatable from the summarized data available from a
study. However, since they are not model-based, it is not clear how they could be used
to assess lifetime risk from a specified exposure pattern, which is an important goal of
the current project. Use of average cohort exposure could cause biases in the
estimates, if, e.g., a large number of subjects were minimally exposed. Also, the
recognized differences between studies in factors, in addition to level of exposure, that
6.6
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
may affect risk will also affect the reliability of conclusions concerning the dose
response shape based on comparisons of results across studies.
Lippmann (1994 and 2000?)
In the most recent of these reviews, although based on a qualitative, anecdotal
treatment of the literature, Lippmann reinforces our general findings that it is longer
fibers (those longer than a minimum of approximately 5 |jm) that contribute to lung
cancer and mesothelioma. He further indicates that, based primarily on the limits
observed for fibers that can be phagocytized, fibers that contribute most to lung cancer
are likely longer than a minimum of 10 fjm. In his review, based on a series of
comparisons of mean and median dimensions reported for the relevant exposures
across a broad range of studies, Lippmann draws several fairly specific conclusions on
the ranges of fiber sizes that may contribute to various diseases (i.e. that the minimum
length fibers that contribute to asbestosis, lung cancer, and mesothelioma are 2, 5, and
10 pm, respectively). He also suggests that fibers that contribute to mesothelioma may
need to be thinner than 0.1 pm while those that contribute to lung cancer may need to
be thicker than 0.15 pm. While it is not clear that drawing such specific conclusions
can be firmly supported by the kinds of qualitative comparisons across reported mean
and median dimensions for exposures in various studies that are described in this
paper, the author indicates that further, more formal study of the dose-response
relationships that he posits is warranted. It is noted that many of the studies reviewed
by Lippmann are the same studies also incorporated in our analysis.
In the earlier review, Lippmann plots lung tumor incidence as a function of inhaled
animal dose for data from a series of broadly varying studies based, respectively, on
fibers longer than 5,10, and 20 fjm (no widths considered) and suggests that the
quality of the fits are comparable. The author further suggests, based on this
evaluation, that PCM seems to provide a reasonable index of exposure. However, no
formal goodness of fit tests were performed in this analysis and, based on visual
inspection, none of the plots would likely show an adequate fit. Moreover, the plot of
the tumor response vs. dose as a function of fibers longer than 5 pm appears to be
substantially worse than the other two plots; if one removes the single highest point in
this plot, it appears that any sign of correlation will largely disappear.
Stavner et al. 1996
In the context of evaluating the "amphibole hypothesis", Stayner et al. computed the
excess relative risk of lung cancer per fiber/ml/yr from 10 studies categorized by the
fiber types to which the cohort was exposed. Each of these studies was also included
8.7
Revwion 1 -B/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
in the present evaluation. Both the lowest and highest excess relative risks came from
cohorts exposed exclusively to chrysotile. Based on their evaluation, they concluded
that the epidemiologic evidence did not support the hypothesis that chrysotile asbestos
is less potent than amphibole for inducing lung cancer. However, based on a review of
the percentage of deaths in various cohorts from mesothelioma, they concluded that
amphiboles were likely to be more potent than chrysotile in the induction of
mesothelioma. They also noted that comparison of the potency of different forms of
asbestos are severely limited by uncontrolled differences in fiber sizes. None of these
conclusions are inconsistent with our general findings.
8.2 RECOMMENDATIONS FOR ASSESSING ASBESTOS-RELATED RISKS
Recommended risk coefficients for lung cancer and mesothelioma (the adjusted KL
values and KM values) are provided in Tables 6-29 and 6-30, which present best-
estimate recommendations and conservative recommendations, respectively. Once it
is decided which of these sets of risk coefficients to adopt for general use, the selected
set can be incorporated into the procedures described below to evaluate asbestos-
related risk. Consistent with the evidence presented elsewhere in this report of a
different potency for chrysotile and amphibole (Chapter 6), separate risk coefficients are
recommended for chrysotile and the amphiboles.
Importantly, the risk coefficients provided in Tables 6-29 and 6-30 are adjusted to be
specifically combined with exposure estimates expressed in terms of the interim
exposure index defined by Equation 7-13. Such exposure estimates include only
asbestos structures longer than 5 pm and thinner than 0.5 pm with those longer than
10 pm weighted more heavily (as described in the equation). Combining the risk
coefficients with measurements expressed in terms of any other index of exposure
(such as the PCME index in current use) wilJ result in invalid estimates of risk. Note, as
indicated in Appendix B, all structures satisfying the dimensional criteria for this Index
(and exhibiting the requisite mineralogy) should be included in analyses for determining
asbestos concentrations; it is not necessary (for example) to distinguish between true
asbestiform fibers and cleavage fragments when the interim index is employed.
Three options are described below for evaluating asbestos-related lung cancer and
mesothelioma risks using the risk coefficients defined in Tables 6-29 or 6-30 with
appropriately derived exposure estimates (expressed in terms of the interim exposure
index defined by Equation 7-13). Additional considerations for determining asbestos
exposure concentrations are also provided.
8.8
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Ontion 1: Using the EPA Dose-Response Models
If there is interest in estimating risk for a particular population that involves time-varying
exposures (or a cohort that potentially experiences background mortality rates
substantially different from the general population), the EPA models for lung cancer
(Equation 6.2) and for mesothelioma (Equation 6.11, for constant exposure, or Equation
6.12 for time-varying exposure) may be directly used to assess risk. This requires
knowledge of both the age-specific background lung cancer mortality rate for the
population of interest and the age-specific total mortality rate for this population in
addition to the data describing the overall pattern of exposure. The mortality rates can
then be used in a lifetable analysis, along with the given exposure pattern and the
appropriate values of Kr and Kw. (either from Table 6-29 or 6-30) to estimate the
additional probability of dying of lung cancer or mesothelioma. Such a lifetable analysis
is described in Appendix D.
Option 2: Estimating Risk from a Risk Tabte
Once it is decided which of the two sets of recommended risk coefficients (Table 6-29
or 6-30) to adopt for general use, a risk table can be constructed (using general U.S.
mortality rates) that can then be applied broadly to assess risk. To illustrate this
approach, Table 8-1 is a risk table (for lifetime, continuous exposure) that was
constructed using the risk coefficients from Table 6-30 (i.e. the "conservative" values).
Note that the values in the table have already been adjusted for lifetime, continuous
exposure (so conversions from an occupational exposure scenario are not required).
The only data required to assess risks using this risk table are estimates of long-term
average exposure {derived from appropriate measurements, as described below) for
each particular exposure scenario and population of interest.
Table 8-1 presents estimates of the additional risk of death from lung cancer and
mesothelioma attributable to lifetime exposure at an asbestos concentration of 0.0005
f/ml (for fibrous structures longer than 5 pm and thinner than 0.5 pm) as determined
using TEM methods recommended below. In the table, separate risk estimates are
provided for males and females and for smokers and non-smokers. Separate estimates
are also presented for exposures containing varying fractions (in percent) of fibrous
structures longer than 10 pm.
e.9
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Separate estimates are presented for smokers and nonsmokers because the lifetime
asbestos-induced risk of both lung cancer and mesothelioma differ between smokers
and non-smokers. The asbestos-induced risk of lung cancer is higher among smokers
because the lung cancer model (Equation 6.2) assumes that the increased mortality
rate from lung cancer risk due to asbestos exposure is proportional to background lung
cancer mortality, which is higher among smokers. This is consistent with the
muitplicative effect between smoking and asbestos exposure reported by several
researchers (see, for example, Hammond et al. 1979).
The asbestos-induced risk of mesothelioma is smaller among smokers because the
mesothelioma model (Equation 6.11) assumes that risk from constant exposure
increases with the cube of age, with the result that the predicted mortality rate is highest
among the elderly. Thus, since smokers have a shorter life span than non-smokers,
their risk of dying from mesothelioma is also predicted to be smaller.
8.10
Revision 1 - 9/3/01
-------
TABLE 8-XXX:
ADDITIONAL RISK PER ONE HUNDRED THOUSAND PERSONS FROM L\pf I\
EXPOSURE TO 0.0005 TEM f/cc LONGER THAN 5.0 pm AND THINNER THAN 0.5 (*n
Pefcent of Fibers Greater Than 10 pm in Length
0.05 0.10
0.50
1.00
2.00
5.00 10.00 20.00 50.00 100 in
MALE NON-SMOKERS
Lung Cancer 0.011 0.013 0.015 0.030 0.05
Mesothelioma 0.004 0.005 0.005 0.011 0.02
Combined 0.015 0.018 0.021 0.D41 0.07
CHRYSOTILE
0.09
0.03
0.12
0.20
0.07
0.27
0.39
0.14
0.53
0.77
0.27
1.04
1.91
0.67
2.58
3 8'
1 ?. '
5 M
FEMALE NON-SMOKERS
Lung Cancer 0.008 0.010 0.011 0.022 0.04 0.06 0.14 0.28 0.55 1.37 2 7a
Mesothelioma 0.004 0.005 0.006 0.012 0.02 0.03 0.08 0.15 0.30 0.74 1 4i
Combined 0.013 0.015 0.017 0.034 0.05 0.10 0.22 0.43 0.85 2.11 4 2;
MALE SMOKERS
Lung Cancer 0.097 0.112 0.128 0.256 0.42 0.74 1.70 3.29 6.49 16.08 32 Oi
Mesothelioma 0.003 0.003 0.004 0.007 0.01 0.02 0.05 0.09 0.18 0.45 0 9(
Combined 0.099 0.116 0.132 0.264 0.43 0.76 1.74 3.39 6.67 16.53 32 «*
FEMALE SMOKERS
Lung Cancer 0.067 0.078 0.089 0.178 0.29 0.51 1.18 2.29 4.51 11.18 22 ?¦>
Mesothelioma 0.004 0.005 0.005 0.011 0.02 0.03 0.07 0.14 0.27 0.66 1 3:
Combined 0.071 0.083 0.095 0.189 0.31 0.54 1.25 2.42 4.78 11.84 23 61
MALE NON-SMOKERS
Lung Cancer 0.04 0.05 0.05 0.11 0.17
Mesothelioma 2.01 2.34 2.67 5.33 8.65
Combined 2.047 2.386 2.725 5.437 8.83
AMPHIBOLE
0.31
15.30
15.61
0.71
35.24
35.94
1.37
68.45
69.82
2.70
134.83
137.53
6.68
333.61
340.28
13 2(<
663 G'.
676 91
FEMALE NON-SMOKERS
Lung Cancer 0.03 0.03 0.04 0.08 0.13 0.22
Mesothelioma 2.23 2.60 2.97 5.92 9.61 16.99
Combined 2.257 2.631 3.005 5.995 9.73 17.21
MALE SMOKERS
Lung Cancer 0.38 0.45 0.51 1.02 1.66 2.93
Mesothelioma 1.36 1.58 1.81 3.61 5.86 10.35
Combined 1.742 2.031 2.319 4.628 7.51 13.29
FEMALE SMOKERS
Lung Cancer 0.27 0.32 0.36 0.72 1.17 2.07
Mesothelioma 1.98 2.31 2.64 5.27 8.55 15.12
Combined 2.255 2.628 3.002 5.989 9.72 17.19
0.52
39.12
39.64
6.75
23.84
30.60
4.76
34.83
39.59
1.00
75.99
77.00
13.12
46.32
59.44
9.25
67.66
76.92
1.98
149.68
151.66
25.84
91.23
117.08
4.89
370.33
375.22
63.91
225.72
289.63
18.23 45.10
133.27 329.68
151.50 374.78
9 71
736 6(,
746 3/
127.01.
449 00
576 00
89 71)
655 6b
745 3?.
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Table 8-1
fi 11
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Separate estimates are provided for different fractions of fibrous structures longer than
10 pm because the model assumes that structures longer than 10 pm are more potent
than structures between 5 and 10 pm in length (in a manner consistent with
Equation 7.13). The derivation of this model is described in detail in Chapters 6 and 7.
Risks from lifetime exposures to asbestos levels other than 0.0005 may be estimated
from the appropriate entry in Table 8-1 by multiplying the value in the selected cell from
the Table by the airborne asbestos concentration of interest and dividing by 0.0005
(i.e., by assuming that the additional risk is proportional to the asbestos exposure level).
Airborne asbestos concentrations to be used in this manner must be estimates of
lifetime average exposure and must be expressed as structures longer than 5 pm and
thinner than 0.5 pm derived as described below. Estimates of the fraction of these
structures that are also longer than 10 pm must also be determined to select the
appropriate cell of the table from which to derive the risk estimate. Note that the two
size fractions that are combined to determine Casb (Equation 7.13) are separately
enumerated (not combined) when they are to be used in conjunction with Table 8-1.
The procedure described above for estimating risks using Table-8-1 should provide
good approximations as long as the projected risk is no greater than 1,000 per 100,000.
Risks greater than 1,000 per 100,000 (i.e. 1 in 100) that are derived from the Table are
likely to be over-estimated. However, for risks associated with short-duration
exposures or exposures that differ radically over time, it may be better to use a lifetable
analysis or a modified version of Table 8-1. This is to avoid substantially under- or
over-estimating risk (depending on how the table might otherwise be applied).
Table 8-1 was derived using the approach described in Appendix D by incorporating the
age-, sex-, and smoking-specific death rates reported for the general U.S. population
and assuming that exposure is constant and continuous at the level indicated in the
table. The underlying models are provided in Chapter 6 for cases in which exposure is
either not constant or not continuous and for which sufficient data exist to characterize
the time-dependence of such exposure. If available, there may also be cases in which
it is advantageous to employ mortality data from a control population that better
matches the exposed population of interest than the U.S. population as a whole.
Option 3: Estimating Risk from a Unit Risk Factor
A third option for estimating risks based on the approach recommended in this
document would be to derive a combined, unit risk factor for asbestos cancer risk
similar to the unit risk factor currently recommended by EPA. The new unit risk factor
would be derived using the selected set of new risk coefficients (either those in Table 6-
29 or Table 6-30) and would be appropriate for use with exposure estimates expressed
8.12
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
in terms of the interim exposure index (rather than the PCME index, as currently
recommended). As with the unit risk factor currently in use, some estimate of the
fraction of smokers in the general population would also need to be provided as an
input. This approach, however, may prove to be somewhat less flexible than the
second option above and therefore may be prone to greater errors.
Renuirements for Asbestos Measurements
One additional advantage of the approach for evaluating asbestos-related risks
recommended in this document (in comparison to the current approach) is that the
procedure for assessing risks is tied unambiguously to a specific index for measuring
and expressing exposure (i.e. the interim index) and this, in turn, is tied unambiguously
to requirements for analyzing asbestos samples.
Estimates of airborne asbestos concentrations that are required to support risk
assessment can be derived either by extrapolation from airborne measurements or by
modeling release and dispersion of asbestos from sources (soils or other bulk
materials). In either case, exposure estimates must be representative of actual (time-
dependent or time-integrated) exposure and must provide measurements of the specific
size fractions of asbestos that are components of the interim exposure index defined by
Equation 7.13. Additional considerations that need to be addressed to assure the
validity of risk estimates derived using this protocol include:
• the array of samples collected for estimating airborne asbestos
concentrations must be representative of the exposure environment;
• the time variation of airborne asbestos concentrations must be properly
addressed;
• airborne samples must to be collected on membrane filters that are
suitable for preparation for analysis by transmission electron microscopy
(TEM). Appropriate procedures for sample collection are described in
Chatfield and Berman (1990) or the ISO Method (ISO 10312)1;
Note that the ISO Method (ISO 10312) is a refinement of the method
originally published as the Interim Superfund Method (Chatfield and
Berman 1990). It incorporates improved rules for evaluating fiber
morphology. Both methods derive from a common development effort
headed by Eric Chatfield.
8.13
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
sample filters must be prepared for analysis using a direct transfer
procedure (e.g. ISO 10312). Should indirect preparation be required (due,
for example, to problems with overloading of sample filters), a sufficient
number of paired samples will need to be collected and analyzed to
establish a site-specific correlation between directly and indirectly
prepared samples;
samples must be analyzed by TEM;
samples must be analyzed using the counting and characterization rules
defined in ISO 10312 with one modification: only structures longer than 5
pm and thinner than 0.5 pm need to be enumerated. Separate scans for
counts of total structures longer than 5 pm and longer than 10 pm are
recommended to increase the precision with which the longest structures
are enumerated. Importantly, ISO Method rules require separate
enumeration and characterization of component fibers and bundles that
are observed within more complex clusters and matrices. Such
components, if they meet the dimensional criteria defined in Equation 7.13
must be included in the structure count;
if risks are to be estimated using the risk models (as described in the first
option), asbestos concentrations derived from the above-described
measurements must be expressed as the weighted sum of structures
between 5 and 10 pm in length and structures longer than 10 pm in
length, per the exposure index defined in Equation 7.13. Only structures
thinner than 0.5 pm are to be included in these counts. Both fibers and
bundles that are isolated structures and fibers and bundles that are
components of more complex structures are to be included in structure
counts (as long as each structure counted satisfies the defined size
criteria for the size category in which it is included);
if risks are to be estimated using the risk models (the first option), the risk
coefficient(s) selected either from Table 6-29 or 6-30 must be appropriate
for the fiber type (i.e. chrysotile or amphibole) and the disease end point
(i.e. lung cancer or mesothelioma) relevant to the situation of interest; and
if risks are to be estimated using Table 8-1 (The second option), rather
than deriving the weighted sum described in Equation 7.13, the
concentration of asbestos structures longer than 10 pm and thinner than
0.5 pm must be derived to determine the appropriate column of the Table
8.14
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
from which to estimate risk and the concentration of total asbestos
structures longer than 5 pm and thinner than 0.5 pm must be derived,
divided by 0.0005, and multiplied by the risk estimate listed in the
appropriate cell of the Table to generate the risk estimate of interest.
Considerations that need to be addressed to assure the validity of risk estimates
derived from soil or bulk measurements combined with release and transport modeling
include:
• the array of samples collected for estimating source concentrations must
be representative of the surface area or volume of source material from
which asbestos is expected to be released and contribute to exposure;
• samples must to be prepared and analyzed using the Modified Elutriator
Method for soils and bulk materials (Berman and Kolk 1997, 2000), which
is the only method capable of providing bulk measurements that-can be
related to risk;
• membrane filters samples prepared using the tumbler and vertical
elutriator per the Superfund method must themselves be prepared for
TEM analysis using a direct transfer procedure;
• TEM analysis must be conducted using the counting and characterization
rules defined in the ISO Method (ISO 10312) in precisely the same
manner that is described above for air measurements. Also, the same
size categories need to be evaluated in the same manner described
above, whether results are to be used to support assessment using risk
models or using the risk table; and
• release and dispersion models that are selected for assessing risks must
be appropriate to the exposure scenario and environmental conditions of
interest. Such models must also be adapted properly so that they accept
input estimates expressed in terms of fiber number concentrations.
Procedures suggested for adapting such models are illustrated in a recent
publication (Berman 2000).
Note, if we design analytical procedures to focus on long structures, we can evaluate
risks cost-effectively. Moreover, even to the extent that shorter structures or other
structures might contribute to risk should generally be addressed by default. The
alternate approach of spending most time and money counting lots of (potentially non-
8.15
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
potent or marginally potent) short structures while not characterizing longer structures
with adequate sensitivity or precision (which is embodied in the current EPA approach)
leaves the much greater and more probable danger of falsely missing potentially
serious hazards because a small population of extremely potent, long fibers were
missed in a particular environment.
8.3 RECOMMENDATIONS FOR FURTHER STUDY
A small number of limited and focused studies (described previously) are previously
recommended in this document because they are likely to provide very cost-effective
improvements to the quality of this document and may support substantial improvement
to the recommended procedures for assessing asbestos-related risks. The
recommended studies are:
(1) a focused study to expand our evaluation of the ability of the current EPA
models to adequately track the time-dependence of disease (Sections
6.2.2 and 6.3.2); and
(2) a focused study to develop the supporting data needed to define
adjustments for potency factors that will allow them to be used with an
exposure index that even more closely captures asbestos characteristics
that determine biological activity than the currently proposed "interim
index" (Section 7.5).
Note that, by properly designing the second of the above-listed studies, it may also be
possible to further address another outstanding issue that was previously identified: the
question of whether dose-response coefficients derived from mining studies are under-
estimated relative to studies involving asbestos products because exposures in the
mining studies may contain large numbers of non-asbestos particles contributed by the
disturbance of host rock (Section 6.2.3).
For the first of the above-recommended studies, we would need to acquire the raw data
from at least one additional cohort exposed primarily to chrysotile and one additoinal
cohort exposed primarily to amphiboles to determine whether our observation that the
current EPA model for lung cancer adequately describes the time dependence of
relative risk associated with exposure to amphiboles but not to chrysotile. Should this
finding be confirmed, the potential magnitude and importance of the deviation would
also be evaluated along with a candidate adjustment to the model that would likely
solve the problem. Simultaneous, additional analysis of the mesothelioma model may
also be useful.
8.16
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Candidate cohorts that may be useful for this evaluation, if they can be acquired,
include: (for amphibole exposures) the lung cancer data and newest mesothelioma data
from Libby (best) or, potentially the lung cancer and mesothelioma data from the
Paterson, New Jersey insulation manufacturing plant studied by Seidman and (for
chrysotile exposures) the lung cancer data from Quebec (best) or, potentially, the lung
cancer and mesothelioma data from the New Orleans asbestos-cement pipe plant
studied by Hughes et al. The possibility of obtaining some or all of these data sets
needs to be further explored.
For the second of the above-listed studies, we would need to obtain suitable bulk
samples of raw ore from the Wittenoom, Libby, and Quebec mines and bulk samples of
textile grade (and several other grade) products from the Quebec mines. We have
already approached individuals who can provide access to these materials and they
have all expressed interest in collaborating.
Ideally, a well-designed field sampling plan would be used to collect and composite raw
ore samples and multiple samples (from widely dispersed time-periods of production)
would be collected and composited for the fiber product samples. The resulting, limited
number of well-constructed composite samples would then be prepared using the
Modified Elutriator Method (Berman and Kolk 2000) and subjected to a rigorous
analysis by TEM (covering counting and characterization of a very broad range of
fibrous structures and non-fibrous particles). The data from these analyses would then
be used both to evaluate questions concerning the quality of exposure at mining sites
and to adjust the dose-response coefficients derived from studies of these mining sites
(and several other studies of facilities that used the products from Quebec). The
coefficients would be adjusted to a better optimized exposure index (in which fibers at
least as long as 20 pm would be separately weighted) and the degree to which use of
this new index improves agreement across certain selected studies would be evaluated.
8.17
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
9.0 REFERENCES
Aalto, M; and Heppleston, AG; "Fibrogenesis by Mineral Fibres: An In-Vitro Study of
the Roles of the Macrophage and Fibre Length." Br J Exp Path. Vol. 65, pp. 91-99.
1984.newnew
Adamson, lYR; "Early Mesothelial Cell Proliferation After Asbestos Exposure: In Vivo
and In Vitro Studies." Environmental Health Perspectives. Vol. 105, Supplement 5, pp.
1205-1208. September, 1997.newnew
Afaq, F; Abidi, P; Matin, R; and Rahman, Q; "Activation of Alveolar Macrophages and
Peripheral Red Blood Cells in Rats exposed to Fibers/Particles." Toxicology Letters,
Vol. 99, pp. 175-182, 1998.newnew
Albin, M; Magnani, C; Krstev, S; Rapiti, E; Shefer, I; "Asbestos and Cancer: An
Overview of Current Trends in Europe." Environ Health Perspectives. Vol. 107, No. 2,
pp. 289-298. May, 1999.newnew
Albin, M; Pooley, FD; Stromberg, U; Attewell, R; Mitha, R; Johansson, L; and Welinder,
H; "Retention Patterns of Asbestos Fibres in Lung Tissue Among Asbestos Cement
Workers." Occup and Environ Med. Vol. 51, pp. 205-211. 1994.newnew
Albin, M; Jakobsson, K; Attewell, R; Johansson, L; Welinder, H; "Mortality and Cancer
Morbidity in Cohorts of Asbestos Cement Workers and Referents." Br J Ind Med. Vol.
79, No. 9, pp. 602-610. September, 1990.newnew
Amandus, H.E., Wheeler, R., "The Morbidity and Mortality of Vermiculite Miners and
Millers Exposed to Tremolite-Actinolite: Part II. Mortality." American Journal of
Industrial Medicine 11:15-26,1987.
Amandus, H.E., Wheeler, R., Jankovic, J., Tucker, J., The Morbidity and Mortality of
Vermiculite Miners and Millers Exposed to Tremolite-Actinolite: Part I. Exposure
Estimates." American Journal of Industrial Medicine 11:1-14,1987
Armstrong, B.K., DeKlerk, N.H., Musk, A.W., Hobbs, M.S.T., "Mortality in Miners and
Millers of Crocidolite in Western Australia." British Journal of Industrial Medicine, Vol.
45, pp. 5-13,1988.
Asgharian, B. and Yu, C.P., Journal of Aerosol Medicine. Vol. 1, pp. 37-50,1988.
9.1
Revision 1 - 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Barchowsky, A; Lannon, BM; Elmore, LC; Treadwell, MD; "Increased Focal Adhesion
Kinase- and Urokinase-type Plasminogen Activator Receptor-Associated Cell Signaling
in Endothelial Cells Exposed to Asbestos." Environmental Health Perspectives. Vol.
105, Supplements, pp. 1131-1137. September, 1997.newnew
Barchowsky, A; Roussel, RR; Krieser, RJ; Mossman, BT; and Treadwell, MD;
"Expression and Activity of Urokinase and its Receptor in Endothelial and Pulmonary
Epithelial Cells Exposed to Asbestos." Toxicology and Applied Pharmacology. Vol.
152, No. 2, pp. 388-396. 1998.
Baris, Y.I., Simonato, L., Artvinli, M., Pooley, F., Saracci, R., Skidmore, J., Wagner, C.,
"Epidemiological and Environmental Evidence of the Health Effects of Exposure to
Erionite Fibres: A Four-Year Study in the Cappadocian Region of Turkey." International
Journal of Cancer, Vol. 39, pp. 10-17. 1987.
Bauman, MD; Jetten, AM; Bonner, JC; Kumar, RK; Bennett, RA; and Brody, AR;
"Secretion of a Platelet-Derived Growth Factor Homologue by Rat Alveolar
Macrophages Exposed to Particulates in Vitro." European Journal of Cell Biology. Vol.
51, pp. 327-334. 1990.newnew
Beckett, S.T., "The Generation and Evaluation of UICC Asbestos Clouds in Animal
Exposure Chambers." Annals of Occupational Hygiene, Vol. 18, pp. 187-198, 1975.
Bellman, B., Konig, H., Muhle, H., Pott, F., "Chemical Durability of Asbestos and of
Man-Made Mineral Fibres In Vivo," Journal of Aerosol Science, Vol. 17, No, 3, pp. 341-
345, 1986.
Bellman, B., Muhle, H., Pott, F., Konig, H., Kloppeel, H., Spumy, K., "Persistence of
Man-Made Fibers (MMF) and Asbestos in Rat Lungs." Annals of Occupational Hygiene,
Vol. 31, pp. 693-709, 1987.
Berman, DW "Asbestos Measurement in Soils and Bulk Materials: Sensitivity,
Precision, and Interpretation - You Can Have It All." In Advances in Environmental
Measurement Methods for Asbestos, ASTM STP 1342, M.E. Beard, H.L. Rook, Eds.
American Society for Testing and Materials. 2000.
Berman, D.W. and Chatfield, E.J.; Interim Superfund Method for the Determination of
Asbestos in Ambient Air. Part 2: Technical Background Document, Office of Solid
Waste and Remedial Response, U.S. EPA, Washington, D.C., EPA/540/2-90/005b,
May, 1990.
9.2
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Berman, DW and Crump KS; "Methodology for Conducting Risk Assessments at
Asbestos Superfund Sites. Part 2: Technical Background Document." Prepared for
Kent Kitchingman, U.S. Environmental Protection Agency, Region 9. Work Assignment
No. 59-06-D800 under Contract No. 68-W9-0059. 1999.
Berman, D.W. and Kolk, A.J.;. Superfund Method for the Determination of Asbestos in
Soils and Bulk Materials, Office of Solid Waste and Emergency Response, U.S. EPA,
Washington, D.C., EPA 540-R-97-028, 1997.
Berman, D.W. and Kolk, A.J.;. Draft: Modified Elutriator Method for the Determination
of Asbestos in Soils and Bulk Materials, unpublished. Aeolus, Inc. 751 Taft St., Albany,
CA 94706. 2000.
Berman, D.W.; Crump, K.S.; Chatfield, E.J.; Davis, J.M.G.; and Jones, A.D.; "The
Sizes, Shapes, and Mineralogy of Asbestos Structures that Induce Lung Tumors or
Mesothelioma in AF/HAN Rats Following Inhalation." Risk Analysis, Vol. 15, No. 2,
1995.
Bernstein, DM; Morscheidt, C; Grimm, H-G; Thevenaz, P; Teichert, U; "Evaluation of
Soluble Fibers Using the Inhalation Biopersistence, a Nine-Fiber Comparison"
Inhalation Toxicology Vol.8, pp.345-385, 1996.newnew
Berry, G., Newhouse, M.L., "Mortality of Workers Manufacturing Friction Materials
Using Asbestos." British Journal of Industrial Medicine, Vol. 40, pp. 1-7,1983.
Bertrand, R., Pezerat, H., "Fibrous Glass: Carcinogenicity and Dimensional
Characteristics." Biological Effects of Mineral Fibres, Wagner, J.C., ed., IARC Scientific
Publications, pp. 901-911,1980.
Bignon, J., Monchaux, G., Sebastien, P., Hirsch, A., Lafuma, J., "Human and
Experimental Data on translocation of Asbestos Fibers Through the Respiratory
System." Annals of New York Academy of Sciences, Vol. 330, pp. 745-750, 1979.
Blake, T; Castranova, V; Schwegler-Berry, D; Baron, P; Deye, GL; Li, C; and Jones, W;
"Effect of Fiber Length on Glass Microfiber Cytotoxity" Journal of Toxicology and
Environmental Health Vol.54, pp.243-259, Part A, 1998.newnew
Bolton, R.E., Davis, J., Donaldson, K., Wright, A., "Variation In the Carcinogenicity of
Mineral Fibres." Annals of Occupational Hygiene, Vol. 26, No. 1-4, pp. 569-582,1982.
9.3
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Bolton, R.E., Davis, Miller, B., Donaldson, K., Wright, A., The Effect of Dose of
Asbestos on Mesothelioma Production in the Laboratory Rat." Proceedings of the Vlth
International Pneumonoconiosis Conference, Vol. 2, pp. 1028-1035,1984.
Bolton, R.E., Addison, J., Davis, J.M.G., Donaldson, K., Jones, A.D., Miller, B.G.,
Wright, A., "Effects of the Inhalation of Dusts from Calcium Silicate Insulation Materials
in Laboratory Rates." Environmental Research, Vol. 39, pp. 26-43, 1986.
Bonneau, L., Malard, C., Pezerat, H., "Studies on Surface Properties of Asbestos,"
Environmental Research, Vol. 41, pp. 268-275,1986.
Boutin, C; Dumortier, P; Rey, F; Viallat, JR; De Vuyst, P; "Black Spots Concentrate
Oncogenic Asbestos Fibers in the Parietal Pleura: Thoracoscopic and Mineralogic
Study." American Journal of Respiratory and Critical Care Medicine. Vol. 153, pp. 444-
449. 1996.newnew
Brass, DM; Hoyle, GW; Poovey, HG; Liu, J-Y; Brody, AR; "Reduced Tumor Necrosis
Factor-Alpha and Transforming Growth Factor-Beta-1 Expression in the Lungs of
Inbred Mice That Fail to Develop Fibroproliferative Lesions Consequent to Asbestos
Exposure." American Journal of Pathology. Vol. 154, No. 3, pp. 853-862.
1999.newnew
Broaddus, VC; Yang, L; Scabo, LM; Ernst, JD; Boylan, AM; "Crocidolite Asbestos
Induces Apoptosis of Pleural Mesothelial Cells: Role of Reactive Oxygen Species and
Poly(ADP-ribosyl) Polymerase." Environmental Health Perspectives. Vol. 105,
Supplements, pp. 1147-1152. September, 1997.newnew
Brody, A.R., Hill, L.H., Adkins, B., O'Connor, R.W., "Chrysotile Asbestos Inhalation in
Rats: Deposition Pattern and Reaction of Alveolar Epithelium and Pulmonary
Macrophages." American Review of Respiratory Diseases, Vol. 123, pp. 670-679, 1981.
Brody, AR; Liu, J-Y; Brass, D; Corti, M; "Analyzing the Genes and Peptide Growth
Factors Expressed in Lung Cells In Vivo Consequent to Asbestos Exposure."
Environmental Health Perpsectives. Vol. 105, Supplement 5, pp. 1165-1171.
September, 1997.newnew
Brown, DM; Fisher, C; and Donaldson, K; "Free Radical Activity of Synthetic Vitreous
Fibers: Iron Chelation Inhibits hydroxyl Radical Generation by Refractory Ceramic
Fiber." Journal of Toxicology and Environmental Health Part A. Vol. 53, No. 7, pp.
545-561. 1998.newnew
9.4
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Campbell, W.J., Huggins, C.W., Wylie, A.G., "Chemical and Physical Characterization of Amosite,
Chrysolite, Crocidolite, and Nonfibrous Tremolite for Oral Ingestion Studies by the National Institute of
Environmental Health Sciences." Report of Investigations 8452, National Library of Natural
Resources, U.S. Department of the Interior. 1980.(3)
Case BW and Dufresne A; "Asbestos, Asbestosis, and Lung Cancer Observations in
Quebec Chrvsotile Workers." Environmental Health Perspectives. Vol. 105,
Supplements, pp. 1113-1120. September, 1997.newnew
Castranova, V, Vallyathan, V; Ramsey, DM; McLaurin, JL; Pack, D; Leonard, S; Barger,
MW; Ma, JUC; Dalai, NS; and Teass, A; "Augmentation of Pulmonary Reactions to
Quartz Inhalation by Trace Amounts of Iron-containing Particles." Environmental Health
Perspectives. Vol. 105, Supplements, pp. 1319-1324. September, 1997.newnew
Chang, L-Y; Overby, LH; Broday, AR; and Crapo, JD; "Progressive Lung Cell Reactions
and Extracellular Matrix Production After a Brief Exposure to Asbestos." Am J Path.
Vol. 131. No. 1, Pp. 156-170. 1989. OLD
NEED CHAOETAL. 1996
CHAP (Chronic Hazard Advisory Panel on Asbestos), "Report to the U.S. Consumer
Product Safety Commission." July 1983.
Chatfield, E.J. and Berman, D.W.; Interim Superfund Method for the Determination of
Asbestos in Ambient Air. Part 1: Method, Office of Solid Waste and Remedial
Response, U.S. EPA, Washington, D.C., EPA/540/2-90/005a, May, 1990.
Chen, Y.K., Ph.D. Dissertation, Mechanical and Aerospace Engineering, State
University of New York at Buffalo, 1992.
Chen, Y.K. and Yu, C.P., "Deposition of Charged Fiber in the Human Lung." Journal of
Aerosol Science Vol 24, Suppl. 1, pp. 5459-5460,1993.
Cherrie, J.W., Dodgson, J., Groat, S., and Carson, M., "Comparison of Optical and
Electron Microscopy for Evaluating Airborne Asbestos." Institute of Occupational
Medicine, Edinburgh, (U.S. Department of Commerce), 1979.
Chesson, J., Rench, J.D., Schultz, B.D. and K.L. Miine, "Interpretation of Airborne
Asbestos Measurements." Submitted for publication, 1989.
9.5
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Cherrie, J.W., Dodgson, J., Groat, S., Carson, M., "Comparison of Optical and Electron
Microscopy for Evaluating Airborne Asbestos." Inst. Occupational Medicine, Edinburgh,
U.S. Dept. of Commerce, 1987.
Choe, N; Zhang, J; Iwagaki, A; Tanaka, S; Hemenway, DR; and Kagan, E; "Asbestos
Exposure Upregulates the Adhesion of Pleural Leukocites to Pleural Mesothelial Cells
via VCAM-1." Am J Physiology Vol. 277, No. 2, Part 1, pp. L292-L300. 1999 newnew
Choe, N; Tanaka, S, Xia, W; Hemenway, DR; Roggli, VI; and Kagan, E; "Pleural
Macrophage Recruitment and Activation in Asbestos-Induced PJeural Injury."
Environmental Health Perspectives. Vol. 105, Supplement 5, pp. 1257-1260.
September, 1997.newnew
Choe, N; Tanaka, S; Kagan, E; "Asbestos Fibers and lnterleukin-1 Upregulate the
Formation of Reactive Nitrogen Species in Rat Pleural Mesothelial Cells." American
Journal of Resp Cell and Molecular Biology. Vol. 19, No. 2, pp. 226-236. 1998.
Churg, A., Wiggs, B., Depaoli, L. Kampe, B., Stevens, B., "Lung Asbestos Content in
Chrysotile Workers with Mesothelioma." American Review of Respiratory Disease, Vol.
130, pp. 1042-1045, 1984.
Coin, PG; Roggli, VL; Brody, AR; "Deposition, Clearance, and Translocation of
Chrysotile Asbestos from Peripheral and Central Regions of the Rat Lung." Environ
Research. Vol. 58, pp. 970-116. 1992.newnew
Coin, PG; Roggli, VL; and Brody, AR; "Persistence of Long, Thin Chrysotile Asbestos
Fibers in the Lungs of Rats" Environmental Health Perspectives Vol.102, Suppl.5,
pp.197-199, 1994.newnew
Costa, DL; Dreher, KL; "Bioavailable Transition Metals in Particulate Matter Mediate
Cardiopulmonary Injury in Healthy and Compromised Animal Models." Environmental
Health Perspectives. Vol. 105, Supplement 5, pp. 1053-1060. September,
1997.newnew
Cox D. and Lindley, D., Theoretical Statistics. Chapman and Hall, London, 1974.
Cox and D. Oakes, Analysis of Survival Data. Chapman and Hall, London, 1984.
9.6
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Crump et al. 1985 - ref description of exposures in textiles (grades of fibers)
NEED
Davis, JMG; Cowie, HA; The Relationship Between Fibrosis and Cancer in
Experimental Animals Exposed to Asbestos and Other Fibers." Environmental Health
Perspectives. Vol. 88, pp. 305-309. 1990.newnew
Davis, Beckett, S.T., Bolton, R.E., Collings, P., Middleton, A .P., "Mass and
Number of Fibres in the Pathogenesis of Asbestos-Related Lung Disease in Rats."
British Journal of Cancer, Vol. 37, pp. 673-688,1978.
Davis, Beckett, S.T., Bolton, R.E., Donaldson, K., "A Comparison of the
Pathological Effects in Rats of the UICC Reference Samples of Amosite and Chrysotile
with Those of Amosite and Chrysotile Collected from the Factory Environment."
Biological Effects of Mineral Fibres, Wagner, J.C., ed., IARC Scientific Publications, pp.
288-292, 1980.
Davis, J.M.G., Addison, J., Bolton, R.E., Donaldson, K., Jones, A.D., Miller, B.G.,
"Inhalation Studies on the Effects of Tremolite and Brucite Dust in Rats."
Carcinogenesis, Vol. 6, No. 5, pp. 667-674, 1985.
Davis, J.M.G., Addison, J., Bolton, R., Donaldson, K., Jones, A.D., Smith, T., The
Pathogenicity of Long Versus Short Fibre Samples of Amosite Asbestos Administered
to Rats by Inhalation and Intraperitoneal Injection." British Journal of Experimental
Pathology, Vol. 67, pp. 415-430, 1986a.
Davis, J.M.G., Gylseth, G., Morgan, A., "Assessment of Mineral Fibres from Human
Lung Tissue." Thorax, Vol. 41, pp. 167-175,1986b.
Davis, J.M.G., Bolton, R.E., Brown, D., Tully, H.E., "Experimental Lesions in Rats
Corresponding to Advanced Human Asbestosis." Exposure Molecular Pathology, Vol.
44, No. 2, pp, 207-221, 1986c.
Davis, J.M.G., Addison, J.t Bolton, R.E., Donaldson, KM Jones, A.D., "Inhalation and
Injection Studies in Rats Using Dust Samples from Chrysotile Asbestos Prepared by a
Wet Dispersion Process." British Journal of Pathology, Vol. 67, pp. 113-129,1986d.
Davis, J.M.G., Jones, A.D., Smith, T., "Comparisons of the Pathogenicity of Long and Short Fibres of
Chrysotile Asbestos in Rats." Institute for Research and Development of Asbestos, Montreal, ed.,
Institute of Occupational Medicine, publ., Report No. TM-87/08,1987.(2)
9.7
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Davis, Bolton, R.E., Douglas, A.N., Jones, A.D., Smith, T., "Effects of
Electrostatic Charge on the Pathogenicity of Chrysotile Asbestos." British Journal of
Industrial Medicine, Vol. 45, No. 5, pp. 292-309, 1988a.
Davis, J.M.G., Jones, A.D., "Comparisons of the Pathogenicity of Long and Short
Fibres of Chrysotile Asbestos in Rats." British Journal of Exposure Pathology, Vol. 69,
No. 5, pp. 171-738, 1988b.
Davis, JMG; Addison, J; Mcintosh, C; Miller, BG; Niven, K; "Variations in the
Carcinogenicity of Tremolite Dust Samples of Differing Morphology." Annals New York
Academy of Sciences." pp. 473-490.1991. newnew
DeKlerk, NH; Musk, AW; Armstrong, BK; and Hobbs, MST; "Diseases in Miners and
Millers of Crocidolite from Wittenoom, Western Australia: A further Followup to
December 1986." Annals of Occupational Hygiene, Vol. 38, Supplement 1, pp.
647-655. 1994.
Dement, J.M., "Estimation of dose and evaluation of dose-response in a retrospective
cohort mortality study of chrysotile asbestos textile workers." Ph.D. Thesis, The
University of North Carolina at Chapel Hill, 1980.
Dement, J.M. and Brown, D.P.; "Lung Cancer Mortality Among Asbestos Textile
Workers: A Review and Update." Annals of Occupational Hygiene, Vol. 38, No. 4, pp.
525-532, 1994.
Dement, J.M., Brown, D.P., "Cohort Mortality and Case-Control Studies of White Male
Chrysotile Asbestos Textile Workers", Journal of Clean Technology, Environmental
Toxicology, and Occupational Medicine 7:1052-1062, 1998.
Dement, J.M.; Brown, D.P.; and Okun, A.; "Follow-up Study of Chrysotile Asbestos
Textile Workers: Cohort Mortality and Case-Control Analysis." American Journal of
Industrial Medicine, Vol 26, pp. 431-447, 1994.
Dement, J.M., Harris, R.L., "Estimates of Pulmonary and Gastrointestinal Deposition for
Occupational Fiber Exposure." NTIS PB80-149644, U.S. HEW Contract #78-2438,
1979.
Dement, J.M., Harris, R.L., Symons, M.J., Shy, C.M., "Estimates of Dose-Response for
Respiratory Cancer Among Chrysotile Asbestos Textile Workers." Annals Occupational
Hygiene, Vol. 26, No. 1-4, pp. 869-887,1982.
9.8
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Dement, J.M., Harris, R.L., Symons, M J., Shy, C.M., "Exposures and Mortality Among
Chrysotile Workers. Part I: Exposure Estimates." American Journal of Industrial
Medicine, Vol. 4, pp. 399-419, 1983a.
Dement, J.M., Harris, R.L., Symons, M.J., Shy, C.M., "Exposures and Mortality Among
Chrysotile Workers. Part II: Mortality." American Journal of Industrial Medicine, Vol. 4,
pp. 421-433, 1983b.
Doll R, and Peto R.; "Cigarette smoking and bronchial carcinoma: dose and time
relationships among regular smokers and lifelong non-smokers." J Epidemiol
Community Health Vol. 32, pp. 303-313. 1978.
Doll, R., Peto, J., "Asbestos: Effects on Health of Exposure to Asbestos." Health and
Safety Commission, London, United Kingdom, 1985.
Dopp, E; Schiffmann, D; "Analysis of Chromosomal Alterations Induced by Asbestos
and Ceramic Fibers." Toxicoloov Letters (Shannonl Vol. 96-97, No Spec. Issue. Pp.
155-162. 1998.newnew
Driscoll, KE; Carter, JM; Hassenbein, DG; Howard, B; "Cytokines and Particle-Induced
Inflammatory Cell Recruitment." Environmental Health Perspectives. Vol. 105,
Supplement 5, pp. 1159-1164. September, 1997.newnew
NEED EASTES AND HADLEY 1994
Eastes, W; Hadley, JG; "Dissolution of Fibers Inhaled by Rats" Inhalation Toxicology
Vol.7, pp.179-196, 1995.newnew
Eastes, W; Hadley, JG; "A Mathematical Model of Fiber Carcinogenicity and Fibrous
in Inhalation and Intraperitoneal Experiments in Rats" Inhalation Toxicoloov Vol.8,
pp.323-343,1996.newnew
Economou, P; Samet, JM; and Lechner, JF; "Familial and Genetic Factors i the
Pathogenesis of Lung Cancer" In: Epidemiology of Luno Cancer: Samet, JM ed.;
Marcel Dekker, Inc., New York., Chapter 14,1994.
Enterline, P.E., Harley, J., Henderson, V., "Asbestos and Cancer - A Cohort Followed
to Death." Graduate School of Public Health, University of Pittsburgh, 1986.
9.9
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Everitt, Jl; Gelzleichter, TR; Bermudez, E; Mangum, JB; Wong, BA; Janszen, DB;
Moss, OR; "Comparison of Pleural Responses of Rats and Hamsters to Subchronic
Inhalation of Refractory Ceramic Fibers." Environmental Health Perspectives. Vol.
105, Supplement 5, pp. 1209-1213. September, 1997.newnew
Finkelstein, M.M., "Mortality Among Long-Term Employees of an Ontario Asbestos-
Cement Factory." British Journal of Industrial Medicine, Vol. 40, pp. 138-144, 1983.
Finkelstein, M.M., "Mortality Among Employees of an Ontario Asbestos-Cement
Factory", American Review of Respiratory Disease 129:754-761,1984.
Finkelstein, MM; and Dufresne, A; "Inferences on the Kinetics of Asbestos Deposition
and Clearance Amoung Chrysotile Miners and Millers." American Journal Ind Med Vol.
35, No. 4, pp. 401-12. April 1999.newnew
Finkelstein, JN; Johnston, C; Barrett, T; Oberdorster, G; "Particulate-Cell Interactions
and Pulmonary Cytokine Expression." Environmental Health Perspectives. Vol. 105,
Supplement 5, pp. 1179-1182. September, 1997.newnew
Floyd, RA; "Role of Oxygen Free Radicals in Carcinogenesis and Brain Ischemia."
The FASEB Journal. Vol. 4, pp. 2587-2597. June,1990.newnew
Fubini, B; "Surface Reactivity in the Pathogenic Response to Particulates."
Environmental Health Perspectives. Vol. 105, Supplement 5, pp. 1013-1020.
September, 1997.newnew
Gehr, P; Geiser, M; Stone, KC; and Crapo, JD; "Morphometric Analysis of the Gas
Exchange Region of the Lung." In Toxicology of the Luna. 2nd edition, Gardner, DE;
Crapo, JD; and McClellan, RO eds. Raven Press, New York. 1993. Newnew
Ghio, AJ; Kadiiska, MB; Xiang, Q-H; Mason, RP; "In Vivo Evidence of Free Radical
Formation After Asbestos Instillation: An ESR Spin Trapping Investigation." Free
Radical Biology and Medicine. Vol. 24, No. 1, pp. 11-17.1998.newnew
Gibbs, G.W., Hwang, C.Y., "Physical Parameters of Airborne Asbestos Fibres in Various Work
Environments - Preliminary Findings." American Industrial Hygiene Association Journal, Vol. 36, No.
6, pp. 459-466,1975.
Gibbs, G.W. and Hwang, C.Y., "Dimensions of Airborne Asbestos Fibers." In Biological
Effects of Mineral Fibers, Wagner, ed., IARC Scientific Publication, pp. 69-78,1980.
9.10
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Gilbert, 0. Statistical Method for Environmental Pollution Monitoring. Van Nostrand
Reinhold, New York. 1987.
Gold, J; Amandusson, H; Krozer, A; Kasemo, B; Ericsson, T; Zanetti, G; Fubini, B;
"Chemical Characterization and Reactivity of Iron Chelator-Treated Amphibole
Asbestos." Environmental Health Perspectives. Vol. 105, Supplements, pp. 1021-
1030.1997 newnew
Goldstein, B., Rendall, R., Webster, I., "A Comparison of the Effects of Exposure of
Baboons to Crocidolite and Fibrous-Glass Dusts." Environmental Research, Vol. 32,
pp. 334-359, 1983.
Golladay, SA; Park, S-H; and Aust, AE; "Efflux of Reduced Glutathione after Exposure
of Human Lung epithelial Cells to Crocidolite Asbestos." Environmental Health
Perspectives. Vol. 105, Supplement 5, pp. 1273-1278.1997 newnew
Governa, M; Camilucci, L Amati, M; Visona, I; Valentino, M; Botta, GC; Campopiano, A;
Fanizza, C; "Wollastonite Fibers In Vitro Generate Reactive Oxygen Species Able to
Lyse Erythrocytes and Activate the Complement Alternate Pathway." Toxicolooical
Sciences. Vol. 44, No. 1, pp. 32-38. 1998.newnew
Gross, TJ; Cobb, SM; Peterson, MW; "Asbestos Exposure Increases Paracellular
Transport of Fibrin Degradation Products Across Human Airway Epithelium." Am J
Physiol. Vol. 266, No. 3, pp. L287-95. Mar, 1994.newnew
Hammond, E.C., Selikoff, I.J., Seidman, H., 1979. "Asbestos Exposure, Cigarette
Smoking and Death Rates." Annals New York Academy of Sciences, Vol. 330, pp.
473-490,1979.
Harris, R.L., Timbrell, V., "The Influence of Fibre Shape in Lung Deposition -
Mathematical Estimates." Inhaled Particles IV, Walton, W.H., ed., Pergamon Press,
1977.
Hart, GA; Kathman, LM; and Hesterberg, TW; "In Vitro Cytotoxicity of Asbestos and
Man-Made Vitreous Fibers: Roles of Fiber Lenght, Diameter and Composition."
Carcinogenesis. Vol. 15, No. 5, pp. 971-977. May, 1994.newnew
Health Effects Institute - Asbestos Research. Asbestos in Public and Commercial
Buildings: A Literature Review and Synthesis of Current Knowledge. HEI-AR, 141
Portland St., Suite 7100, Cambridge, MA. 1991.
9.11
Revision 1 - 9/3A01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Hei, TK; Wu, LJ; Piao. CQ; "Malignant Transformation of Immortalized Human
Bronchial Epithelial Cells by Asbestos Fibers." Environmental Health Perspectives.
Vol. 105, Supplements, pp. 1085-1088. September, 1997.newnew
Heidenreich WF, Luebeck EG, and Moolgavkar SH; "Some properties of the hazard
function of the two-mutation clonal expansion model." Risk Analysis Vol. 17, pp. 391-
399. 1997.
Henderson, V.L., Enterline, P.E., "Asbestos Exposure: Factors Associated with
Excess Cancer and Respiratory Disease Mortality." Annals New York Academy of
Sciences, Vol. 330, pp. 117-126, 1979.
Hesterberg, TW; Chase, G; Axten, C; Miller, WC; Musselman, RP; Kamstrup, O;
Hadley, J; Morscheidt, C; Bernstein, DM; Thevenaz, P; "Biopersistence of Synthetic
Vitreous Fibers and Amosite Asbestos in the Rat Lung Following Inhalation" Toxicology
and Applied Pharmacology Vol. 151, 262-275. 1998a.
Hesterberg, TW; Hart, GA; Chevalier, J; Miller, WC; Hamilton, RD; Bauer, J; Thevenaz,
P; "The Importance of Fiber Biopersistence and Lung Dose in Determining the Chronic
Inhalation Effects of X607, RCF1, and Chrysolite Asbestos in Rats." Toxicology and
Applied Pharmacology. Vol. 153, No. 1, pp. 68-82. 1998b.newnew
NEED HESTERBERG ET AL. 1996, A REF IN HESTERBERG 1998A, WHICH HAS
THE ANALYSIS FOR CROCIDOLITE.
Hesterberg, TW; Axtin, C; McConnell, EE; Oberdorster, G; Everitt, J; Miller, WC;
Chevalier, J; Chase, GR; Thevenaz, P; "Chronic Inhalation Sutdy of Fiber Glass and
Amosite Asbestos in Hamsters: Twelve-month Preliminary Results." Environmental
Health Perspectives. Vol. 105, Supplement 5, pp. 1223-1230. September,
1997.newnew
Hesterberg, TW; Miiller, WC; McConnell, EE; Chevalier, J; Hadley, JG; Bernstein, DM;
Thevenaz, P; Anderson, R; "Chronic Inhalation Toxicity of Size-Separated Glass Fibers
in Fischer 344 Rats" Fundamental and Applied Toxicology Vol.20; pp.464-476,
1993.newnew
Hesterberg, TW; Miiller, WC; Thevenaz, P; Anderson, R; "Chronic Inhalation Studies
of Man-Made Vitreous Fibres: Characterization of Fibres in the Exposure Aerosol and
Lunos" Ann.Occup.Hvg. Vol.39. No.5, pp.637-653, 1995.newnew
9.12
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Hodgson, A.A.; "Fibrous Silicates." Lecture Series No. 4. The Royal Institute of
Chemistry, London, United Kingdom. 1965.
Hodgson, J and Darnton, A; "The Quantitative Risk of Mesothelioma and Lung Cancer
in Relation to Asbestos Exposure." Ann Occup Hyg Vol. 44, No. 8, Pp. 565-601. 2000.
Holian, A; Uthman, MO; Goltsova, T; Brown, SD; Hamilton, RF; "Asbestos and Silica-
Induced Changes in Human Alveolar Macroophage Phenotype." Environmental Health
Perspectives. Vol. 105, Supplement 5, pp. 1139-1142. September, 1997.newnew
Hughes, J.M., Weill, H., "Asbestos Exposure: Quantitative Assessment of Risk."
American Review of Respiratory Disease, Vol. 133, pp. 5-13,1986.
Hughes, J.M., Weill, H., Hammad, Y.Y., "Mortality of Workers Employed at Two
Asbestos Cement Plants." British Journal of Industrial Medicine 44:161-174,1987.
Hume, LA and Rimstidt; "The Biodurability of Chrysotile Asbestos." Am Mineralogist
Vol. 77, pp. 1125-1128. 1992. Newnew
Hwang, C.Y., Gibbs, G.W., "The Dimensions of Airborne Asbestos Fibres -I. Crocklolite
from Kuruman Area, Cape Province, South Africa." Annals Occupational Hygiene, Vol.
24, No. 1, pp. 23-41, 1981.
Ilgren, E; Chatfield, E; "Coalinga Fibre: A Short, Amphibole-f ree Chrysotile ." Part 3:
Lack of Biopersistence. Indoor Built Environ. Vol. 7, pp. 98-100. 1998.newnew
International Agency for Research on Cancer (IARC). Monographs on the Evaluation
of Carcinogenic Risks to Man. Vol 14. IARC, Lyon, France. 1977.
IRIS 1988 FINISH THIS.
Ishizaki, T; Yano, E; Evans, PH; "Cellular Mechanisms of Reactive Oxygen Metabolite
Generation from Human Polymorphonuclear Leukocytes Induced by Crocidolite
Asbestos." Environmental Research. Vol. 75, No. 2, pp. 135-140. 1997.newnew
IS010312: Ambient Air - Determination of Asbestos Fibres - Direct-Transfer
Transmission Electron Microscopy Method. (Developed by Chatfield, E.J. 1993).
Jagirdar, J; Lee, TC; Reibman, J; Gold, LI; Aston. C; Begin, R; Rom, WN;
"Immunohistochemical Localization of Tranforming Growth Factor Beta Isoforms in
9.13
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Asbestos-Related Diseases." Environmental Health Perpsectives. Vol. 105,
Supplement 5, pp. 1197-1203. September, 1997.newnew
Jaurand, M-C; "Mechanisms of Fiber-Induced Genotoxicity." Environmental Health
Perspectives. Vol. 105, Supplements, pp. 1073-1084. September, 1997.newnew
NEED TO ADD JAURAND'S 1991 REVIEW OF MUTAGENICITY ASSAY TESTING
OF ASBESTOS (GET REF FROM JAURAND 1997 OR KEANE ET AL. 1999)
Jesch, NK; Dorger, M; Enders, G; Rieder, G; Vogelmeier, C; Messmer, K; and
Krombach, F; "Expression of the Inducible Nitroc Oxide Synthase and Formation of
Nitric Oxide by Alveolar Macrophages." Environmental Health Perspectives. Vol.105,
Supplements, pp. 1297-1300. September, 1997.newnew
Johnson, N.F., "Asbestos-Induced Changes in Rat Lung Parenchyma." Journal of
Toxicology and Environmental Health, Vol. 21, pp. 193-203,1987.
Johnson, NF; Jaramillo, RJ; "P53, Cip 1, and Gadd 153 Expression Following
Treatment of A549 Cells with Natural and Man-Made Vitreous Fibers." Environmental
Health Perspectives. Vol. 105, Supplement 5, pp. 1143-1145. September,
1997.newnew
Jones, A.D., McMillan, C.H., Johnston, A.M., Mcintosh, C., Cowie, H., Bolton, R.E.,
Borzuki, G., Vincent, J.H., "Pulmonary Clearance of UICC Amosite Fibres Inhaled by
Rats During Chronic Exposure at Low Concentrations." British Journal of Industrial
Medicine, Vol. 45, pp. 300-304, 1988.
Kaiglova, A; Hurbankova, M; Kovacikova, Z; "Impact of Acute and Subchronic
Asbestos Exposure on Some Parameters of Antioxidant Defense System and Lung
Tissue Injury." Industrial Health. Vol. 37, No. 3, pp. 348-351. July, 1999.newnew
Kamp, DW; Greenberger, MJ; Sbalchierro, JS; Preusen, SE; Weitzman, SA; "Cigarette
Smoke Augments Asbestos-Induced Alveolar Epithelial Cell Injury: Role of Free
Radicals." Free Radical Bioloov & Medicine. Vol. 25, No. 6, pp. 728-739.
1998.newnew
Kamp, DW; Dunne, M; Dykewicz, MS; Sbalchiero, JS; Weitzman, SA; and Dunn, MM;
"Asbestos-Induced Injury to Cultured Human Pulmonary Epithelial-like Cells: Role of
Neutroophil Elastase." J Leukocyte Bioloov. Vol. 54, July, 1993.newnew
9.14
Revision 1 * 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Kamp, DW; Graceffa, P; Pryor, WA; Weitzman, SA; The Role of Free Radicals in
Asbestos-Induced Diseases." Free Radical Bioloov & Medicine. Vol. 12, pp. 293-315.
1992.newnew
Kane, AB; Macdonald, JL; "Mechanisms of Mesothelial Cell Injury, Proliferation, and
Neoplasia Induced by Asbestos Fibers." Fiber Toxicology. Chapter 14, pp. 323-347.
1993.newnew
Kauffer, E., Vigneron, J.C., Hesbot, A., Lemonnier, M., "A Study of the Length and
Diameter of Fibres, in Lung and in Broncho-Alveolar Lavage Fluid, Following Exposure
of Rats to Chrysotile Asbestos." Annals of Occupational Hygiene, Vol. 31, No. 2, pp.
233-240, 1987.
Keane, MJ; Miller, WE; Ong, T; Stephens, JW; Wallace, WE; Zhong, B-Z; "A Study of
the Effect of Chrysotile Fiber Surface Composition on Genotoxicity In Vitro." Journal of
Toxicology and Environmental Health Part A. Vol. 57, No. 8, pp. 529-541. Aug,
1999.newnew
Kimizuka, G., Wang, N., Hayashi, Y., "Physical and Microchemical Alterations of
Chrysotile and Amosite Asbestos in the Hamster Lung." Journal of Toxicology and
Environmental Health, Vol. 21, pp. 251-264,1987
Kodama, Y; Boreiko, CJ; Maness, SC; and Hesterberg, TW; "Cytotoxic and Cytogenetic
Effects of Asbestos on Human Bronchial Epithelial Cells in Culture." Carcinogenesis
Vol. 14, No. 4, pp. 691-697. 1993.
Kostyuk, VA; Potapovich, Al; "Antiradical and Chelating Effects in Flavonoid Protection
Against Silica-Induced Cell Injury." Archives of Biochemistry and Biophysics." Vol. 355,
No. 1, pp. 43-48. 1998.newnew
Kravchenko, IV; Furalyov, VA; Vasylieva, LA; and Pylev, LN; "Spontaneous and
Asbestos-Induced Transformation of Mesothelial Cells in Vitro." Teratooenesis
Carcinogenesis and Mutagenesis. Vol. 18, No. 3, pp. 141-151. 1998.newnew
Krombach, F.; Munzing, S.; Allmeling, A.M.; Gerlach, JT; Behr, J; and Dorger, M„ "Cell
Size of Alveolar Macrophages; An Interspecies Comparison.." Environmental Health
Perspectives; Vol. 105, Supplements, pp. 1261-1263. September, 1997.
9.15
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Lacquet, L.M., VanderLinden, L., Lepoutre, J., "Roentgenograph^ Lung Changes,
Asbestosis and Mortality in a Belgian Asbestos-Cement Factory." Biological Effects of
Mineral Fibres, Wagner, J.C., ed., IARC Sci Publ., pp. 783-793, 1980.
Law, BD; Bunn, WB; and Hesterberg, TW; "Dissolution of Natural Mineral and Man-
Made Vitreous Fibers in Kamovsky's and Formalin Fixatives." Inh. Tox. Vol. 3, pp.
309-321, 1991.newnew
Law, BD; Bunn, WB; and Hesterberg, TW; "Solubility of Polymeric Organic Fibers and
Manmade Vitreous Fibers in Gambles Solution." Inh. Tox. Vol. 2, pp. 321-339,
1990.newnew
Leanderson, P., Soderkvist, P., Tagesson, C., Axelson, 0., "Formation of 8-hydroxydeoxyguanosine
by Asbestos and Man-Made Mineral Fibres." British Journal of Industrial Medicine, Vol. 45, pp. 309-
311,1988.(2)
LeBouffant, L., Daniel, H., Heninn, J.P., Martin, J.C., Normand, C., Tichoux, G., Trolard,
F., "Experimental Study on Long-Term Effects of Inhaled MMF on the Lungs of Rats."
Annals of Occupational Hygiene, Vol. 31, No. 4B, pp. 765-790, 1987.
LeBouffant, L., "Physics and Chemistry of Asbestos Dust." Biological Effects of Mineral
Fibres, Wagner, J.C., ed., IARC Scientific Publications, pp. 15-34, 1980.
Lee, K.P., Barras, C.E., Griffith, R.S., Waritz, R.S., Lapin, C.A., "Comparative
Pulmonary Responses to Inhaled Inorganic Fibers with Asbestos and Fiberglass."
Environmental Research, Vol. 24, pp. 167-191, 1981.
Leigh, J; Wang, H; Bonin, A; Peters, M; and Ruan, X; "Silica-induced Apoptosis in
Alveolar and Granulomatous Cells in Vivo." Environmental Health Perspectives. Vol.
105, Supplement5, pp. 1241-1246. September, 1997.newnew
Leikoff, G; and Driscoll, K; "Cellular Approaches in Respiratory Toxicology" In
Toxicology of the Lung. 2nd edition, Gardner, DE; Crapo, JD; and McClellan, RO eds.
Raven Press, New York. 1993. Newnew
Levin, J.L., McLarty, J.W., Hurst, G.A., Smith, A.N., Frank, A.L, "Tyler Asbestos
Workers: Mortality Experience in a Cohort Exposed to Amosite." Occupptional and
Environmental Medicine 55:155-160, 1998.
9.16
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Levresse, V; Reiner, A; Fleury-Feith, J; Levy, F; Moritz, S; Vivo, C; Pilatte, Y; Jaurand,
M-C; "Analysis of Cell Cycle Disruptions in Cultures of Rat Pleural Mesothelial Cells
Exposed toAsbestos Fibers." Amer J Resp Cell and Molecular Biology. Vol. 17, No. 6,
pp. 660-671. 1997.newnew
Li, XY; Gilmour, P.S.; Donaldson, K; and MacNee, W; "In Vivo and in Vitro
Proinflammatory Effects of Particulate Air Pollution (PM10)." Environmental Health
Perspectives. Vol. 105, Supplement 5, pp. 1279-1284. September, 1997.newnew
Liddell, F.D.K., McDonald, A.D., McDonald, J.C., The 1891-1920 Birth Cohort of
Quebec Chrysotile Miners and Millers: Development From 1904 and Mortality to 1992."
Annals of Occupational Hygiene 41:13-36,1997.
Lim, Y; Kim, S-H; Kim, K-A; Oh, M-W; and Lee, K-H; "Involvement of Protein Kinase C,
Phospholipase C, and Protein Tyrosine Kinase Pathways in Oxygen Radical
Generation by Asbestos-stimulated Alveolar Macrophage " Environmental Health
Perspectives. Vol. 105, Supplements, pp. 1325-1328. September, 1997.newnew
Lippmann M; "Asbestos and Other Mineral and Vitreous Fibers." NEED REST OF
CITATION. 2000
Lippmann, M; "Deposition and Retention of Inhaled Fibres: Effects on Incidence of
Lung Cancer and Mesothelioma." Occupational and Environmental Medicine. Vol. 51,
pp.793-798. 1994.newnew
Lippmann, M., Schlesinger, R.B., "Interspecies Comparisons of Particle Deposition and
Mucociliary Clearance in Tracheobronchial Airways." J. Toxicol, Env. Health, Vol. 3, pp.
441 [261 ]-469[289], 1984 (from old bib)
Luster, Ml; Simeonova, PP; "Asbestos Induces Inflammatory Cytokines in the Lung
Through Redox Sensitive Transcription Factors." Toxicology Letters (Shannon). Vol.
102-103, pp. 271-275. 1998.newnew
Lynch, J.R., Ayer, H.E., Johnson, D.J., "The Interrelationships of Selected Asbestos
Exposure Indices." American Industrial Hygiene Association Journal, Vol. 31, No. 5,
pp. 598-604,1970.
Marconi, A., Menichini, E., Paoletti, L., "A Comparison of Light Microscopy and
Transmission Electron Microscopy Results in the Evaluation of the Occupational
9.17
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Exposure to Airborne Chrysotile Fibres." Annals of Occupational Hygiene, Vol. 28, No.
3, pp. 321-331,1984.
Mar?-:11a, JM; Liu, BL; Vaslet, CA; Kane, AB; "Susceptibility of p53-Deficient Mice to
Induction of Mesothelioma by Crocidolite Asbestos Fibers." Environmental Health
Perspectives. Vol. 105, Supplements, pp. 1069-1072. September, 1997.newnew
Martin, LD; Krunkosky, TM; Dye, JA; Fischer, BM; Jiang, NF; Rochelle, LG; Akley, NJ;
Dreher, KL: and Adler, KB; "The Role of Reactive Oxygen and Nitrogen Speices in the
Response of Airway Epithelium to Particulates." Environmental Health Perspectives.
Vol. 105, Supplement 5, pp. 11301-1308. September, 1997.newnew
Mattson, SM; "Glass Fiber Dissolution in Simulated Lung Fluid and Measures Needed
to Improve Consistency and Correspondence in In-Vitro Studies." Presented at the
IARC Conf. "Biopersistence of Respirable Synthetic Fibres and Minerals." Lyon,
France, September 7-9, 1992. Envron Health Persoect. 1994.
McConnell, E., Wagner, J., Skidmore, J., Moore, J., "A Comparative Study of the
Fibrogenic and Carcinogenic Effects of UICC Canadian Chrysotile B Asbestos and
Glass Microfibre (JM 100)." Biological Effects of Man-made Fibres, Proceedings of a
WHO/IARC Conference, pp. 234-252, 1982.
McConnell, EE; Rutter, HA; Ulland, BM; Moore, JA; "Chronic Effects of Dietary
Exposure to Amosite Asbestos and Tremolite in F344 Rats." Environ Health
Perspectives. Vol. 53, pp. 27-44.1983.newnew
NEED TO ADD SECOND MCCONNELL 1983 (FROM ILGREN AND CHATFIELD
1998)
McConnell, E.E.; Mast, R.W.; Hesterberg, T.W.; Chevalier, J.; Kotin, P.; Bernstein,
D.M.; Thevenez, P.; Glass, L.R.; and Anderson, R.; "Chronic Inhalation Toxicity of a
Kaolin-Based Refractory Ceramic Fiber in Syrian Golden Hamsters." Inhalation
Toxicity, Vol. 7, pp. 503-532, 1995.
McDonald, A.D., Fry, J.S., Wooley, A.J., McDonald, J.C., "Dust Exposure and Mortality
in an American Chrysotile Textile Plant." British Journal of Industrial Medicine 39:361-
367,1983a.
9.18
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
McDonald, A.D., Fry, J.S., Woolley, A.J., McDonald, J.C., "Dust Exposure and Mortality
in an American Factory Using Chrysotile, Amosite, and Crocidolite in Mainly Textile
Manufacture." British Journal of Industrial Medicine 40:368-374, 1983b.
McDonald, A.D., Fry, J.S., Woolley, A.J., and McDonald, J.C., "Dust Exposure and
Mortality in an American Chrysotile Asbestos Friction Products Plant." British Journal of
Industrial Medicine, Vol. 41, pp. 151-157, 1984.
McDonald, JC; "Mineral Fibre Persistence and Carcinogenicity." Ind Health Vol.36,
No.4, pp.372-375, October 1998a.newnew reviewed/to be filed.
McDonald, JC; "Invited Editorial: Unfinished Business - The Asbestos Textiles
Mystery." Annals of Occupational Hygiene. Vol. 42, No. 1, pp. 3-5. 1998b.newnew
McDonald, J.C., Gibbs, G.W., Liddell, F.D.K., "Chrysotile Fibre Concentration and Lung
Cancer Mortality: A Preliminary Report." Biological Effects of Mineral Fibres, Wagner,
J.C., ed., IARC Scientific Publications, pp. 811-817, 1980a.
McDonald, J.C., Liddell, F.D.K., Gibbs, G.W., Eyssen, G.E., McDonald, A.D., "Dust
Exposure and Mortality in Chrysotile Mining, 1910-1975." British Journal of Industrial
Medicine, Vol. 37, pp. 11-24, 1980b.
McDonald, J.C.; Liddell, F.D.K., Dufresne, A.; and McDonald, A.D.; "The 1891-1920
Birth Cohort of Quebec Chrysotile Miners and Millers: Mortality 1976-1988." British
Journal of Industrial Medicine, Vol. 50, pp. 1073-1081, 1993.
McDonald J.C., McDonald, A.D., Armstrong, B., Sebastien, P., "Cohort Study of
Mortality of Vermiculite Miners Exposed to Tremolite." British Journal of Industrial
Medicine, Vol. 43, pp. 436-444,1986.
McGavran, PD; Butterick, CJ; and Brody, AR; "Tritiated Thymidine Incorporation and
the Development of an Interstitial Lesion in the Bronchiolar-Alveolar Regions of the
Lungs of Normal and Complement Deficient Mice After Inhalation of Chrysotile
Asbestos." J Environ Pathol Toxicol Oncol (United States^ Vol. 9, No. 5-6, pp. 377-
391. December, 1989.newnew
McGavran, PD; Moore, LB; and Brody, AR; "Inhalation of Chrysotile Asbestos Induces
Rapid Cellular Proliferation in Small Pulmonary Vessels of Mice and Rats." Am J Path
Vol. 136, No. 3, pp. 695-705,1990.newnew.
9.19
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Mercer RR and Crapo, JD; Three-dimensional Analysis of Lung Structure and its
Application to Pulmonary Dosimetry Models." In Toxicology of the Luna. 2nd edition,
Gardner, DE; Crapo, JD; and McClellan, RO eds. Raven Press, New York. 1993.
Newnew
Middleton, A.P., Beckett, S.T., Davis, J.M.G., "Further Observations on the Short-Term
Retention and Clearance of Asbestos by Rats, Using UICC Reference Samples."
Annals of Occupational Hygiene, Vol. 22, pp. 141-152, 1979.
Miller, FJ; Overton, JH; Kimbell, JS; and Russell, ML; "Regional Respiratory Tract
Absorption of Inhaled Reactive Gases." In Toxicology of the Lung. 2nd edition, Gardner,
DE; Crapo, JD; and McClellan, RO eds. Raven Press, New York. 1993. Newnew
Moolgavkar, SH; Dewanji, A; Venzon, DJ; "A Stochastic Two-Stage Model for Cancer
Risk Assessment. 1: The Hazard Function and the Probability of Tumor." Risk
Analysis. Vol. 8, No. 3, pp. 383-392. 1988.newnew
Moolgavkar, SH; Luebeck, EG; De Gunst, M; Port, RE; Schwarz, M; "Quantitative
Analysis of Enzyme-Altered Foci in Rat Hepatocarginogenisis Experiments-!. Single
Agent Regimen" Carcinoaenisis Vol.11. No.8, pp.1271-1278, 1990.newnew
Moolgavkar SH, Luebeck EG, Krewski D, and Zielinski JM; "Radon, cigarette smoke,
and lung cancer: a re-analysis of the Colorado plateau uranium miners' data."
Epidemiology Vol. 4, pp. 204-217. 1993.
Morgan, A., Talbot, R.J., Holmes, A., "Significance of Fibre Length in the Clearance of
Asbestos Fibres from the Lung." British Journal of Industrial Medicine, Vol. 35, pp. 146-
153, 1978.
Morgan, A., Holmes, A., "Concentrations and Dimensions of Coated and Uncoated
Asbestos Fibres in the Human Lung." British Journal of Industrial Medicine, Vol. 37, pp.
25-32, 1980.
Morgan, A., Black, A., Evans, N., Holmes, A., Pritchard, J.N., "Deposition of Sized
Glass Fibres in the Respiratory Tract of the Rat." Annals of Occupational Hygiene, Vol.
23, pp. 353-366, 1980.
Morrow, PE; "Issues: Possible Mechanisms to Explain Dust Overloading of the Lungs."
Fundamental and Applies Toxicology. Vol. 10, pp. 369-384. 1988.newnew
9.20
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Mossman, B.T., "Mechanisms of Asbestos Carciongenesis and Toxicity: the Amphibole
Hypothesis Revisited." British Journal of Industrial Medicine, Vol. 50, pp. 673-676,
1993.
Mossman, BT and Churg, A; "Mechanisms in the Pathogenesis of Asbestosis and
Silicosis." Am J Crit Care Med. Vol. 157, pp. 1666-1680. 1998.newnew
Mossman, B.T., Marsh, J.P., "Role of Active Oxygen Species in Asbestos-induced Cytotoxicity, Cell
Proliferation, and Carcinogenesis." Cellular and Molecular Aspects of Fiber Carcinogenesis. Cold
Spring Harbor Laboratory Press, 0-87969-361-4/91, pp. 159-168,1991.(3)
Mossman, B.T.; Bignon, J.; Corn, M.; Seaton, A.; and Gee, J.B.L., "Asbestos: Scientific
Developments and Implications for Public Policy." Science, Vol. 247 (Jan 19), pp. 294-
301,1990.
Mossman, BT; Kamp, DW; Sigmund, A; and Weitzman, SA; "Mechanisms of
Carcinogenesis and Clinical Features of Asbestos-Associated Cancers." Cancer
Investigation. Vol. 14, No. 5, pp. 466-480. 1996.newnew
Mossman, BT; Faux, S; Janssen, V; Jimenex, LA; Timblin, C; Zanella, C; Goldberg, J;
Walsh, E; Barchowsky, A; Driscoll, K; "Cell Signaling Pathways Elicited by Asbestos."
Environmental Health Perspectives. Vol. 105, Supplement 5, pp. 1121-1125.
September, 1997.newnew
Muhle, H., Bellman, B., Takenata, S., Ziem, Y., "Inhalation and Injection in Rats to Test
the Carcinogenicity of MMMF." Annals Occupational Hygiene, Vol. 31, No. 4B, pp. 755-
764, 1987.
Nehls, P; Seiler, F; Rehn, B; Greferath, R; and Bruch, J; "Formation and Persistence of
8-Oxoguanine Rat Lung Cells as an Important Determinant for Tumor Formation
following Particle Exposure." Environmental Health Perspectives. Vol. 105,
Supplements, pp. 1231-1240. September, 1997.newnew
Nicholson, W.J., "Part III. Recent Approaches to the Control of Carcinogenic
Exposures. Case Study 1: Asbestos - The TLV Approach." Annals New York
Academy of Science 271:152-169,1976.
Nicholson, W.J., Selikoff, I.J., Seidman, H., Lilis, R., Formby, P., "Long-Term Mortality
Experience of Chrysotile Miners and Millers in Thetford Mines, Quebec." Annals New
York Academy of Sciences, Vol. 330, pp. 11-21,1979.
9.21
Revision 1 - 6/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Nikula, KJ; Avila, KJ; Griffith, WC; and Mauderly, JL; "Sites of Particle Retention and
Lung Tissue Responses to Chronically Inhaled Diesel Exhaust and Coal Dust in Rats
and Cynomolgus Monkeys." Environmental Health Perspectives. Vol. 105,
Supplement 5, pp. 1231-1240. September, 1997.newnew
Oberdorster, G; "Macrophage-Associated Responses to Chrysotyle" Annals of
Occupational Hygiene Vol.38. No.4, pp.601 -615, 1994.newnew
Oberdorster, G; Morrow, PE; and Spurny, K; "Size Dependent Lymphatic Short Term
Clearance of Amosite Fibres in the Lung." Annals of Occupational Hygiene. Vol. 32,
(Suppl 1). pp. 149-156. 1988.newnew
Ollikainen, T; Linnainmaa, K; and Kinnula, VL; "DNA Single Strand Breaks Induced by
Asbestos Fibers in Human Pleural Mesothelial Cells In Vitro." Environmental and
Molecular Mutagenesis. Vol. 33, No. 2, pp. 153-160. 1999.newnew
Ontario Royal Commission, "Report of the Royal Commission on Matters of Health and
Safety Arising form the Use of Asbestos in Ontario." Vol. 3, 1984.
Palekar, LD; Spooner, CM; and Coffin, DL; "Influence of Crystallization Habit of
Minerals on In Vitro Cytotoxicity." Annals New York Academy of Sciences, pp. 673-
687. 1979.
Park, S-H; Aust, AE; "Regulation of Nitric Oxide Synthase Induction by Iron and
Glutathione in Asbestos-Treated Human Lung Epithelial Cells." Archives of
Biochemistry and Biophysics. Vol. 360, No. 1, pp. 47-52. 1998.newnew
Peto, J., "Lung Cancer Mortality in Relation to Measured Dust Levels in an Asbestos
Textile Factory." Biological Effects of Mineral Fibres, Wagner, J.C., ed., IARC Scientific
Publications, pp. 829-836, 1980a.
Peto, J., "The Incidence of Pleural Mesothelioma in Chrysotile Asbestos Textile
Workers." Biological Effects of Mineral Fibres, Wagner, J.C., ed., IARC Scientific
Publications, pp. 703-711,1980b.
Peto, J., Doll, R., Howard, S.V., Kinlen, L.J., Lewinsohn, H.C., "A Mortality Study
Among Workers in an English Asbestos Factory." British Journal of Industrial Medicine,
Vol.34, pp. 169-173, 1977.
9.22
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Peto, J., Seidmari, H., Selikoff, I.J.,. "Mesothelioma Mortality in Asbestos Workers:
Implications for Models of Carcinogenesis and Risk Assessment." British Journal of
Cancer, Vol. 45, pp. 124-135, 1982.
Peto, J., Doll, R., Hermon, C., Binns, W., Clayton, R., Goffe, T., "Relationship of
Mortality to Measures of Environmental Asbestos Pollution in an Asbestos Textile
Factory." Annals Occupational Hygiene, Vol. 29, No. 3, pp. 305-355,1985.
Pinkerton, K.E., Brody, A.R., McLauren, D.A., Adkins, B., Jr., O'Connor, R.W., Pratt, P.C., Crapo, J.D.,
"Characterization of the Three Types of Chrysotile Asbestos After Aerosolization." Environmental
Research, Vol. 31, pp. 32-35,1983.
Pinkerton, K.E., Pratt, P.C., Brody, A.R., Crapo, J.D., "Fiber Localization and Its
Relationship to Lung Reaction in Rats After Chronic Inhalation of Chrysotile Asbestos."
American Journal of Pathology, Vol. 117, pp. 484-498,1984.
Pinkerton, K.E., Plopper, C.G., Mercer, R.R., Rogli, V.L., Patra, A.L., Brody, A.R., Crap,
J.D., "Airway Branching Patterns Influence Asbestos Fiber Location and the Extent of
Tissue Injury in the Pulmonary Parenchyma." Laboratory Investigation, Vol. 55, No. 6,
pp. 688-695, 1986.
Piolatto, G., Negri, E., LaVecchia, C., Pira, E., Decarli, A., Peto, J., "An Update of
Cancer Mortality Among Chrysotile Asbestos Miners in Balangero, Northern Italy."
British Journal of Industrial Medicine 47:810-814,1990.
Platek, S.F., Groth, D.H., Ulrich, C.E., Stettler, L.E., Finnell, M.S., Stoll, M., "Chronic
Inhalation of Short Asbestos Fibers." Fundamental and Applied Toxicology, Vol. 5, pp.
327-340. 1985.
Pooley, F.D., "An Examination of the Fiberous Mineral Content of Asbestos Lung
Tissue from the Canadian Chrysotile Mining Industry." Environmental Research, Vol.
12, pp. 281-298,1976.
Pooley, F.D., Tissue Burden Studies." Short and Thin Mineral Fibres: Identification,
Exposure, and Health Effects, Chatfield, E.J., ed., pp. 96-129,1982.
Pott, F., Huth, F., Friedrichs, K.H., "Tumorigenic Effect of Fibrous Dust in Experimental
Animals." Environmental Health Perspectives, Vol. 9, pp. 313-315,1974.
Pott, F., Huth, F., Friedrichs, K.H., Results of Animal Carcinogenesis Studies After
Application of Fibrous Glass and Their Implications Regarding Human Exposure.
9.23
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Occupational Exposure to Fibrous Glass, U.S. HEW Publication No. 76-151, pp. 183-
191, 1976.
Pott, F., "Some Aspects on the Dosimetry of the Carcinogenic Potency of Asbestos and
Other Fibrous Dusts." Staub-Reinhalt, Vol. 38, No. 12, pp. 486-490, 1978.
Pott, F., "Animal Experiments with Mineral Fibers." Short and Thin Mineral Fibers:
Identification, Exposure, and Health Effects, Chatfield, E.J., ed., pp. 133-161, 1982.
Pott, F., Ziem, U„ Reiffer, R.J., Huth, F., Ernst, H., Mohr, U., "Carcinogencity Studies on
Fibres, Metal Compounds, and Some Other Dusts in Rats." Experimental Pathology,
Vol. 32, pp. 129-152, 1987.
Raabe, O., "Deposition and Clearance of Inhaled Particles." Occupational Lung
Diseases, Gee, J.B., et a!., eds., Raven Press, pp. 1047, 1984.
Quinlan, TR; Berube, KA; Hacker, MP; Taatjes, DJ; Timblin, CR; Goldberg, J;
Kimberley, P; O'Shaughnessy, P; Hemenway, D; Torino, J; Jimenez, LA; Mossman, BT;
"Mechanisms of Asbestos-Induced Nitric Oxide Production by Rat Alveolar
Macrophages in Inhalation and In Vitro Models." Free Radical Biology & Medicine. Vol.
24, No. 5, pp. 778-788. 1998.newnew
Rahman, Q; Mahmood, N; Khan, SG; Arif, JM; Athar, M; "Mechanism of Asbestos-
Mediated DNA Damage: Role of Heme and Heme Proteins." Environmental Health
Perspectives. Vol. 105, Supplement 5, pp. 1109-1112. September, 1997.newnew
Roberts, D.R. and Zumwalde, R.D., Industrial Hygiene Summary Report of Asbestos
Exposure Assessment for Brake Mechanics. NIOSH Reports No. IWS-32-4A, Industrial
Hygiene Section, NTIS # PB 87-105433, 1982.
Robledo, R; and Mossman, B; "Cellular and Molecular Mechanisms of Asbestos-
Induced Fibrosis" J of Cell Physiology Vol. 180:pp. 158-166. 1999. Newnew
Roggli, V.L., Brody, A.R., "Changes in Numbers and Dimensions of Chrysotile Asbestos
Fibers in Lungs of Rats Following Short-Term Exposure." Experimental Lung
Research, Vol. 7, pp. 133-147,1984.
Roggli, V.L., George, M.H., Brody, A.R., "Clearance and Dimensional Changes of
Crocidolite Asbestos Fibers Isolated from Lungs of Rats Following Short-Term
Exposure." Environmental Research, Vol. 42, pp. 94-105,1987.
9.24
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
NEED TO ADD ROOD AND SCOTT 1989
Rubino, G.F., Piolatto, G.W., Newhouse, M.L., Scansetti, G., Aresini, G.A., Murray, R.t
"Mortality of Chrysotile Asbestos Workers at the Balangero Mine, Northern Italy."
British Journal of Industrial Medicine, Vol. 36, pp. 187-194,1979.
St. George, JA; Harkema, JR; Hyde, DM; and Plopper, CG; "Cell Populations and
Structure/Function Relationship of Cells in the Airways." In Toxicology of the Luna. 2nd
edition, Gardner, DE; Crapo, JD; and McClellan, RO eds. Raven Press, New York.
1993. Newnew
Sanden, A; Jarvholm, B; Larsson, S; and Thiringer, G; The Risk of Lung Cancer and
Mesothelioma After Cessation of Asbestos Exposure: A Prospective Cohort Study of
Shipyard Workers." Eur Respir J Vol. 5, pp. 281-285. 1992.newnew
Sebastien, P., Plourde, M.( Robb, R., Ross, M., Ambient Air Asbestos Survey in
Quebec Mining Towns - Part 1, Methodological Study. Environmental Protection
Service, Quebec Region, 3/AP/RQ/1E, pp. 1-41, 1984.
Sebastien, P., Plourde, M., Robb, R., Ross, M., Nadon B., and Wypruk, T. Ambient Air
Asbestos Survey in Quebec Mining Towns. Part II: Main Study. Environmental
Protection Service, Environment, Canada. EPS 5/AP/RQ/2E, July 1986.
Sebastien, P; McDonald, JC; McDonald, AD; Case, B; and Harley, R; "Respiratory
Cancer in Chrysotile Textile and Mining Industries: Exposure Inferences from Lung
Analysis." Br J Ind Med, Vol. 46, pp. 180-187, 1989.
Seidman, H., "Short-Term Asbestos Work Exposure and Long-Term Observation — July
1984 Update." Department of Epidemiology, American Cancer Society, 1984.
Seidman, H., Selikoff, I.J. Hammond, E.C., "Short-Term Asbestos Work Exposure and
Long-Term Observation." Annals New York Academy of Sciences, Vol. 330, pp. 61-89,
1979.
Seidman, H., Selikoff, I.J., Gelb, S.K., "Mortality Experience of Amosite Asbestos
Factory Workers: Dose-Response Relationships 5 to 40 Years After Onset of Short-
Term Work Exposure." American Journal of Industrial Medicine, Vol. 10, Nos. 5/6, pp.
479-514, 1986.
9.25
Revision 1 • 8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Selikoff, I.J. Hammond, E.C., Seidman, H., "Mortality Experience of Insulation Workers
in the United States and Canada 1943-1976." Annals New York Academy of Sciences,
Vol. 330, pp. 91-116, 1979.
Selikoff, I.J., Seidman, H., "Asbestos-Associated Deaths among Insulation Workers in
the United States and Canada, 1967-1987." Annals of the New York Academy of
Sciences 643:1-14, 1991.
Smith, D.M., Oritiz, L.W., Archuleta, R.F., Johnson, N.F., "Long-Term Health Effects in
Hamsters and Rats Exposed Chronically to Man-Made Vitreous Fibres." Annals of
Occupational Hygiene, Vol. 34, No. 4B, pp. 731-754, 1987.
Snyder, J., Virta, R., Segreti, J., "Evaluation of the Phase Contrast Microscopy Method
for the Detection of Fibrous and Other Elongated Mineral Particulates by Comparison
with a STEM Technique." American Industrial Hygiene Association Journal, Vol. 48,
No. 5, pp. 471-477, 1987.
Stanton, M., Wrench, C. "Mechanisms of Mesothelioma Induction with Asbestos and
Fibrous Glass." Journal of the National Cancer Institute, Vol. 48, pp. 797-821, 1972.
Stanton, M., Layard, M., Tegeris, A., Miller, E., May, M., Kent, E., "Carcinogenicity of
Fibrous Glass: Pleural Response in the Rat in Relation to Fiber Dimension." Journal of
the National Cancer Institute, Vol. 58, No. 3, pp. 587-597, 1977.
Stanton, M., Layard, M., Tegeris, A., Miller, E., May, M., Morgan, E., "Relation of
Particle Dimension to Carcinogenicity in Amphibole Asbestos and Other Fibrous
Minerals." Journal of the National Cancer Institute, Vol. 67, No. 5, pp. 965-975, 1981.
Stayner, LT; Dankovic, DA; Lemen, RA; "Occupational Exposure to Chrysotile
Asbestos and Cancer Risk: A Review of the Amphibole Hypothesis." Am J Public
Health. Vol. 86, No. 2, pp. 176-86. Feb. 1996.newnew
Stober, W; McClellen, RO; and Morrow, PE; "Approaches to Modeling Disposition of
Inhaled Particles and Fibers i the Lung" In Toxicology of the Luna. 2nd edition, Gardner,
DE; Crapo, JD; and McClellan, RO eds. Raven Press, New York. 1993. Newnew
Strom, K.A. and Yu, C.P., "Mathematical Modeling of Silicon-Carbide Whisker
Deposition in the Lung - Comparison Between Rats and Humans." Aerosol Science
and Technology, Vol. 24, No. 3. (Oct), pp. 193-209,1994.
9.26
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Sussman, RG; Cohen, BS; Lippmann, M; "Asbestos Fiber Deposition in a Human
Tracheobronchial Cast. I. Experimental." Inhalation Toxicology. Vol. 3, pp. 146-160.
1991a.newnew
Sussman, RG; Cohen, BS; Lippmann, M; "Asbestos Fiber Deposition in a Human
Trachebronchial Cast. II. Empirical Model." Inhalation Toxicology. Vol. 3, pp. 161-179.
1991b.newnew
Takeuchi, T; Nakajima, M; and Morimoto, K; "A Human Cell System for Detecting
Asbestos Cytogenotoxicity In Vitro." Mutation Research. Vol. 438, No. 1, pp. 63-70.
1999.newnew
Tanaka, S; Choe, N; Hemenway, DR; Zhu, S; Matalon, S; Kagan, E; "Asbestos
Inhalation Induces Reactive Nitrogen Species and Nitrotyrosine Formation in the Lungs
and Pleura of the Rat." Journal of Clinical Investigation. Vol. 102, No. 2, pp. 445-454.
1998.newnew
Timblin, CR; Guthrie, GD; Janssen, YWM; Walsh, ES; Vacek, P; and Mossman, BT;
"Patterns of C-Fos and C-Jun Proto-Oncogene Expression, Apoptosis, and Proliferation
in Rat Pleural Mesothelial Cells Exposed to Erionite or Asbestos Fibers." Toxicology
and Applied Pharmacology. Vol. 151, No. 1, pp. 88-97. 1998a.newnew
Timblin, CR; Janssen, YMW; Goldberg, JL; and Mossman, BT; "GRP78, HSP72/73,
and CJUN Stress Protein Levels in Lung Epithelial Cells Exposed to Asbestos,
Cadmium or H202." Free Radical Bioloov & Medicine. Vol. 24, No. 4, pp. 632-642.
1998b.newnew
Timbrell, V., Hyett, A.W., Skidmore, J.W., "A Simple Dispenser for Generating Dust
Clouds From Standard Reference Samples of Asbestos." Annals of Occupational
Hygiene, Vol. 11, pp. 273-281,1968.
Timbrell, V., "Deposition and Retention of Fibres in the Human Lung." Annals of
Occupational Hygiene, Vol 26, No. 1-4, pp. 347-369,1982.
U.S. EPA (U.S. Environmental Protection Agency), Airborne Asbestos Health
Assessment Update. Report 600/8-84-003F, U.S. EPA, 1986.
Venzon, D., Moolgavkar, S. "A method for computing profile-likelihood-based
confidence intervals." Appl Stat 37:87-94,1988.
9.27
Revision 1 • 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Vincent, J.H., "On the Practical Significance of Electrostatic Lung Depositon of
Isometric and Fibrous Aerosols." Journal Aerosol Science, Vol. 16, No. 6. pp. 511-519,
1985.
Vincent, JH; Johnston, AM; Jones, AD; Bolton, RE; and Addison, J; "Kinetics of
Deposition and Clearance of Inhaled Mineral Dusts During Chronic Exposure." Br J Ind
Med Vol. 42; pp. 707-715 (1985).
Wagner, J.C., Berry, G., Skidmore, J.W., Timbrell, V., The Effects of the Inhalation of
Asbestos in Rats." British Journal of Cancer, Vol. 29, pp. 252-269,1974.
Wagner, J.C., Berry, G., Skidmore, J.W., "Studies of the Carcinogenic Effects of Fiber
Glass of Different Diameters Following Intrapleural Innoculation in Experimental
Animals." NIOSH 76-151, pp.193-197, 1976.
Wagner, J.C., "Opening Discussion - Environmental and Occupational Exposure to
Natural Mineral Fibres." Biological Effects of Mineral Fibres, Wagner, J.C., ed., IARC
Scientific Publications, pp. 995-998, 1980.
Wagner, J.C., Berry, G.t Hill, R., Munday, D., Skidmore, J., "Animal Experiments with
MMM(V)F - Effects of Inhalation and Intrapleural Inoculation in Rats." Biological Effects
of Man-made Fibres - Proceedings of a WHO/IARC Conference, Copenhagen, pp.
209-233, 1982.
Wagner, J.C., Skidmore, J.W., Hill, R.J., Griffith, D.M., "Erionite Exposure and
Mesotheliomas in Rats." British Journal of Cancer, Vol, 51, pp. 727-730, 1985.
Wagner, J.C. Griffiths, D.M., Munday, D.E., "Experimental Studies with Palygorskite
Dusts." British Journal of Industrial Medicine." Vol. 44, No. 11, pp. 749-763, 1987.
Walton, W.H., "The Nature, Hazards, and Assessment of Occupational Exposure to
Airborne Asbestos Dust: A Review." Annals of Occupational Hygiene, Vol 25, pp. 117-
247, 1982.
Warheit, DB; Snajdr, SI; Hartsky, MA; Frame, SR; "Lung Proliferative and Clearance
Responses to Inhaled para-Aramid RFP in Exposed Hamsters and Rats: Comparisons
with Chrysotiles Asbestos Fibers." Environmental Health Perspectives Vol. 105,
Supplement 5, pp. 1219-1222. September, 1997.newnew
9.28
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Weill, H., Hughes, J., Waggenspack, C., "Influence of Dose and Fibre Type on
Respiratory Malignancy Risk in Asbestos Cement Manufacturing." American Review of
Respiratory Disease, Vol, 120, pp. 345-354,1979.
Winer, A.A., Cossette, M., "The Effect of Aspect Ration on Fiber Counts: A Preliminary
Study." Annals New York Academy of Sciences, Vol. 330, pp. 661-672, 1979.
Weitzman, SA and Graceffa, P; "Communication: Asbestos Catalyzes Hydroxyl and
Superoxide Radical Generation from Hydrogen Peroxide." Archives of Biochemistry
and Biophysics. Vol. 228, No. 1, pp. 373-376. 1984.newnew
World Health Organization: "Reference Methods for Measuring Airborne Man-Made
Mineral Fibers (MMMF), 1985.newnew
Wright, G.W., Kuschner, M., The Influence of Varying Lengths of Glass and Asbestos
Fibres on Tissue Response in Guinea Pigs." Inhaled Parities: Part 2, Walton, W.H.,
McGovern, B., Pergamon Press Oxford, pp. 455-474,1975.
Wylie, A.G., Virta, R.L., Segretti, J.M., "Characterization of Mineral Population by index
Particle: Implications for the Stanton Hypothesis." Environmental Research, Vol, 43,
pp. 427-439, 1987.
Wylie, A.G.; Bailey, K.F.; Kelse, J.W.; and Lee, R.J.; "The Importance of Width in
Asbestos Fiber Carcinogenicity and its Implications for Public Policy." Am Ind Hyg
Assoc J (54) 239-252, 1993.
Wylie, AG; Skinner, HCW; Marsh, J; Snyder, H; Garzione, C; Hodkinson, D; Winters, R;
Mossman, BT; "Mineralogical Features Associated with Cytotoxic and Proliferative
Effects of Fibrous Talc and Asbestos on Rodent Tracheal Epithelial and Pleural
Mesothelial Cells." Toxicol and Applied Pharmacol. Vol. 147, pp. 143-150.
1997.newnew
Yamaguchi, R; Hirano, T; Ootsuyama, Y; Asami, S; Tsurudome, Y; Fukada, S; Yamato,
H; Tsuda, T; Tanaka, I; and Kasai, H; "Increased 8-Hydroxyguanine in DNA and Its
Repair Activity in Hamster and Rat Lung After Intratracheal Instillation of Crocidolite
Asbestos." Japanese J of Cancer Research Vol. 90, No. 5, pp. 505-509,1999.
Newnew
9.29
Revision 1 '8/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Yeh, H-C and Harkema, JR; "Gross Morphometry of Airways." In Toxicology of the
Luna. 2nd edition, Gardner, DE; Crapo, JD; and McClellan, RO eds. Raven Press, New
York. 1993. Newnew
Yu, CP; Yoon, KJ; Investigator's Report: Retention Modeling of Diesel Exhaust
Particles in Rats and Humans. 1991. CHECK THIS
Yu, C.P.; Asgharian, B; and Abraham, J.L.; "Mathematical Modeling of Alveolar
Clearance of Chrysotile Asbestos Fibers from the Rat Lungs." Journal of Aerosol
Science, Vol. 21, pp. 587-594,1990a.
Yu, C.P. and Asgharian, B; "A Kinetic Model of Alveolar Clearance of Amosite Asbestos
Fibers from the Rat Lung at High Lung Burdens." Journal of Aerosol Science, Vol. 21,
pp. 21-27, 1990b.
Yu, C.P.; Asgharian, B; and Pinkerton, K.E.; "Intrapulmonary Deposition and Retention
Modeling of Chrysotile Asbestos Fibers in Rats." Journal of Aerosol Science, Vol. 22,
No. 6, pp. 757-761, 1991.
Yu, C.P.; Zhang, L.; Oberdoster, G.; Mast, R.W.; Glass, L.R..; and Utell, M.J.,
"Deposition Modeling of Refractory Ceramic Fibers in the Rat Lung" Journal of Aerosol
Science, Vol. 25, No. 2 (Mar), pp. 407-417, 1994.
Yu, C.P.; Zhang, L.; Oberdoster, G.; Mast, R.W.; Maxim, D.; and Utell, M.J., "Deposition
of Refractory Ceramic Fibers (RCF) in the Human Respiratory Tract and Comparison
with Rodent Studies." Aerosol Science and Technology, Vol. 23, No. 3 (Oct), pp. 291-
300, 1995a.
Yu, C.P.; Zhang, L.; Oberdoster, G.; Mast, R.W.; Glass, L.R.; and Utell, M.J.,
"Clearance of Refractory Ceramic Fibers (RCF) from the Rat Lung - Development of a
Model." Environmental Research, Vol. 65, No. 2 (May), pp. 243-253,1995b.
Zalma, R; Bonneau, L; Guignard, J; Pezerat, H; Jaurand, M-C; "Formation of Oxy
Radicals by Oxygen Reduction Arising from the Surface Activity of Asbestos." Can J
Chem. Vol. 65, pp. 2338-2341. 1987.newnew
Zanella, CL: Timblin, CR; Cummins, A; Jung, M; Goldberg, J; Raabe, R; Tritton, TR;
and Mossman, BT; "Asbestos-induced Phosphorylation of Epidermal Growth Factor
Receptor is Linked to c-fos and apoptosis." Am J of Physiology Vol. 277, No. 4, Part!,
Pp. L684-L693. 1999 newnew.
9.30
Revision 1 - 9/3/01
-------
PRELIMINARY WORKING DRAFT - DO NOT COPY OR QUOTE
Zhang, Y; Lee, TC; Guillemin, B; Yu, M-C; and Rom, WN; "Enhanced IL-1Beta and
Tumor Necrosis Factor-Alpha Release and Messenger RNA Expression in
Macrophages from Idiopathic Pulmonary Fibrosis or After Asbestos Exposure." J
Immunol (United States). Vol. 150, No. 9, pp. 4188-4196. 1 May, 1993.newnew
Zhu, S; Manuel, M; Tanaka, S; Choe, N; Kagan, E; Matalon, S; "Contribution of
Reactive Oxygen and Nitrogen Species to Particulate-lnduced Lung Injury."
Fnvironmental Health Perspectives. Vol. 106, No. 5, pp. 1157-1163. 1998.newnew
Zoitus, BK; De Meringl, A; Rouyer, E; Thelohan, S; Bauer, J; Law, B; Boymel, PM;
Olson, JR; Christensen, VR; Guldberg, M; Koenig, AR; and Perander, M; "In Vitro
Measurement of Fiber Dissolution Rate Relevant to Biopersistence at Neutral pH: An
Interlaboratorv Round Robin." Inh Tox. Vol. 9. pp. 525-540. 1997.newr>ew.
9.31
Revision 1 - 9/3/01
------- |