HAZARD IDENTIFICATION IN
CARCINOGEN RISK ANALYSIS:
AN INTEGRATIVE APPROACH
PART III
AN APPLICATION OF THE METHODOLOGY:
FORMALDEHYDE IN AIR
by
Douglas J. Crawford-Brown
Jeffrey R. Arnold
Kenneth G. Brown
April, 1994
This project report is part of the U. S. EPA's program of Research to Improve Health Risk
Assessment. It was funded by the Office of Health and Environmental Assessment, Office of
Research and Development, U. S. Environmental Protection Agency, Washington, D. C. 20460.
-------
DISCLAIMER
Although the information in this document has been funded by the U.S. Environmental
Protection Agency, under contract 68-C9-0009, Work Assignment S-l-56 (Kenneth G. Brown,
Ph.D., Inc.), it does not necessarily reflect the views of the Agency and no official endorsement
should be inferred. Mention of trade names or commercial products does not constitute
endorsement or recommendation for use.
-------
CONTENTS
PART III AN APPLICATION OF THE METHODOLOGY: FORMALDEHYDE IN AIR
1. INTRODUCTION 1
1.1. Objective 1
1.2. Identification of Bodies of Evidence 8
1.3. Specification of Contexts 10
1.4. Relevance Strategies and the Assignment of Intellectual Obligation 13
1.5. Outline of Working Tables and Sections of the Report 18
2. OBSERVATIONAL CONTEXT NO. 5 21
2.1. Description of Working Table 1.5 21
2.2. Description of Working Table 2.5 22
2.2.1. Data Categories and Items 22
2.2.2. Judgments of Completeness 23
2.2.3. Judgments of Utility, Strength of Effect, and Exposure-Specific Effect . 24
2.2.4. Exhibits for Working Table 2.5 27
2.3. Description of Working Table 3.5 28
2.3.1. Introduction 28
2.3.2. Epistemic Status of Claims of Carcinogenicity 29
2.3.3. Intellectual Obligation 34
2.3.4. Claims of Carcinogenicity 34
2.3.4.1. Increases Incidence of Cancer 37
2.3.4.2. Classifications 39
2.3.4.2.1. Complete Carcinogen 39
2.3.4.2.2. Partial Carcinogen 39
i
-------
2.3.4.2.3. Mixer 40
2.3.4.2.4. Helper 40
2.3.4.3. Stages 40
2.3.4.3.1. Neoplastic Conversion 40
2.3.4.3.2. Neoplastic Development 41
2.3.4.4. Mechanisms 43
2.3.4.4.1. Genotoxic 43
2.3.4.4.2. Non-Genotoxic 44
2.3.5. Summary Comments for Intra-Context Claims 44
2.4. Description of Working Table 4.5 45
2.4.1. Introduction 45
2.4.2. Column Headings in Working Table 4.5 45
2.4.3. Epistemic Status of Extrapolation Premises 48
2.4.3.1. Exposure to BSDR 48
2.4.3.2. BSDR to Effect 50
2.4.3.3. Host Factors 51
2.4.3.4. Environmental Conditions 52
2.5. Description of Working Table 5.5 53
2.5.1. Introduction 53
2.5.2. Intellectual Obligation 54
2.5.3. Claims of Carcinogenicity 54
2.5.3.1. Increases Incidence of Cancer 55
2.5.3.2. Classifications 56
2.5.3.2.1. Complete Carcinogen 56
ii
-------
2.5.3.2.2. Partial Carcinogen 56
2.5.3.2.3. Mixer 56
2.5.3.2.4. Helper 57
2.5.3.3. Stages 57
2.5.3.3.1. Neoplastic Conversion 57
2.5.3.3.2. Neoplastic Development 58
2.5.3.4. Mechanisms . 59
2.5.3.4.1. Genotoxic 59
2.5.3.4.2. Non-Genotoxic 61
3. OBSERVATIONAL CONTEXT NO. 6 61
3.1. Description of Working Table 1.6 61
3.2. Description of Working Table 2.6 62
3.2.1. Data Categories and Items 62
3.2.2. Judgments of Completeness 63
3.2.3. Judgments of Utility, Strength of Effect, and Exposure-Specific Effect 64
3.2.4. Exhibits for Working Table 2.6 66
3.3. Description of Working Table 3.6 67
3.3.1. Introduction 67
3.3.2. Epistemic Status of Claims of Carcinogenicity 68
3.3.3. Intellectual Obligation 73
3.3.4. Claims of Carcinogenicity 73
3.3.4.1. Increases Incidence of Cancer 76
3.3.4.2. Classifications 77
3.3.4.2.1. Complete Carcinogen 77
iii
-------
3.3.4.2.2. Partial Carcinogen 79
3.3.4.2.3. Mixer 80
3.3.4.2.4. Helper 81
3.3.4.3. Stages 81
3.3.4.3.1. Neoplastic Conversion 81
3.3.4.3.2. Neoplastic Development 83
3.3.4.4. Mechanisms 83
3.3.4.4.1. Genotoxic 83
3.3.4.4.2. Non-Genotoxic 84
3.3.5. Summary Comments for Intra-Context Claims 85
3.4. Description of Working Table 4.6 85
3.4.1. Introduction 85
3.4.2. Column Headings in Working Table 4.6 85
3.4.3. Epistemic Status of Extrapolation Premises 88
3.4.3.1. Exposure to BSDR 89
3.4.3.2. BSDR to Effect 90
3.4.3.3. Host Factors 91
3.4.3.4. Environmental Conditions 92
3.5. Description of Working Table 5.6 93
3.5.1. Introduction 93
3.5.2. Intellectual Obligation 94
3.5.3. Claims of Carcinogenicity 94
3.5.3.1. Increases Incidence of Cancer 95
3.5.3.2. Classifications 96
iv
-------
3.5.3.2.1. Complete Carcinogen 96
3.5.3.2.2. Partial Carcinogen 97
3.5.3.2.3. Mixer 98
3.5.3.2.4. Helper 99
3.5.3.3. Stages 99
3.5.3.3.1. Neoplastic Conversion 99
3.5.3.3.2. Neoplastic Development 100
3.5.3.4. Mechanisms 100
3.5.3.4.1. Genotoxic 100
3.5.3.4.2. Non-Genotoxic 102
4. OBSERVATIONALCONTEXTNO.il 102
4.1. DescriptionofWorkingTablel.il 102
4.2. Description of Working Table 2.11 103
4.2.1. Data Categories and Items 103
4.2.2. Judgments of Completeness 103
4.2.3. Judgments of Utility, Strength of Effect, and Exposure-Specific Effect . 105
4.2.4. Exhibits for Working Table 2.11 107
4.3. Description of Working Table 3.11 107
4.3.1. Introduction 107
4.3.2. Epistemic Status of Claims of Carcinogenicity 108
4.3.3. Intellectual Obligation 113
4.3.4. Claims of Carcinogenicity 114
4.3.4.1. Increases Incidence of Cancer 116
4.3.4.2. Classifications 117
v
-------
4.3.4.2.1. Complete Carcinogen 117
4.3.4.2.2. Partial Carcinogen 118
4.3.4.2.3. Mixer 119
4.3.4.2.4. Helper 120
4.3.4.3. Stages 120
4.3.4.3.1. Neoplastic Conversion 120
4.3.4.3.2. Neoplastic Development 120
4.3.4.4. Mechanisms 121
4.3.4.4.1. Genotoxic 121
4.3.4.4.2. Non-Genotoxic 121
4.3.5. Summary Comments for Intra-Context Claims 121
4.4. Description of Working Table 4.11 121
4.4.1. Introduction 121
4.4.2. Column Headings in Working Table 4.11 122
4.4.3. Epistemic Status of Extrapolation Premises 125
4.4.3.1. Exposure to BSDR 125
4.4.3.2. BSDR to Effect 126
4.4.3.3. Host Factors 127
4.4.3.4. Environmental Conditions 128
4.5. Description of Working Table 5.11 129
4.5.1. Introduction 129
4.5.2. Intellectual Obligation 130
4.5.3. Claims of Carcinogenicity 130
4.5.3.1. Increases Incidence of Cancer 131
vi
-------
4.5.3.2. Classifications 132
4.5.3.2.1. Complete Carcinogen 132
4.5.3.2.2. Partial Carcinogen 132
4.5.3.2.3. Mixer 132
4.5.3.2.4. Helper 132
4.5.3.3. Stages 133
4.5.3.3.1. Neoplastic Conversion 133
4.5.3.3.2. Neoplastic Development 133
4.5.3.4. Mechanisms 133
4.5.3.4.1. Genotoxic 133
4.5.3.4.2. Non-Genotoxic 133
5. OBSERVATIONAL CONTEXT NO. 12 133
5.1. Description of Working Table 1.12 133
5.2. Description of Working Table 2.12 134
5.2.1. Data Categories and Items 135
5.2.2. Judgments of Completeness 135
5.2.3. Judgments of Utility, Strength of Effect, and Exposure-Specific Effect 136
5.2.4. Exhibits for Working Table 2.12 137
5.3. Description of Working Table 3.12 137
5.3.1. Introduction 137
5.3.2. Epistemic Status of Claims of Carcinogenicity 138
5.3.3. Intellectual Obligation 143
5.3.4. Claims of Carcinogenicity 144
5.3.4.1. Increases Incidence of Cancer 145
vii
-------
5.3.4.2. Classifications 146
5.3.4.2.1. Complete Carcinogen 146
5.3.4.2.2. Partial Carcinogen 146
5.3.4.2.3. Mixer 146
5.3.4.2.4. Helper 146
5.3.4.3. Stages 147
5.3.4.3.1. Neoplastic Conversion 147
5.3.4.3.2. Neoplastic Development 147
5.3.4.4. Mechanisms 147
5.3.4.4.1. Genotoxic 147
5.3.4.4.2. Non-Genotoxic 147
5.3.5. Summary Comments for Intra-Context Claims 148
5.4. Description of Working Table 4.12 148
5.5. Description of Working Table 5.12 148
6. OBSERVATIONAL CONTEXT NO. 13 148
6.1. Description of Working Table 1.13 148
6.2. Description of Working Table 2.13 148
6.2.1. Data Categories and Items 148
6.2.2. Judgments of Completeness 150
6.2.3. Judgments of Utility, Strength of Effect, and Exposure-Specific Effect 150
6.2.4. Exhibits for Working Table 2.13 152
6.3. Description of Working Table 3.13 152
6.3.1. Introduction 152
6.3.2. Epistemic Status of Claims of Carcinogenicity 153
viii
-------
6.3.3. Intellectual Obligation 158
6.3.4. Claims of Carcinogenicity 158
6.3.4.1. Increases Incidence of Transformation 160
6.3.4.2. Classifications 161
6.3.4.2.1. Complete Transforming 161
6.3.4.2.2. Partial Transforming 162
6.3.4.2.3. Mixer 163
6.3.4.2.4. Helper 163
6.3.4.3. Stages 163
6.3.4.3.1. Neoplastic Conversion 163
6.3.4.3.2. Neoplastic Development 163
6.3.4.4. Mechanisms 163
6.3.4.4.1. Genotoxic 163
6.3.4.4.2. Non-Genotoxic 164
6.3.5. Summary Comments for Intra-Context Claims 164
6.4. Description of Working Table 4.13 164
6.4.1. Introduction 164
6.4.2. Column Headings in Working Table 4.13 164
6.4.3. Epistemic Status of Extrapolation Premises 167
6.4.3.1. Exposure to BSDR 168
6.4.3.2. BSDR to Effect 169
6.4.3.3. Host Factors 169
6.4.3.4. Environmental Conditions 171
6.5. Description of Working Table 5.13 172
ix
-------
6.5.1. Introduction 172
6.5.2. Intellectual Obligation 173
6.5.3. Claims of Carcinogenicity 173
6.5.3.1. Increases Incidence of Cancer 173
6.5.3.2. Classifications 175
6.5.3.2.1. Complete Carcinogen 175
6.5.3.2.2. Partial Carcinogen 175
6.5.3.2.3. Mixer 175
6.5.3.2.4. Helper 175
6.5.3.3. Stages 175
6.5.3.3.1. Neoplastic Conversion 175
6.5.3.3.2. Neoplastic Development 176
6.5.3.4. Mechanisms 176
6.5.3.4.1. Genotoxic 176
6.5.3.4.2. Non-Genotoxic 176
7. OBSERVATIONAL CONTEXT NO. 14 176
7.1. Description of Working Table 1.14 176
7.2. Description of Working Table 2.14 177
7.2.1. Data Categories and Items 177
7.2.2. Judgments of Completeness 178
7.2.3. Judgments of Utility, Strength of Effect, and Exposure-Specific Effect 179
7.2.4. Exhibits for Working Table 2.14 180
7.3. Description of Working Table 3.14 181
7.3.1. Introduction 181
x
-------
7.3.2. Epistemic Status of Claims of Carcinogenicity 181
7.3.3. Intellectual Obligation 186
7.3.4. Claims of Carcinogenicity 187
7.3.4.1. Increases Incidence of Transformation 189
7.3.4.2. Classifications 190
7.3.4.2.1. Complete Transforming 190
7.3.4.2.2. Partial Transforming 191
7.3.4.2.3. Mixer 192
7.3.4.2.4. Helper 192
7.3.4.3. Stages 193
7.3.4.3.1. Neoplastic Conversion 193
7.3.4.3.2. Neoplastic Development 194
7.3.4.4. Mechanisms 194
7.3.4.4.1. Genotoxic 194
7.3.4.4.2. Non-Genotoxic 195
7.3.5. Summary Comments for Intra-Context Claims 196
7.4. Description of Working Table 4.14 196
7.4.1. Introduction 196
7.4.2. Column Headings in Working Table 4.14 196
7.4.3. Epistemic Status of Extrapolation Premises 199
7.4.3.1. Exposure to BSDR 199
7.4.3.2. BSDR to Effect 201
7.4.3.3. Host Factors 202
7.4.3.4. Environmental Conditions 203
xi
-------
7.5. Description of Working Table 5.14 204
7.5.1. Introduction 204
7.5.2. Intellectual Obligation 205
7.5.3. Claims of Carcinogenicity 205
7.5.3.1. Increases Incidence of Cancer 206
7.5.3.2. Classifications 207
7.5.3.2.1. Complete Carcinogen 207
7.5.3.2.2. Partial Carcinogen 207
7.5.3.2.3. Mixer 208
7.5.3.2.4. Helper 209
7.5.3.3. Stages 209
7.5.3.3.1. Neoplastic Conversion 209
7.5.3.3.2. Neoplastic Development 210
7.5.3.4. Mechanisms 210
7.5.3.4.1. Genotoxic 210
7.5.3.4.2. Non-Genotoxic 212
8. DESCRIPTION OF WORKING TABLE 6 212
8.1. Epistemic Status of the Claims in Working Table 6 213
8.1.1. Increases Incidence of Cancer 213
8.1.2. Classifications 214
8.1.2.1. Complete Carcinogen 214
8.1.2.2. Partial Carcinogen 214
8.1.2.3. Mixer 215
8.1.2.4. Helper 215
xii
-------
8.1.3. Stages 216
8.1.3.1. Neoplastic Conversion 216
8.1.3.2. Neoplastic Development 216
8.1.4. Mechanisms 217
8.1.4.1. Genotoxic 217
8.1.4.2. Non-Genotoxic 218
9. GENERAL CONCLUSIONS AND DISCUSSION 218
9.1 Alterations in the Methodology 223
9.1.1. Numbering of Working Tables 223
9.1.2. Removal of Variability Premise 224
9.1.3. Intellectual Obligation in Working Table 4 224
9.1.4. Columns in Working Table 2 224
9.1.5. Consideration of Burden of Proof 225
9.1.6. Cancer and Cellular Transformation 226
9.2. Remaining Difficulties 227
9.2.1. Theory and Correlation Judgments 227
9.2.2. Separability of Intellectual Obligation and Epistemic Status 228
9.2.3. Pharmacodynamics Data other than on Formaldehyde 229
9.2.4. Methodological Rigor 230
9.2.5. Specification of Contexts 231
9.3. Incorporation of New Data from a Recent Study 233
9.3.1. The Study Conditions and Claims 234
9.3.2. Utility and Strength of Effect Judgments 235
9.3.3. Relevance Strategies Employed 235
xiii
-------
9.3.4. Possible Changes in Working Table Scores 236
9.4. Other Conclusions About Formaldehyde Carcinogenicity 238
9.4.1. The World Health Organization 238
9.4.1.1. Effects on Experimental Animals 238
9.4.1.2. Mutagenicity Effects 239
9.4.1.3. Carcinogenicity Effects in Mice and Rats 240
9.4.1.4. Mechanisms of Carcinogenesis 240
9.4.1.5. Specific Effects on Humans 241
9.4.1.6. General Human Mortality in Epidemiology 242
9.4.1.7. WHO Conclusions About Formaldehyde Carcinogenicity . . . 242
9.4.2. U.S. Occupational Safety and Health Administration 243
9.4.2.1. Human Irritation and Morbidity Effects 244
9.4.2.2. Human Cancer Epidemiology 245
9.4.2.3. Carcinogenic Effects in Experimental Animals 245
9.4.2.4. Concordant Evidence 245
9.4.2.5. OSHA Conclusions About Formaldehyde Carcinogenicity . . . 246
xiv
-------
PART THREE
AN APPLICATION OF THE METHODOLOGY: FORMALDEHYDE IN AIR
1. INTRODUCTION
1.1. Objectives
The present report is supplemental to an earlier report to the U.S.E.P.A. entitled
"Hazard Identification in Carcinogen Risk Analysis: An Integrative Approach" by Crawford-
Brown and Brown. In the previous report, a methodology was developed for analyzing available
evidence for the task of performing a carcinogen hazard identification; for determining the
relevance of that evidence to various claims of carcinogenicity; for assigning a "measure" of the
degree to which that evidence supports any given claim of the carcinogenic properties of the
substance being analyzed; and for identifying key areas in which uncertainty reduced the support
for a claim of carcinogenicity. The resulting methodology yields a series of Working Tables
designed to guide the analyst in assembling the data and determining the role of such data in
reasoning towards the claim that the substance is or is not a carcinogen. In addition, a variety of
more detailed claims, such as the claim that a substance is a partial carcinogen or acts through
genotoxic mechanisms, are elicited within the Working Tables.
The intent of the methodology is to provide an analytic framework within which
judgments of the epistemic status (essentially, degree of evidential support) of a particular claim
of carcinogenicity within hazard identification can be made in a rational fashion without
imposing undue methodological constraints on the many subjective expert judgments that must
underlie a hazard identification for potential carcinogens. Through such a procedure, it becomes
possible to identify key areas of uncertainty in hazard identification, the reasons for those
uncertainties, and the kinds of evidence that would remove the uncertainties and raise the
1
-------
rational basis for regulatory decisions. This outcome ensures that future resources for scientific
research are directed towards the most contentious areas of hazard identification.
A series of analytic tasks are required within the methodology being tested here. The
analyst first is called upon to define the CONTEXTS (defined in the earlier report and in
Section 1.3 in the present report) into which collected data should be placed. Each body of data
then is assigned to a DATA CATEGORY within the context depending upon the role of those
data within lines of reasoning leading to specific claims. The data categories used here are
Tumor Response (containing data on tumorigenicity); Biophysical Effects (containing data on
biological effects leading to, but not yet identical with, the appearance of tumors);
Pharmacodynamics (containing data on the relationship between exposure and biologically
significant dose-rate); Host Factors (containing data on factors modifying the relationship
between production of biophysical effects and appearance of tumors); Concurrent
Environmental Conditions (containing data on external exposure conditions other than the
presence of formaldehyde); and Related Substance Assessment (containing data on the
carcinogenicity of structurally or functionally similar substances).
The second task is to determine the "foundational" quality of the data in each context, by
which we mean the degree to which those data are accurate representations of the properties
they were intended to depict (aside from the question of how the data will be used in lines of
reasoning). This determination is based upon three separate judgments. The judgment of
COMPLETENESS refers to the degree to which the data placed into a data category for a given
context represents an adequate sample of the available data for this data category and context.
A judgment of HI (high epistemic status), ME (medium epistemic status), LO (low epistemic
status) or NO (no epistemic status or "irrelevant") is assigned. The judgment of UTILITY refers
to the degree to which each body of data was collected under conditions appropriate to the
2
-------
properties being measured. Again, the judgment is of HI/ME/LO/NO. The judgment of
STRENGTH OF EFFECT refers to the degree to which any effects noted in a study possess
some measure of "significance", often taken to mean statistical significance. Finally, the results
from each study are summarized as an exhibit so the reader can examine the primary data and
determine if the previous judgments are reasonable.
The third task is to assign measures of the epistemic status (essentially, degree of
evidential support) for the various CLAIMS OF CARCINOGENICITY within each context.
These claims are termed Intra-Context Claims since they are formed for a given context on the
basis only of data assigned to that context in the first task. The broadest claim is that
formaldehyde "Increases the Incidence of Cancer". This claim then is subdivided into 4 separate
claims depending upon the degree to which formaldehyde can increase the incidence of cancer
without the presence of other factors (i.e. antecedent conditions). The 4 possibilities are that
formaldehyde is a COMPLETE CARCINOGEN (producing all transitions to cancer); a
PARTIAL CARCINOGEN (producing only a subset of the necessary transitions to cancer); a
MIXER (producing transitions only after interaction with another substance); or a HELPER
(not producing transitions, but facilitating transitions produced by another substance).
The third task continues with a further subclassification of claims into the STAGE at
which formaldehyde acts to increase the incidence of cancer. The possibilities here are
NEOPLASTIC CONVERSION and NEOPLASTIC DEVELOPMENT. The final
subclassification refers to the MECHANISM by which formaldehyde acts, with the possibilities
being GENOTOXIC and NON-GENOTOXIC.
Each claim of carcinogenicity may be supported by any of the 5 RELEVANCE
STRATEGIES detailed in the original report. The Direct Empirical strategy requires that the
analyst claim a direct observation of the effect of interest under the conditions of interest
3
-------
(including level of exposure and biological context). The Semi-Empirical Extrapolation strategy
is similar to the Direct Empirical strategy with the sole exception that exposures are at levels
other than the level of interest. An empirically-rooted "pattern" noted in the data then is
"followed" to the exposure level of interest. The Theory-Based Inference strategy is invoked
whenever the effect of interest is not observed, but some other effect believed to lead eventually
to the effect of interest is observed and there is an explanatory and predictive theory by which
the observed effect is related to the effect of interest. The Empirical Correlation strategy
requires that the analyst observe an effect other than the effect of interest, and that this
observed effect be correlated (statistically) with the effect of interest. No claim is made that the
observed effect constitutes a step on the way to the observed effect (which IS required by
Theory-Based Inference). The Existential Insight strategy refers to a purely subjective judgment
by a qualified analyst.
Separate judgments of epistemic status are made for each claim of carcinogenicity based
on each of the 5 relevance strategies. Associated with each such judgment is a judgment of the
INTELLECTUAL OBLIGATION assigned by the analyst to each relevance strategy. The term
intellectual obligation refers to the degree to which the analyst judges a particular relevance
strategy to generally provide an adequate basis for justifying claims. For example, the analyst
might determine that empirical correlations are a sufficient basis for justifying claims, even if
understanding does not exist of the reason for the correlation. Conversely, the analyst might also
determine that theory-based inference is required to support the strategy of empirical
correlations, since the former provides a causal explanation of any correlations noted in the
latter.
The analyst then produces a summary judgment for a given claim across the 5 relevance
strategies. This summary judgment incorporates consideration of (i) the epistemic status
4
-------
associated with the claim within a relevance strategy, (ii) the intellectual obligation assigned to
that relevance strategy, and (iii) the coherence of the claim across the full set of 5 relevance
strategies.
The fourth task is to judge the epistemic status of any BACKGROUND PREMISES
needed to extrapolate the separate Claims of Carcinogenicity from the context being examined
to another context. The contexts here are divided into OBSERVATIONAL CONTEXTS (all
contexts for which data are available) and TARGET CONTEXTS (the contexts in which Claims
of Carcinogenicity must eventually be formed if the hazard identification is to be of use in a
regulatory decision. The target context is taken here to be exposure of humans to "low"
concentrations of formaldehyde (less than 2 ppm, which corresponds to concentrations in home
air). The particular background premises required focus on the claim that (1) the relationship
between exposure and biologically significant dose-rate is similar in the two contexts; (2) the
relationship between biologically significant dose-rate and effect is similar in the two contexts;
(3) host factors are similar in the two contexts; and (4) concurrent environmental conditions are
similar in the two contexts. Judgments of epistemic status (HI/ME/LO/NO) are assigned to
each background premise required in each instance of extrapolation between an observational
context and a target context.
The fifth task is to utilize the results of previous tasks to extrapolate claims of
carcinogenicity from each observational context to each target context. Such claims are termed
INTER-CONTEXT CLAIMS since they are formed on the basis of extrapolations between
contexts. A separate judgment is made for each claim of carcinogenicity in the target context as
obtained through extrapolation from each separate observational context. As before, the
judgments are of the form HI/ME/LO/NO. In the sixth and final task the analyst combines the
results of all extrapolations to develop a composite judgment of the carcinogenicity within the
5
-------
target context, taking into account evidence obtained on that context itself (Intra-Context claims)
and by extrapolation from other contexts (Inter-Context claims). A summary judgment for each
claim of carcinogenicity within the target context is formed based on examination of the
coherence across the different modes of extrapolation.
The previous report used a series of examples from past instances of hazard
identification to clarify key points in the framework. That report did not, however, provide a
detailed example of the application of the entire methodology to a single substance. This was
problematic since it might be possible that hazard identification for a given substance might fit
particular parts of the framework quite well, but that the fit would be less perfect in other parts
of the analysis. As a result, the researchers (in consultation with U.S.E.P.A. staff) selected the
present example of formaldehyde as a single example within which the entire methodology could
be tested.
The criteria for testing of the methodology are taken here to be completeness, utility and
practicality. By completeness, we mean the ability of the methodology to incorporate all
important bodies of evidence and to raise all issues of evidential reasoning found in a search of
the extensive literature on formaldehyde and in interviews with experts on the subject.
Formaldehyde was chosen partially because there is a large body of literature available, both in
the form of original studies providing data and in the form of review papers in which the
evidential reasoning of specific individuals and groups is given. This literature base is provided
as a supplementary bibliography in an appendix to this report (Appendix A). There also exists a
large body of literature analyzing the regulatory process by which judgments on the
carcinogenicity of formaldehyde were justified. This literature base also is provided in the
supplementary bibliography.
6
-------
By utility, we mean the ability of the methodology to clarify where key bodies of data are
utilized in the analysis, the role these data play, the issues they raise, and the manner in which
different conceptions of evidential reasoning affect the final judgment of carcinogenicity. Also
essential is the ability of the methodology to aid the analyst in clarifying personal positions on
these matters and to prepare for, and engage in, reasoned debate with other individuals. It is
important to keep in mind that the present authors conceive of rationality as a form of rational
discourse, termed Dialogical Rationality in the earlier report. Such discourse does not require
that judgments be firmly established through universally applicable rules of deductive logic or
even through formal probabilistic techniques (although these are not excluded by the
methodology developed here). It requires, rather, that individuals understand the data at hand,
the quality of those data, their relevance to inferences, and the strengths and weaknesses of the
reasoning leading to those inferences. It requires also that individuals understand how
competing judgments might be supported or critiqued.
By practicality, we mean the ability of the methodology to lead to useful judgments of
carcinogenicity in a reasonable length of time when a reasonable level of resources is provided.
The methodology would fail if (1) it required information typically not available for a substance
when hazard identification is performed; (2) it required scientific or philosophical expertise
beyond the levels normally available in risk analysis and at organizations making judgments; (3)
it required such a long period of time to carry out the analysis that regulatory decisions would
be delayed beyond a reasonable period of time (the idea of "paralysis by analysis"); and/or (4) it
yielded results that were so ambiguous as to be uninterpretable in a regulatory debate or
decision.
The following report details application of the methodology to the single judgment of the
carcinogenicity of formaldehyde. The authors caution that THIS EXAMPLE DOES NOT
7
-------
CONSTITUTE A FULLY DETAILED HAZARD IDENTIFICATION FOR
FORMALDEHYDE, AND SHOULD NOT BE VIEWED AS A SUFFICIENT BASIS FOR
ANY DECISIONS ON THE CARCINOGENICITY OF FORMALDEHYDE. To avoid undue
complexity in the example, and to ensure completion of the project in a timely manner, we have
chosen to focus only on REPRESENTATIVE bodies of evidence. No attempt is made to review
the full body of evidence that would be utilized in a regulatory decision. This is not to say that
the example is therefore flawed, since the intent is to show how the METHODOLOGY
WORKS rather than to perform a complete hazard identification. Care has been taken to
include a cross-section of the KINDS of data that might be available, and to display in all cases
the ROLES of these data in evidential reasoning. As a result, all Working Tables and judgments
called for in the methodology have been completed in the example. Still, the authors have
deliberately utilized only a subset of the available data to demonstrate how the judgments follow
from this admittedly incomplete body of evidence. In a concluding section of the report, an
analysis is presented of the manner in which a selected sample of other data (i.e. data not used
initially) would impact on the judgments made in the body of the report. This approach was
taken to demonstrate the utility of the methodology in incorporating new data as a field of study
develops historically, an important consideration in regulatory debates where scientific
information often is generated during, and in response to, the regulatory debate.
12. Identification of Bodies of Evidence
The literature on formaldehyde is extensive, having accumulated over the course of at
least 100 years. The authors began by preparing a reference list of approximately 250 papers
(see Appendix A) cited most frequently in published summaries of the carcinogenicity of
formaldehyde. This list clearly is too large (by at least an order of magnitude) to provide the
8
-------
base from which the present example should be developed, since the methodological issues being
tested here would be overwhelmed by reflection on the shear mass of data and interpretations of
these data. A subset of 14 studies covering the range of evidential considerations called for in
the methodology was, therefore, culled from the full body of literature.
In producing this subset, two factors were considered. First, the study must be a primary
study rather than a review of other studies or a repetition of some part of a previously published
study. The intent here was to begin the analysis using only the primary bodies of evidence
produced by an experimental or epidemiological study within a given context. Second, there was
a need to include data from each of the data categories identified in the original report. These
categories are:
(a) Tumor Response, consisting of data on various measures of the change in tumor
incidence, time at appearance, etc;
(b) Biophysical Effect, consisting of data on various effects believed to be part of the
chain of cellular changes leading to cancer, such as DNA alterations, cellular
proliferation, etc;
(c) Pharmacodynamics, consisting of data on the relationship between exposure and
biologically significant dose-rate for either the parent molecule or a metabolite;
(d) Host Factors, consisting of data on the relationship between production of
biophysical effects and changes in the rates of transition between stages of cancer,
including data on repair rates (such as repair of DNA) and the health state of an
organism prior to exposure (referred to as the initial state vector);
(e) Concurrent Environmental Conditions, consisting of data on aspects of the external
environment other than the concentration of the substance of interest (i.e.
formaldehyde); and
(f) Related Substance Assessment, consisting of data on the relationship between
exposure and incidence of cancer for substances related structurally and/or functionally
to the substance of interest.
No attempt was made to include each category of data for each context, or all existing
examples of data in a given data category. Rather, the criterion was that each category of data
must be included in at least one context to show its role in lines of reasoning for that context. In
9
-------
addition, for at least one data category and context, there must be an instance in which two
bodies of data were available so as to raise the issue of "meta-analysis" (not taken here to
necessarily mean the meta-analysis associated with statistics and epidemiology). The studies
selected for inclusion in this study are provided in Table 1.
13. Specification of Contexts
CONTEXT means a physical setting within which data are collected. The context is
defined both by the biological properties of the organism exposed to formaldehyde and by the
environmental conditions of exposure. It is necessary to assume that all data assigned to a given
context have been collected under conditions that are sufficiently similar as to justify the belief
that they reflect different aspects of the SAME PHENOMENON LEADING ULTIMATELY
TO THE EFFECT OF INTEREST (e.g., cancer or cellular transformation) IN A
PARTICULAR ORGANISM OR CLASS OF ORGANISMS. For example, exposure of F-344
rats to airborne formaldehyde at a particular laboratory might constitute one context and
exposure of Sprague-Dawley rats to airborne formaldehyde at the same laboratory might
constitute a second context IF it is judged that biological differences between the two strains are
small enough to justify the belief that the two strains are interchangeable with respect to the one
phenomenon of formaldehyde carcinogenicity. Where this belief is not justified, separate
contexts would be specified. Similar comments apply to any decision to divide experiments on
the same strain but exposed in different laboratories.
The intent is to find a balance between recognizing the differences in context that might
exist between separate studies (suggesting a large number of contexts) and the need to simplify
the analysis to a manageable number of contexts so the mass of data may be organized
sequentially into summary statements across subsets of the data. This step requires a judgment
10
-------
TABLE 1. REFERENCES FOR WORKING TABLES 1 AND 2
1. Swenberg, J., Kerns, W., Mitchell, R., Gralla, E. and Pavkov, K, "Induction of Squamous Cell
Carcinomas of the Rat Nasal Cavity by Inhalation Exposure to Formaldehyde Vapor", Cancer
Research. 40, 3398-3402, 1980.
2. Swenberg, J., Gross, E., Martin, J. and Popp J., "Mechanisms of Formaldehyde Toxicity", in
Formaldehyde Toxicity, ed. by J. Gibson, Hemisphere, NY, pp. 132-147, 1983.
3. Steele, V., Arnold, J., Van Arnold, J. and Mass, M., "Evaluation of Rat Tracheal Epithelial Cell
Culture Assay System to Identify Respiratory Carcinogens", Environmental and Molecular
Mutagenesis. 14, 48-54, 1989.
4. Spangler, F. and Ward, J., "Skin Initiation/Promotion Study with Formaldehyde in SENCAR
Mice", in Formaldehyde: Toxicity. Epidemiology. Mechanisms, ed. by J. Chang, J. Gibson and R.
Waritz, pp. 147-158, 1982.
5. Blair, A., Stewart, P. and Hoover, R., "Mortality from Lung Cancer Among Workers Employed
in Formaldehyde Industries", American Journal of Industrial Medicine. 17, 683-699, 1990.
6. Luce, D., Gerin, M., Leclerc, A., Morcet, J., Brugere, J. and Goldberg, M., "Sinonasal Cancer and
Occupational Exposure to Formaldehyde and Other Substances", International Journal of
Cancer. 53, 224-231, 1993.
7. Dykewicz, M., Patterson, R., Cugell, D., Harris, K. and Wu, A., "Serum IgE and IgG to
Formaldehyde-Human Serum Albumin: Lack of Relation to Gaseous Formaldehyde Exposure and
Symptoms", Journal of Allergy and Clinical Immunology. 87, 48-57, 1991.
8. Grafstrom, R., "In Vitro Studies of Aldehyde Effects Related to Human Respiratory
Carcinogenesis", Mutation Research. 238, 175-184, 1990.
9. Heck, H., White, E. and Casanova-Schmitz, M., "Determination of Formaldehyde in Biological
Tissues by Gas Chromatography/Mass Spectrometry", Biomedical Mass Spectrometry. 9, 347-353,
1982.
10. Brusick, D., "Genetic and Transforming Activity of Formaldehyde", in Formaldehyde Toxicity,
ed. by J. Gibson, Hemisphere, NY, pp. 72-84, 1983.
11. Feron, V., Til, H., Vrijer, F., Woutersen, R., Cassee, F. and Van Bladeren, P., "Aldehydes:
Occurrence, Carcinogenic Potential, Mechanism of Action and Risk Assessment", Mutation
Research. 259, 363-385, 1991.
12. Woutersen, R., Appelman, L., Van Garderen-Hoetmer, A. and Feron, V., "Inhalation Toxicity
of Acetaldehyde in Rats. III. Carcinogenicity Study", Toxicology. 41, 213-231, 1986.
13. Casanova-Schmitz, M., Starr, T. and Heck, H.D., "Differentiation between Metabolic
Incorporation and Covalent Binding in the Labeling of Macromolecules in the Rat Nasal Mucosa
and Bone Marrow by Inhaled [14C]- and [3H]-Formaldehyde", Toxicology and Applied Pharmacology.
76, 26-44, 1984.
-------
14. Gibson, J., "Risk Assessment Using a Combination of Testing and Research Results", in
Formaldehyde Toxicity. Hemisphere Publishing, Washington, DC, p. 295-302, 1983.
-------
by the analyst as to a reasonable number of contexts into which the full body of data may be
placed without losing essential information on what may prove to be key differences in the
settings under which studies are conducted.
In the present report, the analysts chose initially to consider 14 contexts based on a
review of the full body of data found in the reference list provided in Appendix A. Organisms
were differentiated only by species, rather than by strain, in recognition of the large inter-species
differences that can exist with respect to carcinogenicity. In addition, separate contexts were
specified for "high" and "low" levels of exposure where available based on a recognition of the
significance of such a distinction in the regulatory arena. Differences in laboratory setting were
not considered in developing the list of contexts, as it was decided to deal with any differences
through the sets of "background premises" used within a context. The contexts located initially
are:
(1) Exposure of unicellular organisms to formaldehyde in solution.
(2) Exposure of invertebrates to formaldehyde in solution.
(3) Exposure of lower order vertebrates to formaldehyde in air.
(4) Exposure of plants to formaldehyde in air.
(5) Exposure of rats to formaldehyde in air.
(6) Exposure of mice to formaldehyde in air.
(7) Exposure of guinea pigs to formaldehyde in air.
(8) Exposure of monkeys to formaldehyde in air.
(9) Exposure of rabbits to formaldehyde in air.
(10) Exposure of dogs to formaldehyde in air.
(11) Exposure of humans to formaldehyde in air at concentrations of 2 ppm and above.
(12) Exposure of humans to formaldehyde in air at concentrations below 2 ppm.
11
-------
(13) Exposure of mammalian cell lines to formaldehyde in solution, corresponding to
high exposures of formaldehyde in air.
(14) Exposure of human cell lines to formaldehyde in solution, corresponding to high
exposures of formaldehyde in air.
A subset of six of the most relevant contexts was selected from the list above, specifically
Contexts 5, 6, 11, 12, 13 and 14. These contexts are described in more detail in the version of
Working Table 1 provided in the various sections of this report (Working Tables 1.5, 1.6, 1.11,
1.12, 1.13 and 1.14); these five Working Tables provide a description of each context and of the
separate studies falling within each context.
Context 12 (exposure of humans to airborne formaldehyde at less than 2 ppm) was
chosen as the TARGET CONTEXT, with the remaining five contexts being designated the
OBSERVATIONAL CONTEXTS. This choice was made based on the desire to include
consideration of extrapolation across exposure levels (or dose), since the five observational
contexts have in common relatively high exposures. As will be noted later, this choice of target
context lowers the epistemic status of the final claim that formaldehyde is a carcinogen in the
target context. If the target context was taken to be Context 11 (where exposures are in line with
those in other contexts), the claim of carcinogenicity would have been stronger. THE
AUTHORS CAUTION THAT THIS CHOICE IN DEFINING THE TARGET CONTEXT IS
A MATTER OF CONVENIENCE FOR TESTING THE METHODOLOGY EXPLORED
HERE, AND SHOULD NOT BE TAKEN TO BE THE ONLY REASONABLE CHOICE OF
TARGET CONTEXT. IN FACT, REGULATORY AGENCIES TYPICALLY VIEW
HAZARD IDENTIFICATION AS REQUIRING ONLY EVIDENCE THAT A SUBSTANCE
IS A CARCINOGEN IN ANY OF THE HUMAN CONTEXTS, AND WOULD CONSIDER
THE TARGET CONTEXT TO INCLUDE BOTH CONTEXT 11 AND CONTEXT 12.
12
-------
The 14 references listed previously in Table 1 fall within these six contexts. Each body of
data from these references then were classified into data categories corresponding to the KIND
of data obtained, as defined in Section 1.1. Data from the 14 primary references were then
divided between the 5 contexts and 6 data categories. Context 5 contains data on Tumor
Response (Reference #1); Biophysical Effects (Reference #2); Pharmacodynamics (References
#9 and #13); and Related Substances (References #11 and #12). Context 6 contains data on
Tumor Response (References #4 and #14) and Biophysical Effects (Reference #10). Context
11 contains data on Tumor Response (References #5 and #6) and Concurrent Environmental
Conditions (Reference #6). Context 12 contains data on Biophysical Effects (Reference #7).
Context 13 contains data on Biophysical Effects (Reference #3) . Context 14 contains data on
Biophysical Effects (Reference #8) and Host Factors (Reference #8).
1.4. Relevance Strategies and the Assignment of Intellectual Obligation
A key task in the present methodology is to assign measures of Intellectual Obligation to
each of the relevance strategies used in bringing the data to bear on judgments. The idea of a
relevance strategy is given full definition in the earlier report and in Section 1.1, but it may be
stated simply as the manner by which an analyst determines that a given piece of information
JUSTIFIES or WARRANTS a particular inference. It is at this stage of the analysis that the
analyst states clearly the degree to which particular broad KINDS of evidence will be counted as
support for an inference. Since the assignment of Intellectual Obligation is common to all
contexts, the discussion of these assignments is provided in this introduction rather than in
conjunction with each specific context. This implies that such broad philosophical positions
should be coherent across contexts.
13
-------
The present analysts have been trained as empirical scientists, and therefore will adopt a
position of strong empiricism in the justification of claims (although other analysts could follow
a different epistemic position). This implies that the stronger the empirical content of a body of
evidence, the higher the epistemic status of claims based on reasoning from that evidence. This
implies further that there is a strong intellectual obligation to employ evidence with as much
empirical content as possible with respect to a specific claim.
The analysts recognize that "empirical content" does not necessarily require direct
observation of a phenomenon (such as tumors), but DOES require at least observational
evidence that can be linked to the phenomenon of interest by either (i) empirically-justified
scientific theories (an instance of Theory-Based Inference), (ii) empirically-established
correlations between the body of evidence and the phenomenon of interest (an instance of
Empirical Correlation), or (iii) empirically-rooted "patterns" in a body of data suggestive of an
effect outside the range of examined data (an instance of Semi-Empirical Extrapolation). By
contrast, Existential Insight does not contain empirical content, even IF the insight was formed
on the basis of experiences. The reason for this is that the analyst cannot re-evaluate the
empirical content of whatever observations have gone into the subjective insight separate from
the subject providing that insight. In other words, the analyst is confronted only with the
statement of the insight itself, rather than with the empirical foundation of that insight (if such a
foundation even exists). Other analysts might claim that elicitation of a purely subjective
judgment from an appropriate expert IS an empirical finding, since it is an empirical
determination of that expert's judgment. We have decided not to take that position here, since
we consider the term "empirical" to apply to the evidence on which an expert might form a
judgment, and not to the fact that the expert HAS made a judgment. A more sociological
14
-------
conception of evidence would lead to a different assignment of Intellectual Obligation for the
Existential Insight relevance strategy.
This leads to the following assignments of Intellectual Obligation, which will be used in
all Working Tables:
DIRECT EMPIRICAL: The judgment here is HIGH since the observation statement in such a
case is directly tied to the property under discussion (although there are background
assumptions about the experimental design). Direct Empirical warrants must, therefore, be
weighted strongly in any Column Summary when available (after considering the foundational
quality of this evidence itself). The lack of Direct Empirical evidence also counts strongly against
a claim.
SEMI-EMPIRICAL EXTRAPOLATION: The judgment here is MEDIUM, just below Theory-
Based Inference, since while observations (empirical evidence) are reported as data for the
outcome of interest, the exposure of interest does not fall within the range of the exposures in
the available data; hence the need for the extrapolation. This is the case, for example, when
dose-response data exist for high concentration exposures but not for the human occupational or
residential exposures usually of interest to the risk analyst. In many cases, semi-empirical
extrapolation does not apply (e.g. when the measurements ail are obtained within the desired
context), and there then is no intellectual obligation to possess such evidence. Semi-Empirical
Extrapolation warrants must, therefore, be weighted moderately in any Column Summary when
available (after considering the foundational quality of this evidence itself). The lack of Semi-
Empirical Extrapolation evidence counts moderately (somewhat less than the lack of Theory-
Based Inference) against a claim.
EMPIRICAL CORRELATION: The judgment here is MEDIUM, still above existential insight
but below semi-empirical correlation. This ranking is justified here because empirical
15
-------
correlation clearly contains empirical content, but the correlation often is asserted with no
attempt to provide an explanation for the correlation. Empirical Correlation warrants must,
therefore, be weighted moderately in any Column Summary when available (after considering
the foundational quality of this evidence itself). The lack of Empirical Correlation evidence
counts weakly against a claim.
THEORY-BASED INFERENCE: The judgment here is MEDIUM, just below DIRECT
EMPIRICAL, since this strategy warrants evidence in two ways: (i) with the strengths of some
actual data from observations (e.g. cell transformation), though not directly about the ultimate
property of interest (e.g. incidence of human cancers); and (ii) with the cohesive force of well-
formed and empirically-tested theories about the importance of these data to the ultimate
property of interest (human cancers). A claim, therefore, will possess no more than medium
epistemic status if a well-established theory backed by empirical observations of effects leading
to (but not equated with) the effect of interest is not available. Theory-Based Inference warrants
must, therefore, be weighted moderately in any Column Summary when available (after
considering the foundational quality of this evidence itself). The lack of Theory-Based Inference
evidence counts moderately against a claim.
EXISTENTIAL INSIGHT: The judgment here is LOW since (as described just above) this
strategy warrants evidence with nothing other than a subjective statement about the degree to
which the data support a given claim, where the concept "degree of support" is not given further
analysis other than to equate it with something like "strength of feeling". This effectively
prevents any re-analysis of the strength of the evidence's empirical content. The analysts do not
wish to assert here that existential insight is useless as a relevance strategy; in fact it is most
likely invoked more often than many scientists and risk analysts realize. Rather these analysts
are claiming that the intellectual obligation to consider evidence warranted by existential insight
16
-------
must be LOW since no supporting information of any sort can be offered. Existential Insight
warrants must, therefore, be weighted weakly in any Column Summary when available (after
considering the foundational quality of this evidence itself). The lack of Existential Insight
evidence counts weakly against a claim.
The present analysts emphasize that the above assignments of Intellectual Obligation are
fully separate from both the contexts and studies that make up the specific bodies of evidence
used in the cells of the various Working Tables of this report, and that more than one of these
relevance strategies can be employed to justify use of any one data set in a context. This
assignment of Intellectual Obligation is meant to be an abstract assessment, meaning that the
analysts have made these assignments without considering the characteristics of actual theories,
correlations, etc, needed in filling out other cells of the Working Tables. For example, it is
assumed that the Intellectual Obligation for Theory-Based Inference is MEDIUM for all
theories, all contexts, all claims and all bodies of evidence used in those theories. Consideration
of the evidence for a SPECIFIC theory is assumed to arise only when the separate CELLS of
the Working Tables are examined.
A counter argument might be that such abstracted assignments of Intellectual Obligation
are not meaningful. For example, it might be argued that one cannot assign a general measure
of Intellectual Obligation to Theory-Based Inference as a class, but rather that this assignment
MUST consider the support for the particular theory being used in a specific inference. In this
case, the analyst would not attempt to provide measures of Intellectual Obligation in the various
Working Tables, but would reflect the assignment of Intellectual Obligation for that
PARTICULAR theory directly in the assignment of epistemic status in a cell of the Working
Table. It may then be the case that a Theory-Based Inference warrant based on some theories
(of a particular stage of carcinogenesis, say) is better established and so carries a higher
17
-------
Intellectual Obligation than do Direct Empirical observations of phenomena pertaining to it (for
example, where the histopathologic form of pre-neoplastic lesions is not clear). The point here is
simply that the judgment of Intellectual Obligation may not be possible in the abstracted sense,
and may be inseparable from the assignment of epistemic status in the other cells of the
Working Table. The analysts have chosen here to view Intellectual Obligation as applying to an
abstracted class of relevance strategy. This decision results in the same assignments of
Intellectual Obligation being made in all Working Tables.
1.5. Outline of Working Tables and Sections of the Report
The report is divided into sections corresponding to the various CONTEXTS (defined
previously) examined in this study. Each section then is subdivided into the analytic tasks called
for in Section 1.1 and resulting in a specific WORKING TABLE summarizing the required
judgments. The various Working Tables are numbered as X.Y, where X refers to the Working
Table and Y refers to the context within which that table has been created. For example,
Working Table 3.6 is Working Table 3 for Context 6.
Working Table l.Y defines Context Y, describes the data placed into that context, and
identifies the data category corresponding to each body of data. Working Table 2.Y summarizes
the "foundational" quality of the data in Context Y, by which we mean the degree to which those
data are accurate representations of the properties they were intended to depict (aside from the
question of how the data will be used in lines of reasoning). Separate judgments are provided
of the Completeness, Utility, and Strength of Effect for each body of data, and the data
themselves are appended as exhibits. Working Table 3.Y contains the judgments of Intellectual
Obligation and Claims of Carcinogenicity for Context Y, and a description of the judgments
associated with the working table appears as subsection 3 in each section of this report. Working
18
-------
Table 4.Y contains the judgments of Intellectual Obligation and Inter-Context Extrapolation
Premises necessary to extrapolate claims of carcinogenicity from Context Y to the Target
Context. A description of the judgments associated with the working table appears as subsection
4 in each section of this report. Working Table 5.Y contains the judgments of Intellectual
Obligation and Claims of Carcinogenicity for extrapolation of the claims from Context Y to the
Target Context. A description of the judgments associated with the working table appears as
subsection 5 in each section of this report. The report closes in Chapter 9 with Working Table 6,
which contains summary judgments for the Claims of Carcinogenicity in the Target Context.
These judgments incorporate the judgments in all previous Working Tables.
The reader should note that Working Table 4 in the present report was listed as
Working Table 5 in the earlier report. The reason for this change is that the original report
divided the collected data into two categories, and placed these into two separate Working
Tables. Working Table 2 contained data necessary for Intra-Context claims, with Working Table
4 containing data necessary for extrapolation FROM the context to the target context. In
performing this example, it became evident that this division was not necessary, and all data on
a context (whether used in generating Intra-context or Inter-context claims) are now placed
directly into Working Table 2. As a result, Working Table 4 from the original report is now
removed, and all Working Table numbers shifted accordingly. The following comparisons may be
made:
(a) Working Table 1 in the original report is Working Table 1 here.
(b) Working Table 2 in the original report is Working Table 2 here.
(c) Working Table 3 in the original report is Working Table 3 here.
(d) Working Table 4 in the original report is now removed, and all associated data
placed into Working Table 2.
(e) Working Table 5 in the original report is Working Table 4 here.
19
-------
(f) Working Table 6 in the original report is Working Table 5 here.
(g) Working Table 7 in the original report is Working Table 6 here.
The report closes with consideration of the lessons drawn from the analysis of
formaldehyde, using the criteria of completeness, utility and practicality mentioned previously.
Key areas in which the methodology functioned less than perfectly are identified, and
recommendations are made for improving the methodology so as to remove these imperfections.
Examples are provided of the manner in which bodies of evidence NOT included in the initial
analysis might have changed judgments, so the reader can develop a sense of the utility of the
methodology in structuring debate on hazard identification within an historically changing
environment of evidence. Throughout the report, examples are given of alternative judgments
and lines of reasoning that might be made; again, this is intended to provide the reader with a
sense of the differences of judgment that still may be found within the constraints of the
methodology. It is important to bear in mind that THIS REPORT REPRESENTS THE
JUDGMENTS OF ONE SET OF ANALYSTS ONLY, IS BASED ON CONSIDERATION OF
ONLY A LIMITED BODY OF EVIDENCE, AND SHOULD NOT BE TAKEN AS
REPRESENTATIVE OF THE JUDGMENTS OF THE SCIENTIFIC COMMUNITY OR OF
THE U.S.E.P.A. A central tenet of the present research is that methodologies for analyzing
uncertainty (and epistemic status) must be sufficiently formal so as to clearly detail the reasons
for judgments being made, but not be so formal as to remove the essential elements of human
judgment that underlie lines of reasoning. This report provides the guidance by which an
intermediate position may be reached, allowing the reader to see why specific judgments have
been made by the current analysts and how different judgments might also be reasonable within
the confines of the methodology.
20
-------
2. OBSERVATIONAL CONTEXT NO. 5
2.1. Description of Working Table 1.5
Working Table 1 for Context 5 is provided below. As in all subsequent sections of this
report, a Working Table is identified by both the general Working Table number (here, 1) and
the context (here, 5). As a result, this Working Table is assigned the number 1.5. The biological
components of this context are given by the fact that all exposed organisms were rats of
relatively similar genetic, anatomical and physiological make-up. Both Fischer 344 and Wistar
rats were judged to fall within this context due to similarities in breathing patterns and lung
histology. Differences in the background incidence of lung cancer, including nasal cancer, are
present, but were not judged to be so significant as to warrant development of two separate
contexts.
The exposures in this context were all to formaldehyde or to a Related Substance
(acetaldehyde) in air at concentrations above 2 ppm, placing them into the category of high
exposures as defined in the Introduction. Incorporation of radionuclides into the formaldehyde
(for purposes of quantifying uptake) was not taken to make the exposures so different as to
warrant development of a separate context, since such incorporation is a minor perturbation to
the molecular structure. Length of exposure varied between the studies, but again this was not
taken to justify development of separate contexts. It was assumed that mechanisms of action of
formaldehyde might be slightly different depending upon length of exposure, but that this
difference would appear in the magnitude of effect rather than the presence and direction of
that effect. This choice is somewhat tenuous since there is evidence that effects such as
promotion may require some minimal period of exposure.
Five bodies of data were identified for this context. These are described in Working
Table 1.5., and are divided into the data categories of:
21
-------
WORKING TABLE 1.5. STUDIES AVAILABLE FOR OBSERVATIONAL CONTEXT NUMBER 5: RATS
EXPOSED TO AIRBORNE FORMALDEHYDE AT CONCENTRATIONS IN EXCESS OF 2 PPM
Study
Number1
Data Category
Description
1
TR
Primary Reference #1 in Table 1
7 week-old Fischer 344 rats were exposed to foimaldehyde in air.
Concentrations were 0, 2, 6 or 15 ppm of foim aldehyde vapor 6
hours/day, 5 days/week, for 18 months of a 24 month study. 10
rats per sex per exposure level were selected randomly for interim
necropsies at 6 and 12 months. 4 cross-sections of nasal turbinates
were examined in each rat. Incidence of nasal carcinomas was
measured in each exposure group.
'The study number refers to ihe order of ihe study within this context. The primary reference number, indexed to Table 1, is provided in the
"Description" column
-------
WORKING TABLE 1.5. Continued
Study
Number1
Data Category
Description
2
BE
Primary Reference #2 in Table 1
7 week-old Fischer 344 rats were exposed to formaldehyde in air.
Numbers of rats in each exposure group are not provided. Exposure
categories consisted of 15 ppm of formaldehyde vapor, 6 hours/day for 5
days; 6 ppm, 6 hours/day for 3 days; 15 ppm, 6 hours/day for 3 days; 0.5
ppm, 6 hours/day for 3 days; and 2 ppm, 6 hours/day for 3 days. Controls
were unexposed to formaldehyde vapor. One day prior to exposure a 7-
day supply of tritiatcd thymidine in an osmotic pump was attached to
each rat. Upon sacrifice, nasal sections were prepared and analyzed for
tritiated thymidine uptake.
'The study number refers to the order of the study within this context. The primary reference number, indexed to Table 1, is provided in the "Description"
column
-------
WORKING TABLE 1.5. Continued
Study
Number1
Data Category
Description
3
PD
Primary Reference #9 in Table 1
Male Fischer 344 rats were exposed to foim aldehyde vapor in air.
Number of exposed rats is not provided. Exposures were to 6 ppm, 6
hours/day for 10 days. Nasal mucosa tissues were analyzed by GC/MS for
each rat. Measurements of fonnaldehyde concentration in units of
umoles/gram are reported for each tissue in exposed and unexposed rats.
'The study number refers to the order of the study within this context. The primary reference number, indexed to Table 1, is provided in the "Description"
column
-------
WORKING TABLE 1.5. Continued
Study
Number1
Data Category
Description
4
RS
Primary Reference #12 in Table 1
Male and female Wistar rats were exposed to acetaldehyde in air.
Exposures were at mean concentrations of 0, 735 and 1412 ppm. A 4th
group was exposed to concentrations decreasing from 3000 ppm to 1100
ppm during the course of the study. All exposures were delivered for 6
hours/day, 5 days/week for up to 28 months. 49 male and 50 female rats
were exposed at 0 ppm; 52 male and 48 female rats were exposed at 735
ppm; 53 male and 53 female rats were exposed at 1412 ppm; and 49 male
and 53 female rats were exposed at 3000/1100 ppm. Each rat was
sacrificed and nasal passages analyzed for papilloma, squamous cell
carcinoma, carcinoma in-situ and adenocarcinoma.
'The study number refers to the order of the study within this context. The primary reference number, indexed to Table 1, is provided in the "Description"
column
-------
WORKING TABLE 1.5. Continued
Study
Number1
Data Category
Description
5
PD
Primary Reference #13 in Table 1
Male Fischer 344 rats were exposed to airborne foimaldehyde labelled
with both radioactive tritium and radioactive carbon (to distinguish
between metabolic incorporation and covalent bonding). Exposures were
for 6 hours at concentrations of 0.3, 2, 6, 10 or 15 ppm. Each exposure
took place 1 day after a prior 6 hour exposure to unlabeled formaldehyde
at the same concentration so as to stimulate cell turnover. Rats were
sacrificed and nasal mucosa removed and tissue from 4 rats in each
exposure category combined, yielding a total sample of 100 mg at each
exposure level. Macro molecules were isolated and analyzed for
radioactivity by liquid scintillation.
'The study number refers to the order of the study within this context. The primary reference number, indexed to Table I, is provided in the "Description"
column
-------
(a) Tumor Response (taken from Reference #1 in Table 1 of this report). Here, the sole
Tumor Response is taken to be incidence of nasal tumors, which clearly is a form of
direct measurement of tumors. It is given the label TR1.
(b) Biophysical Effect (taken from Reference #2 in Table 1). Here, the sole Biophysical
Effect is taken to be cellular proliferation, which is assigned to this data category because
it is an effect leading to, but not identical with, cancer. It is given the label BE1.
(c) Pharmacodynamics (taken from References #9 and #13). Here, there are two bodies
of data, each of which provide information on the relationship between exposure and
biologically significant dose-rate. The first (Reference #9) contains information on tissue
concentration of formaldehyde, which is a measure of the burden of formaldehyde
resulting from exposure. It is given the label PD1. The second (Reference #13) contains
information on DNA adduct formation, which is a measure of biologically significant
dose-rate resulting from exposure. It is given the label PD2. Note that adduct formation
was assigned to the data category Pharmacodynamics, rather than to Biophysical Effect,
since it is taken here to be a measure of dose-rate rather than of effects resulting
ultimately FROM the formation of adducts;
(d) Related Substance Assessment (taken from Reference #13 in Table 1). Here, the
sole body of data is on exposure to acetaldehyde, which is judged to be related
structurally to formaldehyde, leading to nasal tumors in a related manner. It is given the
label RSI. Note that the measurement of nasal tumors might be taken to suggest that
these data are of the form TR (Tumor Response). This suggestion is negated in the
present methodology, however, by the fact that TR data are taken to require tumor
response AS INDUCED BY THE EXPOSURE DEFINING THE CONTEXT (i.e.
formaldehyde).
Data in other data categories were not available for the studies utilized in this context.
22. Description of Working Table 2.5
The judgments for this Working Table are provided in the cells of Working Table 2.5.
The text here describes the reasons for these judgments. The data on which these judgments are
based are provided as various "Exhibits" after Working Table 2.5., with a numbering system
described in Section 2.2.4.
22.1. Data Categories and Items
22
-------
WORKING TABLE 2.5. DATA CHARACTERISTICS FOR OBSERVATIONAL CONTEXT NUMBER 5
Data Categoiy/Item
Description
Completeness
(Hi/Me/Lo/No)
Utility
(Hi/Me/Lo/No)
Strength of Effect
(Hi/Me/Lo/No)
Exposure-
Specific
Effect1
Tumor Response
Nasal tumors (TR1.1)
ME
HI
HI
WT2.C5.
TRI.I
Biophysical Effect
Thymidine uptake (BE1.1)
HI
ME
HI
WT2.C5.
BEI.I
Pharm acodynam ics
Tissue concentration (PD1.1)
Adduct formation (PD2.1)
ME
HI
LO
ME
NO
HI
WT2.C5.
PDI.I
WT2.C5.
PD2.1
Host Factors
No study available
Concurrent Environmental
Conditions
No study available
Related Substances Assessment
Acetaldehyde exposures (RS1.1)
ME
HI
HI
WT2.C5.
RSI.l
'Refers to (he Exhibit Number for the data as described in the text.
-------
WT2.C5.TR1.1
Table 1
Summary of histopathology in the nasal turbinates of rats dying during the first 18 months of exposure to formaldehyde
Diagnosis
Rhinitis, acute Squamous or Squamous Squamous
Exposure No. of rats suppurative or Epithelial Squamous epithelial papillary metaplasia with Squamous Squamous cell Spindle
level (ppm) examined seropunilent Osteomalacia dysplasia metaplasia hyperplasia hyperplasia cellular atypia papilloma carcinoma cell sarcoma
020000000000
220010000000
682086000000
15 44 38 1 1 38 U 6 3 4 28 1
Table 2
Summary of histopathology in the nasal turbinates of rats sacrificed after being exposed to formaldehyde for 18 months
Diagnosis
Rhinitis, acute Squamous or Squamous Squamous
Exposure
level (ppm)
No. of rats
examined
suppurative or
seropurulent
Focal turbinate
atrophy
Epithelial
dysplasia
Squamous
metaplasia
epithelial
hyperplasia
papillary
hyperplasia
metaplasia with
cellular atypia
Adenomatous
polyp
Squamous cell
carcinoma
0
40
0
0
0
0
0
0
0
0
0
2
40
2
0
35
24
0
0
0
1
0
6
40
7
0
37
35
0
0
0
1
0
15
40
39
2
13
39
15
4
7
1
8
-------
WT2.C5.BE1.1
A prominent response to cell loss associated with toxicity is compensatory cell
replication. Surviving cells undergo division in order to replace dead cells and
to increase the thickness of the epithelium. Preliminary studies on the effect of
15 ppm of formaldehyde vapor, 6 h/day for 5 days, have been conducted in
rats. One day prior to exposure the animals had an osmotic mini-pump contain-
ing a 7-day supply of 3H-thymidine implanted in the peritoneal cavity. Follow-
ing the fifth day of exposure the animals were anesthetized with pentobarbital
and killed by vascular perfusion with 10% neutral buffered Formalin; the heads
were decalcified, embedded, and sectioned and slides were prepared for auto-
radiography. Control rats had 5/2686 (0.19 percent) labeled respiratory epi-
thelial cells, while formaldehyde-exposed animals had 634/4712 (13/ percent"),
which represents a 71-fold increase in cell replication. Sequential pulse labeling
studies following 1, 3, 5, or 9 days of formaldehyde exposure demonstrated
that maximum cell proliferation occurred after 3 days. A 10- to 20-fold increase
in cell replication occurred when rats were exposed fnr 3 aavs to 6 or 15 ppm
formaldehyde vapor (Fig. "11; ana when mice were exposea to 15" ppm. Similar
exposures of rats to 0.5 or 2,0 ppm anri rnice to 0.5^. 2.0 or6>0 ppm formalde-
hyde (Fig. 12) did not result in increased^ell turnover in the nasal cavity.
-------
WT2.C5.PD1.1
Concentrations of formaldehyde in tissues of control (i.e. unexposed) rats and of rats exposed to formaldehyde or to
methyl chloride
ToxtCinf
CH,0*
CH3CI6
Tittu*
Nasal
mucosa
Liver
Testes
Brain
rmui CH70 concentration*
Control
(u/no' 0 1 wt)
0.42 = 0.09 (8)
0.201 r 0.017 (6)
0.28 = 0.10 (8)
0.097 = 0.014 (4)
'Exposure regimen: 6 ppm CH70. 6 h day"', 10 days.
"Exposure regimen: 3000 ppm CH^Cl. 6 h day"1. 4 days.
'Mean= SD. Number of samples shown in parentheses.
"One-tailed r-test: NS = not significant, Sig = significant.
Exoo*«d
(umoi g'' wt)
0.39 = 0.12(8)
0.41 =0.14 (7)
0.49 = 0.07 (5)
0.67 = 0.15(4)
Significance
NS( p > 0.25)
Sig (0.01 >p)
Sig (0.01 > p)
Sig (0.01 > p)
-------
WT2.C5.PD2.1
20 f lol «ESI»l«ATORT MUCOSA
¦ 0.3 90
Fig. 2 Concentrations of [l4C] CH20
equivalents in DNA, RNA, and protein of the
respiratory mucosa (a), olfactory mucosa (b), and
bone marrow (femur) (c) in rats exposed for 6 hr
to 0.3, 2, 6, 10, or 15 ppm of [14C] and [-"HJCHiO,
1 day after a single pre-expo sure (6 hr) to the
same concentration of unlabeled CH20.
Bargraphs for aqueous and interfacial DNA from
the respiratory and olfactory mucosa are
overlaid. Values shown are + SD, n = 3 for
nasal mucosa, n = 4 for bone marrow.
-------
WT2.C5.RS1.1
TABLE II
SITE. TYPE AND INCIDENCE OF RESPIRATORY TRACT TUMOURS IN RATS
EXPOSED TO VARIOUS CONCENTRATIONS OF ACETALDEHYDE VAPOUR FOR AT
MOST 28 MONTHS
Site and type of tumours Incidence of tumours a
Males, conc. group FemaJes. conc. group
Control Low Mid Top Control Low Mid Top
Nose
(49)
(52)
(53)
(49)
(50)
(48)
(53)
(53)
Papilloma
0
0
0
0
0
1
0
0
Squamous cell carcinoma
1
1
10*
15—
0
0
5
17"
Carcinoma in situ
0
0
0
1
0
0
3
5
Adenocarcinoma
0
16—
31"
• 21 —
0
6-
26"
• 21"
Larynx
(50)
(50)
(51)
(47)
(51)
(46)
(47)
(49)
Carcinoma in situ
0
0
0
0
0
0
1
0
Lungs
(55)
(54)
(55)
(52)
(53)
(52)
(54)
(54)
Poorly differentiated
0
0
0
0
0
1
0
0
adenocarcinoma
a The number of animals examined is given in brackets.
Statistics: Fisher Exact Test *P < 0.05, "P < 0.01 '"P < 0.001.
-------
The same data categories and items described with respect to Working Table 1.5 are
employed in Working Table 2.5. These are:
(a) Tumor Response, containing TR1 (incidence of nasal tumors);
(b) Biophysical Effects, containing BE1 (cellular proliferation);
(c) Pharmacodynamics, containing PD1 (tissue concentration) and PD2 (DNA adduct
formation);
(d) Related Substance Assessment, containing RSI (incidence of nasal tumors by
acetaldehyde).
These four separate data categories and five bodies of data are displayed in Working
Table 2.5.
222. Judgments of Completeness
In all cases, the judgment of completeness was based on a comparison of the number of
studies used in a data category/item for this context against the number available in the
complete reference list provided in Appendix A. The judgment was HIGH if the data used in
this Working Table was a relatively complete subset of the data indicated in the Appendix,
where "complete" was taken to mean somewhere on the order of 70% or above. The judgment
was MEDIUM if the data used in this Working Table constituted somewhere between 25 and
70% of the total body of potentially available data. The judgment was LOW if the data used
constituted less than 25% of the data available. These assignments also included limited
consideration of the QUALITY of the data, since it should not necessarily count against an
analysis if the analyst fails to collect data of poor quality. This consideration also is dealt with in
the assignment of Utility for the collected studies. Still, it is evident that the judgment of
Completeness might reasonably be expected to include consideration of something more than
the fraction of studies collected in a given data category/item. In the present analysis,
23
-------
Completeness was increased slightly if the collected data for a given data category were the most
heavily cited in the secondary literature on formaldehyde. This leads to the following judgments
of Completeness:
For the data category TR1, the judgment for Completeness is MEDIUM since this study
is 1 of 6 identified in the literature search in Appendix A (which suggests a judgment of LOW),
but the study is heavily cited as the best available. Citations to the other studies are to be found
in the full literature search: Feron et al., (1990); Kerns et al., (1982); Kerns et al., (1983);
Restani and Galli (1991); and Feron et al., (1987).
For the data category BE1, the judgment for Completeness is HIGH as this is the sole
study from Appendix A in which this quantity was measured.
For the data category PD1, the judgment for Completeness is MEDIUM as this is 1 of 2
studies identified in the full literature search.
For the data category PD2, the judgment for Completeness is HIGH as this is the only
study identified in the full literature search.
For the data category RSI, the judgment for Completeness is MEDIUM as this is one of
4 reports containing such information, but two of the others cite this report as the primary
source. The others are Bardana, jr. and Montanaro (1991); Grafstrom (1990); and Tomatis et
al., (1978).
223. Judgments of Utility, Strength of Effect, and Exposure- Specific Effect
The remaining judgments in Working Table 2.5 refer to the qualities of the data
ACTUALLY EXAMINED IN THIS EXAMPLE. By Utility, we mean the quality of the study
DESIGN and its ability to address in rigorous fashion questions pertinent to hazard
identification for formaldehyde. No attempt is made to determine the specific inferences
24
-------
towards which the data will be directed in later Working Tables, since the intent here is only to
determine the degree to which a body of data arose from a study design appropriate for
accurate and precise measurements of the quantity considered in the study. Utility is increased
by high standards of study design, including the ability of the experiment to demonstrate a
causal link between exposure to formaldehyde and the measured effect.
By Strength of Effect, we mean the degree to which the study actually revealed an effect.
Strength is increased when the effect noted possessed statistical significance.
By Exposure-Specific Effect, we mean a summary of the actual effect observed (putting
aside questions of the reliability of this effect, which was dealt with in the assignment of Utility
and Strength of Effect). This summarization is provided in the following text, and the data
themselves are provided in the various Exhibits appearing after Working Table 2.5.
The following judgments then apply to Working Table 2.5:
For the data category TR1, the Utility judgment is HIGH since the study utilized an
appropriate experimental methodology; formaldehyde exposure was well characterized; 120 male
and 120 female rats were exposed, at each exposure level; identification of tumors was by an
appropriate method; the study constituted a deliberate experiment (important to a causal claim);
exposures were long-term (18 month); appropriate controls were utilized; and a variety of other
biological measures of effect were examined to test for coherence of findings. The Strength of
Effect judgment is HIGH for exposures above 6 ppm, but LOW for 2 and 6 ppm. A statistically
significant increase in nasal tumor incidence was observed for the 15 ppm exposure group
(p < 0.05), but not for the lower exposure groups. The Exposure-Specific Effect is summarized in
Exhibit WT2.C5.TR 1.1. Incidence of squamous metaplasia was increased at all exposure levels,
but incidence of nasal carcinomas was increased only at 15 ppm.
25
-------
For the data category BE1, the Utility judgment is MEDIUM since the number of rats
examined is small (10 rats); the measurement technique was performed properly and was
adequate to detect tritiated thymidine uptake; measurement of thymidine uptake is an indicator
of, but not identical to, cellular proliferation; the study constitutes a deliberate experiment
(important to a causal claim); and appropriate controls were utilized. The Strength of Effect
judgment is HIGH for exposure to greater than 6 ppm, as there was a large, and statistically
significant, increase in thymidine uptake in the exposed rats. Judgment is LOW for exposures of
less than or equal to 6 ppm, as there was no increase in thymidine uptake noted. The Exposure-
Specific Effect is summarized in Exhibit WT2.C5.BE1.1. There was a 71-fold increase in tritiated
thymidine uptake at 15 ppm for 6 hrs/day, 5 days. A 10-20-fold increase in uptake for 6 ppm
and 15 ppm, delivered for 3 days at 6 hrs/day, was noted. No increase was noted at lower
exposures.
For the data category PD1, the Utility judgment is LOW due to the small number of rats
examined (8) and insensitivity of the measurement procedure (minimum detection limit within
25% of measured quantities). The Strength of Effect judgment is NO since no statistically-
significant difference in tissue concentrations was noted. The Exposure-Specific Effect is
summarized in Exhibit WT2.C5.PD1.1. Mean tissue concentration in nasal tissue of the exposed
group was 0.39 umol/g. Mean tissue concentration for the unexposed group was 0.42 umol/g.
For the data category PD2, the Utility judgment is MEDIUM due to an adequate mass
of tissue (100 mg) having been examined at each exposure level); there was good sensitivity in
the measurement procedures; an appropriate measurement procedure was employed; but no
information on controls is provided and only 4 rats per exposure level were examined (leading
to concerns over the complicating effects of potential intersubject variability). The Strength of
Effect judgment is HIGH since there was a statistically significant increase in formaldehyde
26
-------
adducts in the 2, 6, 10 and 15 ppm groups relative to the 0.3 ppm group. The Exposure-Specific
Effect is summarized in Exhibit WT2.C5.PD2.1. Mean adduct formation in respiratory mucosa
increased with exposures between 0.3 and 6 ppm, then decreased with exposure above 6 ppm.
The study authors concluded that adduct formation was due to covalent binding rather than
metabolic incorporation, based on the consistency of the ratio of 14C- to 3H-labeled
formaldehyde.
For the data category RSI, the Utility judgment is HIGH since the number of rats
employed was adequate (approximately 50 of each sex at each exposure level); the route of
administration was air; exposures were relatively chronic for 2 years; appropriate techniques
were used to analyze for tumor incidence; appropriate controls were used; the study constitutes
a deliberate experiment (important to a claim of causality). The justification for considering
acetaldehyde to be a "related substance" is given in Primary Reference #9 as "the biological
properties of aldehydes are governed by the presence of a ketone moiety", which is possessed
both by formaldehyde and acetaldehyde. The Strength of Effect judgment is HIGH since the
incidence of nasal carcinomas and adenocarcinomas was elevated with statistical significance in
all exposure groups. The Exposure-Specific Effect is summarized in Exhibit WT2.C5.RS1.1.
There is a dose-related increase in the incidence of nasal cancers when the total of all forms of
cancer is considered. Between the middle and top exposure groups, this incidence was
redistributed from primarily adenocarcinomas (middle exposure group) to an approximately
equal number of adenocarcinomas and squamous cell carcinomas (high exposure group).
22.4. Exhibits for Working Table 25
The data for the various data categories are provided as "Exhibits" after Working Table
2.5. The numbering system for the exhibits was chosen to display information on (in order of
27
-------
appearance) the Working Table; the Context; the specific effect measured; and the study
number within that specific effect. For example, Exhibit WT2.C5.TR1.1 refers to Working Table
2 (all exhibits have this assignment in common); Context 5; Tumor Response of type 1; and data
set #1 within the data category TR1. This same numbering scheme will be employed in all other
Contexts.
23. Description of Working Table 3.5
23. L Introduction
The central task of Working Table 3.5 is to draw inferences on the carcinogenicity of
formaldehyde WITHIN CONTEXT 5, employing data specific to that context. This task requires
that several judgments be made sequentially:
(a) The analyst must determine the particular Claims of Carcinogenicity to be considered
in the analysis. These various claims are discussed in detail in Section 2.3.4.
(b) The available Relevance Strategies by which a given body of data may be related to a
specific Claim of Carcinogenicity must be developed. These strategies were discussed in
detail in Section 1.4., and the same definitions are applied here.
(c) The analyst must assign a measure of Intellectual Obligation to each Relevance
Strategy. This assignment was described in detail in Section 1.4., and the same
assignments are made here.
(d) The analyst must judge the epistemic status of a specific Claim of Carcinogenicity
arising from a specific Relevance Strategy. As described in Section 1.4., this judgment is
separate from the judgment of Intellectual Obligation for that Relevance Strategy. This
judgment is described in more detail in Section 2.3.2. The assignments are entered into
the appropriate cells of Working Table 3.5.
(e) The analyst must produce a summary judgment of the epistemic status of a particular
Claim of Carcinogenicity by combining the judgments from each of the five Relevance
Strategies for that claim. This summary includes consideration of both the epistemic
status of that claim for each Relevance Strategy, as entered into the separate cells of the
Working Table specific to each strategy, as well as the Intellectual Obligation assigned to
that strategy. The assignments are entered into the appropriate COLUMN SUMMARY
cells of Working Table 3.5.
28
-------
WORKING TABLE 3.5. INTRA-CONTEXT SUPPORT FOR CLAIMS OF CARCINOGENICITY IN CONTEXT 5
Claims of Carcinogenicity
Relevance Strategy
I.O.1
Increases
Incidence of
Cancer
Classification(s)
Stage
Mechanism
Complete
Partial
Mixer
Helper
Neo.
Conv.
Neo.
Devel.
Geno-
tox.
Non-
genot
ox.
Direct Empirical (D.E.)
HI
HI
NO
NO
NO
NO
NO
NO
HI
NO
Semi-Empirical
Extrapolation (S.E.E.)
ME
NO
NO
NO
NO
NO
NO
NO
NO
NO
Empirical Correlation
(B.C.)
ME
ME
NO
NO
NO
NO
HI
ME
HI
NO
Theory-based Inference
(T.B.I.)
ME
HI
NO
NO
NO
NO
ME
ME
ME
NO
Existential Insight(E.I.)
LO
HI
NO
NO
NO
NO
ME
ME
HI
NO
Column Summary
HI
NO
NO
NO
NO
ME
LO
HI
NO
Overall Summary
HI
NO
NO
NO
NO
ME
LO
HI
NO
'Intellectual Obligation
-------
(f) The analyst must produce a summary judgment of the COHERENCE of the various
Claims of Carcinogenicity. This task is performed by examining the coherence ACROSS
the various COLUMN SUMMARIES appearing in the Working Table. The assignments
are entered into the appropriate OVERALL SUMMARY cells of Working Table 3.5.
2.3.2. Epistemic Status of Claims of Carcinogenicity
The entries into the various cells of Working Table 3.5. reflect the judgment of the
analyst concerning the epistemic status of the Claim of Carcinogenicity being considered, FOR
A SPECIFIC RELEVANCE STRATEGY. This judgment will have two components:
(a) Each Relevance Strategy calls upon a selected body of data from which the reasoning
will proceed. Specifically, the Direct Empirical strategy requires data on direct measurements of
the effect underlying the Claim of Carcinogenicity, made at the exposure to formaldehyde
defining the context (here, concentrations above 2 ppm). The Semi-Empirical Extrapolation
strategy requires data on direct measurements of the effect underlying the Claim of
Carcinogenicity, made at exposures to formaldehyde other than those defining the context. The
Empirical Correlation strategy requires data on effects other than the effect underlying the
Claim of Carcinogenicity, where this "other" effect is taken to correlate with appearance of the
effect of interest. The Theory-Based Inference strategy requires data on an effect other than the
effect underlying the Claim of Carcinogenicity, where this "other" effect is taken to be a causal
step in the production of the effect of interest as specified by an established theory. The
Existential Insight strategy does not invoke data, except in the sense that a subjective judgment
from an individual might be counted as an instance of data in and of itself.
The analyst must assign the "foundational quality" to the data used in each relevance
strategy. By "foundational quality", we mean the quality of the data with respect to the question
those data were intended to address BY THE ORIGINAL INVESTIGATOR. This judgment is
based on the results of Working Table 2.5 and includes consideration of the Completeness,
29
-------
Utility, Strength of Effect and Exposure-Specific Effect from that table. While the judgment
ultimately is subjective, it is constrained in that it must increase (from NO to LOW to
MEDIUM to HIGH) as the strength of the judgments for Completeness, Utility and Strength of
Effect increase in Working Table 2.5. The summary of Exposure-Specific Effect may also enter
into the assignment of epistemic status to a Claim of Carcinogenicity if the analyst determines
that only effects above a given magnitude will exert an influence on carcinogenic processes (an
issue of importance when possible thresholds are considered).
(b) The analyst then must judge the strength of any "background premises" required by
the Relevance Strategy under consideration. These background premises are defined in detail in
the original report, but may be summarized as the set of (often implicit) premises that must be
introduced into an analysis if the data in Working Table 2.5. are to be used to draw inferences
about the Claims of Carcinogenicity for the context being considered. Presumably, many of these
background premises already are considered valid once Working Tables 1.5. and 2.5. are
complete; otherwise, the data would not have been assigned to this context. But other
background premises concerning the reliability of extrapolation procedures (for the Semi-
Empirical Extrapolation relevance strategy), the strength and specificity of correlations (for the
Empirical Correlation relevance strategy), the degree of verification and validation of theories
(for the Theory-Based Inference relevance strategy), and the credentials of individuals making
subjective judgments (for the Existential Insight relevance strategy) must be warranted prior to
their use.
Having established these two components of epistemic status (foundational quality and
background premises) for a given Claim of Carcinogenicity, the analyst then produces a
composite judgment of the epistemic status of a given Claim of Carcinogenicity and enters this
into the appropriate cell of Working Table 3.5. There are two separate claims that might be
30
-------
made within one of these cells. The first is the claim that formaldehyde DOES produce the
effect underlying that claim (such as increased incidence of cancer or genotoxic effects). In that
case, the assignments of NO/LOW/MED/HIGH refer to the epistemic status of the claim that
formaldehyde produces the effect. This assignment is entered into the area ABOVE the dashed
line in a cell. The second claim is that formaldehyde DOES NOT produce the effect underlying
the claim. In that case, the assignment of NO/LOW/MED/HIGH refer to the epistemic status
of the claim that formaldehyde DOES NOT produce the effect. This assignment is entered into
the area BELOW the dashed line in a cell. THE READER SHOULD NOTE THAT THIS IS A
NEW FEATURE OF THESE WORKING TABLES NOT DISCUSSED IN THE ORIGINAL
REPORT. The reason for this addition to the Working Tables is that the fact that a particular
body of data does not support the claim that formaldehyde does induce a specific effect IS NOT
NECESSARILY evidence that formaldehyde does not induce that effect. It may simply be the
case that the data were obtained under conditions incapable of revealing any effect that might
be present.
This consideration is important due to the issue of "burden of proof. At times, the
burden may lie on the regulatory agency to show that formaldehyde DOES induce cancer.
Attention then would be directed towards the judgments in the upper halves of the various cells
of Working Table 3.5. At other times, the burden may lie on the producer of formaldehyde to
show that formaldehyde DOES NOT induce cancer. Attention then would be directed towards
the judgments in the lower halves of the various cells of Working Table 3.5. IN THE PRESENT
EXAMPLE, ONLY JUDGMENTS THAT AN EFFECT IS PRODUCED BY
FORMALDEHYDE ARE DEVELOPED. THE PROCESS OF ANALYSIS WOULD BE
REPEATED FOR THE LOWER HALVES OF ALL CELLS IF THE BURDEN OF PROOF
WAS SHIFTED.
31
-------
It is important here to consider the broad features of the theory of carcinogenesis to be
used throughout this section. These features structure the interpretation of data used in the
various Relevance Strategies. The theory of carcinogenesis employed here is that cells are
transformed to cancer by three steps or transitions. These are termed initiation, promotion and
progression, each of which may require one or more sub-steps. For each transition to occur, a
biologically-significant dose-rate must reach the targets in appropriate cells so that an interaction
between the substance or a metabolite may take place. This target may or may not have been
identified for a particular substance and tumor type.
The initiation step is taken to involve alteration of DNA structure and function,
producing a new genome. This alteration must then remain intact without being repaired by
processes normally operating in cells or induced by the damage itself. The ability of a substance
to bring a stable alteration about may be determined from an increase in cancer incidence
following (1) application of the initiating agent and (2) subsequent application of a promoting
agent such as TPA. In the absence of such data, the ability of a substance to interact with DNA,
to alter DNA, and/or to induce mutation provides partial but not conclusive evidence of
initiating activity. In addition, the ability of a substance to lower DNA repair rates or fidelity of
repair is partial but not conclusive evidence that the substance will enhance the initiating activity
of other substances or of background events.
The promotion step is taken to involve loss of growth control, represented by an
imbalance between rates of growth and death of cells in a colony and a continued net rate of
expansion even upon reaching confluence. The mechanism for this step is unspecified, but may
result in either a lowering of the rate of removal of cells (as in decreased apoptosis) or an
increase in the rate of division (as in abnormal proliferation). The ability of a substance to bring
this about may be determined from an increase in transformation following (1) prior application
32
-------
of a known initiating agent and (2) subsequent application of the substance of interest. In the
absence of such data, the ability of a substance to alter intercellular communication, the rate of
cellular division, the rate of apoptosis, and/or net rate of growth provides partial but not
conclusive evidence of promoting activity.
The progression step is taken to be poorly understood at present, but probably is related
to the metastatic properties of the promoted cells. There is no assay for progression at present.
Partial but inconclusive evidence of progressing activity might be obtained from a demonstration
that cells have taken on the ability to dislodge from the primary tumor, spread to other organs
or tissues, reattach, and grow to fatal tumors.
If a substance induces only one of the required transitions, it will be necessary for other
transitions to be induced by antecedent conditions if the substance of interest is to be
considered a Partial carcinogen (as defined in Section 2.3.4). The fact that the background
incidence of cancer is not zero may be taken to indicate that each of these other transitions will
always be present. If this assumption is adopted, the ability of a substance to increase the
incidence of cancer cannot be used to differentiate between claims of a Complete or a Partial
carcinogen (as defined in Section 2.3.4). Existing empirical evidence, however, is suggestive of
there being several routes by which cancer may occur. It is not inconsistent to assume that a
substance acts to induce only a subset of transitions, and that the remaining transitions required
are not those associated with the mechanism by which the antecedent conditions of the
experimental context act to bring about background cancer.
The epistemic status of this Theory of Carcinogenesis is taken here to be MED based on
an examination of the available empirical evidence linking DNA alteration, alteration of rates of
growth, and the carcinogenic potential of a range of substances tested in assays for initiation,
promotion and complete cancer. A judgment of HIGH is not considered appropriate since much
33
-------
of the warrant for the theory is subjective judgment of experts in the field based on the
coherence of the explanatory framework and the finding of (unexplained) correlations.
2.3.3. Intellectual Obligation
The issue of Intellectual Obligation was discussed in detail in Section 1.4. Since the
judgments and assignments are common to all Working Tables and Contexts, this discussion is
not repeated here. By way of a summary, the Direct Empirical Relevance Strategy was assigned
an Intellectual Obligation of HIGH; the Semi-Empirical Extrapolation, Theory-Based Inference
and Empirical Correlation Relevance Strategies were assigned an Intellectual Obligation of
MEDIUM; and the Existential Insight Relevance Strategy was assigned an Intellectual
Obligation of LOW. The assignments are entered into the appropriate cells of Working Table
3.5.
23.4. Claims of Carcinogenicity
The framework employed here allows for a wide variety of claims concerning the
carcinogenic action of formaldehyde. In the original report, it was determined that it is desirable
to augment the claim that a substance increases the incidence of cancer, since this sole claim
does not provide full detail on the antecedent conditions under which a substance such as
formaldehyde can exert its carcinogenic potential. It was determined that a more detailed
analysis would subdivide the claim of increased incidence of cancer into more detailed claims
potentially of use in determining the most effective regulatory strategy. The different Claims of
Carcinogenicity employed in Working Table 3.5 (and all subsequent Working Tables discussed in
this report) are:
(a) Increases Incidence of Cancer, implying that formaldehyde raises the incidence of
cancer, raises the multiplicity of cancer at sites, and/or changes the age at which cancer
appears.
34
-------
(b) Complete Carcinogen, implying that formaldehyde increases the incidence of cancer
even when it is not delivered in conjunction with concurrent exposures.
(c) Partial Carcinogen, implying that formaldehyde increases the incidence of cancer only
when delivered in conjunction with concurrent exposures that complete the process of
carcinogenesis.
(d) Mixer, implying that formaldehyde exerts its carcinogenic effect only when it is
combined with another substance, and where this combination results in a new substance
which is at least a partial carcinogen.
(e) Helper, implying that formaldehyde does not in and of itself induce transitions to
cancer, but that it facilitates the carcinogenic action of another substance.
(f) Neoplastic Conversion, implying that formaldehyde acts by converting cells to
neoplastic potential.
(g) Neoplastic Development, implying that formaldehyde acts by producing growth in
colonies of cells already having neoplastic potential.
(h) Genotoxicity, implying that formaldehyde acts through a mechanism involving
interaction with the genetic material of a cell.
(i) Non-genotoxicity, implying that formaldehyde acts through a mechanism other than
interaction with the genetic material of a cell.
The following summary considerations from Working Table 2.5. provide the foundations
for all inferences to Claims of Carcinogenicity drawn in this context:
TR1.1. demonstrates that in vivo exposure of rats to airborne formaldehyde at 15 ppm
increases the incidence of nasal carcinomas and the degree of metaplasia. The epistemic status
of this claim is HIGH since completeness, utility, and strength of effects are HIGH for these
data. At exposures below 15 ppm, the findings are weak and not statistically significant, so the
assignment of epistemic status for the claim of increased incidence is LOW for these exposures.
At both the 2 and 6 ppm exposures, however, there is good evidence that formaldehyde induces
both epithelial dysplasia and squamous metaplasia (relevant to claims of promotional action),
with this claim being assigned an epistemic status of HIGH since completeness, utility, and
strength of effects are each HIGH.
35
-------
BE1.1. demonstrates that in vivo exposure of rats to airborne formaldehyde results in an
increase in incorporation of tritiated thymidine into the cells of nasal passages. The epistemic
status of this claim is scored HIGH for both the 6 and 15 ppm exposure levels, since
completeness was HIGH, utility was MED, and strength of effect was HIGH. While the number
of rats exposed was small (opening the possibility that inter-subject variability accounted for any
measured differences), the analysts believe that such variability will be rather small in this
species owing to the deliberate genetic similarity of the individuals. Additionally, the effect was
so strong that it does not seem reasonable to believe that variability alone might account for the
observed differences between exposed groups and controls.
This finding is understood by the study authors to indicate an increase in cellular
proliferation, which is also relevant to the issue of carcinogenicity. The present analysts assign
this claim an epistemic status somewhere between MED and HIGH since there clearly is
increased incorporation of tritiated thymidine, the theory linking incorporation and proliferation
is rather well established (with MED epistemic status), and there is a clear correlation between
incorporation and proliferation (with HIGH epistemic status).
PD1.1. demonstrates that formaldehyde produces a burden (or concentration) in nasal
tissues of rats exposed in vivo to airborne formaldehyde. This claim is assigned an epistemic
status of NO since the utility and strength of effect were so low as to call into question whether
the study can provide a reliable measure of tissue burden.
PD2.1. demonstrates that in vivo exposure of rats to airborne formaldehyde results in
DNA adducts within the cells of nasal passages. This claim is assigned an epistemic status of
HIGH since the completeness and strength of effects were HIGH and the utility was MED. The
utility score of MED was based on the fact that no control data were reported, but the study
36
-------
does show a statistically significant increase in the 2, 6, 10, and 15 ppm exposure groups relative
to the 0.3 ppm group.
RS1.1. demonstrates that in vivo exposure of rats to airborne acetaldehyde increases the
incidence of nasal carcinomas. This claim is assigned an epistemic status of HIGH since the
utility and strength of effect for the highest exposure groups were HIGH. This is not outweighed
by the MED completeness score, although another analysts might use that to lower the overall
epistemic status to MED.
Acetaldehyde is considered by the present analysts to be a Related Substance to
formaldehyde because, as the next more-complex aldehyde, its structural similarity to
formaldehyde is believed pertinent to the carcinogenic activity of the aldehyde class. The
epistemic status of the claim that acetaldehyde is a Related Substance in this context for the
issue of carcinogenicity is scored MED since there is incomplete information about the
mechanism of carcinogenic action of the aldehyde class. This suggests that one cannot claim
unequivocally that the structural similarities are firmly demonstrated to be the relevant ones for
carcinogenicity.
The separate judgments for Claims of Carcinogenicity for Working Table 3.5. are
provided in the sections that follow. In the case of each claim, explanations are given of the
judgment for each Relevance Strategy; then for the Column Summary across Relevance
Strategies for that claim; then for the Overall Summary across claims.
2.3.4.1. Increases Incidence of Cancer
The judgments for this Claim of Carcinogenicity for the various Relevance Strategies and
for this context are shown in the cells of Working Table 3.5. The basis for these judgments are
given below.
37
-------
A. Direct Empirical: The judgment is HIGH for the 15 ppm exposure group since
completeness was MED, utility was HIGH, and strength of effect was HIGH (cancer incidence
was increased with statistical significance in the 15 ppm group); for the other exposures groups,
this score must be LOW.
B. Semi-Empirical Extrapolation: The judgment is NO since exposures were in the range
that defines this context.
C. Empirical Correlation: The judgment is MED since both increased incorporation of
thymidine and production of DNA adducts correlate (MED epistemic status for the correlation)
with carcinogenicity.
D. Theory-Based Inference: The judgment is HIGH since the studies show (1) that DNA
adducts are being formed upon inhalation (high epistemic status), (2) that a structurally-similar
aldehyde also induces nasal tumors (also high epistemic status) and (3) that proliferation of cells
is stimulated (high epistemic status), which may be understood as some evidence for
promotional effects under the Theory of Carcinogenesis employed here. The final epistemic
status score might well be lowered to MED since there is neither identification of the active
metabolite(s) nor confirmation that significant tissue burden results from exposure in this
context, but the judgment of these analysts is not changed by this.
E. Existential Insight: Judgment is HIGH since these data produce in the analysts strong
confidence in formaldehyde's carcinogenic activity for this context. It is further judged that the
analysts possess the relevant experience, training, and psychological skills necessary to form such
subjective judgments, based on long work experience in the carcinogenesis sciences.
F. Column Summary: The judgment is HIGH since all relevance strategies were either
HIGH or NO, and since Direct Empirical evidence, carrying a HIGH intellectual obligation was
scored HIGH for at least one exposure level.
38
-------
G. Overall Summary: The judgment is HIGH since the analysts' strong foundationalist
epistemological stance would not admit that coherence across claims could affect the epistemic
status of any one claim. If, however, such a stance were not taken and coherence were insisted
upon, then the judgment could still be HIGH because the Direct Empirical score would continue
to dominate.
2.3.4.2. Classifications
2.3.4.2.1. Complete Carcinogen
The judgment in all cells is NO since cancer incidence was measured rather than the
ability of formaldehyde to induce specific steps in carcinogenesis. Since the background
incidence of cancer was non-zero, it is clear that each transition required for cancer occurs with
some non-zero background probability in the population of cells. As a result, it is not possible
to distinguish between the hypothesis that formaldehyde induces ALL transitions in the 15 ppm
group (and is, therefore, a complete carcinogen), and the hypothesis that formaldehyde induces
only a subset of the transitions with the remainder being completed by background transitions
(and is, therefore, a partial carcinogen). The analysts recognize, however, that the definition of a
complete carcinogen COULD be taken to be a substance that does not require the concurrent
application of either a known initiator or promoter. In that case, the present studies could be
taken to indicate (with an epistemic status of HIGH) that formaldehyde is a complete
carcinogen.
2.3.4.2.2. Partial Carcinogen
The judgment in all cells is NO since cancer incidence was measured rather than the
ability of formaldehyde to induce specific steps in carcinogenesis. Since the background
39
-------
incidence of cancer was non-zero, it is clear that each transition required for cancer occurs with
some non-zero background probability in the population of cells. As a result, it is not possible to
distinguish between the hypothesis that formaldehyde induces ALL transitions in the 15 ppm
group (and is, therefore, a complete carcinogen), and the hypothesis that formaldehyde induces
only a subset of the transitions with the remainder being completed by background transitions
(and is, therefore, a partial carcinogen). The analysts recognize, however, that the definition of a
complete carcinogen COULD be taken to be a substance that does not require the concurrent
application of either a known initiator or promoter. In that case, the present studies could be
taken to indicate (with an epistemic status of HIGH) that formaldehyde is a complete
carcinogen rather than a partial carcinogen (i.e. an epistemic status of NO).
2.3.4.2.3. Mixer
The judgment in all cells is NO since the assays did not examine formaldehyde in
conjunction with other exposures.
2.3.4.2.4. Helper
The judgment in all cells is NO since the assays did not examine formaldehyde in
conjunction with other exposures.
2.3.4.3. Stapes
2.3.4.3.1. Neoplastic Conversion
A. Direct Empirical. The judgment is NO since neoplastic conversion was not
determined directly.
40
-------
B. Semi-Empirical Extrapolation. The judgment is NO since exposures were in the range
defining this context.
C. Empirical Correlation. The judgment is HIGH since the production of DNA adducts
(epistemic status of HIGH) correlates with the ability of a substance to induce neoplastic
conversion (epistemic status of HIGH for this correlation).
D. Theory-Based Inference. The judgment is MED since the study of DNA adducts
provides clear evidence that DNA damage is taking place after in-vivo exposure to
formaldehyde. It is not the case, however, that DNA adducts in the form identified here have
been clearly implicated as a causal step in the transition to cancer, or even in the transition to
neoplastic conversion. In addition, the active metabolite is not identified, and tissue burdens
were not found.
E. Existential Insight: The judgment is MED since these data produce in the analysts
moderate confidence in formaldehyde's neoplastic activity for this context. It is further judged
that the analysts possess the relevant experience, training, and psychological skills necessary to
form such subjective judgments, based on long work experience in the carcinogenesis sciences.
F. Column Summary. The judgment is MED since Empirical Correlation was HIGH and
Theory-Based Inference was MED, but Direct Empirical was NO.
G. Overall Summary: The judgment is MED since the analysts' strong foundationalist
epistemological stance would not admit that coherence across claims could affect the epistemic
status of any one claim. If, however, such a stance were not taken and coherence were insisted
upon, then the judgment could still be MED because the lack of Direct Empirical evidence
could not be outweighed by findings in other claims.
2.3.4.3.2. Neoplastic Development
41
-------
A. Direct Empirical. The judgment is NO since neoplastic development was not
determined directly.
B. Semi-Empirical Extrapolation. The judgment is NO since exposures were in the range
defining this context.
C. Empirical Correlation. The judgment is MED since the increase in thymidine uptake
(epistemic status of HIGH) correlates with the ability of a substance to induce cellular
proliferation (epistemic status of HIGH for this correlation and cellular proliferation correlates
with neoplastic development (epistemic status is MED for this correlation).
D. Theory-Based Inference. The judgment is MED since the study of thymidine uptake
suggests that cellular proliferation might be taking place in-vivo, and proliferation is taken within
the Theory of Carcinogenesis used here to be important in promotion. The lack of identified
active metabolite and burden in tissue weakens the epistemic from a potential HIGH value
based on the thymidine studies to the MED selected.
E. Existential Insipht: The judgment is MED since these data produce in the analysts
moderate confidence in formaldehyde's neoplastic activity for this context. It is further judged
that the analysts possess the relevant experience, training, and psychological skills necessary to
form such subjective judgments, based on long work experience in the carcinogenesis sciences.
F. Column Summary. The judgment is LOW since Direct Empirical was NO but
Empirical Correlation and Theory-Based Inference were MED.
G. Overall Summary: The judgment is LOW since the analysts' strong foundationalist
epistemological stance would not admit that coherence across claims could affect the epistemic
status of any one claim. If, however, such a stance were not taken and coherence were insisted
upon, then the judgment could still be LOW because the lack of Direct Empirical evidence could
not be outweighed by findings in other claims.
42
-------
2.3.4.4. Mechanisms
2.3.4.4.1. Genotoxic
A. Direct Empirical: The judgment is HIGH if by "genotoxic" one means simply that
genetic alterations occur since the formation of adducts provide HIGH support for the claim
that formaldehyde is able to induce at least one form of genotoxicity. If, however, "genotoxic" is
taken to mean that a transition to cancer occurs through this genetic alteration, the judgment
would be NO since the studies do not provide observations of the causal role of the adducts.
The former definition has been adopted here.
B. Semi-Empirical Extrapolation: The judgment is NO since exposures were in the range
defining this context.
C. Empirical Correlation: The judgment is HIGH since the production of DNA adducts
(characterized by an epistemic status of HIGH) correlates well (high strength and specificity)
with the ability of a substance to induce genotoxicity. In fact, this could be taken as a tautology
if by "genotoxic" one means simply the ability to induce SOME form of DNA alteration
regardless of whether that alteration is known to be pertinent to cancer. Even if the definition of
genotoxic is taken to imply a genetic alteration leading to a transition between cancer states, the
judgment is MED since adducts correlate (MED epistemic status to correlation) with initiation.
D. Theory-Based Inference: The judgment is MED since the study of adducts formation
provides clear evidence (epistemic status is HIGH) that DNA damage is taking place following
in-vivo exposure to formaldehyde. It is not the case, however, that adduct formation exchange
has been well demonstrated to represent the kind of genotoxicity relevant to cancer in the
transition to neoplastic conversion. In addition, the lack of information on active metabolite and
tissue burden weakens the epistemic status from a potential value of HIGH.
43
-------
E. Existential Insight: The judgment is HIGH since these data produce in the analysts
strong confidence of formaldehyde's ability to induce genotoxicity. It is judged further that the
analysts possess the relevant experience, training and psychological skills necessary to form such
subjective judgments, based on having worked in the science of carcinogenesis for several years.
F. Column Summary: The judgment is HIGH since the Direct Empirical strategy, with
an Intellectual Obligation of HIGH, was HIGH for the data that were available. The presence of
HIGH or MED scores in several other relevance strategies strengthens this assignment.
G. Overall Summary: The judgment is HIGH since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim. If, however, such a stance were not taken and
coherence were insisted upon, the judgment still would be HIGH because the HIGH score in
four out of the five cells could not be outweighed by the inconclusive findings.
2.3.4.4.2. Non-Genotoxic
The judgment in all cells is NO since the studies did not examine mechanisms of action
related to other than genotoxicity.
2.3.5. Summary Comments for Intna-Context Claims
The entries into the cells of Working Table 3.5 indicate that formaldehyde should be
considered a carcinogen in rats when exposure is through inhalation at concentrations at or
above 2 ppm. This judgment is reflected most clearly in the fact that the "Overall Summary" cell
for "Increases Incidence of Cancer" is judged to be HIGH. While the data employed here do not
allow determination of the CLASSIFICATION into which formaldehyde should be placed, they
do provide support for the claim that formaldehyde induces the STAGE "Neoplastic conversion",
44
-------
a finding given additional support by the claim that formaldehyde acts by at least one
"Genotoxic" MECHANISM. The judgment that formaldehyde in this context acts at the STAGE
of "Neoplastic Development" was weakened by the fact that thymidine incorporation was not
taken to be a direct observation of cellular proliferation or neoplastic development, and that
Theory-Based Inference was assigned only MEDIUM Intellectual Obligation. If thymidine
incorporation is considered a relatively Direct Empirical confirmation of neoplastic
development, or if Theory-Based Inference was HIGH Intellectual Obligation, the epistemic
status of the claim of "Neoplastic Development" would be elevated to MEDIUM or HIGH
(which would then be consistent with the judgments made by other analysts in the formaldehyde
literature based on a larger set of data).
2.4. Description of Working Table 4.5
2.4.1. Introduction
Working Table 4.5 is utilized to justify premises necessary for extrapolation of the
findings in Working Table 3.5 (the "Observational Context") to the "Target Context" (Context
12). Four separate claims must be warranted, each of which is described in detail in the original
report and in lesser detail below. The data on which judgments for these premises are based
were described in Section 2.3.4. All judgments of Intellectual Obligation are as described in
Section 1.4.
2.42. Column Headings in Working Table 4.5
These extrapolation premises or claims as they appear in Working Table 4.5 are:
45
-------
WORKING TABLE 4.5. SUPPORT FOR INTER-CONTEXT EXTRAPOLATION PREMISES
CONTEXT NO. 5 TO CONTEXT NO. 12
Relevance Strategy
I.O.1
Exposure to BSDR
Conversion
BSDR to Effect
Conversion
Host Factors
Environmental
Conditions
Direct Empirical (D.E.)
HI
NO
NO
LO
ME
Semi-Empirical Extrapolation
(S.E.E.)
ME
NO
NO
NO
NO
Empirical Correlation (E.C.)
ME
ME
ME
ME
LO
Theory-based Inference (T.B.I.)
ME
ME
HI
ME
ME
Existential Insight (E.I.)
LO
ME
HI
ME
ME
Overall Assessment
LO
LO
LO
ME
'Intellectual Obligation
-------
1. EXPOSURE TO BSDR: Any differences between the BIOLOGICAL components of
the two contexts with respect to the relationship between exposure and BSDR are not so large
as to call into question whether a BSDR might be present in one context but not the other AT
THE SAME LEVEL OF EXPOSURE.
The pertinent considerations here for each context are whether (1) an exposure occurs
through some compartment of the environment (a compartment refers to the medium of the
environment, such as air or water); (2) an intake by some route can occur (a route refers to
inhalation, ingestion, dermal absorption, intubation or injection); (3) this intake results in an
uptake to a target organ or tissue; (4) clearance from the target allows the production of a
burden; and (5) biological transformation to the active metabolite exists. Where extrapolation is
across exposure levels, it is particularly important to establish that there does not exist a
threshold exposure below which a BSDR is not produced (due, e.g., to necessary saturation of
metabolic transformation). Where extrapolation is across species, it is particularly important to
show that any such threshold, if it exists, does not differ between the species in a manner calling
into question this premise.
2. BSDR TO BIOPHYSICAL EFFECT: Any differences between the BIOLOGICAL
components of the two contexts with respect to the relationship between BSDR and the
production of biophysical effects necessary for transitions are not so large as to call into
question whether a biophysical effect might be present in one context but not the other,
ASSUMING THE SAME LEVEL OF BSDR IN BOTH CONTEXTS.
The pertinent considerations here for each context are whether (1) the active metabolite
can interact with causally important biological structures in the organism; (2) this interaction
results in either DNA alterations (for initiation) or stimulation of cellular division (for
promotion); and (3) these biophysical effects are of the type and magnitude necessary to induce
46
-------
transitions designated as either neoplastic development or neoplastic conversion. Where
extrapolation is across exposure levels, it is particularly important to establish that there does
not exist a threshold BSDR below which a biophysical effect is not produced (due, e.g., to
necessary saturation of a pool of DNA damage or a stimulus to cellular proliferation). Where
extrapolation is across species, it is particularly important to show that any such threshold, if it
exists, does not differ between the species in a manner calling into question this premise.
3. ENVIRONMENTAL CONDITIONS: Any differences between the
ENVIRONMENTAL components of the two contexts with respect to the relationship between
exposure (to the substance of interest) and BSDR, or the relationship between BSDR and the
production of biophysical effects necessary for transitions, are not so large as to call into
question whether a biophysical effect might be present in one context but not the other AT THE
SAME LEVEL OF EXPOSURE TO THE SUBSTANCE OF INTEREST.
The pertinent considerations here for each context are whether (1) environmental
conditions increase or decrease the BSDR resulting from a given level of exposure to the
substance of interest, relative to the conditions that exist in the target context; (2) environmental
conditions increase or decrease the probability of biophysical effects from a given level of
BSDR, relative to the target context; and (3) environmental conditions increase or decrease the
probability of transitions from a given level of biophysical effect, relative to the target context.
4. HOST FACTORS: Any differences between the BIOLOGICAL components of the
two contexts with respect to the relationship between the production of biophysical effects, the
probability of transitions necessary for cancer, and the probability of cancer itself are not so
large as to call into question whether an increase in cancer incidence might be present in one
context but not the other AT THE SAME LEVEL OF BIOPHYSICAL EFFECT.
47
-------
The pertinent considerations here for each context are whether (1) a target organ or
tissue exists; (2) biological structures on which relevant biophysical effects are produced exist;
(3) relevant background transitions exist; and (4) repair of biophysical effects, or regression of
states of neoplastic conversion or neoplastic development, exist. Where extrapolation is across
species, it is particularly important to establish that there does not exist a threshold level of
biophysical effect below which a transition is not produced (due, e.g., to necessary saturation of
a stimulus to cellular proliferation before promotion can occur).
In the original report, Working Table 4 (bearing in mind that Working Table 4 in the
present report was numbered Working Table 5 in the original report, as described in Section 1)
also contained an extrapolation premise concerning the role of intersubject and intrasubject
variability. This premise has been removed here and consideration of the role of differences in
intra- and intersubject variability between the two contexts has been incorporated into the 4
separate premises above. This choice was made because variability always refers to variability of
some PROPERTY, and this property would appear in one of the four extrapolation categories
listed above. The analyst then must determine the significance of variability on the relationship
between EXPOSURE AND BSDR, BSDR AND BIOPHYSICAL EFFECT,
ENVIRONMENTAL CONDITIONS, AND HOST FACTORS separately.
2.43. Epistemic Status of Extrapolation Premises
This section summaries the judgments made for each of the cells in Working Table 4.5.
The procedure for assigning epistemic status to a cell is identical to that outlined for Working
Table 3.5 and a description is not repeated here.
2.4.3.1. Exposure to BSDR
48
-------
A. Direct Empirical: The judgment is NO since BSDR was measured in the rats but not
in the humans.
B. Semi-Empirical Extrapolation: The judgment is NO since BSDR was measured in the
rats but not in the humans.
C. Empirical Correlation: The judgment is MED since the production of a nasal dose
following exposure to formaldehyde in rats is taken to correlate strongly with production of a
nasal dose following exposure of humans to formaldehyde, but metabolic transformation has not
been identified in the human context in the studies employed here. Where the active metabolite
has not been identified, the correlation between exposure and BSDR (rather than dose) will be
weakened. The judgment is not HIGH since it has not been demonstrated that there is not a
threshold level of exposure necessary for production of a BSDR, or that this threshold is the
same between contexts.
D. Theory-Based Inference: The judgment is MED since exposure clearly occurs in both
contexts; intake clearly occurs in both cases, although at a lower level in the target context than
in the observational context; the absorption of formaldehyde in nasal passages upon inhalation
in both humans and rats has been established with MED epistemic status regardless of level of
exposure; clearance is present but not infinitely rapid in both contexts; but the active metabolite
has not been shown present in the human context. While the theory of pharmacodynamics
suggests that a dose should be produced in both contexts, it is not clear that a BSDR will be
produced since it is possible that a mechanism of metabolic transformation is stimulated or
saturated in rats but not in humans. The judgment is not HIGH since it has not been
demonstrated that there is not a threshold level of exposure necessary for production of a
BSDR, or that this threshold is the same between contexts.
49
-------
E. Existential Insight: The judgment is MED since these data produce in the analysts
moderate confidence that a BSDR is produced in both contexts.
F. Overall Summary: The judgment is LOW since Empirical Correlation, Existential
Insight and Theory-Based Inference were of MED strength, but none of these has HIGH
Intellectual Obligation. The lack of identification of the active metabolite is troubling
theoretically in cases of extrapolating across species.
2.4.3.2. BSDR to Effect
A. Direct Empirical: The judgment is NO since the same pertinent biophysical effects
were not measured in Context 12 and Context 6.
B. Semi-Empirical Extrapolation: The judgment is NO since this issue was addressed in
A above.
C. Empirical Correlation: The judgment is MED since the ability of the active metabolite
of a substance to induce biophysical effects UPON PRODUCTION OF A BSDR in rats
population (Context 5) is taken to correlate strongly with the same feature in a human
population (Context 12). The judgment is not HIGH since it has not been demonstrated that
there is not a threshold level of BSDR necessary for production of a biophysical effect, or that
this threshold is the same between contexts.
D. Theorv-Based Inference: The judgment is HIGH since similar cellular structures exist
in both contexts; formaldehyde or its active metabolite is assumed to diffuse in a similar fashion
in both contexts; and interaction with pertinent biological structures should occur through this
diffusion. The judgment is not HIGH since it has not been demonstrated that there is not a
threshold level of BSDR necessary for production of a biophysical effect, or that this threshold is
the same between contexts.
50
-------
E. Existential Insight: The judgment is HIGH since these data produce in the analysts
moderate confidence that interaction between formaldehyde (or its active metabolite) and
pertinent biological structures is produced in both contexts.
F. Overall Summary: The judgment is LOW since Empirical Correlation, Existential
Insight and Theory-Based Inference are either MED (E.C. and T.B.I.) or HIGH (E.I.), while the
Intellectual Obligation for all of these is either MED or LOW. The presence of NO epistemic
status for the Direct Empirical warrant prevents this judgment from being either MED or
HIGH.
2.4.3.3. Host Factors
A. Direct Empirical: The judgment is LOW since the presence of target cells is observed
in both species; the presence of cellular targets for biophysical effects is observed in both
species; repair and background transitions are observed to be present in both species; but it has
not been possible to observe whether the background transitions necessary for the particular
damage caused by formaldehyde are present in both species. The judgment is not MED or
HIGH since it has not been demonstrated that there is not a threshold level of biophysical effect
necessary for production of transitions, or that this threshold is the same between contexts.
B. Semi-Empirical Extrapolation: The judgment is NO since the consideration of host
factors does not require extrapolation across exposure levels, and was dealt with in A above.
C. Empirical Correlation: The judgment is MED the presence of appropriate host factors
in rats is taken to correlate with moderate strength with the same factors in a human
population. Judgment is not HIGH since it has not been demonstrated that there is not a
threshold level of biophysical effect necessary for production of transitions, or that this threshold
is the same between contexts.
51
-------
D. Theory-Based Inference: The judgment is MED based on the fact that cellular
structure and function is similar in the two contexts since both involve mammalian species, so
similar targets should exist; the target in Context 12 is taken to be the nasal passages, which
clearly exist in both contexts; repair should be present but not completely effective in both
contexts; and background transitions are taken to occur in both contexts as evidenced by the
non-zero rates of effect in suitable controls. The judgment is not HIGH since it has not been
demonstrated that there is not a threshold level of biophysical effect necessary for production of
transitions, or that this threshold is the same between contexts. In addition, genetic variability
clearly is higher in the target context than in the observational context.
E. Existential Insight: The judgment is MED since the biological similarities between the
two contexts produce in the analysts moderate confidence that the two contexts are similar.
F. Overall Summary: The judgment is LOW since all judgments above were MED (with
the exception of the NO for Semi-Empirical Extrapolation and LOW for Direct Empirical), and
there was HIGH Intellectual Obligation for the Direct Empirical warrant.
2.4.3.4. Environmental Conditions
A. Direct Empirical: The judgment is MED since the exposure is clearly to formaldehyde
in both contexts; the form is taken to be formaldehyde vapor in both contexts; and concurrent
exposures to other substances was controlled in the observational context. The fact that
concurrent environmental exposures in the target context has not been measured and shown to
be identical to the observational context lowers this judgment from HIGH to MED.
B. Semi-Empirical Extrapolation: The judgment is NO since this warrant does not apply
for this extrapolation premise.
52
-------
C. Empirical Correlation: The judgment is LOW since concurrent environmental
exposures in a controlled experiment (Context 5) correlates poorly with concurrent
environmental exposures as defines Context 12.
D. Theorv-Based Inference: The judgment is MED based on the theory that concurrent
exposures will be different in controlled laboratory settings than in uncontrolled human settings,
but that these differences will alter the probability of formaldehyde-induced cancer but not the
presence of carcinogenic action.
E. Existential Insight: The judgment is MED since the environmental differences
between the two contexts produce in the analysts moderate confidence that the two contexts are
similar.
F. Overall Summary: The judgment is MED since most judgments above were either
MED (including Direct Empirical) or NO (Semi-Empirical Extrapolation, which was NO simply
because the same evidence was used in Direct Empirical). The LOW status of Empirical
Correlation is offset partially by the MED Intellectual Obligation of this relevance strategy.
2.5. Description of Working Table 5.5
2.5.1. Introduction
Working Table 5.5 contains judgments of the epistemic status for Inter-Context Claims of
Carcinogenicity. In each entry, the "Observational Context" is Context 5 and the "Target
Context" is Context 12. The judgments in Working Table 5.5 utilize the judgments in Working
Table 3.5 for a particular Claim of Carcinogenicity, as well as the judgments in Working Table
4.5 for the necessary Extrapolation Premises. In each case of an extrapolation, it is the "Overall
Summary" judgment associated with a Claim of Carcinogenicity in Working Table 3.5 that forms
the basis for the extrapolation. The final judgment of epistemic status in Working Table 5.5
53
-------
WORKING TABLE 5.5. INTER-CONTEXT SUPPORT FOR CLAIMS OF CARCINOGENICITY
EXTRAPOLATED FROM CONTEXT 5 TO CONTEXT 12
Claims of Carcinogenicity
Relevance Strategy
I.O.1
Increases
Incidence of
Cancer
Classification(s)
Stage
Mechanism
Complete
Partial
Mixer
Helper
Neo.
Conv.
Neo.
Devel.
Geno-
tox.
Non-
genot
ox.
Direct Empirical (D.E.)
HI
NO
NO
NO
NO
NO
NO
NO
NO
NO
Semi-Empirical
Extrapolation (S.E.E.)
ME
NO
NO
NO
NO
NO
NO
NO
ME
NO
Empirical Correlation
(E.C.)
ME
ME
NO
NO
NO
NO
ME
LO
ME
NO
Theory-based Inference
(T.B.I.)
ME
ME
NO
NO
NO
NO
ME
LO
ME
NO
Existential Insight(E.I.)
LO
ME
NO
NO
NO
NO
ME
LO
ME
NO
Column Summary
LO
NO
NO
NO
NO
ME
LO
ME
NO
Overall Summary
LO
NO
NO
NO
NO
ME
LO
ME
NO
'Intellectual Obligation
-------
cannot be higher than this initial judgment from Working Table 3.5, since the application of the
extrapolation premises can have only the effect of weakening the extrapolation. In other words,
if formaldehyde is judged to "Increase the Incidence of Cancer" with MEDIUM epistemic status
in Working Table 3.5, the highest judgment of epistemic status for the instance of extrapolation
will be MEDIUM, which presumes that the extrapolation premises are judged to each have
HIGH epistemic status in Working Table 4.5.
The analysts have chosen to view warrants from Working Table 3.5 in the following
manner. A particular Claim of Carcinogenicity from Working Table 3.5 may be used as
Empirical Correlation, Theory-Based Inference and/or Existential Insight relevance strategies in
Working Table 5.5 for the same Claim of Carcinogenicity. A particular claim from Working
Table 3.5 may NOT be used as a Direct Empirical relevance strategy in Working Table 5.5,
since there always exists extrapolation across exposure levels. A particular claim from Working
Table 3.5 MAY be used as a Semi-Empirical Extrapolation relevance strategy if it is judged that
the biological and environmental conditions in the two contexts are sufficiently similar to
warrant the claim that these two contexts contain approximately the same biological and
environmental antecedent conditions, differing only in LEVEL of exposure to formaldehyde.
This clearly requires relatively strong epistemic status to each of the four extrapolation premises
found in Working Table 4.5, a condition only weakly met in the present case.
2.5.2. Intellectual Obligation
The assignments of Intellectual Obligation discussed in Section 1.4, and utilized in
Working Table 3.5, are employed in Working Table 5.5.
2.5.3. Claims of Carcinogenicity
54
-------
The same grouping of Claims of Carcinogenicity used as headings in Working Table 3.5
are employed in Working Table 5.5. The judgments of the epistemic status of each of these
claims, for each relevance strategy, are provided in the cells of Working Table 5.5 and are
described below.
2.5.3.1. Increases Incidence of Cancer
A. Direct Empirical: The judgment is NO due to differences in level of exposure and
biological context.
B. Semi-Empirical Extrapolation: The judgment is NO due to LOW epistemic status of
necessary extrapolation premises and differences in context (which are not offset by LOW
strength of extrapolation premises).
C. Empirical Correlation: The judgment is MED since the claim was given HIGH
epistemic status in Working Table 3 but the "Overall Summaries" of extrapolation premises were
all either LOW or MED in Working Table 5.5. While a correlation does exist between
carcinogenicity in rats and humans, the LOW to MED epistemic status of the extrapolation
premises keeps this judgment from being HIGH. The fact that Genotoxicity was judged HIGH
in Working Table 3.5 also supports this judgment of MED.
D. Theory-Based Inference: The judgment is MED since the claim was given HIGH
epistemic status in Working Table 3.5 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.5. While the Theory of Carcinogenesis does
support the claim that carcinogenicity in one mammalian species is indicative of carcinogenicity
in a second, the LOW to MED epistemic status of the extrapolation premises keeps this
judgment from being HIGH. The fact that Genotoxicity was judged HIGH in Working Table 3.5
also supports this judgment of MED. Particularly troubling is the possibility that formaldehyde
55
-------
may act through neoplastic development, which may show a threshold for transition or
biophysical effect; this judgment, however, was assigned an epistemic status of LOW in Working
Table 3.5.
E. Existential Insight: The judgment is MED since the evidence produces in the analysts
moderate confidence that formaldehyde is a carcinogen in Context 12.
F. Column Summary: The judgment is LOW since Direct Empirical and Semi-Empirical
Extrapolation were NO (with HIGH Intellectual Obligation), and the remainder were either
LOW (with MED Intellectual Obligation) or MED (with LOW Intellectual Obligation).
G. Overall Summary: The judgment is LOW since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
2.5.3.2. Classifications
2.5.3.2.1. Complete Carcinogen
The judgment in all cells is NO since the same judgment is given in Working Table 3.5
(for the reasons stated there).
2.5.3.2.2. Partial Carcinogen
The judgment in all cells is NO since the same judgment is given in Working Table 3.5
(for the reasons stated there).
2.5.3.2.3. Mixer
The judgment in all cells is NO since the same judgment is given in Working Table 3.5
(for the reasons stated there).
56
-------
2.5.3.2.4. Helper
The judgment in all cells is NO since the same judgment is given in Working Table 3.5
(for the reasons stated there).
2.5.3.3. Stages
2.5.3.3.1. Neoplastic Conversion
A. Direct Empirical: The judgment is NO due to differences in level of exposure and
biological context.
B. Semi-Empirical Extrapolation: The judgment is NO due to LOW or MED epistemic
status of necessary extrapolation premises and differences in context.
C. Empirical Correlation: The judgment is MED since the claim was given MED
epistemic status in Working Table 3.5 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.5. While a correlation does exist between
neoplastic conversion in these two contexts, the MED epistemic status of the claim in Working
Table 3.5 keeps this judgment from being HIGH. The fact that Genotoxicity was judged HIGH
in Working Table 3.5 elevates the epistemic status above LOW, since genotoxic mechanisms
tend to be less context-specific than non-genotoxic mechanisms.
D. Theory-Based Inference: The judgment is MED since the claim was given MED
epistemic status in Working Table 3 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.5. While the Theory of Carcinogenesis does
support the claim that neoplastic conversion may act by common routes across contexts
containing similar DNA structure, the MED epistemic status of the claim in Working Table 3.5
and several of the extrapolation premises keeps this judgment from being HIGH. The fact that
57
-------
Genotoxicity was judged HIGH in Working Table 3.5 elevates the epistemic status above LOW,
since genotoxic mechanisms tend to be less context-specific than non-genotoxic mechanisms.
E. Existential Insight: The judgment is MED since the evidence produces in the analysts
moderate confidence that formaldehyde acts by neoplastic conversion in Context 12.
F. Column Summary: The judgment is MED since Direct Empirical and Semi-Empirical
Extrapolation were NO (with MED to HIGH Intellectual Obligation), and the remainder were
MED (with MED or LOW Intellectual Obligation). The lack of Direct Empirical and Semi-
Empirical relevance strategies could also reasonably be taken to imply that the judgment here
should be LOW.
G. Overall Summary: The judgment is MED since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
2.5.3.3.2. Neoplastic Development
A. Direct Empirical: The judgment is NO due to differences in level of exposure and
biological context.
B. Semi-Empirical Extrapolation: The judgment is NO due to LOW or MED epistemic
status of necessary extrapolation premises and differences in context.
C. Empirical Correlation: The judgment is LOW since the claim was given LOW
epistemic status in Working Table 3.5 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.5. While a weak to moderate correlation does
exist between neoplastic development in these two contexts, the LOW epistemic status of the
claim in Working Table 3.5 keeps this judgment from being MED.
58
-------
D. Theorv-Based Inference: The judgment is LOW since the claim was given LOW
epistemic status in Working Table 3.5 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.5. While the Theory of Carcinogenesis does
support the claim that neoplastic development may act by common routes across contexts
containing cellular structure and function, the LOW epistemic status of the claim in Working
Table 3.5 and several of the extrapolation premises keeps this judgment from being MED.
E. Existential Insight: The judgment is LOW since the evidence produces in the analysts
low confidence that formaldehyde acts by neoplastic development in Context 12.
F. Column Summary: The judgment is LOW since Direct Empirical and Semi-Empirical
Extrapolation were NO (with MED to HIGH Intellectual Obligation), and the remainder were
LOW (with MED or LOW Intellectual Obligation).
G. Overall Summary: The judgment is LOW since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
2.5.3.4. Mechanisms
2.5.3.4.1. Genotoxic
A. Direct Empirical: The judgment is NO due to differences in level of exposure and
biological context.
B. Semi-Empirical Extrapolation: The judgment is MED due to HIGH epistemic status
of claim in Working Table 3.5 and LOW or MED epistemic status of necessary extrapolation
premises and differences in context. The fact that a genotoxic mechanism is proposed in
Working Table 3.5 raises the epistemic status above LOW to NO since such mechanisms tend to
be less context specific and there is not believed to be a threshold for biophysical effect or
59
-------
initiating transition. The lack of complete agreement between contexts keeps this judgment from
being HIGH.
C. Empirical Correlation: The judgment is MED since the claim was given HIGH
epistemic status in Working Table 3.5 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.5. While a correlation does exist between
genotoxic mechanisms in different mammalian contexts, the LOW to MED epistemic status of
the extrapolation premises keeps this judgment from being HIGH. The fact that Genotoxicity
was judged HIGH in Working Table 3.5 elevates the epistemic status above LOW, since
genotoxic mechanisms tend to be less context-specific than non-genotoxic mechanisms.
D. Theory-Based Inference: The judgment is MED since the claim was given HIGH
epistemic status in Working Table 3.5 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.5. While the Theory of Carcinogenesis does
support the claim that genotoxic mechanisms may act by similar routes across mammalian cells,
the MED to LOW epistemic status of the extrapolation premises keeps this judgment from
being HIGH. The fact that Genotoxicity was judged HIGH in Working Table 3.5 elevates the
epistemic status above LOW, since genotoxic mechanisms tend to be less context-specific than
non-genotoxic mechanisms. The possibility of a threshold for the action of genotoxic substances
is taken to be minimal.
E. Existential Insight: The judgment is MED since the evidence produces in the analysts
moderate confidence that formaldehyde is a carcinogen in Context 12.
F. Column Summary: The judgment is MED since Direct Empirical was NO but other
cells were assigned MED. The presence of MED epistemic status to the Semi-Empirical
Extrapolation cell is taken to keep the overall judgment at MED.
60
-------
G. Overall Summary: The judgment is MED since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
2.5.3.4.2. Non-Genotoxic
The judgment in all cells is NO since the same judgment is given in Working Table 3.5
(for the reasons stated there).
3. OBSERVATIONAL CONTEXT NO. 6
3.1. Description of Working Table 1.6
Working Table 1 for Context 6 is provided below. As in all subsequent sections of this
report, a Working Table is identified by both the general Working Table number (here, 1) and
the context (here, 6). As a result, this Working Table is assigned the number 1.6. The biological
components of this context are given by the fact that all exposed organisms were mice of
relatively similar genetic, anatomical and physiological make-up. SENCAR, CD-I and B6C3F1
mice were judged to fall within this context due to broad similarities in breathing patterns and
lung histology. Differences in the background incidence of lung cancer, including nasal cancer,
are present, but were not judged to be so significant as to warrant development of three
separate contexts.
The exposures in this context were all to formaldehyde in air or (in the case of the skin
initiation assay on SENCAR mice) in liquid solution at concentrations above 2 ppm, placing
them into the category of high exposures as defined in the Introduction. Length of exposure
varied between the studies, but this was not taken to justify development of separate contexts. It
was assumed that mechanisms of action of formaldehyde might be slightly different depending
61
-------
WORKING TABLE 1.6. STUDIES AVAILABLE FOR OBSERVATIONAL CONTEXT NUMBER 6: MICE
EXPOSED TO AIRBORNE FORMALDEHYDE AT CONCENTRATIONS IN EXCESS OF 2 PPM
Study
Number1
Data Category
Description
1
TR
Primary Reference #4 in Table 1
SENCAR mice utilized in skin initiation-promotion assay. Exposures were
from 3.7% solutions of formaldehyde in acetone. 30 female SENCAR
mice in each group. TPA (promoting agent) was applied at a dose of 1.26
ug per mouse per application. Control mice were exposed to formaldehyde
at the same concentration, but with subsequent applications of acetone
(acetone showed no promotional activity). Skin papillomas were counted
by sight and reported as papillomas/mouse.
IThe study number refers to the order of the study within this context. The primary reference number, indexed to Table 1, is provided in the "Description"
column
-------
WORKING TABLE 1.6. Continued
Study
Number1
Data Category
Description
2
TR
Primaiy Reference #14 in Table 1
6 week-old B6C3F1 mice were exposed to foim aldehyde in air.
Concentrations were 0, 2, 6 and 15 ppm for 6 hrs/day, 5 days/week, for
periods of up to 24 months. There were 119-121 mice per exposure group
and for each sex. Between 10 and 30 mice were sacrificed at various
intervals during the 30 month period of the study. Necropsies were
performed and incidence of squamous cell carcinoma of the nasal
passages deteimined.
IThe study number refers to the order of the study within this context. The primary reference number, indexed to Table 1, is provided in the "Description"
column
-------
WORKING TABLE 1.6. Continued
Study
Number1
Data Category
Description
3
BE
Primary Reference #10 in Table 1
CD-I mice were utilized in measurements of sister-chromatid exchange
following exposure to airborne form aldehyde. 5 male and 5 female mice
in each exposure category were exposed to approximately 0, 6 or 12 ppm
of formaldehyde vapor 6 hours/day for 5 days, or 25 ppm 6 hours/day for
4 days. Cells from bone marrow were collected following sacrifice and
analyzed visually for sister chromatid exchange. 50 cells per mouse were
scored.
IThe study number refers to the order of the study within this context. The primary reference number, indexed to Table 1, is provided in the "Description"
column
-------
upon length of exposure, but that this difference would appear in the magnitude of effect rather
than the presence and direction of that effect. This choice is somewhat tenuous since there is
evidence that effects such as promotion may require some minimal period of exposure.
Three bodies of data were identified for this context. These are described in Working
Table 1.6., and are divided into the data categories of:
(a) Tumor Response (taken from Reference #4 and Reference #14 in Table 1 of this
report). Here, the Tumor Response in Reference #4 is taken to be skin papillomas, which
clearly is a form of cancer. It is given the label TR1. The Tumor Response in Reference #14 is
taken to be incidence of nasal tumors, which clearly is a form of direct measurement of tumors.
It is given the label TR2.
(b) Biophysical Effect (taken from Reference #10 in Table 1). Here, the sole Biophysical
Effect is taken to be sister chromatid exchange, which is assigned to this data category because
it is an effect leading to, but not identical with, cancer. It is given the label BE1.
Data in other data categories were not available for the studies utilized in this context.
3.2. Description of Working Table 2.6
The judgments for this Working Table are provided in the
cells of Working Table 2.6. The text here describes the reasons for these judgments. The data on
which these judgments are based are provided as various "Exhibits" after Working Table 2.6.,
with a numbering system described in Section 3.2.4.
3.2.1. Data Categories and Items
The same data categories and items described with respect to Working Table 1.6. are
employed in Working Table 2.6. These are:
62
-------
WORKING TABLE 2.6. DATA CHARACTERISTICS FOR OBSERVATIONAL CONTEXT NUMBER 6
Data Category/Item
Description
Completeness
(Hi/Me/Lo/No)
Utility
(Hi/Me/Lo/No)
Strength of Effect
(Hi/Me/Lo/No)
Exposure-
Specific
Effect1
Tumor Response
Skin initiation CTR1.1)
Nasal tumors CTR2.1)
LO
HI
ME
ME
NO
LO
WT2.C6.
TR1.1
WT2.C6.
TR2.1
Biophysical Effect
Sister chromatid exchange
(BE1.1)
HI
HI
HI
WT2.C6.
BE1.1
Pharm acodynam ics
No study available
Host Factors
No study available
Concurrent Environmental
Conditions
No study available
Related Substances Assessment
No study available
'Refers to the Exhibit Number for the data as described in the text.
-------
WT2.C6.TR1.1
Grp2 (F.T):
ufp I If, I 1. 00000
Grp 4A1 (D,T):»—•
Grp 4B (F,A): no pap.
I onOOPftA
10
20
30
40
50
WEEKS AFTER INITIATION
Fig. 2. Incidence of skin papillomas in SENCAR mice followi
skin application of formaldehyde and TPA, DMBA and TPA, or
formaldehyde and acetone. DMBA in combination with TPA was
much more effective than other treatments 1n inducing
papi1lomas.
-------
WT2.C6.TR2.1
Table 1 Formaldehyde
Inhalation for 24 Months*:
Incidence of Squamous^
Cell Carcinoma of Nasal
Cavity
Number of tumors/naul cavitiai examined . . ..
Formaldehyde
— — : ¦"
concentration (ppm)T
Rat
Mousa - ~
0
0/232 (0%)
.0/223(0%).
2
0/236 (0%)
0/214 (0%)
6
2/235 (0.9%)
0/218 (0%)
• 15
103/232 (44.4%)
2/225 (0.9%)
*6 h/day, 5 days/wk. The study was initiated with 240 Fischer-344 rati and 240 B6C3F1
mice, evenly divided by sex.
^Target concentrations. Actual average concentrations were 0, 2.0, 5.6, and 14.3 ppm. ;
-------
WT2.C6.BE1.1
Table 6 Sister Chromatid Exchange (SCE) Results: Exposure of CD-1 Male and Female Mice to Three Levels of Formaldehyde Vapor
Exposure
Male animal No.
No. of
cells scored
No. of SCE
SCE/Cell
Female animal No.
No. of
cells scored
No. of SCE
SCE/Cell
High dose*
1090-1094
60
269
4.48
1095-1099
93
1056
11.35 S
Mid dose1
0873-0877
100
507
5.07
0898-0902
100
615
6.15 3
Low dose1
0868-0872
98
313
3.19
0893-0897
93
300
3.23
Positive control
(Cyclophosphamide)
0883-0887
85
1298
15.27 §
0908-0912
92
1024
1 1 .1 3 §
Negative control I
1080-1084
100
330
3.30
1085-1089
60
210
3.50
0863-0867
100
549
5.49
0888-0892
100
307
3.07
Combined
200
879
4.40
Combined
160
517
3.23
*Animals exposed 6 h/day X 4 days.
^Animals exposed 6 h/day X 5 days.
J Animals exposed 6 h/day X 5 days and 6 h/day X 4 days.
^Significantly higher than the negative controls.
-------
(a) Tumor Response, containing TR1 (skin papillomas) and TR2 (incidence of nasal
tumors);
(b) Biophysical Effects, containing BE1 (sister chromatid exchange).
These two separate data categories and three separate data items are displayed in
Working Table 2.6.
322. Judgments of Completeness
In all cases, the judgment of completeness was based on a comparison of the number of
studies used in a data category/item for this context against the number available in the
complete reference list provided in Appendix A. The judgment was HIGH if the data used in
this Working Table was a relatively complete subset of the data indicated in the Appendix,
where "complete" was taken to mean somewhere on the order of 70% or above. The judgment
was MEDIUM if the data used in this Working Table constituted somewhere between 25 and
70% of the total body of potentially available data. The judgment was LOW if the data used
constituted less than 25% of the data available. These assignments also included limited
consideration of the QUALITY of the data, since it should not necessarily count against an
analysis if the analyst fails to collect data of poor quality. This consideration also is dealt with in
the assignment of Utility for the collected studies. Still, it is evident that the judgment of
Completeness might reasonably be expected to include consideration of something more than
the fraction of studies collected in a given data category/item. In the present analysis,
Completeness was increased slightly if the collected data for a given data category were the most
heavily cited in the secondary literature on formaldehyde. This leads to the following judgments
of Completeness:
63
-------
For the data category TR1, the judgment for Completeness is LOW since this is only 1
item (Study 1) of 4 items identified in the full literature search. The other 3 studies are Delia
Porta et al., (1968); Federal Panel on Formaldehyde (1982); and Krivanek, et al., (1982).
For the data category TR2, the judgment for Completeness is HIGH since this study is
the only study identified in the literature search.
For the data category BE1, the judgment for Completeness is HIGH since this is the
only in vivo study of sister chromatid exchange identified in the full literature search. It should
be noted, however, that SCE is only 1 of at least 10 cellular biophysical effects measured in
different studies, which might reasonably be taken (but has not been taken here) to weaken the
judgment of Completeness.
3.2.3. Judgments of Utility, Strength of Effect, and Exposure-Specific Effect
The remaining judgments in Working Table 2.6 refer to the qualities of the data
ACTUALLY EXAMINED IN THIS EXAMPLE. By Utility, we mean the quality of the study
DESIGN and its ability to address in rigorous fashion questions pertinent to hazard
identification for formaldehyde. No attempt is made to determine the specific inferences
towards which the data will be directed in later Working Tables, since the intent here is only to
determine the degree to which a body of data arose from a study design appropriate for
accurate and precise measurements of the quantity considered in the study. Utility is increased
by high standards of study design, including the ability of the experiment to demonstrate a
causal link between exposure to formaldehyde and the measured effect.
By Strength of Effect, we mean the degree to which the study actually revealed an effect.
Strength is increased when the effect noted possessed statistical significance.
64
-------
By Exposure-Specific Effect, we mean a summary of the actual effect observed (putting
aside questions of the reliability of this effect, which was dealt with in the assignment of Utility
and Strength of Effect). This summarization is provided in the following text, and the data
themselves are provided in the various Exhibits appearing after Working Table 2.6.
The following judgments then apply to Working Table 2.6:
For the data category TR1, the Utility judgment is MEDIUM since only 1 exposure level
of formaldehyde was used to examine initiation, and this is equivalent to very large doses in
vivo; a marginally adequate number of subjects (30) were employed; appropriate controls were
utilized; the study constitutes a deliberate experiment (important for causal claims); and the
effect is a direct indicator of the ability of formaldehyde to induce transitions between states of
cancer when strong promoters are present. The Strength of Effect judgment is NO for the issue
of formaldehyde as a complete carcinogen, since no papillomas were observed when only
formaldehyde was applied. Judgment is LOW for formaldehyde as an initiator since an increase
in papillomas was observed when TPA was used as a promoting agent, but the increase was not
large. No statistical measure of significance was provided. The Exposure-Specific Effect is
summarized in Exhibit WT2.C6.TR1.1. No papillomas were observed in applications of
formaldehyde in acetone at a concentration of 3.7%. A few papillomas were observed after
application of 3.7% formaldehyde followed by repeated applications of TPA as 1.25 ug
TPA/application/mouse.
For the data category TR2, the Utility judgment is MEDIUM since the study utilized an
appropriate experimental methodology; formaldehyde exposure was well characterized;
approximately 120 male and 120 female mice were exposed at each exposure level; identification
of tumors was by an appropriate method; the study constituted a deliberate experiment
(important for causal claims); exposures were long-term (18-24 months); appropriate controls
65
-------
were utilized; but the number of mice in each exposure category was not sufficient to detect
potential low excess incidence of tumors. The Strength of Effect judgment is LOW for all
exposure categories. No statistically significant increase in nasal tumor incidence was observed in
any exposure category. The opinion of the original investigators was that this study could not be
used to infer any effect on nasal carcinomas in the exposed mice. The Exposure-Specific Effect
is summarized in Exhibit WT2.C6.TR2.1. Incidence of squamous cell carcinoma of the nasal
cavity was 0% at 0, 2 and 6 ppm. Incidence was 0.9% at 15 ppm, which arose from carcinomas
in 2 mice of a total of 225 mice.
For the data category BE1, the Utility judgment is HIGH since multiple levels of
exposure were examined; scoring techniques for cells were appropriate; an adequate number of
animals were examined (10 mice at each exposure level); the exposures were at levels defining
the context; an adequate number of cells were scored; appropriate negative and positive controls
were included; the study constitutes a deliberate experiment (important for causal claims); and
the effect (SCE) demonstrates the ability of formaldehyde to interact with, and cause changes
in, DNA in vivo. The Strength of Effect judgment is HIGH for mid-and-high exposure groups
since SCE was elevated with statistical significance (p<0.05). Judgment is LOW for the low
exposure group since no statistically significant increase in SCE was noted. The Exposure-
Specific Effect is summarized in Exhibit WT2.C6.BE1.1. There is a dose-related increase in
SCE, scored as both total SCE and SCE/mouse. The slope, however is positive only beyond the
lowest non-control exposure group.
32.4. Exhibits for Working Table 26
The data for the various data categories are provided as "Exhibits" after Working Table
2.6. The numbering system for the exhibits was chosen to display information on (in order of
66
-------
appearance) the Working Table; the Context; the specific effect measured; and the study
number within that specific effect. For example, Exhibit WT2.C6.TR1.1 refers to Working Table
2 (all exhibits have this assignment in common); Context 6; Tumor Response of type 1; and data
set #1 within the data category TR1. This same numbering scheme will be employed in all other
Contexts.
33. Description of Working Table 3.6
33.1. Introduction
The central task of Working Table 3.6 is to draw inferences on the carcinogenicity of
formaldehyde WITHIN CONTEXT 6, employing data specific to that context. This task requires
that several judgments be made sequentially:
(a) The analyst must determine the particular Claims of Carcinogenicity to be considered
in the analysis. These various claims are discussed in detail in Section 3.3.4.
(b) The available Relevance Strategies by which a given body of data may be related to a
specific Claim of Carcinogenicity must be developed. These strategies were discussed in
detail in Section 1.4., and the same definitions are applied here.
(c) The analyst must assign a measure of Intellectual Obligation to each Relevance
Strategy. This assignment was described in detail in Section 1.4., and the same
assignments are made here.
(d) The analyst must judge the epistemic status of a specific Claim of Carcinogenicity
arising from a specific Relevance Strategy. As described in Section 1.4., this judgment is
separate from the judgment of Intellectual Obligation for that Relevance Strategy. This
judgment is described in more detail in Section 3.3.2. The assignments are entered into
the appropriate cells of Working Table 3.6.
(e) The analyst must produce a summary judgment of the epistemic status of a particular
Claim of Carcinogenicity by combining the judgments from each of the five Relevance
Strategies for that claim. This summary includes consideration of both the epistemic
status of that claim for each Relevance Strategy, as entered into the separate cells of the
Working Table specific to each strategy, as well as the Intellectual Obligation assigned to
that strategy. The assignments are entered into the appropriate COLUMN SUMMARY
cells of Working Table 3.6.
67
-------
WORKING TABLE 3.6 INTRA-CONTEXT SUPPORT FOR CLAIMS OF CARCINOGENICITY IN CONTEXT 6
Claims of Carcinogenicity
Relevance Strategy
I.O.1
Increases
Incidence of
Cancer
Classification(s)
Stage
Mechanism
Complete
Partial
Mixer
Helper
Neo.
Conv.
Neo.
Devel.
Geno-
tox.
Non-
genot
ox.
Direct Empirical (D.E.)
HI
LO
NO
LO
NO
NO
LO
NO
HI
NO
Semi-Empirical
Extrapolation (S.E.E.)
ME
NO
NO
NO
NO
NO
NO
NO
NO
NO
Empirical Correlation
(E.C.)
ME
ME
NO
LO
NO
NO
ME
NO
HI
NO
Theory-based Inference
(T.B.I.)
ME
ME
LO
LO
NO
NO
ME
NO
HI
NO
Existential lnsight(E.I.)
LO
LO
LO
LO
NO
NO
ME
NO
HI
NO
Column Summary
LO
LO
LO
NO
NO
ME
NO
HI
NO
Overall Summary
LO
LO
LO
NO
NO
ME
NO
HI
NO
'Intellectual Obligation
-------
(f) The analyst must produce a summary judgment of the COHERENCE of the various
Claims of Carcinogenicity. This task is performed by examining the coherence ACROSS
the various COLUMN SUMMARIES appearing in the Working Table. The assignments
are entered into the appropriate OVERALL SUMMARY cells of Working Table 3.6.
33.2. Epistemic Status of Chums of Carcinogenicity
The entries into the various cells of Working Table 3.6. reflect the judgment of the
analyst concerning the epistemic status of the Claim of Carcinogenicity being considered, FOR
A SPECIFIC RELEVANCE STRATEGY. This judgment will have two components:
(a) Each Relevance Strategy calls upon a selected body of data from which the reasoning
will proceed. Specifically, the Direct Empirical strategy requires data on direct measurements of
the effect underlying the Claim of Carcinogenicity, made at the exposure to formaldehyde
defining the context (here, concentrations above 2 ppm). The Semi-Empirical Extrapolation
strategy requires data on direct measurements of the effect underlying the Claim of
Carcinogenicity, made at exposures to formaldehyde other than those defining the context. The
Empirical Correlation strategy requires data on effects other than the effect underlying the
Claim of Carcinogenicity, where this "other" effect is taken to correlate with appearance of the
effect of interest. The Theory-Based Inference strategy requires data on an effect other than the
effect underlying the Claim of Carcinogenicity, where this "other" effect is taken to be a causal
step in the production of the effect of interest as specified by an established theory. The
Existential Insight strategy does not invoke data, except in the sense that a subjective judgment
from an individual might be counted as an instance of data in and of itself.
The analyst must assign the "foundational quality" to the data used in each relevance
strategy. By "foundational quality", we mean the quality of the data with respect to the question
those data were intended to address BY THE ORIGINAL INVESTIGATOR. This judgment is
based on the results of Working Table 2.6 and includes consideration of the Completeness,
68
-------
Utility, Strength of Effect and Exposure-Specific Effect from that table. While the judgment
ultimately is subjective, it is constrained in that it must increase (from NO to LOW to
MEDIUM to HIGH) as the strength of the judgments for Completeness, Utility and Strength of
Effect increase in Working Table 2.6. The summary of Exposure-Specific Effect may also enter
into the assignment of epistemic status to a Claim of Carcinogenicity if the analyst determines
that only effects above a given magnitude will exert an influence on carcinogenic processes (an
issue of importance when possible thresholds are considered).
(b) The analyst then must judge the strength of any "background premises" required by
the Relevance Strategy under consideration. These background premises are defined in detail in
the original report, but may be summarized as the set of (often implicit) premises that must be
introduced into an analysis if the data in Working Table 2.6 are to be used to draw inferences
about the Claims of Carcinogenicity for the context being considered. Presumably, many of these
background premises already are considered valid once Working Tables 1.6 and 2.6 are
complete; otherwise, the data would not have been assigned to this context. But other
background premises concerning the reliability of extrapolation procedures (for the Semi-
Empirical Extrapolation relevance strategy), the strength and specificity of correlations (for the
Empirical Correlation relevance strategy), the degree of verification and validation of theories
(for the Theory-Based Inference relevance strategy), and the credentials of individuals making
subjective judgments (for the Existential Insight relevance strategy) must be warranted prior to
their use.
Having established these two components of epistemic status (foundational quality and
background premises) for a given Claim of Carcinogenicity, the analyst then produces a
composite judgment of the epistemic status of a given Claim of Carcinogenicity and enters this
into the appropriate cell of Working Table 3.6. There are two separate claims that might be
69
-------
made within one of these cells. The first is the claim that formaldehyde DOES produce the
effect underlying that claim (such as increased incidence of cancer or genotoxic effects). In that
case, the assignments of NO/LOW/MED/HIGH refer to the epistemic status of the claim that
formaldehyde produces the effect. This assignment is entered into the area ABOVE the dashed
line in a cell. The second claim is that formaldehyde DOES NOT produce the effect underlying
the claim. In that case, the assignment of NO/LOW/MED/HIGH refer to the epistemic status
of the claim that formaldehyde DOES NOT produce the effect. This assignment is entered into
the area BELOW the dashed line in a cell. THE READER SHOULD NOTE THAT THIS IS A
NEW FEATURE OF THESE WORKING TABLES NOT DISCUSSED IN THE ORIGINAL
REPORT. The reason for this addition to the Working Tables is that the fact that a particular
body of data does not support the claim that formaldehyde does induce a specific effect IS NOT
NECESSARILY evidence that formaldehyde does not induce that effect. It may simply be the
case that the data were obtained under conditions incapable of revealing any effect that might
be present.
This consideration is important due to the issue of "burden of proof'. At times, the
burden may lie on the regulatory agency to show that formaldehyde DOES induce cancer.
Attention then would be directed towards the judgments in the upper halves of the various cells
of Working Table 3.6. At other times, the burden may lie on the producer of formaldehyde to
show that formaldehyde DOES NOT induce cancer. Attention then would be directed towards
the judgments in the lower halves of the various cells of Working Table 3.6. IN THE PRESENT
EXAMPLE, ONLY JUDGMENTS THAT AN EFFECT IS PRODUCED BY
FORMALDEHYDE ARE DEVELOPED. THE PROCESS OF ANALYSIS WOULD BE
REPEATED FOR THE LOWER HALVES OF ALL CELLS IF THE BURDEN OF PROOF
WAS SHIFTED.
70
-------
It is important here to consider the broad features of the theory of carcinogenesis to be
used throughout this section. These features structure the interpretation of data used in the
various Relevance Strategies. The theory of carcinogenesis employed here is that cells are
transformed to cancer by three steps or transitions. These are termed initiation, promotion and
progression, each of which may require one or more sub-steps. For each transition to occur, a
biologically-significant dose-rate must reach the targets in appropriate cells so that an interaction
between the substance or a metabolite may take place. This target may or may not have been
identified for a particular substance and tumor type.
The initiation step is taken to involve alteration of DNA structure and function,
producing a new genome. This alteration must then remain intact without being repaired by
processes normally operating in cells or induced by the damage itself. The ability of a substance
to bring a stable alteration about may be determined from an increase in cancer incidence
following (1) application of the initiating agent and (2) subsequent application of a promoting
agent such as TPA. In the absence of such data, the ability of a substance to interact with DNA,
to alter DNA, and/or to induce mutation provides partial but not conclusive evidence of
initiating activity. In addition, the ability of a substance to lower DNA repair rates or fidelity of
repair is partial but not conclusive evidence that the substance will enhance the initiating activity
of other substances or of background events.
The promotion step is taken to involve loss of growth control, represented by an
imbalance between rates of growth and death of cells in a colony and a continued net rate of
expansion even upon reaching confluence. The mechanism for this step is unspecified, but may
result in either a lowering of the rate of removal of cells (as in decreased apoptosis) or an
increase in the rate of division (as in abnormal proliferation). The ability of a substance to bring
this about may be determined from an increase in transformation following (1) prior application
71
-------
of a known initiating agent and (2) subsequent application of the substance of interest. In the
absence of such data, the ability of a substance to alter intercellular communication, the rate of
cellular division, the rate of apoptosis, and/or net rate of growth provides partial but not
conclusive evidence of promoting activity.
The progression step is taken to be poorly understood at present, but probably is related
to the metastatic properties of the promoted cells. There is no assay for progression at present.
Partial but inconclusive evidence of progressing activity might be obtained from a demonstration
that cells have taken on the ability to dislodge from the primary tumor, spread to other organs
or tissues, reattach, and grow to fatal tumors.
If a substance induces only one of the required transitions, it will be necessary for other
transitions to be induced by antecedent conditions if the substance of interest is to be
considered a Partial carcinogen (as defined in Section 3.3.4). The fact that the background
incidence of cancer is not zero may be taken to indicate that each of these other transitions will
always be present. If this assumption is adopted, the ability of a substance to increase the
incidence of cancer cannot be used to differentiate between claims of a Complete or a Partial
carcinogen (as defined in Section 3.3.4). Existing empirical evidence, however, is suggestive of
there being several routes by which cancer may occur. It is not inconsistent to assume that a
substance acts to induce only a subset of transitions, and that the remaining transitions required
are not those associated with the mechanism by which the antecedent conditions of the
experimental context act to bring about background cancer.
The epistemic status of this Theory of Carcinogenesis is taken here to be MED based on
an examination of the available empirical evidence linking DNA alteration, alteration of rates of
growth, and the carcinogenic potential of a range of substances tested in assays for initiation,
promotion and complete cancer. A judgment of HIGH is not considered appropriate since much
72
-------
of the warrant for the theory is subjective judgment of experts in the field based on the
coherence of the explanatory framework and the finding of (unexplained) correlations.
3 J J. Intellectual Obligation
The issue of Intellectual Obligation was discussed in detail in Section 1.4. Since the
judgments and assignments are common to all Working Tables and Contexts, this discussion is
not repeated here. By way of a summary, the Direct Empirical Relevance Strategy was assigned
an Intellectual Obligation of HIGH; the Semi-Empirical Extrapolation, Theory-Based Inference
and Empirical Correlation Relevance Strategies were assigned an Intellectual Obligation of
MEDIUM; and the Existential Insight Relevance Strategy was assigned an Intellectual
Obligation of LOW. The assignments are entered into the appropriate cells of Working Table
3.6.
3.3.4. Claims of Carcinogenicity
The framework employed here allows for a wide variety of claims concerning the
carcinogenic action of formaldehyde. In the original report, it was determined that it is desirable
to augment the claim that a substance increases the incidence of cancer, since this sole claim
does not provide full detail on the antecedent conditions under which a substance such as
formaldehyde can exert its carcinogenic potential. It was determined that a more detailed
analysis would subdivide the claim of increased incidence of cancer into more detailed claims
potentially of use in determining the most effective regulatory strategy. The different Claims of
Carcinogenicity employed in Working Table 3.6 (and all subsequent Working Tables discussed in
this report) are:
73
-------
(a) Increases Incidence of Cancer, implying that formaldehyde raises the incidence of
cancer, raises the multiplicity of cancer at sites, and/or changes the age at which cancer
appears.
(b) Complete Carcinogen, implying that formaldehyde increases the incidence of cancer
even when it is not delivered in conjunction with concurrent exposures.
(c) Partial Carcinogen, implying that formaldehyde increases the incidence of cancer only
when delivered in conjunction with concurrent exposures that complete the process of
carcinogenesis.
(d) Mixer, implying that formaldehyde exerts its carcinogenic effect only when it is
combined with another substance, and where this combination results in a new substance
which is at least a partial carcinogen.
(e) Helper, implying that formaldehyde does not in and of itself induce transitions to
cancer, but that it facilitates the carcinogenic action of another substance.
(f) Neoplastic Conversion, implying that formaldehyde acts by converting cells to
neoplastic potential.
(g) Neoplastic Development, implying that formaldehyde acts by producing growth in
colonies of cells already having neoplastic potential.
(h) Genotoxicity, implying that formaldehyde acts through a mechanism involving
interaction with the genetic material of a cell.
(i) Non-genotoxicity, implying that formaldehyde acts through a mechanism other than
interaction with the genetic material of a cell.
The following summary considerations from Working Table 2.6. provide the foundations
for all inferences to Claims of Carcinogenicity drawn in this context:
TR1.1 demonstrates that in vivo exposure of mouse skin to 3.7% formaldehyde in
acetone produces a slight increase in the incidence of skin papillomas when followed by the
promoting agent TPA. This increase is not, however, dramatic and there is no measure of
statistical significance provided. The utility of the study was assigned as MED, and the
completeness was assigned as LOW. In addition, no promoting agent other than TPA was
considered, raising the possibility that another promoting agent might reveal a greater initiating
activity for formaldehyde by inducing promoting effects more relevant to the initiating activity of
74
-------
formaldehyde. The claim that formaldehyde alone DOES NOT increase skin papilloma
incidence can also be examined by this study, and this claim is judged to have an epistemic
status of MED since (1) there clearly was no increase in skin papillomas when only
formaldehyde was applied and (2) the completeness of the TR1 data was judged LOW. All
factors considered, the analysts judge that the epistemic status of the claim that formaldehyde
induces skin papillomas when followed by a promoting agent is LOW. It should also be noted
that this study cannot be used as a Direct Empirical warrant in this context since the context is
defined as airborne exposures, not skin applications.
TR2.1 demonstrates that in vivo exposure of mice to formaldehyde in air produces a
non-significant increase in nasal tumors at 15 ppm. It is not judged that the 0.9% incidence in
the 15 ppm exposure group is statistically significant. While completeness was HIGH, the utility
and strength of effect were MED and LOW, respectively. The epistemic status of the claim that
formaldehyde induces nasal tumors in mice at 15 ppm, therefore, is judged to be LOW. The
epistemic status of the claim that formaldehyde induces nasal tumors in mice at 2 or 6 ppm is
judged to be NO since the study population was not sufficiently large to locate any small
increases in incidence that might exist. By contrast, other analysts (such as Dr. Jim Swenberg of
UNC-CH) have judged all claims to be NO based on the lack of statistical significance in the 15
ppm group.
BE1.1 demonstrates that in vivo exposure of mice to airborne formaldehyde induces
statistically-significant increases in sister chromatid exchange within cells. The epistemic status of
this claim is judged HIGH since completeness is HIGH, utility is HIGH and strength of effect is
HIGH. In addition, the analysts judge that the same study supports the claim that exposure to
airborne formaldehyde produces a biologically-significant dose-rate in the cells of mice within
the current context, at least if it is assumed that production of DNA alterations requires the
75
-------
production of such a dose-rate. This claim is judged to have an epistemic status of HIGH due to
(1) the reasons given above for the first claim and (2) the fact that the theory of DNA damage
requiring a biologically-significant dose-rate is judged to have an epistemic status of somewhere
between MED and HIGH.
The separate judgments for Claims of Carcinogenicity for Working Table 3.6. are
provided in the sections that follow. In the case of each claim, explanations are given of the
judgment for each Relevance Strategy; then for the Column Summary across Relevance
Strategies for that claim; then for the Overall Summary across claims.
3.3.4.1. Increases Incidence of Cancer
The judgments for this Claim of Carcinogenicity for the various Relevance Strategies and
for this context are shown in the cells of Working Table 3.6. The basis for these judgments are
given below.
A. Direct Empirical: The judgment is LOW/NO since cancer incidence was increased
slightly in the 15 ppm group but the epistemic status was LOW. The skin initiation assay
indicated weak initiating activity. This increase, however, was following an exposure other than
the one that defines this context.
B. Semi-Empirical Extrapolation: The judgment is NO since exposures were in the range
defining this context.
C. Empirical Correlation: The judgment is MED since initiation is correlated strongly
with carcinogenicity, but the epistemic status of the initiation assay performed was LOW and the
context was incorrect.
D. Theorv-Based Inference: The judgment is MED since the initiation assay does
suggest, within the Theory of Carcinogenesis employed here, that formaldehyde might be acting
76
-------
as an initiating agent, but the finding of LOW epistemic status for the claim that formaldehyde
is an initiating agent weakens this position. The finding of sister chromatid exchange (HIGH
epistemic status) supports the assumption that a biologically significant dose-rate occurs in
pertinent tissue, and also provides a partial explanation of the initiation result.
E. Existential Insight: The judgment is LOW since these data produce in the analysts
only very weak confidence in formaldehyde's carcinogenic ability. It is judged further that the
analysts possess the relevant experience, training and
psychological skills necessary to form such subjective judgments, based on having worked in the
science of carcinogenesis for several years.
F. Column Summary: The judgment is LOW since the Direct Empirical relevance
strategy was LOW (with HIGH intellectual obligation) and other relevance strategies did not
exceed MED.
G. Overall Summary: The judgment is LOW since the analysts' strongly foundational
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim. If, however, such a stance was not taken and coherence were insisted
upon, the judgment still would be LOW because the assignment of LOW in the Direct Empirical
cell could not be outweighed by any combination of scores in all other columns.
3.3.4.2. Classifications
3.3.4.2.1. Complete Carcinogen
A. Direct Empirical: The judgment is NO since cancer incidence was increased slightly in
the 15 ppm group but the epistemic status was LOW, and since the incidence of skin papillomas
was not changed by application of formaldehyde alone. It must also be noted that the initiation
assay was performed using dermal exposure rather than inhalation (the route considered under
77
-------
this context), and there is no support available for the assumption that an active metabolite is
present in both exposure routes. Finally, a positive finding of increased cancer incidence would
still not confirm that formaldehyde is a complete carcinogen since it is also reasonable to
assume that it is a partial carcinogen, with the other transitions being completed by background
events.
B. Semi-Empirical Extrapolation: The judgment is NO since exposures were in the range
defining this context.
C. Empirical Correlation: The judgment is NO since the study effect "cancer incidence" is
identical with the effect to be correlated to, "cancer incidence."
D. Theory-Based Inference: The judgment is LOW since the studies provide only limited
information on any mechanisms of action from which an explanation might be developed other
than that DNA alterations are produced and a strong promoting agent is required for
production of skin papillomas. These last results do suggest, within the Theory of Carcinogenesis
employed here, that formaldehyde might be acting as an initiating agent, but the finding of
LOW epistemic status for the claim that formaldehyde is an initiating agent weakens this
position. In addition, the studies do not provide data relevant to the assumption that the process
of metabolic activation is either present, or similar, in the dermal and inhalation routes of
exposure. A more likely explanation seems to be that formaldehyde is a promoting agent, but
this was not tested in the available studies. It is not the case, therefore, that the data support the
claim that any step in the process of transformation is understood to be the mechanism by which
formaldehyde acts to induce transitions in the mice exposed to airborne formaldehyde. The fact
that formaldehyde needed promotion in the skin assay suggests that if formaldehyde is a
complete carcinogen, it is a weak one.
78
-------
E. Existential Insight: The judgment is LOW since these data produce in the analysts
only very weak confidence of formaldehyde's complete carcinogenic ability. It is judged further
that the analysts possess the relevant experience, training and psychological skills necessary to
form such subjective judgments, based on having worked in the science of carcinogenesis for
several years.
F. Column Summary: The judgment is LOW since all relevance strategies were either
LOW or NO.
G. Overall Summary: The judgment is LOW since the analysts' strongly foundational
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim. If, however, such a stance was not taken and coherence were insisted
upon, the judgment still would be LOW because the assignment of NO in the Direct Empirical
cell could not be outweighed by the inconclusive findings in other columns.
3.3.4.2.2. Partial Carcinogen
A. Direct Empirical: The judgment is LOW since cancer incidence was increased slightly
in the 15 ppm group but the epistemic status was LOW. The skin assay indicates that
formaldehyde does not produce all transitions needed for cancer, suggesting a Partial carcinogen
at most. This study is not in the proper context, however, due to route of exposure.
B. Semi-Empirical Extrapolation: The judgment is NO since exposures were in the range
defining this context.
C. Empirical Correlation: The judgment is LOW since the finding of formaldehyde being
a partial carcinogen in the skin assay has an epistemic status of LOW, and is not outweighed by
the strength and specificity of this correlation in general.
79
-------
D. Theory-Based Inference: The judgment is LOW since the studies do not provide
information on any mechanisms of action from which an explanation might be developed, other
than that DNA alterations are produced and that formaldehyde might be a weak initiating agent
during dermal exposures. This latter result does suggest, within the Theory of Carcinogenesis
employed here, that formaldehyde might be acting as an initiating agent (and therefore a partial
carcinogen), but the finding of LOW epistemic status for the claim that formaldehyde is an
initiating agent weakens this position. In addition, the lack of evidence that active metabolites
are present in both exposure routes weakens any claims concerning inhalation of formaldehyde
(the context of interest here). It is not the case, therefore, that the data support the claim that
any step in the process of transformation is understood to be the mechanism by which
formaldehyde acts to induce transitions in the mice.
E. Existential Insight: The judgment is LOW since these data produce in the analysts
only very weak confidence of formaldehyde's carcinogenic ability. It is judged further that the
analysts possess the relevant experience, training and psychological skills necessary to form such
subjective judgments, based on having worked in the science of carcinogenesis for several years.
F. Column Summary: The judgment is LOW since all relevance strategies were either
LOW or NO.
G. Overall Summary: The judgment is LOW since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim. If, however, such a stance was not taken, insisting upon coherence, the
judgment still would be LOW because the assignment of LOW in the Direct Empirical cell could
not be outweighed by the inconclusive findings in other columns.
3.3.4.2.3. Mixer
80
-------
The judgment in all cells is NO since all assays involved only formaldehyde or application
of formaldehyde with a promoting agent. Such studies can address only the issue of whether a
substance is a Partial or Complete carcinogen.
3.3.4.2.4. Helper
The judgment in all cells is NO since all assays involved only formaldehyde or application
of formaldehyde with a promoting agent. Such studies can address only the issue of whether a
substance is a Partial or Complete carcinogen.
3.3.4.3. Stages
3.3.4.3.1. Neoplastic Conversion
A. Direct Empirical: The judgment is LOW since the skin assay provides LOW support
for the claim that formaldehyde has initiating activity, which is taken to be identical with
neoplastic conversion. In addition, the lack of identification of the active metabolite by either
route of exposure weakens the degree to which it may be assumed that findings from the dermal
route are reliable indicators of effects by inhalation. The measurement of sister chromatid
exchange is pertinent only within the Theory-Based Inference relevance strategy, since it does
not constitute a case of direct observation of neoplastic conversion (but rather a change that
might lead to such a conversion).
B. Semi-Empirical Extrapolation: The judgment is NO since exposures were in the range
defining this context.
C. Empirical Correlation: The judgment is MED since the production of sister chromatid
exchange (characterized by an epistemic status of HIGH) correlates (moderate strength and
specificity) with the ability of a substance to induce neoplastic conversion.
81
-------
D. Theorv-Based Inference: The judgment is MED since the study of sister chromatid
exchange provides clear evidence (epistemic status is HIGH) that DNA damage is taking place
following in vivo exposure to formaldehyde. It is not the case, however, that sister-chromatid
exchange has been clearly implicated as a causal step in the transition to cancer, or even in the
transition to neoplastic conversion. In addition, the skin assay provides only weak evidence that
formaldehyde might be inducing neoplastic conversion, and requires the untested assumption
that the same biologically-significant form of the dose-rate will be present from inhalation. The
SCE data do provide support for the claim that a biologically-significant dose-rate is reaching
the cells of the nasal cavity during inhalation, which are taken here to be the target for the nasal
tumors examined in the cancer assay.
E. Existential Insight: The judgment is MED since these data produce in the analysts
some confidence in formaldehyde's ability to induce neoplastic conversion. It is judged further
that the analysts possess the relevant experience, training and psychological skills necessary to
form such subjective judgments, based on having worked in the science of carcinogenesis for
several years.
F. Column Summary: The judgment is MED since, while the Direct Empirical strategy
with an Intellectual Obligation of HIGH was scored LOW, the presence of MED judgments in
several other relevance strategies together with their MED Intellectual Obligation, suggests that
the final judgment should also be MED, based on the subjective judgment of these analysts.
G. Overall Summary: The judgment is MED since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim. If, however, such a stance were not taken and coherence were to be
insisted upon, the judgment still would be only MED because the assignment of LOW in the
Direct Empirical cell could not be outweighed by the inconclusive findings in other columns.
82
-------
3.3.4.3.2. Neoplastic Development
The judgment in all cells is NO since none of the studies examined neoplastic
development or any phenomenon related causally to neoplastic development. The finding that
promotion was required in the skin assays, however, suggests that there may be some support
for the claim that formaldehyde DOES NOT act through neoplastic development.
3.3.4.4. Mechanisms
3.3.4.4.1. Genotoxic
A. Direct Empirical: The judgment is HIGH if by "genotoxic" one means simply that
genetic alterations occur since the sister-chromatid exchange data provide HIGH support for the
claim that formaldehyde is able to induce at least one form of genotoxicity. If, however,
"genotoxic" is taken to mean that a transition to cancer occurs through this genetic alteration,
the judgment would be NO since the studies do not provide observations of the causal role of
the SCEs. The former definition has been adopted here.
B. Semi-Empirical Extrapolation: The judgment is NO since exposures were in the range
defining this context.
C. Empirical Correlation: The judgment is HIGH since the production of sister
chromatid exchange (characterized by an epistemic status of HIGH) correlates well (high
strength and specificity) with the ability of a substance to induce genotoxicity. In fact, this could
be taken as a tautology if by "genotoxic" one means simply the ability to induce SOME form of
DNA alteration regardless of whether that alteration is known to be pertinent to cancer. Even if
the definition of genotoxic is taken to imply a genetic alteration leading to a transition between
cancer states, the judgment is MED since SCEs correlate (MED epistemic status to correlation)
with initiation.
83
-------
D. Theory-Based Inference: The judgment is HIGH since the study of sister chromatid
exchange provides clear evidence (epistemic status is HIGH) that DNA damage is taking place
following in vivo exposure to formaldehyde. It is not the case, however, that sister-chromatid
exchange has been well demonstrated to represent the kind of genotoxicity relevant to cancer in
the transition to neoplastic conversion. The SCE data do provide support for the claim that a
biologically-significant dose-rate is reaching the cells of the nasal cavity, which are taken here to
be the target for the nasal tumors examined in the cancer assay. Note that this judgment of
epistemic status employs the first definition of genotoxicity given previously.
E. Existential Insight: The judgment is HIGH since these data produce in the analysts
strong confidence of formaldehyde's ability to induce genotoxicity. It is judged further that the
analysts possess the relevant experience, training and psychological skills necessary to form such
subjective judgments, based on having worked in the science of carcinogenesis for several years.
F. Column Summary: The judgment is HIGH since the Direct Empirical strategy, with
an Intellectual Obligation of HIGH, was HIGH for the data that were available. The presence of
HIGH scores in several other relevance strategies and strengthens this assignment.
G. Overall Summary: The judgment is HIGH since the analysts' strongly foundational
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim. If, however, such a stance were not taken and
coherence were insisted upon, the judgment still would be HIGH because the HIGH score in
four out of the five cells could not be outweighed by the inconclusive findings.
3.3.4.4.2. Non-Genotoxic
The judgment in all cells is NO since the studies did not examine mechanisms of action
related to other than genotoxicity.
84
-------
3 J.5. Summary Comments for Intra-Context Claims
The entries into the cells of Working Table 3.6 indicate that formaldehyde should be
weakly considered a carcinogen in mice when exposure is through inhalation at concentrations at
or above 2 ppm. This judgment is reflected most clearly in the fact that the "Overall Summary"
cell for "Increases Incidence of Cancer" is judged to be LOW due to the lack of statistical
significance for increased incidence of cancer in the exposed groups. While the data employed
here allow only very weak determination of the CLASSIFICATION into which formaldehyde
should be placed, they do provide
MEDIUM support for the claim that formaldehyde induces the STAGE "Neoplastic conversion",
a finding given additional support by the HIGH epistemic status of the claim that formaldehyde
acts by at least one "Genotoxic" MECHANISM.
3.4. Description of Working Table 4.6
3.4.1. Introduction
Working Table 4.6 is utilized to justify premises necessary for extrapolation of the
findings in Working Table 3.6 (the "Observational Context") to the "Target Context" (Context
12). Four separate claims must be warranted, each of which is described in detail in the original
report and in lesser detail below. The data on which judgments for these premises are based
were described in Section 3.3.4. All judgments of Intellectual Obligation are as described in
Section 1.4.
3.4.2. Column Headings in Working Table 4.6
These extrapolation premises or claims as they appear in Working Table 4.6 are:
85
-------
WORKING TABLE 4.6. SUPPORT FOR INTER-CONTEXT EXTRAPOLATION PREMISES
CONTEXT NO. 6 TO CONTEXT NO. 12
Relevance Strategy
I.O.1
Exposure to BSDR
Conversion
BSDR to Effect
Conversion
Host Factors
Environmental
Conditions
Direct Empirical (D.E.)
HI
NO
NO
LO
ME
Semi-Empirical Extrapolation
(S.E.E.)
ME
NO
NO
NO
NO
Empirical Correlation (E.C.)
ME
ME
ME
ME
LO
Theory-based Inference (T.B.I.)
ME
ME
ME
ME
ME
Existential Insight (E.I.)
LO
ME
HI
ME
ME
Overall Assessment
LO
LO
LO
ME
'Intellectual Obligation
-------
1. EXPOSURE TO BSDR: Any differences between the BIOLOGICAL components of
the two contexts with respect to the relationship between exposure and BSDR are not so large
as to call into question whether a BSDR might be present in one context but not the other AT
THE SAME LEVEL OF EXPOSURE.
The pertinent considerations here for each context are whether (1) an exposure occurs
through some compartment of the environment (a compartment refers to the medium of the
environment, such as air or water); (2) an intake by some route can occur (a route refers to
inhalation, ingestion, dermal absorption, intubation or injection); (3) this intake results in an
uptake to a target organ or tissue; (4) clearance from the target allows the production of a
burden; and (5) biological transformation to the active metabolite exists. Where extrapolation is
across exposure levels, it is particularly important to establish that there does not exist a
threshold exposure below which a BSDR is not produced (due, e.g., to necessary saturation of
metabolic transformation). Where extrapolation is across species, it is particularly important to
show that any such threshold, if it exists, does not differ between the species in a manner calling
into question this premise.
2. BSDR TO BIOPHYSICAL EFFECT: Any differences between the BIOLOGICAL
components of the two contexts with respect to the relationship between BSDR and the
production of biophysical effects necessary for transitions are not so large as to call into
question whether a biophysical effect might be present in one context but not the other
ASSUMING THE SAME LEVEL OF BSDR IN BOTH CONTEXTS.
The pertinent considerations here for each context are whether (1) the active metabolite
can interact with causally important biological structures in the organism; (2) this interaction
results in either DNA alterations (for initiation) or stimulation of cellular division (for
promotion); and (3) these biophysical effects are of the type and magnitude necessary to induce
86
-------
transitions designated as either neoplastic development or neoplastic conversion. Where
extrapolation is across exposure levels, it is particularly important to establish that there does
not exist a threshold BSDR below which a biophysical effect is not produced (due, e.g., to
necessary saturation of a pool of DNA damage or a stimulus to cellular proliferation). Where
extrapolation is across species, it is particularly important to show that any such threshold, if it
exists, does not differ between the species in a manner calling into question this premise.
3. ENVIRONMENTAL CONDITIONS: Any differences between the
ENVIRONMENTAL components of the two contexts with respect to the relationship between
exposure (to the substance of interest) and BSDR, or the relationship between BSDR and the
production of biophysical effects necessary for transitions, are not so large as to call into
question whether a biophysical effect might be present in one context but not the other AT THE
SAME LEVEL OF EXPOSURE TO THE SUBSTANCE OF INTEREST.
The pertinent considerations here for each context are whether (1) environmental
conditions increase or decrease the BSDR resulting from a given level of exposure to the
substance of interest, relative to the conditions that exist in the target context; (2) environmental
conditions increase or decrease the probability of biophysical effects from a given level of
BSDR, relative to the target context; and (3) environmental conditions increase or decrease the
probability of transitions from a given level of biophysical effect, relative to the target context.
4. HOST FACTORS: Any differences between the BIOLOGICAL components of the
two contexts with respect to the relationship between the production of biophysical effects, the
probability of transitions necessary for cancer, and the probability of cancer itself are not so
large as to call into question whether an increase in cancer incidence might be present in one
context but not the other AT THE SAME LEVEL OF BIOPHYSICAL EFFECT.
87
-------
The pertinent considerations here for each context are whether (1) a target organ or
tissue exists; (2) biological structures on which relevant biophysical effects are produced exist;
(3) relevant background transitions exist; and (4) repair of biophysical effects, or regression of
states of neoplastic conversion or neoplastic development, exist. Where extrapolation is across
species, it is particularly important to establish that there does not exist a threshold level of
biophysical effect below which a transition is not produced (due, e.g., to necessary saturation of
a stimulus to cellular proliferation before promotion can occur).
In the original report, Working Table 4 (bearing in mind that Working Table 4 in the
present report was numbered Working Table 5 in the original report, as described in Section 1)
also contained an extrapolation premise concerning the role of intersubject and intrasubject
variability. This premise has been removed here and consideration of the role of differences in
intra- and intersubject variability between the two contexts has been incorporated into the 4
separate premises above. This choice was made because variability always refers to variability of
some PROPERTY, and this property would appear in one of the four extrapolation categories
listed above. The analyst then must determine the significance of variability on the relationship
between EXPOSURE AND BSDR, BSDR AND BIOPHYSICAL EFFECT,
ENVIRONMENTAL CONDITIONS, AND HOST FACTORS separately.
3.4 J. Epistemic Status of Extrapolation Premises
This section summaries the judgments made for each of the cells in Working Table 4.6.
The procedure for assigning epistemic status to a cell is identical to that outlined for Working
Table 3.6 and a description is not repeated here.
88
-------
3.4.3.1. Exposure to BSDR
A. Direct Empirical: The judgment is NO since BSDR was not determined in either
context at any level of exposure.
B. Semi-Empirical Extrapolation: The judgment is NO since BSDR was not determined
in either context at any level of exposure.
C. Empirical Correlation: The judgment is MED since the production of a nasal dose
following exposure to formaldehyde in mice is taken to correlate strongly with production of a
nasal dose following exposure of humans to formaldehyde, but metabolic transformation has not
been identified in the human context in the studies employed here. Where the active metabolite
has not been identified, the correlation between exposure and BSDR (rather than dose) will be
weakened. The judgment is not HIGH since it has not been demonstrated that there is not a
threshold exposure for production of a BSDR, or that this threshold is the same between
contexts.
D. Theory-Based Inference: The judgment is MED since exposure clearly occurs in both
contexts; intake clearly occurs in both cases, although at a lower level in the target context than
in the observational context; the absorption of formaldehyde in nasal passages upon inhalation
in both humans and mice has been established with MED epistemic status regardless of level of
exposure; clearance is present but not infinitely rapid in both contexts; but the active metabolite
has not been shown present in the human context. While the theory of pharmacodynamics
suggests that a dose should be produced in both contexts, it is not clear that a BSDR will be
produced since it is possible that a mechanism of metabolic transformation is stimulated or
saturated in mice but not in humans. The judgment is not HIGH since it has not been
demonstrated that there is not a threshold exposure for production of a BSDR, or that this
threshold is the same between contexts.
89
-------
E. Existential Insight: The judgment is MED since these data produce in the analysts
moderate confidence that a BSDR is produced in both contexts.
F. Overall Summary: The judgment is LOW since Empirical Correlation, Existential
Insight and Theory-Based Inference were of MED strength, but none of these has HIGH
Intellectual Obligation. The lack of identification of the active metabolite is troubling
theoretically in cases of extrapolating across species.
3.4.3.2. BSDR to Effect
A. Direct Empirical: The judgment is NO since the same pertinent biophysical effects
were not measured in Context 12 and Context 6.
B. Semi-Empirical Extrapolation: The judgment is NO since this issue was addressed in
A above.
C. Empirical Correlation: The judgment is MED since the ability of the active metabolite
of a substance to induce biophysical effects UPON PRODUCTION OF A BSDR in mice
population (Context 6) is taken to correlate strongly with the same feature in a human
population (Context 12). The judgment is not HIGH since it has not been demonstrated that
there is not a threshold BSDR for production of a biophysical effect, or that this threshold is the
same between contexts.
D. Theory-Based Inference: The judgment is MED since similar cellular structures exist
in both contexts; formaldehyde or its active metabolite is assumed to diffuse in a similar fashion
in both contexts; and interaction with pertinent biological structures should occur through this
diffusion. The judgment is not HIGH since it has not been demonstrated that there is not a
threshold BSDR for production of a biophysical effect, or that this threshold is the same
between contexts.
90
-------
E. Existential Insight: The judgment is HIGH since these data produce in the analysts
moderate confidence that interaction between formaldehyde (or its active metabolite) and
pertinent biological structures is produced in both contexts.
F. Overall Summary: The judgment is LOW since Empirical Correlation and Theory-
Based Inference are MED, while the Intellectual Obligation for these is MED. The presence of
NO epistemic status for the Direct Empirical warrant prevents this judgment from being either
MED or HIGH. This is not offset by the HIGH judgment in Existential Insight, given the LOW
Intellectual Obligation of this warrant.
3.4.3.3. Host Factors
A. Direct Empirical: The judgment is LOW since the presence of target cells is observed
in both species; the presence of cellular targets for biophysical effects is observed in both
species; repair and background transitions are observed to be present in both species; but it has
not been possible to observe whether the background transitions necessary for the particular
damage caused by formaldehyde are present in both species. The judgment is not MED or
HIGH since it has not been demonstrated that there is not a threshold level of biophysical effect
necessary for production of a transitions, or that this threshold is the same between contexts.
B. Semi-Empirical Extrapolation: The judgment is NO since the consideration of host
factors does not require extrapolation across exposure levels, and was dealt with in A above.
C. Empirical Correlation: The judgment is MED the presence of appropriate host factors
in mice is taken to correlate with moderate strength with the same factors in a human
population. The judgment is not HIGH since it has not been demonstrated that there is not a
threshold level of biophysical effect necessary for production of a transitions, or that this
threshold is the same between contexts.
91
-------
D. Theory-Based Inference: The judgment is MED based on the fact that cellular
structure and function is similar in the two contexts since both involve mammalian species, so
similar targets should exist; the target in Context 12 is taken to be the nasal passages, which
clearly exist in both contexts; repair should be present but not completely effective in both
contexts; and background transitions are taken to occur in both contexts as evidenced by the
non-zero rates of effect in suitable controls. The judgment is not HIGH since it has not been
demonstrated that there is not a threshold level of biophysical effect necessary for production of
a transitions, or that this threshold is the same between contexts. In addition, genetic variability
clearly is higher in the target context than in the observational context.
E. Existential Insight: The judgment is MED since the biological similarities between the
two contexts produce in the analysts moderate confidence that the two contexts are similar.
F. Overall Summary: The judgment is LOW since judgments for Empirical Correlation,
Theory-Based Inference and Existential Insight were MED, but the judgment was LOW for
Direct Empirical (with HIGH Intellectual Obligation for the Direct Empirical warrant).
3.4.3.4. Environmental Conditions
A. Direct Empirical: The judgment is MED since the exposure is clearly to formaldehyde
in both contexts; the form is taken to be formaldehyde vapor in both contexts; and concurrent
exposures to other substances was controlled in the observational context. The fact that
concurrent environmental exposures in the target context has not been measured and shown to
be identical to the observational context lowers this judgment from HIGH to MED.
B. Semi-Empirical Extrapolation: The judgment is NO since this warrant does not apply
for this extrapolation premise.
92
-------
C. Empirical Correlation: The judgment is LOW since concurrent environmental
exposures in a controlled experiment (Context 6) correlates poorly with concurrent
environmental exposures as defines Context 12.
D. Theory-Based Inference: The judgment is MED based on the theory that concurrent
exposures will be different in controlled laboratory settings than in uncontrolled human settings,
but that these differences will alter the probability of formaldehyde-induced cancer but not the
presence of carcinogenic action.
E. Existential Insight: The judgment is MED since the environmental differences
between the two contexts produce in the analysts moderate confidence that the two contexts are
similar.
F. Overall Summary: The judgment is MED since most judgments above were either
MED (including Direct Empirical) or NO (Semi-Empirical Extrapolation, which was NO simply
because the same evidence was used in Direct Empirical). The LOW status of Empirical
Correlation is offset partially by the MED Intellectual Obligation of this relevance strategy.
3.5. Description of Working Table 5.6
3.5.1. Introduction
Working Table 5.6 contains judgments of the epistemic status for Inter-Context Claims of
Carcinogenicity. In each entry, the "Observational Context" is Context 6 and the "Target
Context" is Context 12. The judgments in Working Table 5.6 utilize the judgments in Working
Table 3.6 for a particular Claim of Carcinogenicity, as well as the judgments in Working Table
4.6 for the necessary Extrapolation Premises. In each case of an extrapolation, it is the "Overall
Summary" judgment associated with a Claim of Carcinogenicity in Working Table 3.6 that forms
the basis for the extrapolation. The final judgment of epistemic status in Working Table 5.6
93
-------
WORKING TABLE 5.6. INTER-CONTEXT SUPPORT FOR CLAIMS OF CARCINOGENICITY
EXTRAPOLATED FROM CONTEXT 6 TO CONTEXT 12
Claims of Carcinogenicity
Relevance Strategy
I.O.1
Increases
Incidence of
Cancer
Classification(s)
Stage
Mechanism
Complete
Partial
Mixer
Helper
Neo.
Conv.
Neo.
Devel.
Geno-
tox.
Non-
genot
ox.
Direct Empirical (D.E.)
HI
NO
NO
NO
NO
NO
NO
NO
NO
NO
Semi-Empirical
Extrapolation (S.E.E.)
ME
NO
NO
NO
NO
NO
NO
NO
ME
NO
Empirical Correlation
(E.C.)
ME
LO
LO
LO
NO
NO
ME
NO
ME
NO
Theory-based Inference
(T.B.I.)
ME
LO
LO
LO
NO
NO
ME
NO
ME
NO
Existential Insight(E.I.)
LO
ME
ME
LO
NO
NO
ME
NO
ME
NO
Column Summary
LO
LO
LO
NO
NO
ME
NO
ME
NO
Overall Summary
LO
LO
LO
NO
NO
ME
NO
ME
NO
'Intellectual Obligation
-------
cannot be higher than this initial judgment from Working Table 3.6, since the application of the
extrapolation premises can have only the effect of weakening the extrapolation. In other words,
if formaldehyde is judged to "Increase the Incidence of Cancer" with MEDIUM epistemic status
in Working Table 3.6, the highest judgment of epistemic status for the instance of extrapolation
will be MEDIUM, which presumes that the extrapolation premises are judged to each have
HIGH epistemic status in Working Table 4.6.
The analysts have chosen to view warrants from Working Table 3.6 in the following
manner. A particular Claim of Carcinogenicity from Working Table 3.6 may be used as
Empirical Correlation, Theory-Based Inference and/or Existential Insight relevance strategies in
Working Table 5.6 for the same Claim of Carcinogenicity. A particular claim from Working
Table 3.6 may NOT be used as a Direct Empirical relevance strategy in Working Table 5.6,
since there always exists extrapolation across exposure levels. A particular claim from Working
Table 3.6 MAY be used as a Semi-Empirical Extrapolation relevance strategy if it is judged that
the biological and environmental conditions in the two contexts are sufficiently similar to
warrant the claim that these two contexts contain approximately the same biological and
environmental antecedent conditions, differing only in LEVEL of exposure to formaldehyde.
This clearly requires relatively strong epistemic status to each of the four extrapolation premises
found in Working Table 4.6, a condition only weakly met in the present case.
3.5.2. Intellectual Obligation
The assignments of Intellectual Obligation discussed in Section 1.4, and utilized in
Working Table 3.6, are employed in Working Table 5.6.
3.5.3. Claims of Carcinogenicity
94
-------
The same grouping of Claims of Carcinogenicity used as headings in Working Table 3.6
are employed in Working Table 5.6. The judgments of the epistemic status of each of these
claims, for each relevance strategy, are provided in the cells of Working Table 5.6 and are
described below.
3.5.3.1. Increases Incidence of Cancer
A. Direct Empirical: The judgment is NO due to differences in level of exposure and
biological context.
B. Semi-Empirical Extrapolation: The judgment is NO due to LOW epistemic status of
necessary extrapolation premises and differences in context (which are not offset by LOW
strength of extrapolation premises).
C. Empirical Correlation: The judgment is LOW since the claim was given LOW
epistemic status in Working Table 3.6 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.6. While a correlation does exist between
carcinogenicity in mice and humans, the LOW epistemic status of the claim in Working Table
3.6 keeps this judgment from being MED. The fact that Genotoxicity was judged HIGH in
Working Table 3.6 could elevate the epistemic status above LOW, since genotoxic mechanisms
tend to be less context-specific than non-genotoxic mechanisms, but the LOW epistemic status of
the claim of carcinogenicity in Working Table 3.6 precludes this.
D. Theory-Based Inference: The judgment is LOW since the claim was given LOW
epistemic status in Working Table 3.6 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.6. While the Theory of Carcinogenesis does
support the claim that carcinogenicity in one mammalian species is indicative of carcinogenicity
in a second, the LOW epistemic status of the claim in Working Table 3.6 and several of the
95
-------
extrapolation premises keeps this judgment from being MED. The fact that Genotoxicity was
judged HIGH in Working Table 3.6 elevates the epistemic status above LOW, since genotoxic
mechanisms tend to be less context-specific than non-genotoxic mechanisms, but the LOW
epistemic status of the claim of transformation in Working Table 3.6 precludes this.
E. Existential Insight: The judgment is MED since the evidence produces in the analysts
moderate confidence that formaldehyde is a carcinogen in Context 12.
F. Column Summary: The judgment is LOW since Direct Empirical and Semi-Empirical
Extrapolation were NO (with HIGH Intellectual Obligation), and the remainder were either
LOW (with MED Intellectual Obligation) or MED (with LOW Intellectual Obligation).
G. Overall Summary: The judgment is LOW since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
3.5.3.2. Classifications
3.5.3.2.1. Complete Carcinogen
A. Direct Empirical: The judgment is NO due to differences in level of exposure and
biological context.
B. Semi-Empirical Extrapolation: The judgment is NO due to LOW epistemic status of
necessary extrapolation premises and differences in context (which are not offset by LOW
strength of extrapolation premises).
C. Empirical Correlation: The judgment is LOW since the claim was given LOW
epistemic status in Working Table 3.6 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.6. While a correlation does exist between a
96
-------
substance being a complete carcinogen in mice and humans, the LOW epistemic status of the
claim in Working Table 3.6 keeps this judgment from being MED.
D. Theory-Based Inference: The judgment is LOW since the claim was given LOW
epistemic status in Working Table 3.6 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.6. While the Theory of Carcinogenesis does
support the claim that a substance being a complete carcinogen in one mammalian species is
indicative of carcinogenicity in a second, the LOW epistemic status of the claim in Working
Table 3.6 and several of the extrapolation premises keeps this judgment from being MED.
E. Existential Insight: The judgment is MED since the evidence produces in the analysts
moderate confidence that formaldehyde is a carcinogen in Context 12.
F. Column Summary: The judgment is LOW since Direct Empirical and Semi-Empirical
Extrapolation were NO (with HIGH Intellectual Obligation), and the remainder were either
LOW (with MED Intellectual Obligation) or MED (with LOW Intellectual Obligation).
G. Overall Summary: The judgment is LOW since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
3.5.3.2.2. Partial Carcinogen
A. Direct Empirical: The judgment is NO due to differences in level of exposure and
biological context.
B. Semi-Empirical Extrapolation: The judgment is NO due to LOW epistemic status of
necessary extrapolation premises and differences in context (which are not offset by LOW
strength of extrapolation premises).
97
-------
C. Empirical Correlation: The judgment is LOW since the claim was given LOW
epistemic status in Working Table 3.6 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.6. While a correlation does exist between a
substance being a partial carcinogen in mice and humans, the LOW epistemic status of the claim
in Working Table 3.6 keeps this judgment from being MED.
D. Theory-Based Inference: The judgment is LOW since the claim was given LOW
epistemic status in Working Table 3.6 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.6. While the Theory of Carcinogenesis does
support the claim that a substance being a partial carcinogen in one mammalian species is
indicative of carcinogenicity in a second, the LOW epistemic status of the claim in Working
Table 3.6 and several of the extrapolation premises keeps this judgment from being MED.
E. Existential Insight: The judgment is LOW since the evidence produces in the analysts
moderate confidence that formaldehyde is a carcinogen in Context 12.
F. Column Summary: The judgment is LOW since Direct Empirical and Semi-Empirical
Extrapolation were NO (with HIGH Intellectual Obligation), and the remainder LOW (with
LOW to MED Intellectual Obligation).
G. Overall Summary: The judgment is LOW since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
3.5.3.2.3. Mixer
The judgment in all cells is NO since the same judgment is given in Working Table 3.6
(for the reasons stated there).
98
-------
3.5.3.2.4. Helper
The judgment in all cells is NO since the same judgment is given in Working Table 3.6
(for the reasons stated there).
3.5.3.3. Stages
3.5.3.3.1. Neoplastic Conversion
A. Direct Empirical: The judgment is NO due to differences in level of exposure and
biological context.
B. Semi-Empirical Extrapolation: The judgment is NO due to LOW or MED epistemic
status of necessary extrapolation premises and differences in context.
C. Empirical Correlation: The judgment is MED since the claim was given MED
epistemic status in Working Table 3.6 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.6. While a correlation does exist between
neoplastic conversion in these two contexts, the MED epistemic status of the claim in Working
Table 3.6 keeps this judgment from being HIGH. The fact that Genotoxicity was judged HIGH
in Working Table 3.6 elevates the epistemic status above LOW, since genotoxic mechanisms
tend to be less context-specific than non-genotoxic mechanisms.
D. Theory-Based Inference: The judgment is MED since the claim was given MED
epistemic status in Working Table 3.6 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.6. While the Theory of Carcinogenesis does
support the claim that neoplastic conversion may act by common routes across contexts
containing similar DNA structure, the MED epistemic status of the claim in Working Table 3.6
and several of the extrapolation premises keeps this judgment from being HIGH. The fact that
99
-------
Genotoxicity was judged HIGH in Working Table 3.6 elevates the epistemic status above LOW,
since genotoxic mechanisms tend to be less context-specific than non-genotoxic mechanisms.
E. Existential Insight: The judgment is MED since the evidence produces in the analysts
moderate confidence that formaldehyde is a carcinogen in Context 12.
F. Column Summary: The judgment is MED since Direct Empirical and Semi-Empirical
Extrapolation were NO (with MED to HIGH Intellectual Obligation), and the remainder were
MED (with MED or LOW Intellectual Obligation). The lack of Direct Empirical and Semi-
Empirical relevance strategies could also reasonably be taken to imply that the judgment here
should be LOW.
G. Overall Summary: The judgment is MED since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
3.5.3.3.2. Neoplastic Development
The judgment in all cells is NO since the same judgment is given in Working Table 3.6
(for the reasons stated there).
3.5.3.4. Mechanisms
3.5.3.4.1. Genotoxic
A. Direct Empirical: The judgment is NO due to differences in level of exposure and
biological context.
B. Semi-Empirical Extrapolation: The judgment is MED due to HIGH epistemic status
of claim in Working Table 3.6 and LOW or MED epistemic status of necessary extrapolation
premises and differences in context. The fact that a genotoxic mechanism is proposed in
100
-------
Working Table 3.6 raises the epistemic status above LOW since such mechanisms tend to be less
context specific and there is not believed to be a threshold for biophysical effect or initiating
transition. The lack of complete agreement between contexts keeps this judgment from being
HIGH.
C. Empirical Correlation: The judgment is MED since the claim was given HIGH
epistemic status in Working Table 3.6 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.6. While a correlation does exist between
genotoxic mechanisms in different mammalian contexts, the LOW to MED epistemic status of
the extrapolation premises keeps this judgment from being HIGH. The fact that Genotoxicity
was judged HIGH in Working Table 3.6 elevates the epistemic status above LOW, since
genotoxic mechanisms tend to be less context-specific than non-genotoxic mechanisms.
D. Theory-Based Inference: The judgment is MED since the claim was given HIGH
epistemic status in Working Table 3.6 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.6. While the Theory of Carcinogenesis does
support the claim that genotoxic mechanisms may act by similar routes across mammalian cells,
the MED to LOW epistemic status of the extrapolation premises keeps this judgment from
being HIGH. The fact that Genotoxicity was judged HIGH in Working Table 3.6 elevates the
epistemic status above LOW, since genotoxic mechanisms tend to be less context-specific than
non-genotoxic mechanisms.
E. Existential Insight: The judgment is MED since the evidence produces in the analysts
moderate confidence that formaldehyde is a carcinogen in Context 12.
F. Column Summary: The judgment is MED since Direct Empirical was NO but other
cells were assigned MED. The presence of MED epistemic status to the Semi-Empirical
Extrapolation cell is taken to keep the overall judgment at MED.
101
-------
G. Overall Summary: The judgment is MED since the analysts' strongly foundational
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
3.5.3.4.2. Non-Genotoxic
The judgment in all cells is NO since the same judgment is given in Working Table 3.6
(for the reasons stated there).
4. OBSERVATIONAL CONTEXT NO. 11
4.1. Description of Working Table 1.11
Working Table 1 for Context 11 is provided below. As in all subsequent sections of this
report, a Working Table is identified by both the general Working Table number (here, 1) and
the context (here, 11). As a result, this Working Table is assigned the number 1.11. The
biological components of this context are given by the fact that all exposed organisms were
humans of adult age and in a state of health sufficient to allow employment.
The exposures in this context were all to formaldehyde in air at concentrations above 2
ppm, placing them into the category of high exposures as defined in the Introduction. Length of
exposure varied between the studies, but this was not taken to justify development of separate
contexts as subjects were classified into exposure groups based on cumulative exposure. It was
assumed that mechanisms of action of formaldehyde might be slightly different depending upon
length of exposure, but that this difference would appear in the magnitude of effect rather than
the presence and direction of that effect. This choice is somewhat tenuous since there is
evidence that effects such as promotion may require some minimal period of exposure.
102
-------
WORKING TABLE 1.11. STUDIES AVAILABLE FOR OBSERVATIONAL CONTEXT NUMBER 11: HUMANS
EXPOSED TO AIRBORNE FORMALDEHYDE AT CONCENTRATIONS IN EXCESS OF 2 PPM
Study
Number1
Data Category
Description
1
TR
Primary Reference #5 in Table 1
An historical cohort of 26,561 workers employed in 10 facilities was
exposed to formaldehyde at cumulative exposures of 0-0.5; 0.5-5.5; and
>5.5 ppm-years. Eight-hour time weighted average historical exposure
levels to formaldehyde vapor were estimated for job/work
area/plant/calendar year combinations. More than 6,000 formaldehyde
measurements constituted the base of exposure data. Exposures to
particulates and other suspect carcinogens were noted but not estimated.
SMRs were calculated using the general U.S. population as the control
group. In addition, SRRs were calculated relative to the lowest exposure
group.
IThe study number refers to the order of the study within this context. The primary reference number, indexed to Table I, is provided in the "Description"
column
-------
WORKING TABLE 1.11. Continued
Study
Number'
Data Category
Description
2
TR and CEC
Primary Reference #6 in Table 1
A case-control study of cancers of the nasal passages and paranasal
sinuses was conducted for workers exposed to formaldehyde, primarily in
air. 207 cases and 409 controls were employed. Exposures to 14 other
substances or groups of substances (dusts, dyes, heavy metals, industrial
fumes) were also studied. Exposure variables were lifetime average
concentration, duration of exposure, and cumulative exposure (<2, 2-3 and
>3 ppm-years; also <30 and >30 ppm-years). Odds ratios were calculated
for sinonasal cancers in the exposed groups and reported for each
exposure or cumulative exposure category.
'The study number refers to the order of the study within this context. The primary reference number, indexed to Table 1, is provided in the "Description"
column
-------
Three bodies of data were identified for this context. These are described in Working
Table 1.11., and are divided into the data categories of:
(a) Tumor Response (taken from Reference #5 and Reference #6 in Table 1 of this
report). Here, the sole Tumor Response is taken to be incidence of lung and nasal tumors,
which clearly is a form of direct measurement of tumors. It is given the label TR1.
(b) Concurrent Environmental Conditions (taken from Reference #6 in Table 1 of this
report). Here, the sole Concurrent Environmental Condition is taken to be exposure to wood
dust. It is given the label CEC1.
Data in other data categories were not available for the studies utilized in this context.
42. Description of Working Table 2.11
The judgments for this Working Table are provided in the cells of Working Table 2.11.
The text here describes the reasons for these judgments. The data on which these judgments are
based are provided as various "Exhibits" after Working Table 2.11., with a numbering system
described in Section 4.2.4.
42.1. Data Categories and Items
The same data categories and items described with respect to Working Table 1.11. are
employed in Working Table 2.11. These are:
(a) Tumor Response, containing TR1 (incidence of nasal tumors);
(b) Concurrent Environmental Conditions, containing CEC1 (exposure to wood dust).
These two separate data categories and three bodies of data are displayed in Working Table
2.11.
V
422. Judgments of Completeness
103
-------
WORKING TABLE 2.11. DATA CHARACTERISTICS FOR OBSERVATIONAL CONTEXT NUMBER 11
Data Category/Item
Description
Completeness
(Hi/Me/Lo/No)
Utility
(Hi/Me/Lo/No)
Strength of Effect
(Hi/Me/Lo/No)
Exposure-
Specific
Effect1
Tumor Response
Nasal/lung tumors (TR1.1)
Nasal/lung tumors (TR1.2)
LO
LO
ME
LO
LO
LO
WT2.C11.
TR1.1
WT2.C11.
TR1.2
Biophysical Effect
No study available
Pharm acodynam ics
No study available
Host Factors
No study available
Concurrent Environmental
Conditions
Exposure to wood dust
(CEC1.1)
LO
LO
HI
WT2.C11.
CEC1.1
Related Substances Assessment
No study available
'Refers to the Exhibit Number for the data as described in the text.
-------
WT2.C11.TR1.1
TABLE II. SMRs for Lung Cancer by Cumulative Exposure to Formaldehyde (Wage White
Men. £20 Years After First Exposure)
Measure
of exposure
ppm-
¦years
X for
trend
>0-<0.5
0.5—< 5.5
>5 5
Total
Cumulative usins
I.J
l.7d
1.6
l.6d
0.08
jobs with more confident
(23' I f .8 )J
(3J/20.2)
(23. U .Si
(80/50.8)
estimates of exposure
Cumulative excluding
1.3
l.6d
1.2
l.4J
-0.89
exposures during last
(48/36.3)
(63 JO.7)
(31.26.9)
(IJ 2/103.9)
15 years
Cumulative for new
1.3
l.Jd
I.I
1.3d
-0.81
hires onlvh
(JJ-32.9)
(35-2J.2)
(19.17.5i
(98/7J.6)
Cumulative for workers
1.3
1.3d
l.7d
i.5d
0.70
employed on Stan dates
1 J
(16/11.9)
(25; 15.0)
(JJ/29.2)
Cumulative fwith potential
I.J
1.3
1.3
1.3 d
-0.24
exposure to formaldehyde
(16/11.2)
(28/22.2)
(3I.2J.6)
(75/58.0)
in solution)
Cumulative (with potential
1.2
1.2
I.J
1.3
0.59
contact to formaldehyde
(32/26.6)
(32/25.6)
(26/18. 6)
(90/70.8)
containing particulates)'
Exposures lagged 5 years'
1.3d
1.2
1.2
l.2d
-0.11
(78.61.0)
(80/67.5)
(49/39.8)
(207/168.3)
Exposures lagged 10 years'
I.J
1.2
1.2
1.2d
-0.05
(75/5J.J)
(76/6J.1)
(41/33.0)
(192/156.5)
Exposures lagged 20 years0
1.2
1.6"
1.2
1.4d
1
f~l
0
1
(56/J5.0)
(66/JI.9)
(20/16.7)
(142/103.7)
Exposures lagged 30 years'
1.2
1.6
I.I
1.4
-0.12
(18/1 J.3)
f 18/11 .J)
(3/2.7)
( 39/28.3)
'Observed/expected Nos.
bHired after Stan date of study.
Total cohort: not restricted to £20 year latency.
dps0.05.
-------
WT2.C11.TR1.2
TABLE HI-ODDS UATIOS1 (OR) ASSOCIATED WITH FORMALDEHYDE
EXPOSURE AMONG MEN: SOL'AMOUS-CELL CARCINOMAS
Formaldehyde
exposure
OR
951 CI;
None
1
Possible
0.96
0.38-2.42
Probable or definite
Average level
<2
>2
0.70
1.32
0.28-1.73
0.54-3.24
Duration
i 20 yr
>20 vr
1.09
0.76
0.48-2.50
0.29-2.01
Cumulative level
<30
>30
1.26
0.68
0.54-2.94
0.27-1.75
Date of 1st exposure
< 1944
>1945
1.47
0.66
0.58-3.71
0.27-1.64
Age at 1st exposure
< 15
16-20
>20
1.08
0.98
0.85
0.32-3.60
0.34-2.85
0.33-2.17
'Adjusted for age (s55. 56-65, >65). exposure to wood dust
(none or possible, probable or definite), exposure to glues and
adhesives (none or possible, probable or definite).-:Conndence
interval.
-------
WT2.C11.CEC1.1
. TABLI vni - ODDS RATIOS (OR) ASSOCIATED WITH EXPOSURE TO GLUES AND ADHESIVES (MEN)
OR1 adjusted for exposure to
Exposure to gJues and adhesive*
Paints, lacquers
and varmsner1
FortMidehvde-*
Wood dust"
Squamous-cell carcinomas
1.7 (0.6-4.3)
Probable or definite, low
1.5(0.5-3.7)
1.5 (0.6-3.7)
, level3
Probable or definite, medi-
23 (0.9-6.3)
2.1 (0.8-5.3)
2.1 (0.8-5.6)
um-high level3
Adenocarcinomas
Probable or definite, low
20.0 (7.2-55.3)
12.6 (4.5-34.9)
8.7 (2.7-28.4)
level
Probable or definite, medi-
88.2 (32.7-237.8)
51.2(19.0-138.0)
11.9 (4.0-35.6)
um-high level
"Other
Probable or definite, low
2.1 (0.6-7.5)
1.8(0.5-6.1)
2.4 (0.7-7.9)
level
Probable or definite, medi-
4.6(1.4-14.6)
3.8(1.2-11,7)
5.7(1.7-18.4)
•um-high level
'95% confidence interval in parentheses.—'Lifetime average level <3.-'Lifetime average level
>3.-4In 2 categories: none or possible, probable or definite.—"In 3 categories: none, possible,
probable or definite.-"In 2 categories: none or possible, probable or definite except for adenocarci-
nomas: none, possible, probable or definite with a lifetime average level £ 3 w. probable or definite
with a lifetime average level >3.
-------
In all cases, the judgment of completeness was based on a comparison of the number of
studies used in a data category/item for this context against the number available in the
complete reference list provided in Appendix A. The judgment was HIGH if the data used in
this Working Table was a relatively complete subset of the data indicated in the Appendix,
where "complete" was taken to mean somewhere on the order of 70% or above. The judgment
was MEDIUM if the data used in this Working Table constituted somewhere between 25 and
70% of the total body of potentially available data. The judgment was LOW if the data used
constituted less than 25% of the data available. These assignments also included limited
consideration of the QUALITY of the data, since it should not necessarily count against an
analysis if the analyst fails to collect data of poor quality. This consideration also is dealt with in
the assignment of Utility for the collected studies. Still, it is evident that the judgment of
Completeness might reasonably be expected to include consideration of something more than
the fraction of studies collected in a given data category/item. In the present analysis,
Completeness was increased slightly if the collected data for a given data category were the most
heavily cited in the secondary literature on formaldehyde. This leads to the following judgments
of Completeness:
For the data category TR1, the judgment for Completeness is LOW since the 2 studies
examined here are 2 of 20 epidemiological analyses identified in the literature search in
Appendix A. The other studies are Babich (1985); Blair and Stewart (1990); Blair et al., (1987);
Dosemeci et al., (1991); Blair et al., (1990); Holmstrom and Lund (1991); Kauppinen and
Partanen (1988); Luce et al., (1993); Malker et al., (1990); Marsh et al., (1992a); Marsh et al.,
(1992); Olsen et al., (1984); Partanen (1993); Siegel (1983); Squire and Cameron (1984); Sterling
and Weinkam (1988); and Stewart et al., (1990).
104
-------
For the data category CEC1, the judgment for Completeness is LOW since essentially all
studies mentioned for TR1 also contained some information on concurrent exposures.
423. Judgments of Utility, Strength of Effect, and Exposure-Specific Effect
The remaining judgments in Working Table 2.11 refer to the qualities of the data
ACTUALLY EXAMINED IN THIS EXAMPLE. By Utility, we mean the quality of the study
DESIGN and its ability to address in rigorous fashion questions pertinent to hazard
identification for formaldehyde. No attempt is made to determine the specific inferences
towards which the data will be directed in later Working Tables, since the intent here is only to
determine the degree to which a body of data arose from a study design appropriate for
accurate and precise measurements of the quantity considered in the study. Utility is increased
by high standards of study design, including the ability of the experiment to demonstrate a
causal link between exposure to formaldehyde and the measured effect.
By Strength of Effect, we mean the degree to which the study actually revealed an effect.
Strength is increased when the effect noted possessed statistical significance.
By Exposure-Specific Effect, we mean a summary of the actual effect observed (putting
aside questions of the reliability of this effect, which was dealt with in the assignment of Utility
and Strength of Effect). This summarization is provided in the following text, and the data
themselves are provided in the various Exhibits appearing after Working Table 2.11.
The following judgments then apply to Working Table 2.11:
For the data category TR1, considering first Study #1, the Utility judgment is MEDIUM
since the study employed 26,561 workers across 10 facilities; measurements of formaldehyde
were utilized; consideration of confounding exposures was appropriate; appropriate
epidemiological methods of determining cause of death are employed; the study does not
105
-------
constitute a deliberate experiment (important for causal claims); and the measured effect is
cancer. The Strength of Effect judgment is LOW since there was no consistent finding across
facilities, and no discernible exposure-response function. The Exposure-Specific Effect is
summarized in Exhibit WT2.C11.TR1.1. The SMR generally was elevated in all exposed groups,
varying between 1.4 and 1.7. No exposure-response slope is evident, regardless of lag time (10,
20 and 30 years were examined). Mortality was more strongly associated with exposures other
than formaldehyde.
For the data category TR1, considering Study #2, The Utility judgment is LOW since
the study was of case-control design with 207 cases and 409 controls; consideration of
confounding exposures was adequate; exposure assignments were based on subjective expert
judgment rather than direct measurements; the study does not constitute a deliberate
experiment (important for causal claims). The Strength of Effect judgment is LOW since
confidence intervals on odds ratios were large, encompassing both 0.0 and 3.0 in most exposure
groups. No statistically significant association was found except in the presence of wood dust.
The Exposure-Specific Effect is summarized in Exhibit WT2.C11.TR1.2. Maximum likelihood
estimates of odds ratios are above 1.0 in all groups except those beginning exposure prior to
1955.
For the data category CEC1, the Utility judgment is LOW since the study was of case-
control design with 207 cases and 409 controls; exposure assignments were based on subjective
expert judgment rather than direct measurements; the study does not constitute a deliberate
experiment (important for causal claims). The Strength of Effect judgment is HIGH since the
odds ratio for formaldehyde exposure is elevated above 1.0 with statistical significance (p<0.05)
in the group exposed to wood dust. The Exposure-Specific Effect is summarized in Exhibit
WT2.C11.CEC1.1. When exposure to medium or high concentrations of wood dust is considered,
106
-------
the odds ratio for formaldehyde exposures at >2 ppm is 5.33 (95% CI of 1.28 to 22.20). This
contrasts with an odds ratio of 3.04 (95% CI of 0.95 to 9.70) when exposure to wood dust is not
considered.
42.4. Exhibits for Working Table 2.11
The data for the various data categories are provided as "Exhibits" after Working Table
2.11. The numbering system for the exhibits was chosen to display information on (in order of
appearance) the Working Table; the Context; the specific effect measured; and the study
number within that specific effect. For example, Exhibit WT2.C11.TR1.2 refers to Working
Table 2 (all exhibits have this assignment in common); Context 11; Tumor Response of type 1;
and data set #2 within the data category TR1. This same numbering scheme will be employed in
all other Contexts.
4 J. Description of Working Table 3.11
43.1. Introduction
The central task of Working Table 3.11 is to draw inferences on the carcinogenicity of
formaldehyde WITHIN CONTEXT 11, employing data specific to that context. This task
requires that several judgments be made sequentially:
(a) The analyst must determine the particular Claims of Carcinogenicity to be considered
in the analysis. These various claims are discussed in detail in Section 4.3.4.
(b) The available Relevance Strategies by which a given body of data may be related to a
specific Claim of Carcinogenicity must be developed. These strategies were discussed in detail in
Section 1.4., and the same definitions are applied here.
107
-------
WORKING TABLE 3.11 INTRA-CONTEXT SUPPORT FOR CLAIMS OF CARCINOGENICITY IN CONTEXT 11
Claims of Carcinogenicity
Relevance Strategy
I.O.1
Increases
Incidence of
Cancer
Classification(s)
Stage
Mechanism
Complete
Partial
Mixer
Helper
Neo.
Conv.
Neo.
Devel.
Geno-
tox.
Non-
genot
ox.
Direct Empirical (D£.)
HI
ME
NO
NO
NO
NO
NO
NO
NO
NO
Semi-Empirical
Extrapolation (S.E.E.)
ME
NO
NO
NO
NO
NO
NO
NO
NO
NO
Empirical Correlation
(E.C.)
ME
NO
NO
NO
NO
NO
NO
NO
NO
NO
Theory-based Inference
(T.B.I.)
ME
NO
NO
NO
NO
NO
NO
NO
NO
NO
Existential Insight(E.I.)
LO
LO
NO
NO
NO
NO
NO
NO
NO
NO
Column Summary
LO
NO
NO
NO
NO
NO
NO
NO
NO
Overall Summary
LO
NO
NO
NO
NO
NO
NO
NO
NO
'Intellectual Obligation
-------
(c) The analyst must assign a measure of Intellectual Obligation to each Relevance
Strategy. This assignment was described in detail in Section 1.4., and the same assignments are
made here.
(d) The analyst must judge the epistemic status of a specific Claim of Carcinogenicity
arising from a specific Relevance Strategy. As described in Section 1.4., this judgment is separate
from the judgment of Intellectual Obligation for that Relevance Strategy. This judgment is
described in more detail in Section 4.3.2. The assignments are entered into the appropriate cells
of Working Table 3.11.
(e) The analyst must produce a summary judgment of the epistemic status of a particular
Claim of Carcinogenicity by combining the judgments from each of the five Relevance Strategies
for that claim. This summary includes consideration of both the epistemic status of that claim
for each Relevance Strategy, as entered into the separate cells of the Working Table specific to
each strategy, as well as the Intellectual Obligation assigned to that strategy. The assignments
are entered into the appropriate COLUMN SUMMARY cells of Working Table 3.11.
(f) The analyst must produce a summary judgment of the COHERENCE of the various
Claims of Carcinogenicity. This task is performed by examining the coherence ACROSS the
various COLUMN SUMMARIES appearing in the Working Table. The assignments are entered
into the appropriate OVERALL SUMMARY cells of Working Table 3.11.
432. Epistemic Status of Claims of Carcinogenicity
The entries into the various cells of Working Table 3.11. reflect the judgment of the
analyst concerning the epistemic status of the Claim of Carcinogenicity being considered, FOR
A SPECIFIC RELEVANCE STRATEGY. This judgment will have two components:
108
-------
(a) Each Relevance Strategy calls upon a selected body of data from which the reasoning
will proceed. Specifically, the Direct Empirical strategy requires data on direct measurements of
the effect underlying the Claim of Carcinogenicity, made at the exposure to formaldehyde
defining the context (here, concentrations above 2 ppm). The Semi-Empirical Extrapolation
strategy requires data on direct measurements of the effect underlying the Claim of
Carcinogenicity, made at exposures to formaldehyde other than those defining the context. The
Empirical Correlation strategy requires data on effects other than the effect underlying the
Claim of Carcinogenicity, where this "other" effect is taken to correlate with appearance of the
effect of interest. The Theory-Based Inference strategy requires data on an effect other than the
effect underlying the Claim of Carcinogenicity, where this "other" effect is taken to be a causal
step in the production of the effect of interest as specified by an established theory. The
Existential Insight strategy does not invoke data, except in the sense that a subjective judgment
from an individual might be counted as an instance of data in and of itself.
The analyst must assign the "foundational quality" to the data used in each relevance
strategy. By "foundational quality", we mean the quality of the data with respect to the question
those data were intended to address BY THE ORIGINAL INVESTIGATOR. This judgment is
based on the results of Working Table 2.11 and includes consideration of the Completeness,
Utility, Strength of Effect and Exposure-Specific Effect from that table. While the judgment
ultimately is subjective, it is constrained in that it must increase (from NO to LOW to
MEDIUM to HIGH) as the strength of the judgments for Completeness, Utility and Strength of
Effect increase in Working Table 2.11. The summary of Exposure-Specific Effect may also enter
into the assignment of epistemic status to a Claim of Carcinogenicity if the analyst determines
that only effects above a given magnitude will exert an influence on carcinogenic processes (an
issue of importance when possible thresholds are considered).
109
-------
(b) The analyst then must judge the strength of any "background premises" required by
the Relevance Strategy under consideration. These background premises are defined in detail in
the original report, but may be summarized as the set of (often implicit) premises that must be
introduced into an analysis if the data in Working Table 2.11. are to be used to draw inferences
about the Claims of Carcinogenicity for the context being considered. Presumably, many of these
background premises already are considered valid once Working Tables 1.11 and 2.11 are
complete; otherwise, the data would not have been assigned to this context. But other
background premises concerning the reliability of extrapolation procedures (for the Semi-
Empirical Extrapolation relevance strategy), the strength and specificity of correlations (for the
Empirical Correlation relevance strategy), the degree of verification and validation of theories
(for the Theory-Based Inference relevance strategy), and the credentials of individuals making
subjective judgments (for the Existential Insight relevance strategy) must be warranted prior to
their use.
Having established these two components of epistemic status (foundational quality and
background premises) for a given Claim of Carcinogenicity, the analyst then produces a
composite judgment of the epistemic status of a given Claim of Carcinogenicity and enters this
into the appropriate cell of Working Table 3.11. There are two separate claims that might be
made within one of these cells. The first is the claim that formaldehyde DOES produce the
effect underlying that claim (such as increased incidence of cancer or genotoxic effects). In that
case, the assignments of NO/LOW/MED/HIGH refer to the epistemic status of the claim that
formaldehyde produces the effect. This assignment is entered into the area ABOVE the dashed
line in a cell. The second claim is that formaldehyde DOES NOT produce the effect underlying
the claim. In that case, the assignment of NO/LOW/MED/HIGH refer to the epistemic status
of the claim that formaldehyde DOES NOT produce the effect. This assignment is entered into
110
-------
the area BELOW the dashed line in a cell. THE READER SHOULD NOTE THAT THIS IS A
NEW FEATURE OF THESE WORKING TABLES NOT DISCUSSED IN THE ORIGINAL
REPORT. The reason for this addition to the Working Tables is that the fact that a particular
body of data does not support the claim that formaldehyde does induce a specific effect IS NOT
NECESSARILY evidence that formaldehyde does not induce that effect. It may simply be the
case that the data were obtained under conditions incapable of revealing any effect that might
be present.
This consideration is important due to the issue of "burden of proof'. At times, the
burden may lie on the regulatory agency to show that formaldehyde DOES induce cancer.
Attention then would be directed towards the judgments in the upper halves of the various cells
of Working Table 3.11. At other times, the burden may lie on the producer of formaldehyde to
show that formaldehyde DOES NOT induce cancer. Attention then would be directed towards
the judgments in the lower halves of the various cells of Working Table 3.11. IN THE
PRESENT EXAMPLE, ONLY JUDGMENTS THAT AN EFFECT IS PRODUCED BY
FORMALDEHYDE ARE DEVELOPED. THE PROCESS OF ANALYSIS WOULD BE
REPEATED FOR THE LOWER HALVES OF ALL CELLS IF THE BURDEN OF PROOF
WAS SHIFTED.
It is important here to consider the broad features of the theory of carcinogenesis to be
used throughout this section. These features structure the interpretation of data used in the
various Relevance Strategies. The theory of carcinogenesis employed here is that cells are
transformed to cancer by three steps or transitions. These are termed initiation, promotion and
progression, each of which may require one or more sub-steps. For each transition to occur, a
biologically-significant dose-rate must reach the targets in appropriate cells so that an interaction
111
-------
between the substance or a metabolite may take place. This target may or may not have been
identified for a particular substance and tumor type.
The initiation step is taken to involve alteration of DNA structure and function,
producing a new genome. This alteration must then remain intact without being repaired by
processes normally operating in cells or induced by the damage itself. The ability of a substance
to bring a stable alteration about may be determined from an increase in cancer incidence
following (1) application of the initiating agent and (2) subsequent application of a promoting
agent such as TPA. In the absence of such data, the ability of a substance to interact with DNA,
to alter DNA, and/or to induce mutation provides partial but not conclusive evidence of
initiating activity. In addition, the ability of a substance to lower DNA repair rates or fidelity of
repair is partial but not conclusive evidence that the substance will enhance the initiating activity
of other substances or of background events.
The promotion step is taken to involve loss of growth control, represented by an
imbalance between rates of growth and death of cells in a colony and a continued net rate of
expansion even upon reaching confluence. The mechanism for this step is unspecified, but may
result in either a lowering of the rate of removal of cells (as in decreased apoptosis) or an
increase in the rate of division (as in abnormal proliferation). The ability of a substance to bring
this about may be determined from an increase in transformation following (1) prior application
of a known initiating agent and (2) subsequent application of the substance of interest. In the
absence of such data, the ability of a substance to alter intercellular communication, the rate of
cellular division, the rate of apoptosis, and/or net rate of growth provides partial but not
conclusive evidence of promoting activity.
The progression step is taken to be poorly understood at present, but probably is related
to the metastatic properties of the promoted cells. There is no assay for progression at present.
112
-------
Partial but inconclusive evidence of progressing activity might be obtained from a demonstration
that cells have taken on the ability to dislodge from the primary tumor, spread to other organs
or tissues, reattach, and grow to fatal tumors.
If a substance induces only one of the required transitions, it will be necessary for other
transitions to be induced by antecedent conditions if the substance of interest is to be
considered a Partial carcinogen (as defined in Section 2.3.4). The fact that the background
incidence of cancer is not zero may be taken to indicate that each of these other transitions will
always be present. If this assumption is adopted, the ability of a substance to increase the
incidence of cancer cannot be used to differentiate between claims of a Complete or a Partial
carcinogen (as defined in Section 4.3.4). Existing empirical evidence, however, is suggestive of
there being several routes by which cancer may occur. It is not inconsistent to assume that a
substance acts to induce only a subset of transitions, and that the remaining transitions required
are not those associated with the mechanism by which the antecedent conditions of the
experimental context act to bring about background cancer.
The epistemic status of this Theory of Carcinogenesis is taken here to be MED based on
an examination of the available empirical evidence linking DNA alteration, alteration of rates of
growth, and the carcinogenic potential of a range of substances tested in assays for initiation,
promotion and complete cancer. A judgment of HIGH is not considered appropriate since much
of the warrant for the theory is subjective judgment of experts in the field based on the
coherence of the explanatory framework and the finding of (unexplained) correlations.
4.3.3. Intellectual Obligation
The issue of Intellectual Obligation was discussed in detail in Section 1.4. Since the
judgments and assignments are common to all Working Tables and Contexts, this discussion is
113
-------
not repeated here. By way of a summary, the Direct Empirical Relevance Strategy was assigned
an Intellectual Obligation of HIGH; the Semi-Empirical Extrapolation, Theory-Based Inference
and Empirical Correlation Relevance Strategies were assigned an Intellectual Obligation of
MEDIUM; and the Existential Insight Relevance Strategy was assigned an Intellectual
Obligation of LOW. The assignments are entered into the appropriate cells of Working Table
3.11.
43.4. Claims of Carcinogenicity
The framework employed here allows for a wide variety of claims concerning the
carcinogenic action of formaldehyde. In the original report, it was determined that it is desirable
to augment the claim that a substance increases the incidence of cancer, since this sole claim
does not provide full detail on the antecedent conditions under which a substance such as
formaldehyde can exert its carcinogenic potential. It was determined that a more detailed
analysis would subdivide the claim of increased incidence of cancer into more detailed claims
potentially of use in determining the most effective regulatory strategy. The different Claims of
Carcinogenicity employed in Working Table 3.11 (and all subsequent Working Tables discussed
in this report) are:
(a) Increases Incidence of Cancer, implying that formaldehyde raises the incidence of
cancer, raises the multiplicity of cancer at sites, and/or changes the age at which cancer
appears.
(b) Complete Carcinogen, implying that formaldehyde increases the incidence of cancer
even when it is not delivered in conjunction with concurrent exposures.
(c) Partial Carcinogen, implying that formaldehyde increases the incidence of cancer only
when delivered in conjunction with concurrent exposures that complete the process of
carcinogenesis.
(d) Mixer, implying that formaldehyde exerts its carcinogenic effect only when it is
combined with another substance, and where this combination results in a new substance
which is at least a partial carcinogen.
114
-------
(e) Helper, implying that formaldehyde does not in and of itself induce transitions to
cancer, but that it facilitates the carcinogenic actibn of another substance.
(f) Neoplastic Conversion, implying that formaldehyde acts by converting cells to
neoplastic potential.
(g) Neoplastic Development, implying that formaldehyde acts by producing growth in
colonies of cells already having neoplastic potential.
(h) Genotoxicity, implying that formaldehyde acts through a mechanism involving
interaction with the genetic material of a cell.
(i) Non-genotoxicity, implying that formaldehyde acts through a mechanism other than
interaction with the genetic material of a cell.
The following summary considerations from Working Table 2.11 provide the foundations
for all inferences to Claims of Carcinogenicity drawn in this context:
TR1.1 demonstrates that in vivo exposure of humans to airborne formaldehyde at
cumulative exposures of (0 to 0.5), (0.5 to 5.5) and (>5.5) ppm-years results in an increase in
the SMR above 1.0. The foundational quality of this claim is judged to be LOW because the
completeness was low, the strength of effect was low (no consistent finding across facilities and
no exposure-response pattern) and the utility was medium (moderate study design and a
statistically significant increase in SMR for the "total exposed group"). It is important to note
that no exposure-response pattern was found, weakening any claim that the observed effect was
connected causally to the formaldehyde exposure.
TR1.2 demonstrates that in vivo exposure of humans to formaldehyde at cumulative
exposures of greater than 30 ppm-years results in an odds ratio in excess of 1.0. The
foundational quality of this claim is judged to be LOW because completeness is low, utility is
low (the elevation in odds ratio was not statistically significant) and strength of effect is low. The
only statistically significant elevation in odds ratio was found in workers exposed concurrently to
wood dust.
115
-------
CEC1.1 demonstrates that in vivo exposure of humans to formaldehyde results, with
statistical significance, in an odds ratio in excess of 1.0 when concurrent exposure to medium or
high concentrations of wood dust occurs. The foundational quality of this claim is MED since
the characteristics of the study are essentially identical to those of TRl-Study 2 above, and
statistical significance was obtained for the elevated SMR.
The separate judgments for Claims of Carcinogenicity for Working Table 3.11 are
provided in the sections that follow. In the case of each claim, explanations are given of the
judgment for each Relevance Strategy; then for the Column Summary across Relevance
Strategies for that claim; then for the Overall Summary across claims.
4.3.4.1. Increases Incidence of Cancer
The judgments for this Claim of Carcinogenicity for the various Relevance Strategies and
for this context are shown in the cells of Working Table 3.11. The basis for these judgments are
given below.
A. Direct Empirical: The judgment is MED for at least the "total exposed group" since
there was a statistically significant increase in SMR above 1.0 for this group in Study 1. The
judgment is not HIGH since there was no exposure-response pattern, and since the results of
Study 2 were not statistically significant.
B. Semi-Empirical Extrapolation: The judgment is NO since exposures were in the range
defining this context.
C. Empirical Correlation: The judgment is NO since "cancer incidence" was measured in
all studies.
D. Theory-Based Inference: The judgment is NO since only cancer incidence was
determined, rather than steps of carcinogenesis.
116
-------
E. Existential Insight: The judgment is LOW since the analysts subjectively feel that
these data provide only marginal evidence of carcinogenic ability. It is judged further that the
analysts possess the relevant experience, training and psychological skills necessary to form such
subjective judgments, based on having worked in the science of carcinogenesis for several years.
F. Column Summary: The judgment is LOW since the Direct Empirical strategy was
MED (with an intellectual obligation of HIGH), the Existential Insight strategy was LOW (but
with LOW intellectual obligation), but all other strategies were NO (and carried MED
intellectual obligation).
G. Overall Summary: The judgment is LOW since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim. If, however, such a stance was not taken, insisting upon coherence, the
judgment still would be LOW because the assignment of LOW in the Direct Empirical cell could
not be outweighed by the inconclusive findings in other columns.
4.3.4.2. Classifications
4.3.4.2.1. Complete Carcinogen
The judgment in all cells is NO since cancer incidence (or SMR or OR) was measured
rather than the ability of formaldehyde to induce specific steps in carcinogenesis. Since the
background incidence of cancer was non-zero, it is clear that each transition required for cancer
occurs with some non-zero background probability in the population of cells. As a result, it is
not possible to distinguish between the hypothesis that formaldehyde induces ALL transitions
(and is, therefore, a complete carcinogen), and the hypothesis that formaldehyde induces only a
subset of the transitions with the remainder being completed by background transitions (and is,
therefore, a partial carcinogen). The analysts recognize, however, that the definition of a
117
-------
complete carcinogen COULD be taken to be a substance that does not require the concurrent
application of either a known initiator or promoter. In that case, the present studies could be
taken to indicate weakly (with an epistemic status of LOW) that formaldehyde is not a complete
carcinogen, since Study 2 suggests that exposure to wood dust also might be required. Study 2,
however, also shows an increase in cancer incidence in groups exposed to formaldehyde when
there is no control for other exposures (including wood dust). This suggests that wood dust is
not necessary for formaldehyde to exert an effect. At most, the data suggest that wood dust
might be a HELPER or MIXER with formaldehyde, or a partial carcinogen itself; the same data
cannot be used to determine WHICH of these hypotheses concerning wood dust is correct, nor
can the reasoning be reversed to make statements about whether FORMALDEHYDE is a
helper, mixer or partial carcinogen (which is more to the point here).
4.3.4.2.2. Partial Carcinogen
The judgment in all cells is NO since cancer incidence (or SMR or OR) was measured
rather than the ability of formaldehyde to induce specific steps in carcinogenesis. Since the
background incidence of cancer was non-zero, it is clear that each transition required for cancer
occurs with some non-zero background probability in the population of cells. As a result, it is
not possible to distinguish between the hypothesis that formaldehyde induces ALL transitions
(and is, therefore, a complete carcinogen), and the hypothesis that formaldehyde induces only a
subset of the transitions with the remainder being completed by background transitions (and is,
therefore, a partial carcinogen). The analysts recognize, however, that the definition of a
complete carcinogen COULD be taken to be a substance that does not require the concurrent
application of either a known initiator or promoter. In that case, the present studies could be
taken to indicate weakly (with an epistemic status of LOW) that formaldehyde is not a complete
118
-------
carcinogen, since Study 2 suggests that exposure to wood dust also might be required. Study 2,
however, also shows an increase in cancer incidence in groups exposed to formaldehyde when
there is no control for other exposures (including wood dust). This suggests that wood dust is
not necessary for formaldehyde to exert an effect. At most, the data suggest that wood dust
might be a HELPER or MIXER with formaldehyde, or a partial carcinogen itself; the same data
cannot be used to determine WHICH of these hypotheses concerning wood dust is correct, nor
can the reasoning be reversed to make statements about whether FORMALDEHYDE is a
helper, mixer or partial carcinogen (which is more to the point here).
4.3.4.2.3. Mixer
The judgment in all cells is NO since all studies involved only formaldehyde or
application of formaldehyde with wood dust. Such studies cannot address whether the
interaction between formaldehyde and wood dust is in the form of a partial carcinogen, a mixer
or a helper. The present studies could be taken to indicate weakly (with LOW epistemic status)
that formaldehyde is NOT a complete carcinogen, since study 2 suggests that exposure to wood
dust might be required. Study 2, however, also shows an increase in cancer incidence in groups
exposed to formaldehyde when there is no control for factors other than formaldehyde
(including wood dust). This suggests that wood dust AMPLIFIES the carcinogenic action of
formaldehyde, but that exposure to wood dust is not necessary for formaldehyde to increase
cancer incidence. At most, the data suggest that wood dust might be a HELPER or MIXER
with formaldehyde, or a partial carcinogen; the same data cannot be used to determine WHICH
of these hypotheses concerning wood dust is correct, nor can the reasoning be reversed to make
claims about whether FORMALDEHYDE is a helper, mixer or partial carcinogen.
119
-------
4.3.4.2.4. Helper
The judgment in all cells is NO since all studies involved only formaldehyde or
application of formaldehyde with wood dust. Such studies cannot address whether the
interaction between formaldehyde and wood dust is in the form of a partial carcinogen, a mixer
or a helper. The present studies could be taken to indicate weakly (with LOW epistemic status)
that formaldehyde is NOT a complete carcinogen, since study 2 suggests that exposure to wood
dust might be required. Study 2, however, also shows an increase in cancer incidence in groups
exposed to formaldehyde when there is no control for factors other than formaldehyde
(including wood dust). This suggests that wood dust AMPLIFIES the carcinogenic action of
formaldehyde, but that exposure to wood dust is not necessary for formaldehyde to increase
cancer incidence. At most, the data suggest that wood dust might be a HELPER or MIXER
with formaldehyde, or a partial carcinogen; the same data cannot be used to determine WHICH
of these hypotheses concerning wood dust is correct, nor can the reasoning be reversed to make
claims about whether FORMALDEHYDE is a helper, mixer or partial carcinogen.
4.3.4.3. Stages
4.3.4.3.1. Neoplastic Conversion
The judgment in all cells is NO since no data were obtained pertinent to the question of
the stage at which formaldehyde acts.
4.3.4.3.2. Neoplastic Development
The judgment in all cells is NO since no data were obtained
pertinent to the question of the stage at which formaldehyde acts.
120
-------
4.3.4.4. Mechanisms
4.3.4.4.1. Genotoxic
The judgment in all cells is NO since no data were obtained pertinent to the question of
the stage at which formaldehyde acts.
4.3.4.4.2. Non-Genotoxic
The judgment in all cells is NO since no data were obtained pertinent to the question of
the stage at which formaldehyde acts.
43.5. Summary Comments for Intra-Context Claims
The entries into the cells of Working Table 3.11 indicate that formaldehyde should be
weakly considered a carcinogen in humans when exposure is through inhalation at
concentrations at or above 2 ppm. This judgment is reflected most clearly in the fact that the
"Overall Summary" cell for "Increases Incidence of Cancer" is judged to be LOW due to the lack
of statistical significance for increased incidence of cancer in the exposed groups. This judgment
is not offset by the fact that the two studies were coherent, each leading to SMRs in excess of
1.0 in some exposure groups, since these studies were characterized by relatively poor Strength
of Effect. In addition, the inability to exclude the hypothesis that wood dust was the primary
carcinogen present weakens the epistemic status of the claims for formaldehyde. The data do
not provide information on any of the other Claims of Carcinogenicity.
4.4. Description or Working Table 4.11
4.4.1. Introduction
121
-------
Working Table 4.11 is utilized to justify premises necessary for extrapolation of the
findings in Working Table 3.11 (the "Observational Context") to the "Target Context" (Context
12). Four separate claims must be warranted, each of which is described in detail in the original
report and in lesser detail below. The data on which judgments for these premises are based
were described in Section 4.3.4. All judgments of Intellectual Obligation are as described in
Section 1.4.
4.4.2. Column Headings in Working Table 4.11
These extrapolation premises or claims as they appear in Working Table 4.11 are:
1. EXPOSURE TO BSDR: Any differences between the BIOLOGICAL components of
the two contexts with respect to the relationship between exposure and BSDR are not so large
as to call into question whether a BSDR might be present in one context but not the other AT
THE SAME LEVEL OF EXPOSURE.
The pertinent considerations here for each context are whether (1) an exposure occurs
through some compartment of the environment (a compartment refers to the medium of the
environment, such as air or water); (2) an intake by some route can occur (a route refers to
inhalation, ingestion, dermal absorption, intubation or injection); (3) this intake results in an
uptake to a target organ or tissue; (4) clearance from the target allows the production of a
burden; and (5) biological transformation to the active metabolite exists. Where extrapolation is
across exposure levels, it is particularly important to establish that there does not exist a
threshold exposure below which a BSDR is not produced (due, e.g., to necessary saturation of
metabolic transformation). Where extrapolation is across species, it is particularly important to
show that any such threshold, if it exists, does not differ between the species in a manner calling
into question this premise.
122
-------
WORKING TABLE 4.11. SUPPORT FOR INTER-CONTEXT EXTRAPOLATION PREMISES
CONTEXT NO. 11 TO CONTEXT NO. 12
Relevance Strategy
I.O.1
Exposure to BSDR
Conversion
BSDR to Effect
Conversion
Host Factors
Environmental
Conditions
Direct Empirical (D.E.)
HI
NO
NO
HI
LO
Semi-Empirical Extrapolation
(S.E.E.)
ME
NO
NO
NO
NO
Empirical Correlation (E.C.)
ME
ME
ME
HI
LO
Theory-based Inference (T.B.I.)
ME
ME
ME
HI
LO
Existential Insight (E.I.)
LO
ME
HI
HI
LO
Overall Assessment
ME
ME
HI
LO
'Intellectual Obligation
-------
2. BSDR TO BIOPHYSICAL EFFECT: Any differences between the BIOLOGICAL
components of the two contexts with respect to the relationship between BSDR and the
production of biophysical effects necessary for transitions are not so large as to call into
question whether a biophysical effect might be present in one context but not the other
ASSUMING THE SAME LEVEL OF BSDR IN BOTH CONTEXTS.
The pertinent considerations here for each context are whether (1) the active metabolite
can interact with causally important biological structures in the organism; (2) this interaction
results in either DNA alterations (for initiation) or stimulation of cellular division (for
promotion); and (3) these biophysical effects are of the type and magnitude necessary to induce
transitions designated as either neoplastic development or neoplastic conversion. Where
extrapolation is across exposure levels, it is particularly important to establish that there does
not exist a threshold BSDR below which a biophysical effect is not produced (due, e.g., to
necessary saturation of a pool of DNA damage or a stimulus to cellular proliferation). Where
extrapolation is across species, it is particularly important to show that any such threshold, if it
exists, does not differ between the species in a manner calling into question this premise.
3. ENVIRONMENTAL CONDITIONS: Any differences between the
ENVIRONMENTAL components of the two contexts with respect to the relationship between
exposure (to the substance of interest) and BSDR, or the relationship between BSDR and the
production of biophysical effects necessary for transitions, are not so large as to call into
question whether a biophysical effect might be present in one context but not the other AT THE
SAME LEVEL OF EXPOSURE TO THE SUBSTANCE OF INTEREST.
The pertinent considerations here for each context are whether (1) environmental
conditions increase or decrease the BSDR resulting from a given level of exposure to the
substance of interest, relative to the conditions that exist in the target context; (2) environmental
123
-------
conditions increase or decrease the probability of biophysical effects from a given level of
BSDR, relative to the target context; and (3) environmental conditions increase or decrease the
probability of transitions from a given level of biophysical effect, relative to the target context.
4. HOST FACTORS: Any differences between the BIOLOGICAL components of the
two contexts with respect to the relationship between the production of biophysical effects, the
probability of transitions necessary for cancer, and the probability of cancer itself are not so
large as to call into question whether an increase in cancer incidence might be present in one
context but not the other AT THE SAME LEVEL OF BIOPHYSICAL EFFECT.
The pertinent considerations here for each context are whether (1) a target organ or
tissue exists; (2) biological structures on which relevant biophysical effects are produced exist;
(3) relevant background transitions exist; and (4) repair of biophysical effects, or regression of
states of neoplastic conversion or neoplastic development, exist. Where extrapolation is across
species, it is particularly important to establish that there does not exist a threshold level of
biophysical effect below which a transition is not produced (due, e.g., to necessary saturation of
a stimulus to cellular proliferation before promotion can occur).
In the original report, Working Table 4 (bearing in mind that Working Table 4 in the
present report was numbered Working Table 5 in the original report, as described in Section 1)
also contained an extrapolation premise concerning the role of intersubject and intrasubject
variability. This premise has been removed here and consideration of the role of differences in
intra- and intersubject variability between the two contexts has been incorporated into the 4
separate premises above. This choice was made because variability always refers to variability of
some PROPERTY, and this property would appear in one of the four extrapolation categories
listed above. The analyst then must determine the significance of variability on the relationship
124
-------
between EXPOSURE AND BSDR, BSDR AND BIOPHYSICAL EFFECT,
ENVIRONMENTAL CONDITIONS, AND HOST FACTORS separately.
4.4 J. Epistemic Status of Extrapolation Premises
This section summaries the judgments made for each of the cells in Working Table 4.11.
The procedure for assigning epistemic status to a cell is identical to that outlined for Working
Table 3.11 and a description is not repeated here.
4.4.3.1. Exposure to BSDR
A. Direct Empirical: The judgment is NO since BSDR was not determined in either
context at any level of exposure.
B. Semi-Empirical Extrapolation: The judgment is NO since BSDR was not determined
in either context at any level of exposure.
C. Empirical Correlation: The judgment is MED since exposure to formaldehyde (which
is well established in both contexts) is taken to correlate with moderate strength with production
of a dose, but metabolic transformation has not been identified in either context. Where the
active metabolite has not been identified, the correlation between exposure and BSDR (rather
than dose) will be weakened. The judgment is not HIGH since it has not been demonstrated
that there is not a threshold exposure necessary for production of a BSDR, or that this
threshold is the same between contexts.
D. Theory-Based Inference: The judgment is MED since exposure clearly occurs in both
contexts; intake clearly occurs in both cases, although at a lower level in the target context than
in the observational context; the absorption of formaldehyde in nasal passages upon inhalation
in humans has been established with MED epistemic status regardless of level of exposure;
125
-------
clearance is present but not infinitely rapid in both contexts; but the active metabolite has not
been shown present in either context. While the theory of pharmacodynamics suggests that a
dose should be produced in both contexts, it is not clear that a BSDR will be produced since it
is possible that a mechanism of metabolic transformation is stimulated or saturated at one level
of exposure but not the other. The judgment is not HIGH since it has not been demonstrated
that there is not a threshold exposure necessary for production of a BSDR, or that this
threshold is the same between contexts.
E. Existential Insight: The judgment is MED since these data produce in the analysts
moderate confidence that a BSDR is produced in both contexts.
F. Overall Summary: The judgment is MED since only Existential Insight and Theory-
Based Inference were of MED strength, but neither of these has HIGH Intellectual Obligation.
The judgment is not LOW since the lack of identification of the active metabolite is not as
troubling theoretically as it would be in extrapolating across species.
4.4.3.2. BSDR to Effect
A. Direct Empirical: The judgment is NO since the biophysical effect of immune
response was measured in one context, but no biophysical effect was measured in the other
context.
B. Semi-Empirical Extrapolation: The judgment is NO since the biophysical effect of
immune response was measured in one context, but no biophysical effect was measured in the
other context.
C. Empirical Correlation: The judgment is MED since the ability of the active metabolite
of a substance to induce biophysical effects UPON PRODUCTION OF A BSDR in one human
population (Context 11) is taken to correlate strongly with the same feature in an essentially
126
-------
identical second population (Context 12). The judgment is not HIGH since it has not been
demonstrated that there is not a threshold BSDR necessary for production of a biophysical
effect, or that this threshold is the same between contexts.
D. Theory-Based Inference: The judgment is MED since similar cellular structures exist
in both contexts; formaldehyde or its active metabolite is assumed to diffuse in a similar fashion
in both contexts; and interaction with pertinent biological structures should occur through this
diffusion. The judgment is not HIGH since it has not been demonstrated that there is not a
threshold BSDR necessary for production of a biophysical effect, or that this threshold is the
same between contexts.
E. Existential Insight: The judgment is HIGH since these data produce in the analysts
strong confidence that interaction between formaldehyde (or its active metabolite) and pertinent
biological structures is produced in both contexts.
F. Overall Summary: The judgment is MED since Empirical Correlation, Existential
Insight and Theory-Based Inference are either MED or HIGH, while the Intellectual Obligation
for all of these is either MED or LOW. The lack of Direct Empirical support prevents this
judgment from being HIGH.
4.4.3.3. Host Factors
A. Direct Empirical: The judgment is HIGH since hosts are observed to be essentially
identical in the two contexts.
B. Semi-Empirical Extrapolation: The judgment is NO since the consideration of host
factors does not require extrapolation across exposure levels, and was dealt with in A above.
C. Empirical Correlation: The judgment is HIGH the presence of appropriate host
factors in one randomly selected human population is taken to correlate strongly with the same
127
-------
factors in a second randomly chosen human population. The threshold level of biophysical effect
for transitions should be the same in the two populations.
D. Theory-Based Inference: The judgment is HIGH based on the fact that cellular
structure and function is similar in the two contexts since both involve humans, so similar targets
should exist; the target in Context 12 is taken to be the nasal passages, which clearly exist in
both contexts; repair should be present but not completely effective in both contexts; and
background transitions are taken to occur in both contexts as evidenced by the non-zero rates of
effect in suitable controls. The threshold level of biophysical effect for transitions should be the
same in the two populations.
E. Existential Insight: The judgment is HIGH since the biological similarities between
the two contexts produce in the analysts strong confidence that the two contexts are similar.
F. Overall Summary: The judgment is HIGH all judgments above were HIGH or NO,
and there was HIGH epistemic status for the Direct Empirical warrant.
4.4.3.4. Environmental Conditions
A. Direct Empirical: The judgment is LOW since the exposure was clearly to
formaldehyde in both contexts; the form was as formaldehyde vapor but with the possible
presence of formaldehyde liquid exposures in some subset of exposures of the observational
context; and concurrent exposures to other substances were possible in the observational context
but not well confirmed.
B. Semi-Empirical Extrapolation: The judgment is NO since this warrant does not apply
for this extrapolation premise.
128
-------
C. Empirical Correlation: The judgment is LOW since concurrent environmental
exposures in one set of occupations does not generally correlate well with concurrent
environmental exposures in a second set of occupations.
D. Theory-Based Inference: The judgment is LOW based on the belief that two sets of
occupations may involve very different mixtures of substances and procedures, as well as
different physical states of formaldehyde.
E. Existential Insight: The judgment is LOW since the environmental differences
between the two contexts produce in the analysts moderate confidence that the two contexts are
similar.
F. Overall Summary: The judgment is LOW since all judgments above were either LOW
or NO.
4.5. Description of Working Table 5.11
4.5.1. Introduction
Working Table 5.11 contains judgments of the epistemic status for Inter-Context Claims
of Carcinogenicity. In each entry, the "Observational Context" is Context 11 and the "Target
Context" is Context 12. The judgments in Working Table 5.11 utilize the judgments in Working
Table 3.11 for a particular Claim of Carcinogenicity, as well as the judgments in Working Table
4.11 for the necessary Extrapolation Premises. In each case of an extrapolation, it is the "Overall
Summary" judgment associated with a Claim of Carcinogenicity in Working Table 3.11 that
forms the basis for the extrapolation. The final judgment of epistemic status in Working Table
5.11 cannot be higher than this initial judgment from Working Table 3.11, since the application
of the extrapolation premises can have only the effect of weakening the extrapolation. In other
words, if formaldehyde is judged to "Increase the Incidence of Cancer" with MEDIUM epistemic
129
-------
WORKING TABLE 5.11. INTER-CONTEXT SUPPORT FOR CLAIMS OF CARCINOGENICITY
EXTRAPOLATED FROM CONTEXT II TO CONTEXT 12
Claims of Carcinogenicity
Relevance Strategy
I.O.1
Increases
Incidence of
Cancer
Classification(s)
Stage
Mechanism
Complete
Partial
Mixer
Helper
Neo.
Conv.
Neo.
Devel.
Geno-
tox.
Non-
genot
ox.
Direct Empirical (D.E.)
HI
NO
NO
NO
NO
NO
NO
NO
NO
NO
Semi-Empirical
Extrapolation (S.E.E.)
ME
LO
NO
NO
NO
NO
NO
NO
NO
NO
Empirical Correlation
(EC.)
ME
LO
NO
NO
NO
NO
NO
NO
NO
NO
Theory-based Inference
CT.B.I.)
ME
LO
NO
NO
NO
NO
NO
NO
NO
NO
Existential Insight(E.I.)
LO
ME
NO
NO
NO
NO
NO
NO
NO
NO
Column Summary
LO
NO
NO
NO
NO
NO
NO
NO
NO
Overall Summary
LO
NO
NO
NO
NO
NO
NO
NO
NO
'Intellectual Obligation
-------
status in Working Table 3.11, the highest judgment of epistemic status for the instance of
extrapolation will be MEDIUM, which presumes that the extrapolation premises are judged to
each have HIGH epistemic status in Working Table 4.11.
The analysts have chosen to view warrants from Working Table 3.11 in the following
manner. A particular cell from that table may be used as Empirical Correlation, Theory-Based
Inference and/or Existential Insight relevance strategies in Working Table 5.11 for the same
Claim of Carcinogenicity. A particular claim from Working Table 3.11 may NOT be used as a
Direct Empirical relevance strategy in Working Table 5.11, since there always exists
extrapolation across exposure levels. A particular claim from Working Table 3.11 MAY be used
as a Semi-Empirical Extrapolation relevance strategy if it is judged that the biological and
environmental conditions in the two contexts are sufficiently similar to warrant the claim that
these two contexts contain approximately the same biological and environmental antecedent
conditions, differing only in LEVEL of exposure to formaldehyde. This clearly requires relatively
strong epistemic status to each of the four extrapolation premises found in Working Table 4.11,
a condition only weakly met in the present case.
4.5.2. Intellectual Obligation
The assignments of Intellectual Obligation discussed in Section 1.4, and utilized in
Working Table 3.11, are employed in Working Table 5.11.
4.5.3. Claims of Carcinogenicity
The same grouping of Claims of Carcinogenicity used as headings in Working Table 3.11
are employed in Working Table 5.11. The judgments of the epistemic status of each of these
130
-------
claims, for each relevance strategy, are provided in the cells of Working Table 5.11 and are
described below.
4.5.3.1. Increases Incidence of Cancer
A. Direct Empirical: The judgment is NO due to differences in level of exposure and
biological context.
B. Semi-Empirical Extrapolation: The judgment is LOW due to LOW epistemic status of
claim in Working Table 3 which is not offset by MED and HIGH strength of extrapolation
premises.
C. Empirical Correlation: The judgment is LOW since the claim was given LOW
epistemic status in Working Table 3.11 and the "Overall Summaries" of extrapolation premises
were all either MED or HIGH in Working Table 4.11. While a correlation does exist for
carcinogenicity between two human contexts, the LOW epistemic status of the claim in Working
Table 3.11 keeps this judgment from being MED or HIGH.
D. Theorv-Based Inference: The judgment is LOW since the claim was given LOW
epistemic status in Working Table 3.11 and the "Overall Summaries" of extrapolation premises
were all either MED or HIGH in Working Table 4.11. While the Theory of Carcinogenesis does
support the claim that carcinogenicity at high levels may indicate carcinogenicity at lower levels
of exposure (particularly for genotoxic mechanisms), the LOW epistemic status of the claim in
Working Table 3.11 keeps this judgment from being MED. The lack of evidence on Genotoxicity
in Context 11 also keeps this epistemic status from being elevated to MED.
E. Existential Insight: The judgment is MED since the evidence produces in the analysts
moderate confidence that formaldehyde is a carcinogen in Context 12.
131
-------
F. Column Summary: The judgment is LOW since Direct Empirical was NO (with HIGH
Intellectual Obligation), Semi-Empirical Extrapolation was LOW (with MED Intellectual
Obligation), and the remainder were either LOW (with MED Intellectual Obligation) or MED
(with LOW Intellectual Obligation).
G. Overall Summary: The judgment is LOW since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
4.5.3.2. Classifications
4.5.3.2.1. Complete Carcinogen
The judgment in all cells is NO since the same judgment is given in Working Table 3.11
(for the reasons stated there).
4.5.3.2.2. Partial Carcinogen
The judgment in all cells is NO since the same judgment is given in Working Table 3.11
(for the reasons stated there).
4.5.3.2.3. Mixer
The judgment in all cells is NO since the same judgment is given in Working Table 3.11
(for the reasons stated there).
4.5.3.2.4. Helper
The judgment in all cells is NO since the same judgment is given in Working Table 3.11
(for the reasons stated there).
132
-------
4.5.3.3. Stages
4.5.3.3.1. Neoplastic Conversion
The judgment in all cells is NO since the same judgment is given in Working Table 3.11
(for the reasons stated there).
4.5.3.3.2. Neoplastic Development
The judgment in all cells is NO since the same judgment is given in Working Table 3.11
(for the reasons stated there).
4.5.3.4. Mechanisms
4.5.3.4.1. Genotoxic
The judgment in all cells is NO since the same judgment is given in Working Table 3.11
(for the reasons stated there).
4.5.3.4.2. Non-Genotoxic
The judgment in all cells is NO since the same judgment is given in Working Table 3.11
(for the reasons stated there).
5. OBSERVATIONAL CONTEXT NO. 12
5.1. Description of Working Table 1.12
Working Table 1 for Context 12 is provided below. As in all subsequent sections of this
report, a Working Table is identified by both the general Working Table number (here, 1) and
the context (here, 12). As a result, this Working Table is assigned the number 1.12. The
133
-------
WORKING TABLE 1.12. STUDIES AVAILABLE FOR OBSERVATIONAL CONTEXT NUMBER 12: HUMANS
EXPOSED TO AIRBORNE FORMALDEHYDE AT CONCENTRATIONS BELOW 2 PPM
Study
Number1
Data Category
Description
1
BE
Primary Reference #7 in Table 1
31 female and 24 male human subjects were grouped into cohorts exposed
to form aldehyde vapor in various occupations characterized by exposure
to form aldehyde in tobacco smoke and other unspecified routes. Subjects
then were evaluated for antibody reaction (serum IgE or IgG) and for
whether the presence of antibodies correlated with a history of respiratory
and conjunctival symptoms. The low level exposure for histology
technicians was 0.64 ppm; for pathology residents from 0.2 to 0.64 ppm;
and for medical residents was unreported. Clinical assessments of
antibody formation were made by EL1SA detection and by skin
challenges, followed by subjective expert judgment of the presence of a
response.
'The study number refers to the order of the study within this context. The primary reference number, indexed to Table 1, is provided in the "Decsription"
column.
-------
biological components of this context are given by the fact that all exposed organisms were adult
humans in a state of health sufficient for employment.
The exposures in this context were all to formaldehyde, with at least a potential for
exposures in air at concentrations above 2 ppm, placing them into the category of high
exposures as defined in the Introduction. Length of exposure varied between the populations
studied, but this was not taken to justify development of separate contexts. It was assumed that
mechanisms of action of formaldehyde might be slightly different depending upon length of
exposure, but that this difference would appear in the magnitude of effect rather than the
presence and direction of that effect. This choice is somewhat tenuous since there is evidence
that effects such as promotion may require some minimal period of exposure.
One body of data was identified for this context. This is described in Working Table
1.12., and is placed into the data category of
(a) Biophysical Effect (taken from Reference #7 in Table 1 of this report). Here, the
sole Biophysical Effect is taken to be allergic response, which is considered a Biophysical Effect
because it is related to immune response which, in turn, is related to transitions necessary for
cancer. It is given the label BE1.
Data in other data categories were not available for the studies utilized in this context.
52. Description of Working Table 2.12
The judgments for this Working Table are provided in the cells of Working Table 2.12.
The text here describes the reasons for these judgments. The data on which these judgments are
based are provided as various "Exhibits" after Working Table 2.12., with a numbering system
described in Section 5.2.4.
134
-------
WORKING TABLE 2.12. DATA CHARACTERISTICS FOR TARGET CONTEXT NUMBER 12
Data Category/Item
Description
Completeness
(Hi/Me/Lo/No)
Utility
(Hi/Me/Lo/No)
Strength of Effect
(Hi/Me/Lo/No)
Exposure-
Specific
Effect1
Tumor Response
No study available
Biophysical Effect
Allergic response (BE1.1)
HI
HI
NO
WT2.C12.
BE1.1
Pharmacodynamics
No study available
Host Factors
No study available
Concurrent Environmental
Conditions
No study available
Related Substances Assessment
No study available
'Refers to the Exhibit Number for the data as described in the text.
-------
WT2.C12.BE1.1
TABLE II. Symptom history, antibody titers, skin testing, and final assessment
Subjects
Symptoms
with F
P1
Physician assessed history
P2
Histology technicians
1
2
3
4
5
6
7
8
9+
10
Medical residents
11
12
13
14
15
16
17
18
19
20
Pathology residents
21
22
23
24
25
26
27
28
29
ENL
EN
NL
NLS
ELS
ENLS
ENLS
ENL
ENL
EN
EN
E
E
EN
EN
N
EN
ENL
E
E
E
E
I. PA
I
I. PA
I. PA
I. PA
I. PA
I
I
N
N
N
N
N
N
I
I
N
I
I
I
I
I. PA
I. PA
I. PA
I
I
I
I
I
I, PA
I
I, PA
I. PA
I, PA
I. PA
I
I
N
N
N
N
N
N
I
I
N
I
I
I
I
I, PA
I
I
I --
I, PA
I
I
I
I. Pa
i
i. pa
i. pa
I, PA
I, PA
I
I
N
N
N
N
N
N
I
I
N
I
I
I
I
I. PA
I
I
I
I, PA
£. eye; /V. nose; L. lung; 5, skin; PI. physician 1; P2, physician 2; C. conference opinion; /. irritant; PA, possible allergic; no symptom
•When a second antibody titer is listed, this was of sera drawn I year after original study sample,
t Subject No. 9 had negative respiratory challenges with F up to 2 ppm.
-------
52.1. Data Categories and Items
The same data category and item described with respect to Working Table 1.12 is
employed in Working Table 2.12. This is:
(a) Biophysical Effect, containing BE1 (allergic response).
This sole data category is displayed in Working Table 2.12.
522. Judgments of Completeness
In all cases, the judgment of completeness was based on a comparison of the number of
studies used in a data category/item for this context against the number available in the
complete reference list provided in Appendix A. The judgment was HIGH if the data used in
this Working Table was a relatively complete subset of the data indicated in the Appendix,
where "complete" was taken to mean somewhere on the order of 70% or above. The judgment
was MEDIUM if the data used in this Working Table constituted somewhere between 25 and
70% of the total body of potentially available data. The judgment was LOW if the data used
constituted less than 25% of the data available. These assignments also included limited
consideration of the QUALITY of the data, since it should not necessarily count against an
analysis if the analyst fails to collect data of poor quality. This consideration also is dealt with in
the assignment of Utility for the collected studies. Still, it is evident that the judgment of
Completeness might reasonably be expected to include consideration of something more than
the fraction of studies collected in a given data category/item. In the present analysis,
Completeness was increased slightly if the collected data for a given data category were the most
heavily cited in the secondary literature on formaldehyde. This leads to the following judgments
of Completeness:
135
-------
For the data category BE1, the judgment of Completeness is HIGH since this was the
only study identified that measured this response.
5.23. Judgments of Utility, Strength of Effect, and Exposure- Specific Effect
The remaining judgments in Working Table 2.12 refer to the qualities of the data
ACTUALLY EXAMINED IN THIS EXAMPLE. By Utility, we mean the quality of the study
DESIGN and its ability to address in rigorous fashion questions pertinent to hazard
identification for formaldehyde. No attempt is made to determine the specific inferences
towards which the data will be directed in later Working Tables, since the intent here is only to
determine the degree to which a body of data arose from a study design appropriate for
accurate and precise measurements of the quantity considered in the study. Utility is increased
by high standards of study design, including the ability of the experiment to demonstrate a
causal link between exposure to formaldehyde and the measured effect.
By Strength of Effect, we mean the degree to which the study actually revealed an effect.
Strength is increased when the effect noted possessed statistical significance.
By Exposure-Specific Effect, we mean a summary of the actual effect observed (putting
aside questions of the reliability of this effect, which was dealt with in the assignment of Utility
and Strength of Effect). This summarization is provided in the following text, and the data
themselves are provided in the various Exhibits appearing after Working Table 2.5.
The following judgments then apply to Working Table 2.12:
For the data category BE1, the Utility judgment is HIGH for the issue of whether
formaldehyde induces allergic response, since measurement techniques are well developed and
properly applied; sample size is adequate; measured response is linked clearly to allergic
response; formaldehyde concentrations were quantified through measurement; appropriate
136
-------
controls were utilized; the study does not constitute a deliberate experiment (important for
causal claims, which might be taken reasonably to weaken the judgment here to MEDIUM).
Judgment is MEDIUM for the issue of whether formaldehyde induces a biophysical effect
relevant to cancer, since allergic response is not clearly established as being related to cancer.
The Strength of Effect judgment is NO since no statistically significant difference in allergic
response between exposed groups was found. The Exposure-Specific Effect is summarized in
Exhibit WT2.C12.BE1.1. No consistent allergic response was found.
5.2.4. Exhibits for Working Table 212
The data for the various data categories are provided as "Exhibits" after Working Table
2.12. The numbering system for the exhibits was chosen to display information on (in order of
appearance) the Working Table; the Context; the specific effect measured; and the study
number within that specific effect. For example, Exhibit WT2.C12.BE1.1 refers to Working
Table 2 (all exhibits have this assignment in common); Context 12; Biophysical Effect of type 1;
and data set #1 within the data category BE1. This same numbering scheme will be employed in
all other Contexts.
53. Description of Working Table 3.12
5.3.1. Introduction
The central task of Working Table 3.12 is to draw inferences on the carcinogenicity of
formaldehyde WITHIN CONTEXT 12, employing data specific to that context. This task
requires that several judgments be made sequentially:
(a) The analyst must determine the particular Claims of Carcinogenicity to be considered
in the analysis. These various claims are discussed in detail in Section 5.3.4.
137
-------
WORKING TABLE 3.12. INTRA-CONTEXT SUPPORT FOR CLAIMS OF CARCINOGENICITY IN CONTEXT 12
Claims of Carcinogenicity
Relevance Strategy
I.O.'
Increases
Incidence of
Cancer
Classification(s)
Stage
Mechanism
Complete
Partial
Mixer
Helper
Neo.
Conv.
Neo.
Devel.
Geno-
tox.
Non-
genot
ox.
Direct Empirical (D.E.)
HI
NO
NO
NO
NO
NO
NO
NO
NO
NO
Semi-Empirical
Extrapolation (S.E.E.)
ME
NO
NO
NO
NO
NO
NO
NO
NO
NO
Empirical Correlation
(E.C.)
ME
NO
NO
NO
NO
NO
NO
NO
NO
NO
Theory-based Inference
(T.B.I.)
ME
NO
NO
NO
NO
NO
NO
NO
NO
NO
Existential Insight(E.I.)
LO
NO
NO
NO
NO
NO
NO
NO
NO
NO
Column Summary
NO
NO
NO
NO
NO
NO
NO
NO
NO
Overall Summary
NO
NO
NO
NO
NO
NO
NO
NO
NO
'Intellectual Obligation
-------
(b) The available Relevance Strategies by which a given body of data may be related to a
specific Claim of Carcinogenicity must be developed. These strategies were discussed in
detail in Section 1.4., and the same definitions are applied here.
(c) The analyst must assign a measure of Intellectual Obligation to each Relevance
Strategy. This assignment was described in detail in Section 1.4., and the same
assignments are made here.
(d) The analyst must judge the epistemic status of a specific Claim of Carcinogenicity
arising from a specific Relevance Strategy. As described in Section 1.4, this judgment is
separate from the judgment of Intellectual Obligation for that Relevance Strategy. This
judgment is described in more detail in Section 5.3.2. The assignments are entered into
the appropriate cells of Working Table 3.12.
(e) The analyst must produce a summary judgment of the epistemic status of a particular
Claim of Carcinogenicity by combining the judgments from each of the five Relevance
Strategies for that claim. This summary includes consideration of both the epistemic
status of that claim for each Relevance Strategy, as entered into the separate cells of the
Working Table specific to each strategy, as well as the Intellectual Obligation assigned to
that strategy. The assignments are entered into the appropriate COLUMN SUMMARY
cells of Working Table 3.12.
(f) The analyst must produce a summary judgment of the COHERENCE of the various
Claims of Carcinogenicity. This task is performed by examining the coherence ACROSS
the various COLUMN SUMMARIES appearing in the Working Table. The assignments
are entered into the appropriate OVERALL SUMMARY cells of Working Table 3.12.
532. Epistemic Status of Claims of Carcinogenicity
The entries into the various cells of Working Table 3.12 reflect the judgment of the
analyst concerning the epistemic status of the Claim of Carcinogenicity being considered, FOR
A SPECIFIC RELEVANCE STRATEGY. This judgment will have two components:
(a) Each Relevance Strategy calls upon a selected body of data from which the reasoning
will proceed. Specifically, the Direct Empirical strategy requires data on direct measurements of
the effect underlying the Claim of Carcinogenicity, made at the exposure to formaldehyde
defining the context (here, concentrations at or below 2 ppm). The Semi-Empirical
Extrapolation strategy requires data on direct measurements of the effect underlying the Claim
of Carcinogenicity, made at exposures to formaldehyde other than those defining the context.
138
-------
The Empirical Correlation strategy requires data on effects other than the effect underlying the
Claim of Carcinogenicity, where this "other" effect is taken to correlate with appearance of the
effect of interest. The Theory-Based Inference strategy requires data on an effect other than the
effect underlying the Claim of Carcinogenicity, where this "other" effect is taken to be a causal
step in the production of the effect of interest as specified by an established theory. The
Existential Insight strategy does not invoke data, except in the sense that a subjective judgment
from an individual might be counted as an instance of data in and of itself.
The analyst must assign the "foundational quality" to the data used in each relevance
strategy. By "foundational quality", we mean the quality of the data with respect to the question
those data were intended to address BY THE ORIGINAL INVESTIGATOR. This judgment is
based on the results of Working Table 2.12 and includes consideration of the Completeness,
Utility, Strength of Effect and Exposure-Specific Effect from that table. While the judgment
ultimately is subjective, it is constrained in that it must increase (from NO to LOW to
MEDIUM to HIGH) as the strength of the judgments for Completeness, Utility and Strength of
Effect increase in Working Table 2.12. The summary of Exposure-Specific Effect may also enter
into the assignment of epistemic status to a Claim of Carcinogenicity if the analyst determines
that only effects above a given magnitude will exert an influence on carcinogenic processes (an
issue of importance when possible thresholds are considered).
(b) The analyst then must judge the strength of any "background premises" required by
the Relevance Strategy under consideration. These background premises are defined in detail in
the original report, but may be summarized as the set of (often implicit) premises that must be
introduced into an analysis if the data in Working Table 2.12 are to be used to draw inferences
about the Claims of Carcinogenicity for the context being considered. Presumably, many of these
background premises already are considered valid once Working Tables 1.12 and 2.12 are
139
-------
complete; otherwise, the data would not have been assigned to this context. But other
background premises concerning the reliability of extrapolation procedures (for the Semi-
Empirical Extrapolation relevance strategy), the strength and specificity of correlations (for the
Empirical Correlation relevance strategy), the degree of verification and validation of theories
(for the Theory-Based Inference relevance strategy), and the credentials of individuals making
subjective judgments (for the Existential Insight relevance strategy) must be warranted prior to
their use.
Having established these two components of epistemic status (foundational quality and
background premises) for a given Claim of Carcinogenicity, the analyst then produces a
composite judgment of the epistemic status of a given Claim of Carcinogenicity and enters this
into the appropriate cell of Working Table 3.12. There are two separate claims that might be
made within one of these cells. The first is the claim that formaldehyde DOES produce the
effect underlying that claim (such as increased incidence of cancer or genotoxic effects). In that
case, the assignments of NO/LOW/MED/HIGH refer to the epistemic status of the claim that
formaldehyde produces the effect. This assignment is entered into the area ABOVE the dashed
line in a cell. The second claim is that formaldehyde DOES NOT produce the effect underlying
the claim. In that case, the assignment of NO/LOW/MED/HIGH refer to the epistemic status
of the claim that formaldehyde DOES NOT produce the effect. This assignment is entered into
the area BELOW the dashed line in a cell. THE READER SHOULD NOTE THAT THIS IS A
NEW FEATURE OF THESE WORKING TABLES NOT DISCUSSED IN THE ORIGINAL
REPORT. The reason for this addition to the Working Tables is that the fact that a particular
body of data does not support the claim that formaldehyde does induce a specific effect IS NOT
NECESSARILY evidence that formaldehyde does not induce that effect. It may simply be the
140
-------
case that the data were obtained under conditions incapable of revealing any effect that might
be present.
This consideration is important due to the issue of "burden of proof. At times, the
burden may lie on the regulatory agency to show that formaldehyde DOES induce cancer.
Attention then would be directed towards the judgments in the upper halves of the various cells
of Working Table 3.12. At other times, the burden may lie on the producer of formaldehyde to
show that formaldehyde DOES NOT induce cancer. Attention then would be directed towards
the judgments in the lower halves of the various cells of Working Table 3.12. IN THE
PRESENT EXAMPLE, ONLY JUDGMENTS THAT AN EFFECT IS PRODUCED BY
FORMALDEHYDE ARE DEVELOPED. THE PROCESS OF ANALYSIS WOULD BE
REPEATED FOR THE LOWER HALVES OF ALL CELLS IF THE BURDEN OF PROOF
WAS SHIFTED.
It is important here to consider the broad features of the theory of carcinogenesis to be
used throughout this section. These features structure the interpretation of data used in the
various Relevance Strategies. The theory of carcinogenesis employed here is that cells are
transformed to cancer by three steps or transitions. These are termed initiation, promotion and
progression, each of which may require one or more sub-steps. For each transition to occur, a
biologically-significant dose-rate must reach the targets in appropriate cells so that an interaction
between the substance or a metabolite may take place. This target may or may not have been
identified for a particular substance and tumor type.
The initiation step is taken to involve alteration of DNA structure and function,
producing a new genome. This alteration must then remain intact without being repaired by
processes normally operating in cells or induced by the damage itself. The ability of a substance
to bring a stable alteration about may be determined from an increase in cancer incidence
141
-------
following (1) application of the initiating agent and (2) subsequent application of a promoting
agent such as TPA. In the absence of such data, the ability of a substance to interact with DNA,
to alter DNA, and/or to induce mutation provides partial but not conclusive evidence of
initiating activity. In addition, the ability of a substance to lower DNA repair rates or fidelity of
repair is partial but not conclusive evidence that the substance will enhance the initiating activity
of other substances or of background events.
The promotion step is taken to involve loss of growth control, represented by an
imbalance between rates of growth and death of cells in a colony and a continued net rate of
expansion even upon reaching confluence. The mechanism for this step is unspecified, but may
result in either a lowering of the rate of removal of cells (as in decreased apoptosis) or an
increase in the rate of division (as in abnormal proliferation). The ability of a substance to bring
this about may be determined from an increase in transformation following (1) prior application
of a known initiating agent and (2) subsequent application of the substance of interest. In the
absence of such data, the ability of a substance to alter intercellular communication, the rate of
cellular division, the rate of apoptosis, and/or net rate of growth provides partial but not
conclusive evidence of promoting activity.
The progression step is taken to be poorly understood at present, but probably is related
to the metastatic properties of the promoted cells. There is no assay for progression at present.
Partial but inconclusive evidence of progressing activity might be obtained from a demonstration
that cells have taken on the ability to dislodge from the primary tumor, spread to other organs
or tissues, reattach, and grow to fatal tumors.
If a substance induces only one of the required transitions, it will be necessary for other
transitions to be induced by antecedent conditions if the substance of interest is to be
considered a Partial carcinogen (as defined in Section 2.3.4). The fact that the background
142
-------
incidence of cancer is not zero may be taken to indicate that each of these other transitions will
always be present. If this assumption is adopted, the ability of a substance to increase the
incidence of cancer cannot be used to differentiate between claims of a Complete or a Partial
carcinogen (as defined in Section 2.3.4). Existing empirical evidence, however, is suggestive of
there being several routes by which cancer may occur. It is not inconsistent to assume that a
substance acts to induce only a subset of transitions, and that the remaining transitions required
are not those associated with the mechanism by which the antecedent conditions of the
experimental context act to bring about background cancer.
The epistemic status of this Theory of Carcinogenesis is taken here to be MED based on
an examination of the available empirical evidence linking DNA alteration, alteration of rates of
growth, and the carcinogenic potential of a range of substances tested in assays for initiation,
promotion and complete cancer. A judgment of HIGH is not considered appropriate since much
of the warrant for the theory is subjective judgment of experts in the field based on the
coherence of the explanatory framework and the finding of (unexplained) correlations.
533. Intellectual Obligation
The issue of Intellectual Obligation was discussed in detail in Section 1.4. Since the
judgments and assignments are common to all Working Tables and Contexts, this discussion is
not repeated here. By way of a summary, the Direct Empirical Relevance Strategy was assigned
an Intellectual Obligation of HIGH; the Semi-Empirical Extrapolation, Theory-Based Inference
and Empirical Correlation Relevance Strategies were assigned an Intellectual Obligation of
MEDIUM; and the Existential Insight Relevance Strategy was assigned an Intellectual
Obligation of LOW. The assignments are entered into the appropriate cells of Working Table
3.12.
143
-------
5.3.4. Claims of Carcinogenicity
The framework employed here allows for a wide variety of claims concerning the
carcinogenic action of formaldehyde. In the original report, it was determined that it is desirable
to augment the claim that a substance increases the incidence of cancer, since this sole claim
does not provide full detail on the antecedent conditions under which a substance such as
formaldehyde can exert its carcinogenic potential. It was determined that a more detailed
analysis would subdivide the claim of increased incidence of cancer into more detailed claims
potentially of use in determining the most effective regulatory strategy. The different Claims of
Carcinogenicity employed in Working Table 3.12 (and all subsequent Working Tables discussed
in this report) are:
(a) Increases Incidence of Cancer, implying that formaldehyde raises the incidence of
cancer, raises the multiplicity of cancer at sites, and/or changes the age at which cancer
appears.
(b) Complete Carcinogen, implying that formaldehyde increases the incidence of cancer
even when it is not delivered in conjunction with concurrent exposures.
(c) Partial Carcinogen, implying that formaldehyde increases the incidence of cancer only
when delivered in conjunction with concurrent exposures that complete the process of
carcinogenesis.
(d) Mixer, implying that formaldehyde exerts its carcinogenic effect only when it is
combined with another substance, and where this combination results in a new substance
which is at least a partial carcinogen.
(e) Helper, implying that formaldehyde does not in and of itself induce transitions to
cancer, but that it facilitates the carcinogenic action of another substance.
(f) Neoplastic Conversion, implying that formaldehyde acts by converting cells to
neoplastic potential.
(g) Neoplastic Development, implying that formaldehyde acts by producing growth in
colonies of cells already having neoplastic potential.
(h) Genotoxicity, implying that formaldehyde acts through a mechanism involving
interaction with the genetic material of a cell.
144
-------
(i) Non-genotoxicity, implying that formaldehyde acts through a mechanism other than
interaction with the genetic material of a cell.
The following summary considerations from Working Table 2.12 provide the foundations
for all inferences to Claims of Carcinogenicity drawn in this context:
BE1.1 demonstrates that in vivo exposure of humans to airborne formaldehyde in a
variety of occupations does not result in significant allergic response. This claim is given an
epistemic status of HIGH since completeness was high, utility was high, but no statistically
significant effect was found.
The separate judgments for Claims of Carcinogenicity for Working Table 3.12 are
provided in the sections that follow. In the case of each claim, explanations are given of the
judgment for each Relevance Strategy; then for the Column Summary across Relevance
Strategies for that claim; then for the Overall Summary across claims.
5.3.4.1. Increases Incidence of Cancer
The judgments for this Claim of Carcinogenicity for the various Relevance Strategies and
for this context are shown in the cells of Working Table 3.12. The basis for these judgments are
given below.
The judgment in all cells is NO since (1) the study measured allergic response, (2) no
response was found and (3) even if a response had been found, it is not judged that allergic
response is correlated with cancer or a causal factor in cancer induction. While the lack of
statistical significance of the findings doesn't necessarily indicate that formaldehyde does NOT
induce allergic response (the definition of "significance" could be made less stringent), it must be
borne in mind that allergic response is not given a role in the theory of carcinogenesis used by
these analysts. Allergic response also has not been shown to be correlated with cancer. As a
result, the analysts have chosen to retain the judgment that these data provide no basis for
145
-------
addressing the question of carcinogenicity. It should be noted that this issue (of the lack of
relevance of these data) could have been addressed in earlier Working Tables by assigning a
utility of NO to the data and dropping them from further consideration. In that case, there
would be no Intra-context warrants for the claims in Context 12.
5.3.4.2. Classifications
5.3.4.2.1. Complete Carcinogen
The judgment in all cells is NO since (1) the study measured allergic response, (2) no
response was found and (3) even if a response had been found, it is not judged that allergic
response is correlated with cancer or a causal factor in cancer induction.
5.3.4.2.2. Partial Carcinogen
The judgment in all cells is NO since (1) the study measured allergic response, (2) no
response was found and (3) even if a response had been found, it is not judged that allergic
response is correlated with cancer or a causal factor in cancer induction.
5.3.4.2.3. Mixer
The judgment in all cells is NO since (1) the study measured allergic response, (2) no
response was found and (3) even if a response had been found, it is not judged that allergic
response is correlated with cancer or a causal factor in cancer induction.
5.3.4.2.4. Helper
146
-------
The judgment in all cells is NO since (1) the study measured allergic response, (2) no
response was found and (3) even if a response had been found, it is not judged that allergic
response is correlated with cancer or a causal factor in cancer induction.
5.3.4.3. Stapes
5.3.4.3.1. Neoplastic Conversion
The judgment in all cells is NO since (1) the study measured allergic response, (2) no
response was found and (3) even if a response had been found, it is not judged that allergic
response is correlated with cancer or a causal factor in cancer induction.
5.3.4.3.2. Neoplastic Development
The judgment in all cells is NO since (1) the study measured allergic response, (2) no
response was found and (3) even if a response had been found, it is not judged that allergic
response is correlated with cancer or a causal factor in cancer induction.
5.3.4.4. Mechanisms
5.3.4.4.1. Genotoxic
The judgment in all cells is NO since (1) the study measured allergic response, (2) no
response was found and (3) even if a response had been found, it is not judged that allergic
response is correlated with cancer or a causal factor in cancer induction.
5.3.4.4.2. Non-Genotoxic
147
-------
The judgment in all cells is NO since (1) the study measured allergic response, (2) no
response was found and (3) even if a response had been found, it is not judged that allergic
response is correlated with cancer or a causal factor in cancer induction.
5.3.5. Summary Comments for Intm-Context Claims
The entries into the cells of Working Table 3.12 indicate that no claims concerning the
carcinogenicity of formaldehyde when exposure is through inhalation at concentrations below 2
ppm can be warranted by the data employed here.
5.4. Description of Working Table 4.12
Working Table 4.12 is utilized to justify premises necessary for extrapolation of the
findings in Working Table 3.12 (the "Observational Context") to the "Target Context" (Context
12). Since Context 12 is also the "Target Context", the extrapolation premises are not required
and there is no version of Working Table 4 for this context.
5.5. Description of Working Table 5.12
Working Table 5.12 contains judgments of the epistemic status for Inter-Context Claims
of Carcinogenicity. Since Context 12 is also the "Target Context", there is no version of Working
Table 5 for this context.
6. OBSERVATIONAL CONTEXT NO. 13
6.1. Description of Working Table 1.13
Working Table 1 for Context 13 is provided below. As in all subsequent sections of this
report, a Working Table is identified by both the general Working Table number (here, 1) and
148
-------
WORKING TABLE 1.13. STUDIES AVAILABLE FOR OBSERVATIONAL CONTEXT NUMBER 13: MAMMALIAN CELL
LINES EXPOSED TO FORMALDEHYDE IN SOLUTION AT EQUIVALENT CONCENTRATIONS IN EXCESS OF 2 PPM
Study
Number1
Data Category
Description
1
BE
Primary Reference #3 in Table 1
Primary cultures of rat tracheal epithelial cells isolated from 8-12 week
old specific-pathogen free Fischer 344 rats were exposed to liquid
formaldehyde at concentrations ranging from 0.01 (non-toxic) to 0.15
(toxic) ul/1 for 24 hours. 30 dishes of 20,000 cells each were exposed at
each concentration level. 30 days after exposure the cultures were fixed,
stained and scored for colonies having a density greater than 1,300
cells/mm2. Cell density within a colony was measured using a calibrated
eyepiece with a dissecting microscope. Positive and solvent control groups
were included with each assay to assure assay functionality.
Transformation frequency as a function of exposure level was reported.
'The study number refers to the order of the study within this context. The primary reference number, indexed to Table 1, is provided in the "Description"
column.
-------
the context (here, 13). As a result, this Working Table is assigned the number 1.13. The
biological components of this context are given by the fact that all exposed organisms were
mammalian cells cultured in vitro from rat tracheal epithelium.
The exposures in this context were all to formaldehyde in solution at concentrations
representative of high exposures.
One body of data was identified for this context. This is described in Working Table
1.13., and represents the data category of
(a) Biophysical Effect (taken from Reference #3 in Table 1). Here, the sole Biophysical
Effect is taken to be cellular transformation, which is assigned to this data category because it is
an effect leading to, but not identical with, cancer. It is given the label BE1.
Data in other data categories were not available for the studies utilized in this context.
62. Description of Working Table 2.13
The judgments for this Working Table are provided in the cells of Working Table 2.13.
The text here describes the reasons for these judgments. The data on which these judgments are
based are provided as various "Exhibits" after Working Table 2.13., with a numbering system
described in Section 6.2.4.
6.2.1. Data Categories and Items
The same data categories and items described with respect to Working Table 1.13 are
employed in Working Table 2.13. These are:
(a) Biophysical Effects, containing BE1 (cellular transformation);
This data category is displayed in Working Table 2.13.
149
-------
WORKING TABLE 2.13. DATA CHARACTERISTICS FOR OBSERVATIONAL CONTEXT NUMBER 13
Data Category/Item
Description
Completeness
(Hi/Me/Lo/No)
Utility
(Hi/Me/Lo/No)
Strength of Effect
(Hi/Me/Lo/No)
Exposure-
Specific
Effect'
Tumor Response
No study available
Biophysical Effect
Cellular transformation (BE 1.1)
ME
HI
ME
WT2.C13.
BE1.1
Pharm acodynam ics
No study available
Host Factors
No study available
Concurrent Environmental
Conditions
No study available
Related Substances Assessment
No study available
'Refers to the Exhibit Number for the data as described in the text.
-------
WT2.C13.BE1.1
Formaldehyde
2.5
0.5
0.2
0
0.1
0.3
0.5
0.4
Concentration (pl/liter)
Fig. 1. Typical rat tracheal epithelial cell transformation assay dose-response curves for four compounds:
pvrene. 3-methylchoIanthrene. formaldehyde, and benzo(alpyrene.
-------
622. Judgments of Completeness
In all cases, the judgment of completeness was based on a comparison of the number of
studies used in a data category/item for this context against the number available in the
complete reference list provided in Appendix A. The judgment was HIGH if the data used in
this Working Table was a relatively complete subset of the data indicated in the Appendix,
where "complete" was taken to mean somewhere on the order of 70% or above. The judgment
was MEDIUM if the data used in this Working Table constituted somewhere between 25 and
70% of the total body of potentially available data. The judgment was LOW if the data used
constituted less than 25% of the data available. These assignments also included limited
consideration of the QUALITY of the data, since it should not necessarily count against an
analysis if the analyst fails to collect data of poor quality. This consideration also is dealt with in
the assignment of Utility for the collected studies. Still, it is evident that the judgment of
Completeness might reasonably be expected to include consideration of something more than
the fraction of studies collected in a given data category/item. In the present analysis,
Completeness was increased slightly if the collected data for a given data category were the most
heavily cited in the secondary literature on formaldehyde. This leads to the following judgments
of Completeness:
For the data category BE1, the judgment for Completeness is MEDIUM as this is 1 of 3
items identified in the full literature search shown in Appendix A. The other 2 studies are
Boreiko and Ragan (1983); and Ragan and Boreiko (1982).
623. Judgments of Utility, Strength of Effect, and Exposure- Specific Effect
The remaining judgments in Working Table 2.13 refer to the qualities of the data
ACTUALLY EXAMINED IN THIS EXAMPLE. By Utility, we mean the quality of the study
150
-------
DESIGN and its ability to address in rigorous fashion questions pertinent to hazard
identification for formaldehyde. No attempt is made to determine the specific inferences
towards which the data will be directed in later Working Tables, since the intent here is only to
determine the degree to which a body of data arose from a study design appropriate for
accurate and precise measurements of the quantity considered in the study. Utility is increased
by high standards of study design, including the ability of the experiment to demonstrate a
causal link between exposure to formaldehyde and the measured effect.
By Strength of Effect, we mean the degree to which the study actually revealed an effect.
Strength is increased when the effect noted possessed statistical significance.
By Exposure-Specific Effect, we mean a summary of the actual effect observed (putting
aside questions of the reliability of this effect, which was dealt with in the assignment of Utility
and Strength of Effect). This summarization is provided in the following text, and the data
themselves are provided in the various Exhibits appearing after Working Table 2.13.
The following judgments then apply to Working Table 2.13:
For the data category BE1, the Utility judgment is HIGH since the study employed
measured concentrations of formaldehyde; it utilized the same rat line (F344) for developing the
cell culture as in the original CUT study; the study constituted a deliberate experimental
manipulation of formaldehyde concentration (important in causal claims); an adequate number
of cells were analyzed; scoring techniques were appropriate; and transformation of cells is
considered an indicator of the ability of formaldehyde to induce at least partial transitions to
cancer.
The Strength of Effect judgment is MEDIUM since the increase in transformation
frequency in the highest non-toxic exposure category appears to be significant, although no
statistical measure is provided. In addition, the slope of the transformation-exposure curve is
151
-------
positive until the onset of cytotoxicity, although again there is no statistical test of significance
developed. This judgment might have been HIGH if statistical tests had been reported. The
Exposure-Specific Effect is summarized in Exhibit WT2.C13.BE1.1. Transformation frequency
increased with increasing formaldehyde concentration up to 0.15 ul/1, followed by a decrease
presumably due to cytotoxicity.
62.4. Exhibits for Working Table 2.13
The data for the various data categories are provided as "Exhibits" after Working Table
2.13. The numbering system for the exhibits was chosen to display information on (in order of
appearance) the Working Table; the Context; the specific effect measured; and the study
number within that specific effect. For example, Exhibit WT2.C13.BE1.1 refers to Working
Table 2 (all exhibits have this assignment in common); Context 13; Biophysical Effect of type 1;
and data set #1 within the data category BE1. This same numbering scheme will be employed in
all other Contexts.
63. Description of Working Table 3.13
63.1. Introduction
The central task of Working Table 3.13 is to draw inferences on the carcinogenicity of
formaldehyde WITHIN CONTEXT 13, employing data specific to that context. This task
requires that several judgments be made sequentially:
(a) The analyst must determine the particular Claims of Carcinogenicity to be considered
in the analysis. These various claims are discussed in detail in Section 6.3.4.
(b) The available Relevance Strategies by which a given body of data may be related to a
specific Claim of Carcinogenicity must be developed. These strategies were discussed in
detail in Section' 1.4., and the same definitions are applied here.
152
-------
WORKING TABLE 3.13. INTRA-CONTEXT SUPPORT FOR CLAIMS OF CARCINOGENICITY IN CONTEXT 13
Claims of Carcinogenicity
Relevance Strategy
I.O.1
Increases
Incidence of
Transform.
Classification(s)
Stage
Mechanism
Complete
Partial
Mixer
Helper
Neo.
Conv.
Neo.
Devel.
Geno-
tox.
Non-
genot
ox.
Direct Empirical (D.E.)
HI
ME
NO
NO
NO
NO
NO
NO
NO
NO
Semi-Empirical
Extrapolation (S.E.E.)
ME
NO
NO
NO
NO
NO
NO
NO
NO
NO
Empirical Correlation
(E.C.)
ME
NO
NO
NO
NO
NO
NO
NO
NO
NO
Theory-based Inference
(T.B.I.)
ME
NO
NO
NO
NO
NO
NO
NO
NO
NO
Existential Insight(E.l.)
LO
ME
NO
NO
NO
NO
NO
NO
NO
NO
Column Summary
ME
NO
NO
NO
NO
NO
NO
NO
NO
Overall Summary
ME
NO
NO
NO
NO
NO
NO
NO
NO
'intellectual Obligation
-------
(c) The analyst must assign a measure of Intellectual Obligation to each Relevance
Strategy. This assignment was described in detail in Section 1.4., and the same
assignments are made here.
(d) The analyst must judge the epistemic status of a specific Claim of Carcinogenicity
arising from a specific Relevance Strategy. As described in Section 1.4., this judgment is
separate from the judgment of Intellectual Obligation for that Relevance Strategy. This
judgment is described in more detail in Section 6.3.2. The assignments are entered into
the appropriate cells of Working Table 3.13.
(e) The analyst must produce a summary judgment of the epistemic status of a particular
Claim of Carcinogenicity by combining the judgments from each of the five Relevance
Strategies for that claim. This summary includes consideration of both the epistemic
status of that claim for each Relevance Strategy, as entered into the separate cells of the
Working Table specific to each strategy, as well as the Intellectual Obligation assigned to
that strategy. The assignments are entered into the appropriate COLUMN SUMMARY
cells of Working Table 3.13.
(f) The analyst must produce a summary judgment of the COHERENCE of the various
Claims of Carcinogenicity. This task is performed by examining the coherence ACROSS
the various COLUMN SUMMARIES appearing in the Working Table. The assignments
are entered into the appropriate OVERALL SUMMARY cells of Working Table 3.13.
632. Epistemic Status of Claims of Carcinogenicity
The entries into the various cells of Working Table 3.13 reflect the judgment of the
analyst concerning the epistemic status of the Claim of Carcinogenicity being considered, FOR
A SPECIFIC RELEVANCE STRATEGY. This judgment will have two components:
(a) Each Relevance Strategy calls upon a selected body of data from which the reasoning
will proceed. Specifically, the Direct Empirical strategy requires data on direct measurements of
the effect underlying the Claim of Carcinogenicity, made at the exposure to formaldehyde
defining the context (here, concentrations in solution equivalent to airborne concentrations
above 2 ppm). The Semi-Empirical Extrapolation strategy requires data on direct measurements
of the effect underlying the Claim of Carcinogenicity, made at exposures to formaldehyde other
than those defining the context. The Empirical Correlation strategy requires data on effects
153
-------
other than the effect underlying the Claim of Carcinogenicity, where this "other" effect is taken
to correlate with appearance of the effect of interest. The Theory-Based Inference strategy
requires data on an effect other than the effect underlying the Claim of Carcinogenicity, where
this "other" effect is taken to be a causal step in the production of the effect of interest as
specified by an established theory. The Existential Insight strategy does not invoke data, except
in the sense that a subjective judgment from an individual might be counted as an instance of
data in and of itself.
The analyst must assign the "foundational quality" to the data used in each relevance
strategy. By "foundational quality", we mean the quality of the data with respect to the question
those data were intended to address BY THE ORIGINAL INVESTIGATOR. This judgment is
based on the results of Working Table 2.13 and includes consideration of the Completeness,
Utility, Strength of Effect and Exposure-Specific Effect from that table. While the judgment
ultimately is subjective, it is constrained in that it must increase (from NO to LOW to
MEDIUM to HIGH) as the strength of the judgments for Completeness, Utility and Strength of
Effect increase in Working Table 2.13. The summary of Exposure-Specific Effect may also enter
into the assignment of epistemic status to a Claim of Carcinogenicity if the analyst determines
that only effects above a given magnitude will exert an influence on carcinogenic processes (an
issue of importance when possible thresholds are considered).
(b) The analyst then must judge the strength of any "background premises" required by
the Relevance Strategy under consideration. These background premises are defined in detail in
the original report, but may be summarized as the set of (often implicit) premises that must be
introduced into an analysis if the data in Working Table 2.1. are to be used to draw inferences
about the Claims of Carcinogenicity for the context being considered. Presumably, many of these
background premises already are considered valid once Working Tables 1.13 and 2.13 are
154
-------
complete; otherwise, the data would not have been assigned to this context. But other
background premises concerning the reliability of extrapolation procedures (for the Semi-
Empirical Extrapolation relevance strategy), the strength and specificity of correlations (for the
Empirical Correlation relevance strategy), the degree of verification and validation of theories
(for the Theory-Based Inference relevance strategy), and the credentials of individuals making
subjective judgments (for the Existential Insight relevance strategy) must be warranted prior to
their use.
Having established these two components of epistemic status (foundational quality and
background premises) for a given Claim of Carcinogenicity, the analyst then produces a
composite judgment of the epistemic status of a given Claim of Carcinogenicity and enters this
into the appropriate cell of Working Table 3.13. There are two separate claims that might be
made within one of these cells. The first is the claim that formaldehyde DOES produce the
effect underlying that claim (such as increased incidence of cancer or genotoxic effects). In that
case, the assignments of NO/LOW/MED/HIGH refer to the epistemic status of the claim that
formaldehyde produces the effect. This assignment is entered into the area ABOVE the dashed
line in a cell. The second claim is that formaldehyde DOES NOT produce the effect underlying
the claim. In that case, the assignment of NO/LOW/MED/HIGH refer to the epistemic status
of the claim that formaldehyde DOES NOT produce the effect. This assignment is entered into
the area BELOW the dashed line in a cell. THE READER SHOULD NOTE THAT THIS IS A
NEW FEATURE OF THESE WORKING TABLES NOT DISCUSSED IN THE ORIGINAL
REPORT. The reason for this addition to the Working Tables is that the fact that a particular
body of data does not support the claim that formaldehyde does induce a specific effect IS NOT
NECESSARILY evidence that formaldehyde does not induce that effect. It may simply be the
155
-------
case that the data were obtained under conditions incapable of revealing any effect that might
be present.
This consideration is important due to the issue of "burden of proof'. At times, the
burden may lie on the regulatory agency to show that formaldehyde DOES induce cancer.
Attention then would be directed towards the judgments in the upper halves of the various cells
of Working Table 3.13. At other times, the burden may lie on the producer of formaldehyde to
show that formaldehyde DOES NOT induce cancer. Attention then would be directed towards
the judgments in the lower halves of the various cells of Working Table 3.13. IN THE
PRESENT EXAMPLE, ONLY JUDGMENTS THAT AN EFFECT IS PRODUCED BY
FORMALDEHYDE ARE DEVELOPED. THE PROCESS OF ANALYSIS WOULD BE
REPEATED FOR THE LOWER HALVES OF ALL CELLS IF THE BURDEN OF PROOF
WAS SHIFTED.
It is important here to consider the broad features of the theory of transformation to be
used throughout this section. These features structure the interpretation of data used in the
various Relevance Strategies. The theory of transformation employed here is that cells are
transformed by two steps or transitions. These are termed initiation and promotion, each of
which may require one or more sub-steps. The initiation step is taken to involve alteration of
DNA structure and function, producing a new genome. This alteration must then remain intact
without being repaired by processes normally operating in cells or induced by the damage itself.
The ability of a substance to bring a stable alteration about may be determined from an increase
in transformation following (1) application of the initiating agent and (2) subsequent application
of a promoting agent such as TPA. In the absence of such data, the ability of a substance to
interact with DNA, to alter DNA, and/or to induce mutation provides partial but not conclusive
evidence of initiating activity. In addition, the ability of a substance to lower DNA repair rates
156
-------
or fidelity of repair is partial but not conclusive evidence that the substance will enhance the
initiating activity of other substances or of background events.
The promotion step is taken to involve loss of growth control, represented by an
imbalance between rates of growth and death of cells in a colony and a continued net rate of
expansion even upon reaching confluence. The mechanism for this step is unspecified, but may
result in either a lowering of the rate of removal of cells (as in decreased apoptosis) or an
increase in the rate of division (as in abnormal proliferation). The ability of a substance to bring
this about may be determined from an increase in transformation following (1) prior application
of a known initiating agent and (2) subsequent application of the substance of interest. In the
absence of such data, the ability of a substance to alter the rate of cellular division, the rate of
apoptosis, and/or net rate of growth provides partial but not conclusive evidence of promoting
activity.
If a substance induces only one of the required transitions, it will be necessary for other
transitions to be induced by antecedent conditions if the substance of interest is to be
considered a Partial transforming agent. The fact that the background frequency of
transformation is not zero may be taken to indicate that each of these other transitions will
always be present. If this assumption is adopted, the ability of a substance to increase the
frequency of transformation cannot be used to differentiate between claims of a Complete or a
Partial transforming agent. Existing empirical evidence, however, is suggestive of there being
several routes by which transformation may occur. It is not inconsistent to assume that a
substance acts to induce only a subset of transitions, and that the remaining transitions required
are not those associated with the mechanism by which the antecedent conditions of the
experimental context act to bring about background transformation.
157
-------
The epistemic status of this Theory of Transformation is taken here to be MED based
on an examination of the available empirical evidence linking DNA alteration, alteration of rates
of growth, and the transforming potential of a range of substances tested in assays for initiation,
promotion and transformation. A judgment of HIGH is not considered appropriate since much
of the warrant for the theory is subjective judgment of experts in the field based on the
coherence of the explanatory framework and on limited experimental tests of key assumptions.
6 J J. Intellectual Obligation
The issue of Intellectual Obligation was discussed in detail in Section 1.4. Since the
judgments and assignments are common to all Working Tables and Contexts, this discussion is
not repeated here. By way of a summary, the Direct Empirical Relevance Strategy was assigned
an Intellectual Obligation of HIGH; the Semi-Empirical Extrapolation, Theory-Based Inference
and Empirical Correlation Relevance Strategies were assigned an Intellectual Obligation of
MEDIUM; and the Existential Insight Relevance Strategy was assigned an Intellectual
Obligation of LOW. The assignments are entered into the appropriate cells of Working Table
3.13.
63.4. Claims of Carcinogenicity
The framework employed here allows for a wide variety of claims concerning the
carcinogenic action of formaldehyde. In the original report, it was determined that it is desirable
to augment the claim that a substance increases the incidence of cancer, since this sole claim
does not provide full detail on the antecedent conditions under which a substance such as
formaldehyde can exert its carcinogenic potential. It was determined that a more detailed
analysis would subdivide the claim of increased incidence of cancer into more detailed claims
158
-------
potentially of use in determining the most effective regulatory strategy. The different Claims of
Carcinogenicity employed in Working Table 3.13 (and all subsequent Working Tables discussed
in this report) are:
(a) Increases Incidence of Transformation, implying that formaldehyde raises the
incidence of transformation.
(b) Complete Carcinogen, implying that formaldehyde increases the incidence of
transformation even when it is not delivered in conjunction with concurrent exposures.
(c) Partial Carcinogen, implying that formaldehyde increases the incidence of
transformation only when delivered in conjunction with concurrent exposures that
complete the process of carcinogenesis.
(d) Mixer, implying that formaldehyde exerts its carcinogenic effect only when it is
combined with another substance, and where this combination results in a new substance
which is at least a partial carcinogen.
(e) Helper, implying that formaldehyde does not in and of itself induce transitions to
transformation, but that it facilitates the carcinogenic action of another substance.
(f) Neoplastic Conversion, implying that formaldehyde acts by converting cells to
neoplastic potential.
(g) Neoplastic Development, implying that formaldehyde acts by producing growth in
colonies of cells already having neoplastic potential.
(h) Genotoxicity, implying that formaldehyde acts through a mechanism involving
interaction with the genetic material of a cell.
(i) Non-genotoxicity, implying that formaldehyde acts through a mechanism other than
interaction with the genetic material of a cell.
The following summary considerations from Working Table 2.13 provide the foundations
for all inferences to Claims of Carcinogenicity drawn in this context:
BE1.1 demonstrates that in vitro exposure to formaldehyde of a colony of cells derived
from tissue of the F344 rat line induces cellular transformation. This form of cellular
transformation is identical to that defining the various claims made in Working Table 3 for this
context. The statement that this transformation is dose-dependent and caused by the action of
formaldehyde or a metabolite is given an epistemic status of MED since, as demonstrated in
159
-------
Working Table 2 for this context, the utility of the study is HIGH and the strength of effect is at
least MED. The fact that the completeness of the BE1 data set has only been judged MED is
taken to reduce the overall judgment of epistemic status to below HIGH, since the other
(unexamined) studies might have yielded different results. This is an important consideration
since there often are unexplained differences in transformation assays performed by different
experimental groups even when the same procedures supposedly are being used.
The separate judgments for Claims of Carcinogenicity for Working Table 3.13 are
provided in the sections that follow. In the case of each claim, explanations are given of the
judgment for each Relevance Strategy; then for the Column Summary across Relevance
Strategies for that claim; then for the Overall Summary across claims.
6.3.4.1. Increases Incidence of Transformation
The judgments for this Claim of Carcinogenicity for the various Relevance Strategies and
for this context are shown in the cells of Working Table 3.13. The basis for these judgments are
given below.
A. Direct Empirical: The judgment is MED since transformation clearly is present in this
one study with good foundational characteristics as described above, but the completeness of the
data set examined is only MED.
B. Semi-Empirical Extrapolation: The judgment is NO since transformation was not
measured.
C. Empirical Correlation: The judgment is NO since the measured effect
(transformation) is identical to the claimed effect (transformation). This evidence does not,
therefore, constitute a case of correlation, but rather one of Direct Empirical observation.
160
-------
D. Theory-Based Inference: The judgment is NO since the study does not provide
information on any mechanisms of action, from which an explanation might be developed. It is
not the case, therefore, that the data in BEl-Study 1 support the claim that any step in the
process of transformation is understood to be the mechanism by which formaldehyde acts to
transform these cells.
E. Existential Insight: The judgment is MED since the data produce in the analysts
moderate confidence in the transformation ability of formaldehyde. It is judged further that the
analysts possess the relevant experience, training and psychological skills necessary to form such
subjective judgments, based on having worked in the science of carcinogenesis for several years.
F. Column Summary: The judgment is MED since Direct Empirical data are available
and indicate transforming ability with MED epistemic status, and this warrant carries high
intellectual obligation. It is worrying that Theory-Based Inference is NO, since this carries MED
Intellectual Obligation. Still, this does not dictate that the overall judgment cannot be MED; it
only dictates that the judgment cannot be above MED.
G. Overall Summary: The judgment is MED since the analysts' strongly foundational
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim. This indicates that nothing can be said about the mechanism of action
within this context, so there is no issue of coherence that arises in this case.
6.3.4.2. Classifications
6.3.4.2.1. Complete Transformation
The judgment in all cells is NO since transformation was measured rather than the
ability of formaldehyde to induce specific steps in transformation. Since the background
incidence of transformation was non-zero, it is clear that each transition required for
161
-------
transformation occurs with some non-zero background probability in the population of cells. As
a result, it is not possible to distinguish between the hypothesis that formaldehyde induces ALL
transitions (and is, therefore, a complete transformation agent), and the hypothesis that
formaldehyde induces only a subset of the transitions with the remainder being completed by
background transitions (and is, therefore, a partial transforming agent). The analysts recognize,
however, that the definition of a complete transforming agent COULD be taken to be a
substance that does not require the concurrent application of either a known initiator or
promoter. In that case, the present study could be taken to indicate that formaldehyde is a
complete transforming agent and not a partial transforming agent.
6.3.4.2.2. Partial Transformation
The judgment in all cells is NO since transformation was measured rather than the
ability of formaldehyde to induce specific steps in transformation. Since the background
incidence of transformation was non-zero, it is clear that each transition required for
transformation occurs with some non-zero background probability in the population of cells. As
a result, it is not possible to distinguish between the hypothesis that formaldehyde induces ALL
transitions (and is, therefore, a complete transformation agent), and the hypothesis that
formaldehyde induces only a subset of the transitions with the remainder being completed by
background transitions (and is, therefore, a partial transforming agent). The analysts recognize,
however, that the definition of a complete transforming agent COULD be taken to be a
substance that does not require the concurrent application of either a known initiator or
promoter. In that case, the present study could be taken to indicate that formaldehyde is a
complete transforming agent and not a partial transforming agent.
162
-------
6.3.4.2.3. Mixer
The judgment in all cells is NO since transformation was measured rather than the
ability of formaldehyde to induce specific steps in transformation or to interact with other
agents.
6.3.4.2.4. Helper
The judgment in all cells is NO since transformation was measured rather than the
ability of formaldehyde to induce specific steps in transformation or to interact with other
agents.
6.3.4.3. Stages
6.3.4.3.1. Neoplastic Conversion
The judgment in all cells is NO since transformation was measured rather than the
ability of formaldehyde to induce specific steps in transformation or to interact with other
agents.
6.3.4.3.2. Neoplastic Development
The judgment in all cells is NO since transformation was measured rather than the
ability of formaldehyde to induce specific steps in transformation or to interact with other
agents.
6.3.4.4. Mechanisms
6.3.4.4.1. Genotoxic
163
-------
The judgment in all cells is NO since transformation was measured rather than the
ability of formaldehyde to induce specific steps in transformation or to interact with other
agents.
6.3.4.4.2. Non-Genotoxic
The judgment in all cells is NO since transformation was measured rather than the
ability of formaldehyde to induce specific steps in transformation or to interact with other
agents.
63.5. Summary Comments for Intra-Context Claims
The entries into the cells of Working Table 3.113 indicate that formaldehyde should be
moderately considered capable of inducing transformation in non-human mammalian cell lines
when exposed in vitro to formaldehyde in solution at concentrations representative of airborne
exposures at or above 2 ppm. This judgment is reflected most clearly in the fact that the
"Overall Summary" cell for "Increases Incidence of Transformation" is judged to be MED due to
the presence of Direct Empirical data showing transformation. The only feature of the data
reducing this judgment below HIGH was the low Completeness of the data.
6.4. Description of Working Table 4.13
6.4.1. Introduction
Working Table 4.13 is utilized to justify premises necessary for extrapolation of the
findings in Working Table 3.13 (the "Observational Context") to the "Target Context" (Context
12). Four separate claims must be warranted, each of which is described in detail in the original
report and in lesser detail below. The data on which judgments for these premises are based
164
-------
WORKING TABLE 4.13. SUPPORT FOR INTER-CONTEXT EXTRAPOLATION PREMISES
CONTEXT NO. 13 TO CONTEXT NO. 12
Relevance Strategy
I.O."
Exposure to BSDR
Conversion
BSDR to Effect
Conversion
Host Factors
Environmental
Conditions
Direct Empirical (D.E.)
HI
NO
NO
LO
LO
Semi-Empirical Extrapolation
(S.E.E.)
ME
NO
NO
NO
NO
Empirical Correlation (E.C.)
ME
LO
ME
ME
ME
Theory-based Inference (T.B.I.)
ME
ME
ME
ME
ME
Existential Insight (E.I.)
LO
ME
ME
LO
ME
Overall Assessment
LO
ME
LO
ME
'Intellectual Obligation
-------
were described in Section 6.3.4. All judgments of Intellectual Obligation are as described in
Section 1.4.
6.42. Column Headings in Working Table 4.13
These extrapolation premises or claims as they appear in Working Table 4.13 are:
1. EXPOSURE TO BSDR: Any differences between the BIOLOGICAL components of
the two contexts with respect to the relationship between exposure and BSDR are not so large
as to call into question whether a BSDR might be present in one context but not the other AT
THE SAME LEVEL OF EXPOSURE.
The pertinent considerations here for each context are whether (1) an exposure occurs
through some compartment of the environment (a compartment refers to the medium of the
environment, such as air or water); (2) an intake by some route can occur (a route refers to
inhalation, ingestion, dermal absorption, intubation or injection); (3) this intake results in an
uptake to a target organ or tissue; (4) clearance from the target allows the production of a
burden; and (5) biological transformation to the active metabolite exists. Where extrapolation is
across exposure levels, it is particularly important to establish that there does not exist a
threshold exposure below which a BSDR is not produced (due, e.g., to necessary saturation of
metabolic transformation). Where extrapolation is across species, it is particularly important to
show that any such threshold, if it exists, does not differ between the species in a manner calling
into question this premise.
2. BSDR TO BIOPHYSICAL EFFECT: Any differences between the BIOLOGICAL
components of the two contexts with respect to the relationship between BSDR and the
production of biophysical effects necessary for transitions are not so large as to call into
165
-------
question whether a biophysical effect might be present in one context but not the other
ASSUMING THE SAME LEVEL OF BSDR IN BOTH CONTEXTS.
The pertinent considerations here for each context are whether (1) the active metabolite
can interact with causally important biological structures in the organism; (2) this interaction
results in either DNA alterations (for initiation) or stimulation of cellular division (for
promotion); and (3) these biophysical effects are of the type and magnitude necessary to induce
transitions designated as either neoplastic development or neoplastic conversion. Where
extrapolation is across exposure levels, it is particularly important to establish that there does
not exist a threshold BSDR below which a biophysical effect is not produced (due, e.g., to
necessary saturation of a pool of DNA damage or a stimulus to cellular proliferation). Where
extrapolation is across species, it is particularly important to show that any such threshold, if it
exists, does not differ between the species in a manner calling into question this premise.
3. ENVIRONMENTAL CONDITIONS: Any differences between the
ENVIRONMENTAL components of the two contexts with respect to the relationship between
exposure (to the substance of interest) and BSDR, or the relationship between BSDR and the
production of biophysical effects necessary for transitions, are not so large as to call into
question whether a biophysical effect might be present in one context but not the other AT THE
SAME LEVEL OF EXPOSURE TO THE SUBSTANCE OF INTEREST.
The pertinent considerations here for each context are whether (1) environmental
conditions increase or decrease the BSDR resulting from a given level of exposure to the
substance of interest, relative to the conditions that exist in the target context; (2) environmental
conditions increase or decrease the probability of biophysical effects from a given level of
BSDR, relative to the target context; and (3) environmental conditions increase or decrease the
probability of transitions from a given level of biophysical effect, relative to the target context.
166
-------
4. HOST FACTORS: Any differences between the BIOLOGICAL components of the
two contexts with respect to the relationship between the production of biophysical effects, the
probability of transitions necessary for cancer, and the probability of cancer itself are not so
large as to call into question whether an increase in cancer incidence might be present in one
context but not the other AT THE SAME LEVEL OF BIOPHYSICAL EFFECT.
The pertinent considerations here for each context are whether (1) a target organ or
tissue exists; (2) biological structures on which relevant biophysical effects are produced exist;
(3) relevant background transitions exist; and (4) repair of biophysical effects, or regression of
states of neoplastic conversion or neoplastic development, exist. Where extrapolation is across
species, it is particularly important to establish that there does not exist a threshold level of
biophysical effect below which a transition is not produced (due, e.g., to necessary saturation of
a stimulus to cellular proliferation before promotion can occur).
In the original report, Working Table 4 (bearing in mind that Working Table 4 in the
present report was numbered Working Table 5 in the original report, as described in Section 1)
also contained an extrapolation premise concerning the role of intersubject and intrasubject
variability. This premise has been removed here and consideration of the role of differences in
intra- and intersubject variability between the two contexts has been incorporated into the 4
separate premises above. This choice was made because variability always refers to variability of
some PROPERTY, and this property would appear in one of the four extrapolation categories
listed above. The analyst then must determine the significance of variability on the relationship
between EXPOSURE AND BSDR, BSDR AND BIOPHYSICAL EFFECT,
ENVIRONMENTAL CONDITIONS, AND HOST FACTORS separately.
6.4 J. EpLstemic Status of Extrapolation Premises
167
-------
This section summaries the judgments made for each of the cells in Working Table 4.13.
The procedure for assigning epistemic status to a cell is identical to that outlined for Working
Table 3.13 and a description is not repeated here.
6.4.3.1. Exposure to BSDR
A. Direct Empirical: The judgment is NO since BSDR was not determined in either
context at any level of exposure.
B. Semi-Empirical Extrapolation: The judgment is NO since BSDR was not determined
in either context at any level of exposure.
C. Empirical Correlation: The judgment is LOW since exposure to formaldehyde (which
is well established in both contexts) is taken to correlate with moderate strength with production
of a dose, but metabolic transformation has not been identified in the observational context.
Where the active metabolite has not been identified, the correlation between exposure and
BSDR (rather than dose) will be weakened. In addition, the lack of existence of a threshold has
not been established in either context.
D. Theorv-Based Inference: The judgment is MED since exposure clearly occurs in both
contexts; intake clearly occurs for the observational context and the presence of inhalation
strongly suggests that intake occurs in the target context, although at a lower level than in the
observational context; uptake clearly occurs for the observational context and the absorption of
formaldehyde in nasal passages upon inhalation in humans has been established with MED
epistemic status; clearance is present but not infinitely rapid in both contexts; but the active
metabolite has not been shown present in either context. While the theory of pharmacodynamics
suggests that a dose should be produced in both contexts, it is not clear that a BSDR will be
produced or that it will exceed any threshold necessary in either context.
168
-------
E. Existential Insight: The judgment is MED since these data produce in the analysts
moderate confidence that a BSDR is produced in both contexts.
F. Overall Summary: The judgment is LOW since only Existential Insight and Theory-
Based Inference were of MED strength, but neither of these has HIGH Intellectual Obligation.
6.4.3.2. BSDR to Effect
A. Direct Empirical: The judgment is NO since the same biophysical effect was not
determined in these contexts at any level of exposure.
B. Semi-Empirical Extrapolation: The judgment is NO since the same biophysical effect
was not determined in these contexts at any level of exposure.
C. Empirical Correlation. The judgment is MED since the ability of the active metabolite
of a substance to induce biophysical effects in vitro UPON PRODUCTION OF A BSDR is
taken to correlate with moderate strength with the ability of that same active metabolite to
induce biophysical effects in vivo. It has not, however, been shown that a threshold for
production of biophysical effects does not exist or is similar between contexts.
D. Theorv-Based Inference: The judgment is MED since similar cellular structures exist
in both contexts; formaldehyde or its active metabolite is assumed to diffuse in a similar fashion
in both contexts; and interaction with pertinent biological structures should occur through this
diffusion. It has not been established that a threshold for biophysical does not exist or is similar
between contexts.
E. Existential Insight: The judgment is MED since these data produce in the analysts
moderate confidence that interaction between formaldehyde (or its active metabolite) and
pertinent biological structures is produced in both contexts.
169
-------
F. Overall Summary: The judgment is MED since Empirical Correlation and Existential
Insight and Theory-Based Inference are MED, while the Intellectual Obligation for Empirical
Correlation and Theory-Based Inference is MED. The lack of Direct Empirical support prevents
this judgment from being HIGH.
6.4.3.3. Host Factors
A. Direct Empirical: The judgment is LOW since the target for Context 12 has been
observed to be the nasal passages and has been observed to be the entire cell in Context 13; the
pertinent structures for production of biophysical effects were not observed directly; background
transitions clearly were present in some form since the background incidence of effect was non-
zero in both contexts; the effect of repair was not observed.
B. Semi-Empirical Extrapolation: The judgment is NO since the consideration of host
factors does not require extrapolation across exposure levels, and was dealt with in A above.
C. Empirical Correlation: The judgment is MED since the ability of a substance to
induce transformation in vitro correlates with moderate strength with the ability to induce
carcinogenesis in humans, suggesting that there also is a correlation in the necessary host
factors. This judgment is not HIGH since it has not been demonstrated that there is not a
threshold level of biophysical effect necessary for transitions or that this threshold is similar
between contexts.
D. Theory-Based Inference: The judgment is MED based on the fact that cellular
structure and function is similar in the two contexts since both involve mammalian cells, so
similar targets should exist; the target in Context 12 is taken to be the nasal passages, which
clearly exist; repair should be present but not completely effective in both contexts; and
background transitions are taken to occur in both contexts as evidenced by the non-zero rates of
170
-------
effect in suitable controls. This judgment is not HIGH since it has not been demonstrated that
there is not a threshold level of biophysical effect necessary for transitions or that this threshold
is similar between contexts.
E. Existential Insight: The judgment is LOW since the biological differences between the
two contexts produce in the analysts moderate confidence that the two contexts are similar.
F. Overall Summary: The judgment is LOW since Empirical Correlation and Theory-
Based Inference are MED, while the Intellectual Obligations for these are MED. The LOW
judgment for Direct Empirical lowers this overall judgment.
6.4.3.4. Environmental Conditions
A. Direct Empirical: The judgment is LOW since the exposure was clearly to
formaldehyde in Context 13, but exposures other than to formaldehyde will be present in
Context 12.
B. Semi-Empirical Extrapolation: The judgment is NO since this warrant does not apply
for this extrapolation premise.
C. Empirical Correlation: The judgment is MED since the ability of a substance to
induce transformation in vitro correlates with moderate strength with the ability to induce
carcinogenesis in humans, suggesting that there also is a correlation in the necessary
environmental conditions.
D. Theory-Based Inference: The judgment is MED based on the theory that the
presence of substances other than formaldehyde in Context 12 would primarily enhance the
transitions induced by only formaldehyde in Context 13; this theory is given MED epistemic
status.
171
-------
E. Existential Insight: The judgment is MED since the environmental differences
between the two contexts produce in the analysts moderate confidence that the two contexts are
similar.
F. Overall Summary: The judgment is MED since Empirical Correlation and Theory-
Based Inference are MED, while the Intellectual Obligations for these are MED. The presence
of some Direct Empirical support would tend to enhance this epistemic status, but the judgment
for that warrant was LOW.
6.5. Description of Working Table 5.13
6.5.1. Introduction
Working Table 5.13 contains judgments of the epistemic status for Inter-Context Claims
of Carcinogenicity. In each entry, the "Observational Context" is Context 13 and the "Target
Context" is Context 12. The judgments in Working Table 5.13 utilize the judgments in Working
Table 3.13 for a particular Claim of Carcinogenicity, as well as the judgments in Working Table
4.13 for the necessary Extrapolation Premises. In each case of an extrapolation, it is the "Overall
Summary" judgment associated with a Claim of Carcinogenicity in Working Table 3.13 that
forms the basis for the extrapolation. The final judgment of epistemic status in Working Table
5.13 cannot be higher than this initial judgment from Working Table 3.13, since the application
of the extrapolation premises can have only the effect of weakening the extrapolation. In other
words, if formaldehyde is judged to "Increase the Incidence of Cancer" with MEDIUM epistemic
status in Working Table 3.13, the highest judgment of epistemic status for the instance of
extrapolation will be MEDIUM, which presumes that the extrapolation premises are judged to
each have HIGH epistemic status in Working Table 4.13.
172
-------
The analysts have chosen to view warrants from Working Table 3.13 in the following
manner. A particular cell from that table may be used as Empirical Correlation, Theory-Based
Inference and/or Existential Insight relevance strategies in Working Table 5.13 for the same
Claim of Carcinogenicity. A particular claim from Working Table 3.13 may NOT be used as a
Direct Empirical relevance strategy in Working Table 5.13, since there always exists
extrapolation across exposure levels. A particular claim from Working Table 3.13 MAY be used
as a Semi-Empirical Extrapolation relevance strategy if it is judged that the biological and
environmental conditions in the two contexts are sufficiently similar to warrant the claim that
these two contexts contain approximately the same biological and environmental antecedent
conditions, differing only in LEVEL of exposure to formaldehyde. This clearly requires relatively
strong epistemic status to each of the four extrapolation premises found in Working Table 4.13,
a condition only weakly met in the present case.
6.5.2. Intellectual Obligation
The assignments of Intellectual Obligation discussed in Section 1.4, and utilized in
Working Table 3.11, are employed in Working Table 5.13.
6.53. Claims of Carcinogenicity
The same grouping of Claims of Carcinogenicity used as headings in Working Table 3.13
are employed in Working Table 5.13. The judgments of the epistemic status of each of these
claims, for each relevance strategy, are provided in the cells of Working Table 5.13 and are
described below.
6.5.3.1. Increases Incidence of Cancer
173
-------
WORKING TABLE 5.13. INTER-CONTEXT SUPPORT FOR CLAIMS OF CARCINOGENICITY
EXTRAPOLATED FROM CONTEXT 13 TO CONTEXT 12
Claims of Carcinogenicity
Relevance Strategy
I.O.1
Increases
Incidence of
Cancer
Classification(s)
Stage
Mechanism
Complete
Partial
Mixer
Helper
Neo.
Conv.
Neo.
Devel.
Geno-
tox.
Non-
genot
ox.
Direct Empirical (D.E.)
HI
NO
NO
NO
NO
NO
NO
NO
NO
NO
Semi-Empirical
Extrapolation (S.E.E.)
ME
NO
NO
NO
NO
NO
NO
NO
NO
NO
Empirical Correlation
(E.C.)
ME
LO
NO
NO
NO
NO
NO
NO
NO
NO
Theory-based Inference
(TB.I.)
ME
LO
NO
NO
NO
NO
NO
NO
NO
NO
Existential Insight(E.I.)
LO
ME
NO
NO
NO
NO
NO
NO
NO
NO
Column Summary
LO
NO
NO
NO
NO
NO
NO
NO
NO
Overall Summary
LO
NO
NO
NO
NO
NO
NO
NO
NO
'Intellectual Obligation
-------
A. Direct Empirical: The judgment is NO due to differences in level of exposure and
biological context.
B. Semi-Empirical Extrapolation: The judgment is NO due to LOW epistemic status of
necessary extrapolation premises and differences in context (which are not offset by LOW
strength of extrapolation premises).
C. Empirical Correlation: The judgment is LOW since the claim was given MED
epistemic status in Working Table 3.13 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.13. While a correlation does exist between in
vitro transformation and carcinogenicity, the MED epistemic status of the claim in Working
Table 3.13 keeps this judgment from being HIGH. The lack of evidence on Genotoxicity in
Context 13 keeps this epistemic status from being elevated to MED.
D. Theory-Based Inference: The judgment is LOW since the claim was given MED
epistemic status in Working Table 3.13 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.13. While the Theory of Carcinogenesis does
support the claim that in vitro transformation may indicate a transition necessary for
carcinogenesis, the MED epistemic status of the claim in Working Table 3.13 and the LOW to
MED epistemic status of the extrapolation premises keeps this judgment from being MED. The
lack of evidence on Genotoxicity in Context 13 keeps this epistemic status from being elevated
to MED.
E. Existential Insight: The judgment is MED since the evidence produces in the analysts
moderate confidence that formaldehyde is a carcinogen in Context 12.
F. Column Summary: The judgment is LOW since Direct Empirical and Semi-Empirical
Extrapolation were NO (with HIGH Intellectual Obligation), and the remainder were either
LOW (with MED Intellectual Obligation) or MED (with LOW Intellectual Obligation).
174
-------
G. Overall Summary: The judgment is LOW since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
6.5.3.2. Classifications
6.5.3.2.1. Complete Carcinogen
The judgment in all cells is NO since the same judgment is given in Working Table 3.13
(for the reasons stated there).
6.5.3.2.2. Partial Carcinogen
The judgment in all cells is NO since the same judgment is given in Working Table 3.13
(for the reasons stated there).
6.5.3.2.3. Mixer
The judgment in all cells is NO since the same judgment is given in Working Table 3.13
(for the reasons stated there).
6.5.3.2.4. Helper
The judgment in all cells is NO since the same judgment is given in Working Table 3.13
(for the reasons stated there).
6.5.3.3. Stages
6.5.3.3.1. Neoplastic Conversion
175
-------
The judgment in all cells is NO since the same judgment is given in Working Table 3.13
(for the reasons stated there).
6.5.3.3.2. Neoplastic Development
The judgment in all cells is NO since the same judgment is given in Working Table 3.13
(for the reasons stated there).
6.5.3.4. Mechanisms
6.5.3.4.1. Genotoxic
The judgment in all cells is NO since the same judgment is given in Working Table 3.13
(for the reasons stated there).
6.5.3.4.2. Non-Genotoxic
The judgment in all cells is NO since the same judgment is given in Working Table 3.13
(for the reasons stated there).
7. OBSERVATIONAL CONTEXT NO. 14
7.1. Description of Working Table 1.14
Working Table 1 for Context 14 is provided below. As in all subsequent sections of this
report, a Working Table is identified by both the general Working Table number (here, 1) and
the context (here, 14). As a result, this Working Table is assigned the number 1.14. The
biological components of this context are given by the fact that all exposed organisms were
human cells established as cultures from fibroblasts.
176
-------
WORKING TABLE 1.14. STUDIES AVAILABLE FOR OBSERVATIONAL CONTEXT NUMBER 14: HUMAN CELL LINES
EXPOSED TO FORMALDEHYDE IN SOLUTION AT EQUIVALENT CONCENTRATIONS IN EXCESS OF 2 PPM
Study
Number1
Data Category
Description
1
BE and HF
Primary Reference #10 in Table 1
Human tracheobronchial tissues taken from patients without clinical
evidence of cancer were maintained in culture as explants and used to
establish primary outgrowths of fibroblasts serially cultured in serum- and
growth factor- supplemented medium. Cell lines then were exposed to
various concentrations of formaldehyde for 1 hour to determine the level
at which a 50% decrease in colony forming efficiency was found. This
exposure then was utilized throughout subsequent measurements. In
addition to measurements of clonal growth rate,; trypan blue exclusion
ability; cross-linked envelope formation; free-total low molecular weight
thiol content; DNA single strand breaks; and DNA cross-links; the
primary endpoints of interest here are:
A) Frequency of 6-Thioguanine resistence mutations (BE)
B) Alterations of DNA repair activity (HF)
'The study number refers to the order of the study within this context. The primary reference number, indexed to Table 1, is provided in the "Description"
column
-------
The exposures in this context were all to formaldehyde in solution at concentrations
representative of high exposures.
Two bodies of data were identified for this context. These are described in Working
Table 1.14., and are divided into the data categories of
(a) Biophysical Effect (taken from Reference #8 in Table 1). Here, the sole Biophysical
Effect is taken to be cellular mutation, which is assigned to this data category because it is an
effect leading to, but not identical with, cancer. It is given the label BE1.
(b) Host Factor (taken from Reference #8 in Table 1). Here, the sole Host Factor
measured is alteration of DNA repair rates, which is assigned to this data category because it
stands intermediate between the production of biophysical effects (such as DNA damage) and
transitions to cancer. It is given the label HF1.
Data in other data categories were not available for the studies utilized in this context.
12. Description of Working Table 2.14
The judgments for this Working Table are provided in the cells of Working Table 2.14.
The text here describes the reasons for these judgments. The data on which these judgments are
based are provided as various "Exhibits" after Working Table 2.14., with a numbering system
described in Section 7.2.4.
72.1. Data Categories and Items
The same data categories and items described with respect to Working Table 1.14 are
employed in Working Table 2.14. These are:
(a) Biophysical Effects, containing BE1 (cellular mutation);
(b) Host Factors, containing HF1 (DNA repair).
177
-------
WORKING TABLE 2.14. DATA CHARACTERISTICS FOR OBSERVATIONAL CONTEXT NUMBER 14
Data Category/Item
Description
Completeness
(Hi/Me/Lo/No)
Utility
(Hi/Me/Lo/No)
Strength of Effect
(Hi/Me/Lo/No)
Exposure-
Specific
Effect'
Tumor Response
No study available
Biophysical Effect
6-Thioguanine mutation (BE1.1)
HI
HI
ME
WT2.C14.
BE1.1
Pharm acodynam ics
No study available
Host Factors
DNA repair (HFI.l)
HI
HI
ME
WT2.C14.
HFI.l
Concurrent Environmental
Conditions
No study available
Related Substances Assessment
No study available
'Refers to the Exhibit Number for the data as described in the text.
-------
WT2.C14.BE1.1
TABLE 2
pathobiological effects of formaldehyde and acrolein in cultured human BRONCHIAL CELLS
Aldehyde Colony-
Clonal
Trypan
Gluta-
Cross-
Cytosolic DNA single-
DNA-protein 6-TWoguanmT"
forming
growth
blue
thioneso d
linked
[CaJ*]f strand
cross-links 4
resistance
effi-
rate30 "
exclu-
(mM)
envelope c
(mM) breaks8
(no. per
mutations h
ciency^ "
(mM)
sion 30'
-------
WT2.C14.HF1.1
table 3
INHIBITORY EFFECTS OF FORMALDEHYDE AND ACROLEIN ON DNA REPAIR IN HUMAN BRONCHIAL FIBRO-
BLASTS
Activity related to DNA repair Aldehyde Dose DNA repair activity in the
presence of aldehyde
(percent of control)
O'-Methylguanine-DNA
Acrolein
10 m-M
79
Methyltransferaje activity *
30 pM.
60
Formaldehyde
100 mM
89
300 mM
84
Removal of O'-raethylguanine b
Formaldehyde
100 mM
82
300 /iM
61
Resealing of ionizing radiation-
Formaldehyde
100 tiM
85
induced single-strand breaks c
Following exposure of cells to the respective aldehyde for 1 h. cell extracts were prepared and analyzed for enzyme activity (Krokan
et al.. 1985).
Ratios of 06-meihylguanine to 7-methyIguanine were used as index for 04-methylguanine repair in cells exposed to 200 p-M. MNU
for 1 h and subsequently left for 5 h repair incubation with and without the indicated concentrations of formaldehyde (GrafstrSm
et al.. 1985 a).
Cells were exposed to 800 rad of ionizing radiation and removal of SSB was measured with and without the presence of
formaldehyde for 1 h (Grafstrom et al., 1983).
-------
These two separate data categories are displayed in Working Table 2.14.
122. Judgments of Completeness
In all cases, the judgment of completeness was based on a comparison of the number of
studies used in a data category/item for this context against the number available in the
complete reference list provided in Appendix A. The judgment was HIGH if the data used in
this Working Table was a relatively complete subset of the data indicated in the Appendix,
where "complete" was taken to mean somewhere on the order of 70% or above. The judgment
was MEDIUM if the data used in this Working Table constituted somewhere between 25 and
70% of the total body of potentially available data. The judgment was LOW if the data used
constituted less than 25% of the data available. These assignments also included limited
consideration of the QUALITY of the data, since it should not necessarily count against an
analysis if the analyst fails to collect data of poor quality. This consideration also is dealt with in
the assignment of Utility for the collected studies. Still, it is evident that the judgment of
Completeness might reasonably be expected to include consideration of something more than
the fraction of studies collected in a given data category/item. In the present analysis,
Completeness was increased slightly if the collected data for a given data category were the most
heavily cited in the secondary literature on formaldehyde. This leads to the following judgments
of Completeness:
For the data category BE1, the judgment for Completeness is HIGH since this was the
only study of this kind identified in the literature (because of the use of human cell lines).
For the data category HF1, the judgment for Completeness is HIGH since this was the
only study of this kind identified in the literature (because of the use of human cell lines).
178
-------
123. Judgments of Utility, Strength of Effect, and Exposure-Specific Effect
The remaining judgments in Working Table 2.14 refer to the qualities of the data
ACTUALLY EXAMINED IN THIS EXAMPLE. By Utility, we mean the quality of the study
DESIGN and its ability to address in rigorous fashion questions pertinent to hazard
identification for formaldehyde. No attempt is made to determine the specific inferences
towards which the data will be directed in later Working Tables, since the intent here is only to
determine the degree to which a body of data arose from a study design appropriate for
accurate and precise measurements of the quantity considered in the study. Utility is increased
by high standards of study design, including the ability of the experiment to demonstrate a
causal link between exposure to formaldehyde and the measured effect.
By Strength of Effect, we mean the degree to which the study actually revealed an effect.
Strength is increased when the effect noted possessed statistical significance.
By Exposure-Specific Effect, we mean a summary of the actual effect observed (putting
aside questions of the reliability of this effect, which was dealt with in the assignment of Utility
and Strength of Effect). This summarization is provided in the following text, and the data
themselves are provided in the various Exhibits appearing after Working Table 2.14.
The following judgments then apply to Working Table 2.14:
For the data category BE1, the Utility judgment is HIGH since multiple measures of
mutation frequency are reported; complete controls were used; CFE 50% level was utilized for
measurement of mutation frequency; the study constitutes a deliberate experiment (important
for causal claims); and mutation has been suggested as indicative of the ability of a substance to
induce DNA alterations. The Strength of Effect judgment is MEDIUM since although mutation
clearly is shown, and is reported as "significant", no information is provided on sample size and
179
-------
no measure of statistical significance is provided. The Exposure-Specific Effect is summarized in
Exhibit WT2.C14.BE 1.1. Formaldehyde exposure at a level of CFE 50% results in a 3-fold
increase in mutation frequency.
For the data category HF1, the Utility judgment is HIGH since the study included 2
exposure levels and 3 measures of DNA repair activity; this activity also was correlated with
other pathobiological and genotoxic effects of formaldehyde at the same exposure; methods and
materials are appropriate to this assay; control cultures were utilized; the study constitutes a
deliberate experiment (important for causal claims); and DNA repair is related clearly to the
ability of cells to repair damage leading to transitions to a cancerous state following induction of
biophysical effects. The Strength of Effect judgment is MEDIUM since although the effect is
clearly shown for some DNA, there is no information provided on number of samples assayed
or the statistical significance of results. The Exposure-Specific Effect is summarized in Exhibit
WT2.C14.HF1.1. Formaldehyde increasingly inhibits DNA repair in both compromised and
intact human cells as a function of increasing exposure.
12 A. Exhibits for Working Table 2.14
The data for the various data categories are provided as "Exhibits" after Working Table
2.14. The numbering system for the exhibits was chosen to display information on (in order of
appearance) the Working Table; the Context; the specific effect measured; and the study
number within that specific effect. For example, Exhibit WT2.C14.HF1.1 refers to Working
Table 2 (all exhibits have this assignment in common); Context 14; Host Factor of type 1; and
data set #1 within the data category HF1. This same numbering scheme will be employed in all
other Contexts.
180
-------
73. Description of Working Table 3.14
7.3.1. Introduction
The central task of Working Table 3.14 is to draw inferences on the carcinogenicity of
formaldehyde WITHIN CONTEXT 14, employing data specific to that context. This task
requires that several judgments be made sequentially:
(a) The analyst must determine the particular Claims of Carcinogenicity to be considered
in the analysis. These various claims are discussed in detail in Section 7.3.4.
(b) The available Relevance Strategies by which a given body of data may be related to a
specific Claim of Carcinogenicity must be developed. These strategies were discussed in
detail in Section 1.4, and the same definitions are applied here.
(c) The analyst must assign a measure of Intellectual Obligation to each Relevance
Strategy. This assignment was described in detail in Section 1.4, and the same
assignments are made here.
(d) The analyst must judge the epistemic status of a specific Claim of Carcinogenicity
arising from a specific Relevance Strategy. As described in Section 1.4, this judgment is
separate from the judgment of Intellectual Obligation for that Relevance Strategy. This
judgment is described in more detail in Section 7.3.2. The assignments are entered into
the appropriate cells of Working Table 3.14.
(e) The analyst must produce a summary judgment of the epistemic status of a particular
Claim of Carcinogenicity by combining the judgments from each of the five Relevance
Strategies for that claim. This summary includes consideration of both the epistemic
status of that claim for each Relevance Strategy, as entered into the separate cells of the
Working Table specific to each strategy, as well as the Intellectual Obligation assigned to
that strategy. The assignments are entered into the appropriate COLUMN SUMMARY
cells of Working Table 3.14.
(f) The analyst must produce a summary judgment of the COHERENCE of the various
Claims of Carcinogenicity. This task is performed by examining the coherence ACROSS
the various COLUMN SUMMARIES appearing in the Working Table. The assignments
are entered into the appropriate OVERALL SUMMARY cells of Working Table 3.14.
132. Epistemic Status of Claims of Carcinogenicity
181
-------
WORKING TABLE 3.14. INTRA-CONTEXT SUPPORT FOR CLAIMS OF CARCINOGENICITY IN CONTEXT 14
Claims of Carcinogenicity
Relevance Strategy
I.O.'
Increases
Incidence of
Transform.
Classification(s)
Stage
Mechanism
Complete
Partial
Mixer
Helper
Neo.
Conv.
Neo.
Devel.
Geno-
tox.
Non-
genot
ox.
Direct Empirical (D.E.)
HI
NO
NO
NO
NO
NO
NO
NO
HI
NO
Semi-Empirical
Extrapolation (S.E.E.)
ME
NO
NO
NO
NO
NO
NO
NO
NO
NO
Empirical Correlation
(E.C.)
ME
ME
NO
HI
NO
NO
ME
NO
HI
NO
Theory-based Inference
(T.B.I.)
ME
ME
NO
ME
NO
LO
ME
NO
HI
NO
Existential Insight(E.I.)
LO
LO
NO
LO
NO
ME
LO
NO
HI
NO
Column Summary
LO
NO
ME
NO
LO
ME
NO
HI
NO
Overall Summary
LO
NO
ME
NO
LO
ME
NO
HI
NO
'Intellectual Obligation
-------
The entries into the various cells of Working Table 3.14 reflect the judgment of the
analyst concerning the epistemic status of the Claim of Carcinogenicity being considered, FOR
A SPECIFIC RELEVANCE STRATEGY. This judgment will have two components:
(a) Each Relevance Strategy calls upon a selected body of data from which the reasoning
will proceed. Specifically, the Direct Empirical strategy requires data on direct measurements of
the effect underlying the Claim of Carcinogenicity, made at the exposure to formaldehyde
defining the context (here, concentrations in solution equivalent to airborne concentrations
above 2 ppm). The Semi-Empirical Extrapolation strategy requires data on direct measurements
of the effect underlying the Claim of Carcinogenicity, made at exposures to formaldehyde other
than those defining the context. The Empirical Correlation strategy requires data on effects
other than the effect underlying the Claim of Carcinogenicity, where this "other" effect is taken
to correlate with appearance of the effect of interest. The Theory-Based Inference strategy
requires data on an effect other than the effect underlying the Claim of Carcinogenicity, where
this "other" effect is taken to be a causal step in the production of the effect of interest as
specified by an established theory. The Existential Insight strategy does not invoke data, except
in the sense that a subjective judgment from an individual might be counted as an instance of
data in and of itself.
The analyst must assign the "foundational quality" to the data used in each relevance
strategy. By "foundational quality", we mean the quality of the data with respect to the question
those data were intended to address BY THE ORIGINAL INVESTIGATOR. This judgment is
based on the results of Working Table 2.14 and includes consideration of the Completeness,
Utility, Strength of Effect and Exposure-Specific Effect from that table. While the judgment
ultimately is subjective, it is constrained in that it must increase (from NO to LOW to
MEDIUM to HIGH) as the strength of the judgments for Completeness, Utility and Strength of
182
-------
Effect increase in Working Table 2.14. The summary of Exposure-Specific Effect may also enter
into the assignment of epistemic status to a Claim of Carcinogenicity if the analyst determines
that only effects above a given magnitude will exert an influence on carcinogenic processes (an
issue of importance when possible thresholds are considered).
(b) The analyst then must judge the strength of any "background premises" required by
the Relevance Strategy under consideration. These background premises are defined in detail in
the original report, but may be summarized as the set of (often implicit) premises that must be
introduced into an analysis if the data in Working Table 2.14 are to be used to draw inferences
about the Claims of Carcinogenicity for the context being considered. Presumably, many of these
background premises already are considered valid once Working Tables 1.13 and 2.13 are
complete; otherwise, the data would not have been assigned to this context. But other
background premises concerning the reliability of extrapolation procedures (for the Semi-
Empirical Extrapolation relevance strategy), the strength and specificity of correlations (for the
Empirical Correlation relevance strategy), the degree of verification and validation of theories
(for the Theory-Based Inference relevance strategy), and the credentials of individuals making
subjective judgments (for the Existential Insight relevance strategy) must be warranted prior to
their use.
Having established these two components of epistemic status (foundational quality and
background premises) for a given Claim of Carcinogenicity, the analyst then produces a
composite judgment of the epistemic status of a given Claim of Carcinogenicity and enters this
into the appropriate cell of Working Table 3.14. There are two separate claims that might be
made within one of these cells. The first is the claim that formaldehyde DOES produce the
effect underlying that claim (such as increased incidence of cancer or genotoxic effects). In that
case, the assignments of NO/LOW/MED/HIGH refer to the epistemic status of the claim that
183
-------
formaldehyde produces the effect. This assignment is entered into the area ABOVE the dashed
line in a cell. The second claim is that formaldehyde DOES NOT produce the effect underlying
the claim. In that case, the assignment of NO/LOW/MED/HIGH refer to the epistemic status
of the claim that formaldehyde DOES NOT produce the effect. This assignment is entered into
the area BELOW the dashed line in a cell. THE READER SHOULD NOTE THAT THIS IS A
NEW FEATURE OF THESE WORKING TABLES NOT DISCUSSED IN THE ORIGINAL
REPORT. The reason for this addition to the Working Tables is that the fact that a particular
body of data does not support the claim that formaldehyde does induce a specific effect IS NOT
NECESSARILY evidence that formaldehyde does not induce that effect. It may simply be the
case that the data were obtained under conditions incapable of revealing any effect that might
be present.
This consideration is important due to the issue of "burden of proof'. At times, the
burden may lie on the regulatory agency to show that formaldehyde DOES induce cancer.
Attention then would be directed towards the judgments in the upper halves of the various cells
of Working Table 3.14. At other times, the burden may lie on the producer of formaldehyde to
show that formaldehyde DOES NOT induce cancer. Attention then would be directed towards
the judgments in the lower halves of the various cells of Working Table 3.14. IN THE
PRESENT EXAMPLE, ONLY JUDGMENTS THAT AN EFFECT IS PRODUCED BY
FORMALDEHYDE ARE DEVELOPED. THE PROCESS OF ANALYSIS WOULD BE
REPEATED FOR THE LOWER HALVES OF ALL CELLS IF THE BURDEN OF PROOF
WAS SHIFTED.
It is important here to consider the broad features of the theory of transformation to be
used throughout this section. These features structure the interpretation of data used in the
various Relevance Strategies. The theory of transformation employed here is that cells are
184
-------
transformed by two steps or transitions. These are termed initiation and promotion, each of
which may require one or more sub-steps. The initiation step is taken to involve alteration of
DNA structure and function, producing a new genome. This alteration must then remain intact
without being repaired by processes normally operating in cells or induced by the damage itself.
The ability of a substance to bring a stable alteration about may be determined from an increase
in transformation following (1) application of the initiating agent and (2) subsequent application
of a promoting agent such as TPA. In the absence of such data, the ability of a substance to
interact with DNA, to alter DNA, and/or to induce mutation provides partial but not conclusive
evidence of initiating activity. In addition, the ability of a substance to lower DNA repair rates
or fidelity of repair is partial but not conclusive evidence that the substance will enhance the
initiating activity of other substances or of background events.
The promotion step is taken to involve loss of growth control, represented by an
imbalance between rates of growth and death of cells in a colony and a continued net rate of
expansion even upon reaching confluence. The mechanism for this step is unspecified, but may
result in either a lowering of the rate of removal of cells (as in decreased apoptosis) or an
increase in the rate of division (as in abnormal proliferation). The ability of a substance to bring
this about may be determined from an increase in transformation following (1) prior application
of a known initiating agent and (2) subsequent application of the substance of interest. In the
absence of such data, the ability of a substance to alter the rate of cellular division, the rate of
apoptosis, and/or net rate of growth provides partial but not conclusive evidence of promoting
activity.
If a substance induces only one of the required transitions, it will be necessary for other
transitions to be induced by antecedent conditions if the substance of interest is to be
considered a Partial transforming agent. The fact that the background frequency of
185
-------
transformation is not zero may be taken to indicate that each of these other transitions will
always be present. If this assumption is adopted, the ability of a substance to increase the
frequency of transformation cannot be used to differentiate between claims of a Complete or a
Partial transforming agent. Existing empirical evidence, however, is suggestive of there being
several routes by which transformation may occur. It is not inconsistent to assume that a
substance acts to induce only a subset of transitions, and that the remaining transitions required
are not those associated with the mechanism by which the antecedent conditions of the
experimental context act to bring about background transformation.
The epistemic status of this Theory of Transformation is taken here to be MED based
on an examination of the available empirical evidence linking DNA alteration, alteration of rates
of growth, and the transforming potential of a range of substances tested in assays for initiation,
promotion and transformation. A judgment of HIGH is not considered appropriate since much
of the warrant for the theory is subjective judgment of experts in the field based on the
coherence of the explanatory framework and on limited experimental tests of key assumptions.
7.3.3. Intellectual Obligation
The issue of Intellectual Obligation was discussed in detail in Section 1.4. Since the
judgments and assignments are common to all Working Tables and Contexts, this discussion is
not repeated here. By way of a summary, the Direct Empirical Relevance Strategy was assigned
an Intellectual Obligation of HIGH; the Semi-Empirical Extrapolation, Theory-Based Inference
and Empirical Correlation Relevance Strategies were assigned an Intellectual Obligation of
MEDIUM; and the Existential Insight Relevance Strategy was assigned an Intellectual
Obligation of LOW. The assignments are entered into the appropriate cells of Working Table
3.14.
186
-------
7 J.4. Claims of Carcinogenicity
The framework employed here allows for a wide variety of claims concerning the
carcinogenic action of formaldehyde. In the original report, it was determined that it is desirable
to augment the claim that a substance increases the incidence of cancer, since this sole claim
does not provide full detail on the antecedent conditions under which a substance such as
formaldehyde can exert its carcinogenic potential. It was determined that a more detailed
analysis would subdivide the claim of increased incidence of cancer into more detailed claims
potentially of use in determining the most effective regulatory strategy. The different Claims of
Carcinogenicity employed in Working Table 3.14 (and all subsequent Working Tables discussed
in this report) are:
(a) Increases Incidence of Transformation, implying that formaldehyde raises the
incidence of transformation.
(b) Complete Carcinogen, implying that formaldehyde increases the incidence of
transformation even when it is not delivered in conjunction with concurrent exposures.
(c) Partial Carcinogen, implying that formaldehyde increases the incidence of
transformation only when delivered in conjunction with concurrent exposures that
complete the process of carcinogenesis.
(d) Mixer, implying that formaldehyde exerts its carcinogenic effect only when it is
combined with another substance, and where this combination results in a new substance
which is at least a partial carcinogen.
(e) Helper, implying that formaldehyde does not in and of itself induce transitions to
transformation, but that it facilitates the carcinogenic action of another substance.
(f) Neoplastic Conversion, implying that formaldehyde acts by converting cells to
neoplastic potential.
(g) Neoplastic Development, implying that formaldehyde acts by producing growth in
colonies of cells already having neoplastic potential.
(h) Genotoxicity, implying that formaldehyde acts through a mechanism involving
interaction with the genetic material of a cell.
(i) Non-genotoxicity, implying that formaldehyde acts through a mechanism other than
interaction with the genetic material of a cell.
187
-------
The following summary considerations from Working Table 2.14 provide the foundations
for all inferences to Claims of Carcinogenicity drawn in this context:
BE1.1 demonstrates that in vitro exposure to formaldehyde of a colony of cells derived
from human tissue lowers the repair of damage induced in DNA. The statement that this
alteration in repair effectiveness is dose-dependent and caused by the action of formaldehyde or
a metabolite is assigned an epistemic status of MED since, as demonstrated in Working Table 2
for this context, the utility of the study is HIGH but the strength of effect is MED due to the
lack of reported confidence limits (from which statistical significance might be determined) and
the fact that the repair parameters measured did not change dramatically in the exposed cells.
While the results appear significant, this lack of statistical rigor in a setting where the change in
repair was not large is worrying and lowers the assigned epistemic status from what would have
been an assignment of HIGH if statistical significance had been determined.
BE2.1 demonstrates that in vitro exposure to formaldehyde of a colony of cells derived
from human tissue increases the frequency of mutation and induces alterations in DNA
structure. The statement that this increase is dose-dependent and caused by the action of
formaldehyde is assigned an epistemic status of MED since, as demonstrated in Working Table
2 for this context, the utility of the study is HIGH but the strength of effect is MED due to the
lack of reported confidence limits (from which statistical significance might be determined).
While the results appear significant, this lack of statistical rigor is worrying and lowers the
assigned epistemic status from what would have been an assignment of HIGH if statistical
significance had been determined.
The separate judgments for Claims of Carcinogenicity for Working Table 3.14 are
provided in the sections that follow. In the case of each claim, explanations are given of the
188
-------
judgment for each Relevance Strategy; then for the Column Summary across Relevance
Strategies for that claim; then for the Overall Summary across claims.
7.3.4.1. Increases Incidence of Transformation
The judgments for this Claim of Carcinogenicity for the various Relevance Strategies and
for this context are shown in the cells of Working Table 3.14. The basis for these judgments are
given below.
A. Direct Empirical: The judgment is NO since transformation was not measured.
B. Semi-Empirical Extrapolation: The judgment is NO since transformation was not
measured in either the concentration range of interest nor one from which an extrapolation
could be drawn.
C. Empirical Correlation: The judgment is MED since BE2 (mutation frequency)
indicates that formaldehyde induces mutations in these cells, and mutation is taken to correlate
(with medium strength and specificity) with transformation. This correlation could have been
strengthened if statistical rigor on
the mutation frequencies had been included (as described just above).
D. Theory-Based Inference: The judgment is MED since BE2 indicates that
formaldehyde induces mutations in cells, and the analysts are working under the theory that
mutation of DNA is a necessary step in at least the initiation stage of cell transformations. In
addition, BE1 indicates that formaldehyde slightly inhibits DNA repair, suggesting (rather
weakly) that formaldehyde may be capable of bringing about initiation through inhibition of
repair normally directed at background DNA damage. Under the theory of transformation
employed here, such inhibition of repair would be expected to lead to increased transformation.
This study does not, however, establish that the form of mutation measured is clearly identical
189
-------
to the kind of mutation associated with the initiation step of transformation; nor does it
establish that the necessary transitions other than initiation are present in the cell line. The
judgment of MED is marginal, and the analysts also would have found it reasonable to assign an
epistemic status of LOW.
E. Existential Insight: The judgment is LOW since these data produce in the analysts
weak feelings of support for the claim that they are indications of transformation.
F. Column Summary: The judgment is LOW since Direct Empirical data are not
available, and this warrant carries high intellectual obligation. Furthermore, although the
Empirical Correlation and Theory-Based Inference warrants are of MED epistemic status, they
are very low MED scores: the Empirical Correlation data are thin and not supported by
statistical measures; the data warranted with a Theory-Based Inference strategy are also thin,
and do not indicate direct support for the central theory used in this justification.
G. Overall Summary: The judgment is LOW since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim. If, however, such a stance were not taken, the analysts' still would
leave the judgment as it stands in the Column Summary since the epistemic status of no other
claim is above a low MED, and NO Direct Empirical evidence is available.
7.3.4.2. Classifications
7.3.4.2.1. Complete Transformation
The judgment in all cells is NO since transformation was not measured and it is not
judged proper to establish a correlation between mutation ability and the ability of a substance
to act as a complete transformation agent. At best, this correlation may establish a substance as
a partial transformation agent, as reflected in other cells of this Working Table.
190
-------
7.3.4.2.2. Partial Transformation
A. Direct Empirical: The judgment is NO since the study did not measure directly either
the initiation or promotion transitions leading to transformation.
B. Semi-Empirical Extrapolation: The judgment is NO since the study did not measure
directly either the initiation or promotion transitions leading to transformation.
C. Empirical Correlation: The judgment is HIGH since BE2 (mutation frequency)
indicates that formaldehyde induces mutations in these cells, and mutation is taken to correlate
(high strength and specificity) with initiation activity.
D. Theory-Based Inference: The judgment is MED since BE2 indicates that
formaldehyde induces mutations in cells, and the analysts are working under the theory that
mutation of DNA is a necessary step in at least initiation of cells. The study does not, however,
establish that the form of mutation measured is clearly identical to the kind of mutation
associated with the initiation step of transformation. The judgment of MED is marginal, though,
and the analysts also would have found it reasonable to assign an epistemic status of LOW.
E. Existential Insight: The judgment is LOW since these data produce in the analysts
only very weak confidence that some potential for transformation might be indicated.
F. Column Summary: The judgment is MED since Direct Empirical data are not
available, and these carry high intellectual obligation. Empirical Correlation is HIGH, however,
and there is a MED judgment of Theory-Based Inference.
G. Overall Summary: The judgment is MED since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim. If, however, such a stance were not taken, the analysts' still would
leave the judgment as it stands in the Column Summary since although the epistemic status of
191
-------
one other claim (Empirical Correlation) is above MED, its status applies to only one of the two
outcomes (BE2, mutation frequency) and is only thinly established in the data.
7.3.4.2.3. Mixer
The judgment in all cells is NO since transformation was not measured either for
formaldehyde alone or in conjunction with other substances. The data do not provide any
indication of whether or not formaldehyde requires the presence of other agents to exert its
effect at any stage of transformation.
7.3.4.2.4. Helper
A. Direct Empirical: The judgment is NO since the studies did not examine the ability of
formaldehyde to alter the transforming potential of any other agent.
B. Semi-Empirical Extrapolation: The judgment is NO since the studies did not examine
the ability of formaldehyde to alter the transforming potential of any other agent.
C. Empirical Correlation: The judgment is NO since there are no data from which to
develop a measure of the required correlation.
D. Theory-Based Inference: The judgment is LOW since BE1 indicates that
formaldehyde may lower DNA repair, which would tend (under the theory used here) to lower
the ability of cells to repair DNA damage induced by other agents. It is not established,
however, that the KIND of DNA repair inhibited by formaldehyde is identical to that operating
for other agents, and the strength of effect noted in BE1 is only MED.
E. Existential Insight: The judgment is MED since the BE1 data produce in the analysts
moderate confidence that there is some potential for formaldehyde to act as a Helper.
192
-------
F. Column Summary: The judgment is LOW since only Theory-Based Inference and
Existential Insight warrants are available, with the TBI warrant being Low and the EI warrant
being MED (but with LOW Intellectual Obligation).
G. Overall Summary: The judgment is LOW since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim. If, however, such a stance was not taken, the analysts still would leave
the judgment as it stands in the Column Summary since the epistemic status of only one other
claim is above LOW, and that (Existential Insight) carries little Intellectual Obligation.
7.3.4.3. Stages
7.3.4.3.1. Neoplastic Conversion
A. Direct Empirical: The judgment is NO since the BE2 study did not examine the
ability of formaldehyde to induce neoplastic conversion.
B. Semi-Empirical Extrapolation: The judgment is NO since the BE2 study did not
examine the ability of formaldehyde to induce neoplastic conversion.
C. Empirical Correlation: The judgment is HIGH since BE2 (mutation frequency)
indicates that formaldehyde induces mutations in these cells, and mutation is taken to correlate
(high strength and specificity) with the ability of an agent to act through neoplastic conversion.
D. Theorv-Based Inference: The judgment is MED since BE2 indicates that
formaldehyde induces mutations in cells, and the analysts are working under the theory that
mutation of DNA is a necessary step in neoplastic conversion of cells. The study does not,
however, establish that the form of mutation measured is clearly identical to the kind of
mutation associated with the neoplastic conversion step of transformation. It does, however,
193
-------
indicate that formaldehyde is at least capable of inducing several forms of neoplastic conversion
associated with mutation.
E. Existential Insight: The judgment is LOW since these data produce in the analysts
only very weak confidence that some potential for transformation exists.
F. Column Summary: The judgment is MED since Direct Empirical data are not
available, and these carry high intellectual obligation. Empirical Correlation is HIGH, however,
and there is a MED judgment of Theory-Based Inference.
G. Overall Summary: The judgment is MED since the analysts'
strongly foundationaList epistemological stance suggests that coherence across claims should not
affect the epistemic status of any given claim. If, however, such a stance was not taken, the
analysts still would leave the judgment as it stands in the Column Summary since the epistemic
status of other claims is not HIGH.
7.3.4.3.2. Neoplastic Development
The judgment in all cells is NO since the analysts do not believe that either body of data
is relevant to the issue of neoplastic development.
7.3.4.4. Mechanisms
7.3.4.4.1. Genotoxic
A. Direct Empirical: The judgment is HIGH since the BE2 study data indicate strongly
that formaldehyde induces a variety of DNA damage in cells, including cross-links and strand
breaks, at the same time that it induces mutation. These data cannot be taken to indicate that
formaldehyde induces the kinds of genotoxic effects necessary for transformation, but the issue
here is only whether formaldehyde is able to produce effects on DNA.
194
-------
B. Semi-Empirical Extrapolation: The judgment is NO since the data reported here are
at the exposure levels defining the context.
C. Empirical Correlation: The judgment is HIGH since the effects reported here
(mutation, DNA breaks, etc) correlate (with high strength and specificity) with genotoxic action.
D. Theory-Based Inference: The judgment is HIGH since under the theory employed
here, genotoxic action arises from the ability of an agent to induce alterations in DNA structure
and function, which has been clearly established here. In addition, the underlying theory is taken
to be well established empirically.
E. Existential Insight: The judgment is HIGH since the analysts are confident that these
data represent the ability of formaldehyde to induce genotoxic effects (which is the issue here;
see above) that may then provide a path to transformation.
F. Column Summary: The judgment is HIGH since strong Direct Empirical data are
available, and these carry HIGH Intellectual Obligation. Also, HIGH epistemic status is assigned
to the Empirical Correlation, Theory-Based Inference, and Existential Insight warrants, thus
supporting this determination, though with lower Intellectual Obligations.
G. Overall Summary: The judgment is HIGH since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim. If, however, such a stance was not taken, the analysts still would leave
the judgment as it stands in the Column Summary since the epistemic status of several claims is
HIGH.
7.3.4.4.2. Non-Genotoxic
The judgment in all cells is NO since the analysts do not believe that either body of data
is relevant to the issue of non-genotoxic action.
195
-------
73.5. Summary Comments for Intm-Context Claims
The entries into the cells of Working Table 3.14 indicate that formaldehyde should be
slightly considered capable of inducing transformation in human cell lines when exposed in vitro
to formaldehyde in solution at concentrations representative of airborne exposures at or above 2
ppm. This judgment is reflected most clearly in the fact that the "Overall Summary" cell for
"Increases Incidence of Transformation" is judged to be LOW due to the lack of presence of
Direct Empirical data showing transformation (although the presence of mutation is related and
suggestive). The data also support strongly the claim that formaldehyde acts through at least a
genotoxic mechanism in this context, and support moderately the claim that formaldehyde
induces neoplastic conversion. The data also supply moderate support for the claim that
formaldehyde is a partial carcinogen in this context, at least when an initiation assay is
performed.
7.4. Description of Working Table 4.14
7.4.1. Introduction
Working Table 4.14 is utilized to justify premises necessary for extrapolation of the
findings in Working Table 3.14 (the "Observational Context") to the Target Context" (Context
12). Four separate claims must be warranted, each of which is described in detail in the original
report and in lesser detail below. The data on which judgments for these premises are based
were described in Section 7.3.4. All judgments of Intellectual Obligation are as described in
Section 1.4.
7.4.2. Column Headings in Working Table 4.14
196
-------
WORKING TABLE 4A4. SUPPORT FOR INTER-CONTEXT EXTRAPOLATION PREMISES
CONTEXT NO. 14 TO CONTEXT NO. 12
Relevance Strategy
I.O.1
Exposure to BSDR
Conversion
BSDR to Effect
Conversion
Host Factors
Environmental
Conditions
Direct Empirical (D.E.)
HI
NO
NO
LO
LO
Semi-Empirical Extrapolation
(S.E.E.)
ME
NO
NO
NO
NO
Empirical Correlation (E.C.)
ME
LO
ME
ME
ME
Theory-based Inference (T.B.I.)
ME
ME
ME
ME
ME
Existential Insight (E.I.)
LO
ME
ME
LO
ME
Overall Assessment
LO
ME
LO
ME
'Intellectual Obligation
-------
These extrapolation premises or claims as they appear in Working Table 4.14 are:
1. EXPOSURE TO BSDR: Any differences between the BIOLOGICAL components of
the two contexts with respect to the relationship between exposure and BSDR are not so large
as to call into question whether a BSDR might be present in one context but not the other AT
THE SAME LEVEL OF EXPOSURE.
The pertinent considerations here for each context are whether (1) an exposure occurs
through some compartment of the environment (a compartment refers to the medium of the
environment, such as air or water); (2) an intake by some route can occur (a route refers to
inhalation, ingestion, dermal absorption, intubation or injection); (3) this intake results in an
uptake to a target organ or tissue; (4) clearance from the target allows the production of a
burden; and (5) biological transformation to the active metabolite exists. Where extrapolation is
across exposure levels, it is particularly important to establish that there does not exist a
threshold exposure below which a BSDR is not produced (due, e.g., to necessary saturation of
metabolic transformation). Where extrapolation is across species, it is particularly important to
show that any such threshold, if it exists, does not differ between the species in a manner calling
into question this premise.
2. BSDR TO BIOPHYSICAL EFFECT: Any differences between the BIOLOGICAL
components of the two contexts with respect to the relationship between BSDR and the
production of biophysical effects necessary for transitions are not so large as to call into
question whether a biophysical effect might be present in one context but not the other
ASSUMING THE SAME LEVEL OF BSDR IN BOTH CONTEXTS.
The pertinent considerations here for each context are whether (1) the active metabolite
can interact with causally important biological structures in the organism; (2) this interaction
results in either DNA alterations (for initiation) or stimulation of cellular division (for
197
-------
promotion); and (3) these biophysical effects are of the type and magnitude necessary to induce
transitions designated as either neoplastic development or neoplastic conversion. Where
extrapolation is across exposure levels, it is particularly important to establish that there does
not exist a threshold BSDR below which a biophysical effect is not produced (due, e.g., to
necessary saturation of a pool of DNA damage or a stimulus to cellular proliferation). Where
extrapolation is across species, it is particularly important to show that any such threshold, if it
exists, does not differ between the species in a manner calling into question this premise.
3. ENVIRONMENTAL CONDITIONS: Any differences between the
ENVIRONMENTAL components of the two contexts with respect to the relationship between
exposure (to the substance of interest) and BSDR, or the relationship between BSDR and the
production of biophysical effects necessary for transitions, are not so large as to call into
question whether a biophysical effect might be present in one context but not the other AT THE
SAME LEVEL OF EXPOSURE TO THE SUBSTANCE OF INTEREST
The pertinent considerations here for each context are whether (1) environmental
conditions increase or decrease the BSDR resulting from a given level of exposure to the
substance of interest, relative to the conditions that exist in the target context; (2) environmental
conditions increase or decrease the probability of biophysical effects from a given level of
BSDR, relative to the target context; and (3) environmental conditions increase or decrease the
probability of transitions from a given level of biophysical effect, relative to the target context.
4. HOST FACTORS: Any differences between the BIOLOGICAL components of the
two contexts with respect to the relationship between the production of biophysical effects, the
probability of transitions necessary for cancer, and the probability of cancer itself are not so
large as to call into question whether an increase in cancer incidence might be present in one
context but not the other AT THE SAME LEVEL OF BIOPHYSICAL EFFECT.
198
-------
The pertinent considerations here for each context are whether (1) a target organ or
tissue exists; (2) biological structures on which relevant biophysical effects are produced exist;
(3) relevant background transitions exist; and (4) repair of biophysical effects, or regression of
states of neoplastic conversion or neoplastic development, exist. Where extrapolation is across
species, it is particularly important to establish that there does not exist a threshold level of
biophysical effect below which a transition is not produced (due, e.g., to necessary saturation of
a stimulus to cellular proliferation before promotion can occur).
In the original report, Working Table 4 (bearing in mind that Working Table 4 in the
present report was numbered Working Table 5 in the original report, as described in Section 1)
also contained an extrapolation premise concerning the role of intersubject and intrasubject
variability. This premise has been removed here and consideration of the role of differences in
intra- and intersubject variability between the two contexts has been incorporated into the 4
separate premises above. This choice was made because variability always refers to variability of
some PROPERTY, and this property would appear in one of the four extrapolation categories
listed above. The analyst then must determine the significance of variability on the relationship
between EXPOSURE AND BSDR, BSDR AND BIOPHYSICAL EFFECT,
ENVIRONMENTAL CONDITIONS, AND HOST FACTORS separately.
7.4 J. Epistemic Status of Extrapolation Premises
This section summaries the judgments made for each of the cells in Working Table 4.14.
The procedure for assigning epistemic status to a cell is identical to that outlined for Working
Table 3.14 and a description is not repeated here.
7.4.3.1. Exposure to BSDR
199
-------
A. Direct Empirical: The judgment is NO since BSDR was not determined in either
context at any level of exposure.
B. Semi-Empirical Extrapolation: The judgment is NO since BSDR was not determined
in either context at any level of exposure.
C. Empirical Correlation: The judgment is LOW since exposure to formaldehyde (which
is well established in both contexts) is taken to correlate with moderate strength with production
of a dose, but metabolic transformation has not been identified in the observational context.
Where the active metabolite has not been identified, the correlation between exposure and
BSDR (rather than dose) will be weakened. In addition, it has not been demonstrated that there
is not a threshold exposure for production of a BSDR or that this threshold is the same between
contexts.
D. Theory-Based Inference: The judgment is MED since exposure clearly occurs in both
contexts; intake clearly occurs for the observational context and the presence of inhalation
strongly suggests that intake occurs in the target context, although at a lower level than in the
observational context; uptake clearly occurs for the observational context and the absorption of
formaldehyde in nasal passages upon inhalation in humans has been established with MED
epistemic status; clearance is present but not infinitely rapid in both contexts; but the active
metabolite has not been shown present in either context. While the theory of pharmacodynamics
suggests that a dose should be produced in both contexts, it is not clear that a BSDR will be
produced. In addition, it has not been demonstrated that there is not a threshold exposure for
production of a BSDR or that this threshold is the same between contexts.
E. Existential Insight: The judgment is MED since these data produce in the analysts
moderate confidence that a BSDR is produced in both contexts.
200
-------
F. Overall Summary: The judgment is LOW since only Existential Insight and Theory-
Based Inference were of MED strength, but neither has HIGH Intellectual Obligation.
7.4.3.2. BSDR to Effect
A. Direct Empirical: The judgment is NO since the same biophysical effect was not
determined in these contexts at any level of exposure.
B. Semi-Empirical Extrapolation: The judgment is NO since the same biophysical effect
was not determined in these contexts at any level of exposure.
C. Empirical Correlation: The judgment is MED since the ability of the active metabolite
of a substance to induce biophysical effects in vitro UPON PRODUCTION OF A BSDR is
taken to correlate with moderate strength with the ability of that same active metabolite to
induce biophysical effects in vivo. The judgment is not HIGH since it has not been demonstrated
that there is not a threshold BSDR for production of a biophysical effect or that this threshold
is the same between contexts.
D. Theory-Based Inference: The judgment is MED since similar cellular structures exist
in both contexts; formaldehyde or its active metabolite is assumed to diffuse in a similar fashion
in both contexts; and interaction with pertinent biological structures should occur through this
diffusion. The judgment is not HIGH since it has not been demonstrated that there is not a
threshold BSDR for production of a biophysical effect or that this threshold is the same between
contexts.
E. Existential Insight: The judgment is MED since these data produce in the analysts
moderate confidence that interaction between formaldehyde (or its active metabolite) and
pertinent biological structures is produced in both contexts.
201
-------
F. OveraU Summary: The judgment is MED since Empirical Correlation, Theory-Based
Inference and Existential Insight are MED, while the Intellectual Obligation for Empirical
Correlation and Theory-Based Inference is MED. The lack of Direct Empirical support prevents
this judgment from being HIGH.
7.4.3.3. Host Factors
A. Direct Empirical: The judgment is LOW since while human cells are present in
Context 14, the target for Context 12 has been observed to be the specific cells of the nasal
passages; the pertinent structures for production of biophysical effects were not observed
directly; background transitions clearly were present in some form since the background
incidence of effect was non-zero in both contexts; the effect of repair was not observed. The
judgment is not MED since it has not been demonstrated that there is not a threshold
biophysical effect for production of transitions or that this threshold is the same between
contexts.
B. Semi-Empirical Extrapolation: The judgment is NO since the consideration of host
factors does not require extrapolation across exposure levels, and was dealt with in A above.
C. Empirical Correlation: The judgment is MED since the ability of a substance to
induce transformation in vitro correlates with moderate strength with the ability to induce
carcinogenesis in humans, suggesting that there also is a correlation in the necessary host
factors. The judgment is not HIGH since it has not been demonstrated that there is not a
threshold biophysical effect for production of transitions or that this threshold is the same
between contexts.
D. Theory-Based Inference: The judgment is MED based on the fact that cellular
structure and function is similar in the two contexts since both involve human cells, so similar
202
-------
targets should exist; the target in Context 12 is taken to be the nasal passages, which clearly
exist; repair should be present but not completely effective in both contexts; and background
transitions are taken to occur in both contexts as evidenced by the non-zero rates of effect in
suitable controls. The judgment is not HIGH since it has not been demonstrated that there is
not a threshold biophysical effect for production of transitions or that this threshold is the same
between contexts.
E. Existential Insight: The judgment is LOW since the biological differences between the
two contexts produce in the analysts moderate confidence that the two contexts are similar.
F. Overall Summary: The judgment is LOW since Empirical Correlation and Theory-
Based Inference are MED, while the Intellectual Obligations for these are MED. The LOW
judgment for Direct Empirical lowers this overall judgment.
7.4.3.4. Environmental Conditions
A. Direct Empirical: The judgment is LOW since the exposure was clearly to
formaldehyde in Context 13, but exposures other than to formaldehyde will be present in
Context 12.
B. Semi-Empirical Extrapolation: The judgment is NO since this warrant does not apply
for this extrapolation premise.
C. Empirical Correlation: The judgment is MED since the ability of a substance to
induce transformation in vitro correlates with moderate strength with the ability to induce
carcinogenesis in humans, suggesting that there also is a correlation in the necessary
environmental conditions.
D- Theory-Based Inference: The judgment is MED based on the theory that the
presence of substances other than formaldehyde in Context 12 would primarily enhance the
203
-------
transitions induced by only formaldehyde in Context 13; this theory is given MED epistemic
status.
E. Existential Insight: The judgment is MED since the environmental differences
between the two contexts produce in the analysts moderate confidence that the two contexts are
similar.
F. Overall Summary: The judgment is MED since Empirical Correlation and Theory-
Based Inference are MED, while the Intellectual Obligations for these are MED. The presence
of some Direct Empirical support would tend to enhance this epistemic status, but the judgment
for that warrant was LOW.
7.5. Description of Working Table 5.14
7.5.1. Introduction
Working Table 5.14 contains judgments of the epistemic status for Inter-Context Claims
of Carcinogenicity. In each entry, the "Observational Context" is Context 14 and the 'Target
Context" is Context 12. The judgments in Working Table 5.14 utilize the judgments in Working
Table 3.14 for a particular Claim of Carcinogenicity, as well as the judgments in Working Table
4.14 for the necessary Extrapolation Premises. In each case of an extrapolation, it is the "Overall
Summary" judgment associated with a Claim of Carcinogenicity in Working Table 3.14 that
forms the basis for the extrapolation. The final judgment of epistemic status in Working Table
5.14 cannot be higher than this initial judgment from Working Table 3.14, since the application
of the extrapolation premises can have only the effect of weakening the extrapolation. In other
words, if formaldehyde is judged to "Increase the Incidence of Cancer" with MEDIUM epistemic
status in Working Table 3.14, the highest judgment of epistemic status for the instance of
204
-------
WORKING TABLE 5.14. INTER-CONTEXT SUPPORT FOR CLAIMS OF CARCINOGENICITY
EXTRAPOLATED FROM CONTEXT 14 TO CONTEXT 12
Claims of Carcinogenicity
Relevance Strategy
I.O.1
Increases
Incidence of
Cancer
Classification(s)
Stage
Mechanism
Complete
Partial
Mixer
Helper
Neo.
Conv.
Neo.
Devel.
Geno-
tox.
Non-
genot
ox.
Direct Empirical (D.E.)
HI
NO
NO
NO
NO
NO
NO
NO
NO
NO
Semi-Empirical
Extrapolation (S.E.E.)
ME
NO
NO
NO
NO
NO
LO
NO
ME
NO
Empirical Correlation
(EC.)
ME
LO
NO
ME
NO
NO
ME
NO
ME
NO
Theory-based Inference
(T.B.I.)
ME
LO
NO
ME
NO
NO
ME
NO
ME
NO
Existential lnsighl(E.I.)
LO
ME
NO
ME
NO
NO
ME
NO
ME
NO
Column Summary
LO
NO
LO
NO
NO
ME
NO
ME
NO
Overall Summary
LO
NO
LO
NO
NO
ME
NO
ME
NO
'Intellectual Obligation
-------
extrapolation will be MEDIUM, which presumes that the extrapolation premises are judged to
each have HIGH epistemic status in Working Table 4.14.
The analysts have chosen to view warrants from Working Table 3.14 in the following
manner. A particular cell from that table may be used as Empirical Correlation, Theory-Based
Inference and/or Existential Insight relevance strategies in Working Table 5.14 for the same
Claim of Carcinogenicity. A particular claim from Working Table 3.14 may NOT be used as a
Direct Empirical relevance strategy in Working Table 5.14, since there always exists
extrapolation across exposure levels. A particular claim from Working Table 3.14 MAY be used
as a Semi-Empirical Extrapolation relevance strategy if it is judged that the biological and
environmental conditions in the two contexts are sufficiently similar to warrant the claim that
these two contexts contain approximately the same biological and environmental antecedent
conditions, differing only in LEVEL of exposure to formaldehyde. This clearly requires relatively
strong epistemic status to each of the four extrapolation premises found in Working Table 4.14,
a condition only weakly met in the present case.
7.5.2. Intellectual Obligation
The assignments of Intellectual Obligation discussed in Section 1.4, and utilized in
Working Table 3.11, are employed in Working Table 5.14.
7.53. Claims of Carcinogenicity
The same grouping of Claims of Carcinogenicity used as headings in Working Table 3.14
are employed in Working Table 5.14. The judgments of the epistemic status of each of these
claims, for each relevance strategy, are provided in the cells of Working Table 5.14 and are
described below.
205
-------
7.5.3.1. Increases Incidence of Cancer
A. Direct Empirical: The judgment is NO due to differences in level of exposure and
biological context.
B. Semi-Empirical Extrapolation: The judgment is NO due to LOW epistemic status of
necessary extrapolation premises and differences in context (which are not offset by LOW
strength of extrapolation premises).
C. Empirical Correlation: The judgment is LOW since the claim was given LOW
epistemic status in Working Table 3.14 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.14. While a correlation does exist between in
vitro transformation and carcinogenicity, the LOW epistemic status of the claim in Working
Table 3.14 keeps this judgment from being MED. The fact that Genotoxicity was judged HIGH
in Working Table 3.14 could elevate the epistemic status above LOW, since genotoxic
mechanisms tend to be less context-specific than non-genotoxic mechanisms, but the LOW
epistemic status of the claim of transformation in Working Table 3.14 precludes this.
D. Theory-Based Inference: The judgment is LOW since the claim was given LOW
epistemic status in Working Table 3.14 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.14. While the Theory of Carcinogenesis does
support the claim that in vitro transformation may indicate a transition necessary for
carcinogenesis, the LOW epistemic status of the claim in Working Table 3.14 and several of the
extrapolation premises keeps this judgment from being MED. The fact that Genotoxicity was
judged HIGH in Working Table 3.14 could elevate the epistemic status above LOW, since
genotoxic mechanisms tend to be less context-specific than non-genotoxic mechanisms, but the
LOW epistemic status of the claim of transformation in Working Table 3.14 precludes this.
206
-------
E. Existential Insight: The judgment is MED since the evidence produces in the analysts
moderate confidence that formaldehyde is a carcinogen in Context 12.
F. Column Summary: The judgment is LOW since Direct Empirical and Semi-Empirical
Extrapolation were NO (with HIGH Intellectual Obligation), and the remainder were LOW
(with LOW to MED intellectual obligation).
G. Overall Summary: The judgment is LOW since the analysts' strongly foundational
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
7.5.3.2. Classifications
7.5.3.2.1. Complete Carcinogen
The judgment in all cells is NO since the same judgment is given in Working Table 3.14
(for the reasons stated there).
7.5.3.2.2. Partial Carcinogen
A. Direct Empirical: The judgment is NO due to differences in level of exposure and
biological context.
B. Semi-Empirical Extrapolation: The judgment is NO due to LOW epistemic status of
necessary extrapolation premises and differences in context (which are not offset by LOW
strength of extrapolation premises).
C. Empirical Correlation: The judgment is MED since the claim was given MED
epistemic status in Working Table 3.14 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.14. While a correlation does exist between in
vitro transformation and carcinogenicity, the LOW epistemic status of the claim in Working
207
-------
Table 3.14 keeps this judgment from being HIGH. The fact that Genotoxicity was judged HIGH
in Working Table 3.14 elevates the epistemic status above LOW, since genotoxic mechanisms
tend to be less context-specific than non-genotoxic mechanisms.
D. Theorv-Based Inference: The judgment is MED since the claim was given MED
epistemic status in Working Table 3.14 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.14. While the Theory of Carcinogenesis does
support the claim that in vitro transformation may indicate a transition necessary for
carcinogenesis, the MED epistemic status of the claim in Working Table 3.14 and several of the
extrapolation premises keeps this judgment from being HIGH. The fact that Genotoxicity was
judged HIGH in Working Table 3.14 elevates the epistemic status above LOW, since genotoxic
mechanisms tend to be less context-specific than non-genotoxic mechanisms.
E. Existential Insight: The judgment is MED since the evidence produces in the analysts
moderate confidence that formaldehyde is a carcinogen in Context 12.
F. Column Summary: The judgment is LOW since Direct Empirical and Semi-Empirical
Extrapolation were NO (with HIGH Intellectual Obligation), and the remainder were MED
(with MED or LOW Intellectual Obligation).
G. Overall Summary: The judgment is LOW since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
7.5.3.2.3. Mixer
The judgment in all cells is NO since the same judgment is given in Working Table 3.14
(for the reasons stated there).
208
-------
7.5.3.2.4. Helper
The judgment in all cells is NO since the claim in Working Table 3.14 is LOW, the
necessary extrapolation premises are either LOW or MED, and the finding of genotoxic route of
action in Working Table 3.14 is not judged relevant to this issue.
7.5.3.3. Stages
7.5.3.3.1. Neoplastic Conversion
A. Direct Empirical: The judgment is NO due to differences in level of exposure and
biological context.
B. Semi-Empirical Extrapolation: The judgment is LOW due to LOW or MED epistemic
status of necessary extrapolation premises and differences in context. The fact that a genotoxic
mechanism is proposed in Working Table 3.14 raises the epistemic status above NO since such
mechanisms tend to be less context specific and there is not believed to be a threshold for
biophysical effect or initiating transition.
C. Empirical Correlation: The judgment is MED since the claim was given MED
epistemic status in Working Table 3.14 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.14. While a correlation does exist between
neoplastic conversion in these two contexts, the MED epistemic status of the claim in Working
Table 3.14 keeps this judgment from being HIGH. The fact that Genotoxicity was judged HIGH
in Working Table 3.14 elevates the epistemic status above LOW, since genotoxic mechanisms
tend to be less context-specific than non-genotoxic mechanisms.
D. Theory-Based Inference: The judgment is MED since the claim was given MED
epistemic status in Working Table 3.14 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.14. While the Theory of Carcinogenesis does
209
-------
support the claim that neoplastic conversion may act by common routes across contexts
containing similar DNA structure, the MED epistemic status of the claim in Working Table 3.14
and several of the extrapolation premises keeps this judgment from being HIGH. The fact that
Genotoxicity was judged HIGH in Working Table 3.14 elevates the epistemic status above LOW,
since genotoxic mechanisms tend to be less context-specific than non-genotoxic mechanisms.
E. Existential Insight: The judgment is MED since the evidence produces in the analysts
moderate confidence that formaldehyde is a carcinogen in Context 12.
F. Column Summary: The judgment is MED since Direct Empirical was NO and Semi-
Empirical Extrapolation was LOW (with HIGH Intellectual Obligation), and the remainder were
MED (with MED or LOW Intellectual Obligation). The presence of at least LOW epistemic
status to the Semi-Empirical Extrapolation cell is taken to keep the overall judgment at MED,
although the judgment might also reasonably be given as LOW.
G. Overall Summary: The judgment is MED since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
7.5.3.3.2. Neoplastic Development
The judgment in all cells is NO since the same judgment is given in Working Table 3.14
(for the reasons stated there).
7.5.3.4. Mechanisms
7.5.3.4.1. Genotoxic
A. Direct Empirical: The judgment is NO due to differences in level of exposure and
biological context.
210
-------
B. Semi-Empirical Extrapolation: The judgment is MED due to HIGH epistemic status
of claim in Working Table 3.14 and LOW or MED epistemic status of necessary extrapolation
premises and differences in context. The fact that a genotoxic mechanism is proposed in
Working Table 3.14 raises the epistemic status above LOW since such mechanisms tend to be
less context specific and there is not believed to be a threshold for biophysical effect or initiating
transition. The lack of complete agreement between contexts keeps this judgment from being
HIGH.
C. Empirical Correlation: The judgment is MED since the claim was given HIGH
epistemic status in Working Table 3.14 and the "Overall Summaries'1 of extrapolation premises
were all either LOW or MED in Working Table 4.14. While a correlation does exist between
genotoxic mechanisms in different mammalian contexts, the MED epistemic status of the claim
in Working Table 3.14 keeps this judgment from being HIGH. The fact that Genotoxicity was
judged HIGH in Working Table 3.14 elevates the epistemic status above LOW, since genotoxic
mechanisms tend to be less context-specific than non-genotoxic mechanisms.
D. Theory-Based Inference: The judgment is MED since the claim was given HIGH
epistemic status in Working Table 3.14 and the "Overall Summaries" of extrapolation premises
were all either LOW or MED in Working Table 4.14. While the Theory of Carcinogenesis does
support the claim that genotoxic mechanisms may act by similar routes across mammalian cells,
the MED epistemic status of the claim in Working Table 3.14 and several of the extrapolation
premises keeps this judgment from being HIGH. The fact that Genotoxicity was judged HIGH
in Working Table 3.14 elevates the epistemic status above LOW, since genotoxic mechanisms
tend to be less context-specific than non-genotoxic mechanisms.
E. Existential Insight: The judgment is MED since the evidence produces in the analysts
moderate confidence that formaldehyde is a carcinogen in Context 12.
211
-------
F. Column Summary: The judgment is MED since Direct Empirical was NO but other
cells were assigned MED. The presence of MED epistemic status to the Semi-Empirical
Extrapolation cell is taken to keep the overall judgment at MED.
G. Overall Summary: The judgment is MED since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
7.5.3.4.2. Non-Genotoxic
The judgment in all cells is NO since the same judgment is given in Working Table 3.14
(for the reasons stated there).
8. DESCRIPTION OF WORKING TABLE 6
Working Table 6 summarizes the results of each extrapolation from an "Observational
Context" (Contexts 5, 6, 11, 13 and 14) to the Target Context" (Context 12). The cell entries in
the row termed Intra-Context are identical to those in the "Overall Summary" row of Working
Table 3.12. The cell entries in each Inter-Context row are identical to the those in the "Overall
Summary" row of the version of Working Table 5 associated with that "Observational Context".
For example, the cell entries in the row for "Inter-Context No. 6" in Working Table 6 are
identical to those in the "Overall Summary" cells of Working Table 5.6.
Summary judgments were made "down" a column for each Claim of Carcinogenicity in
Working Table 6, taking into account all cells in that column. This resulted in one "Column
Summary" entry for each Claim of Carcinogenicity. An "Overall Summary" judgment then was
prepared as in earlier Working Tables. The reasons for these various judgments are provided in
the following sections. The assignments of Intellectual Obligation do not apply to this Working
212
-------
WORKING TABLE 6. SUMMARY SUPPORT FOR CLAIMS OF CARCINOGENICITY BY CONTEXT
Claims of Carcinogenicity1
Context Number
Increases
Incidence of
Cancer
Classification(s)
Sta,
ge(s)
Mechanism(s)
Complete
Partial
Mixer
Helper
Neo.
Conv.
Neo.
Devel.
Genotoxic
Non-genotoxic
Intra-Context1
From Context 12
NO
NO
NO
NO
NO
NO
NO
NO
NO
Inter-Context2
From Context 5
LO
NO
NO
NO
NO
ME
LO
ME
NO
Inter-Context2
From Context 6
LO
LO
LO
NO
NO
ME
NO
ME
NO
Inter-Context2
From Context 11
LO
NO
NO
NO
NO
NO
NO
NO
NO
Inter-Context2
From Context 13
LO
NO
NO
NO
NO
NO
NO
NO
NO
Inter-Context2
From Context 14
LO
NO
LO
NO
NO
ME
NO
ME
. NO
Column
Summary
LO
NO
NO
NO
NO
ME
NO
ME
NO
Overall
Summary
LO
NO
NO
NO
NO
ME
NO
ME
NO
'This is an instance of Intra-Context Extrapolation
2This is an instance of Inter-Context Extrapolation from the Observational Context indicated
-------
Table since the judgments of epistemic status are based on reflection upon the different routes
of extrapolation to the "Target Context", rather than on different Relevance Strategies.
8.1. Epistemic Status of the Claims in Working Table 6
Explanation of assignments of epistemic status to cells of Working Table 7. The entries
in cells other than "Column Summary" and "Overall Summary" are taken directly from the
"Overall Summary" cells in Working Table 6 (for Observational Contexts) or Working Table 3
(for the Target Context). The assignments of epistemic status for the "Column Summary" and
"Overall Summary" are explained in the entries below.
8.1.1. Increases Incidence of Cancer
A. Column Summary: The judgment is LOW since the majority of cells (5 of 6) are
LOW and the remainder are NO. Of particular concern is (1) the fact that there is NO intra-
context support for the claim (which removes the possibility of Direct Empirical warrants, taken
in this analysis as having HIGH Intellectual Obligation), (2) the judgment that the intercontext
claims are weakened seriously by the lack of identification of the active metabolite as required
by the extrapolation premises (particularly the Exposure-to-BSDR premise) and (3) the
judgment that intercontext claims also are weakened seriously by the lack of support for the
claim that a threshold for biophysical effects does not exist. If the active metabolite was shown
to be produced, and if it could be established FROM THE DATA BASE EMPLOYED IN
THIS ANALYSIS that neoplastic conversion was not the dominant mechanism of action in mice
and rats, the second 2 concerns above would be removed partially and the Column Summary
might be elevated to MED.
213
-------
B. Overall Summary: The judgment is LOW since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim. It should be noted, however, that the presence of MED epistemic
status to the claims of neoplastic conversion and genotoxic mechanism might reasonably be
taken to elevate this judgment to MED.
8.12. Classifications
8.1.2.1. Complete Carcinogen
A. Column Summary: The judgment is NO since the majority of cells (5 of 6) are NO
and the remainder is LOW. The sole intercontext claim arises from the skin studies in mice (as
described in previous Working Tables), but the relevance of these results to the claim in Context
12 (the Target Context) is weakened by the lack of identification of the active metabolite as
required by the extrapolation premises (particularly the Exposure-to-BSDR premise) and by the
lack of support for the claim that a threshold for biophysical effects does not exist. Given the
presence of the LOW judgment based on the mice data (Context 12), a case could be made that
the Column Summary here should be LOW, but the analysts do not feel that a LOW epistemic
status to a single Inter-Context claim justifies the formation of belief about this claim in the
target context.
B. Overall Summary: The judgment is NO since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
8.1.2.2. Partial Carcinogen
214
-------
A. Column Summary: The judgment is NO since the majority of cells (4 of 6) are NO
and the remainder are LOW. The only support comes from Contexts 6 (mice) and 14 (in vitro
human cell lines), and these judgments were LOW as explained in Working Tables 5.6 and 5.14
for these two contexts. The relevance of these results to the claim in Context 12 (the Target
Context) is weakened by the lack of identification of the active metabolite as required by the
extrapolation premises (particularly the Exposure-to-BSDR premise) and by the lack of support
for the claim that a threshold for biophysical effects does not exist. Given the presence of the
LOW judgment based on the mice data (Context 12) and in vitro data (Context 14), a case could
be made that the Column Summary here should be LOW, but the analysts do not feel that a
LOW epistemic status to a single in vivo Inter-Context claim justifies the formation of belief
about this claim in the target context, given that the background transition rates are very
different in the in vitro context.
B. Overall Summary: The judgment is NO since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
8.1.2.3. Mixer
A. Column Summary: The judgment is NO since all cells are NO.
B. Overall Summary: The judgment is NO since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
8.1.2.4. Helper
A. Column Summary: The judgment is NO since all cells are NO.
215
-------
B. Overall Summary: The judgment is NO since the analysts' strongly foundational
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
8.1.3. Stages
8.1.3.1. Neoplastic Conversion
A. Column Summary: The judgment is MED since half the cells are MED and the
remainder are NO (bearing in mind that NO does not mean evidence AGAINST the claim).
The reasoning is that neoplastic conversion has been demonstrated with MED epistemic status
in both the rat and mouse. The relevance of these results to the target context admittedly is
weakened by the lack of identification of the active metabolite as required by the extrapolation
premises (particularly the Exposure-to-BSDR premise) and by the lack of support for the claim
that a threshold for biophysical effects does not exist. It is assumed here, however, that
neoplastic conversion is not, in general, a transition for which there is a distinct threshold.
Rather, decreases in BSDR are assumed to result in a continuous decrease in the
PROBABILITY of neoplastic conversion. Given the lack of identification of the active
metabolite, concerns remain over the premise concerning Exposure-to-BSDR, but these were
addressed already in the assignment of MED epistemic status to the claim in Working Tables
5.5 and 5.6.
B. Overall Summary: The judgment is NO since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
8.1.3.2. Neoplastic Development
216
-------
A. Column Summary: The judgment is NO since all but one cell is NO, with the
remaining cell being LOW. The relevance of the results in mice to the claim in Context 12 (the
Target Context) is weakened by the lack of identification of the active metabolite as required by
the extrapolation premises (particularly the Exposure-to-BSDR premise) and by the lack of
support for the claim that a threshold for biophysical effects does not exist. Particularly worrying
is the possibility that formaldehyde acts in the mice primarily by promotion, and that promotion
might be a process displaying a threshold (which may not be exceeded in humans). This
concerns was already addressed in producing the judgment for this claim in Working Table 5.6.
Given the presence of the LOW judgment based on the mice data (Context 12), a case could be
made that the Column Summary here should be LOW, but the analysts do not feel that a LOW
epistemic status to a single Inter-Context claim justifies the formation of belief about this claim
in the target context.
B. Overall Summary: The judgment is NO since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
8.1.4. Mechanisms
8.1.4.1. Genotoxic
A. Column Summary: The judgment is MED since half of the cells are MED and the
remainder are NO (bearing in mind that NO does not mean evidence AGAINST the claim).
The reasoning is that genotoxicity has been demonstrated with MED epistemic status in both
the rat and mouse. The relevance of these results to the target context admittedly is weakened
by the lack of identification of the active metabolite as required by the extrapolation premises
(particularly the Exposure-to-BSDR premise) and by the lack of support for the claim that a
217
-------
threshold for biophysical effects does not exist. It is assumed here, however, that genotoxicity is
not, in general, a mechanism for which there is a distinct threshold. Rather, decreases in BSDR
are assumed to result in a continuous decrease in the PROBABILITY of genotoxic effects.
Given the lack of identification of the active metabolite, concerns remain over the premise
concerning Exposure-to-BSDR, but these were addressed already in the assignment of MED
epistemic status to the claim in Working Tables 5.5 and 5.6.
B. Overall Summary: The judgment is MED since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
8.1.4.2. Non-Genotoxic
A. Column Summary: The judgment is NO since all cells are NO.
B. Overall Summary: The judgment is NO since the analysts' strongly foundationalist
epistemological stance suggests that coherence across claims should not affect the epistemic
status of any given claim.
9. GENERAL CONCLUSIONS AND DISCUSSION
The three central criteria for assessing the present methodology were described as utility,
completeness and practicality in Section 1.1. These same criteria are used in the present section
to examine the development of the Working Tables for the instance of formaldehyde exposures.
It is essential to bear in mind that these conclusions are drawn with respect to the particular
subset of available data on formaldehyde used in the earlier sections of this report, rather than
on the full body of data provided in Appendix A.
218
-------
Turning first to completeness, no problems arose in the sense of having to exclude bodies
of data because these found no place in the analysis. Data from the categories Tumor Response,
Biophysical Effect, Pharmacodynamics, Host Factors, Concurrent Environmental Conditions and
Related Substance Assessments each were collected and given a distinct role within the analysis.
Of equal importance is the fact that the methodology explicitly called for each of these
categories of data at some point in the analysis, providing support for the claim that the
methodology aids in both identifying the necessary bodies of data and locating their role within
the analysis. In a search of the larger body of literature in Appendix A, no body of data was
identified that would have been excluded from the analysis by the methodology. This conclusion
applied to data on (a) breathing frequencies, (b) tidal volumes, (c) formaldehyde partitioning
coefficients between air and tissue, (d) nasal histology, (e) background incidence of cancer, (f)
bioactivation mechanisms, (g) clearance rates for formaldehyde and active metabolites, (h)
effects of concurrent exposures, (i) apparent thresholds in promotional effects, etc. The ability to
incorporate such data was confirmed in the present study because such considerations have
played a central role in many of the debates concerning regulation of formaldehyde (although
the considerations have been more important in the estimation of risk rather than in hazard
identification).
In considering utility, the methodology also functioned as intended. A large body of data
was abstracted into the various Working Tables, allowing the analysts to determine the point in
lines of reasoning at which each body of data was needed to support a particular inference.
While the process of reasoning could have been conducted in a purely textual manner, relying
on descriptions rather than Working Tables, the presence of the Working Tables organized the
data into clearly defined tasks of reasoning. This ensured both that particular claims of
219
-------
carcinogenicity were related directly to the data on which they were based, and that all data
related to a particular claim were brought to bear on that claim.
The authors found the methodology useful in clarifying the reasoning behind particular
judgments made in the cells of the Working Tables. A characteristic of the secondary (policy-
related) literature on formaldehyde provided in Appendix A is a reliance on textual descriptions
for presenting evidence and lines of reason. Readers unfamiliar with the topic of formaldehyde
carcinogenicity, and of carcinogenic hazard identification in general, are not led to examine the
underlying and often unstated assumptions by which data are asserted as relevant to the tasks of
inference. The present methodology avoided that problem by showing explicitly the "background
premises" that must be supported if a given body of data was to made relevant and employed in
lines of reasoning. This aided the analysts in clarifying their own reasoning on formaldehyde
carcinogenicity, explicating both the strengths and weaknesses of each inference. These strengths
and weaknesses were separated clearly into issues of the foundational quality of the data
(Completeness, Utility, Strength of Effect, Exposure-Specific Effect); the relevance of the data
(Direct Empirical, Semi-Empirical Extrapolation, Empirical Correlation, Theory-Based
Inference, Existential Insight); the background premises needed to employ those data in a given
relevance strategy; and the relative merits of different lines of reasoning (examined in the cells
for "Column Summary" and "Overall Summary" for Working Tables 3, 5 and 6).
As to practicality, this judgment may be made from both an "internal" and "external"
perspective. An internal perspective examines practicality by considering only the efforts
required for the present analysis. The project took place over a period of approximately 5
months, encompassing the tasks of collecting primary data through producing the final report.
Much of this time was spent refining the methodology, with approximately 3 months being
devoted to the actual analysis of data and lines of reasoning. Once familiar with the
220
-------
methodology, the analysts were able to incorporate new data rapidly into the analysis. While the
present analysis used only a subset of the available data, therefore, it is evident that use of the
full body of data in Appendix A would not have lengthened the analysis beyond a time frame
that is reasonable for regulatory decisions.
From an external perspective, by which we mean a comparison between the present
methodology and previously existing analyses, the length of time and amount of resources
necessary are comparable. The present methodology required an initial period of approximately
one month to become familiar with the manner in which data were employed in the Working
Tables and in which inferences were extrapolated between Working Tables (where necessary).
Once that training period passed, however, the analysis proceeded more smoothly than would
have been the case without the framework of the methodology. This training period would not
require repetition for different analyses, since the methodology is the same for all instances of
hazard identification.
The most useful aspect of the methodology was the degree to which it clarified the
position of the analyst with respect to the various Claims of Carcinogenicity. Complex judgments
were brought out in sequence, allowing one judgment to be formed before its use in larger lines
of reasoning. For example, the judgment of the quality of each body of epidemiological data was
addressed first and shown to result from separate considerations of Completeness, Utility,
Strength of Effect and Exposure-Specific Response. This judgment then was combined with a
meta-analysis of the coherence across the two epidemiological studies to produce a judgment of
the claim that cancer incidence was increased in occupational populations in general. This
judgment then was combined with a judgment of the applicable category of evidence (Relevance
Strategy) in which the data should be placed (Direct Empirical), of the quality of the
background premises required to utilize this relevance strategy, and of the Intellectual
221
-------
Obligation associated with that relevance strategy (HIGH). This resulted in the judgment that
formaldehyde increases the incidence of cancer in human populations exposed at above 2 ppm,
at least when this one relevance strategy is considered. This judgment then was combined with
similar judgments from the other four relevance strategies to yield a "Column Summary"
judgment of the claim that formaldehyde increases the incidence of cancer in human populations
exposed at concentrations above 2 ppm. This judgment then was combined with the necessary
"Extrapolation Premises" (each of which had been previously judged for foundational quality of
available data; applicable Relevance Strategy,; Intellectual Obligation; and coherence across
Relevance Strategies) to yield an Inter-Context judgment of the claim that formaldehyde
increases the incidence of cancer in the target context. Finally, the judgments for each route of
reasoning to the target context were examined for coherence and a composite judgment made of
the claim that formaldehyde increases the incidence of cancer in the target context.
Without the methodology, and its ability to "deconstruct" the multiple lines of reasoning,
the analysts would have been faced with the task of making composite judgments in potentially
large "steps". In other words, the composite judgments (such as in Working Table 6) would not
have been constructed from earlier judgments, complicating the defense of any such judgments.
By breaking all judgments into their components, the analysts were better able to ensure that
separate considerations (of foundational quality, Intellectual Obligation, etc) were both
examined in detail and weighted properly into the analysis. This opened the process of reasoning
to greater scrutiny and re-analysis, placing each judgment into the framework of judgments from
which the claims in Working Table 6 were constructed. The result was an increased ability of the
analysts to understand the composite judgments of epistemic status in Working Table 6; to
identify the key areas in which those composite judgments were dependent upon judgments in
earlier Working Tables; to identify the reasoning leading to these earlier judgments; to identify
222
-------
the strengths and weaknesses of these earlier judgments; and to identify the kinds of
information that would have weakened or strengthened all judgments. One goal of the
methodology was to prepare analysts for regulatory debates, helping the analyst (1) to clarify his
or her own position by understanding the basis for that position, (2) to understand the manner
in which that position might reasonably be challenged from other positions, (3) to understand
the degree to which that position rests on consideration of data, on consideration of theories,
and on consideration of cognitive values needed to assign Intellectual Obligation, and (4) to
understand where new bodies of data potentially introduced into a regulatory debate would
affect all judgments. It is the conclusion of the present study that the methodology developed
here aided in reaching all of these goals.
9.1. Alterations in the Methodology
While all of the goals discussed above were met satisfactorily, the example of
formaldehyde brought out several issues regarding the structure and use of the framework of
Working Tables. Some of these were discussed in previous sections of the report as they arose
in specific Working Tables. All such considerations are described here.
9.1.1. Numbering of Working Tables
In the original report, there were 7 Working Tables. Working Table 4 contained data
necessary for the various extrapolation premises. These extrapolation premises utilized data
from the data categories Biophysical Effects, Pharmacodynamics, Host Factors, and
Environmental Conditions (analogous to Concurrent Environmental Conditions). It was
determined during this project that each of these bodies of data could also be placed into
Working Table 2 since the appropriate data categories existed in that Working Table. As a
223
-------
result, Working Table 4 from the earlier report was removed and all Working Tables
renumbered. The new numbering system for the remaining 6 Working Tables, in relation to the
older numbering system, is described in Section 1.4.
9.1.2. Removal of Variability Premise
In the original report, the Working Table for Inter-Context extrapolation premises (now
Working Table 4) contained a premise concerning intra- and intersubject variability. It was
determined during this project that this premise always referred to variability of some quantity
contained within the other 4 premises. This extrapolation premise was, therefore, removed from
the Working Table. The analysts now determines the effect of variability on each of the
remaining 4 premises separately.
9.13. Intellectual Obligation in Working Table 4
In the original report, the Working Table for Inter-Context extrapolation premises did
not contain a column for assignments of Intellectual Obligation to the different Relevance
Strategies (as is present in Working Tables 3 and 5). This was an oversight, since the assignment
of Intellectual Obligation is necessary whenever a Relevance Strategy is used. This has been
corrected by adding this column to Working Table 4.
9.1.4. Columns in Working Table 2
In the original report, Working Table 2 contained a column labeled "Causality" and a
column labeled "Organism-Specific Effect". It was determined during this project that these
columns were not required. The judgment of "Causality" was intended originally to reflect
224
-------
whether the data had been collected under conditions in which a causal link could be established
firmly. This judgment may also be considered part of the "Utility" of a study, and the so the
judgment of causality was removed as a separate column and placed into the "Utility" column.
It also was determined that "Organism-Specific Effect" could be placed under "Exposure-
Specific Effect" without loss of information. The same kinds of data were found to be present in
both columns, the only difference arising from the fact that exposure was zero in the former and
non-zero in the latter. The "Organism-Specific Effect" column was removed from Working Table
2 to simplify the analysis.
9.1.5. Consideration of Burden of Proof
In the original report, the cells of the Working Tables containing the Claims of
Carcinogenicity and the Extrapolation Premises were divided by a dashed line. The judgments
above that line were intended to refer to a judgment based on data specific to the substance of
interest (here, formaldehyde) and the judgments below that line were to be based on more
generic data on a species (termed "floater data" in that report). It was determined during this
project that this division of judgments should be altered. First, the use of "floater data" now is
explicitly addressed in Working Table 2 and factored into all judgments in conjunction with data
on formaldehyde exposure. Second, it was realized during the study that the regulatory debate
described in the secondary literature (see Appendix A) at times revolved around judgments that
a particular effect (such as cancer or genotoxicity) was established with insufficient strength, and
at other times around judgments that the effect was shown NOT to occur. This required
separate judgments that formaldehyde DOES induce the effect and that formaldehyde DOES
NOT induce the effect. It is important to note here that a judgment of NO in a particular cell of
a Working Table means only that the evidence does not provide SUPPORT for a claim. It does
225
-------
not also mean that the evidence is AGAINST the claim. To deal with this issue, the dashed line
has been re-interpreted to divide judgments FOR a given claim (above the dashed line) and
AGAINST the same claim (below the dashed line).
Each of these judgments may play a role in the final analysis. In the example developed
in this report, only the judgments FOR a given claim were examined. A judgment of NO in a
cell then meant that the claim was not supported by the available data; it did not mean that the
claim was falsified by the data. For example, consider Working Table 3.6 (for the context of
exposure of mice). The judgment of LOW for the claim of a Partial Carcinogen was based
primarily on the initiation assay, which was by application of formaldehyde in solution to the
skin of the mice. This study suggested that formaldehyde induced primarily initiation, but it was
not structured to test for promotional activity. The judgment for Neoplastic Development was
NO since a promotional assay was not performed and all judgments were only for the claim that
the effect DOES occur. IF a promotional assay had been done and the result was negative, the
judgment in this cell would still have been NO, indicating no effect on the judgment in the cell
for the claim of a Partial Carcinogen. IF a promotional assay had been done and the result was
positive, this would have strengthened the claim of a Complete Carcinogen and weakened the
claim of a Partial Carcinogen. And IF a promotional assay had been done and the effect was
negative, a judgment would have been made that formaldehyde is NOT a promotional agent in
this context (and entered into the bottom half of the cells for the claim of Neoplastic
Development). This would have lowered the claim that formaldehyde is a Partial Carcinogen
and raised the claim that formaldehyde is a Complete Carcinogen.
9.1.6. Cancer and Cellular Transformation
226
-------
As originally designed, Working Table 3 was intended for use in justifying the claim that
a substance increased the incidence of cancer. In completing these Working Tables for
formaldehyde, the investigators encountered the problem that two of the contexts (13 and 14)
were in vitro exposures of cell lines, a context within which cancer has no meaning. The closest
approximation to cancer in vitro is cellular transformation. In response, the column headings in
Working Tables 3.13 and 3.14 were altered to represent cellular transformation. As a result, the
column headed "Increases Incidence of Cancer" was changed to "Increases Incidence of
Transformation".
92. Remaining Difficulties
The current project also identified several areas in which analysts might be expected to
encounter difficulties. While these difficulties also arise in all attempts to judge the quality of
inference (and are not, therefore, unique to the present methodology), they are of sufficient
interest to warrant discussion here.
92.1. Theory and Correlational Judgments
The present analysis was intended only to examine the use of data on formaldehyde
within the framework of Working Tables. At several points in the analysis (particularly in
Working Tables 3 and 4), use was made of explicit theories of carcinogenesis and
transformation, as well as empirical correlations between measured biophysical effects (such as
transformation) and the Claims of Carcinogenicity.
These theories and correlations were discussed very briefly in the text with minimal
justification. It was not possible in this study to provide a detailed analysis of the evidential
support for theories and for measures of correlation. Such an analysis would require another
227
-------
series of studies focusing entirely on these theories and correlations. Clearly, this additional
analysis would require the development of Working Tables to assess the evidential support
available. Since such an analysis is common to all substances being considered for hazard
identification, rather than to a particular substance, it is not included in the current version of
the methodology. It is recommended that the USEPA consider developing additional Working
Tables for the justification of theories and correlations, and that these be used in conjunction
with expert panels to determine a common set of judgments of epistemic status to be applied
uniformly throughout the field of carcinogen hazard identification.
922. Separability of Intellectual Obligation and Epistemic Status
As discussed in Section 1.4, an issue arose as to the degree to which Intellectual
Obligation for the different Relevance Strategies could be separated from the judgments of
epistemic status required to complete the cells for Claims of Carcinogenicity. In the current
example developed here, this assignment of Intellectual Obligation was an abstracted assessment,
meaning that the analysts have made these assignments without considering the characteristics
of actual theories, correlations, etc, needed in filling out other cells of the Working Tables. For
example, it was assumed that the Intellectual Obligation for Theory-Based Inference is
MEDIUM for all theories, all contexts, all claims and all bodies of evidence used in those
theories. Consideration of the evidence for a SPECIFIC theory was assumed to arise only when
the separate CELLS of the Working Tables were completed.
A counter argument might be that such abstracted assignments of Intellectual Obligation
are not meaningful. It might be argued that one cannot assign a general measure of Intellectual
Obligation to Theory-Based Inference as a class, but rather that this assignment MUST consider
the support for the particular theory being used in a specific inference. For example, the
228
-------
Intellectual Obligation towards the theory relating increased incorporation of tritiated thymidine
and Neoplastic Development (invoked in Working Table 3.5) might be considered higher than
that for the theory relating allergic response to carcinogenicity (invoked in Working Table 3.12).
In this case, the analyst might choose not to attempt to provide measures of Intellectual
Obligation in the various Working Tables, but would reflect the assignment of Intellectual
Obligation for that PARTICULAR theory directly in the assignment of epistemic status in a cell
of the Working Table. It may then be the case that a Theory-Based Inference warrant based on
some theories (of a particular stage of carcinogenesis, say) is better established and so carries a
higher Intellectual Obligation than do Direct Empirical observations of phenomena pertaining to
it (for example, where the histopathologic form of pre-neoplastic lesions is not clear). The point
here is simply that the judgment of Intellectual Obligation may not be possible in the abstracted
sense, and may be inseparable from the assignment of epistemic status in the other cells of the
Working Table.
923. Pharmacodynamics Data Other than on Formaldehyde
One problem that arose in the re-structured Working Tables is that there was a large
mass of data on the general anatomical, physiological and histological properties of the different
organisms constituting the various contexts. These data did not necessarily appear in any
particular paper on formaldehyde, since those papers typically were concerned only with effects
produced through the formaldehyde exposure. The extrapolation premises appearing in Working
Table 4, however, require information on the relative values of such factors as breathing
frequency, tidal volume, retention half-times, etc, for both the Observational and Target
Contexts.
229
-------
The present analysts found themselves continuously having to move outside of the
literature explicitly on formaldehyde to obtain the more generic Pharmacodynamics and Host
Factor data. This problem was dealt with in the present study by using summary judgments of
the effects of any generic Pharmacodynamics and Host Factor differences between species, with
these judgments being taken directly from the literature in Appendix A (and resulting in the
judgments of epistemic status in Working Table 4 for each context). These summary judgments
were not given further examination in this study, since such judgments of generic factors were
not the focus of this study. As in the case of the discussion in Section 9.2.2., it is recommended
that the USEPA codify these generic judgments for the contexts typically encountered in hazard
identification, so they may be examined more completely and applied consistently across
instances of hazard identification. Otherwise, each hazard identification would be delayed
through lengthy preparation of these generic judgments.
92.4. Methodological Rigor
As described in the original report, two of the criteria for judging the rationality of an
analysis are Logical Soundness and Methodological Rigor. Taken together, these mean the
degree to which a judgment may be deduced from smaller components through a clearly
articulated procedure capable of being scrutinized by external reviewers. A goal of the
framework of analysis examined here was to avoid undue methodological constraints on
judgments that contain a significant component of subjective judgment in the form of cognitive
values, requiring only that judgments be made after full reflection on the various analytic
components going into those judgments.
Still, it might be argued that more methodological rigor is to be desired. It is not
possible, for example, for an outside examiner to fully understand the judgment of LOW for the
230
-------
claim of "Increases Incidence of Cancer" in the "Column Summary" cell of Working Table 3.6,
since that judgment involved a subjective response to the five separate components of
Intellectual Obligation and of Relevance Strategy in the cells above the "Column Summary". A
similar problem arises in explaining the judgment of MED for the claim of "Genotoxicity" as a
"Column Summary" in Working Table 6. For an analyst valuing Logical Soundness and
Methodological Rigor, the lack of an algorithm for combining five judgments of Intellectual
Obligation and five judgments of the epistemic status associated with specific Relevance
Strategies may be worrying.
There is no direct way to address this concern. It is a natural consequence of recognizing
that judgments of epistemic status involve both methodological considerations (summarized in
Working Table 2 for each context) and issues of cognitive values. The present analysis DID
employ constraints on the generation of composite judgments as detailed in Section 1.4. Even
within these constraints, however, there will be latitude in assigning final judgments. If anything,
a merit of the current framework of analysis is that it allows the analyst to understand this
latitude, and to see how other analysts might reach different conclusions even within the
constraints imposed in Section 1.4.
9.2.5. Specification of Contexts
A judgment must be made as to when a given set of studies were conducted under
conditions sufficiently similar to constitute the same context. For Context 12, this issue did not
arise since there is a single study. In Contexts 13 and 14, again only a single study for each
context was included in the set of 14 references examined. Still, it might prove reasonable to
assume that Context 13 and Context 14 should be combined into a single context focusing on
mammalian cell lines, since human cells satisfy this definition. It was decided here to keep these
231
-------
two contexts separate because (1) it is possible that a substance might produce transformation in
human cells but not cell lines derived from other species (or the inverse) due to bioactivation
properties present only in human cells (or absent only in those cells) and (2) the added
complexity introduced by keeping the cell lines in separate contexts did not appear to be
significant given the small number of studies examined for this example. The reader should
note that with the cells lines kept in separate contexts, it is possible to draw separate inferences
about the ability of formaldehyde to induce transformation in "human cells" and in "other
mammalian cells" (as is done in the remainder of this report). These inferences then are
extrapolated separately to the target context using "extrapolation premises" (defined in the
earlier report and in later sections of this report) unique to each of the two cases of
extrapolation. If the contexts were combined, all intra-context inferences would refer only to the
ability of formaldehyde to induce transformation in the broader class of "all mammalian cell
lines", followed by extrapolation of this broader claim to the target context. This might result in
a loss of information on differences between human and non-human cell lines at some point
during the subsequent analysis and extrapolation to the target context. For example, there might
exist different empirical correlations between in vitro transformation and in vivo carcinogenicity
for the different cell lines, a difference that might be lost if the intra-context claim for
transformation did not distinguish between the various mammalian cell lines.
A related (but reverse) example of the issue of defining contexts may be found in the
case of the data on mice (Context 6). Reference #14 in Table 1 contains data on exposure of
mice to airborne formaldehyde, while Reference #4 contains data on initiation assays from
exposure of mouse skin to formaldehyde in solution. The analysts chose in this case to place
both studies in the same context since it was judged that the ability of formaldehyde TO
PRODUCE INITIATION did not depend upon the route of exposure, the physical form, or the
232
-------
tissue being exposed (skin versus nasal passages). The important defining feature of the data
was that the tissues exposed were part of the larger category of "in vivo mouse tissue". Other
analysts might argue (quite reasonably) that the mechanisms of action of formaldehyde are so
different in skin and nasal tissue as to require two separate contexts, necessitating repetition of
all Working Tables described later in this report for an additional context. It still might be
possible to deal with the concerns over context by placing both studies into the single context
used here (Context 6) and raising issues of the background assumptions needed to infer
initiation action in mouse nasal passages from the initiating action of formaldehyde in mouse
skin. In any event, it should be clear from this brief discussion that specification of contexts
remains an issue in which expert judgment is both required and a source of potential differences
between analysts.
93. Incorporation of a Recent Study
In the course of performing this evaluation of a framework for evidential epistemic
analysis, the present analysts have been made aware of newly finished work on formaldehyde
unavailable at the time the initial Contexts and studies (see Table 1) were set. This newer work
may be pertinent to the conclusions reached here insofar as it concerns specific results about a
particularly important Context, rat exposure by inhalation (Context 5). Since the purpose of the
present evaluation is to test the epistemic framework, though, and is not to establish conclusions
about the data going into that framework, the present analysts decided against incorporating
data released during the course of the project. Continual redefinition of the project's Contexts
and of the data sets within them would suit neither the purpose of this evaluation of the
framework, nor the ultimate purpose for which it is intended in environmental health decisions.
233
-------
Still, a key test of the methodology is it's ability to incorporate new data as these become
available during an analysis (and, perhaps, in response to that analysis). Without reconstructing
an entirely new set of the various Working Tables, it can still be useful to examine the claims of
carcinogenicity and judgments of epistemic status made previously for one particular exposure
context and in the extrapolation from that Context to the Target Context to see whether and/or
how they might be changed by the addition of new evidence. To demonstrate whether the
determinations made here might be altered by new data, the present analysts have incorporated
data reported in T.M. Monticello, et al., "Sustained Cell Proliferation: A Critical Factor in
Formaldehyde Nasal Cancer," (obtained as a preprint from one of the co-authors, Dr. James
Swenberg at UNC, and cited by him as a of particular interest with respect to hazard
identification for formaldehyde).
9.3.1. The Study Conditions and Claims
Of the several studies newly finished which might have been chosen for examination, the
Monticello work was preferred since it fits neatly into an established Context, Rat Exposure by
Inhalation, and also contains a claim about extrapolation of the reported results to the Target
Context, Human Exposure by Inhalation. Moreover, the study's design was familiar to the
present analysts, thereby eliminating any problems with interpreting outcomes from a study
design dramatically unlike the ones previously considered.
The chief claims made in this new study, taken to be supported by data reported as
evidence for evaluation in the present framework, are these:
A. "Our results [of concentration-dependent increases in nasal cell proliferation that
parallel exposure concentration-tumor response curves for the same strain of rats at the
same exposure conditions] support the proposal that sustained increases in cell
proliferation are crucial for the induction of nasal carcinomas by formaldehyde."
234
-------
B. "The lack of increased cell proliferation over basal control levels in the lower
concentration groups (0.7, 2, or 6 ppm formaldehyde) suggests that the increased risk for
malignant transformation observed at 10 and 15 ppm may not exist for the lower
exposure concentrations studied."
C. "Results from the present study imply that exposure of humans to levels of
formaldehyde that do not increase cell proliferation would pose a much smaller risk of
cancer than that currently estimated by linear extrapolation from the rodent bioassay
data."
932. Utility and Strength of Effect Judgments
The present analysts judged the Utility of the Monticello study to be HIGH since it
included and adhered to strong materials and methods descriptions, and since these were in fact
replicated from those of a study that was already included in the present analysis (Swenberg, et
al. 1980). There were no questions about exposure conditions, animal husbandry, DNA labeling,
histoautoradiography, cell scoring, control animals, or statistical measures of results that might
lower this utility score.
The Strength of Effect was also judged here to be HIGH since the study reports
statistically significant results for the correlation of tumor incidence and cell proliferation with
exposure, each being significant at the 95% confidence level. The correlation between exposure
and cellular proliferation was nonlinear, as had been the correlation between exposure and nasal
tumor response reported in 1980 and 1983.
933. Relevance Strategies Employed
The study authors report evidence of increased cell proliferation by association with the
uptake of radio-labeled thymidine delivered to the animals during formaldehyde exposure; thus,
as before when this technique was used [see the description under Context 5 above], this cannot
count as Direct Empirical evidence for increased proliferation, but only as Empirical Correlation
235
-------
supplemented by Theory-Based Inference. The theory necessary to support this correlation is the
one relating thymidine uptake uniquely with proliferation, a theory understood above to have
been established with only MED epistemic status.
In addition, the study authors must employ a refinement of the standard Theory of
Carcinogenesis used throughout the present analysis to account for their claim that nasal tumor
response is a product of a combination of high formaldehyde concentration effects to increase
cell proliferation together with the (merely stated) weak mutagenic character of formaldehyde
exposure. This assumption of theory forms the warrant for the authors' claim that their
"findings indicate that formaldehyde-induced preneoplastic nasal lesions exhibit disturbed cell
maturation and/or an alteration in growth control, consistent with the hypothesis that
preneoplastic lesions represent clonal expansion of transformed cells." The present analysts judge
this warrant and evidence as established with MED epistemic status.
In making their last claim, for the extrapolation from results reported for rat exposure to
the target context of human exposure, the study authors provide no warrant other than
Existential Insight. But they do rely here on a variant of Semi-Empirical Extrapolation to suggest
the possibility of a threshold exposure for nasal cancer induction, since the lowest exposures in
the study were not seen to induce increased cell proliferation. The present analysts judge this
claim to have MED epistemic status based on the weak support offered as warrant.
93.4. Possible Changes in Earlier Working Table Scores
Had these data from the Monticello study been included in the direct evaluation of
evidence for claims of carcinogenicity in Context 5, the most important effect they might have
had would be to alter slightly the present analysts' conclusions expressed in Working Table 6.5
about data for "Exposure to BSDR" and "BSDR to Effect". The study suggests that either
236
-------
"Exposure to BSDR" or "BSDR to Effect" may be characterized by a threshold exposure, at least
for the promotional transition. This would tend to decrease the epistemic status of the judgment
that an increase in nasal tumors noted at high exposures in rats (Context 5, with exposure above
2 ppm) could be taken as Inter-Context support for the claim of carcinogenicity in the Target
Context (humans at exposures of less than 2 ppm). A related impact would be to strengthen the
Inter-Context support for the claim that formaldehyde is NOT a carcinogen at low levels of
exposure. This impact is not considered further here since the previous versions of the Working
Tables examined only claims that a given effect takes place (the "upper half of the cells in the
Working Table), not the inverse or negation of such claims (the "lower half' of the cells).
It might also have been possible that the data from the Monticello results would have
strengthened the present analysts' determinations about the "Neoplastic Development Stage of
Carcinogenesis" in Working Table 3.5. This might have come about since the Utility and
Strength of Effect reported for the new data both are HIGH. However, without any Direct
Empirical evidence the Column Summary and Overall Summary - both MED -- in the previous
version of the Working Table are unlikely to have changed due to slight up-grades in the
judgments of epistemic status for the Theory-Based Inference and Semi-Empirical Extrapolation
relevance strategy cells.
The present analysts are less-sure than the authors of this study that their results are
sufficient for down-grading the epistemic status of a claim for extrapolating from Context 5 to
Context 12, for the reason that no Direct Empirical evidence has been offered in the
observational context, and that the theory necessary for extrapolation is supported with only
MED epistemic status. Thus the scores in Working Table 6 would remain relatively unchanged
despite the new evidence. The "Column Summary" judgment of "Neoplastic Conversion" in
Working Table 6 might be downgraded slightly since (1) the original assignment included a
237
-------
judgment of LOW for Inter-Context extrapolations from Context 5 and (2) the new data seem to
suggest that this latter judgment might be downgraded to NO. The "Column Summary" judgment
of "Increases Incidence of Cancer" in Working Table 6 might also be downgraded slightly for the
same reasons.
9.4. Other Conclusions About Formaldehyde Carcinogenicity
9.4.1. World Health Organization
The World Health Organization International Programme on Chemical Safety issued two
documents on formaldehyde in 1989: (1) WHO Environmental Health Criteria 89:
Formaldehyde (its "Criteria Document"); and (2) WHO Health and Safety Guide 57 (its
recommendations for action on human exposures to formaldehyde). The WHO conclusions
about formaldehyde carcinogenicity were based on data drawn from non-human animal bioassay
studies and from human epidemiology, just as the ones in the present analysis have been.
However, the WHO criteria document contains significantly more data on non-cancer, non-
mortality endpoints than does the present report, and these data appear to have had strong
influence on the WHO analysts' final conclusions about formaldehyde. Additionally, the WHO
analysts have also considered the coherence of data from studies looking at routes of exposure
other than inhalation, while the present report deals only with inhalation exposure.
More specifically on the issue of formaldehyde's carcinogenicity, the WHO analysts
assembled data in categories and contexts not significantly different from the Data Categories
used in the framework of the present analysis, and listed results from much of the same data
discovered in the literature search done for the present report. The findings by the WHO
analysts for the categories/contexts and the data compiled under them were:
9.4.1.1. Effects on Experimental Animals
238
-------
A. At 120 mg/m3 both mice and rats show morbidity (ulceration or necrosis of the nasal
turbinate mucosa, pulmonary congestion, and bacterial septicemia owing to a damaged
respiratory mucosa) and mortality;
B. At 1.2 mg/m3 both mice and rats show morbidity (increased eye irritation, increased
airway resistance, decreased respiratory rate);
C. Rabbits and Guinea pigs show acute toxicity from dermal and oral administration;
mice and rats by oral, inhalation, intravenous injection, and subcutaneous exposures;
D. At repeated exposures of 7-25 mg/m3, rats show histological changes in nasal
epithelium, cell degradation, inflammation, necrosis, and increased cell proliferation;
E. At concentrations below 1 mg/m3 no cell damage or hyperplasia is seen in rats;
F. Dose-related lesions (dysplasia, squamous metaplasia) observed with repeated
exposures of rats at 2.4, 6.7, or 17.2 mg/m3 in respiratory and olfactory epithelia;
G. Clinical abnormalities (dyspnoea, listlessness, hunched posture) seen after inhalation
during repeated exposures of rats to 2.4, 4.8, 12, 24, or 48 mg/m3
H. Oral administration in drinking water at 100-150 mg/kg body weight over 13 weeks to
rats and dogs results in slightly depressed growth rates, but no effects on the stomach;
I. DNA-protein cross-links formed in vitro and in vivo in rat nasal mucosa at in vivo
exposure concentration equivalents of as little as 1.1 mg/m3.
9.4.1.2. Mutagenicity Effects
A. Formaldehyde has been shown to be mutagenic in many in vitro test systems (E. coli,
S. typhimurium, nematodes, Drosophila, and mammalian lymphoma and lymphoblasts) using
DNA repair, chromatid exchange, and chromosome aberration measures at varying levels of
exposure;
239
-------
B. Some in vitro cell transformation data are species specific;
C. Local DNA adducts are observed in rats without simultaneous systemic genetic
effects;
D. Human lymphocyte cultures show no chromatid exchange below a formaldehyde
concentration of 5 ml culture medium;
E. There is no definite evidence that formaldehyde induces mutations in vivo.
9.4.1.3. Carcinogenic Effects in Mice and Rats
A. Nasal squamous cell carcinomas and severe tissue damage seen in rats exposed at
17.2 mg/m3;
B. Low, statistically non-significant incidence of tumors seen at exposures of 6.7 mg/m3;
C. No tumors are found at other sites;
D. Long-term oral administration of formaldehyde (0.02-5%) in drinking water found to
induce papillomas in rat forestomach;
E. Exposure by inhalation in a population of rats with previous nasal mucosa damage
resulted in a significant number of nasal squamous cell carcinomas after 28 months at 12 mg/m3;
F. Dermal tests for initiation/promotion (with PAHs) in mice were either negative or
inconclusive;
G. No observation of co-carcinogenic effects with inhalation exposure to formaldehyde
and to a proven initiator (coal-tar aerosol).
9.4.1.4. Mechanisms of Carcinogenesis
A. Formaldehyde reacts readily with many cellular nucleophiles, forming variously-stable
adducts;
240
-------
B. DNA adducts are produced in reaction with formaldehyde in vitro.
9.4.1.5. Specific Effects in Humans
A. Formaldehyde is a known skin and eye irritant at concentrations between 5% and
25%, and is corrosive at concentrations >25%;
B. Various and highly variable sensory reactions are reported from exposures at
concentrations > 1 mg/m3;
C. Skin sensitization following contact dermatitis or contact eczema is induced only by
direct skin contact with formaldehyde solutions >2%, and not by formaldehyde gaseous
exposure;
D. Death by ingestion, owing to pulmonary edema, respiratory failure, or circulatory
collapse, has been reported from exposures ranging from "a few drops" up to 89 ml; recovery has
been made, though, after ingestion of as much as 120 ml;
E. No acute deaths have been reported from formaldehyde inhalation;
F. Formaldehyde appears not to be teratogenic;
G. Non-asthma, non-Type I and non-Type IV reactions may be induced by exposure;
H. Formaldehyde can induce reversible obstruction of airways at irritant exposure
concentrations;
I. There is perhaps some increase in airway resistance from long-term exposure to
concentrations as low as 0.5 mg/m3;
J. There has been no increase in the incidence of chromosomal aberrations or in the
frequency of chromatid exchange demonstrated for occupationally exposed subgroups
(formaldehyde manufacturers, pathology staff, autopsy workers), but some increase for other
subgroups (anatomy students) who are additionally exposed to phenol and other agents;
241
-------
K. Some increase in chromosomal aberrations has been reported for hemodialysis
patients exposed to as much as 125 mg +/- 50 mg formaldehyde;
L. Formaldehyde interferes with DNA repair in vitro, but there are no data relating this
to mutagenic outcomes;
M. Increased incidence of dysmenorrhoea and problem pregnancies (anemia, toxemia,
and low birth-weight babies) has been reported in workers exposed by inhalation to
concentrations ranging from 0.005-4.3 mg/m3, but no difference in fertility or spontaneous
abortions has been demonstrated;
N. No adverse effects demonstrated in sperm counts from occupational inhalation
exposures ranging from 0.7-1.4 mg/m3 over 11 months;
O. There are some self-reported and uncontrolled data reporting adverse psychological
effects (depression, irritability, memory loss, decreased attention span) from inhalation
exposures at indoor residential concentrations of 0.1-4.4 mg/m3.
9.4.1.6. General Human Mortality in Epidemiology
Upper respiratory tract cancers are thought to be predominant, owing to the rapid
metabolism of formaldehyde, although epidemiology of several occupational subgroups suggests
that other increases in cancer incidence (Hodgkin's disease, leukemia, buccal cavity cancers,
cancers of the pharynx, lung, nose, prostate, bladder, brain, colon, skin, and kidney) may be
associated with formaldehyde exposure in the occupational matrix; however, the odds ratios for
these studies vary greatly, and the degree of statistical control and power are variable as well.
9.4.1.7. WHO Conclusions About Formaldehyde Carcinogenicity
242
-------
Although several of the data categories used by WHO were identical with those used in
the present report, it can be seen from the form of the summary of WHO data given here that
the WHO analysts did not explicitly evaluate each evidence claim that underlay the data drawn
from the studies they considered. Moreover, there seems not to have been any determination of
the Completeness or Utility of the WHO studies, and only informal descriptions of the Strength
of Effect were made. Without making these determinations explicitly for each result drawn from
each study (as was done in the present report) it is not possible to understand exactly how the
WHO decision on formaldehyde was reached, or to establish an epistemic status for its ultimate
claim about formaldehyde carcinogenicity.
In addition to consideration of the summary data repeated just above, WHO's final
conclusion about formaldehyde carcinogenicity in humans may also have been partly determined
by its analysts' acceptance of IARC conclusions about formaldehyde. The WHO criteria
document repeats (without further comment) the two IARC conclusions that (1) sufficient
evidence exists for a determination of formaldehyde's carcinogenicity in animals, and (2) there is
limited evidence for the carcinogenicity of formaldehyde in humans. Without a fuller accounting
of how individual elements such as these statements and other data results were weighted in
their decision, it is not possible to reconstruct how the WHO analysts ultimately concluded that
"there is experimental evidence [within the general epidemiological context] providing a
relatively clear suggestion of a possible cancer risk for human beings from exposure to
formaldehyde."
9.42. U.S. Occupational Safety and Health Administration
In May 1992 the U.S. Occupational Safety And Health Administration, U.S. Department
of Labor published its Rule on Formaldehyde exposure: "Occupational Exposure to
243
-------
Formaldehyde" Federal Register 57,102 at 22289-22328. That Rule followed on OSHA's
determination in December 1987 of a time-weighted exposure limit for formaldehyde and
included a summary of the evidence used by OSHA analysts in making their determination on
formaldehyde carcinogenicity. The evidence considered by OSHA analysts was largely drawn
from studies concentrated on the toxic and irritant effects of inhalation, direct dermal contact,
and ingestion exposures to formaldehyde. Hence the evidentiary basis for the OSHA
conclusions is not the same as the one used in the present report where only inhalation exposure
was considered and where possible carcinogenic effects and mortality were emphasized to the
near exclusion of irritants and other morbidity effects.
Owing to OSHA's administrative mission, much of the evidence compiled for its
determination of formaldehyde carcinogenicity was drawn from epidemiological studies of
human occupational exposures. Thus the data categories used by the OSHA analysts tend not to
repeat those of the present report to the extent that the WHO formaldehyde report did. This
makes examination of the OSHA conclusion on formaldehyde useful for judging the
completeness of the categories set out under the framework evaluated in the present report.
However, as was the case with the evidence assembled for the WHO conclusion, the individual
results listed by OSHA have not been evaluated by the OSHA analysts as to their Completeness
or Utility or Strength of Effect, making intercomparisons of judgments concerning data quality
given in the present report difficult. The evidence listed in the OSHA Rule were these:
9.4.2.1. Human Irritation and Morbidity Effects
A. Inhalation of formaldehyde at >50 ppm produces pulmonary edema, pneumonia, and
bronchial irritation "which can result in death";
B. Inhalation at concentrations >5 ppm cause lower airway irritation and constriction;
244
-------
C. "Formaldehyde can produce symptoms of bronchial asthma in humans" either through
sensitization by formaldehyde exposure or direct irritation by formaldehyde in a pre-existing
case;
D. Tolerance at low level exposures (< 1 ppm) permits continued exposure to increasing
level of formaldehyde;
E. Eye irritation (burning, itching, redness, increased tearing) follows on exposures of
0.05-0.5 ppm, though no irreversible tissue damage is observed;
F. Dermal exposure "can cause" allergic contact dermatitis at levels below 1 ppm;
G. Gastrointestinal toxicity and death can follow ingestion of as little as 30 mis of 37%
formaldehyde solution;
H. Damage to liver, kidney, spleen, stomach, pancreas, brain, CNS can result from
ingestion of even lesser concentrations.
9.4.2.2. Human Cancer Epidemiology
"Long term exposure to formaldehyde has been shown to be associated with an increased
risk of cancer of the nose and accessory sinuses, nasopharyngeal and oropharyngeal cancer, and
lung cancer in humans."
9.4.2.3. Carcinogenic Effects in Experimental Animals
"Animal experiments provide conclusive evidence of a
causal relationship between nasal cancer in rats and formaldehyde exposure."
9.4.2.4. Concordant Evidence
245
-------
"Concordant evidence of carcinogenicity includes DNA binding, genotoxicity in short-
term tests, and cytotoxic changes in the cells of the target organ suggesting both preneoplastic
changes and a dose-rate effect."
9.4.2.5. OSHA Conclusions About Formaldehyde Carcinogenicity
The present analysts recognize that both the form of the OSHA report - a Rule for a
time-weighted average exposure duration — and its contents do not permit a direct comparison
to the evaluation performed with the framework used here. However, it is clear that without
some framework for evaluating both the individual claims about study results used as evidence
in a determination and the strength of the data on which those claims rest, the OSHA
conclusion about formaldehyde (that "formaldehyde is a complete carcinogen and appears to
exert an effect on at least two stages of the carcinogenic process) appears incompletely
warranted. This is, of course, not to say that the conclusion IS unwarranted, only that without a
more explicit accounting of the evaluation elements (Completeness, Utility, Strength of Effect,
Relevance Strategy) no reconstruction of the conclusion is possible. Hence the bare judgment
that stands as OSHA's conclusion about formaldehyde appears to be incompletely warranted
even by the data it lists as evidence.
246
-------
APPENDIX A. CITATIONS FOR FORMALDEHYDE LITERATURE SEARCH
Adams, D. O., T. A. Hamilton, L. D. Lauer, and J. H. Dean. The Effect of Formaldehyde
Exposure Upon the Mononuclear Phagocyte System of Mice." Toxicology and Applied
Pharmacology 88 (1987): 165-74.
Albert, R. E., A. R. Sellakumar, S. Laskin, M. Kuschner, N. Nelson, and C. A. Snyder. "Gaseous
Formaldehyde and Hydrogen Chloride Induction of Nasal Cancer in the Rat." Journal of
the National Cancer Institute 68, no. 597-603 (1982).
AMA Council on Scientific Affairs. "Formaldehyde." Journal of the American Medical
Association 261, no. 8 (February 1989): 1183- 1187.
Andersen, K. E., A. Boman, A. Volund, and J. E. Wahlberg. "Induction of Formaldehyde
Contact Sensitivity: Dose-Response Relationship in the Guinea-pig Maximization Test."
Acta Dermato- venereologica (Stockholm) 65, no. 472-78 (1985).
Ashby, J., and P. Lefevre. "Genetic Toxicology Studies with Formaldehyde and Closely Related
Chemicals Including Hexamethylphosphoramide." in Proceedings of the Third CUT
Conference on Toxicity: Formaldehyde Toxicity. 85-97. (ed.) J. E. Gibson. New York:
Hemisphere Publishers, 1983.
Auerbach, C., M. Moutschen-Dahmen, and J. Moutschen. "Genetic and Cytotoxic Effects of
Formaldehyde and Related Compounds." Mutation Research 39 (1977): 317-362.
Babich, H. "Reproductive and Carcinogenic Health Risks to Hospital Personnel from Chemical
Exposure - A Literature Review." Journal of Environmental Health 48, no. 2 (1985):
52-56.
Ballarin, C., F. Sarto, L. Giacomelli, G. B. Bartolucci, and E. Clonfero. "Micronucleated Cells in
Nasal Mucosa of Formaldehyde- Exposed Workers." Mutation Research 280, no. 1
(1992): 1-7.
Bardana jr, E. J., and A. Montanaro. "Formaldehyde: An Analysis of Its Respiratory, Cutaneous,
and Immunologic Effects." Annals of Allergy 66, no. 6 (1991): 441-52.
Barrow, C. S., W. H. Steinhagen, and J. C. F. Chang. "Formaldehye Sensory Irritation." in
Proceedings of the Third CUT Conference on Toxicity: Formaldehyde Toxicity. 16-24.
(ed.) J. E. Gibson. New York: Hemisphere Publishers, 1982.
Bartnik, F. G., C. Gloxhuber, and V. Zimmermann. "Percutaneous Absortption of Formaldehyde
in Rats." Toxicology Letters 25 (1985): 167-172.
Basler, A., W. V. D. Hude, and M. Scheutwinkel-Reich. "Formaldehyde Induced Sister
Chromatid Exchanges in vitro and the Influence of the Exogenous Metabolizing Systems
S9 Mix and Primary Rat Hepatocytes." Archives of Toxicology 58 (1985): 10- 13.
247
-------
Bauchinger, M., and E. Schmid. "Cytogenic Effects in Lymphocytes of Formaldehyde Workers of
a Paper Factory." Mutation Research 158 (1985): 195-199.
Benyajati, C., A. R. Place, and W. Sofer. "Formaldehyde Mutagenesis in Drosophila Molecular
Analysis of ADH-negative Mutants." Mutation Research 111 (1983): 1-7.
Bermudez, E., and L. L. Delehanty. "The Effects of in vitro Formaldehyde Treatment on the
Cells of the Rat Nasal Epithelium." Environmental Mutagenesis 8(Suppl.6) (1986): 11.
Billings, R. E., M. E. Ku, M. E. Brower, C. E. Dallas, and J. C. Theiss. "Disposition of
Formaldehyde in Mice." Toxicologist 4, no. 29 (1984): 466.
Blair, A., A. Linos, P. A. Stewart, L. F. Burmeister, R. Gibson, G. Everett, L. Schuman, and K.
P. Cantor. "Comments on Occupational and Environmental Factors in the Origin of Non-
Hodgkin's Lymphoma." Cancer Research 52, no. 19 Suppl. (1992): 5501s-5502s.
Blair, A., R. Saracci, P. A. Stewart, R. B. Hayes, and C. Shy. "Epidemiologic Evidence on the
Relationship Between Formaldehyde Exposure and Cancer." Scandinavian Journal of
Work. Environment and Health 16, no. 6 (1990): 381-93.
Blair, A., P. Stewart, R. N. Hoover, J. F. Fruameni, J. Walrath, M. O'Berg, and W. Gaffey.
"Cacner of the Nasopharynx and Oropharynx and Formaldehyde Exposure." Journal of
the Nutional Cancer Institute 78 (1987): 191-193.
Blair, A., P. A. Stewart, R. N. Hoover et al. "Cancer of the Nasopharynx and Oropharynx and
Formaldehyde Exposure." Journal of the National Cancer Institute 78 (1987): 191-93.
Blair, A., and P. A. Stewart. "Correlation Between Different Measures of Occupational Exposure
to Formaldehyde." American Journal of Epidemiology 131, no. 3 (1990): 510-516.
Blair, A., P. Stewart, M. O'Berg, W. Gaffay, J. Walrath, J. Ward, R. Bales, S. Kaplan, and D.
Cubit. "Mortality Among Industrial Workers Exposed to Formaldehyde." Journal of the
National Cancer Institute 76 (1986): 1071-1084.
Blair, A., P. A. Stewart, and R. N. Hoover. "Mortality from Lung Cancer Among Workers
Employed in Formaldehyde Industries." American Journal of Industrial Medicine 17, no.
6 (1990): 683- 699.
Bolt, H. M. "Experimental Toxicology of Formaldehyde." Journal of Cancer Research and
Clinical Oncology 113 (1987): 305-09.
Boreiko, C. J., and D. L. Ragan. "Formaldehyde Effects in the C3H/10T1/2 Cell Transformation
Assay." in Proceedings of the Third CUT Conference on Toxicity: Formaldehyde
Toxicity. 63-71. (ed.) J. E. Gibson. New York: Hemisphere Publishers, 1983.
Brendel, R. "In German: [Hexamethylenetetramine tolerance in rats]." Forschritte der
Arzneimittelforschung 14 (1964): 51-53.
248
-------
Brinton, L. A., W. J. Blot, J. A. Becker, D. M. Winn, J. R. Browder, J. C. Farmer, and J. F.
Fraumeni. "A Case-control Study of Cancers of the Nasal Cavity and Paranasal Sinuses."
American Journal of Epidemiology 119, no. 896-906 (1984):
Brusick, D. J. "Genetic and Transforming Activity of Formaldehyde." in Proceedings of the Third
CIIT Conference on Toxicity: Formaldehyde Toxicity. 72-84. (ed.) J. E. Gibson. New
York: Hemisphere Publishers, 1983.
Brusick, D., D. Jagannath, B. Myhr, and D. Stetka. "The Genetic Activity of Paraformaldehyde
in the Ames Assay, the L5178Y Mouse Lymphoma Assay, the CHO SCE Assay, and in
an in vitro CHO Chromosome Analysis." Environmental Mutagenesis 2 (1980): 253.
Casanova-Schmitz, M., and H. d.'A Heck. "Effects of Formaldehyde Exposure on the
Extractability of DNA from Proteins in the Rat Nasal Mucosa." Toxicology and Applied
Pharmacology 70 (1983): 121-132.
Casanova-Schmitz, M., and H. d.'A Heck. "Further Studies of the Metabolic Incorporation and
Covalent Binding of Inhaled [3HJ- and [14C] Formaldehyde in Fischer 344 Rats: Effects
of Glutation Depletion." Toxicology and Applied Pharmacology 89 (1987): 105-121.
Casanova-Schmitz, M., T. B. Starr, and H. d'A Heck. "Differentiation Between Metabolic
Incorporation and Covalent Binding in the Labeling of Macromolecules in the Rat Nasal
Mucosa and Bone Marrow by Inhaled [14C] and [3H] Formaldehyde." Toxicology and
Applied Pharmacology 76 (1984): 26-44.
Casteel, S. W., R. J. Vernon, and E. M. Bailey jr. "Formaldehyde: Toxicology and Hazards."
Veterinary and Human Toxicology 29, no. 1 (1987): 31-33.
Chang, C. C., and M. E. Gershwin. "Perspectives on Formaldehyde Toxicity: Separating Fact
from Fantasy." Regulatory Toxicology and Pharmacology 16, no. 2 (1992): 150-160.
Chang, J. C. F., E. A. Gross, J. A. Swenberg, and C. S. Burrow. "Nasal Cavity Deposition,
Histopathology, and Cell Proliferation After Single or Repeated Formaldehyde Exposure
in B6C3F1 Mice and F-344 Rats." Toxicology and Applied Pharmacology 68 (1983): 161-
76.
Formaldehyde: Toxicology. Epidemiology. Mechanisms, (eds.) John J. Clary, James E. Gibson,
and Richard S. Waritz. New York, USA: Marel Dekker, 1983.
Coggon, D., B. Pannett, and E. D. Acheson. "Use of Job-Exposure Matrix in an Occupational
Analysis of Lung and Bladder Cancers on the Basis of Death Certificates." Journal of the
National Cancer Institute 72, no. 1 (1984): 61-65.
Connor, T. H. "Absence of Mutagenicity in the Urine of Autopsy Service Workers Exposed to
Formaldehyde: Factors Influencing Mutagenicity Testing of Urine." International
Archives of Occupational and Environmental Health 56 (1985): 225-237.
249
-------
Consensus Workshop on Formaldehyde. "Report on the Consensus Workshop on
Formaldehyde." Environmental Health Perspectives 58 (1984): 323-381.
Coyne, L. B., R. E. Cook, J. R. Mann, S. Bouyoucos, O. F. McDonald, and C. L. Baldwin.
"Formaldehyde: A Comparative Evaluation of Four Monitoring Methods." American
Industrial Hygiene Association Journal 46, no. 10 (1985): 609-619.
Craft, T. R., E. Bermudez, and T. R. Skopek. "Formaldehyde Mutagenesis and Formation of
DNA-Protein Crosslinks in Human Lymphoblasts in vitro." Mutation Research 176
(1987): 147-155.
Dalbey, W. E. "Formaldehyde and Tumors in Hamster Respiratory Tract." Toxicology 24 (1982):
9-14.
Dally, K. A., L. P. Hamrahan, M. A. Woodbury, and M. S. Kanarek. "Formaldehyde Exposure in
Non-Occupational Environments." Archives of Environmental Health 36 (1981): 277-284.
Dean, J. H., L. D. Lauer, R. H. House, M. J. Murray, W. S. Stillman, R. D. Irons, W. H.
Steinhagen, M. C. Phelps, and D. O. Adams. "Studies of Immune Function and Host
Resistance in B6C3F1 Mice Exposed to Formaldehyde." Toxicology and Applied
Pharmacology 72 (1984): 519-29.
Delia Porta, G., F. R. Cabral, and G. Parmiani. "In Italian: [Transplacental Toxicity and
Carcinogenesis Studies in Rats with Hexamethylenetetramine." Tumori 56 (1970):
325-334.
Delia Porta, G., M. I. Colnaghi, and G. Parmiani. "Noncarcinogenicity of
Hexamethylenetetramine in Mice and Rats." Food and Cosmetics Toxicology 6 (1968):
707-715.
Dooley, J. F., G. R. Blackburn, C. A. Schreiner, and C. R. MacKerer. "Inhibition or the
Mutagenicity of Formaldehyde (HCHO) in the L5178Y TK+/- Mouse Lymphomas
Assay by Formaldehyde Dehydrogenase (FDH): Application to Characterization of
Mutagenic Components." Environmental Mutagenesis 7 (1985): 9.
Dosemeci, M., A. Blair, P. A. Stewart, J. Chandler, and M. A. Trush. "Mortality Among
Industrial WOrkers Exposed to Phenol." Epidemiology 2, no. 3 (1991): 188-93.
Dykewicz, M. S., R. Patterson, D. W. Cugell, K. E. Harris, and A. F. Wu. "Serum IgE and IgG
to Formaldehyde-Human Serum Albumin: Lack of Relation to Gaseous Formaldehyde
Exposure and Symptoms." Journal of Allergy and Clinical Immunology 87, no. l(Pt 1)
(1991): 48-57.
Escalante-Semerena, J. C., and R. S. Wolfe. "Formaldehyde Oxidation and Methanogenesis."
Journal of Bacteriology 158, no. 2 (1984): 721-26.
250
-------
Federal Panel on Formaldehyde. "Report of the Federal Panel on Formaldehyde."
Environmental Health Perspectives 43 (1982): 139- .168.
Formaldehyde Sensitivity and Toxicology, (ed.) Susan E. Feinman. Boca Raton, FL USA: CRC
Press, 1988.
Feron, V. J., J. P. Bruyntjes, J. J. Woutersen, H. R. Jmmel, and L. M. Appelman. Nasal
Tumours in Rats After Short-term Exposure to a Cytotoxic Concentration of
Formaldehyde. Zeist, The Netherlands: CIVO Institute TNO, 1987.
Feron, V. J., H. R. Jemmel, J. W. G. M. Wilmer, R. A. Woutersen, and A. Zwart. "Nasal tumors
in rats after severe injury to the nasal mucosa and exposure to formaldehyde vapor:
Preliminary Results." in The Role of Formaldehyde in Biological Systems. 73- 77. (eds.)
E. Tyihak, and G. Gullner. Budapest: Sote Press, 1987.
Feron, V. J., H. P. Til, F. de Vrijer, R. A. Woutersen, F. R. Cassee, and P. J. van Bladeren.
"Aldehydes: Occurrence, Carcinogenic Potential, Mechanism of Action, and Risk
Assessment." Mutation Research 259, no. 3-4 (1991): 363-85.
Florence, E., and D. F. Milner. "Determination of Free and Loosely Protein-Bound
Formaldehyde in the Tissues of Pigs Fed Formalin-Treated Skim Milk as a Protein
Supplement." Journal Sci Food Agric 32 (1981): 288-92.
Fontignie-Houbrechts, N. "Genetic Effects of Formaldehyde in the Mouse." Mutation Research
88 (1981): 109-114.
Frazier, T. M., N. R. Lalich, and O. H. Pedersen. "Uses of Computer-Generated Maps in
Occupational Hazard and Mortality Surveillance." Scandinavian Journal of Work-
Environment. and Health 9, no. 2 (1983): 148-154.
Frigas, E., W. V. Filley, and C. E. Reed. "Bronchial Challenge with Formaldehyde Gas: Lack of
Bronchoconstriction in 13 Patients Suspected of Having Formaldehyde-Induced Asthma."
Mayo Clinic Proceedings 59 (1984): 295-99.
Formaldehyde Toxicity, (ed.) James E. Gibson. New York, USA: Hemisphere Publishing, 1983.
Goldmacher, V. S., and W. G. Thilly. "Formaldhyde is Mutagenic for Cultured Human Cells."
Mutation Research 116 (1983): 417-22.
Goldmacher, V. S., P. Temcharoen, and W. G. Thilly. "Mutagenic Effects of Formaldehyde in
Bacterial and Human Cells." in Formaldehyde: Toxicology. Epidemiology. Mechanisms.
173-192. (eds.) J. J. Clary, J. E. Gibson, and R. S. Waritz. New York: Marcel Dekker,
1982.
Gottschling, L. M., H. J. Beaulieu, and W. W. Melvin. "Monitoring of Formic Acid in Urine of
Humans Exposed to Low Levels of Formaldehyde." American Industrial Hygiene
Association Journal 45 (1984): 19-23.
251
-------
Grafstrom, R. C. "In Vitro Studies of Aldehyde Effects Related to Human Respiratory
Carcinogenesis." Mutation Research 238, no. 3 (1990): 175-84.
Grafstrom, R. C., A. J. Fornace, H. Autrup, J. F. Lechner, and C. C. Harris. "Formaldehyde
Damage to DNA and Inhibition of DNA Repair in Human Bronchial Cells." Science 220
(1983): 216-18.
Grafstrom, R. C., A. J. Fornace, and C. C. Harris. "Repair of DNA Damage Caused by
Formaldehyde in Human Cells." Cancer Research 44 (1984): 4323-4327.
Harrington, J. M., and D. Oakes. "Mortality Study of British Pathologists (1974-1980)." British
Journal of Industrial Medicine 41 (1984): 188-91.
Harving, H., J. Korsgaard, R. Dahl, O. F. Pedersen, and L. Molhave. "Low Concentrations of
Formaldehyde in Bronchial Asthma: A Study of Exposure Under Controlled Conditions."
British Medical Journal 293 (1986): 310.
Harving, H., J. Korsgaard, O. F. Pedersen, L. Molhave, and R. Dahl. "Pulmonary Function and
Bronchial Reactivity in Asthmatics During Low-level Formaldehyde Exposure." Lung 168,
no. 1 (1990): 15-21.
Hattis, D. "Pharmacokinetic Principle for Dose-Rate Extrapolation of Carcinogenic Risk from
Genetically Active Agents." Risk Analysis 10, no. 2 (1990): 303-316.
Hayes, R. B., J. W. Raatgever, A. D. Bruyn, and M. Gerin. "Cancer of the Nasal Cavity and
Paranasal Sinuses, and Formaldehyde Exposure." International Journal of Cancer 37
(1986): 487-92.
Heck, H. d'A. "Summary of Final Report on Formaldehyde Study (Abstracts): Biochemical
Toxicology of Inhaled Formaldehyde." Chemical Industry Institute of Toxicology Acts 2
(1982): 1-7.
Heck, H. d'A., and M. Casanova-Schmitz. "Reaction of Formaldehyde in the Rat Nasal Mucosa."
in Formaldehyde: Toxicology. Epidemiology. Mechanisms. 211-224. (eds.) J. J. Clary, J.
E. Gibson, and R. S. Waritz. New York: Marcel Dekker, 1982.
Heck, H. d'A., and M. Casanova-Schmitz.. "The Relevance of Disposition Studies to the
Toxicology of Formaldehyde." Chemical Industry Institute of Toxicology Acts 4, no. 5
(1984): 1-6.
Heck, H. d'A., T. Y. Chin, and M. Casanova-Schmitz. "Distribution of [14C] Formaldehyde in
Rats after Inhalation Exposure." in Proceedings of the Third CUT Conference on
Toxicity: Formaldehyde Toxicity. 26-36. (ed.) J. E. Gibson. New York: Hemisphere
Publishers, 1982.
Heck, H. d'A, Casanova-Schmitz, M., P. B. Dodd, E. N. Schachter, T. J. Witek, and T. Tosun.
"Formaldehyde Concentrations in the Blood of Humans and Fischer-344 Rats Exposed
252
-------
to Formaldehyde Under Controlled Conditions." American Industrial Hygiene
Association Journal 46 (1985): 1-3.
Hendrick, D. J., R. J. Rando, D. J. Lane, and M. J. Morris. "Formaldehyde Asthma: Challenge
Exposure Levels and Fate After Five Years." Journal of Occupational Medicine 24
(1982): 893-97.
Henry, C. J., L. Fishbein, W. J. Meggs, N. A. Ashford, P. A. Schulte, H. Anderson, J. S.
Osborne, and D. W. Sepkovic. "Approaches for Assessing Health Risks from Complex
Mixtures in Indoor Air: A Panel Overview." Environmental Health Perspectives 95, no.
Nov (1991): 135-143.
Holmstrom, M., and V. J. Lund. "Malignant Melanomas of the Nasal Cavity after Occupational
Exposure to Formaldehyde." British Journal of Industrial Medicine 48, no. 1 (1991): 9-11.
Hooper, K., J. LaDou, J. S. Rosenbaum, and S. A. Book. "Regulation of Priority Carcinogens
and Reproductive or Developmental Toxicants [erratum in 23,4(1993)673] ." American
Journal of Industrial Medicine 22, no. 6 (1992): 793-808.
Iversen, O. H. "Formaldehyde and Skin Carcinogenesis." Environmental Int 12 (1986): 541-44.
Jackson, Silas. Formaldehyde Toxicology Literature Search: January 1977 - June 1982. Bethesda,
MD: USDHHS, PHS, NLM, 1982.
Jeffcoat, A. R., F. Chasalow, D. B. Feldman, and H. Marr. "Disposition of [14C] Formaldehyde
after Topical Exposure to Rats, Guinea Pigs, and Monkeys." in Proceedings of the Third
CIIT Conference on Toxicity: Formaldehyde Toxicity. 38-50. (ed.) J. E. Gibson. New
York: Hemisphere Publishers, 1982.
Jensen, N. J., and K. H. Cohr. Testing of Formaldehyde in the Mammalian Spot Test by
Inhalation." Mutation Research 113, no. 266 (1983):
Johannsen, F. R., G. J. Levinskas, and A. S. Tegeris. "Effects of Formaldehyde in the Rat and
Dog Following Oral Exposure." Toxicology Letters 30 (1986): 1-6.
Kauppinen, T., and T. Partanen. "Use of Plant- and Period- Specific Job-Exposure Matrices in
Studies on Occupational Cancer." Scandinavian Journal of Work. Environment, and
Health 14, no. 3 (June 1988): 161-167.
Kerns, W. D., D. J. Donofrio, and K. L. Pavkov. "The Chronic Effects of Formaldehyde Vapor
Inhalation in Rats and Mice: A Preliminary Report." in Preceedings of the Third CIIT
Conference on Toxicology: Formaldehyde Toxicology. 111-131. (ed.) J. E. Gibson. New
York: Hemisphere Publishers, 1982.
Kerns, W. D., K. L. Pavkov, D. J. Donofrio, E. J. Gralla, and J. A. Swenberg. "Carcinogenicity of
Formaldehyde in Rats and Mice after Long-term Inhalation Exposure." Cancer Research
43, no. 4382-4392 (1983):
253
-------
Khamgaonkar, M. B., and M. B. Fulare. "Pulmonary effects of Formaldehyde Exposure « An
Environmental Epidemiological Study." Indian Journal of Chest Diseases and Allied
Sciences 33, no. 1 (1991): 9-13.
Kilburn, K. H., R. Warshaw, C. T. Boylen, S. J. S. Johnson, B. Seidman, R. Sinclair, and T.
Takaro jr. "Pulmonary and Neurobehavioral Effects of Formaldehyde Exposure."
Archives of Environmental Health 40 (1985): 254-60.
Klein-Szanto, A. J., H. Ura, S. Momiki, D. Bonfil, and S. Litwin. "Effects of Formaldehyde on
Xenotransplanted Human Respiratory Epithelium." Respiratory Reports of the Health
Effects Institute 51, no. July (1992): 1-17, 19-27.
Kligerman, A. D., M. C. Phelps, and G. L. Erexson. "Cytogenetic Analysis of Lymphocytes from
Rats Following Formaldehyde Inhalation." Toxicology Letters 21 (1984): 241-46.
Kreiger, R. A., and V. F. Garry. "Formaldehyde-Induced Cytotoxicity and Sister Chromatid
Exchanges in Human Lymphocytes Cultures." Mutation Research 120 (1983): 51-55.
Krivanek, N. D., N. C. Chromey, and J. W. McAlack. "Skin Initiation/Promotion Study with
Formaldehyde in CD-I Mice." in Formaldehyde: Toxicology. Epidemiology. Mechanisms.
159-172. (eds.) J. J. Clary, J. E. Gibson, and R. S. Waritz. New York: Marcel Dekker,
1982.
Lam, CH W., M. Casanova-Schmitz, and H. d'A Heck. "Depletion of Nasal Mucosal Glutathione
to Acrolein and Enhancement of Formaldehyde Induced DNA-Protein Cross Linking by
Simultaneous Exposure to Acrolein." Archives of Toxicology 58 (1985): 67-71.
Lebowitz, M. D., J. J. Quackenboss, Krzyzanowski. M., M. K. O'Rourke, and C. Hayes.
"Multipoint Exposures and Health Responses to Particulate Matter." Archives of
Environmental Health 47. no. 1 (1992): 71-75.
Levine, R. J., D. A. Andjelkovich, and L. K. Shaw. The Mortality of Ontario Undertakers and a
Review of Formaldehyde-Related Mortality Studies." Journal of Occupational Medicine
26 (1984): 740-746.
Levy, S., S. Nocentini, and C. Billardon. "Induction of Cytogenic Effects in Human Fibroblast
Cultures After Exposure to Formaldehyde or X-rays." Mutation Research 119 (1983):
309-317.
Loden, M. "The in vitro Permeability of Human Skin to Benzene, Ethylene Glycol,
Formaldehyde, and N-hexane." Acta Pharmacologica et Toxicologica 58 (1986): 382-89.
Luce, D., M. Gerin, A. Leclerc, J. F. Morcet, J. Brugere, and M. Goldberg. "Sinonasal Cancer
and Occupational Exposure to Formaldehyde and Other Substances." International
Jounal of Cancer 53, no. 2 (1993): 224-231.
254
-------
Main, D. M., and T. J. Hogan. "Health Effects of Low-Level Exposure to Formaldehyde."
Journal of Occupational Medicine 25 (1983): 896-900.
Malaka, T., and A. M. Kodama. "Respiratory Health of Plywood Workers Occupationally
Exposed To Formaldehyde." Archives of Environmental Health 45, no. 5 (1990): 288-294.
Malker, H. S., J. K. McLaughlin, J. A. Weiner, D. T. Silverman, W. J. Blot, J. L. Ericsson, and J.
F. Fraumeni jr. "Occupational Risk Factors for Nasopharyngeal Cancer in Sweden."
British Journal of Industrial Medicine 47, no. 3 (1990): 213-214.
Malorny, G., N. Rietbrock, and M. Schneider. "Die Oxydation des Formaldehyds zu
Ameisensaeure im Blut, ein Beitrag zum Stoffwechsel des Formaldehyds." Archives of
Experimental Pathology and Pharmacology 250 (1965): 419-436.
Marinari, U. M., M. Ferro, L. Sciaba, R. Finollo, A. M. Bassi, and G. Brambilla.
"DNA-Damaging Activity of Biotic and Xenobiotic Aldehydes in Chinese Hamster Ovary
Cells." Cell Biochemistry and Function 2 (1984):243-48.
Marsh, G. M., R. A. Stone, and V. L. Henderson. "A Reanalysis of the National Cancer Institute
Study on Lung Cancer Mortality Among Industrial Workers Exposed to Formaldehyde."
Journal of Occupational Medicine 34, no. 1 (1992): 42-44.
Marsh, G. M., R. A. Stone, and V. L. Henderson.. "Lung Cancer Mortality Among Industrial
Workers Exposed to Formaldehyde: A Poisson Regression Analysis of the National
Cancer Institute Study." American Industrial Hygiene Association Journal 53, no. 11
(1992): 681-91.
Maurice, F., J. P. Rivory, P. H. Larsson, S. G. O. Johannson, and J. Bousquet. "Anaphylactic
Shock Caused by Formaldehyde in a Patient Undergoing Long-Term Hemodialysis."
Journal of Allergy and Clinical Immunology 77 (1986): 594-97.
McMartin, K. E., G. Martin-Amat, P. E. Noker, and T. R. Tephly. "Lack of Role for
Formaldehyde in Methanol Poisoning in the Monkey." Biochemistry and Pharmacology
28 (1979): 645-49.
Mitchell, R. I., K. L. Ravkov, R. L. Persing, and D. A. Holzworth. Final Report on a 90-Day
Inhalation Toxicity Study in Rats and Mice Exposed to Formaldehyde. RTP, NC: CIIT,
1979.
Monteiro-Riviere, N. A., and J. A. Popp. "Ultrastructural Evaluation of Acute Nasal Toxicity in
the Rat Respiratory Epithelium in Response to Formaldehyde Gas." Fundamental and
Applied Toxicology 6, no. 251-62 (1986):
Monticello, T. M., K. T. Morgan, J. I. Everitt, and J. A. Popp. "Effect of Formaldehyde Gas on
the Respiratory Tract of Rhesus Monkeys: Pathology and Cell Proliferation." American
Journal of Pathology 134, no. 3 (1989): 515-20.
255
-------
Morgan, K. T., D. L. Patterson, and E. A. Gross. "Formaldehyde and the Nasal Mucociliary
Apparatus." in Formaldehyde: Toxicology. Epidemiology. Mechanisms. 193-210. (eds.) J.
J. Clary, J. E. Gibson, and R. S. Waritz. New York: Marcel Dekker, 1982.
National Research Council. Formaldehyde: An Assessment of Its Health Effects. Washington,
D.C.: National Research Council, 1980.
National Research Council. Formaldehyde and Other Aldehydes. Washington, D.C.: National
Research Council/National Academy Press, 1981.
Nettesheim, P. Respiratory Carcinogenesis: Report of Progress jn FY 1976 and 1977.
Washington, D.C.: NCI-DOE Cancer Research Program, 1977.
NIOSH. NIOSH Criteria For A Recommended Standard Occupational Exposure to
Formaldehve. Cincinnati, OH: National Institute of Occupational Safety and Health,
1976.
Nordman, H., H. Keskinen, and M. Tuppurainen. "Formaldehyde Asthma: Rare or
Overlooked?" Journal of Allergy and Clinical Immunology 75 (1985): 91-8.
Nunn, A. J., A. A. Craigen, J. H. Darbyshire, and K. M. Venables. "Six Year Follow Up of Lung
Function in Men Occupationally Exposed to Formaldehyde." British Journal of Industrial
Medicine 47 (1990): 747-52.
Occupational Safety and Health Administration. Occupational Exposure to Formaldehyde: Final
Rule. Washington, D.C.: USDoL, 27 May 1992.
Olsen, J. H., and S. Asnaes. "Formaldehyde and the Risk of Squamous Cell Carcinoma of the
Sinonasal Cavities." British Journal of Industrial Medicine 43 (1986): 769-774.
Olsen, J. H., S. P. Jensen, M. Hink, K. Faurbo, N. O. Breum, and O. M. Jensen. "Occupational
Formaldehyde Exposure and Increased Nasal Cancer Risk in Man." International Journal
of Cancer 34, no. 5 (1984): 639-644.
Overman, O. "Absence of Embryonic Effects of Formaldehyde After Percutaneous Exposure in
Hamsters." Toxicology Letters 24 (1985): 107-110.
Owen, B. A., C. S. Dudney, E. L. Tan, and C. E. Easterly. "Formaldehyde in Drinking Water:
Comparative Hazard Evaluation and An Approach to Regulation." Regulatory
Toxicology and Pharmacology 11, no. 3 (1990): 220-36.
Partanen, T. "Formaldehyde Exposure and Respiratory Cancer - A Meta-analysis of the
Epidemiological Evidence." Scandinavian Journal of Work. Environment, and Health 19,
no. 1 (1993): 8-15.
Partanen, T., T. Kauppinen, R. Luukkonen, T. Hakulinen, and E. Pukkala. "Malignant
Lymphomas and Leukemias, and Exposures in the Wood Industry: An Industry-Based
256
-------
Case-Referent Study." International Archives of Occupational and Environmental Health
64, no. 8 (1993): 593-6.
Patterson, R. M., M. I. Bornstein, M. S. Bogdanffy, and K. T. Morgan. "Retention of
Formaldehyde Gas by the Nasal Passages of F-344 Rats." Toxicologist 6 (1986): 55.
Perera, Frederica, and Catherine Petito. "Formaldehyde: A Question of Cancer Policy." Science
216, no. 18 June (1982): 1285-1290.
Perkins, J. L., and J. D. Kimbrough. "Formaldehyde Exposure in a Gross Anatomy Laboratory."
Journal of Occupational Medicine 27. no. 11 (1985): 813-815.
Plesner, B. H., and K. Henseon. "Formaldehyde and Hexamethylenetetramine in Styles Cell
Transformation Assay." Carcinogenesis 4 (1983): 457-59.
Pross, H. F., J. H. Day, R. H. Clark, and R. E. Lees. "Immunologic Studies of Subjects with
Asthma Exposed to Formaldehyde and Urea-Formaldehyde Foam Insulation off
Products." Journal of Allergy and Clinical Immunology 79, no. 5 (1987): 797- 810.
Ragan, D. L., and C. J. Boreiko. "Initiation of C3H/10T1/2 Cell Transformation by
Formaldehyde." Cancer Letters 13 (1982): 325- 331.
Restani, P., and C. L. Galli. "Oral Toxicology of Formaldehyde and Its Derivatives." Critical
Reviews in Toxicology 21, no. 5 (1991): 315-328.
Robbins, J. D., W. P. Norred, A. Bathija, and A. G. Ulsamer. "Bioavailability in Rabbits of
Formaldehyde From Durable-Press Textiles." Journal of Toxicology and Environmental
Health 14 (1984): 453-63.
Ross, W. E., D. R. McMillan, and C. F. Ross. "Comparison of DNA Damage by
Methylenetetramine and Formaldehyde." Journal of the National Cancer Institute 67
(1981): 217-221.
Ross, W. E., D. R. McMillan, and C. F. Ross. "Comparison of DNA Damage By
Methylmelamines and Formaldehyde." Journal of the National Cancer Institute 67
(1981): 217-221.
Ross, W. E., and N. Shipley. "Relationship Between DNA Damage and Survival in
Formaldehyde-Treated Mouse Cells." Mutation Research 79 (1980): 277-283.
Rusch, G. M., J. J. Clary, W. E. Rinehart, and H. F. Bolte. "A 26-Week Inhalation Toxicity
Study with Formaldehyde in the Monkey, Rat, and Hamster." Toxicology and Applied
Pharmacology 68 (1983): 329-43.
Rutala, W. A. "APIC Guideline for Selection and Use of Disinfectants." American Journal of
Infection Control 18, no. 2 (1990): 99-117.
257
-------
Rutala, W. A.. "Draft Guidelines for Selection and Use of Disinfectants." American Journal of
Infection Control 17, no. 1 (February 1989): 24A-38A-
Sanner, T., P. Eker, and E. Riverdal. "Cell Transformation by Inhalation of Formaldehyde."
Mutation Research 113 (1983): 303.
Schachter, E. N., T. J. Witek, T. Tosun, B. P. Leaderer, and G. J. Beck. "A Study of Respiratory
Effects from Exposure to 2 ppm Formaldehyde in Healthy Subjects." Archives of
Environmental Health 41 (1986): 229-39.
Schmid, E., W. Goggelmann, and M. Bauchinger. "Formaldehyde- Induced Cytotoxic, Genotoxic,
and Mutagenic Response in Human Lymphocytes and Salmonella typhimurium."
Mutagenesis 1 (1986): 427-31.
Scott, M. J., J. B. Ward, C. E. Dallas, and J. C. Theiss. "Chromosome Damage Observed in Lung
but Not Bone Marrow of Sprague-Dawley Rats Exposed to Formaldehyde By Inhalation."
Environmental Mutagenesis 7 (1985): 53-4.
Sellakumar, A. R., C. A. Snyder, J. J. Solomon, and R. E. Albert. "Carcinogenicity of
Formaldehyde and Hydrogen Chloride in Rats." Toxicology and Applied Pharmacology
81 (1985): 401-406.
Siegel, David M., Vasilios H. Frankos, and Marvin A. Schneiderman. "Formaldehyde Risk
Assessment for Occupationally Exposed Workers." Regulatory Toxicology and
Pharmacology 3 (1983): 355-371.
Sielken jr., Robert L. "Some Capabilities, Limitations, and Pitfalls in the Quantitative Risk
Assessment of Formaldehyde." in Risk Analysis in the Chemical Industry: Proceedings of
a Symposium 23-24 September. 1985. 74-122. Chemical Manufacturers Association.
Rockville, MD: Government Institutes, 1985.
Smith, A. E. "Formaldehye." Occupational Medicine. Oxford 42, no. 2 (1992): 83-88.
Snyder, R. D., and B. Van Houten. "Genotoxicity of Formaldehyde and An Evaluation of Its
Effects on the DNA Repair Process in Human Diploid Fibroblasts." Mutation Research
165 (1986): 21-30.
Somers, E. "Risk Estimation for Environmental Chemicals as a Basis for Decision Making."
Regulations ]n Toxicology and Pharmacology 4, no. 2 (1984): 99-105.
Spangler, E. F., and J. M. Ward. "Skin Initiation/Promotion Study with Formaldehyde in
SENCAR Mice." in Formaldehyde: Toxicology. Epidemiology. Mechanisms. 147-158.
(eds.) J. J. Clary, J. E. Gibson, and R. S. Waritz. New York: Marcel Dekker, 1982.
Squire, Robert A., and Cameron Lorraine L. "An Analysis of Potential Carcinogenic Risk from
Formaldehyde." Regulatory Toxicology and Pharmacology 4 (1984): 107-129.
258
-------
Stankowski, L. F., W. G. Tuman, E. G. Godek, and G. J. Kasper. "Induction of Mammalian Cell
Mutations by Formaldehyde." Environmental Mutagenesis 8(Suppl 6) (1986): 81.
Starr, T. B. "Quantitative Cancer Risk Estimation for Formaldehyde." Risk Analysis 10, no. 1
(1990): 85-91.
Starr, T. B., and J. E. Gibson. "Mechanistic Toxicology of Formaldehyde and Its Implications for
Quantitative Risk Estimation." Annual Review of Pharmacology and Toxicology 25
(1985): 745-767.
Starr, T. B. "Improving Quantitative Risk Estimates with Mechanistic Data: The Case of
Formaldehyde." in Risk Analysis ]n the Chemical Industry: Proceedings of a Symposium
23-24 September. 1985. 141-153. Chemical Manufacturers Association. Rockville, MD:
Government Institutes, 1985.
Stayner, L., L. Elliot, L. Blade, R. Keenlyside, and W. Halperin. "A Retrospective Cohort
Mortality Study of Workers Exposed to Formaldehyde in the Garment Industry."
American Journal of Industrial Medicine 13 (1988): 796.
Steele, V. E., J. T. Arnold, J. Van-Arnold, and M. J. Mass. "Evaluation of a Rat Tracheal
Epithelial Cell Culture Assay System to Identify Respiratory Carcinogens."
Environmental and Molecular Mutagenesis 14, no. 1 (1989): 48-54.
Sterling, T. D., and J. J. Weinkam. "Reanalysis of Lung Cancer Mortality in a National Cancer
Institute Study on "Mortality Among Industrial Workers Exposed to Formaldehyde" [see
comments]." Journal of Occupational Medicine 30. no. 11 (November 1988): 895-890.
Stewart, P. A., C. Schairer, and A. Blair. "Comparison of Jobs, Exposures, and Mortality Risks
for Short-term and Long-term Workers." Journal of Occupational Medicine 32, no. 8
(1990): 703- 08.
Stock, T. H., and S. R. Mendez. "A Survey of Typical Exposures to Formaldehyde in Houston
Area Residences." American Industrial Hygiene Association Journal 46, no. 6 (1985):
313-317.
Stroup, N., A. C. Blair, and G. Erickson. "Brain Cancer and Other Causes of Death in
Anatomists." American Journal of Epidemiology 120 (1984): 500.
Sun, H. W., R. J. Feigal, and H. H. Messer. "Cytotoxicity of Glutaraldehyde and Formaldehyde
in Relation to Time of Exposure and Concentration." Pediatric Dentistry 12, no. 5
(1990): 303- 307.
Swenberg, J. A., E. A. Gross, H. W. Randall, and C. S. Barrow. "The Effect of Formaldehyde
Exposure in Cytotoxicity and Cell Proliferation." in Formaldehyde: Toxicology.
Epidemiology. Mechanisms. 225-236. (eds.) J. J. Clary, J. E. Gibson, and R. S. Waritz.
New York: Marcel Dekker, 1982.
259
-------
Swenberg, J. A., E. A. Gross, J. Martin, and J. A. Popp. "Mechanisms of Formaldehyde
Toxicity." in Proceedings of the Third CIIT Conference on Toxicity: Formaldehyde
Toxicity. 132- 150. Gibson, J. E. New York: Hemisphere Publishers, 1982.
Swenberg, J. A., H. d'A Heck, K. T. Morgan, and T. B. Starr. "A Scientific Approach to
Formaldehyde Risk Assessment." in Risk Quantitation and Regulatory Policy, pp.
255-281. (eds.) D. G. Hoel, R. A. Merrill, and F. P. Perera. Cold Spring Harbor, NY
USA: Cold Spring Harbor Laboratory, 1985.
Swenberg, J. A., W. D. Kerns, R. I. Mitchell, E. J. Gralla, and K. L. Pavkov. "Introduction of
Squamous Cell Carcinomas of the Rat Nasal Cavity by Inhalation Exposure to
Formaldehyde Vapor." Cancer Research 40 (1980): 3398.
Swiecichowski, A. L., K. J. Long, M. L. Miller, and G. D. Leikauf. "Formaldehyde-Induced
Airway Hyperreactivity in vivo and ex vivo in guinea pigs." Environmental Research 61,
no. 2 (1993): 185-99.
Thomson, E. J., S. Shackleton, and J. M. Harrington. "Chromosome Aberrations and
Sister-Chromatid Exchange Frequencies in Pathology Staff Occupational^ Exposed to
Formaldehyde." Mutation Research 141 (1984): 89-93.
Til, H. P., R. A. Woutersen, V. J. Feron, and J. J. Clary. "Evaluation of the Oral Toxicity of
Acetaldehyde and Formaldehyde in a 4-week Drinking Water Study in Rats." Food and
Chemical Toxicology 26 (1988): 447-452.
Tobe, M., K. Naito, and Y. Kurokawa. "Chronic Toxicity Study on Formaldehyde Administed
Orally to Rats." Toxicology 56 (1989): 79-86.
Tomatis, L., C. Agthe, H. Bartsch, J. Huff, R. Montesana, R. Saracci, E. Walder, and J.
Wilbourn. "Evaluation of the Carcinogenicity of Chemicals: A Review of the Monograph
Program of the IARC (1971-1977)." Cancer Research 38 (1978): 877-885.
Formaldehyde: Analytical Chemistry and Toxicology, (ed.) Victor Turoski. Washington, D.C.:
American Chemical Society, 1985.
Upreti, R. K., M. Y. H. Farooqui, A. E. Ahmed, and G. A. S. Ansari. Toxicokinetics and
Molecular Interaction of [14-CJ Formaldehyde in Rats." Archives of Environmental
Contamination and Toxicology 16 (1987): 263-73.
U.S. House of Representatives Committee on Science and Technology, 98th Congress, 1st
Session 1983. Review of the Scientific Basis of the Environmental Protectional Agency's
Carcinogenic Risk Assessment on Formaldehyde: Together with Additional and Minority
Views. Washington, D.C.: U.S. Government Printing Office, 1983.
VanVoorhees, Robert F. "EPA Assessment of Formaldehyde under the Toxic Substances
Control Act." in Risk Analysis in the Chemical Industry: Proceedings of a Symposium
260
-------
23-24 September. 1985. 123- 141. Chemical Manufacturers Association. Rockville, MD:
Government Institutes, 1985.
Vaughan, T. L., C. Strader, S. Davis, and J. R. Dating. "Formaldehyde and Cancers of the
Pharynx, Sinus, and Nasal Cavity: I. Occupational Exposures." International Journal of
Cancer 38 (1986): 677-83.
Vaughan, T. L., C. Strader, S. Davis, and J. R. Dating.. "Formaldehyde and Cancers of the
Pharynx, Sinus, and Nasal Cavity: II. Residential Exposures." International Journal of
Cancer 38 (1986): 685-88.
Walker jr., B., P. Fox, V. Li, and G. Parker. "The Massachusetts Program for Reducing the Risk
of Formaldehyde Exposure." Public Health Reporter 102, no. 3 (1987): 290-94.
Walrath, J., and J. F. Fraumeni. "Cancer and Other Causes of Death Among Embalmers."
Cancer Research 44 (1984): 4638-4641.
Ward jr., J. B., J. A. Hokanson, E. R. Smith, L. W. Chang, M. A. Pereira, E. B. Whorton jr., an^l
M. S. Legator. "Sperm Count, Morphology, and Fluorescent Body Frequency in Autopsy
Service Workers Exposed to Formaldehyde." Mutation Research 130 (1984): 417-424.
Watanabe, F., T. Matsunaga, T. Soejima, and Y. Iwata. "Study on Aldehyde Carcinogenicity:
Communication I. Experimentally Induced Rat Sarcomas by Repeated Injection of
Formalin. Seven Cases of Trnasplantable Sarcomas of Rats Developed in the Area of
repeated Subcutaneous Injections of Urotropin (hexamethylenetetramine)." Gann. The
Japanese Journal of Cancer Research 45 (1954): 451-452.
Weisburger, J. H., and G. M. Williams. "Chemical Carcinogenesis." in Cancer Medicine, (eds.) J.
F. Holland, and E. Frei. Philadelphia, PA: Lea and Febiger, 1980.
WHO. Recommended Health-Based Occupational Exposure Limits for Respiratory Irritants:
Report of a WHO Study Group. Geneva, SZ: World Health Organization, 1984.
Wilhelmsson, B., and M. Holmstrom. "Positive Formaldehyde-RAST After Prolonged
Formaldehyde Exposure by Inhalation." Lancet 8551 (1987): 164.
Wilmer, J. W., L. M. Appelman, and R. A. Woutersen. Subchronic (13-week) Inhalation Toxicity
Study of Formaldehyde in Male Rats: 8-Hour Continuous versus 8-Hour Intermittent
Exposure. Zeist, The Netherlands: CIVO Institutes Inc., 1986.
Wimer, J. W. G. M., R. A. /Appelman, L. M. Woutersen, W. R. Leeman, and V. J. Feron.
"Subacute (4-week) Inhalation Toxicity Study of Formaldehyde in Male Rats: 8-Hour
Intermittent versus 8- Hour Continuous Exposures." Journal of Applied Toxicology 7
(1987): 15-16.
Witek, T. J., E. N. Schachrer, T. Tosun, B. P. Leaderer, and G. J. Beck. "Controlled Human
Studies on the Pulmonary Effects of Indoor Air Pollution: Experiences with Sulfur
Dioxide and Formaldehyde." Environment International 12 (1986): 129-135.
261
-------
Witek, T. J., E. N. Schachrer, T. Tosun, G. J. Beck, and B. P. Leaderer. "An Evaluation of
Respiratory Effects Following Exposure to 2.0 ppiri Formaldehyde in Asthmatics: Lung
Function, Symptoms, and Airway Reactivity." Archives of Environmental Health 42
(1987): 230-37.
World Health Organization. Environmental Health Criteria 89: Formaldehyde. Geneva, SZ:
World Health Organization, 1989.
World Health Organization, International Programme on Chemical Safety. Formaldehyde
Health and Safety Guide: A Companion Volume to Environmental Health Criteria 89,
Formaldehyde. Series no. 57. Stuttgart, FDR: World Health Organization, 1991.
Woutersen, R. A., L. M. Appelman, A. Van Garderen-Hoetmer, and V. J. Feron. "Inhalation
Toxicity of Acetaldehyde in Rats: III. Carcinogenicity Study." Toxicology 41 (1986):
213-231.
Woutersen, R. A., L. M. Appelmann, J. M. G. M. Wilmer, H. E. Flake, and V. J. Feron.
"Subchronic (13-week) Inhalation Toxicity Study of Formaldehyde in Rats." Journal of
Applied Toxicology 7 (1987): 43-49.
Yager, J. W., K. L. Cohn, R. C. Spear, J. M. Fisher, and L. Morse. "Sister-Chromatid Exchanges
in Lymphocytes of Anatomy Students Exposed to Formaldehyde-Embalming Solution."
Mutation Research 174 (1986): 133-139.
Yasamura, R., I. Naruse, and R. Shoji. Teratogenic and Embryocidal Effects of Formaldehyde
in Mice." Teratology 28, no. 1 (1983): 37.
Zwart, A., R. A. Woutersen, J. W. G. M. Wilmer, and B. J. Spit. Subchronic (13-week)
Inhalation Toxicity Study with Formaldehyde in Rats. Zeist, The Netherlands: CIVO
Institutes, 1987.
262
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO. 2.
EPA/600/R-94/204b
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
Hazard Identification in Carcinogen Risk Analysis:
An Integrative Approach
Part III An Application of the Methodoloqy:
FnrmaInphvde in Air
5. REPORT DATE
April 1994
6. PERFORMING ORGANIZATION CODE
j.auwoW) * Ml Mir
Douglas J. Crawford-Brown, Jeffrey R. Arnold,
Kenneth G. Brown
8. PERFORMING ORGANIZATION REPORT NO.
OHEA-C-554; NCEA-W-013
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Kenneth G. Brown, Ph.D., Inc.
P.O. Box 16608
Chapel Hill, NC 27516-6608
10. PROGRAM ELEMENT NO. j
11. CONTRACT/GRANT NO.
68-C9-0009; W.A. S-l-56 1
12. SPONSORING AGENCY NAME AND ADDRESS
National Center for Environmental Assessment*
Washington Laboratory
U.S. EPA
Washington, DC 20460
13. TYPE OF REPORT AND PERIOD COVERED
14. SPONSORING AGENCY CODE
i
EPA/600/021
15. SUPPLEMENTARY NOTES
This project report is part of the U.S. EPA's program, Research and Improve
Health Risk Assessment (RIHRA). ^Formerly Ofc. of Health & Environmental Assessmt.
16. ABSTRACT
This report applies and tests the methodology developed in an earlier report (Crawford-Brown and
Brown, 1992) to formaldehyde as an example. The methodology requires the performance of a series of
analytical tasks using the extensive literature on formaldehyde. The tasks are: 1) define a series of contexts, or |
sets of empirical test conditions, into which the available data are placed; 2) determine the quality of the data for '
each context (i.e., appropriate test system, adequacy of experimental conditions, strength of measured effect); 3)
evaluate the degree to which the data in each context support statements (or claims) of carcinogenicity. This
involves the ennumeration of type of evidence (e.g. direct observations, theory-based inferences, empirical
correlations, existential insights), the degree to which that type of evidence is compelling and overall judgments
of degree of support across all types of evidence; 4) examine the degree of support for background premises
necessary for extrapolating results from one context to the context of interest (e.g. high dose rat to low dose
human contexts); 5) using the results of previous tasks, extrapolate to the context of interest and form a
composite judgment of carcinogenicity. These tasks were performed for the formaldehyde data using a series of
working tables. The methodology was then characterized in terms of completeness (ability to incorporate all
relevant information), utility (ability to aid the analyst in clarifying personal positions and displaying the
judgments made) and practicality (whether it demands more information than is actually available, makes
unrealistic demands on time, and yields relatively unambiguous results). Suggestions for improvement are made
based on the experience of analyzing the formaldehyde information.
17. KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
c. COSATl Field/Group
18. DISTRIBUTION STATEMENT
Release to the Public
19. SECURITY CLASS (This Report)
Unclassified
21. NO. OF PAGES
S3 I
20. SECURITY CLASS (This page)
Unclassified
22. PRICE
EPA Form 2220-1 (Rev. 4-77) previous edition is obsolete
------- |