DRAFT - DO NOT CITE OR QUOTE EPA/635/R-10/003D
ww.epa.gov/iris
vvEPA
TOXICOLOGICAL REVIEW
OF
DICHLOROMETHANE
(METHYLENE CHLORIDE)
(CAS No. 75-09-2)
In Support of Summary Information on the
Integrated Risk Information System (IRIS)
June 2011
NOTICE
This document is an Final Agency/Interagency Science Discussion Review draft. This
information is distributed solely for the purpose of pre-dissemination peer review under
applicable information quality guidelines. It has not been formally disseminated by EPA. It does
not represent and should not be construed to represent any Agency determination or policy. It is
being circulated for review of its technical accuracy and science policy implications.
U.S. Environmental Protection Agency
Washington, DC
-------
DISCLAIMER
This document is a preliminary draft for review purposes only. This information is
distributed solely for the purpose of pre-dissemination peer review under applicable information
quality guidelines. It has not been formally disseminated by EPA. It does not represent and
should not be construed to represent any Agency determination or policy. Mention of trade
names or commercial products does not constitute endorsement or recommendation for use.
DRAFT - DO NOT CITE OR QUOTE
-------
CONTENTS—TOXICOLOGICAL REVIEW OF DICHLOROMETHANE
(CAS No. 75-09-2)
LIST OF TABLES ix
LIST OF FIGURES xvii
LIST OF ABBREVIATIONS AND ACRONYMS xxi
FOREWORD xxiii
AUTHORS, CONTRIBUTORS, AND REVIEWERS xxiv
1. INTRODUCTION 1
2. CHEMICAL AND PHYSICAL INFORMATION 3
3. TOXICOKINETICS 5
3.1. ABSORPTION 5
3.1.1. Oral—Gastrointestinal Tract Absorption 5
3.1.2. Inhalation—Respiratory Tract Absorption 5
3.2. DISTRIBUTION 7
3.3. METABOLISM 9
3.3.1. The CYP2E1 Pathway 11
3.3.2. The GST Pathway 14
3.4. ELIMINATION 20
3.5. PHYSIOLOGICALLY BASED PHARMACOKINETIC MODELS 21
3.5.1. Probabilistic Mouse PBPK Dichloromethane Model (Marino et al., 2006) 27
3.5.2. Probabilistic Human PBPK Dichloromethane Model (David etal., 2006) 31
3.5.3. Evaluation of Rat PBPK Dichloromethane Models 40
3.5.4. Comparison of Mouse, Rat, and Human PBPK Models 42
3.5.5. Uncertainties in PBPK Model Structure for the Mouse, Rat and Human 46
4. HAZARD IDENTIFICATION 51
4.1. STUDIES IN HUMANS 51
4.1.1. Introduction—Case Reports, Epidemiologic, and Clinical Studies 51
4.1.2. Noncancer Studies 51
4.1.2.1. Case Reports of Acute, High-dose Exposures 51
4.1.2.2. Controlled Experiments Examining Acute Effects 52
4.1.2.3. Observational Studies Focusing on Clinical Chemistries, Clinical
Examinations, and Symptoms 53
4.1.2.4. Observational Studies Using Workplace Medical Program Data 60
4.1.2.5. Studies of Ischemic Heart Disease Mortality Risk 63
4.1.2.6. Studies of Suicide Risk 64
4.1.2.7. Studies of Infectious Disease Risk 65
4.1.2.8. Studies of Reproductive Outcomes 65
4.1.2.9. Summary of Noncancer Studies 68
4.1.3. Cancer Studies 71
4.1.3.1. Identification and Selection of Studies for Evaluation of Cancer Risk 71
4.1.3.2. Description of the Selected Studies 71
4.1.3.3. Cellulose Triacetate Film Base Production Cohorts 71
4.1.3.4. Cellulose Triacetate Fiber Production Cohorts 79
iii DRAFT - DO NOT CITE OR QUOTE
-------
4.1.3.5. Solvent-Exposed Workers—Hill Air Force Base, Utah 84
4.1.3.6. Case-Control Studies of Specific Cancers and Dichloromethane 85
4.1.3.7. Summary of Cancer Studies by Type of Cancer 97
4.2. SUBCHRONIC AND CHRONIC STUDIES AND CANCER BIO ASSAYS IN
ANIMALS—ORAL AND INHALATION 107
4.2.1. Oral Exposure: Overview of Noncancer and Cancer Effects 107
4.2.1.1. Toxicity Studies ofSubchronic Oral Exposures: Hepatic Effects 107
4.2.1.2. Toxicity Studies of Chronic Oral Exposures: Hepatic Effects and
Carcinogenicity 112
4.2.2. Inhalation Exposure: Overview of Noncancer and Cancer Effects 119
4.2.2.1. Toxicity Studies ofSubchronic Inhalation Exposures: General, Renal,
and Hepatic Effects 120
4.2.2.2. Toxicity Studies from Chronic Inhalation Exposures 124
4.3. REPRODUCTIVE/DEVELOPMENTAL STUDIES—ORAL AND
INHALATION 143
4.3.1. Reproductive Toxicity Studies 145
4.3.1.1. Gavage and Subcutaneous Injection Studies 145
4.3.1.2. Inhalation Studies 146
4.3.2. Developmental Toxicity Studies 147
4.3.2.1. Gavage Studies and Culture Studies 147
4.3.2.2. Inhalation Studies 148
4.4. OTHER DURATION-OR ENDPOINT-SPECIFIC STUDIES 150
4.4.1. Short-term (2-Week) Studies of General and Hepatic Effects in Animals 150
4.4.2. Immunotoxicity Studies in Animals 150
4.4.3. Neurotoxicology Studies in Animals 153
4.4.3.1. Neurotoxicology Studies—Oral Exposures 159
4.4.3.2. Neurotoxicology Studies—InhalationalExposure 160
4.5. MECHANISTIC DATA AND OTHER STUDIES IN SUPPORT OF THE
MODE OF ACTION 167
4.5.1. Genotoxicity Studies 167
4.5.1.1. In Vitro Genotoxicity Assays 168
4.5.1.2. In Vivo Genotoxicity Assays 179
4.5.2. Mechanistic Studies of Liver Effects 189
4.5.3. Mechanistic Studies of Lung Effects 193
4.5.4. Mechanistic Studies of Neurological Effects 198
4.6. SYNTHESIS OF MAJOR NONCANCER EFFECTS 200
4.6.1. Oral 200
4.6.1.1. Summary of Human Data 200
4.6.1.2. Summary of Animal Data 200
4.6.2. Inhalation 204
4.6.2.1. Summary of Human Data 204
4.6.2.2. Summary of Animal Studies 205
4.6.3. Mode of Action Information 213
4.6.3.1. Mode of Action for Nonneoplastic Liver Effects 213
4.6.3.2. Mode of Action for Nonneoplastic Lung Effects 214
4.6.3.3. Mode of Action for Neurological Effects 214
4.6.3.4. Mode of Action for Neurodevelopmental Effects 216
4.6.3.5. Mode of Action for Immunotoxicity 216
iv DRAFT - DO NOT CITE OR QUOTE
-------
4.7. EVALUATION OF CARCINOGENICITY 216
4.7.1. Summary of Overall Weight of Evidence 216
4.7.2. Synthesis of Human, Animal, and Other Supporting Evidence 218
4.7.3. Mode of Action Information 230
4.7.3.1. Hypothesized Mode of Action 230
4.7.3.2. General Conclusions About the Mode of Action for Tumors in Rodents
and Relevance to Humans 238
4.8. SUSCEPTIBLE POPULATIONS AND LIFE STAGES 242
4.8.1. Possible Childhood Susceptibility 242
4.8.2. Possible Gender Differences 243
4.8.3. Other 244
5. DOSE-RESPONSE ASSESSMENTS 246
5.1. ORAL REFERENCE DOSE (RfD) 246
5.1.1. Choice of Principal Study and Critical Effect—with Rationale and
Justification 246
5.1.2. Derivation Process for Noncancer Reference Values 249
5.1.3. Evaluation of Dose Metrics for Use in Noncancer Reference Value
Derivations 253
5.1.4. Methods of Analysis—Including Models (PBPK, HMD, etc.) 255
5.1.5. RfD Derivation—Including Application of Uncertainty Factors (UFs) 260
5.1.6. Previous RfD Assessment 261
5.1.7. RfD Comparison Information 261
5.2. INHALATION REFERENCE CONCENTRATION (RfC) 264
5.2.1. Choice of Principal Study and Critical Effect—with Rationale and
Justification 264
5.2.2. Derivation Process for RfC Values 269
5.2.3. Methods of Analysis—Including Models (PBPK, BMD, etc.) 269
5.2.4. RfC Derivation—Including Application of Uncertainty Factors (UFs) 273
5.2.5. Previous RfC Assessment 275
5.2.6. RfC Comparison Information 275
5.3. UNCERTAINTIES IN THE ORAL REFERENCE DOSE AND INHALATION
REFERENCE CONCENTRATION 280
5.4. CANCER ASSESSMENT 290
5.4.1. Cancer OSF 290
5.4.1.1. Choice of Study/Data—with Rationale and Justification 290
5.4.1.2. Derivation of OSF 291
5.4.1.3. Dose-Response Data 293
5.4.1.4. Dose Conversion and Extrapolation Methods: Cancer OSF 294
5.4.1.5. Oral Cancer Slope Factor 300
5.4.1.6. AIternative Derivation Based on Route-to-Route Extrapolation 300
5.4.1.7. Alternative Based On Administered Dose 302
5.4.1.8. Previous IRIS Assessment: Cancer OSF 303
5.4.1.9. Comparison of Cancer OSFs Using Different Methodologies 303
5.4.2. Cancer IUR 305
5.4.2.1. Choice of Study/Data—with Rationale and Justification 305
5.4.2.2. Derivation of the Cancer IUR 306
5.4.2.3. Dose-Response Data 306
5.4.2.4. Dose Conversion and Extrapolation Methods: Cancer IUR 307
v DRAFT - DO NOT CITE OR QUOTE
-------
5.4.2.5. CancerIUR 315
5.4.2.6. Comparative Derivation Based on Rat Mammary Tumor Data 318
5.4.2.7. Alternative Based on Administered Concentration 318
5.4.2.8. Previous IRIS Assessment: Cancer IUR 319
5.4.2.9. Comparison of Cancer IUR Using Different Methodologies 320
5.4.3. Differences Between Current Assessment and Previous IRIS PBPK-based
Assessment 322
5.4.4. Application of Age-Dependent Adjustment Factors (ADAFs) 324
5.4.4.1. Application of ADAFs in Oral Exposure Scenarios 324
5.4.4.2. Application of ADAFs in Inhalation Exposure Scenarios 325
5.4.5. Uncertainties in Cancer Risk Values 326
6. MAJOR CONCLUSIONS IN THE CHARACTERIZATION OF HAZARD AND
DOSE RESPONSE 342
6.1. HUMAN HAZARD POTENTIAL 342
6.2. DOSE RESPONSE 345
6.2.1. OralRfD 345
6.2.2. Inhalation RfC 346
6.2.3. Uncertainties in RfD and RfC Values 348
6.2.4. Oral Cancer Slope Factor 351
6.2.5. Cancer IUR 355
6.2.6. Uncertainties in Cancer Risk Values 358
7. REFERENCES 361
APPENDIX A: SUMMARY OF EXTERNAL PEER REVIEW AND PUBLIC COMMENTS
AND DISPOSITION A-l
APPENDIXB: HUMANPBPKDICHLOROMETHANEMODEL B-l
B.I. HUMAN MODEL DESCRIPTION 1
B.2. REVISIONS TO PARAMETER DISTRIBUTIONS OF DAVID ET AL. (2006) 2
B.3. CY2E1 AND GST-T1 5
B.4. ANALYSIS OF HUMAN PHYSIOLOGICAL DISTRIBUTIONS FOR PBPK
MODELING 10
B.4.1. Age 10
B.4.2. Gender 11
B.4.3. BW 12
B.4.4. Alveolar Ventilation 14
B.4.5. QCC 15
B.4.6. Fat Fraction 16
B.4.7. Liver Fraction 17
B.4.8. Tissue Volume Normalization 18
B.5. SUMMARY OF REVISED HUMAN PBPK MODEL 18
APPENDIX C. RAT DICHLOROMETHANE PBPK MODELS C-l
C.I. METHODS OF ANALYSIS 1
C.I.I. Select! on of Evaluation Data Sets and PBPK Models 1
C.I.2. Analysis 4
vi DRAFT - DO NOT CITE OR QUOTE
-------
C.2. RESULTS 6
C.2.1. Evaluation of Model Structure for Description of Carboxyhemoglobin
Levels 8
C.2.2. Evaluation of Prediction of Uptake, Blood and Liver Concentrations, and
Expiration of Dichloromethane 10
C.2.3. Evaluation of Relative Flux of CYP and GST Metabolism of
Dichloromethane 20
C.2.4. Evaluation of Model Predictions of Oral Absorption of Dichloromethane 22
C.3. MODEL OPTION SUMMARY 28
APPENDIX D. SUMMARY OF BENCHMARK DOSE (BMD) MODELING OF
NONCANCER ENDPOINTS D-l
D. 1. ORAL RfD: BMD MODELING OF LIVER LESION INCIDENCE DATA FOR
RATS EXPOSED TO DICHLOROMETHANE IN DRINKING WATER FOR 2
YEARS (SEROTA ET AL., 1986a) 1
D.2. INHALATION RfC: BMD MODELING OF LIVER LESION INCIDENCE
DATA FOR RATS EXPOSED TO DICHLOROMETHANE VIA INHALATION
FOR 2 YEARS (NITSCHKE ET AL., 1988a) 5
APPENDIX E: SUMMARY OF BENCHMARK DOSE (BMD) MODELING OF CANCER
ENDPOINTS E-l
E. 1. ORAL CANCER SLOPE FACTORS: BMD MODELING OF LIVER TUMOR
INCIDENCE DATA FOR MICE EXPOSED TO DICHLOROMETHANE IN
DRINKING WATER FOR 2 YEARS (SEROTA ET AL., 1986b; HAZLETON
LABORATORIES, 1983) 1
E. 1.1. Modeling Results for the Internal Liver Metabolism Metric 3
E. 1.2. Modeling Results for the Whole Body Metabolism Metric 6
E.2. CANCER IUR: BMD MODELING OF LIVER AND LUNG TUMOR
INCIDENCE DATA FOR MALE MICE EXPOSED TO
DICHLOROMETHANE VIA INHALATION FOR 2 YEARS (MENNEAR ET
AL., 1988; NTP, 1986) 9
E.2.1. Modeling Results for the Internal Liver Metabolism Metric, Liver Tumors.
Mennear et al. (1988); NTP (1986): Internal Liver Dose-Response for Liver
Tumors in Male Mice 11
E.2.2. Modeling Results for the Internal Lung Metabolism Metric, Lung Tumors.
Mennear et al. (1988); NTP (1986): Internal Lung Dose-Response for Lung
Tumors in Male Mice 14
E.2.3. Modeling Results for the Whole Body Metabolism Metric, Liver Tumors.
Mennear et al. (1988); NTP (1986): Internal Whole-Body Metabolism
Dose-Response for Liver Tumors in Male Mice 16
E.2.4. Modeling Results for the Whole Body Metabolism Metric, Lung Tumors.
Mennear et al. (1988); NTP (1986): Internal Whole-Body Metabolism
Dose-Response for Lung Tumors in Male Mice 19
vii DRAFT - DO NOT CITE OR QUOTE
-------
APPENDIX F. COMPARATIVE CANCER IUR BASED ON FEMALE MICE DATA F-1
APPENDIX G. COMPARATIVE CANCER IUR BASED ON BENIGN MAMMARY GLAND
TUMORS IN RATS G-l
APPENDIX H: SOURCE CODE AND COMMAND FILES FOR DICHLOROMETHANE
PBPK MODELS H-l
viii DRAFT - DO NOT CITE OR QUOTE
-------
LIST OF TABLES
Table 2-1. Physical properties and chemical identity of dichloromethane...
Table 3-1. Distribution of radioactivity in tissues 48 hours after inhalation exposure of mature
male Sprague-Dawley rats (n = 3) for 6 hours 7
Table 3-2. Brain and perirenal fat dichlorom ethane and blood CO concentrations in male Wistar
rats exposed by inhalation to dichloromethane at constant exposure concentrations
compared with intermittently high exposure concentrations 9
Table 3-3. Mean prevalences of the GST-T1 null (-/-) genotype in human ethnic groups 16
Table 3-4. GST-T1 enzyme activities toward dichloromethane in human, rat, mouse, and
hamster tissues (liver, kidney, and erythrocytes) 18
Table 3-5. Values for parameter distributions in a B6C3Fi mouse probabilistic PBPK model for
dichloromethane compared with associated values for point parameters in earlier
deterministic B6C3Fi mouse PBPK models for dichloromethane 29
Table 3-6. Internal daily doses for B6C3Fi mice exposed to dichloromethane for 2 years
(6 hours/day, 5 days/week) calculated with different PBPK models 31
Table 3-7. Results of calibrating metabolic parameters in a human probabilistic PBPK model for
dichloromethane with individual kinetic data for 42 exposed volunteers and MCMC
analysis 33
Table 3-8. Parameter distributions used in human Monte Carlo analysis for dichloromethane by
David et al. (2006) 35
Table 3-9. Parameter distributions for the human PBPK model for dichloromethane used by
EPA 38
Table 3-10. Parameter values for the rat PBPK model for dichloromethane used by EPA 42
Table 3-11. Parameters in the mouse, rat, and human PBPK model for dichloromethane used by
EPA 44
Table 4-1. Percentage of male General Electric plastic polymer workers reporting neurologic
symptoms or displaying abnormal values in measures of neurological function,
hepatic function, and cardiac function 62
Table 4-2. Ischemic heart disease mortality risk in four cohorts of dichloromethane-exposed
workers 64
Table 4-3. Suicide risk in two cohorts of dichloromethane-exposed workers 65
ix DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-4. Mortality risk in Eastman Kodak cellulose triacetate film base production workers,
Rochester, New York 74
Table 4-5. Mortality risk by cumulative exposure in Eastman Kodak cellulose triacetate film
base production workers, Rochester, New York 76
Table 4-6. Mortality risk in Imperial Chemical Industries cellulose triacetate film base
production workers, Brantham, United Kingdom: 1,473 men employed 1946-1988,
followed through 1994 79
Table 4-7. Mortality risk in Hoechst Celanese Corporation cellulose triacetate fiber production
workers, Rock Hill, South Carolina: 1,271 men and women employed 1954-1977,
followed through 1990 81
Table 4-8. Cancer mortality risk in Hoechst Celanese Corporation cellulose triacetate fiber
production workers, Cumberland, Maryland: 2,909 men and women employed
1970-1981, followed through 1989 83
Table 4-9. Summary of cohort studies of cancer risk and dichloromethane exposure 98
Table 4-10. Summary of case-control studies of cancer risk and dichloromethane exposure... 100
Table 4-11. Incidences of histopathologic changes in livers of male and female F344 rats
exposed to dichloromethane in drinking water for 90 days 109
Table 4-12. Incidences of histopathologic changes in livers of male and female B6C3Fi mice
exposed to dichloromethane in drinking water for 90 days Ill
Table 4-13. Studies of chronic oral dichloromethane exposures (up to 2 years) 112
Table 4-14. Incidences of nonneoplastic liver changes and liver tumors in male and female F344
rats exposed to dichloromethane in drinking water for 2 years 114
Table 4-15. Incidences for focal hyperplasia and tumors in the liver of male B6C3Fi mice
exposed to dichloromethane in drinking water for 2 years 117
Table 4-16. Studies of chronic inhalation dichloromethane exposures 125
Table 4-17. Incidences of nonneoplastic histologic changes in male and female F344/N rats
exposed to dichloromethane by inhalation (6 hours/day, 5 days/week) for 2 years 127
Table 4-18. Incidences of selected neoplastic lesions in male and female F344/N rats exposed to
dichloromethane by inhalation (6 hours/day, 5 days/week) for 2 years 129
Table 4-19. Incidences of nonneoplastic histologic changes in B6C3Fi mice exposed to
dichloromethane by inhalation (6 hours/day, 5 days/week) for 2 years 131
Table 4-20. Incidences of neoplastic lesions in male and female B6C3Fi mice exposed to
dichloromethane by inhalation (6 hours/day, 5 days/week) for 2 years 133
x DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-21. Incidences of selected nonneoplastic and neoplastic histologic changes in male and
female Sprague-Dawley rats exposed to dichloromethane by inhalation (6 hours/day,
5 days/week) for 2 years 137
Table 4-22. Incidences of selected nonneoplastic histologic changes in male and female
Sprague-Dawley rats exposed to dichloromethane by inhalation (6 hours/day,
5 days/week) for 2 years 140
Table 4-23. Incidences of selected neoplastic histologic changes in male and female Sprague-
Dawley rats exposed to dichloromethane by inhalation (6 hours/day, 5 days/week)
for 2 years 141
Table 4-24. Summary of studies of reproductive and developmental effects of dichloromethane
exposure in animals 144
Table 4-25. Reproductive outcomes in F344 rats exposed to dichloromethane by inhalation for
14 weeks prior to mating and from GDs 0-21 146
Table 4-26. Studies of neurobehavioral changes from dichloromethane, by route of exposure and
type of effect 154
Table 4-27. Studies of neurophysiological changes as measured by evoked potentials resulting
from dichloromethane, by route of exposure 156
Table 4-28. Studies of neurochemical changes from dichloromethane, by route of exposure.. 158
Table 4-29. Results from in vitro genotoxicity assays of dichloromethane with bacteria, yeast, or
fungi 169
Table 4-30. Results from in vitro genotoxicity assays of dichloromethane with mammalian
systems, by type of test 173
Table 4-31. Results from in vivo genotoxicity assays of dichloromethane in insects 179
Table 4-32. Results from in vivo genotoxicity assays of dichloromethane in mice 181
Table 4-33. Results from in vivo genotoxicity assays of dichloromethane in rats and hamsters
185
Table 4-34. Comparison of in vivo dichloromethane genotoxicity assays targeted to lung or liver
cells, by species 187
Table 4-35. NOAELs and LOAELs in selected animal studies involving oral exposure to
dichloromethane for short-term, subchronic, or chronic durations 202
Table 4-36. NOAELs and LOAELs in animal studies involving inhalation exposure to
dichloromethane for subchronic or chronic durations, hepatic, pulmonary, and
neurologic effects 207
xi DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-37. NOAELs and LOAELs in selected animal studies involving inhalation exposure to
dichloromethane, reproductive and developmental effects 211
Table 4-38. Incidence of liver tumors in male B6C3Fi mice exposed to dichloromethane in a 2-
year oral exposure (drinking water) studya 221
Table 4-39. Incidences of liver tumors in male and female F344 rats exposed to dichloromethane
in drinking water for 2 years 222
Table 4-40. Incidences of selected neoplastic lesions in B6C3Fi mice exposed to
dichloromethane by inhalation (6 hours/day, 5 days/week) for 2 years 224
Table 4-41. Incidences of selected neoplastic lesions in F344/N rats exposed to dichloromethane
by inhalation (6 hours/day, 5 days/week) for 2 years 225
Table 4-42. Incidences of mammary gland tumors in two studies of male and female Sprague-
Dawley rats exposed to dichloromethane by inhalation (6 hours/day, 5 days/week)
for 2 years 227
Table 4-43. Comparison of internal dose metrics in inhalation and oral exposure scenarios in
male mice and rats 229
Table 4-44. Experimental support for mutagenic mode of action for dichloromethane 237
Table 5-1. Incidence data for liver lesions and internal liver doses based on various metrics in
male and female F344 rats exposed to dichloromethane in drinking water for 2 years
256
Table 5-2. BMD modeling results for incidence of liver lesions in male and female F344 rats
exposed to dichloromethane in drinking water for 2 years, based on liver-specific
CYP metabolism dose metric (mg dichloromethane metabolism via CYP pathway
per liter liver tissue per day) 258
Table 5-3. RfD for dichloromethane based on PBPK model-derived probability distributions of
human drinking water exposures extrapolated from liver lesion incidence data for
male rats exposed via drinking water for 2 years, based on liver-specific CYP
metabolism dose metric (mg dichloromethane metabolized via CYP pathway per
liter liver tissue per day) 259
Table 5-4. Potential points of departure with applied UFs and resulting candidate RfDs 262
Table 5-5. Incidence data for liver lesions (hepatic vacuolation) and internal liver doses based on
various metrics in female Sprague-Dawley rats exposed to dichloromethane via
inhalation for 2 years 270
Table 5-6. BMD modeling results for incidence of noncancer liver lesions in female Sprague-
Dawley rats exposed to dichloromethane by inhalation for 2 years, based on liver
xii DRAFT - DO NOT CITE OR QUOTE
-------
specific CYP metabolism metric (mg dichloromethane metabolized via CYP
pathway per liter liver tissue per day) 272
Table 5-7. Inhalation RfC for dichloromethane based on PBPK model-derived probability
distributions of human inhalation exposure extrapolated from liver lesion data for
female rats exposed via inhalation for 2 years, based on liver-specific CYP
metabolism dose metric (mg dichloromethane metabolized via CYP pathway per
liter liver tissue per day) 273
Table 5-8. Potential points of departure with applied UFs and resulting candidate RfCs 278
Table 5-9. Statistical characteristics of human equivalent doses in specific populations of the
GST-T1"A group 287
Table 5-10. Statistical characteristics of HECs in specific populations of the GST-T1"A group289
Table 5-11. Incidence data for liver tumors and internal liver doses, based on GST metabolism
dose metrics in male B6C3Fi mice exposed to dichloromethane in drinking water for
2 years 294
Table 5-12. BMD modeling results and tumor risk factors for internal dose metric associated
with 10% extra risk for liver tumors in male B6C3Fi mice exposed to
dichloromethane in drinking water for 2 years, based on liver-specific GST
metabolism and whole body GST metabolism dose metrics 296
Table 5-13. Cancer OSFs for dichloromethane based on PBPK model-derived internal liver
doses in B6C3Fi mice exposed via drinking water for 2 years, based on liver-specific
GST metabolism and whole body metabolism dose metrics, by population genotype
299
Table 5-14. Alternative route-to-route cancer OSFs for dichloromethane extrapolated from male
B6C3Fi mouse inhalation liver tumor incidence data using a tissue-specific GST
metabolism dose metric, by population genotype 301
Table 5-15. Cancer OSF based on a human BMDLio using administered dose for liver tumors in
male B6C3Fi mice exposed to dichloromethane in drinking water for 2 years 302
Table 5-16. Comparison of OSFs derived using various assumptions and metrics, based on
tumors in male mice 304
Table 5-17. Incidence data for liver and lung tumors and internal doses based on GST
metabolism dose metrics in male and female B6C3Fi mice exposed to
dichloromethane via inhalation for 2 years 307
Table 5-18. BMD modeling results and tumor risk factors associated with 10% extra risk for
liver and lung tumors in male and female B6C3Fi mice exposed by inhalation to
dichloromethane for 2 years, based on liver-specific GST metabolism and whole
body GST metabolism dose metrics 311
xiii DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-19. lURs for dichloromethane based on PBPK model-derived internal liver and lung
doses in B6C3Fi male mice exposed via inhalation for 2 years, based on liver-
specific GST metabolism and whole body metabolism dose metrics, by population
genotype 314
Table 5-20. Upper bound estimates of combined human lURs for liver and lung tumors resulting
from lifetime exposure to 1 ug/m3 dichloromethane based on liver-specific GST
metabolism and whole body metabolism dose metrics, by population genotype.... 317
Table 5-21. Inhalation units risks based on human BMDLio values using administered
concentration for liver and lung tumors in B6C3Fi mice exposed by inhalation to
dichloromethane for 2 years 319
Table 5-22. Comparison of lURs derived by using various assumptions and metrics 321
Table 5-23. Comparison of key B6C3Fi mouse parameters differing between prior and current
PBPK model application 322
Table 5-24. Application of ADAFs to dichloromethane cancer risk following a lifetime (70-
year) oral exposure 325
Table 5-25. Application of ADAFs to dichloromethane cancer risk following a lifetime (70-
year) inhalation exposure 326
Table 5-26. Summary of uncertainty in the derivation of cancer risk values for dichloromethane
327
Table 5-27. Statistical characteristics of human internal doses for 1 mg/kg-day oral exposures in
specific populations 339
Table 5-28. Statistical characteristics of human internal doses for 1 mg/m3 inhalation exposures
in specific subpopulations 340
Table 6-1. Comparison of OSFs derived by using various assumptions and metrics, based on
liver tumors in male mice 354
Table 6-2. Comparison of lURs derived by using various assumptions and metrics 357
Table B-l. Parameter distributions used in human Monte Carlo analysis for dichloromethane by
David et al. (2006) B-3
Table B-2. Parameters forBW distributions as functions of age and gender B-13
Table B-3. Parameter distributions for the human PBPK model for dichloromethane used by
EPA B-20
Table C-l. Parameter values used in rat PBPK models C-7
xiv DRAFT - DO NOT CITE OR QUOTE
-------
Table C-3. Effect of PBPK model variation on predicted dichloromethane metabolite production
in the liver of (male) rats from inhalation exposures3 C-21
Table C-4. Observations and predictions of total expired dichloromethane resulting from gavage
doses in ratsa C-27
Table D-l. Incidence data for liver lesions and internal liver doses based on various metrics in
male and female F344 rats exposed to dichloromethane in drinking water for 2 years
(Serota et al., 1986a) D-l
Table D-2. BMD modeling results for incidence of liver lesions in male and female F344 rats
exposed to dichloromethane in drinking water for 2 years, based on liver-specific
CYP metabolism dose metric (mg dichloromethane metabolism via CYP pathway
per liter liver tissue per day) D-2
Table D-3. Incidence data for liver lesions (hepatic vacuolation) and internal liver doses based
on various metrics in female Sprague-Dawley rats exposed to dichloromethane via
inhalation for 2 years (Nitschke etal., 1988a) D-5
Table D-4. BMD modeling results for incidence of liver lesions in female Sprague-Dawley rats
exposed to dichloromethane by inhalation for 2 years, based on liver specific CYP
metabolism metric (mg dichloromethane metabolized via CYP pathway per liter
liver tissue per day) D-6
Table E-l. Incidence data for liver tumors and internal liver doses, based on GST metabolism
dose metrics, in male B6C3Fi mice exposed to dichloromethane in drinking water
for 2 years E-2
Table E-2. BMD modeling results and tumor risk factors for internal dose metric associated with
10% extra risk for liver tumors in male B6C3Fi mice exposed to dichloromethane in
drinking water for 2 years, based on liver-specific GST metabolism and whole body
GST metabolism dose metrics E-2
Table E-3. Incidence data for liver and lung tumors and internal doses based on GST
metabolism dose metrics in male B6C3Fi mice exposed to dichloromethane via
inhalation for 2 years E-9
Table E-4. BMD modeling results and tumor risk factors associated with 10% extra risk for liver
and lung tumors in male B6C3Fi mice exposed by inhalation to dichloromethane for
2 years, based on liver-specific GST metabolism and whole body GST metabolism
dose metrics E-10
Table F-l. Incidence data for liver and lung tumors and internal doses based on GST
metabolism dose metrics in female B6C3Fi mice exposed to dichloromethane via
inhalation for 2 years F-l
Table F-2. BMD modeling results and tumor risk factors associated with 10% extra risk for liver
and lung tumors in female B6C3Fi mice exposed by inhalation to dichloromethane
xv
DRAFT - DO NOT CITE OR QUOTE
-------
for 2 years, based on liver-specific GST metabolism and whole body GST
metabolism dose metrics F-3
Table F-3. lURs for dichloromethane based on PBPK model-derived internal liver and lung
doses in B6C3Fi female mice exposed via inhalation for 2 years, based on liver-
specific GST metabolism and whole body metabolism dose metrics, by population
genotype F-5
Table F-4. Upper bound estimates of combined human lURs for liver and lung tumors resulting
from lifetime exposure to 1 ug/m3 dichloromethane based on liver-specific GST
metabolism and whole body metabolism dose metrics, by population genotype, using
female mouse data for derivation of risk factors F-7
Table G-l. Incidence data for mammary gland tumors and internal doses based on different dose
metrics in male and female F344 rats exposed to dichloromethane via inhalation for
2 years G-l
Table G-2. BMD modeling results associated with 10% extra risk for mammary gland tumors in
F344 rats exposed by inhalation to dichloromethane for 2 years based on AUC for
dichloromethane in slowly perfused tissue G-3
Table G-3. lURs for dichloromethane based on benign mammary tumors and PBPK model-
derived internal doses in F344N rats exposed via inhalation for 2 years based on
AUC for dichloromethane in slowly perfused tissue dose metric G-4
xvi
DRAFT - DO NOT CITE OR QUOTE
-------
LIST OF FIGURES
Figure 3-1. Proposed pathways for dichloromethane metabolism 10
Figure 3-2. Schematics of PBPK models (1986-2006) used in the development of estimates for
dichloromethane internal dosimetry 23
Figure 3-3. Schematic of mouse PBPK model used by Marino et al. (2006) 28
Figure 3-4. Schematic of human PBPK used by David et al. (2006) 32
Figure 3-5. Schematic of rat PBPK model used in current assessment 41
Figure 3-6. Comparison of dichloromethane oxidation rate data with alternate kinetic models. 49
Figure 5-1. Exposure response array for oral exposure to dichloromethane 248
Figure 5-2. Process for deriving noncancer oral RfDs and inhalation RfCs using rodent and
human PBPK models 250
Figure 5-3. PBPK model-derived internal doses (mg dichloromethane metabolized via the CYP
pathway per liter liver per day) in rats and humans and their associated external
exposures (mg/kg-day), used for the derivation of RfDs 257
Figure 5-4. Comparison of candidate RfDs derived from selected points of departure for
endpoints presented in Table 5-4 263
Figure 5-5. Exposure response array for chronic (animal) or occupational (human) inhalation
exposure to dichloromethane (log Y axis) 265
Figure 5-6. Exposure response array for subacute to subchronic inhalation exposure to
dichloromethane (log Y axis) 267
Figure 5-7. PBPK model-derived internal doses (mg dichloromethane metabolized via the CYP
pathway per liter liver per day) in rats and humans versus external exposures (ppm).
271
Figure 5-8. Comparison of candidate RfCs derived from selected points of departure for
endpoints presented in Table 5-8 279
Figure 5-9. Sensitivity coefficients for long-term mass CYP- and GST-mediated metabolites per
liver volume from a daily drinking water concentration of 10 mg/L in rats 284
Figure 5-10. Sensitivity coefficients for long-term mass CYP- and GST-mediated metabolites
per liver volume from a long-term average daily inhalation concentration of 500 ppm
in rats 284
xvii DRAFT - DO NOT CITE OR QUOTE
-------
Figure 5-11. Frequency density of human equivalent doses in specific populations in comparison
to a general population (0.5- to 80-year-old males and females) estimate for an
internal dose of 12.57 mg dichloromethane metabolized by CYP per liter liver per
day 286
Figure 5-12. Frequency density of HECs in specific populations in comparison to a general
population (0.5- to 80-year-old males and females) estimate for an internal dose of
130.0 mg dichloromethane metabolized by CYP per liter liver per day 288
Figure 5-13. Process for deriving cancer OSFs and lURs by using rodent and human PBPK
models 292
Figure 5-14. PBPK model-derived internal doses (mg dichloromethane metabolized via the GST
pathway per liter liver per day) in mice and humans and their associated external
exposures (mg/kg-day) used for the derivation of cancer OSFs based on liver tumors
in mice 295
Figure 5-15. PBPK model-derived internal doses (mg dichloromethane metabolized via the GST
pathways per liter tissue per day) for liver (A) and lung (B) in mice and humans and
their associated external exposures (ppm) used for the derivation of cancer lURs. 309
Figure 5-16. PBPK-model-predicted exposure-response relationships for hepatic CYP and GST
metabolism for continuous inhalation exposure to dichloromethane in 30-year-old
GST+/+women 333
Figure 5-17. Sensitivity coefficients for long-term mass GST-mediated metabolites per liver
volume from a long-term average daily inhalation concentration of 2,000 ppm in
mice 336
Figure 5-18. Sensitivity coefficients for long-term mass GST-mediated metabolites per liver
volume from a long-term average daily drinking water concentration of 500 mg/L in
mice 337
Figure 5-19. Sensitivity coefficients for long-term mass GST-mediated metabolites per lung
volume from a long-term average daily inhalation concentration of 500 ppm in mice.
337
Figure 5-20. Histograms for a liver-specific dose of GST metabolism (mg GST metabolites per
liter liver per day) for the general population (0.5- to 80-year-old males and females),
and specific age/gender groups within the population of GST-T1+/+ genotypes, given
a daily oral dose-rate of 1 mg/kg-day dichloromethane 339
Figure 5-21. Histograms for liver-specific dose of GST metabolism (mg GST metabolites per
liter liver per day) for the general population (0.5- to 80-year-old males and females),
and specific age/gender groups within the population of GST-T1+/+ genotypes, given
a continuous inhalation exposure to 1 mg/m3 dichloromethane 340
Figure B-l. Schematic of the David et al. (2006) PBPK model for dichloromethane in the
human B-l
xviii DRAFT - DO NOT CITE OR QUOTE
-------
Figure B-2. Total CYP2E1 activity (Vmax) normalized to the average total activity in 14-18
year-old individuals (Vmax[14-18]) plotted against normalized BW for individuals
ranging from 6 months to 18 years of age B-7
Figure B-3. Body-weight scaled CYP2E1 activity (Vmaxc) normalized to the average scaled
activity in 14-18 year-old individuals (Vmaxc[14-18]) plotted against age individuals
ranging from 6 months to 18 years of age B-9
Figure B-4. U.S. age distribution, 6 months to 80 years (values from U.S. Census Bureau).. B-l 1
Figure B-5. U.S. age-specific gender distribution (values from U.S. Census Bureau) B-l 1
Figure B-6. Function fits to age-dependent data for BW mean and SDs for males and females in
the United States (values from Portier et al. [2007]) B-13
Figure B-7. Example BW histogram from Monte Carlo simulation for 0.5- to 80-year-old males
and females in the United States (simulated n = 10,000) B-14
Figure B-8. Mean value respiration rates for males and females as a function of age (values from
Clewelletal. [2004]) B-15
Figure B-9. GSDs for respiration rates for males and females as a function of age (values from
Arcus-Arth and Blaisdell [2007]) B-15
Figure B-10. Fraction body fat (VFC) over various age ranges in males and females (data from
Clewelletal. [2004]) B-17
Figure C-1. Schematic of the PBPK model for dichloromethane in the rat C-2
Figure C-2. Observations of exhaled [14C]-labelled dichloromethane (DCM) (left y-axis) and
CO (right y-axis) after a bolus oral dose of 200 mg/kg [14C]-dichloromethane in rats
(data of Angeloetal., 1986b) C-9
Figure C-3. Observations of Gargas et al. (1986; data points) and predictions for Variations A
and C (curves) for respiratory uptake by three rats of 100-3,000 ppm
dichloromethane in a 9-L closed chamber C-12
Figure C-4. Observations (data points) of Angelo et al. (1986b) and predictions for model
Variations A and C (curves) of dichloromethane (DCM) blood concentrations (upper
panel) and amount of dichloromethane exhaled (lower panel) following 10 and 50
mg/kg intravenous DCM injections in rats C-14
Figure C-5. Observations of Andersen et al. (1987; data points) and simulations (curves) for
models A and C for dichloromethane (DCM) in rat blood from inhalation of 200 and
1000 ppm DCM for 4 hours C-16
xix DRAFT - DO NOT CITE OR QUOTE
-------
Figure C-8. Simulation results using model Variations A and C for weekly average metabolic
rates by the GST and CYP pathways for 6 hours/day, 5 days/week inhalation
exposures C-22
Figure D-l. Predicted (logistic model) and observed incidence of noncancer liver lesions in male
F344 rats exposed to dichloromethane in drinking water for 2 years (Serota et al.,
1986a) D-3
Figure D-2. Predicted (log-probit model) and observed incidence of noncancer liver lesions in
female Sprague-Dawley rats inhaling dichloromethane for 2 years (Nitschke 1988a).
D-7
Figure E-l. Predicted and observed incidence of animals with hepatocellular carcinoma or
adenoma in male B6C3Fi mice exposed to dichloromethane in drinking water for 2
years, using liver-specific metabolism dose metric (Serota et al., 1986b; Hazleton
Laboratories, 1983) E-3
Figure E-2. Predicted and observed incidence of animals with hepatocellular carcinoma or
adenoma in male B6C3Fi mice exposed to dichloromethane in drinking water for 2
years, using whole-body metabolism dose metric (Serota et al., 1986b; Hazleton
Laboratories, 1983) E-6
Figure E-3. Predicted and observed incidence of animals with hepatocellular carcinoma or
adenoma in male B6C3Fi mice exposed by inhalation to dichloromethane for 2
years, using liver-specific metabolism dose metric (Mennear et al., 1988; NTP,
1986) E-ll
Figure E-4. Predicted and observed incidence of animals with carcinoma or adenoma in the lung
of male B6C3Fi mice exposed by inhalation to dichloromethane for 2 years, using
liver-specific metabolism dose metric (Mennear et al., 1988; NTP, 1986) E-14
Figure E-5. Predicted and observed incidence of animals with hepatocellular carcinoma or
adenoma in male B6C3Fi mice exposed by inhalation to dichloromethane for 2
years, using whole-body metabolism dose metric (Mennear et al., 1988; NTP, 1986).
E-16
Figure E-6. Predicted and observed incidence of animals with carcinoma or adenoma in the lung
of male B6C3Fi mice exposed by inhalation to dichloromethane for 2 years, using
whole-body metabolism dose metric (Mennear et al., 1988; NTP, 1986) E-19
Figure G-l. PBPK model-derived internal doses (daily average AUC for dichloromethane in
slowly perfused tissue) in rats and humans and their associated external exposures
(ppm) used for the derivation of cancer lURs based on mammary tumors in rats.. G-2
xx DRAFT - DO NOT CITE OR QUOTE
-------
LIST OF ABBREVIATIONS AND ACRONYMS
Al ratio of lung Vmaxc to liver Vmaxc
A2 ratio of lung kfc to liver kfc
ABCOC background amount of CO
ADAF age-dependent adjustment factor
AIC Akaike's Information Criterion
ALT alanine aminotransferase
AP alkaline phosphatase
AST aspartate aminotransferase
AUC area under the curve
BAER brainstem-auditory evoked response
BMD benchmark dose
BMDL10 95% lower bound on the BMD
BMDS benchmark dose software
BMR benchmark response
BW body weight
CAEP cortical-auditory-evoked potential
CASRN Chemical Abstracts Service Registry Number
CHO Chinese hamster ovary
CI confidence interval
CNS central nervous system
CO carbon monoxide
COHb carboxyhemoglobin
CV coefficient of variation
CYP cytochrome P450
DNA deoxyribonucleic acid
FEP flash-evoked potential
FOB functional observational battery
FracR fraction of Vmaxc in rapidly perfused tissues
GD gestational day
GM geometric mean
GSD geometric standard deviation
GSH reduced glutathione
GST glutathione S-transferase
GST-T1 GST-thetal-1
HEC human equivalent concentration
HPRT hypoxanthine-guanine phosphoribosyl transferase
ICD-9 International Classification of Diseases 9* ed.
IgM immunoglobulin M
IRIS Integrated Risk Information System
IUR inhalation unit risk
ka first-order oral absorption rate constant
Km Michaelis-Menten kinetic constant
kfc first-order GST metabolic rate constant
LOAEL lowest-observed-adverse-effect level
LOH loss of heterozygosity
MCHC mean corpuscular hemoglobin concentration
xxi
DRAFT - DO NOT CITE OR QUOTE
-------
MCMC
mRNA
NADPH
NIOSH
NOAEL
NRC
NTP
OR
OSF
OSHA
PSO
PB
PBPK
POD
PND
QAlvC
QCC
QSC
REnCOC
RfC
RfD
SD
SEM
SEP
SMR
SSB
TWA
UF
U.S. EPA
VFC
VLC
VPR
VSC
Markov Chain Monte Carlo
messenger ribonucleic acid
nicotinamide adenine dinucleotide phosphate
National Institute of Occupational Safety and Health
no-observed-adverse-effect level
National Research Council
National Toxicology Program
odds ratio
oral slope factor
Occupational Safety and Health Administration
partial oxygen pressure
blood:air partition coefficient
physiologically based pharmacokinetic
point of departure
postnatal day
allometric alveolar ventilation constant
cardiac output
fractional flow rate of slowly perfused tissues (fraction of QCC)
endogenous rate of CO production
reference concentration
reference dose
standard deviation
standard error of the mean
somatosensory-evoked potential
standardized mortality ratio
single strand break
time-weighted average
uncertainty factor
U.S. Environmental Protection Agency
fractional tissue volume of fat (fraction of BW)
fractional tissue volume of liver (fraction of BW)
CYP maximum velocity
ventilation:perfusion ratio
fractional tissue volume of slowly perfused tissues (fraction of BW)
xxn
DRAFT - DO NOT CITE OR QUOTE
-------
FOREWORD
The purpose of this Toxicological Review is to provide scientific support and rationale
for the hazard and dose-response assessment in IRIS pertaining to exposure to dichloromethane.
It is not intended to be a comprehensive treatise on the chemical or toxicological nature of
di chl or om ethane.
The intent of Section 6, Major Conclusions in the Characterization of Hazard and Dose
Response, is to present the major conclusions reached in the derivation of the reference dose,
reference concentration and cancer assessment, where applicable, and to characterize the overall
confidence in the quantitative and qualitative aspects of hazard and dose response by addressing
the quality of data and related uncertainties. The discussion is intended to convey the limitations
of the assessment and to aid and guide the risk assessor in the ensuing steps of the risk
assessment process.
For other general information about this assessment or other questions relating to IRIS,
the reader is referred to EPA's IRIS Hotline at (202) 566-1676 (phone), (202) 566-1749 (fax), or
hotline.iris@epa.gov (email address).
xxiii DRAFT - DO NOT CITE OR QUOTE
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS
CHEMICAL MANAGERS
Glinda S. Cooper, Ph.D.
Ambuja S. Bale, Ph.D., DABT
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
AUTHORS
Glinda S. Cooper, Ph.D.
Ambuja S. Bale, Ph.D., DABT
Paul Schlosser, Ph.D.
Gene (Ching-Hung) Hsu, Ph.D., DABT
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
John C. Lipscomb, Ph.D., DABT
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Cincinnati, OH
CONTRIBUTORS
Andrew Rooney, Ph.D.
Allan Marcus, Ph.D.
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection
David Eastmond, Ph.D.
Environmental Toxicology Graduate Program
University of California, Riverside
Riverside, CA
CONTRACTOR SUPPORT
Peter McClure, Ph.D., DABT
Michael Lumpkin, Ph.D.
Fernando Llados, Ph.D.
Mark Osier, Ph.D., DABT
Daniel Plewak, B.S.
XXIV
DRAFT - DO NOT CITE OR QUOTE
-------
Syracuse Research Corporation
Syracuse, NY
Elizabeth Dupree Ellis, Ph.D.
Oak Ridge Institute for Science and Education
Center for Epidemiologic Research
Oak Ridge, TN
REVIEWERS
This document has been provided for review to EPA scientists, interagency reviewers
from other federal agencies and White House offices, and the public, and peer reviewed by
independent scientists external to EPA. A summary and EPA's disposition of the comments
received from the independent external peer reviewers and from the public is included in
Appendix A.
INTERNAL EPA REVIEWERS
Ghazi Dannan, Ph.D.
Catherine Gibbons, Ph.D.
Karen Hogan, M.S.
Jennifer Jinot, Ph.D.
Paul White
Samantha Jones, Ph.D.
Jamie Strong, Ph.D.
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
David Herr, Ph.D.
National Health and Environmental Effect Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
EXTERNAL PEER REVIEWERS
James V. Bruckner, Ph.D.
Department of Pharmaceutical and Biomedical Sciences, College of Pharmacy
University of Georgia
David W. Gaylor, Ph.D.
Gaylor and Associates, LLC
Lisa M. Kamendulis, Ph.D.
Department of Pharmacology and Toxicology, Division of Toxicology
Indiana University School of Medicine
xxv DRAFT - DO NOT CITE OR QUOTE
-------
Kannan Krishnan, Ph.D.
Departement de sante environnementale et sante au travail, Faculte de medicine
Universite de Montreal
Harihara M. Mehendale, Ph.D.
Department of Toxicology, College of Pharmacy
The University of Louisiana at Monroe
Martha M. Moore, Ph.D.
National Center for Toxicological Research
U.S. Food and Drug Administration
Andrew G. Salmon, D.Phil.
Air Toxicology and Epidemiology Branch, Office of Environmental Health Hazard Assessment
California Environmental Protection Agency
xxvi DRAFT - DO NOT CITE OR QUOTE
-------
1. INTRODUCTION
This document presents background information and justification for the Integrated Risk
Information System (IRIS) Summary of the hazard and dose-response assessment of
dichloromethane. IRIS Summaries may include oral reference dose (RfD) and inhalation
reference concentration (RfC) values for chronic and other exposure durations, and a
carcinogenicity assessment.
The RfD and RfC, if derived, provide quantitative information for use in risk assessments
for health effects known or assumed to be produced through a nonlinear (presumed threshold)
mode of action. The RfD (expressed in units of mg/kg-day) is defined as an estimate (with
uncertainty spanning perhaps an order of magnitude) of a daily exposure to the human
population (including sensitive subgroups) that is likely to be without an appreciable risk of
deleterious effects during a lifetime. The inhalation RfC (expressed in units of mg/m3) is
analogous to the oral RfD, but provides a continuous inhalation exposure estimate. The
inhalation RfC considers toxic effects for both the respiratory system (portal-of-entry) and for
effects peripheral to the respiratory system (extrarespiratory or systemic effects). Reference
values are generally derived for chronic exposures (up to a lifetime), but may also be derived for
acute (<24 hours), short-term (>24 hours up to 30 days), and subchronic (>30 days up to 10% of
lifetime) exposure durations, all of which are derived based on an assumption of continuous
exposure throughout the duration specified. Unless specified otherwise, the RfD and RfC are
derived for chronic exposure duration.
The carcinogenicity assessment provides information on the carcinogenic hazard
potential of the substance in question and quantitative estimates of risk from oral and inhalation
exposure may be derived. The information includes a weight-of-evidence judgment of the
likelihood that the agent is a human carcinogen and the conditions under which the carcinogenic
effects may be expressed. Quantitative risk estimates may be derived from the application of a
low-dose extrapolation procedure. If derived, the oral slope factor is a plausible upper bound on
the estimate of risk per mg/kg-day of oral exposure. Similarly, an inhalation unit risk is a
plausible upper bound on the estimate of risk per ug/m3 air breathed.
Development of these hazard identification and dose-response assessments for
dichloromethane has followed the general guidelines for risk assessment as set forth by the
National Research Council (1983). EPA Guidelines and Risk Assessment Forum Technical
Panel Reports that may have been used in the development of this assessment include the
following: Guidelines for the Health Risk Assessment of Chemical Mixtures (U.S. EPA, 1986a),
Guidelines for Mutagenicity Risk Assessment (U.S. EPA, 1986b), Recommendations for and
Documentation of Biological Values for Use in Risk Assessment (U.S. EPA, 1988a), Guidelines
for Developmental Toxicity Risk Assessment (U.S. EPA, 1991), Interim Policy for Particle Size
1 DRAFT - DO NOT CITE OR QUOTE
-------
and Limit Concentration Issues in Inhalation Toxicity Studies (U.S. EPA, 1994a), Methods for
Derivation of Inhalation Reference Concentrations and Application of Inhalation Dosimetry
(U.S. EPA, 1994b), Use of the Benchmark Dose Approach in Health Risk Assessment (U.S. EPA,
1995), Guidelines for Reproductive Toxicity Risk Assessment (U.S. EPA, 1996), Guidelines for
Neurotoxicity Risk Assessment (U.S. EPA, 1998), Science Policy Council Handbook: Risk
Characterization (U.S. EPA, 2000a), Benchmark Dose Technical Guidance Document (U.S.
EPA, 2000b), Supplementary Guidance for Conducting Health Risk Assessment of Chemical
Mixtures (U.S. EPA, 2000c), A Review of the Reference Dose and Reference Concentration
Processes (U.S. EPA, 2002), Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a),
Supplemental Guidance for Assessing Susceptibility from Early-Life Exposure to Carcinogens
(U.S. EPA, 2005b), Science Policy Council Handbook: Peer Review (U.S. EPA, 2006a), and A
Framework for Assessing Health Risk of Environmental Exposures to Children (U.S. EPA,
2006b).
The literature search strategy employed for this compound was based on the Chemical
Abstracts Service Registry Number (CASRN) and at least one common name. Any pertinent
scientific information submitted by the public to the IRIS Submission Desk was also considered
in the development of this document. The relevant literature was reviewed through December
2010. It should be noted that references have been added to the Toxicological Review after the
external peer review in response to peer reviewers' comments and for the sake of completeness.
These references have not changed the overall qualitative and quantitative conclusions. In
Section 7, references added after peer review are noted with an asterisk.
DRAFT - DO NOT CITE OR QUOTE
-------
2. CHEMICAL AND PHYSICAL INFORMATION
Dichloromethane is a colorless liquid with a penetrating, ether-like odor (Lewis, 1997).
Selected chemical and physical properties of dichloromethane are listed in Table 2-1.
Table 2-1. Physical properties and chemical identity of dichloromethane
CAS number
Synonyms
Molecular weight
Chemical formula
Boiling point
Melting point
Vapor pressure
Density
Vapor density
Water solubility
Other solubility
Partition coefficient
Flash point
Auto ignition temperature
Latent heat of vaporization
Heat of fusion
Critical temperature
Critical pressure
Viscosity
Henry's law constant
OH reaction rate constant
Chemical structure
Physical property/chemical identity
75-09-2
Methylene chloride, methylene dichloride,
methyl bichloride
84.93
CH2C12
40°C
-95.1°C
1.15 x 102mmHgat25°C
1.3266g/mLat20°C
2.93 (air= 1.02)
1.30 x 104 mg/L at 25°C
Miscible in ethanol, ether, and
dimethylformamide; soluble in carbon
tetrachloride
log Kow= 1.25
Not flammable
640°C
3.30 x 105J/kg
16.89 cal/g
245.0°C
6.171 x !06Pa
0.430 cP at 20°C
3.25 x 10-3atmm3/molat25°C
1.42 x 10'13 cnrVmolecule sec at 25°C
H
f^\ r* r*\
Ul U Ul
H
Reference
Lide (2000)
O'Neiletal. (2001)
O'Neiletal. (2001)
O'Neiletal. (2001)
Lide (2000)
Lide (2000)
Boubliketal. (1984)
Lide (2000)
Holbrook (2003)
Horvath(1982)
IARC (1999)
Hanschetal. (1995)
U.S. Coast Guard (1999)
Holbrook (2003)
U.S. Coast Guard (1999)
U.S. Coast Guard (1999)
Holbrook (2003)
Holbrook (2003)
Lewis (1997)
Leighton and Calo (1981)
Atkinson (1989)
Dichloromethane is produced by two methods of manufacturing (IARC, 1999). The
older method involves the direct reaction of methane with chlorine either at high temperatures or
:To avoid confusion, "dichloromethane" is used throughout this summary even if a specific paper used the term
"methylene chloride."
DRAFT - DO NOT CITE OR QUOTE
-------
at lower temperatures under catalytic or photolytic conditions (Holbrook, 2003). The more
common method used today involves an initial reaction of hydrochloric acid with methanol to
yield methyl chloride. Excess methyl chloride is then reacted in the gas phase thermally with
chlorine to produce dichloromethane (Holbrook, 2003). This process can also be carried out
catalytically or photolytically.
Dichloromethane became an important industrial chemical in the United States during
World War II (Hardie, 1964). Dichloromethane has been used in paint strippers and removers,
as a propellant in aerosols, in the manufacture of drugs, pharmaceuticals, film coatings,
electronics, and polyurethane foam, and as a metal-cleaning solvent. Dichloromethane can also
be used in the decaffeination process of coffee and tea (ATSDR, 2000). The U.S. production
was 3.8 million pounds in 1941 and 8.3 million pounds in 1944 (Searles and McPhail, 1949).
Dichloromethane production rose sharply in the decades following the war due to the increased
demand for this substance for use mainly in paint strippers (Hardie, 1964; Searles and McPhail,
1949). U.S. production in 1947, 1955, 1960, and 1962 was approximately 19, 74, 113, and
144 million pounds, respectively (Hardie, 1964; Searles and McPhail, 1949). As other solvent
uses and its use in aerosol propellants became important, demand for this substance increased
further (Anthony, 1979). Dichloromethane production continued to rise dramatically through the
1970s; production capacities were 520 million pounds in 1973 and 830 million pounds in 1979
(CMR, 1979, 1973).
After 1980, production of dichloromethane began to decline. Production capacities fell
from 722 million pounds in 1982 to 465 million pounds in 1997 (CMR, 1997, 1982). The total
U.S. production capacity for dichloromethane in 2000 was 535 million pounds (CMR, 2000).
The demand for dichloromethane decreased from 600 million pounds in 1979 to 200 million
pounds in 1999 (CMR, 2000, 1979). The decline in production of and demand for
dichloromethane over the past 2 decades has been attributed to increased regulation, the use of
alternative chemicals in aerosol spray cans, and concern over dichloromethane carcinogenicity
(Holbrook, 2003; ATSDR, 2000).
Dichloromethane in the environment will partition mainly to air (NLM, 2003). In air,
dichloromethane exists as a vapor. Some of the dichloromethane released to soil or water is
expected to volatilize to air. In soil, dichloromethane is expected to be highly mobile and may
migrate to groundwater. The potential for dichloromethane to bioconcentrate in aquatic or
marine organisms is low. Dichloromethane may biodegrade in soil or water under both aerobic
and anaerobic conditions.
DRAFT - DO NOT CITE OR QUOTE
-------
3. TOXICOKINETICS
3.1. ABSORPTION
3.1.1. Oral—Gastrointestinal Tract Absorption
There are currently no data available on absorption of dichloromethane following oral
intake in humans. However, after oral administration in animals, dichloromethane is rapidly and
nearly completely absorbed in the gastrointestinal tract (Angelo et al., 1986a, b; McKenna and
Zempel, 1981). Angelo et al. (1986b) reported that, following administration of single
radiolabeled oral doses (10, 50, or 200 mg/kg) to mature male F344 rats, 97% of the label was
detected in the exhaled air within 24 hours, indicating nearly complete absorption. At several
time points within 40 minutes of dose administration, <2% of the dose was found in the lower
part of the gastrointestinal tract, indicating that the majority of dichloromethane absorption
occurs in the upper gastrointestinal tract (Angelo et al., 1986b). Similar results were reported in
mature male B6C3Fi mice exposed to up to 50 mg/kg (Angelo et al., 1986a). In mature male
Sprague-Dawley rats administered a single dose (1 or 50 mg/kg) of radiolabeled
dichloromethane, <1% of the label was found in feces collected for 48 hours after dose
administration (McKenna and Zempel, 1981). Absorption of dichloromethane generally follows
first-order kinetics (Angelo et al., 1986a), and freely passes through phospholipid cell
membranes by passive diffusion. The vehicle appears to affect the rate but not the extent of
gastrointestinal absorption, with an aqueous vehicle resulting in a more rapid absorption of
dichloromethane than an oil-based vehicle (Angelo et al., 1986a).
3.1.2. Inhalation—Respiratory Tract Absorption
Several studies in humans have demonstrated the absorption of dichloromethane
following inhalation exposure. In a study by Astrand et al. (1975), 14 male volunteers (ages 19-
29) were exposed to about 870 mg/m3 (250 ppm) or 1,740 mg/m3 (500 ppm) for 30 minutes
while resting or exercising on a bicycle ergometer. There was a pause of about 20 minutes
without exposure between rest and exercise periods. Uptake of dichloromethane was estimated
at about 55% while resting and about 40, 30, and 35% at respective workloads of 50, 100, and
150 watts2. Blood levels of dichloromethane correlated directly with exposure concentrations,
and did not appear to increase when a workload was applied (Astrand et al., 1975). Similar
reports of rapid uptake and a direct correlation between dichloromethane exposure level and
blood levels in humans have been presented by other groups (DiVincenzo and Kaplan, 1981;
DiVincenzo et al., 1971).
2 A watt is the International System Unit of power and is equal to 1 joule of energy per second. It is a measure of
the rate of energy use or production (i.e., the exercise effort that was exerted by the individuals in the study).
5 DRAFT - DO NOT CITE OR QUOTE
-------
With extended (>l-2 hours) exposure, uptake tends to reach a steady-state level, at which
point blood dichloromethane levels remain more or less constant (DiVincenzo and Kaplan, 1981;
DiVincenzo et al., 1972; Riley et al., 1966). DiVincenzo et al. (1972) reported that in humans
exposed to 100 or 200 ppm of dichloromethane for 2 hours (without physical exercise),
dichloromethane was rapidly absorbed, reaching an approximate steady state, as assessed by
levels of unchanged dichloromethane in the expired air, within the first 15-30 minutes of
exposure. A later study by the same group (DiVincenzo and Kaplan, 1981) similarly reported a
rapid absorption of dichloromethane in volunteers exposed to 50-200 ppm for 7.5 hours on each
of 5 consecutive days. A steady-state level, as assessed by levels of unchanged dichloromethane
in the expired air, was reached quickly (1-2 hours), with exhaled dichloromethane levels
increasing with increasing exposure level. A similar pattern was seen with blood
dichloromethane levels. Estimated pulmonary uptake was 69-75% and did not vary appreciably
with exposure concentration. In another experiment in which one of the investigators was seated
during exposure to 100 ppm dichloromethane for 2 hours, concentrations of dichloromethane in
expired air reached an apparent plateau of about 70 ppm within the first hour of exposure (Riley
etal., 1966).
Body fat may influence absorption of dichloromethane, as evidenced by data from an
experiment involving 12 men ages 21-35, divided into two groups (n = 6 per group) based on
percent body fat (Engstrom and Bjurstrom, 1977). The mean percent body fat in the leaner
group was 7.8% (standard error of the mean [SEM] 1.9), range 2.3-13.6%, compared with
25.1% (SEM 2.8), range 18.3-36.2%, in the more overweight group. Total uptake of
dichloromethane during a light exercise period (50 watts) for 1 hour with an exposure level of
750 ppm was positively correlated with percent body fat (r = 0.81), and the estimated amount of
dichloromethane in fat storage was also correlated with percent body fat (r = 0.84).
A pattern of absorption similar to that seen in humans has been seen in animals. Initially,
dichloromethane is readily absorbed following inhalation exposure, as evidenced by rapid
appearance of dichloromethane in blood, tissues, and expired air (Withey and Karpinski, 1985;
Stott and McKenna, 1984; Anders and Sunram, 1982; Carlsson and Hultengren, 1975; Roth et
al., 1975). For example, absorption of inhaled 500 ppm dichloromethane in anesthetized, mature
male F344 rats reached an apparent plateau within 10-20 minutes and was relatively constant for
up to 2 hours (Stott and McKenna, 1984). In these experiments, absorption was calculated from
measurements of exposure (nose only) and effluent concentrations and ventilation flow rate in
intact animals; double tracheostomized rats were used to measure absorption in the isolated
upper respiratory tract and the lower respiratory tract. At a ventilation rate of 53 mL/minute,
absorption expressed as mean percentage of dichloromethane available for absorption was 44%
(standard deviation [SD] 10) in intact rats, 13.2% (SD 3.6) in the upper respiratory tract, and
37% (SD 4.1) in the lower respiratory tract.
DRAFT - DO NOT CITE OR QUOTE
-------
3.2. DISTRIBUTION
Results from studies of animals show that, following absorption, dichloromethane is
rapidly distributed throughout the body and has been detected in all tissues that have been
evaluated. Twenty minutes after a single intravenous dose of 10 mg [14C]-dichloromethane/kg to
mature male B6C3Fi mice (Angelo et al., 1986a), total label was greatest in the liver
(6.72 ug-equivalents/g tissue), with lower levels reported in the lung (1.82 ug-equivalents/g
tissue), kidney (1.84 ug-equivalents/g tissue), and the remainder of the carcass
(1.90 ug-equivalents/g tissue). By 4 hours post administration, levels in the liver had fallen to
3.08 ug-equivalents/g tissue, lung levels were 0.64 ug-equivalents/g tissue, and carcass levels
were 0.23 ug-equivalents/g tissue. The levels in the kidney rose sharply in the first hour
postexposure but then fell and remained steady at -1.60 ug-equivalents/g tissue for the
remaining 3 hours of the study (Angelo et al., 1986a). McKenna et al. (1982) exposed groups of
mature male Sprague-Dawley rats to 50, 500, or 1,500 ppm [14C]-labeled dichloromethane for
6 hours and examined tissues at 48 hours for presence of radiolabel; results are shown in
Table 3-1. The greatest concentration of label was found in the liver, followed by the kidney and
lung.
Table 3-1. Distribution of radioactivity in tissues 48 hours after inhalation
exposure of mature male Sprague-Dawley rats (n = 3) for 6 hours
Tissue
Liver
Kidney
Lung
Brain
Epididymal fat
Skeletal muscle
Testes
Whole blood
Remaining carcass
Mean ± SD, jig-equivalent dichloromethane/g tissue, by exposure level
50 ppm
8.4 ±1.5
3.3 ±0.1
1.9 ±0.2
0.8 ±0.3
0.5 ±0.2
l.liO.l
1.1 ±0.2
1.1 ±0.2
1.3 ±0.2
500 ppm
35.6 ±7.5
16.2 ±2.4
11.0±1.3
4.2 ±1.3
6.5 ±0.5
4.4 ±1.9
5.5 ±1.3
8.1 ±1.9
5.9 ±0.9
1,500 ppm
44.2 ±3.5
30.5 ±0.2
16.5 ±1.6
6.7 ±0.2
4.1 ±0.9
7.7 ±0.7
8.1 ±0.5
8.9 ±1.7
8.6 ±1.4
Source: McKenna etal. (1982).
As noted in the preceding section, body fat may affect the uptake of dichloromethane,
and there is also evidence of a relationship between adiposity and dichloromethane storage. In
the study by Engstrom and Bjurstrom (1977) involving 12 men ages 21-35 exposed to 750 ppm
dichloromethane during a 1-hour light exercise (50 watts) period, dichloromethane was measured
in body fat biopsy specimens at 1, 2, 3, and 4 hours postexposure. All specimens were taken
from the buttocks. The concentration of dichloromethane (per gram tissue) was negatively
DRAFT - DO NOT CITE OR QUOTE
-------
correlated with percent body fat, but the total estimated amount of dichloromethane in fat tissue
4 hours postexposure was higher in subjects with a higher amount of fat (r = 0.84).
Carlsson and Hultengren (1975) exposed groups of 10 mature male Sprague-Dawley rats
to [14C]-dichloromethane for 1 hour at a mean concentration of 1,935 mg/m3 (557 ppm) and SD
of 90 mg/m3 (26 ppm). The initial levels were highest in the white adipose tissue (approximately
80 jig dichloromethane per gram tissue) compared with approximately 35, 20, and 5 jig-
equivalent dichloromethane/g tissue in the liver, kidney and adrenal glands, and brain,
respectively. These initial levels in the adipose quickly fell to <10 jig-equivalent
dichloromethane/g tissue; more moderate declines were seen in the other tissues.
With acute 6-hour exposure scenarios, peak exposure concentrations during the exposure
period may have a greater influence on dichloromethane levels in the brain and perirenal fat than
time-weighted average (TWA) concentrations (Savolainen et al., 1981). In rats exposed over a
6-hour period for 5 days/week to a TWA of 1,000 ppm dichloromethane consisting of two 1-hour
peak concentrations (2,800 ppm) interspersed with exposure to 100 ppm, levels of
dichloromethane in the brain was 3-fold higher (p < 0.001) than corresponding levels in rats
exposed to constant levels of 1,000 ppm; a two-fold increased risk was seen in the
dichloromethane levels in perirenal fat after one week of exposure (p < 0.001), but this
difference was much smaller after two weeks of exposure. This difference was not seen with
blood carbon monoxide (CO) levels (Table 3-2). With constant exposure concentrations of 500
or 1,000 ppm, perirenal fat levels of dichloromethane approximately doubled following 2 weeks
of exposure compared with 1 week of exposure, indicating that some storage of dichloromethane
in fat tissue can occur with repeated exposure scenarios (Table 3-2). In contrast, brain levels of
dichloromethane in rats exposed for 1 week were higher than brain levels in rats exposed for
2 weeks. One possible explanation of these observations is that there is an induction of enzymes
involved in dichloromethane metabolism in liver and other tissues with repeated exposure, and
dichloromethane in fat is poorly metabolized.
DRAFT - DO NOT CITE OR QUOTE
-------
Table 3-2. Brain and perirenal fat dichloromethane and blood CO
concentrations in male Wistar rats exposed by inhalation to
dichloromethane at constant exposure concentrations compared with
intermittently high exposure concentrations
Exposure level3
(TWA, ppm)
Control
500, constant
1,000, constant
1,000, with two 1-hr
peaks of 2,800 ppm
Exposure wks
1
2
Brain (nmol/g)
0
30 ±7
33 ±2
lll±18b
0
9±3
14 ±3
50±15b
1
2
Perirenal fat (nmol/g)
0
436 ± 47
1,316 ±209
2,295 ± 147b
0
918 ±215
2,171 ±219
2,431 ±146
1
2
Blood CO (nmol/g)
40 ±15
675 ± 195
876 ± 80
728 ± 84
30 ±10
781 ±62
825 ± 56
873 ± 90
"Groups of five rats were exposed to 0, 50, or 1,000 ppm 6 hrs/d or 100 ppm interspersed with two 1-hr peaks of
2,800 ppm for 5 d/wk for 1 or 2 wks. Tissue concentration values are mean ± SD.
bDifference between 1,000 ppm TWA constant exposure, p < 0.001; t-test calculated by EPA using sample size,
mean and standard deviation as provided by Savolainen et al. (1981).
Source: Savolainen etal. (1981).
Placental transfer. Dichloromethane is capable of crossing the placental barrier and
entering the fetal circulation. Anders and Sunram (1982) reported that when pregnant Sprague-
Dawley rats (n = 3) were exposed to 500 ppm dichloromethane for 1 hour on gestational day
(GD) 21, mean maternal blood levels were 176 nmol/mL (SEM 50), while fetal levels were
115 nmol/mL (SEM 40); interestingly, the levels of CO, a metabolite of dichloromethane, were
similar in both the maternal blood (167 nmol/mL, SEM 12) and fetal blood (160 nmol/mL, SEM
31). Withey and Karpinski (1985) also reported higher maternal compared with fetal
dichloromethane levels based on a study of five pregnant Sprague-Dawley rats exposed to 107-
2,961 ppm of dichloromethane. Maternal blood levels of dichloromethane were 2-2.5-fold
higher than those found in the fetal circulation.
Blood-brain barrier transfer. Dichloromethane is thought to readily transfer across the
blood-brain barrier by passive diffusion, as evidenced by the detection of radioactivity in brain
tissue 48 hours after exposures of rats to radiolabeled dichloromethane at concentrations of 50,
500, or 1,500 ppm for 6 hours (McKenna et al., 1982) (see Table 3-1), and the historical
demonstrations that dichloromethane has transient sedative and anesthetic properties in humans
(for review of these reports, see Mattsson et al. [1990] and Winneke [1974]). Dichloromethane
is no longer used as an anesthetic gas because the margin between anesthetic and lethal doses is
narrow (Winneke, 1974).
3.3. METABOLISM
Metabolism of dichloromethane involves two primary pathways, outlined in Figure 3-1
(ATSDR, 2000; Guengerich, 1997; Hashmi et al., 1994; Gargas et al., 1986). Dichloromethane
DRAFT - DO NOT CITE OR QUOTE
-------
is metabolized to CO in a cytochrome P450 (CYP)-dependent oxidative pathway that is
predominant at low exposure levels. The CYP-related pathway results in the addition of oxygen,
followed by spontaneous rearrangement to formyl chloride, and then to CO; each spontaneous
rearrangement releases H+ and Cl" ions. At higher exposure levels, the CYP pathway becomes
saturated and a second pathway begins to predominate. Glutathione S-transferase
(GST)-catalyzed addition of glutathione (GSH) is the initial step in this pathway. The
replacement of one of the chlorine atoms with the S-glutathione group results in formation of
S-(chloromethyl)glutathione and the release of H+ and Cl" ions. Hydration of
S-(chloromethyl)glutathione results in an S-glutathionyl methanol molecule, which can
spontaneously form formaldehyde or rearrange to form an S-glutathione formaldehyde molecule,
and then further rearrange to formate. Both formaldehyde and formate can then be further
metabolized to CO2.
OH
H-C-H
GS
GSTT1
Cl ^
H-C-H
GS
S-(chloromethyl)
glutathione
0
Dichloromethane
Cl
1
H-C-H
^ JL \
Cl x
CYP2E1
X
O
A
Cl^ H
Formyl Chloride
1
(minor pathway) Q
*=* H^CX
H
II
Formaldehyde G - S **'' ^ X H
S-glutathionyl methanol y
O
1
H |
1
CO
Carbon Monoxide
I
COHb
Carboxyhemoglobin
H |
'CO,
^ U v
G-S H C02
\L O
A Formic acid —> CO2
^ O N!
OH H
Adapted from: ATSDR (2000); Guengerich (1997); Hashmi et al. (1994); Gargas
etal. (1986).
Figure 3-1. Proposed pathways for dichloromethane metabolism.
As described in the following discussion of the CYP- and GST-mediated metabolism,
these two pathways effectively compete for the available dichloromethane. Because
dichloromethane binds to the CYP reaction site with higher affinity than the GST site, most of
the dichloromethane is metabolized by CYP at lower exposure levels. As the available CYP
enzyme becomes saturated at higher exposure levels more dichloromethane is left available for
10 DRAFT - DO NOT CITE OR QUOTE
-------
binding to the lower-affinity GST metabolic site, and the proportion metabolized by GST
increases. Both pathways are expected to operate, however, even at low exposures. At low
substrate concentrations, the rates of enzyme-catalyzed reactions are determined by the
probability of molecular collision between the substrate and enzyme active site, and the
probability of the subsequent reaction step (versus possible release of the un-reacted substrate
molecule from the site). This joint probability is directly proportional to, or first order in, the
concentrations of the substrate(s) and enzyme. In particular, as long as GSH, GST, and
dichloromethane are present at non-zero concentrations, the probability of these molecular
collisions and transformations will be non-zero, hence the rate of the GST-mediated reaction will
be non-zero. Exposure to other agents may shift the balance between the pathways. For
example, pretreatment with compounds that deplete GSH (e.g., buthionine sulfoximine,
diethylmaleate, phorone) resulted in an increase in blood carboxyhemoglobin (COHb) levels
following a single injection of dichloromethane relative to animals that did not receive GSH
depletion, indicating a shift to the CYP pathway (Oh et al., 2002). Similarly, co-exposure to
agents that compete for CYP2E1 results in a shift toward the GST pathway and away from CO
production (Lehnebach et al., 1995; Pankow and Jagielki, 1993; Pankow et al., 1991a, b; Glatzel
etal., 1987; Roth etal., 1975).
3.3.1. The CYP2E1 Pathway
There is considerable evidence of the importance of the CYP2E1 metabolic pathway in
studies in animals (Oh et al., 2002; Wirkner et al., 1997; Kim and Kim, 1996; Lehnebach et al.,
1995; Pankow et al., 1991a, b; Pankow and Hoffmann, 1989; Pankow, 1988; Glatzel et al., 1987;
Angelo et al., 1986a, b; Landry et al., 1983; Anders and Sunram, 1982; McKenna et al., 1982;
McKenna and Zempel, 1981; Rodkey and Collison, 1977; Carlsson and Hultengren, 1975; Roth
et al., 1975; Fodor et al., 1973) and humans (Takeshita et al., 2000; DiVincenzo and Kaplan,
1981; Astrand et al., 1975). These studies demonstrate that exposure to dichloromethane,
regardless of exposure route, results in the formation of CO, as assessed by direct measurements
of elevated levels of CO in expired air and increased levels of COHb in the blood.
The first step in the CYP2E1 pathway is the formation of formyl chloride (Figure 3-1).
Watanabe and Guengerich (2006) conducted a series of studies to investigate the downstream
metabolites of formyl chloride and reported only marginal (3% maximum at pH 9) formation of
^-formyl GSH from formyl chloride in the presence of GSH. Therefore, most (>97%) of the
formyl chloride is metabolized further to CO. Furthermore, CO formation from formyl chloride
was independent of GSH presence in the assay.
Results from numerous studies in rats in which CYP2E1 metabolism was blocked or
induced indicate that the generation of CO occurs as a result of metabolism of dichloromethane
by the CYP2E1 pathway (Figure 3-1). Co-exposure of rats to a high dose of ethanol
(174 mmol/kg), which is metabolized by CYP2E1, and dichloromethane (1.6, 6.2,
11 DRAFT - DO NOT CITE OR QUOTE
-------
15.6 mmol/kg) resulted in no increase in blood COHb, indicating that the metabolic pathway for
CO formation had been either blocked or saturated (Glatzel et al., 1987). Similar results have
been seen with coadministration of other known CYP substrates, including diethyldithio-
carbamate (Lehnebach et al., 1995), methanol (Pankow and Jagielki, 1993), benzene, toluene,
and three xylene isomers (Pankow et al., 1991b). Pretreatment of animals with CYP inducers
(e.g., benzene, toluene, xylenes, methanol, isoniazid), particularly those that induce CYP2E1,
resulted in an increased level of CO formation, as assessed by COHb formation or measurement
in expired air, following single exposures to dichloromethane (Kim and Kim, 1996; Pankow and
Jagielki, 1993; Pankow et al., 1991b; Pankow and Hoffmann, 1989; Pankow, 1988).
Pretreatment with disulfuram, a CYP2E1 blocker, resulted in a complete lack of formation of
COHb following dichloromethane exposure, indicating that CYP2E1 is the isozyme responsible
for metabolism of dichloromethane (Kim and Kim, 1996).
Evidence in hamster and rat studies suggests that the CYP2E1 pathway becomes
saturated at high dichloromethane exposure levels; comparable data from studies in mice were
not found. In hamsters, mean COHb percentages were elevated to a similar degree (about 28-
30%, compared with <1% in controls) in three groups exposed by inhalation to 500, 1,500, or
3,500 ppm dichloromethane for 6 hours (Burek et al., 1984). After 21 months of exposure by
this protocol, mean COHb percentages in the three exposure groups remained similarly elevated,
indicative of saturation of the CYP2E1 pathway in hamsters at exposure levels >500 ppm and a
lack of accumulation of dichloromethane and CYP2E1 metabolites with chronic exposure.
McKenna et al. (1982) found that blood COHb levels in rats increased when inhalation exposure
concentration was increased from 50 to 500 ppm but that similar levels of COHb were reported
following exposure to 1,500 ppm as following exposure to 500 ppm; the peak blood COHb
percentages were approximately 10%. In rats exposed to 0, 50, 200, or 500 ppm for 6 hours/day,
5 days/week for 2 years, mean COHb percentages were 2.2, 6.5, 12.5, and 13.7%, respectively,
suggesting that saturation of the CYP2E1 pathway is approached at 200 ppm (Nitschke et al.,
1988a). In male F344 rats exposed for 4 hours to dichloromethane concentrations of about 150,
300, 600, 1,000, and 2,000 ppm, mean COHb percentages (estimated from a figure) were about
4% at 150 ppm and about 8% at each of the four higher exposure concentrations (Gargas et al.,
1986). McKenna and Zempel (1981) reported that increasing the oral dose of labeled
dichloromethane from 1 to 50 mg/kg in rats resulted in a lower fraction of the total dose being
metabolized to CO. Single injections of 3 and 6 mmol/kg of dichloromethane in rats resulted in
nearly identical levels of blood COHb (Oh et al., 2002).
In human subjects exposed to dichloromethane in the workplace, saturation of CYP
metabolism appears to be approached by the 400-500 ppm range (Ott et al., 1983e). Blood
samples were drawn during working hours from 136 fiber production workers who were exposed
to dichloromethane, acetone, and methanol. A comparison group of acetate production workers
from another plant exposed only to acetone was also included in this study. TWA exposure
12 DRAFT - DO NOT CITE OR QUOTE
-------
concentrations for the workers were determined by personal monitoring techniques, and percent
COHb levels in the blood samples were determined. Estimated TWA concentrations in the
exposed workers followed a bimodal distribution, with a lower mode of exposure concentrations
in the 150-200 ppm range and the higher mode in the range of 450-500 ppm; only 21% (29 out
of 136 workers) were in the 200-400 ppm range. In categorical analyses, the workers were
divided into three groups with 22, 68, and 46 individuals, respectively, in the <100, 100 to 299,
and >300-ppm groups. Plots of percent COHb against TWA exposure concentrations showed
the appearance of saturation at around 400 ppm, with the beginning of the plateauing occurring
around 300 ppm.
The liver is the tissue most enriched in CYP2E1 catalytic activity, but CYP2E1 protein
and messenger ribonucleic acid (mRNA) have been detected in other human tissues, including
the lung, brain, kidney, pancreas, bladder, small intestine, and blood lymphocytes (Nishimura et
al., 2003). As such, the liver is expected to be the main site of CYP metabolism of
dichloromethane, but other tissues are also expected to metabolize dichloromethane via this
pathway. Of particular relevance given the neurologic effects seen with dichloromethane are the
distribution and inducibility of CYP2E1 in different areas of the brain (Miksys and Tyndale,
2004). Individuals with decreased CYP2E1 activity may experience decreased generation of CO
and an increased level of GST-related metabolites following exposure to dichloromethane. As a
result, these individuals may be more susceptible to the chronic effects of dichloromethane from
GST-related metabolites than individuals with higher levels of CYP2E1 activity. Conversely,
individuals with higher CYP2E1 activity may experience relatively increased generation of CO
at a given dichloromethane exposure level and, therefore, may be more susceptible to the acute
toxicity of dichloromethane (from CO). However, there are currently no studies that have
evaluated dichloromethane and CO effects with variable levels CYP2E1 activity.
Results from studies examining human interindividual variation in CYP2E1 activities
(e.g., catalytic activities, protein levels, or mRNA levels) indicate that individuals may vary in
their ability to metabolize dichloromethane through the CYP2E1 pathway. In a study of liver
samples from 30 Japanese and 30 Caucasian individuals, two- to threefold variation was found in
the levels of CYP2E1 protein, whereas catalytic activity toward substrates associated with
CYP2E1 (e.g., 7-ethoxycoumarin) displayed a wider range of values, approximately 25-fold; no
clear gender-specific or ethnic differences were found in hepatic levels of CYP2E1 protein or
enzymatic activities associated with CYP2E1 (Shimada et al., 1994). In a study of
interindividual variation in 70 healthy human subjects (40 men and 30 women) given an oral
dose of chlorzoxazone, a therapeutic agent whose metabolism and blood clearance has been
related to CYP2E1 levels, a three- to fourfold range in plasma half-life and clearance values was
observed, with no clear or dramatic age- or gender-specific differences (Kim et al., 1995). A six-
to sevenfold range in chlorzoxazone hydroxylation activity was reported for a group of
69 healthy, smoking and nonsmoking male and female volunteers with mixed ethnic
13 DRAFT - DO NOT CITE OR QUOTE
-------
backgrounds; the range was markedly increased when a group of 72 alcoholic inpatients was
included (Lucas et al., 1999). In studies of human liver microsomes, four- to sixfold ranges in
CYP2E1-dependent oxidation of trichloroethylene have been reported (Lipscomb et al., 2003,
1997). CYP2E1 protein levels in 50 specimens of human lymphocytes from healthy individuals
showed an approximate fivefold range (Bernauer et al., 2000), and a 3.7-fold range in liver
CYP2E1 mRNA levels was reported for a group of 24 patients with chronic hepatitis (Haufroid
et al., 2003). More recently, a threefold range was reported for maximal rates of hepatic
CYP2E1-catalyzed metabolism of dichloromethane, which were estimated with a modified
physiologically based pharmacokinetic (PBPK) model originally developed by Andersen et al.
(1987) and kinetic data (e.g., dichloromethane breath and blood concentrations) for 13 volunteers
(10 males and 3 females) exposed to one or more concentrations of dichloromethane by
inhalation for 7.5 hours (Sweeney et al., 2004). In summary, most studies indicate a three- to
sevenfold variability in CYP2E1 activity, as assessed by various types of measurements, among
"healthy" volunteers. However, various clinical factors (i.e., obesity, alcoholism, use of specific
medications) or co-exposures (i.e., to various solvents) (Lucas et al., 1999) may result in greater
variation and thus the potential for saturation at lower exposures within the general population.
Several genetic polymorphisms for the human CYP2E1 gene have been described, but
clear and consistent correlations with interindividual variation in CYP2E1 protein levels or
associated enzyme activities have not been identified (Ingelman-Sundberg, 2004; Lucas et al.,
2001; Kim et al., 1995; Shimada et al., 1994). The most frequently studied CYP2E1
polymorphisms, Rsal/Pstl, are located in the 5'-flanking region of the gene, and mutations are
thought to lead to increased CYP2E1 protein expression via transcription (Lucas et al., 2001).
Available data indicate that the frequency of this polymorphism, as well as other CYP2E1
polymorphisms, varies among ethnic groups. For example, Stephens et al. (1994) examined
blood samples from 126 African-Americans, 449 European Americans, and 120 Taiwanese
subjects and found frequencies for a rare Rsal allele (C2) of 0.01 in African-Americans, 0.04 in
European Americans, and 0.28 in Taiwanese subjects. In a study of 102 Mexicans, the reported
mutation frequency at the Rsal C2 allele was 0.30 (Mendoza-Cantu et al., 2004).
3.3.2. The GST Pathway
The other major pathway for dichloromethane metabolism involves the conjugation of
dichloromethane to GSH, catalyzed by GST. This results in the formation of a GSH conjugate
that is eventually metabolized to CC>2 (Figure 3-1). The conjugation of dichloromethane to GSH
results in formation of two reactive intermediates that have been proposed to be involved in
dichloromethane toxicity, S-(chloromethyl)glutathione and formaldehyde. In studies with rat,
mouse, and human liver cytosol preparations in the presence of GSH, examination of metabolites
with [13C]-NMR indicated that S-(chloromethyl)glutathione was an intermediate in the pathway
to formaldehyde (Hashmi et al., 1994). Formaldehyde formation from dichloromethane has been
14 DRAFT - DO NOT CITE OR QUOTE
-------
noted in human (Bruhn et al., 1998; Hallier et al., 1994; Hashmi et al., 1994), rat, and mouse
(Casanova et al., 1997; Hashmi et al., 1994) cells in vitro. Formation of a free hydrogen ion is
also hypothesized, although no direct evidence supporting this has been presented.
The GST pathway has approximately a 10-fold lower affinity for dichloromethane than
the CYP pathway (Reitz et al., 1989; Andersen et al., 1987). Although both pathways are
assumed to be operating at all exposures, at lower exposure concentrations the CYP pathway is
expected to predominate, and as exposure concentrations increase, the GST pathway is expected
to gain in relative importance as a dispositional pathway for absorbed dichloromethane. Based
on in vitro studies with liver preparations, the estimated Michaelis-Menten kinetic constant (Km)
values in GST assays with dichloromethane were about 137 mM in a B6C3Fi mouse preparation
and about 44 mM in two human preparations (Reitz et al., 1989). In contrast, estimated Km
values in CYP assays were about 1.8, 1.4, and 2.0 mM in B6C3Fi mouse, F344 rat, and Syrian
golden hamster preparations, respectively. In four human liver preparations, estimated CYP Km
values were about 2.6, 2.0, 0.9, and 2.8 mM (Reitz et al., 1989). A possible resolution of these
apparent in vitro versus in vivo discrepancies is discussed in Section 3.5.5 (in particular, see
Figure 3-6).
Early investigations indicated that in humans, GSTs of the a-, u-, and 7i-classes were not
responsible for the metabolism of dichloromethane (Bogaards et al., 1993). Tissue samples that
metabolized substrates specific to those GST classes did not conjugate dichloromethane to GSH.
Later investigations identified the recently-characterized GST theta class (Meyer et al., 1991),
specifically GST-thetal-1 (GST-T1), as the GST isoenzyme responsible for the metabolism of
dichloromethane (Mainwaring et al., 1996; Blocki et al., 1994). In the absence of the GST-T1
gene, no deoxyribonucleic acid (DNA)-protein cross-links were formed by human liver cells
exposed to dichloromethane (Casanova et al., 1997), and formaldehyde production was not
detected in human erythrocytes (Hallier et al., 1994). In a mouse model with a disrupted
GST-T1 gene, GST activity with dichloromethane in liver and kidney cytosol samples was
substantially lower compared with wild-type GST mice (Fujimoto et al., 2007).
A polymorphism of the GST-T1 gene has been demonstrated in humans. People with
two functional copies of the gene (+/+) readily conjugate GSH to dichloromethane. Individuals
having only one working copy of the gene (+/-) display relatively decreased conjugation ability.
Individuals with no functional copy of the gene (-/-) do not express active GST-T1 protein and
do not metabolize dichloromethane via a GST-related pathway (Thier et al., 1998). Results from
studies of GST-T1 genotypes in human blood samples indicate that average prevalences of the
GST-T1 null (-/-) genotype are higher in Asian ethnic groups (47-64%) than in other groups,
including Caucasians (19-20%), African-Americans (22%), and mixed groups (19%) (Raimondi
et al., 2006; Garte et al., 2001; Nelson et al., 1995) (see Table 3-3). Although information on the
age distribution of study subjects was not generally reported in these analyses, there is little
reason to expect effect modification by age since this is not a gene linked to early mortality.
15 DRAFT - DO NOT CITE OR QUOTE
-------
Based on data collected by Nelson et al. (1995) and ethnicity data from U.S. 2000 census data
(and assuming Hardy-Weinberg equilibrium), Haber et al. (2002) calculated U.S. average
distributions of GST-T1 genotypes as follows: 32% +/+; 48% +/-; and 20% -/-.
Table 3-3. Mean prevalences of the GST-T1 null (-/-) genotype in human
ethnic groups
Ethnic group
Chinese
Korean
Caucasian
Asian
African-American
Mexican American
Other
Reference
Nelson et al. (1995)a
64.4% (n = 45)
60.2% (n= 103)
20.4% (n= 442)
Not reported
21. 8% (n= 119)
9.7%(n=73)
Not reported
Garte et al. (2001)b
Not reported
Not reported
19.7% (n= 5,577)
47.0% (n= 575)
Not reported
Not reported
Not reported
Raimondi et al. (2006)c
Not reported
Not reported
19.0% (n= 6,875)
53.6% (n= 1,727)
Not reported
Not reported
19.4% (n= 1,485)
"Nelson et al. (1995) examined prevalence of the null GST-T1 genotype from analysis of blood samples from
subjects of various ethnicities as noted above.
bGarte et al. (2001) collected GST-T1 genotype data in Caucasian (29 studies; 5,577 subjects) and Asian (3 studies,
575 subjects) ethnic groups; subjects were controls in case-control studies of cancer and various polymorphisms in
genes for bioactivating enzymes.
°Raimondi et al. (2006) collected GST-T1 genotype data from 35 case-control studies of cancer and GST-T1
genotype; data in this table are for control subjects. The "othef group in this study is defined as Latino, African-
American, and mixed ethnicities.
Results from a study of the distribution of activity levels for in vitro conjugation of
dichloromethane with GSH in 22 human liver samples are roughly reflective of these estimates
of the distribution of this polymorphism (Bogaards et al., 1993). No activity was found in
3/22 of the liver samples. Eleven of the samples showed low activity levels (0.21-0.41 nmol
product/minute/mg protein), and eight samples showed high activity levels ranging from 0.82 to
1.23 nmol/minute/mg protein. In another study of seven human subjects, lysates of erythrocytes
showed high activities for producing formaldehyde from dichloromethane (presumably via
GST-T1) in three subjects (15.4, 17.7, and 17.8 nmol product/minute/mg hemoglobin) and lower
activity in the other four subjects (4.3, 6.0, 7.2, and 7.6 nmol product/minute/mg hemoglobin)
(Hallieretal., 1994).
Comparisons of mice, rats, humans, and hamsters for the ability to metabolize
dichloromethane via the GST pathway in liver and lung tissues indicate that mice appear to be
the most active at metabolizing dichloromethane (Sherratt et al., 2002; Thier et al., 1998;
Casanova et al., 1997, 1996; Hashmi et al., 1994; Reitz et al., 1989). Reitz et al. (1989) reported
mean (± SD) GST enzymatic activity levels with dichloromethane as substrate (in units of nmol
product formed/minute/mg protein) in liver cytosol preparations to be: 25.9 ± 4.2 units in
B6C3Fi mice (n = 15 determinations per preparation); 7.05 ±1.7 nmol/minute/mg in F344 rats
16
DRAFT - DO NOT CITE OR QUOTE
-------
(n = 6); and 1.27 ± 0.21 nmol/minute/mg in Syrian golden hamsters (n = 6). Mean GST activity
levels in liver preparations from four human subjects (accident victims screened for human
immunodeficiency virus and hepatitis B and C and obtained through a transplant center) were
2.62 ± 0.44 (n = 10), -0.01 ± 0.04 (n = 6), 2.71 ± 0.45 (n = 6), and 3.03 ± 0.44 nmol/minute/mg
(n = 6) (Reitz et al., 1989). The finding that one of the four individuals was unable to conjugate
dichloromethane with GST was reflective of the estimated frequency of the GST-T1 null
genotype in the U.S. population (approximately 20% in Caucasians and African-Americans see
Table 3-3 and Haber et al., 2002). Mean GST activity levels in lung cytosol preparations showed
a similar rank order among species: 7.3 ± 1.4 nmol/minute/mg in mice (n = 4), 1.0 ± 0.1
nmol/minute/mg in rats (n = 4), 0.0 ± 0.2 nmol/minute/mg in hamsters (n = 4), and 0.37 ± 0.25
nmol/minute/mg in a pooled lung preparation from humans (n = 2).
Thier et al. (1998) conducted a study evaluating the activity of GST-T1 after treatment of
dichloromethane in the cytosol of liver and kidney homogenates from hamsters (pooled male and
females), rats (pooled male and female), male mice, and female mice and for humans classified
as nonconjugators, low conjugators, or high conjugators of GST to dichloromethane. Little
information is provided about the human samples other than that 13 kidney cancer patients were
the source of the kidney samples; normal tissue identified by pathological exam was used. Blood
samples from 10 of these patients were collected, and enzyme activities measured in erythrocytes
from 9 of these samples were reported. Results of conjugation of dichloromethane to GSH from
these studies are presented in Table 3-4. As can be seen from the table, activity levels (expressed
as nmol/minute per mg of cytosolic protein) of humans varied considerably, with nonconjugators
(presumed to be GST-Tl"7") having no detectable activity, low conjugators (presumed to be
GST-T1+") having moderate activity, and high conjugators (presumed to be GST-T1+ +) having
approximately twice the activity seen in low conjugators. In the liver, the activity of rat GST
conjugation was over twofold that seen in human high conjugators, while levels in mice were
>11-fold (males) or 18-fold (females) greater than those of human high conjugators. In the
kidney, the activity of high-conjugator humans was approximately 1.8-fold that of rats and was
comparable to the activity of both male and female mice. The data in Table 3-4 show the
following order for GST-T1 activities with dichloromethane as substrate: in liver preparations,
mouse » rat > human high conjugators > human low conjugators > hamster > human
nonconjugators and, in kidney preparations, female mouse ~ male mouse ~ human high
conjugators > rat ~ human low conjugators > hamster > human nonconjugators. In addition, the
data indicate that activity levels in liver, kidney, and erythrocytes of human subjects are in
correspondence with the nonconjugator, low conjugator, and high conjugator designations.
17 DRAFT - DO NOT CITE OR QUOTE
-------
Table 3-4. GST-T1 enzyme activities toward dichloromethane in human,
rat, mouse, and hamster tissues (liver, kidney, and erythrocytes)
Human, nonconjugators
Human, low conjugators
Human, high conjugators
Rat
Mouse, male
Mouse, female
Hamster
Activity (nmol/min per mg protein)3
Liver
Not detectable (2)
0.62 ±0.30 (11)
1.60 ±0.48 (12)
3.71 ±0.28 (8)
18.2 ±2.22 (5)
29.7 ±6.3 1(5)
0.27 ± 0.20 (6)
Kidney
Not detectable (1)
1.38 ±0.52 (8)
3.05 ±0.72 (4)
1.71 ±0.28 (8)
3. 19 ±0.46 (5)
3. 88 ±0.90 (5)
0.25 ±0.21 (6)
Activity (nmol/min per mL)a
Erythrocytes
Not detectable (1)
9.67 ± 2.49 (5)
18.28 ±0.46 (3)
Not measured
Not measured
Not measured
Not measured
aMean ± SD with number of samples noted in parentheses.
Source: Adapted from Thier et al. (1998).
Sherratt et al. (2002) reported that, on a per mg basis, native recombinant mouse GST-T1
(purified after expression in Escherichia coif) was approximately twofold more active toward
dichloromethane than native recombinant human enzyme, as well as being approximately
fivefold more efficient (as assessed by the ratio of kcat/Km), where kcat is the maximum rate of the
reaction catalyzed by the enzyme per enzyme molecule; i.e., Vmax/Et where Et is the total enzyme
concentration).
The distribution of GST-T1 in human tissues has been examined with antibodies raised
against recombinant human GST-T1 (Sherratt et al., 2002, 1997). Immunoblotting of sodium
dodecyl sulfate polyacrylamide gel electrophoresis gels loaded with tissue extracts from a
73-year-old man who had died with brochopneumonia and atherosclerosis indicated the
following order of expression of GST-T1: liver ~ kidney > prostate ~ small intestine > cerebrum
~ pancreas ~ skeletal muscle > lung ~ spleen ~ heart ~ testis (Sherratt et al., 1997). It was
estimated that the levels of cross-reacting materials in the cerebrum, pancreas, or skeletal muscle
extracts were about 10% of those in the liver, whereas levels in the lung, spleen, heart, and testis
were <5% of the levels in the liver. Comparison of the amounts of cross-reacting material in
soluble liver extracts from a B6C3Fi mouse and five human subjects (i.e., normal liver tissue
samples from biopsies of secondary liver tumors) found that levels of GST-T1 protein were
higher in the mouse extracts than in any of the human liver extracts (Sherratt et al., 2002).
Densitometer analysis indicated that the GST-T1 level in the mouse liver extract was about
fivefold higher than those in human liver extracts displaying the highest level. Cross-reacting
material was not detectable in liver extracts from one of the five human subjects, indicating that
this individual may have been GST-T1 null (Sherratt et al., 2002).
Results from in situ hybridization with oligonucleotide antisense probes for GST-T1
mRNA levels and immunohistochemical studies with antibodies to GST-T1 have indicated that
18
DRAFT - DO NOT CITE OR QUOTE
-------
there may be subtle differences between mice and humans in the intracellular localization of
GST-T1 in the liver. Mainwaring et al. (1996) reported that staining for GST-T1 mRNA was
higher in liver slices from B6C3Fi mice than in liver slices from F344 rats and that staining in
human liver samples was very low. Although the number of mouse and rat liver samples
examined in this study was not indicated in the available report, it was reported that slices from
five human liver samples were examined. No information was provided regarding the clinical
history of the sources of the human samples. In mouse liver, staining for GST-T1 mRNA was
enhanced in the limiting plate hepatocytes, in nuclei, in bile-duct epithelial cells, and in lesser
amounts in the centrilobular cells in general. In rat liver, a similar pattern was observed, except
no enhanced staining was observed in the limiting plate hepatocytes or in nuclei. Staining for
GST-T1 mRNA in the human liver samples showed an even distribution throughout the liver
lobule, and no mention of a specific nuclear localization was made (Mainwaring et al., 1996).
Quondamatteo et al. (1998), using antibodies to GST-T1, subsequently reported a similar
localization of GST-T1 protein in nuclei of cells in mouse liver slices. In another study using
antibodies raised against recombinant human GST-T1 or a peptide derived from the deduced
mouse GST-T1 primary sequence, Sherratt et al. (2002) reported that nuclear staining was
observed in all cells in mouse liver slices (from five individual B6C3Fi mice) showing the
presence of mouse GST-T1; staining in the cytoplasm was only detected in cells with very high
levels of GST-T1. In liver slices obtained from two human subjects (males, ages 60 and
61 years, with a secondary liver tumor and what was described as a "cavernous hemangioma"
without malignancy, respectively), the most intense nuclear staining was associated with bile
duct epithelial cells, but there was heterogeneity of staining within hepatocytes; some cells
showed nuclear staining, but others only exhibited cytoplasmic staining (Sherratt et al., 2002).
In summary, the relative amount of dichloromethane metabolized via the GST pathway
increases with increasing exposure concentrations. As the high affinity CYP pathway becomes
saturated (either from high exposure levels of genetic or other factors that decrease CYP2E1
activity), the GST pathway increases in relative importance as a dispositional pathway for
dichloromethane. Two reactive metabolites (S-(chloromethyl)glutathione and formaldehyde)
resulting from this pathway have been identified. GST-T1 is the GST isozyme that catalyzes
conjugation of dichloromethane with GST. Interindividual variation in the ability to metabolize
dichloromethane via GST-T1 is associated with genetic polymorphisms in humans. Estimated
U.S. population prevalence of nonconjugators (-/- at the GST-T1 locus) is about 20%, but higher
prevalences (47-64%) have been reported for Asians (Raimondi et al., 2006; Haber et al., 2002;
Garte et al., 2001; Nelson et al., 1995). The prevalences for low (+/- at the GST-T1 locus) and
high (+/+) conjugators have been estimated at 48 and 32%, respectively (Haber et al., 2002).
The liver and kidney are the most enriched tissues in GST-T1, but evidence is available for the
presence of GST-T1 in other tissues at lower levels, including the brain and lung. In humans,
GST-T1 expression in the brain is lower than that seen in the liver or kidney but higher than in
19 DRAFT - DO NOT CITE OR QUOTE
-------
the lung. Comparisons of mice, rats, humans, and hamsters for the ability to metabolize
dichloromethane via the GST pathway in liver (based on measurement of tissue-specific enzyme
activity) indicate the following rank order: mice > rats > or ~ humans > hamsters. In mouse
liver tissue, GST-T1 appears to be localized in the nuclei of hepatocytes and bile-duct
epithelium, but rat liver does not show preferential nuclear localization of GST-T1. In human
liver tissue, some hepatocytes show nuclear localization of GST-T1 and others show localization
in cytoplasm, as well as in bile duct epithelial cells. The apparent species differences in
intracellular localization of GST-T1 may play a role in species differences in susceptibility to
dichloromethane carcinogenicity if nuclear production of S-(chloromethyl)glutathione is more
likely to lead to DNA alkylation than cytoplasmic production.
3.4. ELIMINATION
Dichloromethane is eliminated mainly through exhalation either of the parent compound
or as the two primary metabolites CC>2 and CO (Angelo et al., 1986a, b; McKenna et al., 1982;
DiVincenzo and Kaplan, 1981; DiVincenzo et al., 1972, 1971). In human studies,
dichloromethane is rapidly eliminated from the body following the cessation of exposure, with
much of the parent compound completely removed from the bloodstream and expired air by
5 hours postexposure in experiments using exposure levels of 90, 100, or 210 ppm (DiVincenzo
et al., 1972, 1971; Riley et al., 1966). Studies in rats have similarly demonstrated that
elimination from the blood is rapid, with elimination half-times in F344 rats on the order of 4-
6 minutes following intravenous doses in the range of 10-50 mg/kg (Angelo et al., 1986a). In a
study using Sprague-Dawley rats, Carlsson and Hultengren (1975) demonstrated variability in
elimination rates between different types of tissues, with the most rapid elimination seen in the
adipose and brain tissue, while elimination from liver, kidneys, and adrenals proceeded more
slowly.
In a study using human volunteers, DiVincenzo and Kaplan (1981) reported a dose-
related increase in CO in the expired breath after inhalation exposure to 50-200 ppm of
dichloromethane, with a net elimination as CO on the order of 25-35% of the absorbed dose.
Similar results have been reported in animal studies. Following gavage administration of 50 or
200 mg/kg-day doses of [14C]-labeled dichloromethane in water to groups of six mature male
F344 rats for up to 14 days, >90% of the label was recovered in the expired air within 24 hours
of dose administration (Angelo et al., 1986b). Following administration of the first of 14 daily
50 mg/kg-day doses, radioactivity in parent compound, CO2, and CO in the 24-hour expired
breath accounted for 66, 17, and 16% of the administered radioactivity, respectively; similar
patterns were reported for 24-hour periods following administration of the seventh and
fourteenth 50 mg/kg-day dose. Following administration of the first 200 mg/kg-day dose,
radioactivity in parent compound, CO2, and CO in the 24-hour expired breath accounted for 77,
9, and 6%, respectively, of the administered radioactivity (Angelo et al., 1986b). In mature, male
20 DRAFT - DO NOT CITE OR QUOTE
-------
Sprague-Dawley rats given a smaller dose (1 mg/kg) of [14C]-labeled dichloromethane,
radioactivity in parent compound, CO2, and CO in 48-hour expired breath accounted for 12, 35,
and 31%, respectively; these data indicate that, at lower dose levels, a greater percentage of the
administered dose was metabolized by the CYP pathway and eliminated in the expired breath,
compared with higher dose levels (McKenna and Zempel, 1981). Similar patterns of
radioactivity distribution in parent compound, CC>2, and CO in expired breath were found in
mature, male B6C3Fi mice following gavage administration of 50 mg/kg-day (in water) or
500 or 1,000 mg/kg-day (in corn oil) [14C]-labeled dichloromethane (Angelo et al., 1986a). For
example, radioactivity in parent compound, CO2, and CO in 24-hour expired breath accounted
for 61, 18, and 11% of the administered radioactivity, following administration of a single
50 mg/kg dose to a group of six mice (Angelo et al., 1986a). Exhalation rates were similarly
high following inhalation exposure of mature, male Sprague-Dawley rats (>90%) (McKenna et
al., 1982) or following intravenous administration of dichloromethane to mature, male F344 rats
(Angelo et al., 1986b).
Elimination of dichloromethane in the urine of exposed humans is generally small, with
total urinary dichloromethane levels on the order of 20-25 or 65-100 ug in 24 hours following a
2-hour inhalation exposure to 100 or 200 ppm, respectively (DiVincenzo et al., 1972). However,
a direct correlation between urinary dichloromethane and dichloromethane exposure levels was
found in volunteers, despite the comparatively small urinary elimination (Sakai et al., 2002).
Following administration of a labeled dose in animals, regardless of exposure route, generally
<5-8% of the label is found in the urine and <2% in the feces (McKenna et al., 1982; McKenna
and Zempel, 1981; DiVincenzo et al., 1972, 1971).
3.5. PHYSIOLOGICALLY BASED PHARMACOKINETIC MODELS
Several PBPK models for dichloromethane in animals and humans have been developed
from 1986 to 2006. These models are mathematical representations of the body and its
absorption, distribution, metabolism, and elimination of dichloromethane and select metabolites,
based on the structure of the Ramsey and Andersen (1984) model for styrene. The models'
equations are designed to mimic actual biological behavior of dichloromethane, incorporating in
vitro and in vivo data to define physiological and metabolic equation parameters. As such, the
models can simulate animal or human dichloromethane exposures and predict a variety of
dichloromethane and metabolite internal dosimeters (i.e., instantaneous blood and tissue
concentration, area under the curve [AUC] of concentration versus time plots, rate of metabolite
formation), allowing for the extrapolation of toxicity data across species, route of exposure, and
high to low exposure levels. The development of dichloromethane PBPK models has resulted in
either increased biological detail and functionality or refinement of model parameters with newly
available data. The former type of development provides more options for toxicity data
extrapolation, while the latter serves to increase confidence in model predictions and decrease
21 DRAFT - DO NOT CITE OR QUOTE
-------
uncertainty in risk assessments for which the models were, or will be, applied. This section of
the document describes each of the models reported in the scientific literature and/or used by the
regulatory community (i.e., Occupational Safety and Health Administration [OSHA], EPA) and
their contribution to the advancement of predictive dosimetry and data extrapolation for
dichloromethane. In some instances, model development was accomplished by the addition of
new biological compartments (e.g., tissue systems). Diagrams of the compartmental structure of
the models are shown in Figure 3-2. Significant statistical advances in parameter estimation also
have been incorporated in model development. For this reason, some animal and human PBPK
models may be described as deterministic (Sweeney et al., 2004; Casanova et al., 1996; Reitz et
al., 1988a, b; U.S. EPA, 1988b, 1987a, b; Andersen et al., 1987; Gargas et al., 1986) in which
point estimates for each model parameter are used, resulting in point estimates for dosimetry.
Others may be described as probabilistic (Jonsson and Johanson, 2001; El-Masri et al., 1999;
OSHA, 1997), in which probability distributions for each parameter were defined, resulting in
probability distributions for dosimetry. The latter approach, particularly utilizing a Bayesian
hierarchical statistical model structure (described below) (David et al., 2006; Marino et al., 2006)
to estimate parameter values, allows for the introduction of intra- and interspecies variability into
model predictions and quantitative assessment of model uncertainty. Both deterministic (U.S.
EPA, 1988b, 1987a, b) and probabilistic (OSHA, 1997) applications have been used to develop
regulatory values. As discussed below, subsequent applications of the developed models for
cancer risk assessment have resulted in significantly different estimates of human cancer risk.
22 DRAFT - DO NOT CITE OR QUOTE
-------
© i t
Gas
Exchange
ST •« ' ' > CYP —
1 Gsrnc
ST< 1 I > CYP
i t Gsinr
1 Gsinn
GST 4 '11 > CYF
Formaldehyde
I
DNA-protein crosslinks
1 Gsinn
Models C-G all build on the structure in model B. Models E and G have been applied in humans;
all others have been applied in humans and rodents (mice and/or rats). CYP = CYP pathway
metabolites; GST = GST pathway metabolites.
Adapted from: Model A—Gargas et al. (1986); B—Andersen et al. (1987);
C—Andersen et al. (1991); D—Casanova et al. (1996); E—Sweeney et al. (2004); F—OSHA
(1997); G—Jonsson and Johanson (2001).
Figure 3-2. Schematics of PBPK models (1986-2006) used in the
development of estimates for dichloromethane internal dosimetry.
23
DRAFT - DO NOT CITE OR QUOTE
-------
The deterministic rat model of Gargas et al. (1986), based on previous work by Ramsey
and Andersen (1984) examining inhalation pharmacokinetics of styrene in rats, was the first
PBPK model for dichloromethane. It comprised four compartments (fat, liver, richly perfused
tissues, and slowly perfused tissues [Figure 3-2A]) and described flows and partitioning of parent
material and metabolites through the compartments with differential equations. Metabolism,
which was restricted to the liver compartment, was described as two competing pathways: the
GST pathway, described with a linear first-order kinetic model, and the CYP pathway, described
with a saturable Michaelis-Menten kinetic model. Rate constants for the CYP and GST
pathways in rats were determined by optimization of the model with in vivo gas uptake data.
COHb production was modeled both endogenously and from CYP-mediated metabolism of
dichloromethane. This model demonstrated the dose-dependent flux through the competing CYP
and GST metabolic pathways and the effect of CYP inhibition on COHb generation.
Andersen et al. (1987) extended the rat model of Gargas et al. (1986) to include a lung
compartment, including CYP and GST metabolism pathways within the lung, in rats, mice,
hamsters, and humans (Figure 3-2B). Physiological flow rates were allometrically scaled among
species by 3/4 power of body weight (BW). Rate constants for the CYP and GST pathways in
rodents were determined by optimization of the model with in vivo gas uptake data. CYP rate
constants for humans were derived from data on dichloromethane uptake in human subjects
(number of subjects not reported). Human GST rate constants were derived by allometric
scaling of the animal GST rate constants. Model predictions compared favorably with kinetic
data for human subjects exposed by inhalation to dichloromethane (Andersen et al., 1987).
Using the mouse cancer bioassay data from NTP (1986), Andersen et al. (1987) compared the
linear body surface area-derived or the PBPK model-derived human liver and lung dose
surrogates associated with tumor development (mg dichloromethane metabolized via GST
pathway/volume tissue/day). They reported that PBPK model-extrapolated human liver and lung
internal doses were 167- and 144-fold lower for inhalation exposure and 45- and 213-fold lower
for drinking water exposure, respectively, than body surface area scaled internal doses. The
study authors suggested that the lower model-predicted human internal dose surrogates were due
to the need to saturate the CYP pathway before appreciable tumorigenic metabolite levels could
be attained, which is not captured by extrapolation based on body surface area.
U.S. EPA (1988b, 1987a, b) slightly modified the Andersen et al. (1987) model for mice
by using different alveolar ventilation and cardiac flow rates and used the mouse and human
models to derive human cancer risks from animal tumor incidence data. The flow rate
parameters in the Andersen et al. (1987) model were based on a human breathing rate of
12.5 m3/day (reflecting a resting rate), compared with the EPA value of 20 m3/day (reflecting
average daily activity level), and a mouse breathing rate of 0.084 m3/day (based on allometric
scaling of bioassay-specific BWs), compared with the rate commonly used by EPA,
24 DRAFT - DO NOT CITE OR QUOTE
-------
0.043 m3/day (U.S. EPA, 1987a). The internal dose metric used in the applications of the model
to cancer risk assessment was reflective of the amount of dichloromethane metabolized by the
GST pathway. In addition to using the mouse and human PBPK models to account for species
differences in dosimetry, a body surface area correction factor of 12.7 was applied to low-dose
slopes of estimated dose-response relationships for liver and lung tumors in mice to account for
presumed higher human responsiveness, relative to mice, to dichloromethane-induced cancer
(U.S. EPA, 1987a). The factor of 12.7 is the cube root of the ratio of human to mouse reference
BWs; this BW scaling factor was applied to adjust for differences in the lifetime impact in mice
and humans resulting from both processes leading to differences in internal doses (e.g., clearance
of a given daily amount of dichloromethane metabolically activated per liter of tissue) and
differences in pharmacodynamics or response (Rhomberg, 1995). A human cancer IUR of
4.7 x 10"7 per (jig/m3), based on this analysis, was placed on IRIS in September 1990.
The Andersen et al. (1987) models were also modified by addition of submodel structures
for estimation of new dosimeters of interest. Andersen et al. (1991) added the capability to
specifically describe the kinetics of dichloromethane, CO, and COHb in rats and humans with
the addition of the Coburn-Forster-Kane equation to describe CO and COHb kinetics
(Figure 3-2C). However, equations were not added for metabolism of dichloromethane to CO in
the lung. Casanova et al. (1996) extended the Andersen et al. (1987) mouse model to include a
submodel that predicted the formation of formaldehyde and DNA-protein cross-links in the liver
(Figure 3-2D).
Further refinements of the Andersen et al. (1987) models allowed for incorporation of
new data. New in vitro measurements of metabolic rate constants in human and animal tissues
were incorporated into the Andersen et al. (1987) models by Reitz and coworkers (Reitz, 1991;
Reitz et al., 1988a, b). Sweeney et al. (2004) modified the Andersen et al. (1987) human PBPK
model, adding extrahepatic CYP metabolism in richly perfused tissues (Figure 3-2E) to obtain a
better fit of the model to kinetics data for humans. Data for 13 volunteers (10 men and
3 women) who were exposed to one or more concentrations of dichloromethane for 7.5 hours
included dichloromethane concentrations in breath and blood, COHb concentrations in blood,
and CO concentrations in exhaled breath. Individual CYP maximal velocity (Vmaxc) values were
obtained by optimizing model predictions to match time-course data simultaneously for
dichloromethane concentrations in blood and exhaled breath for each individual. Resultant
individual values of CYP Vmaxc ranged from 7.4 to 23.6 mg/hour/kg0'7, indicating an
approximate threefold range in maximal CYP metabolic activity.
The significance of metabolic variability for the kinetics of dichloromethane in animals
and humans was explored by several investigators using PBPK models. Dankovic and Bailer
(1994) used the updated human model presented by Reitz et al. (1988a, b) to explore the
consequences of interindividual variability for in vitro kinetic constants with the CYP and GST
pathways (based on data for four human subjects) and reported that predicted GST-metabolized
25 DRAFT - DO NOT CITE OR QUOTE
-------
doses to the lung and liver could range from about zero to up to fivefold greater than those
predicted with the values of these rate constants used in the Reitz et al. (1988a, b) model.
El-Masri et al. (1999) replaced parameter estimates in the mouse and human PBPK models
presented by Casanova et al. (1996) with probability distributions, including published
information on the distribution of GST-T1 polymorphism in human populations, and used Monte
Carlo simulations to estimate distributions of cancer potency of dichloromethane in mice,
distributions of the amount of DNA-protein cross-links formed in the liver of humans, and
distributions of human cancer risks at given exposure levels of dichloromethane. The analysis
showed that, at exposure levels of 1, 10, 100, and 1,000 ppm dichloromethane, average and
median cancer risk estimates were 23-30% higher when GST-T1 polymorphism was not
included in the model.
Given the demonstrated influence of population variability in dichloromethane
metabolism on PBPK model-derived cancer risk estimates (El-Masri et al., 1999; Dankovic and
Bailer, 1994), PBPK model development has included a more formal statistical treatment of data
for physiological and metabolic variability. Bayesian statistical approaches have been applied to
develop probabilistic PBPK models for dichloromethane. Probabilistic models account for
variability between individuals in model parameters by replacing point estimates for the model
parameters with probability distributions. Calibration or fitting of probabilistic PBPK models to
experimental toxicokinetic data is facilitated by a Bayesian technique called Markov Chain
Monte Carlo (MCMC) simulation, which quantitatively addresses both variability and
uncertainty in PBPK modeling (Jonsson and Johanson, 2003).
OSHA (1997) used MCMC simulation to fit probabilistic versions of the Reitz et al.
(1988a, b) and Andersen et al. (1991, 1987) mouse and human models, which included
probability distributions for all model parameters. GST- and CYP-mediated metabolism
occurred in the liver and lung compartments (see Figure 3-2F). The model parameters were
modified to focus on occupational exposure scenarios; that is, a parameter distribution for work
intensity (using data from Astrand et al. [1975]) was added, which adjusted physiological flow
rates as a function of work intensity as measured in watts. In addition, updated measurements of
blood:air and tissue:air partition coefficients (Clewell et al., 1993) were used to describe
distributions for these parameters. The Clewell et al. (1993) blood:air partition coefficient (PB)
of 23 for mice is higher than the value of 8.29 reported by Andersen et al. (1987) and used by
EPA (1988b, 1987a, b). The newer Clewell et al. (1993) value for mice is the preferred value,
since it is much closer to the values for rats (19.4) and hamsters (22.5) rather than humans (9.7),
as reported by Andersen et al. (1987). Distributions of metabolic, physiological, and partitioning
parameters in the mouse and human models were updated by using Bayesian methods with data
for mice and humans in published studies of mouse and human physiology and dichloromethane
kinetic behavior.
26 DRAFT - DO NOT CITE OR QUOTE
-------
Jonsson et al. (2001) used additional human kinetics data to expand the PBPK model of
Reitz et al. (1988a, b) and added new model compartments (Figure 3-2G). These investigators
used MCMC simulation to develop a probabilistic model from the Reitz et al. (1988a, b) human
model by using published in vitro measurements of liver Vmaxc for the CYP pathway (Reitz et
al., 1989) and kinetic data for five human subjects exposed by inhalation to dichloromethane
(Astrand et al., 1975). A working muscle compartment was added to the basic Andersen et al.
(1987) and Reitz et al. (1988a, b) structure (see Figure 3-2G). Jonsson and Johanson (2001)
refined and extended this probabilistic model by including an additional fat compartment (to
provide a better description of the experimental data for the time course of dichlorom ethane in
subcutaneous fat), incorporating (with MCMC simulation) kinetic data for dichloromethane in an
additional 21 human subjects and including three GST-T1 genotypes/phenotypes
(nonconjugators -/-, low conjugators +/-, high conjugators +/+). Monte Carlo simulations were
then used with the refined probabilistic model to predict human liver cancer risk estimates at
several dichloromethane exposure levels using an algorithm similar to the one used by El-Masri
et al. (1999), using DNA-protein cross-links as the internal dose metric. The mean, 50*, 90* ,
and 95* percentile human cancer risk values from Jonsson et al. (2001) and El-Masri et al.
(1999) were very similar, within onefold of one another for simulated exposure levels up to
100 ppm.
The most statistically rigorous and data-intensive PBPK model development was
performed by Marino et al. (2006) for mice and David et al. (2006) for humans. Development of
these models used multiple mouse and human data sets in a Bayesian hierarchical statistical
structure to quantitatively capture population variability and reduce uncertainty in model
dosimetry and the resulting risk values. EPA used these models in the derivation of reference
values and cancer risk estimates in the current assessment, and these models are described in
more detail below.
3.5.1. Probabilistic Mouse PBPK Dichloromethane Model (Marino et al., 2006)
Marino et al. (2006) used MCMC analysis to develop a probabilistic PBPK model for
dichloromethane in mice, using the Andersen et al. (1987) model structure as a starting point
(Figure 3-3). Metabolic kinetic parameters (Vmaxc, Km, kfc, ratio of lung Vmax to liver Vmax [Al],
and ratio of lung kfc to liver kfc [A2]) (Table 3-5) were calibrated with this Bayesian
methodology by using several experimental data sets. Distribution parameters (i.e., means and
coefficients of variation [CVs]) for other physiological parameters (i.e., BW, fractional flow
rates, and fractional tissue volumes) and partition coefficients were taken from the general
literature as noted by Clewell et al. (1993). Marino et al. (2006) noted that using distributions for
these latter parameters from the general literature (based on a large number of animals) was
better than updating them based on the relatively smaller number of animals in the available
dichloromethane kinetic studies. Clewell et al. (1993) determined blood:air and tissue:air
27 DRAFT - DO NOT CITE OR QUOTE
-------
partition coefficients (means and CVs) with tissues from groups of male and female B6C3Fi
mice. These partition coefficients were derived by using a vial equilibration method similar to
that used by prior investigators (Andersen et al., 1987; Gargas et al., 1986). Tissue:air partition
coefficients were approximately 2-3 times lower than previously utilized values with the
exception of the liver coefficient, which was similar to previous values (Table 3-5). The PB (23)
from Clewell et al. (1993) is higher than the previously reported value of 8.3 (Gargas et al.,
1986). The higher value is more in line with values measured in rats (19.4) and hamsters (22.5)
and, thus, is more reasonable than the older value of 8.3. Table 3-5 shows mean and CVs for
physiological parameters and partition coefficients in the Marino et al. (2006) mouse model as
well as values used in earlier deterministic PBPK mouse models for dichloromethane.
i t
Gas
exchange
nsT <
Fat
Richly
perfused
Slowly
perfused
Liver
Lung
fc. HYP
CO sub
model
Endogenous
production
Figure 3-3. Schematic of mouse PBPK model used by Marino et al. (2006).
28
DRAFT - DO NOT CITE OR QUOTE
-------
Table 3-5. Values for parameter distributions in a B6C3Fi mouse probabilistic PBPK model for
dichloromethane compared with associated values for point parameters in earlier deterministic B6C3Fi mouse
PBPK models for dichloromethane
Parameter
Fractional flow rates (fraction ofQCC)0
QFC Fat
QLC Liver
QRC Rapidly perfused tissues
QSC Slowly perfused tissues
Fractional tissue volumes (fraction ofBW)°
VFC Fat
VLC Liver
VLuC Lung
VRC Rapidly perfused tissues
VSC Slowly perfused tissues
Partition coefficients'^
PB Blood:air
PF Fatblood
PL Liverblood
PLu Lung:blood
PR Rapidly perfused:blood
PS Slowly perfused:blood
Flow rates
QCC Cardiac output (L/hr/kg° 74)
VPR ventilation:perfusion ratio
Metabolism parameters
VmaxC Maximum CYP metabolic rate (mg/hr/kg0 7)
Km CYP affinity (mg/L)
kfc First-order GST metabolic rate constant (kg0 3/hr)
Al Ratio of lung VmaxC to liver VmaxC
A2 Ratio of lung kfc to liver kfc
Marino et al. (2006)a
Prior mean
0.05
0.24
0.52
0.19
0.04
0.04
0.0115
0.05
0.78
23
5.1
1.6
0.46
0.52
0.44
28.0
1.52
11.1
0.396
1.46
0.462
0.322
Prior CV
0.60
0.96
0.50
0.40
0.30
0.06
0.27
0.30
0.30
0.15
0.30
0.20
0.27
0.20
0.20
0.58
0.75
2
2
2
0.55
0.55
Final posterior
mean
Final posterior
CV
These parameters were taken from an
extensive literature database derived from a
large number of animals; therefore, further
Bayesian updating does not inform on the
true mean and variance for these values.
24.2
1.45
9.27
0.574
1.41
0.207
0.196
0.19
0.20
0.21
0.42
0.28
0.36
0.37
EPA (1988b,
1987a, b)
0.05
0.24
0.52
0.19
0.04
0.04
0.0119
0.05
0.78
8.29
14.5
1.71
1.71
1.71
0.96
14.3d
1.0
11.1
0.396
1.46
0.416
0.137
Andersen
etal.
(1987)
0.05
0.24
0.52
0.19
0.04
0.04
0.0119
0.05
0.78
8.29
14.5
1.71
1.71
1.71
0.96
28.0e
1.0
11.1
0.396
1.46
0.416
0.137
aMCMC analysis was used to update prior distributions (means and CVs) for flow rate and metabolic parameters in a sequential process with three sets of kinetic
data from mouse studies, as explained further in the text. Final values for posterior distributions are given in this table.
bSource: Andersen et al. (1991, 1987).
"Source: Clewelletal. (1993).
dBased on a mouse breathing rate of 0.043 mVd.
eBased on a mouse breathing rate of 0.084 mVd.
29
DRAFT - DO NOT CITE OR QUOTE
-------
The Bayesian calibration of the cardiac output (QCC) constant, ventilation:perfusion ratio
(VPR), and metabolic parameters was divided into three sequential steps: using kinetic data
from closed chamber studies with mice treated with an inhibitor of CYP2E1 (trans-l,2-dichloro-
ethylene) in order to minimize the oxidative pathway and enable a more precise estimate of
parameters for the GST pathway, followed by kinetic data for mice given intravenous injections
of dichloromethane to estimate metabolism parameters in the absence of pulmonary absorption
processes and, finally, kinetic data for naive mice exposed to dichloromethane in closed
chambers (Marino et al., 2006). The initial prior distributions were based on mean values used
by Andersen et al. (1987) for the metabolic parameters and by OSHA (1997) for the parameters
for VPR, Al, and A2. Posterior distributions from the first Bayesian analysis were used as prior
distributions for the second step, and posterior distributions from the second step were used as
prior distributions for the final updating. Final results from the Bayesian calibration of the
mouse probabilistic model are shown in Table 3-5.
Marino et al. (2006) used the Bayesian-calibrated mouse model to calculate internal dose
metrics associated with exposure conditions in the NTP (1986) B6C3Fi mouse cancer inhalation
bioassay. The internal dose metric selected was mg dichloromethane metabolized by the GST
pathway per liter tissue per day. This is the same dose metric used in earlier applications of
PBPK models to derive human cancer IUR estimates based on cancer responses in mice (OSHA,
1997; Andersen et al., 1987; U.S. EPA, 1987a, b). Its use is consistent with evidence that
dichloromethane metabolism via GST-T1 results in the formation of a reactive metabolite that
damages DNA and results in the formation of tumors (see Section 4.7). The model was used to
calculate values for this internal dose metric in the lung and liver of mice in the NTP (1986)
study, using the mean values of the final distributions for the parameters in the model. Resultant
values were three- to fourfold higher than values calculated with the Andersen et al. (1987) and
U.S. EPA (1987a, b) versions of the model (Table 3-6). Marino et al. (2006) noted that the
difference could be primarily attributed to the changes in the partition coefficients based on
Clewell et al. (1993) as well as to the Bayesian updating of the metabolic parameters (see
Table 3-5).
30 DRAFT - DO NOT CITE OR QUOTE
-------
Table 3-6. Internal daily doses for B6C3Fi mice exposed to dichloromethane
for 2 years (6 hours/day, 5 days/week) calculated with different PBPK
models
Target organ
Liverb
Lungb
NTP (1986)
exposure level"
Control
2,000 ppm
4,000 ppm
Control
2,000 ppm
4,000 ppm
PBPK model
Marino et al. (2006)
0
2,359.99
4,869.85
0
474.991
973.343
EPA (1987a, b)
0
727.8
1,670
0
111.4
243.7
Andersen et al. (1987)
0
851
1,811
0
123
256
a2,000 ppm = 6,947 mg/m3; 4,000 ppm = 13,894 mg/m3.
blnternal dose expressed as mg dichloromethane metabolized by the GST pathway per liter tissue per d.
Marino et al. (2006) noted that inclusion of extrahepatic CYP metabolism in the slowly
perfused tissue compartment in the mouse model had little impact on the formation of GST
metabolites in the liver and lung, especially at exposure levels used in the mouse NTP (1986)
bioassay. To support this contention, the Andersen et al. (1987) model was modified to include
10% of the liver rate of oxidative metabolism in the slowly perfused tissue compartment (as
suggested by Sweeney et al. [2004]), and the modified model was used to calculate the formation
of GST metabolites. If extrahepatic metabolism was included in the slowly perfused tissue
compartment, there was a 5-6% reduction in the formation of GST metabolites in the lung and
liver at an exposure level of 50 ppm. At 2,000 or 4,000 ppm, however, there was only a 0.77 or
0.37% reduction, respectively. Marino et al. (2006) did not discuss the impact of including
extrahepatic metabolism in the rapidly perfused tissue compartment; the same group of
investigators developed a human PBPK model that included CYP metabolism in the richly
perfused compartment (David et al., 2006).
3.5.2. Probabilistic Human PBPK Dichloromethane Model (David et al., 2006)
The basic model structure used by David et al. (2006) was that of Andersen et al. (1987)
with the addition of the CO submodel of Andersen et al. (1991), refinements from the Marino et
al. (2006) mouse model, and an inclusion of CYP metabolism in richly perfused tissue
(Figure 3-4). David et al. (2006) used Bayesian analysis to develop and calibrate metabolic
parameters in a human probabilistic PBPK model for dichloromethane, using kinetic data from
several studies of volunteers exposed to dichloromethane (n = 13 from DiVincenzo and Kaplan
[1981]; n = 12 from Engstrom and Bjurstrom [1977]; n = 14 from Astrand et al. [1975]; n = 3
from Stewart et al. [1972a], and group means for metabolism parameters from Andersen et al.
[1991]). Exhaled dichloromethane and CO and blood levels of dichloromethane and COHb were
available in the studies by Andersen et al. (1991) and DiVincenzo and Kaplan (1981). The other
three studies included two or three of these measures. The only available data for levels of
31
DRAFT - DO NOT CITE OR QUOTE
-------
dichloromethane in fat came from the study of Engstrom and Bjurstrom (1977) (described in
Section 3.2).
1
> Ga
Exchs
A GST*
s ^
nge
Richly
Perfused
1
Slowly
Perfused
. II .
nrc
CYP
•fP —
CO Sub
Model
1 t
Air
1 t
V 1
t
1
Endogenous
Production
Figure 3-4. Schematic of human PBPK used by David et al. (2006).
Values (means and SDs or CVs) for the model parameter distributions were selected from
multiple sources considered to provide the most current scientific evidence for each parameter
(David et al., 2006). Mean values for QCC, VPR, and all fractional tissue volumes and blood
flow rates were based on mean values used by U.S. EPA (2000d) in a PBPK model for vinyl
chloride, as were values for CVs for all physiological parameters, except CVs for VPR and
fractional lung volume, which were set to those used by OSHA (1997). Means for the CO
submodel parameters were set equal to those in Andersen et al. (1991), except for those for the
endogenous rate of CO production (REnCOC) and the background amount of CO (ABCOC),
which were based on data collected by DiVincenzo and Kaplan (1981). Means for partition
coefficients, the Al ratio and the A2 ratio were those used by Andersen et al. (1987), whereas
prior means for Vmaxc and Km were those used by Andersen et al. (1991). The prior mean for the
metabolic parameter for CYP metabolism in the rapidly perfused tissue was set at 0.03, slightly
lower than the value suggested by Sweeney et al. (2004). Prior CVs for the metabolic
parameters were set at 200%.
MCMC analysis was used to calibrate metabolic parameters in the human model in a
two-step approach: (1) posterior distributions were estimated separately by using data from each
of the five studies with kinetic data for humans exposed to dichloromethane (with durations
ranging from 1 to 8 hours and concentrations ranging from 50 to 1,000 ppm), and (2) posterior
distributions were estimated with combined data from the 42 individual subjects from the four
studies with individual subject data (DiVincenzo and Kaplan, 1981; Engstrom and Bjurstrom,
1977; Astrand et al., 1975; Stewart et al., 1972a). Estimates of the population mean values for
the fitted parameters from the Bayesian calibration with the combined kinetic data for individual
subjects are shown in Table 3-7. This analysis resulted in a narrowing of the distribution for the
32
DRAFT - DO NOT CITE OR QUOTE
-------
CYP2E1 metabolism parameters Vmaxc and Km from a fairly broad prior distribution with a CV
of 200% for both parameters to 13.1 and 33.6%, respectively, for Vmaxc and Km. It should be
noted that the CV values only represent the uncertainties in the corresponding population mean
values and do not include the estimated interindividual variability (Harvey Clewell, to Paul
Schlosser, U.S. EPA, email dated October 14, 2009). Thus, that narrowing should only be
interpreted as indicating a high degree of confidence in the population mean. As will be
discussed in detail later, other data which better characterize the variability in CYP2E1 activity
among the human population should then be used in conjunction with these uncertainties to
characterize the full range of uncertainty and variability.
Table 3-7. Results of calibrating metabolic parameters in a human
probabilistic PBPK model for dichloromethane with individual kinetic data
for 42 exposed volunteers and MCMC analysis
Parameter
VmaxC — maximal CYP metabolic rate (mg/hr/kg07)
Km— CYP affinity (mg/L)
kfc— first-order GST metabolic rate (kg0 3/hr)
Al— ratio of lung VmaxC to liver VmaxC
A2 — ratio of lung kfc to liver kfc
FracR — fraction of VmaxC in rapidly perfused tissues
Prior distributions
Mean
(arithmetic)
6.25
0.75
2
0.00143
0.0473
0.03
CV
2
2
2
2
2
2
Posterior distributions
Mean
(arithmetic)
9.42
0.433
0.852
0.000993
0.0102
0.0193
CV
0.131
0.336
0.711
0.399
0.728
0.786
Source: David et al. (2006).
A component of quantitative uncertainty arises in examining the results of David et al.
(2006), specifically for the GST metabolic parameter, kfc. The authors reported Bayesian
posterior statistics for the population mean parameters when calibration was performed either
with specific published data sets or the entire combined data set. While one would generally
expect that the values obtained from the combined data set should be a weighted average of the
values from individual data sets, the population mean for the liver GST activity (coefficient), kfc,
was 0.852 from the combined data set while the values from the individual data sets ranged from
1.92 to 34.0 kg°3/hour.
A clarification provided by Marino (Dale Marino, to Glinda Cooper, U.S. EPA, email
dated April 25, 2007) is that the parameter bounds stated in the text of David et al. (2006) were
only applied for the analysis of the DiVincenzo and Kaplan (1981) and the combined data set.
But according to the text and distribution prior statistics specified, the upper bound for kfc would
have been 12 kg°3/hour (mean + 2.5 SDs, with mean = 2 and SD = mean x CV = 2x2 = 4). The
data of Andersen et al. (1991) were not used in the combined analysis because only group
average values were available from that source, rather than individual data. Since the remaining
33
DRAFT - DO NOT CITE OR QUOTE
-------
study-specific mean kfc values were 7.95, 5.87, 34.0, and 1.92, with CVs of <2, it seems unlikely
that application of this upper bound would result in a value of kfc = 0.852 kg0 3/hour. Given that
there had been convergence problems with the combined data set when parameter values were
unbounded, it is possible that convergence had not actually been reached after parameter bounds
were introduced, and a higher value for kfc would have been obtained had the chain been
continued longer. The implications of this parameterization uncertainty are discussed further in
Section 5.3 for noncancer toxicity modeling and Section 5.4.5 for cancer dose-response
modeling.
Setting this uncertainty aside, since the parameter statistics shown in Table 3-7 (values
reported by David et al., 2006) represent population means and the level of uncertainty in those
means, their correct interpretation requires further consideration. As noted above, there is a
known range of variability in CYP2E1 expression among the human population which should be
incorporated when estimating overall population variability. For the remaining parameters
except kfc (i.e., for Km, Al, A2, and FracR), there is not a known equivalent level of variability
and it will be assumed that there is, in fact, a single true value for the population which is
estimated as the population mean by David et al. (2006). In that case one needs only to include
the uncertainty in the mean represented by the CV values in Table 3-7 in a statistical (Monte
Carlo) sampling in order to fully characterize model uncertainty and variability. While the
analysis of David et al. (2006) may have included variability among individual-specific estimates
for those parameters, this treatment effectively assumes that this variability was only an apparent
artifact of attributes such as the limited data and measurement noise.
David et al. (2006) further refined the human probabilistic model to reflect
polymorphisms in the GST pathway: homozygous positive (+/+) GST-T1, heterozygous (+/-)
GST-T1, and homozygous negative (-/-) GST-T1 individuals with no GST activity.
Distributions of GST activities for these genotypes in a group of 208 healthy male and female
subjects from Sweden were scaled to obtain distributions of kfc for each genotype (Warholm et
al., 1994). When weighted by estimated frequencies of the genotypes in the U.S. population and
appropriately scaled, these genotype-specific activity distributions would result in an overall
population mean equal to the kfc mean for the posterior distribution shown in Table 3-7
(0.852 kg°3/hour). The resultant mean kfc values were 0.676 kg°3/hour (SD 0.123) for
heterozygous individuals and 1.31 kg°3/hour (SD 0.167) for homozygous positive individuals, as
indicated in Table 2 of David et al. (2006). The final parameter distributions used by David et al.
(2006) are summarized in Table 3-8.
34 DRAFT - DO NOT CITE OR QUOTE
-------
Table 3-8. Parameter distributions used in human Monte Carlo analysis for
dichloromethane by David et al. (2006)
Parameter
BW
QCC
VPR
QFC
QLC
QRC
QSC
Body weight (kg)
Cardiac output (L/hr/kg° 74)
Ventilation:perfusion ratio
Fat
Liver
Rapidly perfused tissues
Slow perfused tissues
Distribution
Mean
(arithmetic)
70.0
16.5
1.45
0.05
0.26
0.50
0.19
SD
21.0
1.49
0.203
0.0150
0.0910
0.10
0.0285
Source
Humans3
Humans3
Humans3
Humans3
Humans3
Humans3
Humans3
Tissue volumes (fraction BW)
VFC
VLC
VLuC
VRC
VSC
Fat
Liver
Lung
Rapidly perfused tissues
Slowly perfused tissues (muscle)
0.19
0.026
0.0115
0.064
0.63
0.0570
0.00130
0.00161
0.00640
0.189
Humans3
Humans3
Humans3
Humans3
Humans3
Partition coefficients
PB
PF
PL
PLu
PR
PS
Blood:air
Fatblood
Liverblood
Lung: arterial blood
Rapidly perfused tissue :blood
Slowly perfused tissue (muscle :blood)
9.7
12.4
1.46
1.46
1.46
0.82
0.970
3.72
0.292
0.292
0.292
0.164
Humansb
Ratsb
Ratsb
Ratsb
Ratsb
Ratsb
Metabolism parameters
Vmaxc
Km
Al
A2
FracR
Maximum metabolism rate (mg/hr/kg0 7)
Affinity (mg/L)
Ratio of lung Vmax to liver Vmax
Ratio of lung KF to liver KF
Fractional CYP2E1 capacity in rapidly perfused tissue
9.42
0.433
0.000993
0.0102
0.0193
1.23
0.146
0.000396
0.00739
0.0152
Calibration0
Calibration0
Calibration0
Calibration0
Calibration0
First order metabolism rate (/hr/kg° 3)
kfC
Homozygous (-/-)
Heterozygous (+/-)
Homozygous (+/+)
0
0.676
1.31
0
0.123
0.167
Hybridd
Hybridd
Hybridd
SD = standard deviation.
3U. S. EPA, 2000d. Human PBPK model used for vinyl chloride.
bAndersen et al. (1987). Blood:air partition measured using human samples; other partition coefficients based on
estimates from tissue measures in rats.
°Bayesian calibration based on five data sets (see text for description); posterior distributions presented in this table.
dThe overall population mean for kfc as determined by Bayesian calibration; the distribution of activity among the
three genotypes and variability in activity for each genotype (SD values) were then scaled from the ex vivo data of
Warholmetal. (1994).
Source: David et al. (2006).
35
DRAFT - DO NOT CITE OR QUOTE
-------
Since the measurements of GST-T1 activity by Warholm et al. (1994) were performed ex
vivo using blood samples, it is reasonable to assume that those measurements are accurate and,
hence, that the characterization of the distributions for each genotype as being normal with the
reported level of variance represents the true level of human variability for each genotype.
However that characterization does not include the uncertainty in the overall population mean for
the rate of hepatic GST activity towards dichloromethane, indicated by the CV for kfc in
Table 3-7 (from Table 4 of David et al. [2006]). Thus, to fully account for both the population
variability and parameter uncertainty, a Monte Carlo statistical sampling should first sample the
population mean from a distribution with mean = 0.852 kg0 3/hour and CV = 0.711 (thus
accounting for uncertainty) and then reweight the population-specific distributions listed in
Table 3-8 to have that sampled population mean, before selecting (sampling) an individual value
of kfc from those weighted distributions (thus accounting for variability). EPA incorporated this
change in the PBPK modeling used in this assessment.
As described in detail in Appendix B, EPA evaluated the adequacy of all the parameter
distributions used by David et al. (2006) to characterize variability among the full human (U.S.)
population. EPA's conclusion is that the reported distributions for many of the physiological
parameters in particular, as well as Vmaxc (CYP2E1) as described above, only represented a
narrow set of adults and did not represent the full range of variability. EPA therefore chose to
use supplemental data sources to define these distributions in a way that should fully characterize
the variability in the human population for individuals between 6 months and 80 years of age.
Specifically, while the BW distribution used in the David et al. (2006) PBPK model used ranges
from 7 to 130 kg, thus covering 6-month-old children to obese adults, there are age-dependent
changes and gender-dependent differences in ventilation rates and body fat that are not explicitly
included. To more accurately reflect the distribution of physiological parameters in the entire
population, EPA replaced the unstructured distributions of David et al. (2006) with distributions
based on available information that specifically account for population variability in age, gender,
and age- and gender-specific distributions or functions for BW, QCC, alveolar ventilation, body
fat (fraction), and liver fraction (see Appendix B for more details of the evaluation of each of
these parameters).
For Vmaxc (CYP2E1), EPA also incorporated additional data for the variability in
CYP2E1 activity among humans based on Lipscomb et al. (2003). The Lipscomb et al. (2003)
study used in vitro analysis of liver samples from 75 human tissue donors (activity towards
trichloroethylene and measurements of protein content) to estimate a distribution of activity in
the population. These data support a wider distribution in CYP2E1 activity than had been used
in the David et al. (2006) dose-metric and unit-risk calculations, with approximately a sixfold
range observed for CYP2E1 in Lipscomb et al. (2003) and a twofold range used by David et al.
(2006). After sampling the population mean for CYP2E1 from the distribution indicated by the
parameters in Table 3-8 to capture the uncertainty in the population mean, EPA assumed a log-
36 DRAFT - DO NOT CITE OR QUOTE
-------
normal distribution for human variability around that mean and sampled the individual value
using that population mean with geometric standard deviation (GSD) = 1.73. Further, since even
the data available to Lipscomb et al. (2003) were limited, and the log-normal distribution is
naturally bounded to be greater than zero, EPA chose to use a nontruncated log-normal
distribution in the second (variability) sampling step for this parameter. (In sampling the
population-mean value for Vmaxc from its range of uncertainty, EPA did truncate the distribution
as indicated by David et al. [2006], since that mean is expected to be bounded away from zero.)
Finally, the scaling of CYP2E1 for individuals under the age of 18 was adjusted based on the
data of Johnsrud et al. (2003); EPA's analysis of these data indicate CYP2E1 activity in children
is better predicted when assumed to scale with BW raised to the 0.88 power, as compared to the
more general power of 0.7, used by David et al. (2006). (Details of EPA's analysis are given in
Appendix B, Section B.3.) CYP2E1 activity for individuals over the age of 18 is still assumed to
scale as BW°7.
The resulting set of parameter distribution characteristics, including those used as defined
by David et al. (2006) are described in Table 3-9. Using this revised set of distributions,
including the CYP and GST activity distributions, and other distributions used as defined by
David et al. (2006) or revised by EPA, the model as applied should reflect the full variability in
the (U.S.) human population.
37 DRAFT - DO NOT CITE OR QUOTE
-------
Table 3-9. Parameter distributions for the human PBPK model for dichloromethane used by EPA
Parameter
BW
Body weight (kg)
Distribution
Shape
Normal
(Geometric)
mean"
SD/GSD"
/ (age, gender)
Lower
bound
1st %tile
Upper
bound
99th %tile
Section or source
B.4.3;NHANESIV
Flow rates
QAlvC
vprv
QCC
Alveolar ventilation (L/hr/kg075)
Variability in ventilation:perfusion ratio
Cardiac output (L/hr/kg0 75)
Normal
Log-normal
f(age, gender)
1.00
QCCmem=/QAlvC)
f(age)
0.203
5th %tile
0.69
95* %tile
1.42
QCC = QCCmem/vprv
B.4.4; mean: Clewell et al. (2004);
SD: Arcus-Arth and Blaisdell (2007)
VPR/VPRmem of David et al. (2006)
B.4.5; Clewell et al. (2004) (mean)
Fractional flow rates (fraction of QCC)
QFC
QLC
QRC
QSC
Fat
Liver
Rapidly perfused tissues
Slow perfused tissues
Normal
Normal
Normal
Normal
0.05
0.26
0.50
0.19
0.0150
0.0910
0.10
0.0285
0.0050
0.010
0.20
0.105
0.0950
0.533
0.80
0.276
David et al. (2006); after sampling from
these distributions, normalize:
Ql-QC-Q>c
Z2/c
Tissue volumes (fraction BW)
VFC
VLC
VLuC
VRC
VSC
Fat
Liver
Lung
Rapidly perfused tissues
Slowly perfused tissues
Normal
Normal
Normal
Normal
Normal
/ (age, gender)
f(age)
0.0115
0.064
0.63
0.3 -mean
0.05 -mean
0.00161
0.00640
0.189
0.1-mean
0.85-mean
0.00667
0.0448
0.431
1.9-mean
1.15-mean
0.0163
0.0832
0.829
Fat mean: B.4.6 (Clewell et al., 2004);
liver mean: B.4.7 (Clewell et al., 2004);
otherwise, David et al. (2006); after
sampling from these distributions,
normalize:
0.9215 -BW-ViC
Z-JC
Partition coefficients
PB
PF
PL, PLu,
&PR
PS
Blood:air
Fat:blood
Liverblood, lung:arterial blood, and
rapidly perfused tissue:blood
Slowly perfused tissue (muscle) :blood
Log-normal
Log-normal
Log-normal
Log-normal
9.7
11.9
1.43
0.80
1.1
1.34
1.22
1.22
7.16
4.92
0.790
0.444
13.0
28.7
2.59
1.46
Geometric mean (GM) & GSD values
listed here, converted from arithmetic
mean and SD values of David et al.
(2006)
(Table 3-9 continues on next page)
38
DRAFT - DO NOT CITE OR QUOTE
-------
Table 3-9. Parameter distributions for the human PBPK model for dichloromethane used by EPA
Parameter
Distribution
Shape
(Geometric)
mean"
SD/GSD3
Lower
bound
Upper
bound
Section or source
Metabolism parameters (based on Monte Carte calibration from five human data sets)
* maxCmean
/
Vmaxc
Km
Al
A2
FracR
Population mean /
individual maximum metabolism rate
(mg/hr/kgxvmax)
Affinity
(mg/L)
Ratio of lung Vmax to liver Vmax
Ratio of lung KF to liver KF
Fractional MFO capacity in rapidly
perfused tissue
Log-normal
Log-normal
Log-normal
Log-normal
Log-normal
Log-normal
9.34
* maxCmean
0.41
0.00092
0.0083
0.0152
1.14
1.73
1.39
1.47
1.92
2.0
6.96
(none)
0.154
0.000291
0.00116
0.00190
11.88
(none)
1.10
0.00292
0.0580
0.122
B.3 mean: David et al. (2006);
Individual GSD: Lipscomb et al.
(2003); Xvmax = 0.88 for age <18;
Xvmax = 0.70 for age >18
Geometric mean (GM) & GSD values
listed here, converted from arithmetic
mean and SD values of David et al.
(2006)
First order metabolism rate ([hr/kg03]"1)
kfCmean
Population average
kfC/kfCmean
Homozygous (-/-)
Heterozygous (+/-)
Homozygous (+/+)
Log-normal
Normal
Normal
Normal
0.6944
0
0.8929
1.786
1.896
0
0.1622
0.2276
0.1932
0
0
0
2.496
0
1.704
2.924
Adapted from David et al. (2006);
kfcmean is first sampled, then the relative
individual value, kfc/kfCmean, given the
genotype; kfc is then the product
""Arithmetic mean and SD listed for normal distributions; GM and GSD listed for log-normal distributions.
39
DRAFT - DO NOT CITE OR QUOTE
-------
3.5.3. Evaluation of Rat PBPK Dichloromethane Models
Several deterministic PBPK rat models have been reported in the scientific literature
(Sweeney et al., 2004; Andersen et al., 1991, 1987; Reitz, 1991; Reitz et al., 1988a, b; U.S. EPA
1988b, 1987a, b; Gargas et al., 1986). Unlike the mouse (Marino et al., 2006) and human (David
et al., 2006), no hierarchical population model for dichloromethane in the rat exists in which
parameter uncertainty is quantitatively integrated into model calibration. Rat data are not
available that would allow for Bayesian calibration of individual metabolic parameters for the
CYP or GST pathways. Thus, EPA assessed modified versions of deterministic rat PBPK
models to select the most appropriate model for use in extrapolating internal dosimetry from rats
to humans, for example in the determination of RfDs and RfCs based on effects seen in the rat.
This work is described in detail in Appendix C and is based on evaluation of blood levels of
dichloromethane, the percent saturation of hemoglobin as COHb (%COHb), and expired
dichloromethane following intravenous injection (Angelo et al., 1986b) and closed chamber gas
uptake (Gargas et al., 1986), as well as evaluation of dichloromethane and %COHb blood levels
from a 4-hour inhalation exposure (Andersen et al., 1991, 1987). Based on this work, the basic
model structure of Andersen et al. (1991) was chosen, with the inclusion of lung
dichloromethane metabolism via CYP (0.2% of liver metabolite production) and GST (14.9% of
liver metabolite production) pathways (estimated from Reitz et al., 1989) (Figure 3-5) with
metabolic parameters recalibrated against data of Andersen et al. (1991), based on prediction
agreement of the various parameters with the available rat data sets. Table 3-10 presents the
parameter distribution data for this model.
40 DRAFT - DO NOT CITE OR QUOTE
-------
DCM model
\\GST <«—| |-»CYP
Air
Alveolar
air
Oral _
exposure
nQT -.
Fat
Richly
perfused
Slowly
perfused
Lung
tissue
—> Gl tract
a
I
Liver
t —
». HYP
CO model
Endogenous
production
Figure 3-5. Schematic of rat PBPK model used in current assessment.
41
DRAFT - DO NOT CITE OR QUOTE
-------
Table 3-10. Parameter values for the rat PBPK model for dichloromethane
used by EPA
Parameter
Mean
Flow rates
QCC (L/hr/kg074)
VPR
15.9
0.94
Fractional flow rates (percent of QCC)
Fat
Liver
Rapidly perfused tissues
Slowly perfused tissues
9
20
56
15
Tissue volumes (percent BW)
Fat
Liver
Lung (scaled as BW° ")
Rapidly perfused tissues
Slowly perfused tissues
7
4
1.15
5
75
Partition coefficients
Blood:air
Fat:blood
Liverblood
Lung:arterial blood
Rapidly perfused tissue :blood
Slowly perfused tissue (muscle) :blood
19.4
6.19
0.732
0.46
0.732
0.408
Metabolism parameters
VmaxC: maximum metabolism rate (mg/hr/kg07)
Km: affinity (mg/L)
kfc: 1st order GST metabolic rate in liver (kg°3/hr)
Al : ratio of lung Vmax to liver Vmax
A2: ratio of lung kf to liver kf
PI : yield of CO from CYP metabolism
Other elimination/absorption constants
Fl : correction factor for CO exhalation rate
kj,: first-order oral absorption rate constant, ka (1/hr)
3.97
0.510
2.47
0.002
0.149
0.68
1.21
4.31
3.5.4. Comparison of Mouse, Rat, and Human PBPK Models
The comparison of various parameters across species (Table 3-11) primarily shows the
modest interspecies differences that are known to occur in physiological parameters, also
including the approximately twofold differences in partition coefficients which occur because of
differences in rodent versus human blood lipid content. The 2.5-fold lower Vmaxc (CYP activity)
in rats versus mice is also typical. The most striking difference is the variation in Al and A2.
Those values, however, reflect the in vitro differences originally quantified by Lorenz et al.
(1984) and Reitz et al. (1989), with the mouse and human values being those used in the
42
DRAFT - DO NOT CITE OR QUOTE
-------
dichloromethane PBPK modeling of Andersen et al. (1987). In examining the derivation of the
rat values, however, it appeared that Andersen et al. (1987) must have failed to adjust the relative
activity per mg microsomal protein to account for the difference in total microsome content of
the lung versus liver. Once this adjustment was performed, the rat value was found to be 20-fold
lower than the value used by Andersen et al. (1987) and about twice that of the human. These
interspecies differences in Al and A2 are based on independent measurements of tissue-specific
metabolic capacity; while the specific values for mouse and human were refined through
Bayesian analysis, the ultimate (posterior) values used are within a reasonable range of the in
vitro measurements and so do not appear to be artifactual. (Since in vivo kinetics often indicate
some differences from what would be predicted without adjustment from in vitro, it is not
surprising that such differences occur here.) These differences do explain why lung-specific
metrics in particular lead to lower internal dose and hence risk predictions in humans compared
to whole-body metrics.
43 DRAFT - DO NOT CITE OR QUOTE
-------
Table 3-11. Parameters in the mouse, rat, and human PBPK model for dichloromethane used by EPA
Parameter
Fractional flow rates (fraction of cardiac output)0
QFC Fat
QLC Liver
QRC Rapidly perfused tissues
QSC Slowly perfused tissues
Fractional tissue volumes (fraction ofBW)b
VFC Fat
VLC Liver
VluC Lung
VRC Rapidly perfused tissues
VSC Slowly perfused tissues
Partition coefficients'^
PB Blood/air
PF Fatftlood
PL Liver/blood
PLu Lung/blood
PR Rapidly perfused^lood
PS Slowly perfused/blood
Flow rates
QCC Cardiac output (L/hr/kg° 74)
VPR Ventilation/perfusion ratio
QAlvC
Mouse"
mean
0.05
0.24
0.52
0.19
0.04
0.04
0.0115
0.05
0.78
23.0
5.1
1.6
0.46
0.52
0.44
24.2
1.45
QCC/VPR
Ratb
value
0.09
0.20
0.56
0.15
0.07
0.04
0.0115
0.05
0.75
19.4
6.19
0.73
0.46
0.73
0.41
14.99
0.94
QCC/VPR
Human0
Mean
0.05
0.26
0.50
0.19
f(age, gender)
f(age)
0.0115
0.064
0.63
9.7
11.9
1.43
1.43
1.43
0.80
QCCmean=/QAlvC)
(variable)
f(age, gender)
CV/GSD (shape, bounds)
0.3 (N, 0.1-1.9)
0.35 (N, 0.0385-2.05)
0.2 (N, 0.4-1.6)
0. 15 (N, 0.553-1.453)
0.3 (N, 0.1-1.9)
0.05 (N, 0.85-1.15)
0.14 (N, 0.58-1.42)
0.1 (N, 0.7-1.3)
0.3 (N, 0.684-1.32)
7J(LN, 0.738-1.34)
1.34 (LN, 0.413-2.41)
1.22 (LN, 0.552-1.81)
II
II
1.22 (LN, 0.555-1.83)
QCC = QCCmean/vprv
(varies) (LN, 0.69-1.42)
f(age) (N, 5^-95*%)
Sources
David et al. (2006); then
normalized:
oc.g/c
& 2>c
Fat mean: §2.2.3.6;
Liver mean: §2.2.3.7;
otherwise David et al.
(2006); then normalized:
Jr 0.9215 -BW- ViC
'• I.**
Geometric mean (GM) &
GSD/GM values
converted from arithmetic
mean & SDs of David et
al. (2006)
QCC: §2.2.3.5;
vprv = VPR/VPRmean:
David et al. (2006);
QAlvC: §2.2.3.4;
(Table 3-11 continues on next page)
44
DRAFT - DO NOT CITE OR QUOTE
-------
Table 3-11. Parameters in the mouse, rat, and human PBPK model for dichloromethane used by EPA
Parameter
Metabolism parameters
VmaxC Maximum CYP metabolic rate (mg/hr/kgXvmax)
Xvmax CYP allometric scaling power
Km CYP affinity (mg/L)
kfc(mean) First-order GST metabolic rate constant
(kg°3/hr)
kfC/kfc(mean)
Al Ratio of lung VmaxC to liver VmaxC
A2 Ratio of lung kfc to liver kfc
Mouse"
mean
9.27
0.7
0.574
1.41
1
0.207
0.196
Ratb
value
3.97
0.7
0.510
2.47
1
0.002
0.149
Human0
Mean
LN(m =9.34, s = 1.14
Ib = 6.96, ub = 11.88)
0.88 for age <18;
0.7 for age >18)
0.41
LN(m = 0.6944,
s = 1.896 Ib =0.1932,
ub = 2.496)
0 (-/-)e
0.8929 (+/-)e
1.7896 (+/+)e
0.00092
0.0083
CV/GSD (shape, bounds)
7. 73 (LN, [unbounded])
1.39 (LN, 0.376-2.68)
-/-: NA
+/-: 0. 182 (N, 0-1.91)
+/+: 0. 127 (N, 0-1.64)
7.47 (LN, 0.316-3. 17)
7.P2(LN, 0.140-6.99)
Sources
VmaxC: §2.2.2;
others: David et al. (2006)
(GM&GSD/GM values
converted from arithmetic
mean & SDs)
kfc(mean): Combined data
set posterior, Table 4 of
David et al. (2006)
kfc/kfc(mean): rescaled
from Table 2 of David et
al. (2006)
aBased on Marino et al. (2006) (source for all mouse parameters).
''Based on Andersen et al. (1991), with the addition of lung metabolism of dichloromethane via the CYP (4% of liver metabolite production) and GST (14% of liver
metabolite production) pathways. Physiological parameters and partition coefficients are from Andersen et al. (1991). The values for dichloromethane metabolism
in the lung (as a fractional yield of liver metabolism for each pathway) were estimated from the in vitro ratios of enzyme activity (nmol/min/mg protein) in lung and
liver cytosolic (GST) and microsomal (CYP) tissue fractions (Reitz et al., 1989). Metabolic parameters were re-optimized against the inhalation data of Andersen et
al. (1991) using a heteroscedasticity parameter value of 2, which uses relative error for the model fitting algorithm. See Appendix C for further details.
°Based on David et al. (2006), with changes as noted. Additional sources include Clewell et al. (2004), Arcus-Arth and Blaisdell (2007), and Lipscomb et al. (2003).
See identified sections for details. Distribution values (mean and a measure of dispersion) are provided with the CV (mean/SD) presented for normal (N)
distributions and the GSD (italicized) presented for log-normal (LN) distributions. Distributions were truncated, bounds are (upper-lower bound)/mean.
eValues for the homozygous (-/-), heterozygous (+/-), and homozygous (+/+) GST-T1 genotypes, respectively.
45
DRAFT - DO NOT CITE OR QUOTE
-------
3.5.5. Uncertainties in PBPK Model Structure for the Mouse, Rat and Human
There is uncertainty in the dichloromethane PBPK modeling because there are parts of
the entire data set for which the existing model and parameters fit very poorly and where the
discrepancies appear to be structural (i.e., cannot be resolved simply by re-fitting model
parameters unless fits to other data are degraded). The data used for model parameter estimation
are primarily measurements of parent dichloromethane kinetics (e.g., blood or closed-chamber
air concentrations over time), rather than measurements of metabolite levels which can be
unambiguously attributed to one of the two principal metabolic pathways (GST and CYP). For
the mouse model in particular, only parent dichloromethane data were used, though exhaled
amounts of CC>2 and CO are available. Because only dichloromethane measurements are used,
estimation of the fraction of dichloromethane metabolized by the GST versus CYP pathway
depends strongly on the assumed equations describing those rates and how the pharmacokinetic
data are interpreted. Specifically, if there is some degree of saturation in the GST kinetics or the
CYP kinetics are not accurately described by Michaelis-Menten kinetics (see next paragraph),
then the estimation of the fraction of metabolism going down each pathway would shift. Further,
while Marino et al. (2006) used data from mice pretreated with ^ram'-l,2-dichloroethylene
(tDCE), a specific CYP2E1 inhibitor (mice were exposed to 100 ppm tDCE for 1.5 h prior to
dichloromethane exposure), the authors assumed without verification that 100% of the CYP2E1
activity was eliminated by the inhibitor when using those data. In contrast, Mathews et al.
(1997) found that pretreatment of F344 rats by tDCE (100 mg/kg intraperitoneally) only yielded
65% inhibition of CYP2E1. If a significant fraction of the CYP2E1 activity was not eliminated
in the dichloromethane experiments, then that activity is erroneously assigned to the GST
pathway in the parameter estimation of Marino et al. (2006).
In addition to the possibility of incomplete inhibition of CYP2E1 affecting the data
interpretation, the Michaelis-Menten rate equation used in all of the published PBPK models for
dichloromethane, including that of Marino et al. (2006), has in fact not been shown to accurately
describe the CYP2E1-mediated metabolism of dichloromethane in the relevant concentration
range. While the Michaelis-Menten equation usually describes CYP-mediated oxidation data
quite well, if there is some departure of the actual kinetics from this equation and incomplete
inhibition from the tDCE treatment, the model parameters obtained under those assumptions
would be compromised. If pathway-specific metabolite data were used to define or bound the
ratio of GST to CYP metabolism, the resulting estimates would be less sensitive to errors in the
CYP rate equation.
EPA compared model predictions of total CYP metabolism in mice, which should match
with CO elimination at 24 hours after bolus exposures to dichloromethane as measured by
Angelo et al. (1986a), since only the CYP pathway produces CO. At 50 mg/kg (in water), the
model predicts that 10% of the dichloromethane is metabolized by the CYP pathway, which
agrees with the observed values (11-12%; Angelo et al., 1986a). However, at 500 and 1,000
46 DRAFT - DO NOT CITE OR QUOTE
-------
mg/kg, the model predicts that only 1.8 and 1.0% will be metabolized by the CYP pathway,
while Angelo et al. (1986a) observed 9-10 and 3.6-7%, respectively. Thus, the extent of CYP
saturation predicted by the model does not match with these pathway-specific metabolite data.
However, as shown in Appendix C, the rat model is able to predict total exhaled CO quite well,
indicating an error in the fraction of metabolism via the GST pathway of less than 13% in that
species. That the mouse model does not describe well the dose-dependent shift in metabolism
shown by those CO data suggests that the dose-dependence of the CYP Michaelis-Menten rate-
equation may not be adequate. As will be shown, an alternative equation for CYP kinetics may
fit the existing dichloromethane data better than Michaelis-Menten kinetics, with the result that a
higher portion of total dichloromethane metabolism would be interpreted as being CYP-
mediated. Thus, there is some uncertainty in the choice of equation for the CYP pathway, which
leads to some uncertainty in the estimated GST:CYP metabolic ratio, upon which current risk
predictions are based. However the extent of the error appears quite limited in the rat and more
predominant at high exposures versus low exposures in the mouse.
The potential error in assuming Michaelis-Menten kinetics for CYP-mediated oxidation
of dichloromethane is reinforced by examining the in vitro oxidative (i.e., CYP-specific) kinetics
of dichloromethane reported by Reitz et al. (1989). When extrapolated from in vitro to in vivo,
the apparent values of the oxidative saturation constant, Km, identified by Reitz et al. (1989) for
mice, rats, and humans are over 2 orders of magnitude greater than those obtained in vivo with
the PBPK model. This apparent discrepancy is partly explained by the disparate concentration
ranges investigated: Reitz et al. (1989) used much higher dichloromethane concentrations in
vitro than those observed in or predicted for the various in vivo pharmacokinetic studies. In
particular, the oxidation of dichloromethane could involve two oxidative processes, one with a
high affinity (low Km) corresponding to the nonlinearity observed in vivo and one with a low
affinity (high Km) corresponding to the nonlinearity observed in vitro. Such a low-affinity
process might account for the higher CO production observed in vivo (see above) than predicted
by the current model. Further, a low-affinity process would have nearly linear kinetics in the
exposure range used for the in vivo dosimetry studies and hence would be difficult to distinguish
from GST-mediated metabolism unless pathway-specific metabolite data are used. If this second
oxidative process is not inhibited by tDCE, then it may correspond to the 35% of oxidative
metabolism which was observed to remain in rats after tDCE treatment by Mathews et al. (1997).
The data of Reitz et al. (1989) could simply indicate a second CYP with low-affinity
dichloromethane activity. However that possibility is contradicted by the results of Kim and
Kim (1996) who observed that another CYP2E1-specific inhibitor, disulfiram, completely
abolished dichloromethane-induced increases on COHb in rats. Another possible explanation
which would support the findings observed in Kim and Kim (1996) as well as Reitz et al. (1989)
and the various in vivo data is that a number of CYPs exhibit "atypical" kinetics, not described
by the classic Michaelis-Menten equation, consistent with the enzymes having dual binding sites
47 DRAFT - DO NOT CITE OR QUOTE
-------
as proposed by Korzekwa et al. (1998). (Korzekwa et al. [1998] demonstrated atypical kinetics
for several CYP-isozyme/substrate pairs, but not specifically for CYP2E1.) The application of
this alternate kinetic model to dichloromethane dosimetry in mice has been explored by Evans
and Caldwell (2010), who demonstrate that all the dichloromethane gas-uptake data in mice can
be explained with this model in the hypothetical case where the GST pathway is not included.
The alternate PBPK model of Evans and Cal dwell (2010) is not considered further here because
GST-mediated metabolism of dichloromethane clearly occurs in mice, rats, and humans based on
the in-vitro observations of Reitz et al. (1989) and is mechanistically linked to dichloromethane-
induced cancer as discussed in Section 4.5. Thus a model which excludes GST-mediated
metabolism is not consistent with the overall database concerning dichloromethane metabolism
and carcinogenesis research.
Figure 3-6 shows kinetic model fits to the in vitro mouse dichloromethane oxidation
kinetic data of Reitz et al. (1989), after expressing those data on a per gram of liver basis. Both
the standard Michaelis-Menten kinetic equation (solid line) and the dual-binding equation
(dashed line) given by Korzekwa et al. (1998) are shown. In particular, the high-affinity (low)
Km for the dual-binding equation was set equal to that obtained by Marino et al. (2006) from
their PBPK modeling. This figure shows that the dual binding model is not only consistent with
the apparent high-affinity saturation obtained from in vivo PBPK modeling (Km of Marino et al.
[2006]), but also with the apparent low-affinity (high Km) data of Reitz et al. (1989), and
describes those in vitro data better than the standard Michaelis-Menten equation. Reitz et al.
(1989) used classic Lineweaver-Burk plots to display their kinetic data; i.e., I/reaction rate
versus I/concentration. The systematic discrepancy between their data and Michaelis-Menten
kinetics evident in Figure 3-6 is much less obvious with that scaling, which likely explains why
they made no note of it.
48 DRAFT - DO NOT CITE OR QUOTE
-------
0)
2.5 -
2
)
c
B
Q.
>; 0.5
1 -
Reitz et al. (1989) data
Michaelis-Menten kinetics
Dual-binding CYP kinetics
100 200 300
[DCM] (tr^/L)
400
500
Dichloromethane oxidation data obtained with mouse liver microsomes by Reitz
et al. (1989) (points), expressed on a per gram of liver basis, are shown with a
fitted Michaelis-Menten equation (solid line) or a fitted dual-binding-site equation
as described by Korzekwa et al. (1998) (dashed line), where the high affinity
saturation constant of the dual-binding-site equation set equal to the mean Km
determined for mice via PBPK modeling by Marino et al. (2006). The Km for the
Michaelis-Menten equation (108 mg/L) is inconsistent with the in vivo
dichloromethane dosimetry data, while the in vitro data shown here are
inconsistent with the Km estimated in vivo (0.42 mg/L) if that equation is used.
Figure 3-6. Comparison of dichloromethane oxidation rate data with
alternate kinetic models.
In summary regarding model equations, the current PBPK model used the standard
Michaelis-Menten equation to describe CYP2E1-catalyzed oxidation of small volatile organic
compounds. Analysis of the dichloromethane (pharmaco)kinetic data and evaluation of the
inconsistencies described above suggest that an alternate equation, which would impact risk
predictions, may better represent CYP2E1-induced oxidation of dichloromethane. The analysis
provided here demonstrates shortcomings in the existing model which the alternate model may
address, indicating that this is a substantial model uncertainty. However, the hypothesis that
CYP2E1 kinetics for dichloromethane should be described by this alternate rate equation
requires further laboratory testing. For example, dichloromethane oxidation in a bacterial
expression system where only CYP2E1 is expressed could be measured over a concentration
range sufficient to firmly distinguish between the two kinetic forms indicated in the figure above.
Such experiments would clearly show that the metabolic kinetics are due to atypical kinetics
occurring with a single enzyme (CYP2E1), versus involvement of a second, low-affinity
enzyme. Also, the alternate equation would need to be incorporated into a PBPK model which
also included the GST pathway, and the resulting model calibrated not only for the mouse, but
49
DRAFT - DO NOT CITE OR QUOTE
-------
also the rat and the human. Until such additional experiments and modeling are available, the
existing PBPK model remains the best available science for dose- and risk-extrapolation from
rodents to humans despite this uncertainty. Analysis of the GST-mediated metabolism of
dichloromethane measured by Reitz et al. (1989) shows that those results are within a factor of
three of the GST kinetic parameters used in the current PBPK model, indicating that any error in
the GST:CYP balance is in that range.
50 DRAFT - DO NOT CITE OR QUOTE
-------
4. HAZARD IDENTIFICATION
4.1. STUDIES IN HUMANS
4.1.1. Introduction—Case Reports, Epidemiologic, and Clinical Studies
There has been considerable interest in the influence of occupational exposure to
dichloromethane in relation to a variety of conditions. The recognition that dichloromethane can
be metabolized and bound to hemoglobin to form COHb, resulting in a reduction in the oxygen
carrying capacity of the blood (Stewart et al., 1972b), prompted investigations into risk of
ischemic heart disease and other cardiovascular effects. Reports of neurological effects from
acute, high-exposure situations contributed to concern about neurological effects of chronic
exposure to lower levels of dichloromethane. A general interest in potential cancer risk became
more focused on lung and liver cancer because of the observation of these specific tumors in the
NTP (1986) experiments in mice. Details of the studies pertaining to the experimental and
epidemiologic studies of noncancer outcomes (e.g., cardiac, neurologic, hepatic, reproductive)
are presented in Section 4.1.2, and studies of cancer risk are presented in Section 4.1.3.
4.1.2. Noncancer Studies
4.1.2.1. Case Reports of Acute, High-dose Exposures
Numerous published case reports describe health effects resulting from acute exposure to
dichloromethane. Most describe health effects resulting from inhalation of dichloromethane or
dermal contact, but a few involve ingestion. The COHb levels in some of these cases were
relatively low (7.5-13%), so the initial toxic effects of acute dichloromethane exposure appear to
be due to its anesthetic properties as opposed to metabolic conversion of dichloromethane to CO.
Bakinson and Jones (1985) reported on a series of 33 cases of acute inhalation exposures
to dichloromethane that occurred in the workplace over the period 1961-1980. Thirteen had lost
consciousness, and one of the workers died. Nineteen cases reported general neurological
effects, 13 reported gastrointestinal symptoms, 4 reported respiratory symptoms, and 1 reported
hepatic symptoms. Of the 19 with general neurological symptoms, all reported headache, and
dizziness was reported by 11 workers. Five workers reported one of the following symptoms:
drunkenness, confusion, lack of coordination, or paresthesia.
Rioux and Myers (1988) summarized the health effects reported for 26 cases of
dichloromethane poisoning published in the literature between 1936 and 1986. Three cases
resulted from abuse-related exposures, 2 from chronic exposures, and 21 from acute exposures.
The most common effects involved the central nervous system (CNS) (unconsciousness,
drowsiness, headache, and behavioral symptoms), pulmonary edema and dyspnea, and
dermatologic symptoms. Even severe symptoms could be reversed, but four deaths occurred.
51 DRAFT - DO NOT CITE OR QUOTE
-------
More than 10 other case reports of fatalities or poisonings have been published since the
summaries by Rioux and Myers (1988) and Bakinson and Jones (1985), and many of these
incidents involve inadequately ventilated occupational settings (Jacubovich et al., 2005; Raphael
et al., 2002; Fechner et al., 2001; Zarrabeitia et al., 2001; Goulle et al., 1999; Mahmud and
Kales, 1999; Kim et al., 1996; Tay et al., 1995; Manno et al., 1992; Leikin et al., 1990;
Shusterman et al., 1990). CNS depression and resulting narcosis, respiratory failure, and heart
failure are common features of these reports. In a survey of workers in furniture stripping shops,
10 of the 21 workers stated that they sometimes experienced dizziness, nausea, or headache
during furniture stripping operations (Hall and Rumack, 1990).
Chang et al. (1999) reported details of six patients who had ingested dichloromethane
(four in a suicide attempt and two from accidental ingestion during intoxication). The estimated
amounts ingested were <350 mL. COHb levels, measured in only two of the cases, were 8.4 and
35% (with the latter being seen in a fatal case). As in exposures resulting from inhalation, the
most common symptoms involved CNS depression, ranging from somnolence and weakness to
deep coma. Tachypnea (n = 6) and corrosive gastrointestinal tract injury (n = 3) were also
reported. Hepatic and renal failure and pancreatitis were found in the two most severe cases.
4.1.2.2. Controlled Experiments Examining Acute Effects
Several controlled experiments were conducted in the 1970s examining
neurophysiological effects and levels of COHb resulting from short-term (1-4 hours) exposures
to dichloromethane at levels up to 1,000 ppm, or longer-term exposures at levels up to 500 ppm.
The 8-hour threshold limit value before 1975 was 500 ppm (NIOSH, 1986). These studies are
described below. With the exception of Putz et al. (1979), there is no description in the
published reports of the informed consent and other human subjects research ethics procedures
undertaken in these studies, but there is no evidence that the conduct of the research was
fundamentally unethical or significantly deficient relative to the ethical standards prevailing at
the time the research was conducted.
In 1972, Stewart et al. (1972a, b) reported results from four experiments that were
initiated based on the chance observation of an elevation in COHb saturation levels by one of the
investigators the morning after he had spent 2 hours working with varnish remover. Participants
were medical students and faculty (including at least one of the coauthors). A total of 11 healthy
nonsmoking volunteers were placed in an exposure chamber with mean concentrations of
dichloromethane of 213 ppm for 1 hour (n = 1), 514 ppm for 1 hour and then 869 ppm for 1 hour
(n = 3), 514 ppm for 1 hour (n = 8), or 986 ppm for 1 hour (n = 3). The COHb saturation post-
exposure peaked at 2.4% in the one subject exposed to 213 ppm, 4-8.5% in the 3 subjects
exposed to 514 and 869 ppm, 3.4% (mean) in the 8 subjects expose to 514 ppm, and 10% (mean)
in the 3 subjects exposed to 986 ppm. Stewart et al. (1972) noted that these experiments
indicated that dichloromethane exposure above 500 ppm resulted in COHb saturation levels that
52 DRAFT - DO NOT CITE OR QUOTE
-------
exceeded and were more prolonged than those seen with the threshold limit value exposures to
CO. The exposures also resulted in symptoms of CNS depression indicated by visual evoked
response changes and reports of light-headedness. Although return of COHb levels to
background levels could take >24 hours, all of the other symptoms were reversible within a few
hours after exposure ceased. Because only one subject participated in the lower exposure (213
ppm) protocol, this study provides little information regarding effects at levels below 500 ppm.
Winneke (1974) measured auditory vigilance, visual flicker fusion frequency, and
14 psychomotor tasks in a total of 38 women exposed to dichloromethane levels of 300-800 ppm
for 4 hours in an exposure chamber. A comparison group (nine females, nine males) exposed to
100 ppm CO for 5 hours was also included. Exposure to 800 ppm dichloromethane resulted in a
statistically significant decrease in the performance of 10 of the 14 psychomotor tasks. In tests
of auditory vigilance and visual flicker fusion, depressed response was seen at 300 ppm and was
further depressed at 800 ppm. These effects were not seen with CO exposure.
Forster et al. (1974) exposed four healthy young men to dichloromethane levels ranging
from 0 to 500 ppm for 7.5 hours/day for a total of 26 days over a 6-week period to investigate
alterations in hemoglobin affinity for oxygen and altered pulmonary function. While no changes
were observed in pulmonary function, the percent oxyhemoglobin saturation was increased,
suggesting an increased hemoglobin affinity for oxygen (as occurs with increasing CO levels),
with no indication of adaptation to restore this affinity for oxygen to normal.
Putz et al. (1979) examined the behavioral effects seen after exposure to dichloromethane
and to CO. Twelve healthy volunteers (six men and six women) each acted as his/her own
control in separate 4-hour exposures to 70 ppm CO and 200 ppm dichloromethane. These levels
were chosen so that the COHb level would reach 5% from each of these exposures. These were
conducted as double-blind experiments so that neither the investigators nor the participant knew
the exposure condition under study. Informed consent was obtained, and the study was reviewed
by the National Institute of Occupational Safety and Health (NIOSH) Human Subject Review
Board. The performance tests were dual tasks (an eye-hand coordination task in conjunction
with a tracking task), with five measures of performance assessed at six time points over the 4-
hour test period and an auditory vigilance task. Two levels of difficulty were assessed for each
task to allow assessment of whether the exposure effect was similar in low and high difficulty
tasks. The tests of eye-hand coordination, tracking tasks, and auditory vigilance revealed
significant impairment with both exposures under the more difficult task conditions. Effects
were similar or stronger in magnitude for dichloromethane compared with CO.
4.1.2.3. Observational Studies Focusing on Clinical Chemistries, Clinical Examinations, and
Symptoms
Studies in currently exposed workers. Ott et al. (1983a, c, d) evaluated several
parameters of hepatic, hematopoietic, and cardiac function in workers exposure to
53 DRAFT - DO NOT CITE OR QUOTE
-------
dichloromethane in a triacetate fiber production plant in Rock Hill, South Carolina. Two
hundred sixty-six Rock Hill workers and a comparison group of 251 workers in an acetate fiber
production plant in Narrows, Virginia, were included in the examination of urinary and blood
measures. These groups included men and women, blacks and whites, and smokers and
nonsmokers. The median 8-hour TWA exposure for dichloromethane ranged from 60 to
475 ppm in Rock Hill. Acetone at levels up to >1,000 ppm was present in both plants, but
dichloromethane and acetone exposures were inversely related.
There were differences in blood collection procedures between the two plants and in the
age, sex, race, and smoking history distribution of the study groups. The demographic and
smoking differences were accounted for in the analysis by stratification. Statistically significant
differences were seen between the workers in the two plants for COHb, serum alanine
aminotransferase (ALT), total bilirubin, and mean corpuscular hemoglobin concentration
(MCHC) (although the direction and magnitude of these differences were not reported, and the
authors stated that the difference in serum ALT could be due to the differences in blood
collection procedures, which involved a sitting versus recumbent position of the subjects at the
exposed and nonexposed plants, respectively) (Ott et al., 1983c). Within the Rock Hill plant,
analyses were also conducted to examine associations between dichloromethane exposure and
the clinical parameters within specific race-sex groups by using multiple regression to control for
smoking status, age, and time of venipuncture. Positive associations were seen with COHb in all
race-sex groups (increases of 0.7-2.1% per 100 ppm increase in dichloromethane) and with total
bilirubin (increases of 0.05-0.08 mg/dL per 100 ppm increase in dichloromethane) in all groups
except nonwhite men (which was a much smaller group, n = 20, than the other groups). Red cell
count, hematocrit, hemoglobin, and aspartate aminotransferase (AST) were also positively
associated with dichloromethane exposure in white females. The increase in total bilirubin level
was not supported by parallel changes in other measures of liver function or red blood cell
turnover, suggesting that this measure was not reflecting liver damage or hemolysis.
The increased red cell count, hemoglobin, and hematocrit in women exposed to high
levels of dichloromethane (up to 475 ppm, 8-hour TWA) may indicate a compensatory
hematopoietic effect. The fact that these changes were not significant among men may be due to
higher baseline hemoglobin, which was observed when comparisons were made between
nonsmoking men and women. No such difference in the baseline values was observed among
the smoking men and women, suggesting that the compensatory advantage may be lost among
smokers.
Ott et al. (1983e) presented results from a further investigation of changes in COHb,
alveolar CO, and oxygen half-saturation pressure in relation to dichloromethane exposure.
Blood samples were collected before and after shifts from 136 Rock Hill and 132 Narrows
workers. For the Rock Hill workers, personal monitoring for dichloromethane exposure was
done during the shift. The TWA for dichloromethane ranged from 0 to 900 ppm, with a bimodal
54 DRAFT - DO NOT CITE OR QUOTE
-------
distribution (peaks around 150 and 500 ppm) resulting from the layout of the plant. The blood
samples were used to determine blood COHb, alveolar CO levels, and the partial oxygen
pressure (Pso; that is, the pressure required to keep 50% of the blood oxygen-carrying capacity
saturated with oxygen at pH 7.4 and 37°C). Separate analyses were conducted for smokers and
nonsmokers to account for the smoking-related effects on COHb. Linear relationships were seen
between dichloromethane exposure and the before-shift COHb and alveolar CO levels, reflecting
residual metabolism from the previous day's exposure. There were significant quadratic
relationships between dichloromethane exposure and the postshift COHb and alveolar CO levels,
indicating a partial saturation of the enzyme system metabolizing dichloromethane. The
PSO group means were lower among the exposed compared with the referents, among smokers
compared with nonsmokers, and among men compared with women. Given the relationship
between COHb and PSO, an expected decrease in P50 during the shift was observed among the
exposed.
Continuous 24-hour cardiac monitoring was also evaluated in a smaller sample of
24 dichloromethane-exposed workers from the triacetate fiber production plant in Rock Hill,
South Carolina, and 26 workers from the comparison plant in Narrows, Virginia. This study (Ott
et al., 1983d) was limited to white men ages >35 years. Special efforts were made to recruit men
with a history of heart disease, because this group was postulated to be most likely to
demonstrate positive findings. The estimated TWA dichloromethane exposure ranged from
60 to 475 ppm in the exposed group. The evaluation examined ventricular and supraventricular
ectopic activity and S-T segment depression in the exposed and nonexposed groups.
Comparisons were also made between cardiac performance during work hours and nonwork
hours to discern possible short-term effects of recent exposure. Comparing the findings for the
24 exposed and 26 referent volunteers indicated no difference in ventricular or supraventricular
ectopic activity or S-T-segment depression. There was no difference comparing work and
nonwork hours among exposed volunteers.
Soden et al. (1996) studied all active male workers exposed to dichloromethane at a
Hoechst Celanese triacetate film production plant in Belgium. The production process was the
same as the process at the Hoechst Celanese Rock Hill plant, except the Belgium plant was
newer with better engineering controls to significantly reduce overall levels of the
dichloromethane, acetone, and methanol used in the process. The objectives of the study were to
determine the impact of varying levels of dichloromethane exposure on COHb levels, whether
successive days of dichloromethane exposure affected the COHb levels, and what impact
smoking had on COHb levels in conjunction with dichloromethane exposure. Workers were
monitored semiannually for COHb at the end of the work shift and were personally monitored
for exposure to the three solvents. Smoking status was defined based on a health assessment
questionnaire, with smokers smoking at least one cigarette per day. Among nonsmokers, a dose
response was found among COHb levels and average dichloromethane exposure levels in the
55 DRAFT - DO NOT CITE OR QUOTE
-------
range of 7-90 ppm. The maximum COHb was 4.00% at an average exposure of 90 ppm
(correlation coefficient = 0.58,/> < 0.05). Smokers' COHb levels were elevated when compared
with those of nonsmokers with similar dichloromethane air levels, but the dose-response
correlation between dichloromethane air levels and COHb levels was weaker and not statistically
significant (correlation coefficient 0.20). The maximum COHb level for smokers was 6.35% at
an average dichloromethane air level of 99 ppm. The authors concluded that dichloromethane
exposures up to the levels observed did not produce COHb levels that are likely to cause cardiac
symptoms.
Cherry et al. (1983, 1981) reported the results of health evaluations of two studies of
triacetate film production workers. Cherry et al. (1981) recruited 46 of the 76 male workers at a
triacetate film factory, where workers were exposed to dichloromethane and methanol in a ratio
of 9:1 at air levels of dichloromethane ranging from 75 to 100 ppm. A small comparison group
(n = 12) of workers at this factory who worked a similar shift pattern (rapidly rotating shifts) but
who were not exposed to dichloromethane was also included. The men were asked whether they
had ever experienced cardiac symptoms (pain in the arms, chest pain sitting or lying, or chest
pain when walking or hurrying) and were asked about the presence in the past 12 months of
neurological disorders (frequent headaches, dizziness, loss of balance, difficulty remembering
things, numbness and tingling in the hands or feet), affective symptoms (irritability, depression,
tiredness), and stomachache (as an indicator of symptom over reporting). No difference in
response was found in history of stomachache (reported by 15% of exposed workers compared
with 17% nonexposed workers). Six of the exposed and none of the unexposed men responded
positively to the cardiac symptoms. The exposed group reported an excess of neurological
symptoms; the number (and proportion) reporting zero, one, two, and three or more symptoms
were 26 (0.56), 8 (0.17), 9 (0.20), and 3 (0.07), respectively, in exposed workers compared with
11 (0.92), 1 (0.12), 0 (0.00), and 0 (0.00), respectively, in controls (p < 0.02 for %2 test of linear
trend). With respect to affective symptoms, the number (and proportion) reporting zero, one,
two, and three symptoms were 28 (0.61), 6 (0.13), 7 (0.15), and 5 (0.11), respectively, among the
exposed workers, and 9 (0.75), 2 (0.17), 1 (0.08), and 0 (0.0), respectively, among the unexposed
workers. The authors concluded that there was no difference between exposed and nonexposed
in reporting of affective symptoms based on a y2 test of linear trend. There was no discussion of
the statistical power of this test or of tests of the proportion reporting a specified number of
symptoms (which may be a more appropriate test given the sample size), but the statistical power
of this test was very low. For example, taking the simple case of the comparison of the
proportion reporting two or more symptoms and using the approximate estimates from this study
(25 and 10% in the exposed and unexposed, respectively), EPA calculated that approximately 75
exposed and 300 unexposed workers would be needed for a power of 0.80 (i.e., an 80% chance
of rejecting the null hypothesis when the null hypothesis was false); the actual power with the
sample size of 46 and 12 is <0.10.
56 DRAFT - DO NOT CITE OR QUOTE
-------
Based on these results, a follow-up study was conducted which included a larger referent
group. This study included the symptom list described in the previous paragraph, a standardized
clinical exam (including an electrocardiograph), and neurological and psychological tests of
nerve conduction, motor speed and accuracy, intelligence, reading, and memory (Cherry et al.,
1981). Twenty-nine of the original 46 exposed workers participated in the follow-up. The men
who did not participate in the follow-up were similar in age and symptoms to the participants.
The new referent group was recruited from another plant with the same owner and a very similar
process but without dichloromethane exposure. One control, age-matched within 3 years, was
selected for each exposed worker. No differences between the groups were found in the clinical
exam, electrocardiogram, or nerve conduction tests. A statistically significant (p < 0.05) deficit
among the exposed workers was found for coarse motor speed. On two tests of overall
intelligence, the exposed group did significantly better than the referent, but on a reading ability
test designed to assess premorbid educational level, scores for the exposed group were slightly
lower than for the referent group. (Only one of these three differences, the trail making
intelligence test, was statistically significant.) With respect to the report of neurological
symptoms in the past year, the number (and proportion) reporting zero, one, two, and three
symptoms were 17 (0.59), 4 (0.14), 6 (0.21), and 2 (0.07), respectively, among the exposed
workers, and 21 (0.72), 6 (0.21), 0 (0.0), and 2 (0.07), respectively, among the unexposed
workers, with a test of linear trend that was not statistically significant. The authors interpret the
results as indicating that the differences in neurological symptoms seen in the initial study were
due to chance and that, taken as a whole, the exposed workers had no detrimental effect
attributable to dichloromethane exposure. Again, the limitations of the statistical power of the
analysis and alternative interpretations that might have resulted from approaches taken to
improve the power were not discussed. These approaches include combining the unexposed
groups from the two analyses, using the full sample of the exposed group instead of the subset of
29 who completed the clinical exam, or using a different test (i.e., of a proportion rather than a
linear trend).
Cherry et al. (1983) compared dichloromethane-exposed workers at an acetate film
factory to nonexposed workers (from the same plant but from areas without solvent contact or
from another film production factory in which solvents were not used). The 56 exposed and
36 unexposed workers were matched to within 3 years of age. Both factories were on rapid
rotating shifts. Exposure to dichloromethane ranged from 28 to 173 ppm, using individual air
sampling pumps. Blood samples were taken to monitor dichloromethane levels at the beginning
and end of the shift. Study participants were asked to rate sleepiness, physical and mental
tiredness, and general health on visual analog scales with the extreme responses at either end.
Participants were also given a digit symbol substitution test and a test of simple reaction time.
No differences were seen between exposed and unexposed groups at the beginning of the shift on
the four visual analog scales, but the exposed deteriorated more on each of the scales than did the
57 DRAFT - DO NOT CITE OR QUOTE
-------
controls. This difference in deterioration was statistically significant (p < 0.05) during the
morning shift but was not statistically significant during the afternoon or night shifts. A
significant correlation was shown between change in mood over the course of the shift and level
of dichloromethane in the blood. No difference was seen between the exposed and referents on
the tests of reaction time or digit substitution. However, among the exposed, deterioration in the
digit substitution tests at the end of the shift was significantly related to blood dichloromethane
levels.
Anundi et al. (1993) studied 12 men who worked in a graffiti-removing company. Each
worker filled out a questionnaire about previous occupational and nonoccupational exposure to
solvents and use of protective equipment. Half-day breathing zone samples were taken for each
of the 12 workers, and 15-minute samples were also taken for 10 workers. On the day the air
sampling was done, a structured interview pertaining to recent diseases or symptoms related to
allergies, asthma, diseases of the skin, respiratory organs, gastrointestinal tract, urinary organs,
neurological trauma and disease, and neuropsychiatric symptoms was conducted by a physician,
and blood and urine samples were collected. The results were compared with those of 233 men
from the area population. The 12 men (mean age 23 years) had worked between 3 months and
4.5 years cleaning graffiti from underground stations. No respiratory protection was used, and
the leather gloves were frequently soaked with solvent. While mixed solvent was used to do the
cleaning, dichloromethane was the predominant component, as confirmed by the air samples.
The geometric mean (GM) of the TWA calculated from the half-day samples was
127 mg/m3 (range 18-1,188 mg/m3), with half of the samples exceeding the Swedish permissible
exposure limit of 120 mg/m3. The GM of the 15-minute samples was 400 mg/m3 (range 6-
5,315 mg/m3), with most samples exceeding the Swedish short-time exposure limit of
300 mg/m3. Two workers had clinical laboratory data outside the normal range (urinary ai- or
p2-microglobulin, serum ALT, y-glutamyl transpeptidase), which could indicate possible kidney
and liver damage. The authors stated that in both cases, factors other than the solvent exposure
(i.e., urinary tract medical condition preceding employment, history of renal stones) could have
influenced these laboratory results. The prevalence of irritation of the eyes and upper respiratory
tract (blocked nose and nasal catarrh) was much higher in the graffiti-cleaning workers compared
with the referent group (e.g., >70% of the workers compared with 18% of the comparison group
reported a blocked nose; -50% of workers and 15% of the comparison group reported eye
irritation), but there were no or much smaller differences in abnormal tiredness, headache,
nausea, or irritative cough. No acute effects on the CNS were noted.
Studies in retired workers. Lash et al. (1991) examined the hypothesis that long-term
exposure to dichloromethane produces lasting CNS effects as measured by long-term impairment
on memory and attention centers. Retired aircraft maintenance workers employed in at least 1 of
14 targeted jobs with dichloromethane exposure for >6 years between 1970 and 1984 were
58 DRAFT - DO NOT CITE OR QUOTE
-------
compared to a like group of workers without dichloromethane exposure. The unexposed workers
were also retired aircraft mechanics at the same base and held 1 of 10 jobs in the jet shop where
little solvent was used. The exposed group, made up of painters and mechanics in the overhaul
department, was chosen to maximize the exposure contrast yet minimize differences in potential
confounders between exposed and nonexposed groups. Exposures were typically within state
and federal guidelines for dichlorom ethane exposure. From 1974 to 1986, when
155 measurements for dichloromethane exposure were made, mean breathing zone TWAs
ranged from 82 to 236 ppm and averaged 225 ppm for painters and 100 ppm for mechanics. The
mean length of retirement among the study participants was 5.3 years in the exposed group and
5.1 years in the unexposed group.
Data collection occurred in three phases: (1) an initial questionnaire was given to all
retired members of the airline mechanics union to identify eligible workers, (2) a telephone
survey was conducted to collect medical, demographic, and general employment criteria, and
(3) subjects who qualified were then recruited to participate in the medical evaluation. Sixty
percent of the 1,758 retirees responded to the questionnaire, and 259 of these retirees met the
eligibility criteria. Ninety-one men qualified for the medical evaluation based on the telephone
survey; 25 retirees exposed to solvents and 21 unexposed retirees participated in the evaluation.
All were men between the ages of 55 and 75 without a history of alcoholism or any neurological
disorder. The 25 exposed participants worked an average of 11.6 years in dichloromethane-
exposed jobs during the target period and 23.8 years in the industry.
The medical evaluation included a questionnaire about the occurrence of 33 different
symptoms in the past year, physiological measurement of odor and color vision senses, auditory
response potential, hand grip strength, and measures of reaction time (simple, choice, and
complex), short-term visual memory and visual retention, attention, and spatial ability. The only
large differences (i.e., effect size, or mean difference between groups divided by the SD of the
outcome measure, of >0.4) between the two groups were a higher score on verbal memory tasks
(effect size approximately OA5,p = 0.11) and lower score on attention tasks (effect size
approximately -0.55, p = 0.08) and complex reaction time (effect size approximately -0.40,
p = 0.18) in the exposed compared with the control group. (Although not noted by the authors,
the power to detect a statistically significant difference between the groups given this sample size
was low [i.e., approximately 0.30 for an effect size of 0.40, using a two-tailed alpha of 0.05]).
The authors investigated the possibility of response bias, given the low initial response to the
mailed questionnaire recruiting retirees and the small number of workers from the entire pool of
eligible participants who actually participated in the medical evaluation. Attempts were made to
contact 30% of the questionnaire nonrespondents, with 46% contacted and 31% completing the
telephone interview. The only difference found between those who responded to the mailed
questionnaire and those who did not was a higher percentage of diagnosed heart disease among
the nonrespondents who were 2.5 years older and had been retired 1.7 more years than the
59 DRAFT - DO NOT CITE OR QUOTE
-------
respondents. Those who were eligible but did not participate in the medical evaluation were
similar to the exam participants on all characteristics included in the interview. The only
difference was a higher prevalence of gout among the unexposed who did not participate
compared to the unexposed who did participate.
4.1.2.4. Observational Studies Using Workplace Medical Program Data
Kolodner et al. (1990) investigated the effect of occupational exposure to
dichloromethane on six health outcomes identified in the literature or based on biological
plausibility. Participants in the study were male workers at least 19 years old at two
General Electric plastic polymer plants where dichloromethane was one of the chemicals used.
Four dichloromethane exposure categories were established based on full-shift personal air
monitoring data (8-hour TWA) collected in 1979-1985, job titles, and industrial hygienists'
knowledge of plant operations. The mean 8-hour TWA and number of workers in each of the
four exposure groups were 49.0 ppm for the 19 workers in the highest, 10.9 ppm for the
49 workers in the intermediate, 3.3 ppm for the 56 workers in the low, and <1.0 ppm for the
772 workers in the minimal/no exposure group.
Data from 1984 annual medical exams and 1985 absence data from payroll records were
evaluated for possible health effects resulting from occupational exposure to dichloromethane.
A high percentage of workers participated in the annual medical exams, with only 5 of the
896 eligible for inclusion in the study refusing the exam completely in 1984. Six hypotheses
were specifically tested regarding dichloromethane exposure in relation to different health
outcomes: absence due to illness, hepatotoxicity (manifested by nausea, weakness and fatigue,
palpable liver, abdominal tenderness, jaundice, hepatomegaly, abnormal serum y-glutamyl
transferase, ALT, AST, or bilirubin), diabetes mellitus (manifested by weight loss, weakness and
fatigue, polydypsia, polyuria, impaired vision, excessive weight loss, elevated fasting blood
sugar, and abnormal urinary glucose or urinary acetone), CNS toxicity (manifested by headache,
lightheadedness, dizziness and vertigo, ataxia, weakness and fatigue, and abnormalities detected
in the central motor, central sensory, cranial nerve, gait, neurocoordination, or Bibinski reflex
examinations), cardiovascular abnormalities (manifested by fatigue, dyspnea, chest pain with
exertion, palpitations, or abnormalities detected in the point maximum impulse exam, blood
pressure measurements, or electrocardiogram), and neoplastic breast changes (154 women were
included in this portion of the study—manifested by painful breast, breast swelling, lump, nipple
discharge, or abnormalities detected in the breast examination).
Workers were placed in exposure categories based on their current jobs. In addition,
exposure to high noise levels occurred in both plants, and workers in each plant had exposure to
another chemical, either phenol or phosgene. The authors noted that workers tended to move
from entry-level jobs with high dichloromethane exposure to supervisory jobs with lower
dichloromethane exposure, based on the seniority system in place at both plants. Thus, current
60 DRAFT - DO NOT CITE OR QUOTE
-------
exposure levels reported did not necessarily reflect cumulative exposure. Because of the way the
seniority system moved workers through jobs and the fact that workers were assigned to
dichloromethane exposure categories based on their current job, age was inversely related to
exposure and was controlled in the analysis of some of the continuous variables using analysis of
covariance. Age adjustment was not employed in the analysis of dichotomous variables. The
mean age was 35.3, 39.7, 37.1, and 29.5 years in the minimal/no, low, medium3, and high
exposure groups, respectively. The small number of workers in the exposed groups limited the
ability to evaluate the effects of dichloromethane exposure on health outcomes related to age,
since age had to be adjusted in these analyses. The racial distribution did not differ among the
exposure groups.
The authors indicated the only null hypothesis that could not be accepted based on the
data was the hypothesis of CNS symptoms. However, it should be noted that the small size and
younger age distribution in the high exposure group and the lack of adjustment for age in most of
the analyses make it difficult to interpret the statistical testing that was performed. Data
pertaining to neurological, hepatic, and cardiac function are shown in Table 4-1. Among the six
neurological symptoms evaluated, a statistically significant positive exposure-effect relationship
between dizziness/vertigo and dichloromethane exposure was identified. This trend was driven
most strongly by the low frequency of this reported symptom in the minimal/no exposure group
(1.2%), but there was no linear trend across the higher levels of exposure (7.5, 2.1, and 5.3% in
the low, medium, and high exposure groups, respectively).
3The "medium" exposure group is also referred to as the "intermediate" exposure group in Kolodner et al. (1990).
61 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-1. Percentage of male General Electric plastic polymer workers
reporting neurologic symptoms or displaying abnormal values in measures
of neurological function, hepatic function, and cardiac function
Exposure group3
Minimal/no
(n = 772)
Low
(n = 56)
Medium
(n = 49)
High
(n = 19)
Neurological
Headache
Lightheadedness
Dizziness/vertigo
Ataxia
Babinski
Gait
Faintness/syncopeb
Seizures'3
Paresis/paralysis'3
Parasthesisb
Head trauma/concussion13
Peripheral motor examb'°
Peripheral sensory examb>0
Rhomberg examb'°
8.7
2.9
1.2
0.0
0.0
0.0
0.1
0.4
0.7
4.0
0.8
0.5
1.1
0.0
7.5
3.8
7.5
1.9
0.0
0.0
0.0
0.0
0.0
7.5
1.9
0.0
2.4
0.0
10.4
4.2
2.1
0.0
0.0
0.0
2.1
2.1
0.0
14.6
0.0
0.0
5.1
2.6
5.3
5.3
5.3
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Hepatic
Serum gamma glutamyl transferase
Serum total bilirubin
Serum AST
Serum ALT
8.0
3.0
1.8
9.1
16.1
1.8
3.6
10.7
12.2
2.0
4.1
8.2
5.3
10.0
0.0
5.3
Cardiac11
Palpitations: percent abnormal
1.2
9.1
2.1
0.0
Electrocardiogram
Borderline/abnormal
Bradycardia/tachycardia abnormalities13
General rhythm abnormalities
Atrial, atrioventricular, or sinus abnormalities
Bundle blocks or ventricular abnormalities
Axis deviations
Wave abnormalities
Hypertrophy
Evidence of infarction
18.5
20.2
12.0
0.8
3.9
2.6
4.0
3.8
2.3
16.7
16.7
11.1
0.0
5.6
1.9
3.7
3.7
5.6
19.1
25.5
17.0
0.0
10.6
2.1
10.6
6.4
2.1
8.3
0.0
8.3
0.0
8.3
8.3
0.0
0.0
0.0
"Mean 8-hr TWA exposure was < 1.0, 3.3, 0.9, and 49.0 ppm and mean age 35.3, 39.7, 37.1, and29.5yrs in the
minimal/no, low, medium, and high groups, respectively.
bThe authors considered these to be screening variables rather than hypothesis-testing variables.
°n = 629, 42, 39, and 14 in the minimal/no, low, medium, and high groups, respectively.
dFor all cardiac outcomes except bradycardia/tachycardia, n = 728, 54, 47, and 12 in the minimal/no, low, medium,
and high groups, respectively. For bradycardia/tachycardia, n = 727 in the minimal/no group.
Source: Kolodneretal. (1990).
62
DRAFT - DO NOT CITE OR QUOTE
-------
Soden (1993) compared health-monitoring data from dichloromethane-exposed workers
in the Rock Hill triacetate fiber production plant to workers from another plant making polyester
fibers owned by the same company in the same geographic area. Exposed and control workers
were chosen from among workers who had worked at least 10 years in their respective areas and
who participated in the company's health-monitoring program between 1984 and 1986 and were
still employed on December 31, 1986. Controls were matched by race, age, and gender to each
Rock Hill worker for a sample size of 150 and 260 in the exposed and control groups,
respectively. (The aim of the study had been 1:2 matching.) The 8-hour TWAs among the Rock
Hill workers were those reported by Lanes et al. (1990), namely 475 ppm for dichloromethane,
900 ppm for acetone, and 100 ppm for methanol. None of these exposures occurred at the
polyester plant. There was a 90% participation rate in the health-monitoring program. Six
questions in the health history portion of the health-monitoring program concerned cardiac and
neurological symptoms (chest discomfort with exercise; racing, skipping, or irregular heartbeat;
recurring severe headaches; numbness/tingling in hands or feet; loss of memory; dizziness). Part
of this program included blood samples used for standard clinical hepatic and hematologic
parameters: serum ALT, AST, total bilirubin, and hematocrit. The clinical measures were
available for 90 (60%) of the exposed and 120 (46%) of the control group; some participants
declined this part of the health-monitoring program because similar tests had been part of recent
personal medical care.
There was little difference in the frequency of reported symptoms between exposed
workers and controls: chest discomfort reported by 2.0% of exposed and 4.0% of the controls,
irregular heartbeat reported by 5.5% of exposed and 6.0% of the controls, recurring severe
headaches reported by 3.5% of exposed and 5.5% of the controls, numbness/tingling in hands
and feet reported by 6.4% of exposed and 8.1% of the controls, loss of memory reported by 1.3%
of exposed and 0.4% of the controls, and dizziness reported by 2.7% of exposed and 4.8% of the
controls (Soden, 1993). The levels of the blood values were similar in the exposed and control
groups, except for a 3.1 IU/L decrease in serum AST activity (p = 0.06). The authors concluded
that this difference was not clinically significant, but they did not discuss the potential bias
introduced by the selective participation in this part of the study.
4.1.2.5. Studies oflschemic Heart Disease Mortality Risk
Several cohort studies examined the relation between dichloromethane exposure and risk
of cardiovascular-related mortality. The methodological details of these studies are described in
Section 4.1.3.2. No evidence of increased risk of ischemic heart disease mortality was seen in
two triacetate film production cohort studies (Hearne and Pifer, 1999; Tomenson et al., 1997) or
in two triacetate fiber production cohort studies (Gibbs et al., 1996; Lanes et al., 1993).
Information on this outcome was not included in the dichloromethane analysis of civilian Air
63 DRAFT - DO NOT CITE OR QUOTE
-------
Force base workers (Radican et al., 2008; Blair et al., 1998). The standardized mortality ratios
(SMRs) for ischemic heart disease mortality were <1.0 in all of the cohorts and dose groups
examined (Table 4-2). The "healthy worker effect" may have contributed to these observations.
No case-control studies of ischemic heart disease and dichloromethane exposure were identified.
Table 4-2. Ischemic heart disease mortality risk in four cohorts of
dichloromethane-exposed workers
Obs
Exp
SMR
95% CI
Triacetate film production
Hearne and Pifer( 1999)
Tomenson et al. (1997)
Cohort 1 (men)
Cohort 2 (men)
Men
117
122
114
136.7
143.3
123.9
0.86
0.85
0.92
0.71-1.03
0.71-1.02
0.76-1.10
Triacetate fiber production
Lanes etal. (1993)
Gibbsetal. (1996)
Men and women
43
47.8
0.90
65-121
Men
50-100 ppm
350-700 ppm
96
98
100.1
106.8
0.96
0.92
0.78-1.2
0.75-1.1
Women
50-100 ppm
350-700 ppm
32
0
45.8
3.4
0.70
-
0.48-0.99
0.0-1.1
CI = confidence interval; Exp = number of expected deaths; Obs = number of observed deaths
4.1.2.6. Studies of Suicide Risk
Suicide risk is not an outcome that was a primary hypothesis or motivation of the cohort
studies, but it may be relevant given the potential neuropsychological effects of dichloromethane,
as evidenced from studies of acute and chronic exposure scenarios described previously. In a
triacetate film production cohort in Rochester, New York, Hearne and Pifer (1999) reported
14 observed deaths from suicide compared with 7.8 expected, for an SMR of 1.8 (95%
confidence interval [CI] 0.98-3.0) (Table 4-3). This cohort ("Cohort 1") consisted of 1,311 men
who were first employed between 1946 and 1970 and were followed through 1994. Similar
results were seen in a different, but somewhat overlapping, cohort in this study ("Cohort 2") of
1,013 men employed between 1964 and 1970 and followed through 1994 (see Section 4.1.3.3.1).
There was also evidence of increasing suicide risk with dichloromethane exposure, particularly
in the highest exposure group, in the study of triacetate fiber production workers in Maryland
(Gibbs, 1992). The triacetate fiber production cohort study in Rock Hill, South Carolina, has
published what appears to be erroneous information about suicide risk. In the 1993 paper (Lanes
et al., 1993), 4 observed and 5.21 expected cases were reported (SMR 0.77), but the SMR that
was reported with these data was 1.19 (95% CI 0.39-2.8). This ratio would correspond to
6 observed and around 5.2 expected cases. Information on suicide was not included in the other
64
DRAFT - DO NOT CITE OR QUOTE
-------
film and fiber cohort studies (Tomenson et al., 1997) or in the analysis of civilian Air Force base
workers (Radican et al., 2008; Blair et al., 1998). There are no case-control studies of suicide
risk and dichloromethane exposure.
Table 4-3. Suicide risk in two cohorts of dichloromethane-exposed workers
Obsa
Expb
SMR
95% CI
Triacetate film production
Hearne and Pifer( 1999)
Cohort 1
Cohort 2
14
9
7.8
5.1
1.8
1.8
0.98-3.0
0.81-3.4
Triacetate fiber production0
Gibbs (1992)
50-100 ppm
350-700 ppm
8
8
6.4
4.4
1.3
1.8
0.54-2.5
0.78-3.6
aObs = number of observed deaths.
bExp = number of expected deaths.
°One additional study provided data on suicide risk, but an error seems to be present: 4 observed and 5.21 expected
cases were reported in Lanes et al. (1993), which would be an SMR of 0.77, but the SMR reported with these data
was 1.19 (95% CI 0.39-2.8). This ratio would correspond to 6 observed and around 5.2 expected cases.
4.1.2.7. Studies of Infectious Disease Risk
There is limited information pertaining to infectious disease risk in relation to
dichloromethane exposure. Only one of the cohort studies (Hearne and Pifer, 1999) reported
data for the broad category of infectious and parasitic disease mortality. In Cohort 1 of this
analysis, there were no observed deaths in this category (5.6 expected), and in Cohort 2 there
were 3 observed and 4.7 expected deaths, for an SMR of 0.64. The detailed report by Gibbs
(1992) of the cellulose triacetate fiber production cohorts in Maryland (Gibbs et al., 1996) also
contained information on the facility in South Carolina that was the site of the report by Lanes et
al. (1993, 1990). Slightly elevated risks of mortality due to influenza and pneumonia were seen
among the male workers in the high exposure group in Maryland (7 observed, 5.62 expected,
SMR 1.25) and in South Carolina (3 observed, 1.33 expected, SMR 2.26). Among females, there
were few observed or expected cases (in Maryland, 1 observed, 0.23 expected, SMR 4.36; in
South Carolina, 0 observed, 0.74 expected). In the extended follow-up a cohort of civilian Air
Force base workers, an increased risk of mortality due to bronchitis, based on only four exposed
cases, was seen among the men who had been exposed to dichloromethane (Radican et al.,
2008). The hazards ratio adjusting for age and race, with the referent group consisting of male
workers with no chemical exposure, was 9.21 (95% CI 1.03-82.69).
4.1.2.8. Studies of Reproductive Outcomes
Pregnancy outcomes in women exposed to dichloromethane have been investigated in
two studies. Taskinen et al. (1986) studied spontaneous abortions among women employed in
65
DRAFT - DO NOT CITE OR QUOTE
-------
eight pharmaceutical factories between 1973 and 1980. Data on pregnancy outcomes were
collected from a national hospital and clinic discharge registry in Finland from 1973 to 1981 by
matching the worker rosters to the registry. Exposure to dichloromethane was one of eight
solvents or classes of solvents included in the study. The study consisted of two parts. The first
investigated the rate of spontaneous abortions (number of spontaneous abortions divided by the
sum of spontaneous abortions and births) during, before, or after employment in the
pharmaceutical industry. One hundred and forty-two spontaneous abortions and 1,179 births
were identified among the female workers at the eight plants. Employment hire and termination
dates were obtained from plant records. The spontaneous abortion rate was 10.9% during
employment compared with 10.6% before and after employment. These results compared to a
rate of 8.5% in the general population in the geographic area where the factories were located.
The rate of spontaneous abortions among workers declined over the period of the study, with a
3-year moving average of 15% at the beginning declining to 9.5% at the end of the study. Over
the same period, the industrial hygiene allegedly improved in the plants. Ten congenital
malformations of different types were identified among the women (five among those who were
employed in the pharmaceutical industry during the pregnancy and five among those whose
pregnancies occurred before or after this employment).
The second part of the study by Taskinen et al. (1986) was a case-control study of the risk
of spontaneous abortions in relation to workplace exposures during pregnancy. The source
population consisted of women who were employed in one of the eight Finnish pharmaceutical
factories during at least 1 week of the first trimester of pregnancy during the study period. Cases
(n = 44) were selected from this population based on hospital or clinic records indicating a
spontaneous abortion, and 130 controls (women who had given birth) were age-matched
(3:1 matching; age within 2.5 years) to each case. Occupational exposure data were obtained by
questionnaires completed by the plant physician or the nursing staff, blinded to the case status of
the study member, in consultation with labor protection chiefs and department foremen. The
questionnaire requested information about job history and job tasks, exposure to eight specific
solvents or classes of solvents (aliphatic solvents, alicyclic solvents, toluene, xylene, benzene,
chloroform, dichloromethane, and other solvents), antineoplastic agents, carcinogens, hormones,
antibiotics, heavy lifting, known chronic diseases, acute diseases during pregnancy, smoking
status, and previous pregnancies. Exposure frequency to each solvent was based on the
cumulative weighted sum of the number of days/week the woman was exposed to the solvent.
While overall response to the questionnaire was 93%, less than half the questionnaires contained
information about smoking or previous pregnancies, precluding inclusion of these variables in
the analysis. The distribution of broad categories of occupations (i.e., pharmaceutical workers
and packers, laboratory assistants) was similar in both groups. However, exposure to each of the
solvents was higher in the cases compared with controls, and the results for dichloromethane
were relatively strong. For dichloromethane, the prevalence of exposure was 28.9 and 14.3% in
66 DRAFT - DO NOT CITE OR QUOTE
-------
cases and controls, respectively, resulting in an odds ratio (OR) of 2.3 (95% CI 1.0-5.7). There
was also evidence of an increasing risk with higher exposure frequency, with an OR of 2.0 (95%
CI 0.6-6.6) with exposures of less than once a week and 2.8 (95% CI 0.8-9.5) with exposures of
once a week or more. An association was also seen with exposure to four or more solvents
(OR 3.4, [95% CI 1.0-12.5]). Weaker associations were seen with other specific solvents (e.g.,
chloroform, toluene).
Bell et al. (1991) investigated the relation between birth weight and maternal exposure to
airborne dichloromethane as a result of living around the triacetate film facility in Rochester,
New York. For this population-based cross-sectional study, birth certificates were obtained for
all births in 1976-1987 in Monroe County, where the triacetate film facility is located. Multiple
births and births of infants weighing <750 grams were excluded. Data abstracted from the
certificate included date of birth, census tract of residence, age, race, educational level of the
mother and father, sex, gestational age, multiple births, month of the pregnancy that prenatal care
began, total previous births, total previous live births, and conditions present during the
pregnancy. An air dispersion modeling system for 250 air emissions, including dichloromethane
and predicting average annual ground level concentrations in the surrounding community, was
used to assign dichloromethane exposure levels to each birth mother. One of four levels of
exposure was assigned to each census tract based on the isopleth of exposure in which more than
half of the census tract population resided. Because of the few births among nonwhites that
occurred in areas of higher exposure, the study was restricted to whites (n = 91,302). The
number of births that occurred in each of the four exposure levels was n = 1,085 in the high-
exposure group (50 ug/m3 [0.014 ppm]), n = 1,795 in the moderate-exposure group (25 ug/m3
[0.007 ppm]), n = 6,044 in the low-exposure group (10 ug/m3 [0.003 ppm]), and n = 82,076 in
the no-exposure group. At the levels of dichloromethane exposure in this population, no
significant adverse effect on birth weight was found. There was an 18.7 g decrease in
birthweight (95% CI -51.6 to 14.2) in the high- compared with the no-exposure group, adjusting
for maternal age, maternal education, parity, previous pregnancy loss, late start of prenatal care,
sex of the child, and pregnancy complications. No significant association was found between
any combination of exposure levels and birth weight. There was no association between
exposure group and risk of a low birthweight infant (i.e., <2,500 g, OR 1.0 [95% CI 0.81-1.2] in
the high- compared with the no-exposure group). The authors point out a number of problems
with assignment of dichloromethane exposure. It is possible that the dichloromethane exposure
was overestimated using the model. Comparisons to ambient air sampling levels collected
6 times/year resulted in the dichloromethane exposure derived from the model being twice as
high as the ambient air samples. There was also inaccuracy in the assignment of
dichloromethane exposure level to each birth because the exposure assignment was made using
the predominant value of the isopleth for a census tract.
67 DRAFT - DO NOT CITE OR QUOTE
-------
Two studies have investigated the occurrence of oligospermia among men occupationally
exposed to dichloromethane exposure. Kelly (1988) studied 34 men employed in an automotive
plant as bonders, finishers, and press operators. These men were self-referred to a health center
for a variety of complaints, including neurological symptoms, musculoskeletal symptoms, and
shortness of breath. Twenty-six of the men were bonders and eight were finishers or press
operators. The job as bonder consisted of dipping hands into an open bucket of dichloromethane
and splashing it onto plastic automobile parts. The dichloromethane exposure for bonders
averaged 68 ppm with a range of 3.3-154.4 ppm. Eight men, all of whom were bonders,
reported symptoms of testicular and epididymal tenderness, with confirmation on medical exam.
They ranged in age from 20 to 47 years old and had been bonders for up to 2.9 years. The COHb
levels for the eight workers with genital symptoms ranged from 1.2 to 17.3%, with an average of
6.9% anywhere from 4 to 90 hours postexposure. The COHb levels for the two men who
smoked were among the highest, namely 7.3 and 17.3%. Four of the eight workers agreed to
provide semen samples; their sperm counts were 2-26 x 106/cm3. The authors stated that men
with sperm counts as low as 25 x 106/cm3 may still be fertile, but none of these men had had any
children since working with dichloromethane despite not using contraceptives. There was one
miscarriage. All four men reported dipping their hands into open buckets of dichloromethane
without any protective equipment, and two men reported feeling dizzy, giddy, and high at work.
Based on the results of the Kelly (1988) case report, Wells et al. (1989) planned to do a
study of oligospermia among 20 exposed workers and 20 unexposed workers to
dichloromethane. The exposed workers were unvasectomized men who had worked for the
3 months prior to recruitment in furniture stripping shops. Eleven men were recruited from
among 14 eligible workers at six different shops where dichloromethane was utilized. Names of
acquaintances of the exposed were solicited as potential referents. Only one exposed man
provided any names. Therefore, the study was redirected as a case report on the 11 exposed
men. The mean TWA dichloromethane exposure was 122 ppm (range 15-366 ppm) with a mean
COHb of 5.8% (range 2.2-13.5%). The mean COHb for smokers, 10.2% (range 8.1-13.5), was
higher than for nonsmokers, 3.9% (range 2.2-5.9), and the nonsmoker levels were higher than
the 2% level considered to be the upper limit of normal in nonsmoking populations. The mean
sperm count was 54 x 106/cm3 (range 23-128 x 106/cm3) compared to a population value of
47 x 106/cm3 for the same geographic area based on samples analyzed at the same laboratory.
Using the standard definition for oligospermia of 20 x 106/cm3, none of the 11 workers had
oligospermia.
4.1.2.9. Summary of Noncancer Studies
The clinical and workplace studies of noncancer health effects of dichloromethane
exposure have examined markers of disease and specific clinical endpoints relating to cardiac,
neurological disease, hepatic function, and reproductive health.
68 DRAFT - DO NOT CITE OR QUOTE
-------
Cardiac effects. The effect of dichloromethane on the formation of COHb (Stewart et al.,
1972b) raised concerns about potential risk of cardiovascular damage. To date, there is little
evidence of cardiac damage related to dichloromethane exposure in the cohort studies of
dichloromethane-exposed workers that examined ischemic heart disease mortality risk (Hearne
and Pifer, 1999; Tomenson et al., 1997; Gibbs et al., 1996; Lanes et al., 1993) or in two small
cardiac monitoring studies (Ott et al., 1983d; Cherry et al., 1981). However, limitations in these
studies should be noted, including the healthy worker effect and the absence of data pertaining to
workers who died before the establishment of the analytic cohort (Gibbs et al., 1996; Gibbs,
1992).
Neurological effects. The acute effects of dichloromethane exposure on neurological
function seen in numerous case reports have also been established in experimental studies in
humans (Putz et al., 1979; Winneke, 1974; Stewart et al., 1972a, b). Relatively less is known
about the potential long-term effects of chronic exposures in humans. Some data from studies of
workers suggest that the effects of dichloromethane are relatively short-lived. For example, in
the study by Cherry et al. (1983) of 56 exposed and 36 unexposed workers, alterations in mood
or in digit substitution test results were seen during the course of a work shift but were not seen
at the beginning of a shift. No difference in four neurological symptoms was seen in an analysis
of exposed workers (average exposure 475 ppm, > 10-year duration) and an unexposed
comparison group by Soden (1993). Other data suggest an increase in prevalence of
neurological symptoms among workers (Cherry et al., 1981) and possible detriments in attention
and reaction time in complex tasks among retired workers (Lash et al., 1991). These latter two
studies are limited by the small sample size. Thus, Cherry et al. (1981) and Lash et al. (1991)
have low power for detecting statistically significant results and consequently should not be
interpreted as definitive analyses showing no effects. Rather, these analyses provide evidence of
an increased prevalence of neurological symptoms among workers with average exposures of
75-100 ppm (Cherry et al., 1981) and long-term effects on specific neurological measures (i.e.,
attention and reaction time) in workers whose past exposures, at least for part of their work
history, were in the 100-200 ppm range (Lash et al., 1991). The increased risk of suicide
(approximately a twofold increased risk) seen in two of the worker cohort studies (Hearne and
Pifer, 1999; Gibbs, 1992) is an additional indication of potential neurological consequences of
dichloromethane exposure. Adequate studies addressing these specific issues are not available.
Thus, given the suggestions from the currently available studies, the statement that there are no
long-term neurological effects of chronic exposures to dichloromethane cannot be made with
confidence.
69 DRAFT - DO NOT CITE OR QUOTE
-------
Hepatic effects. Three studies provide data pertaining to markers of hepatic damage (i.e.,
serum enzymes and bilirubin levels) (Soden, 1993; Kolodner et al., 1990; Ott et al., 1983c). Two
of these studies were based in the Rock Hill, South Carolina, cellulose triacetate fiber plant
(Soden, 1993; Ott et al., 1983c), with the most recent of the studies focusing on workers with
>10 years duration in a high exposure area (average exposure estimated as 475 ppm). There is
some evidence of increasing levels of serum bilirubin with increasing dichloromethane exposure
in Ott et al. (1983c) and Kolodner et al. (1990), but there are no consistent patterns with respect
to the hepatic enzymes examined (serum y-glutamyl transferase, serum AST, serum ALT).
These studies do not provide clear evidence of hepatic damage in dichloromethane-exposed
workers, to the extent that this damage could be detected by these serologic measures; however,
these data are limited and thus the absence, presence, or extent of hepatic damage is not known
with certainty.
Infectious disease and immune-related effects. Only limited and somewhat indirect
evidence pertaining to immune-related effects of dichloromethane in humans is available. No
risk was seen in the broad category of infectious and parasite-related mortality reported by
Hearne and Pifer (1999), but there was some evidence of an increased risk for influenza and
pneumonia-related mortality at two cellulose triacetate fiber production work sites in Maryland
and South Carolina (Gibbs, 1992). In the Maryland facility, an increased risk of cervical cancer
was seen among the 938 female workers, with an SMR of 3.0 (95% CI 0.96-6.9) in the 50-100
ppm group and 5.4 (95% CI 0.13-30.1) in the 350-700 ppm group (Gibbs et al., 1996). Cervical
cancer is viral mediated (human papilloma virus), and immunosuppression is a risk factor for
development of this disease, as seen by the increased risk in immunocompromised patients and
people taking immunosuppressant medications (Leitao et al., 2008; Ognenovski et al., 2004). In
a cohort study of civilian Air Force base workers, an increased risk of bronchitis-related
mortality, based on four exposed cases, was seen among the men who had been exposed to
dichloromethane, with a hazard ratio of 9.21 (95% CI 1.03-82.69) (Radican et al., 2008). This
collection of studies indicates that immune suppression, and a potentially related susceptibility to
specific types of infectious diseases, may be a relevant health outcome for consideration with
respect to dichloromethane exposure.
Reproductive effects. Studies pertaining to various reproductive effects and
dichloromethane exposure from workplace settings (Wells et al., 1989; Kelly, 1988; Taskinen et
al., 1986) or environmental settings (Bell et al., 1991) have examined possible associations with
spontaneous abortion (Taskinen et al., 1986), low birth weight (Bell et al., 1991), and
oligospermia (Wells et al., 1989; Kelly, 1988). Of these, the data pertaining to spontaneous
abortion provide the strongest evidence of an adverse effect of dichloromethane exposure,
particularly with respect to the case-control study in which the strongest association was seen
70 DRAFT - DO NOT CITE OR QUOTE
-------
specifically with the higher frequency category of dichloromethane exposure. However, it is a
small study (44 cases, 130 controls) with limited quantitative exposure assessment and multiple
exposures (although the association seen with dichloromethane was among the highest seen
among the solvents) and so cannot be considered to firmly establish the role of dichloromethane
in induction of miscarriage. Nevertheless, the high exposure scenario, including the potential for
substantial dermal exposure in the study of Kelly (1988), also suggests the potential for adverse
male reproductive effects.
4.1.3. Cancer Studies
4.1.3.1. Identification and Selection of Studies for Evaluation of Cancer Risk
Seventeen epidemiologic studies of cancer risk were identified and included in this
evaluation: four cohorts for which the primary solvent exposure was to dichloromethane (two in
film production settings and two in cellulose triacetate fiber production), one large cohort of
civilian employees at a military base with exposures to a variety of solvents but that included an
assessment specifically of dichloromethane exposure, and twelve case-control studies of specific
cancers with data on dichloromethane exposure. One additional study (Ott et al., 1985), a cohort
of 1,919 men employed at Dow Chemical facilities, was identified but was not included in the
summary. The analysis was based on exposure to a combined group of chlorinated methanes
(e.g., carbon tetrachloride, chloroform, methyl chloride, and dichloromethane), and it was not
possible from the data presented to assess the individual effects of dichloromethane.
4.1.3.2. Description of the Selected Studies
In this section, the study setting, methods (including exposure assessment techniques),
results pertaining to incidence of mortality from specific cancers, and a brief summary of
primary strengths and limitations are summarized for each of the identified studies. When two
papers of the same cohort were available, the results from the longer period of follow-up are
emphasized in the summary. Information from earlier reports is used when these reports contain
more details regarding working conditions, study design, and exposure assessment. The
description of individual studies is followed by a summary of the evidence available from these
studies relating to specific types of cancer.
4.1.3.3. Cellulose Triacetate Film Base Production Cohorts
4.1.3.3.1. Cellulose triacetate film base production—Rochester, New York (Eastman Kodak).
Friedlander et al. (1978) reported a cohort mortality study of workers in an Eastman Kodak
facility in Rochester, New York. This study was expanded and extended several times during
the next 20 years (Hearne and Pifer, 1999; Hearne et al., 1990, 1987). The latest analysis
provided data on two overlapping cohorts. The first cohort (Cohort 1) consisted of 1,311 male
workers employed in the roll coating division (n = 1,070) or the dope and distilling departments
71 DRAFT - DO NOT CITE OR QUOTE
-------
(n = 241) of the Eastman Kodak facility in Rochester, New York. Men who began working in
these areas after 1945 and were employed in these areas for at least 1 year (including seasonal or
part-time work that equaled 1 full-time year equivalent) from 1946 to 1970 were included.
Follow-up time was calculated from the end of the first year of employment in the study area
through December 31, 1994. The mean duration of work in Cohort 1 was 17 years. The total
number of person-years of follow-up was 46,112, and the mean duration of follow-up was
35.2 years (range 25-49 years). The second cohort (Cohort 2) included 1,013 male workers in
the roll coating division who were employed for at least 1 year in this division between 1964 and
1970. Follow-up time was calculated from January 1, 1964, for those who were employed there
before 1964, or the date of first employment in the roll coating division for those who began in
1964 or later. Follow-up continued through December 31, 1994. The mean duration of work in
Cohort 2 was 24 years. Total follow-up time was 26,251 person-years, and the mean duration of
follow-up was 25.9 years (range 25-31 years). Cohort 2 was the focus of previous analyses by
Friedlander et al. (1978) and Hearne et al. (1990, 1987).
For both cohorts, causes of death were based on the underlying causes of death recorded
on the death certificates, which were routinely obtained by the company for the processing of life
insurance claims. The expected number of deaths was calculated using appropriate age-, sex-,
calendar time-, and cause-specific death rates for men in New York State (excluding New York
City). In addition, another referent group was also used in the analysis of the second cohort.
This other referent was based on the age-, sex-, calendar time-, and cause-specific death rates of
other hourly male workers employed at the Eastman Kodak plant in Rochester, New York. (An
internal referent group was also described for Cohort 1, but data for that analysis were not
presented.)
Dichloromethane was first used in the film production process at the Eastman Kodak
facility around 1944 (Hearne et al., 1987). Cellulose triacetate was dissolved in dichloromethane
and then cast into a thin film onto revolving wheels. The film was then cured by circulating hot
air in the coating machines, and the solvent was recovered and redistilled. 1,2-Dichloropropane
and 1,2-dichloroethane were also used as solvents from the 1930s to the 1960s, but
dichloromethane was predominant (ratio 17:2:1 for dichloromethane:l,2-dichloropropane:
1,2-dichloroethane in general workplace air measurements) (Hearne et al., 1987).
The exposure assessment in the Rochester, New York, Eastman Kodak cohort studies
was based on employment records (start and stop dates for specific jobs in the relevant areas of
the company) in combination with air monitoring data used to estimate the exposure level for a
given job, location, and time period (Hearne et al., 1987). Air monitoring began in the 1940s,
but few data are available before 1959. In the most recent update (Hearne and Pifer, 1999), more
than 1,500 area and 2,500 breathing zone air samples were used in the exposure assessment
process. Reductions in exposures in the dope department and the distilling department began
after 1965. The highest exposure jobs were operator and maintenance workers (dope
72 DRAFT - DO NOT CITE OR QUOTE
-------
department) and filter washing and waste operator (distilling department), with estimated 8-hour
TWA exposures of 100-520 ppm between 1946 and 1985. There was little change in estimated
exposures for jobs in the roll coating division from the 1940s through 1985, but some reduction
was seen from 1986 to 1994. The mean 8-hour TWA exposures were 39 ppm for Cohort 1 and
26 ppm for Cohort 2. These data were used to estimate a cumulative exposure index (i.e., the
summation across all jobs held by an individual of the product of the average dichloromethane
concentration as ppm and the duration of employment in that job). The authors refer to this as a
"career exposure index." Additional adjustment in these estimates was made for respiratory
protection, but the details of this adjustment were not described. For Cohort 1, the cumulative
exposure categories used in exposure-effect analyses were <150, 150-349, 350-799, and
>800 ppm. For Cohort 2, the cumulative exposure categories were <400, 401-799, 800-1,199,
and >1,200 ppm. The cut points were chosen to produce an approximately equal number of
expected total deaths in these categories.
There was no increased risk of mortality for all sites of cancer or for lung cancer in either
cohort analysis (Table 4-4). Data pertaining to smoking history, obtained from a survey of
workers in the New York film production facility, indicate that smoking rates were similar in the
exposed group, the internal comparison group, and the general population; therefore, it is
unlikely that differences in smoking could be masking an effect of dichloromethane.
73 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-4. Mortality risk in Eastman Kodak cellulose triacetate film base production workers, Rochester, New
York
Cancer type
Cancer, all sites
Liver3
Pancreas
Lung3
Brain3
Lymphatic system
Non-Hodgkin's
Hodgkin's
Multiple myeloma
Leukemia
Cohort 1:
1,311 men employed 1946-1970, followed through 1994
New York referent group
Obsb
93
1
5
27
6
5
2
2
1
8
Expb
105.8
2.4
5.5
36.0
2.8
6.6
4.1
1.1
1.5
3.9
SMR
0.88
0.42
0.92
0.75
2.16
0.75
0.49
1.82
0.68
2.04
95% CI
0.71-1.08
0.01-2.36
0.30-2.14
0.49-1.09
0.79-4.69
0.24-1.78
0.06-1.76
0.20-6.57
0.01-3.79
0.88-4.03
Cohort 2:
1,013 men employed 1964-1970, followed through 1994
New York referent group
Obs
91
1
8
28
4
6
3
2
1
6
Exp
102.0
2.4
5.3
34.2
2.1
5.7
3.5
0.6
1.5
3.5
SMR
0.89
0.42
1.51
0.82
1.88
1.06
0.85
3.13
0.65
1.73
95% CI
0.72-1.10
0.01-2.33
0.65-2.98
0.55-1.19
0.51-4.81
0.39-2.30
0.17-2.50
0.35-11.30
0.01-3.62
0.63-3.76
Kodak referent group
Exp
94.7
1.8
5.1
31.3
2.7
5.7
3.6
0.9
1.3
4.4
SMR
0.96
0.55
1.55
0.89
1.46
1.05
0.84
2.23
0.79
1.37
95% CI
0.77-1.18
0.01-3.07
0.67-3.06
0.59-1.29
0.39-3.75
0.38-2.28
0.17-2.46
0.25-8.05
0.01^.39
0.50-2.98
3Liver includes liver and biliary duct; lung includes lung, trachea, and bronchus; brain includes brain and CNS.
bObs = number observed deaths, Exp = number of expected deaths.
Source: Hearne and Pifer (1999).
74
DRAFT - DO NOT CITE OR QUOTE
-------
The only specific sites for which the SMRs were >1.0 in both cohorts were brain and
CNS cancer, Hodgkin's lymphoma, and leukemia. Pancreatic cancer mortality risk was
increased in Cohort 2 but not in Cohort 1. None of these associations were statistically
significant, and the Hodgkin's lymphoma observations were based on a total of only four cases
in both cohorts combined, and so were imprecise. Within Cohort 2, there was little difference in
results for most sites using the different referent groups, but the point estimates for the SMRs for
brain and CNS cancer, Hodgkin's lymphoma, and leukemia were somewhat higher using the
New York State referent group compared with the internal Eastman Kodak referent group. An
attenuation of the dichloromethane association seen in the analyses using the internal Kodak
referent group would be expected if the risk of specific cancers was increased in this comparison
group, possibly because of other workplace exposures.
The authors presented the exposure-effect analysis based on the estimated cumulative
dichloromethane exposure groups for all sites of cancer, pancreatic cancer, lung cancer, brain
cancer, and leukemia (Table 4-5). There is no evidence of an exposure-effect for all site cancer
mortality or lung cancer mortality risk. The relatively sparse number of deaths for the other
specific cancer types makes it difficult to interpret the data. The patterns for pancreatic cancer
differ between the two cohorts, with increased risk at the higher dose in Cohort 1 and a U-shaped
curve seen in Cohort 2. For brain cancer mortality, a higher SMR was seen in the groups with
cumulative exposure levels of >800 ppm-years compared with lower exposure groups. For
leukemia in both cohorts, an increased mortality risk is seen in the highest exposure group (mean
approximately 1,700 ppm-years).
75 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-5. Mortality risk by cumulative exposure in Eastman Kodak
cellulose triacetate film base production workers, Rochester, New York
Cohort, cancer,
referent group
SMRs (number of observed deaths)
Cohort la
Cumulative exposure
(ppm yrs)
Cancer, all sites
Internal
New York
Pancreas
Internal
New York
Lungb
Internal
New York
Brainb
Internal
New York
Leukemia
Internal
New York
<150
0.81 (20)
0.67
0.74 (1)
0.68
0.78 (5)
0.52
0.58(1)
1.10
0.83 (2)
1.61
150-349
1.02(19)
0.93
0.00 (0)
0.00
1.07 (6)
0.90
0.78(1)
1.77
0.00 (0)
0.00
350-799
1.10(28)
0.95
0.77 (1)
0.65
1.25 (9)
0.86
1.65 (3)
3.99
0.48(1)
0.98
>800
1.07 (26)
1.00
2.34 (3)
2.18
0.90 (7)
0.77
0.85 (1)
1.78
2.73 (5)
5.79
Cohort T
Cumulative exposure
(ppm yrs)
Cancer, all sites
Internal
New York
Pancreas
Internal
New York
Lungb
Internal
New York
Brainb
Internal
New York
Leukemia
Internal
New York
<400
0.89(18)
0.76
2.58 (4)
2.86
0.95 (6)
0.80
0.00 (0)
0.00
0.00 (0)
0.00
400-799
0.96 (33)
0.93
0.00 (0)
0.00
1.15(12)
1.00
1.13 (2)
2.02
0.84 (2)
1.26
800-1,199
1.11(23)
1.13
0.95 (2)
1.83
0.94 (6)
0.89
1.37(1)
1.75
0.75 (1)
1.10
>1,200
1.08(17)
1.12
1.43 (2)
2.67
0.82 (4)
0.79
1.49(1)
2.50
2.70 (3)
4.84
"Cohort 1: 1,311 men employed 1946-1970 in the roll coating division, dope department, or distilling department,
followed through 1994; mean exposure (cumulative exposure yrs) 66, 244, 543, and 1,782 ppm-yrs in the four dose
groups, respectively.
bLung includes lung, trachea, and bronchus; brain includes brain and CNS.
°Cohort 2: 1,013 men employed 1964-1970 in the roll coating division, followed through 1994; mean exposure
(cumulative exposure yrs) 168, 581, 981, and 1,670 ppm-yrs in the four dose groups, respectively.
Source: Hearne and Pifer (1999).
A strength of the Eastman Kodak cohort studies was the sampling data for
dichloromethane that allowed an assessment of each worker's exposure using the monitoring
data and the worker's job history, making exposure-effect analyses possible. Follow-up of the
76
DRAFT - DO NOT CITE OR QUOTE
-------
vital status of the cohort was >99% (Hearne and Pifer, 1999). There was also some information
on smoking history for workers in the plant, based on a survey conducted in 1986 (Hearne et al.,
1987). A difficulty in interpreting the data, however, is that there was some overlap between the
cohorts: 707 of the men were included in both Cohort 1 and Cohort 2. Data are not presented in
a way that would allow the reader to eliminate duplicate cases and person-years so that cases are
only counted once when examining both cohorts. A strength of the Cohort 1 sampling strategy,
compared with that of Cohort 2, is that Cohort 1 is limited to workers who began work at the
plant after 1945. These workers would not have had workplace exposure to methanol and
acetone, which were used at the plant in the film production process prior to that time. Also,
follow-up began with the beginning of employment in the relevant area. In contrast,
Cohort 2 was limited to workers who were employed from 1964 to 1970, so exposed workers
who left or died before 1964 were not included. The relatively small number of cases with
specific low incidence cancers (e.g., brain, leukemia) is also a limitation of the analyses of both
of the cohorts in this study. In addition, the exposure levels in both cohorts (mean 8-hour TWA
39 and 26 ppm in Cohorts 1 and 2, respectively) is relatively low compared with values seen in
other workplaces, including the cellulose triacetate fiber production cohorts described in Ott et
al. (1983a) and Gibbs et al. (1996). Also, the outcome assessment is based on mortality
(underlying cause from death certificates) rather than incidence data, and, because the Kodak
studies were limited to men, there is no information on risk of breast cancer or other female
reproductive cancers.
4.1.3.3.2. Cellulose triacetate film base production—Brantham, United Kingdom (Imperial
Chemical Industries). Tomenson et al. (1997) reported the results of a retrospective cohort
mortality study of 1,473 men who worked at a film-base production facility in Brantham,
England, anytime between 1946 and 1988 in jobs that were considered to have dichloromethane
exposure. The start of the follow-up period was not specified by the authors but is likely to have
been 1946 or the date of first employment at the plant. Follow-up of the cohort continued
through December 31, 1994, and vital status was based on national records (United Kingdom
National Health Service Central Register and the Department of Social Security). Cause of death
was based on the underlying causes of death recorded on the death certificates. The expected
number of deaths was calculated using age-, sex-, calendar time-, and cause-specific death rates
for England and Wales. In addition, a comparison using mortality rates for the local areas
(Tendring and Samford) for 1968-1978 and analyses limited to workers who had been employed
for at least 3 months were also made, but the results of these analyses were not presented. The
mean duration of work in the cohort was 9 years, the total number of person-years was 39,759,
and the mean duration of follow-up was 27.0 years (7-49 years).
This facility produced cellulose diacetate film from 1950 to 1988, with other types of
films also manufactured beginning in the 1960s. Dichloromethane was the solvent used in this
77 DRAFT - DO NOT CITE OR QUOTE
-------
process, and exposure occurred in the production of the triacetate film base and the casting of the
film into rolls. The exposure assessment was based on >2,700 personal or air monitoring
samples collected since 1975. An exposure matrix was constructed, assigning jobs to 1 of
20 work groups with similar exposure potential for each of four different time periods (before
1960, 1960-1969, 1970-1979, and 1980-1988). For the 1980-1988 period, exposure estimates
for specific jobs were based on about 330 personal monitoring samples. For the earlier time
periods, information about work tasks and location was used in combination with the information
about the number of, use of, speed of, and problems with casting machines at different times
from their initial introduction in 1950. The highest exposures were estimated to be in the casting
machine operators and cleaners. Lifetime cumulative exposure to dichloromethane was
calculated as the product of the mean level of exposure for the assigned work group and the
duration of employment in that job summed across all jobs. Three categories of cumulative
exposure were used for the analysis of ever-exposed workers: <400, 400-700, and >800 ppm-
years. Approximately 30% of the workers in the cohort were classified as "unassigned" for the
calculation of exposure group because sufficient information needed to determine exposures (i.e.,
the location and tasks assigned to laborers and maintenance workers) was not available. The
mean 8-hour TWA exposure was estimated at 19 ppm for the cohort.
There was no increased risk of mortality for all sites of cancer (Table 4-6), and the SMRs
for most of the specific cancer sites examined (stomach, colon, rectum, liver, pancreas, lung, and
prostate) were <1.0. The only specific sites for which there was an increased SMR (i.e., 1.1 or
higher) were brain and CNS cancer and leukemia, and these estimates were based on few (less
than five) observed cases (Table 4-6). Tomenson et al. (1997) present the exposure-effect
analysis based on the estimated cumulative dichloromethane exposure groups for all sites of
cancer, pancreatic cancer, and lung cancer, and there is no evidence of an increasing effect with
increasing exposure level in these analyses. A formal exposure-effect analysis for brain cancer
or leukemia was not presented. However, the authors described two of the brain cancer cases as
having "minimal" exposure to dichloromethane (and thus presumably would have been in the
<400 ppm-year cumulative exposure group). One case was estimated as having 572 ppm-years
cumulative exposure, and the other case was an electrician classified in the unassigned exposure
group. He had worked for 21 years at an exposure level "that was unlikely to have exceeded
15 ppm 8-hour TWA."
78 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-6. Mortality risk in Imperial Chemical Industries cellulose
triacetate film base production workers, Brantham, United Kingdom:
1,473 men employed 1946-1988, followed through 1994
Cancer type
Cancer, all sites
Liver and biliary duct
Pancreas
Lung, trachea, bronchus
Brain and CNS system
Lymphatic and hematopoietic
Leukemia
Observed
68
0
3
19
4
6
3
Expected"
104.6
1.5
4.4
41.3
2.8
7.1
2.7
SMR
0.65
-
0.68
0.46
1.45
0.85
1.11
95% CI
0.51-0.82
-
0.14-1.99
0.29-0.75
0.40-3.72
0.31-1.84
0.23-1.84
"Expected, calculated from observed and SMR data reported by the authors by using the following formula:
expected = 100 x observed •*• SMR; SMRs and CIs were not calculated for categories with zero observed cases.
Source: Tomenson et al. (1997).
A strength of this study was the monitoring data available that allowed assignment of
cumulative exposure categories for use in exposure-effect analyses. However, 30% (439) of
exposed workers had insufficient work histories to determine lifetime cumulative exposure. Air
measurements were not available until 1975, and personal measures were not available until
1980. In addition, the duration of exposure was relatively low (mean, 9 years), the mean
exposure level was relatively low (mean 8-hour TWA, 19 ppm), and there were very few deaths
from specific types of cancer, which limit the statistical power of the study to examine
associations among dichloromethane and specific cancers. Other limitations, as were also noted
in the Kodak cohort studies, include the use of mortality rather than incidence to define risk, the
reliance solely on underlying causes of death from death certificates to classify specific cancer
types, and the lack of information on breast cancer risk.
4.1.3.4. Cellulose Triacetate Fiber Production Cohorts
4.1.3.4.1. Cellulose triacetate fiber production—Rock Hill, South Carolina (Hoechst Celanese
Corporation). Two cohorts of cellulose triacetate fiber workers have been studied in Rock Hill,
South Carolina (Lanes et al., 1993, 1990; Ott et al., 1983a, b), and Cumberland, Maryland (Gibbs
et al., 1996; Gibbs, 1992). Workers were exposed to dichloromethane, methanol, and acetone in
both facilities.
Ott et al. (1983a, b) conducted a retrospective cohort mortality study of 1,271 acetate
fiber production workers (551 men and 720 women) employed at least 3 months from 1954 to
1977 at Dow Chemical Company, Rock Hill, South Carolina. This analysis focused on ischemic
heart disease mortality risk, and there was no presentation of cancer risk. The Rock Hill cohort
study was updated through September 30, 1986 (Lanes et al., 1990), and December 31,
1990 (Lanes et al., 1993), and analyses of cancer mortality risks were included in these later
79
DRAFT - DO NOT CITE OR QUOTE
-------
reports. Causes of death information was obtained from death certificates with coding based on
the underlying and contributing causes (Ott et al., 1983a). The referent used in the updates was
the general population of York County, South Carolina, and analyses were adjusted for age, race,
gender, and calendar period. Because the results of the mortality risk analyses were similar for
both updates, those from the 1993 paper are presented here. The mean duration of work in the
cohort was not reported, but 56% worked for <5 years (calculated from Tables 3 and 4 of Ott et
al., 1983b). The mean duration of follow-up was 23.6 years in the analysis through 1986 (Lanes
et al., 1990) but was not reported in the later paper (Lanes et al., 1993). The 1993 report added
approximately 4.25 years of follow-up, which would result in an estimate of approximately
28 years of follow-up for this report.
The Rock Hill, South Carolina, plant began producing cellulose triacetate fiber in 1954.
Dichloromethane was used as the solvent for the initial mixing with cellulose triacetate flakes.
This mixture was then filtered and transferred to the extrusion area for drying and winding. Air
measurements taken in 1977-1978 were assumed to be representative of operations since
dichloromethane use began in 1954, based on review of processing operations. The median
8-hour TWA exposures were estimated at 140, 280, and 475 ppm in the low, moderate, and high
categories of exposure (Ott et al., 1983a). Employment records provided information on jobs
held and dates employed, and this was used in conjunction with the exposure estimates for
specific jobs and work areas to classify individual exposures. However, detailed work history
information was only available for 475 (37%) of the workers (Lanes et al., 1990), and it is not
clear how the exposure assessment was applied to workers with missing job history data.
Methanol was also used in the cellulose triacetate fiber production process, and methanol
exposure was estimated as 1/10 that of dichloromethane. Acetone exposure was used in the
production of acetate (cellulose diacetate) fiber at an adjacent part of the plant. The exposure to
acetone was inversely related to that of dichloromethane, with estimated median 8-hour TWAs
of 1,080 ppm acetone in the low dichloromethane exposure group and 110 ppm acetone in the
moderate and high dichloromethane groups in the Rock Hill plant (Ott et al., 1983a).
In the latest follow-up (Lanes et al., 1993), there was no increase in mortality risk from
cancer (all sites) or from cancer of the lung or pancreas (Table 4-7). The SMR for liver and bile
duct cancer, based on four observed cases, was 2.98 (95% CI 0.81-7.63). This was lower than
the SMR of 5.75 (95% CI 1.82-13.8) that was reported in the 1990 analysis based on these same
four cases but on a shorter follow-up period (and thus lower number of expected cases). Three
of these cases were bile duct cancers. This was the first cohort study that included women, and
this study provided data on breast cancer risk. There were 3 observed breast cancer deaths
compared with 5.59 expected, yielding an SMR of 0.54 (95% CIO. 11-1.57). No data were
provided pertaining to reproductive risk factors (e.g., pregnancy history) for breast cancer among
the women in this cohort, so it is difficult to assess whether these potential confounders are likely
to have been distributed differently in the cohort compared with the referent group. Information
80 DRAFT - DO NOT CITE OR QUOTE
-------
about brain cancer, Hodgkin's lymphoma, and leukemia (Table 4-7) was not included in this
report but was included in the report by Gibbs (1992) (see Table 11 of that report).
Table 4-7. Mortality risk in Hoechst Celanese Corporation cellulose
triacetate fiber production workers, Rock Hill, South Carolina: 1,271 men
and women employed 1954-1977, followed through 1990
Cancer type
Cancer, all sites
Liver and biliary duct
Pancreas
Lung, trachea, bronchus
Brain and CNSC
Hodgkin's lymphoma0
Leukemia0
Breast cancer (women)
Observed
39
4
2
13
1
0
1
o
J
Expected
47.7
1.34
2.42
16.21
1.5
0.24
1.11
5.59
SMRa
0.82
2.98
0.83
0.80
0.67
-
0.90
0.54
95% CPb
0.58-1.52
0.81-7.63
0.10-2.99
0.43-1.37
0.2-3.71
-
0.02-5.0
0.11-1.57
aSMRs and CIs were not calculated for categories with zero observed cases.
bCIs were calculated from Breslow and Day (1987, Table 2.10).
°Data for brain and CNS cancer, Hodgkin's lymphoma, and leukemia in the Rock Hill study were reported in Gibbs
(1992).
Source: Lanes et al. (1993), except as noted in footnote c.
There are a number of limitations in this study including the small size of the cohort,
small number of observed cancer deaths, availability of detailed work history information for
only 37% of the workers, and use of mortality rather than incidence data. The exposure levels at
this plant were high, but the duration of exposure for most of the cohort was relatively short
(<5 years). It is the first cohort study, however, that included women and provided information
on breast cancer risk.
4.1.3.4.2. Cellulose triacetate fiber production—Cumberland, Maryland (Hoechst Celanese
Corporation). Gibbs et al. (1996) studied a cohort of 2,909 cellulose triacetate fiber production
workers (1,931 men and 978 women) at a Hoechst Celanese plant in Cumberland, Maryland.
This retrospective cohort mortality study included all workers who were employed on or after
January 1, 1970, and who worked at least 3 months. This study also included a very small
comparison group (256 men, 46 women) that was described as a "0" or "no" exposure group of
workers at the plant who worked in jobs that were not considered to have had dichloromethane
exposure; totals for this study were 2,187 men and 1,024 women in the exposed and nonexposed
groups combined.
The plant closed in 1981, and mortality was followed through 1989. Since 1955,
employees of this plant were exposed to dichloromethane, methanol, acetone, and finishing oils
used as lubricants. Before 1955, acetone was the only exposure. Industrial hygiene monitoring
81
DRAFT - DO NOT CITE OR QUOTE
-------
focusing on dichloromethane, acetone, and methanol began in the late 1960s. Exposure
groupings (low = 50-100 ppm and high = 350-700 ppm) were assigned by area in which
employees worked. The extrusion and spinning workers and jet wipers were among the high
exposure group (300-1,250 ppm 8-hour TWA). The SMR analysis that was reported used
Allegany County, Maryland, as the comparison group. Cause of death information was obtained
from death certificates, but the authors did not state whether they used underlying or underlying
and contributing cause of death information. The mean duration of work in the cohort was not
reported. The total follow-up period included 49,828 person-years (16,292 years in the high
exposure group and 33,536 years in the low exposure group), and the mean duration of follow-up
was 17.2 years (range 8-20 years). These data were found in Hearne and Pifer (1999, Table 7).
There was little evidence of an increase in mortality risk from cancer (all sites) or from
cancer of the liver and bile duct, pancreas, or brain in men or women (Table 4-8). An increasing
risk with increasing exposure level was seen for prostate cancer mortality in men. The/>-value
for the trend was not given, but the authors describe it as a "nonstatistically significant dose-
response relationship." A statistically significant SMR for prostate cancer death was seen in the
350-700 ppm group when latency (at least 20 years since first exposure) was included in the
analysis (SMR = 2.08,/> < 0.05). Cervical cancer mortality risk was increased, but the small
number of cases in the high exposure group did not allow a precise assessment of the pattern
with respect to exposure level. There was no increased risk of breast cancer.
82 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-8. Cancer mortality risk in Hoechst Celanese Corporation cellulose
triacetate fiber production workers, Cumberland, Maryland: 2,909 men
and women employed 1970-1981, followed through 1989
Cancer type,
exposure level"
Cancer, all sites
50-100 ppm
350-700 ppm
Liver
50-100 ppm
350-700 ppm
Pancreas
50-100 ppm
350-700 ppm
Lung
50-100 ppm
350-700 ppm
Brain3
50-100 ppm
350-700 ppm
Hodgkin's3
50-100 ppm
350-700 ppm
Leukemia3
50-100 ppm
350-700 ppm
Prostate
50-100 ppm
350-700 ppm
Cervical
50-100 ppm
350-700 ppm
Breast3
50-100 ppm
350-700 ppm
Men (n = 1,931)
Obsb
121
64
57
2
1
1
3
2
1
35
20
15
2
1
1
1
0
4
1
22
9
13
Expb
70.0
75.6
1.33
1.24
2.24
2.90
25.7
27.3
1.88
1.94
0.4
0.41
2.14
2.28
6.41
7.26
SMR
0.91
0.75
0.75
0.81
0.89
0.35
0.78
0.55
0.53
0.52
2.5
1.9
0.44
1.4
1.8
95% CIC
0.70-1.2
0.57-0.98
0.02-1.2
0.02-1.5
0.1-3.2
0.01-1.9
0.48-1.2
0.31-0.91
0.01-2.96
0.01-2.87
0.06-13.9
0.51-1.8
0.01-2.4
0.64-2.7
0.95-3.1
Not applicable
0
0
0
0.03
0.02
Women (n = 978)
Obsb
42
37
5
0
0
0
1
1
0
11
9
2
2
2
0
0
0
0
0
Expb
44.79
4.61
1.04
0.10
1.73
0.18
8.24
0.87
0.66
0.07
0.23
0.02
1.25
0.13
SMR
0.83
1.1
0.58
1.1
2.3
3.1
95% CIC
0.58-1.1
0.35-2.5
-
-
0.01-3.2
-
0.50-2.1
0.28-8.3
0.37-10.9
Not applicable
6
5
1
10
9
1
1.69
0.19
9.8
1.07
3.0
5.4
0.92
0.93
0.96-6.9
0.13-30.1
0.42-1.7
0.02-5.2
3Data for brain and CNS cancer, Hodgkin's lymphoma, leukemia, and breast cancer reported in Gibbs (1992).
bObs = number of observed deaths, Exp = number of expected deaths. Referent group = Allegany County,
Maryland. SMRs and CIs were not calculated for categories with zero observed cases.
cCIs were calculated fromBreslow and Day (1987, Table 2.10).
Sources: Gibbs et al. (1996); Gibbs (1992).
A primary limitation of this study is that workers who were exposed before 1970 but
were not working at the plant in 1970 were not included in the cohort. The authors had
83
DRAFT - DO NOT CITE OR QUOTE
-------
attempted to create a cohort of all workers who were employed on or after January 1, 1954, but
problems with the completeness of the personnel file made it impossible to use this study design.
From what the author (Gibbs, 1992) was able to determine, the records of workers who had died,
left the company, or retired before the mid to late 1960s (when a new personnel system was
developed) were not available. Additional limitations include the small size of the cohort, small
number of observed cancer deaths, and use of mortality (death certificate) data. This is
particularly problematic for cancers with relatively high survival rates (such as prostate cancer
and cervical cancer), since incidence rates are not estimated well by mortality rates in this
situation.
4.1.3.5. Solvent-Exposed Workers—Hill Air Force Base, Utah
Spirtas et al. (1991), Blair et al. (1998), and Radican et al. (2008) evaluated exposure to
dichloromethane in relation to mortality risk in successive retrospective cohort studies of
14,457 civilian workers employed at Hill Air Force Base in Utah for at least 1 year from 1952 to
1956. The analysis was limited to the workers who were white or who had missing data on race,
resulting in a sample size of 14,066 (10,461 men and 3,605 women). Spirtas et al. (1991)
examined mortality through 1982 (3,832 deaths), Blair et al. (1998) updated mortality through
1990 (5,727 deaths), and Radican et al. (2008) extended the follow-up period through 2000
(8,580 deaths). The underlying and contributing causes of death information from death
certificates was used to classify cause-specific mortality. SMRs were calculated by using
mortality rates from the Utah population, and an internally standardized life table method was
used to adjust for age at entry into the cohort and competing causes of death. In the Radican et
al. (2008) analysis, adjusted relative risks (rate ratios) were estimated from a Poisson regression
analysis with unexposed workers as the referent. The mean duration of work was not reported.
In the analysis through 1982 (Spirtas et al., 1991), there were 22,770 person-years of follow-up
in men and 3,091 person-years of follow-up in women who were classified as exposed to
dichloromethane. The total number of workers classified as exposed to dichloromethane was
1,222 (Stewart et al., 1991), which would yield an estimated mean of approximately 21 years of
follow-up through 1982. The additional follow-up of 18 years in Radican et al. (2008) would be
expected to increase the mean follow-up time among those still alive to approximately 40 years.
Two industrial hygienists developed the exposure assessment based on walkthrough
surveys, interviews with management and labor representatives, review of historical records, job
descriptions, monitoring data and other information pertaining to chemicals used, and
organization of the work site (Radican et al., 2008; Blair et al., 1998; Spirtas et al., 1991). Each
worker was assigned exposure by using information on the worker's job history, which included
job titles, department codes, and dates of employment. The most detailed exposure assessment
was done for trichloroethylene, the primary focus of the study. Dichloromethane, one of
84 DRAFT - DO NOT CITE OR QUOTE
-------
25 other exposures analyzed, was classified as a dichotomous exposure (ever exposed, never
exposed).
Radican et al. (2008) presented the mortality risk for three specific cancers in relation to
15 of the 25 chemicals classified as dichotomized exposures. The rate ratios for non-Hodgkin's
lymphoma and multiple myeloma in relation to dichloromethane in men were 2.2 (95% CI 0.76-
5.42, based on eight observed cases) and 2.58 (95% CI 0.86-7.72, based on seven observed
cases), respectively. These rate ratios (particularly those for multiple myeloma) were
considerably higher than the rate ratios for any of the other chemicals examined in which the
next highest observed rate ratios were 2.0 and 2.1 for Freon, o-dichlorobenzene, and the "other
alcohols" categogy. No cases of either of these cancers were observed in women with
dichloromethane exposure, but the rate ratio for breast cancer in these women was 2.35 (95% CI
0.98-5.65, based on six observed cases). Associations of similar magnitude (rate ratios of 2.3-
2.8) were also seen among breast cancer and some other exposures (Freon, solder flux, isopropyl
alcohol, and 1,1,1-trichloroethane). All of these risk estimates were slightly attenuated from the
estimates in Blair et al. (1998).
This is the largest of the identified cohort studies that included women and specifically
reported data pertaining to breast cancer risk. The major limitation of this study is that the
exposure assessment for dichloromethane was based on a dichotomized classification. In
addition, exposure to many different types of solvents was common; thus, in some analyses it
can be difficult to assess the independent effects of individual exposures. Some aspects of
reproductive history, such as age at first pregnancy, are known risk factors for breast cancer.
Reproductive history was not included in this analysis, but Blair et al. (1998) noted that it is
unlikely that these factors would confound the results of a few specific chemicals, since the
referent group was an internal group within the cohort (and thus would be expected to be similar
in terms of socioeconomic status) and there was no association overall between solvent
exposures and breast cancer mortality.
4.1.3.6. Case-Control Studies of Specific Cancers and Dichloromethane
Twelve site-specific cancer case-control studies included dichloromethane as an exposure
of interest. These studies involve six types of cancer: brain and CNS (Cocco et al., 1999;
Heineman et al., 1994), breast (Cantor et al., 1995), kidney (Dosemeci et al., 1999), pancreas
(Kernan et al., 1999), rectum (Dumas et al., 2000), and various forms of lymphoma and
leukemia, including childhood leukemia (Gold et al., 2011; Wang et al., 2009; Seidler et al.,
2007; Miligi et al., 2006; Costantini et al., 2008; Infante-Rivard et al., 2005).
4.1.3.6.1. Case-control studies of brain cancer. Heineman et al. (1994) studied the association
between astrocytic brain cancer (International Classification of Diseases 9* ed. [ICD-9] codes
191, 192, 225, and 239.7) and occupational exposure to chlorinated aliphatic hydrocarbons.
85 DRAFT - DO NOT CITE OR QUOTE
-------
Cases were identified by using death certificates from southern Louisiana, northern New Jersey,
and the Philadelphia area. This analysis was limited to white males who died between 1978 and
1981. Controls were randomly selected from the death certificates of white males who died of
causes other than brain tumors, cerebrovascular disease, epilepsy, suicide, and homicide. The
controls were frequency matched to cases by age, year of death, and study area.
Next of kin were successfully located for interview for 654 cases and 612 controls, which
represents 88 and 83% of the identified cases and controls, respectively. Interviews were
completed for 483 cases (74%) and 386 controls (63%). There were 300 cases of astrocytic
brain cancer (including astrocytoma, glioblastoma, mixed glioma with astrocytic cells). The
ascertainment of type of cancer was based on review of hospital records, which included
pathology reports for 229 cases and computerized tomography reports for 71 cases. After the
exclusion of 66 controls with a possible association between cause of death and occupational
exposure to chlorinated aliphatic hydrocarbons (some types of cancer, cirrhosis of the liver), the
final analytic sample consisted of 300 cases and 320 controls.
In the next-of-kin interviews, the work history included information about each job held
since the case (or control) was 15 years old (job title, description of tasks, name and location of
company, kinds of products, employment dates, and hours worked per week). Occupation and
industry were coded based on four-digit Standard Industrial Classification and Standard
Occupational Classification (Department of Commerce) codes. The investigators developed
matrices linked to jobs with likely exposure to dichloromethane, five other chlorinated aliphatic
hydrocarbons (carbon tetrachloride, chloroform, methyl chloroform, tetrachloroethylene, and
trichloroethylene), and organic solvents (Gomez et al., 1994; Dosemeci et al., 1994). This
assessment was done blinded to case-control status. Exposure was defined as the probability of
exposure to a substance for a specific occupation and industry, duration of employment in the
exposed occupation and industry, specific exposure intensity categories, average intensity score
(the three-level semiquantitative exposure concentration assigned to each job multiplied by
duration of employment in the job, summed across all jobs), and cumulative exposure score
(weighted sum of years in all exposed jobs with weights based on the square of exposure
intensity [1, 2, 3] assigned to each job). Secular trends in the use of specific chemicals were
considered in the assignment of exposure potential. Exposures were lagged 10 or 20 years to
account for latency. In addition, a more detailed coding of specific jobs within the 4-digit
industry and occupation code categories was used, for example, to distinguish production of
paint removers from production of paints, varnishes, lacquers, enamels, and allied products
(Dosemici et al., 1994). Thus, this exposure assessment procedure was quite detailed.
Adjusting for age and study area, the OR for the association between any exposure to
dichloromethane and risk of astrocytic brain cancer was 1.3 (95% CI 0.9-1.8). There was a
statistically significant trend (p < 0.05) with increasing probability of exposure to
dichloromethane with an OR = 1.0 (95% CI 0.7-1.6) for low probability, OR = 1.6 (95% CI 0.8-
86 DRAFT - DO NOT CITE OR QUOTE
-------
3.0) for medium probability, and OR = 2.4 (95% CI 1.0-5.9) for high probability compared with
the referent group of unexposed men. An increased risk with higher duration of exposure was
also observed with OR =1.7 (95% CI 0.9-3.6) for >21 years of work in exposed jobs for all
exposed workers and OR = 6.1 (95% CI 1.1-43.8) for the combination of >21 years of work in a
high probability of exposure job. Similar results were seen in additional analyses controlling for
age, study area, employment in electronics occupations and industries, and exposure to carbon
tetrachloride, tetrachloroethylene, and trichloroethylene. In the analyses adjusting for these other
exposures, only dichloromethane exhibited a trend with increasing probability of exposure.
There was also evidence of an association between astrocytic brain cancer risk and
dichloromethane exposure, based on the average intensity score, with an OR =1.1 (95% CI 0.7-
1.7) for the low-medium intensity group and an OR = 2.2 (95% CI 1.1-4.1) for the high intensity
group (trends-value < 0.05). The combination of high intensity and high duration (>21 years)
was strongly associated with risk (OR = 6.1 [95% CI 1.5-28.3]), and a weaker association (OR =
1.4 [95% CI 0.6-3.2]) was seen for high intensity and shorter duration (2-20 years). The
association between cumulative exposure score (low, medium, and high) and astrocytic brain
cancer risk was nonlinear (ORs of 0.9, 1.9, and 1.2 in the low, medium, and high exposure
categories, respectively).
The strengths of this case-control study include a large sample size, detailed work
histories (including information not just about usual or most recent industry and occupation but
also about tasks and products for all jobs held since age 15), and comprehensive exposure
assessment and analysis along several different dimensions of exposure. The major limitations
were the lack of direct exposure information and potential inaccuracy of the descriptions of work
histories that were obtained from next-of-kin interviews. Heineman et al. (1994) acknowledge
these limitations in the report, and in response to a letter by Norman and Boggs (1996) criticizing
the methodology and interpretation of the study, Heineman et al. (1996) noted that while the lack
of direct exposure information must be interpreted cautiously, it does not invalidate the results.
Differential recall bias between cases and controls was unlikely because work histories came
from next-of-kin for both groups and the industrial hygienists made their judgments blinded to
disease status. Exposure to other solvents, including carbon tetrachl oride, tetrachloroethylene
and trichloroethylene were also common in cases and controls. These co-exposures provide
additional evidence that recall bias was not distorting the observed effects, as the strong
associations that were seen with the exposure measures for dichloromethane were not seen with
the other solvents included in the analysis. In addition, adjustment for these other solvent
exposures did not result in any attenuation in the observed association with dichloromethane.
The relatively strong and statistically significant associations between dichloromethane and
astrocytic brain tumors were seen along multiple measures of exposure, suggesting that the
results were unlikely to be spurious. Nondifferential misclassification would, on average,
87 DRAFT - DO NOT CITE OR QUOTE
-------
attenuate true associations and would be unlikely to result in the types of exposure-response
relationships that were observed in this study.
Norman and Boggs (1996) described an apparent inconsistency in the estimated trends in
dichloromethane and carbon tetrachloride exposure based on the methodology used in this case-
control study (described in more detail in Gomez et al. [1994]). In response, Gomez (1996)
noted that the apparent inconsistency was actually due to an error in the labeling of the lines on
one of the figures in the report rather than an inconsistency with the estimated trends. Another
point raised by Norman and Boggs (1996) was that the Heineman et al. (1994) findings were
surprising in light of the lack of brain carcinogenesis in animals. In response, Heineman et al.
(1996) pointed out that carcinogens commonly cause different cancers in animals and humans. It
can also be noted that brain tumors are exceedingly rare in animal bioassays (Sills et al., 1999).
Norman and Boggs (1996) also suggested that the results of the Heineman et al. (1994) study be
given no weight when compared with the results of the cohort studies. The authors responded by
pointing out that the cohort studies had low statistical power and large CIs around their point
estimates but were not inconsistent with an association between dichloromethane and brain
cancer (Heineman et al., 1996). This point is strengthened further by the more recent results
from the Rochester, New York, Eastman Kodak cohort (Hearne and Pifer, 1999), described
previously, since an increased SMR for brain and CNS cancers was seen in the longer follow-up
period of this cohort. Dell et al. (1999) also noted the limitation of the lack of direct exposure
measures in the Heineman et al. (1994) study, and the difficulty in categorizing jobs with
occasional exposure to a specific solvent. As noted by Heineman et al. (1994), the "high
probability" category for dichloromethane in this study included painting, paint or varnish
manufacture, ship or boat building and repair, and electronics manufacture. The "high intensity"
category included these jobs and jobs in roofing and the pharmaceutical industry. Dell et al.
(1999) specifically questioned the inclusion of the roofing jobs as a job with potential for high
intesnsity dichloromethane exposure. The number of individuals who were categorized based on
this type of job, and more specific information pertaining to the more detailed coding within
these job categories that may have contributed to the classification decisions, were not discussed
in Dell et al. (1999), Heineman et al. (1994), Gomez et al. (1994), or Dosemici et al. (1994).
In another case-control study of brain cancer and dichloromethane exposure, Cocco et al.
(1999) identified 12,980 female cases of cancer of the brain and CNS through the underlying
cause of death listing (ICD-9 codes 191 and 192) on death certificates from 24 states from
1984 to 1992. (This collection of death certificates is a data set created by the National Center
for Health Statistics, NIOSH, and the National Cancer Institute to facilitate research on
occupational exposures and mortality risk.) The cases included 161 women with meningioma
(ICD-9 codes 192.1, 192.3). Four women who died of nonmalignant diseases, excluding
neurological disorders, were chosen as controls for each case. The controls were frequency
matched to the cases by state, race, and 5-year age group. Occupation data were based on the
88 DRAFT - DO NOT CITE OR QUOTE
-------
occupation fields in the death certificates. This job was coded based on the three-digit industry
and three-digit occupation (Department of Commerce) codes. The investigators developed job
exposure matrices that were applied to these industry/occupation codes. The job exposure
matrices included probability and intensity scores for 11 occupational hazards, one of which was
dichloromethane, but also included other solvents, electromagnetic fields, chlorinated aliphatic
hydrocarbons, benzene, lead, nitrosamines, insecticides, herbicides, and public contact. The
investigators used logistic regression models to estimate ORs, adjusting for each workplace
exposure, marital status, three levels of socioeconomic status (based on occupation), and age at
death. For each chemical, four levels of intensity and probability were defined (unexposed, low,
medium, and high).
A weak association between dichloromethane exposure and brain/CNS cancer was seen
(OR 1.2 [95% CI 1.1-1.3]) (Cocco et al., 1999). There was no exposure-related trend in the
association between probability or intensity of exposure and brain cancer. A similar but more
imprecise association was seen with meningioma cancer (OR 1.2 [95% CI 0.7-2.2]). There were
too few cases of meningioma to stratify by exposure probability and intensity.
The major limitations of this study are the use of mortality rather than incidence data and
the reliance on occupation data from death certificates. The death certificate occupation data are
based on "usual" occupation, which may be more prone to misclassification in studies of women
because of gender-related differences in work patterns (i.e., shorter duration jobs for women
compared with men). A relatively broad job exposure matrix was applied to the job information,
and typically more generic job exposure matrices result in less sensitive assessment with limited
ability to detect exposure-response trends (Teschke et al., 2002). Nondifferential
misclassification of outcome and exposure would generally result in attenuated effect estimates.
4.1.3.6.2. Case-control studies of breast cancer. Cantor et al. (1995) conducted a case-control
study of occupational exposures and breast cancer using the 24-state (1984-1989) death
certificate data described in the previous section. Cases were women with breast cancer coded as
the underlying cause of death (ICD-9 code 174). Four female controls per case were selected
from all noncancer deaths, frequency matched by age (5-year age groups) and ethnicity (black,
white). The occupation listed on the death certificate was coded based on the three-digit industry
and three-digit occupation (Department of Commerce) codes, and this was used with a job
exposure matrix developed by the investigators to assess 31 workplace exposures, one of which
was dichloromethane. Four exposure probability and three exposure level scores were assigned.
ORs for probability and level were calculated for each ethnic group, adjusting for age at death
and a measure of socioeconomic status (based on occupation). After excluding subjects whose
death certificate occupations were listed as homemaker, there were 29,397 white cases and
4,112 black cases (total 33,509) and 102,955 white controls and 14,839 black controls (total
117,794).
89 DRAFT - DO NOT CITE OR QUOTE
-------
There was little evidence of an association between exposure probability and breast
cancer mortality using the probability exposure metric. The ORs were 1.05 (95% CI 0.97-1.1)
and 0.76 (95% CI 0.3-2.0) in probability level 3 and level 4, respectively, for white women and
1.13 (95% CI 0.9-1.4) in probability level 3 for black women. (There were too few black
women in exposure probability level 4 for analysis.) Weak associations were seen with exposure
level. In white women, an OR of 1.17 (95% CI 1.1-1.3) was seen with the highest exposure
level, and in black women the OR in this exposure group was 1.46 (95% CI 1.2-1.7). In the
analysis that jointly considered exposure level and probability ratings but excluded the lowest
probability of exposure, the OR for the highest category of exposure level was 1.28 in whites
(p< 0.05) and 1.21 in blacks.
As with the Cocco et al. (1999) case-control study that used a similar methodology, the
limitations of this study include the use of an outcome defined by mortality rather than incidence,
use of usual occupation information as recorded in death certificates, and use of a very broad job
exposure matrix to classify 31 different exposures. Although information on pregnancy and
lactation history (known risk factors for breast cancer) was not available, the authors did adjust
for socioeconomic status by using the occupation data, which may have corrected for some of the
potential confounding due to reproductive history.
4.1.3.6.3. Case-control studies of pancreatic cancer. Kernan et al. (1999) conducted a case-
control study of 63,097 pancreatic cancer cases using the 24-state (1984-1993) death certificate
data. The diagnosis of pancreatic cancer was based on underlying cause of death (ICD-9 code
157). Four controls who had died during the same time period of causes other than cancer were
selected for each case, frequency-matched by state, race, gender, and 5-year age group
(n = 252,386). Usual occupation and industry, based on the occupation data in the death
certificate, were coded by using the three-digit (Department of Commerce) codes. A job-
exposure matrix was used with the industry and occupation codes to evaluate exposure intensity
and probability (each categorized as high, medium, or low) for formaldehyde, dichloromethane,
10 other solvents, and a combined "organic solvents" measure. Race- and gender- specific
analyses were conducted by using logistic regression to estimate ORs and 95% CIs, adjusting for
age, marital status (ever, never married), residential area (metropolitan, nonmetropolitan), and
region (east, south central, south, and west).
The point estimates for the ORs in the low, medium, and high intensity categories in the
four race-gender groups ranged from 0.8 to 1.3, with no exposure-effect trend seen in any group.
The only statistically significant OR was for high exposure intensity in white females (OR 1.3
[95% CI 1.1-1.6]), with ORs of 1.0 (95% CI 0.9-1.1) for medium intensity and 1.1 (95% CI 1.0-
1.2) for low intensity in this group. An elevated OR was seen with high exposure probability in
black males (OR 2.2 [95% CI 1.0-4.8]) but not in white females (OR 1.0 [95% CI 0.8-1.4]) or
white males (OR 1.0 [05% CI 0.8-1.3]), and the ORs were 0.9 for medium exposure probability
90 DRAFT - DO NOT CITE OR QUOTE
-------
in these three groups. There were relatively few black females in this study, resulting in
imprecise estimates (OR 2.0 [95% CI 0.8-5.4] for medium exposure and OR 1.5 [95% CI 0.6-
3.6] for high exposure).
The limitations of this study, as with the other case-control studies that used the 24-state
death certificate data set, include the reliance on cause of death data from death certificates rather
than medical-record validated incidence data and the use of death certificate occupation data.
The job exposure matrix used with the occupation data was more focused than those used in
Cocco et al. (1999) and Cantor et al. (1995). Although the analysis adjusted for some
sociodemographic characteristics, it did not include measures of smoking history or diabetes,
which are known risk factors for pancreatic cancer (Lowenfels and Maisonneuve, 2005).
4.1.3.6.4. Case-control studies of renal cancer. Dosemeci et al. (1999) reported data from a
population-based case-control study of the association between occupational exposures and renal
cancer risk. The investigators identified newly diagnosed patients with histologically confirmed
renal cell carcinoma from the Minnesota Cancer Surveillance System from July 1, 1988, to
December 31, 1990. The study was limited to white cases, and age and gender-stratified controls
were ascertained by using random digit dialing (for subjects ages 20-64) and from Medicare
records (for subjects 65-85 years). Of the 796 cases and 796 controls initially identified,
438 cases (273 men, 165 women) and 687 controls (462 men, 225 women) with complete
personal interviews were included in the occupational analysis.
Data were obtained through in-person interviews that included demographic variables,
residential history, diet, smoking habits, medical history, and drug use. The occupational history
included information about the most recent and usual industry and occupation (coded using the
standard industrial and occupation codes, Department of Commerce), job activities, hire and
termination dates, and full- and part-time status. A job exposure matrix developed by the
National Cancer Institute was used with the coded job data to estimate exposure status to
dichloromethane and eight other chlorinated aliphatic hydrocarbons. ORs were adjusted for age,
smoking, hypertension and use of drugs for hypertension, and body mass index. No association
between renal cell carcinoma and exposure to dichloromethane was observed in men (OR 0.85
[95% CI 0.6-1.2]), women (OR 0.95 [95% CI 0.4-2.2]), or both sexes combined (OR 0.87 [95%
CI 0.6-1.2]).
Strengths of this study include the use of incident cases of renal cancer from a defined
population area and confirmation of the diagnosis using histology reports. The occupation
history was based on usual and most recent job in combination with a relatively focused job
exposure matrix. In contrast to the type of exposure assessment that can be conducted in cohort
studies within a specific workplace, however, exposure measurements based on personal or
workplace measurements were not used, and a full lifetime job history was not obtained.
91 DRAFT - DO NOT CITE OR QUOTE
-------
4.1.3.6.5. Case-control studies of rectal cancer. Dumas et al. (2000) reported data from a case-
control study of occupational exposures and rectal cancer conducted in Montreal, Quebec,
Canada. The investigators identified 304 newly diagnosed cases of primary rectal cancer,
confirmed on the basis of histology reports, between 1979 and 1985; 257 of these participated in
the study interview. One control group (n = 1,295) consisted of patients with other forms of
cancer (excluding lung cancer and other intestinal cancers), recruited through the same study
procedures and time period as the rectal cancer cases. A population-based control group
(n = 533), frequency matched by age strata, was drawn by using electoral lists and random digit
dialing. The occupational assessment consisted of a detailed description of each job held during
the working lifetime, including the company, products, nature of work at site, job activities, and
any additional information from the interviews that could furnish clues about exposure. The
percentage of proxy respondents was 15.2% for cases, 19.7% for other cancer controls, and
12.6% for the population controls.
A team of industrial hygienists and chemists blinded to subjects' disease status translated
jobs into potential exposure to 294 substances with three dimensions (degree of confidence that
exposure occurred, frequency of exposure, and concentration of exposure). Each of these
exposure dimensions was categorized into none, any, or substantial exposure. Logistic
regression models adjusted for age, education, proxy versus subject responder status, cigarette
smoking, beer drinking, and body mass index. Using the cancer control group, the OR for any
exposure to dichloromethane was 1.2 (95% CI 0.5-2.8) and the OR for substantial exposure
(confident that exposure occurred with >5 years of exposure at medium or high frequency and
concentration) was 3.8 (95% CI 1.1-12.2). The results using the population-based control group
for this exposure were not presented.
The strengths of this study were the large number of incident cases, specific information
about job duties for all jobs held, and a definitive diagnosis of rectal cancer. However, the use of
the general population (rather than a known cohort of exposed workers) reduced the likelihood
that subjects were exposed to dichloromethane, resulting in relatively low statistical power for
the analysis. The job exposure matrix applied to the job information was very broad since it was
used to evaluate 294 chemicals.
4.1.3.6.6. Case-control studies oflymphoma, leukemia, and multiple myeloma. Several case-
control studies examined risk of different types of hemato- and lymphopoietic cancer and
occupational exposure to dichloromethane in Germany (Seidler et al., 2007), Italy (Costantini et
al., 2008; Miligi et al., 2006), and the United States (Barry et al., 2011; Gold et al., 2011; Wang
et al., 2009). Another study examined maternal occupational exposure in relation to risk of
childhood acute lymphoblastic leukemia in Quebec (Infante-Rivard et al., 2005). Each of these
population-based studies used a fairly extensive exposure assessment protocol, as described
below.
92 DRAFT - DO NOT CITE OR QUOTE
-------
Seidler et al. (2007) examined occupational solvent exposure in relation to lymphoma
risk in a population-based case control study in 6 areas of Germany. Incident cases (ages 18 to
80 years) were identified through hospitals and physician practices. The sample (n = 710; 87%
of the identified cases) was primarily made up of non-Hodgkin lymphoma (n = 554 B cell and n
= 35 T cell) patients. Controls drawn from population registries were individually matched to
cases by age (within 1 year), gender, and area. The participation rate among identified controls
was 44% (total n = 710). Data were collected through a structured in-person interview covering
demographics, medical history, alcohol and tobacco use, and a work history including details of
all jobs lasting at least one year. This information included job and task information, and 14
specific supplementary questionnaires for jobs with likely exposure to solvents and other
chemicals. An industrial hygienist, blinded to case-control status, reviewed the work history data
and developed measures of intensity (low = 1 to 10 ppm, medium = 10 to 100 ppm, and high =
>200 ppm) and frequency of exposure (low = 1 to 5% of working time, medium = >5 to 30% of
working time, high = >30% of working time) to dichloromethane. Measures of three other
chlorinated solvents and four aromatic hydrocarbons were also developed in this manner.
Cumulative exposure was calculated as the sum, across all jobs, of the product of the intensity
and frequency and job duration measure for each solvent. Conditional logistic regression,
adjusting for smoking (pack years) and alcohol use, was used to examine associations between
exposure measures and all lymphoma. Unconditional logistic regression, with additional
adjustment for the matching factors, was used to examine associations within specific subsets of
disease (e.g., B-non-Hodgkin lymphoma), using all controls as the comparison group. The
adjusted ORs for all lymphoma and dichloromethane exposure was 0.4 (95% CI 0.2-1.0), 0.8
(95% CI 0.3-1.9), and 2.2 (95% CI 0.4-11.6) for cumulative exposures of >0 to <26.3 ppm-
years, >26.3 to <175 ppm-years, and >175 ppm-years, respectively, compared to 0 ppm-years
(trend/? = 0.40). A similar pattern was seen in the analysis limited to B-non-Hodgkin lymphoma
cases: OR = 0.4 (95% CI 0.2-1.1), 0.9 (95% 0.3-2.3), and 2.7 (95% CI 0.5-14.5) for cumulative
exposures of >0 to <26.3 ppm-years, >26.3 to <175 ppm-years, and >175 ppm-years,
respectively, compared to 0 ppm-years (trend/? = 0.29).
Another population-based case-control study focusing on solvent exposure and non-
Hodgkin lymphoma risk was conducted in Connecticut (Barry et al., 2011; Wang et al., 2009).
This study was limited to women, ages 21-84 years, identified through the Yale University
Rapid Case Acertainment system, with diagnosis between 1996 and 2000. A total of 832 eligible
cases were identified, and 601 (72%) participated in the study. Controls were identified through
random digit dialing (ages <65 years) and Medicare file (ages >65 years). The participation rate
was 69% among controls identified through random digit dialing and 47% among controls
selected through the Medicare files. The final sample included 717 controls. Barry et al. (2011)
is limited to 518 cases and 597 controls with blood or buccal cell samples available for
genotyping. Six variants in four genes involved in the metabolism of benzene (or other solvents,
93 DRAFT - DO NOT CITE OR QUOTE
-------
including dichloromethane) were selected for analysis. These genes were CYP2E1 (rs2070673
and rs2031920), EPHX1 (rs2234922 and rs!051740), NQO1 (rs!800566), and MPO
(rs2333227). Data pertaining to work history were collected through structured interviews,
focusing on all jobs held for at least one year. The information collected included job title,
company name, and duties. A job exposure matrix was used to link the job data to measures of
intensity and probability (a 4-category scale used for each: none, low, medium, and high) for
general classes of solvents and for dichloromethane and seven other specific solvents.
The association seen with ever exposure to dichloromethane and non-Hodgkin lymphoma
was OR = 1.5 (95% CI 1.0-2.3) (Wang et al., 2009). There was little difference in risk by
probability or intensity score. In the subset of patients and controls used in the genetic analysis,
the association with ever exposure to dichloromethane (OR 1.69, 95% CI 1.06-2.69) was similar
to that seen in the full sample (Barry et al., 2011). The increased risk was seen among carriers of
the CYP2E1 rs20760673 TT genotype (OR 4.42, 95% CI 2.03-9.62), but not among carriers ot
the TA or AA genotype (OR 0.80, 95% CI 0.36-1.75) (interaction^ < 0.01). A similar
interaction was seen between this genotype and exposure to carbon tetrachloride and methyl
chloride. The functional significance of this variant, which occurs in the promoter region, is not
known. No association was seen between any of the solvents and CYP2E1 rs2031920, EPHX1
rs 1051740, NQO1 rs!800566, or MPO rs2333227.
The studies in Italy were part of a large study of hemato-lymphopoietic cancers
conducted in 11 geographic areas chosen based on the historical presence of solvent-based
industries or farming activities (Costantini et al., 2001; Vineis et al., 1999). The number of areas
varied depending on the specific disease in the analysis: 8 for non-Hodgkin lymphoma (Miligi et
al., 2006), 7 for leukemia, and 6 for multiple myeloma (Costantini et al., 2008). Incident cases,
ages 20-70 years, were identified through hospital and hematology centers, with classification
and confirmation of non-Hodgkin lymphoma diagnosis based on pathology review and
consideration of cell type of origin. The participation rate among identified cases was 83% for
non-Hodgkin lymphoma (Miligi et al., 2006), 85% for leukemia, and 83% for multiple myeloma
(Costantini et al., 2008). Controls, frequency matched by area, gender and 5-year age groups to
the cases, were identified through computerized demographic files or National Health Service
files covering each of the study areas. The participation rates among the randomly selected
controls were 73, 72, and 76% for those included in the non-Hodgkin lymphoma, leukemia, and
multiple myeloma analyses, respectively (Costantini et al., 2008; Miligi et al., 2006). Data were
collected through a structured in-person interview covering demographics, medical history,
family medical history, and alcohol and tobacco use (mean length, approximately 60 minutes).
The interview also included a detailed work history with the inclusion of job- and industry-
specific modules focusing on solvents and agricultural exposures; jobs lasting 5 or more years
were included in this assessment (Constantini et al., 2001). These data were reviewed by
industrial hygienists, blinded to case-control status, and used to develop measures of exposure
94 DRAFT - DO NOT CITE OR QUOTE
-------
duration, probability (low, medium, and high) and intensity (very low, low, medium, and high)
for five general classes of hydrocarbon solvents (aromatic, chlorinated, technical, aliphatic, and
derivative oxygenate) and 8 specific solvents (including dichloromethane). Analyses were
conducted adjusting for gender, age, education and geographic area, and the reference group
consisted of individuals with no occupational exposure to any solvent.
The analysis of non-Hodgkin lymphoma included 1428 cases and 1530 controls (Miligi et
al., 2006). No association was seen with the general exposure category of "any" occupational
solvent exposure using the intensity measure (adjusted OR 1.0 and 1.1, respectively, for very
low/low and medium/high intensity, compared with unexposed) or with duration of medium-high
exposure (adjusted OR 1.0 and 1.0, respectively, for <= 15 and >15 years, compared with
unexposed). A more pronounced pattern was seen between dichloromethane exposure intensity
and non-Hodgkin lymphoma risk, with adjusted OR = 0.9 (95% CI 0.5-1.6; 23 exposed cases)
for the very low/low intensity and adjusted OR =1.7 (95% CI 0.7-4.3; 13 exposed cases) for the
medium/high intensity groups (trend p-value = 0.46). The sparse number of exposed cases
precluded analysis by duration. In analysis by disease subtype, medium/high dichloromethane
exposure intensity was associated with small lymphocytic non-Hodgkin lymphoma (adjusted OR
3.2, 95% CI 1.0-10.1; 6 exposed cases; total n = 285).
The leukemia analysis included 586 cases and 1278 controls (Costantini et al., 2008). No
association was seen with dichloromethane exposure intensity: for very low/low intensity (7
exposed cases), adjusted OR = 0.7 (95% CI 0.3-1.7) and for medium/high intensity (2 exposed
cases) adjusted OR = 0.5 (95% CI 0.1-2.3). There were too few exposed cases of multiple
myeloma (4 exposed to dichloromethane; total n = 263) for analysis.
Gold et al. (2011) conducted a population-based case-control study of multiple myeloma
in Seattle, Washington and Detroit, Michigan using the Surveillance, Epidemiology and End
Results (SEER) cancer registries to identify newly diagnosed cases. Eligible cases were ages
35-74 years, diagnosed between January 1, 2000 and March 31, 2002; 365 cases were identified
who met these criteria. Of these, 64 (18%) died before being contacted, 28 (8%) could not be
located, and 18 (5%) were not contacted because consent for the recruitment process was not
obtained from their physician. The remaining 255 patients were contacted, and 181 (71% of
those contacted and confirmed eligible) participated in the study. Controls were identified
through random digit dialing (ages 35-64 years) and Medicare file (ages 65-74 years). The
participation rate among the 1133 potential controls identified was 52%, for a total of 481
controls. Data were collected through a computer-assisted personal interview that included a
detailed work history covering all jobs held since age 18. Job-specific modules for 20 solvent-
related jobs were also included, with details on tasks performed and chemicals used. A job
exposure matrix was developed to classify exposure to each of six chlorinated solvents
(including dichloromethane), with measures of probability, frequency, intensity, and confidence.
Ratings were conducted by two reviewers, blinded to case-control status. Measures of duration
95 DRAFT - DO NOT CITE OR QUOTE
-------
of exposure and cumulative exposure were also developed. Analyses were conducted adjusting
for gender, age, race, education, and SEER site. One set of analyses classified any possible
exposure among the exposed, and a second set of analyses included the low-confidence exposure
jobs with the unexposed group. In general, somewhat stronger associations or patterns were seen
in the second analyses. For example, the OR for ever exposed was 1.5 (95% CI 0.9-2.3) in the
first analysis and 2.0 (95% CI 1.2-3.2) with the reclassification of the low confidence jobs. In
the second set of analyses, a non-monotonic increasing trend was seen with duration of exposure
(OR 1.0, 2.0, 1.1, 2.7, and 2.1, respectively, in unexposed, 1-4 years, 5-7 years, 8-24 years, and
25-47 years, trend/? = 0.01). Similar patterns were seen with cumulative exposure (trend/? =
0.08) and cumulative exposure lagged by 10 years (trend/? = 0.06).
Infante-Rivard et al. (2005) examined the association between maternal occupational
exposures, before and during pregnancy, and risk of childhood acute lymphoblastic leukemia
(ICD-9 code 204.0) using data from a population-based case-control study in Quebec, Canada.
Incident cases diagnosed from 1980 to 2000 were identified from the cancer hospitals in the
province, and diagnosis was confirmed based on clinical records from an oncologist or
hematologist. Between 1980 and 1993, cases ages 0-9 years at diagnosis were included, and
from 1994 to 2000 the age range was expanded to 14 years. The number of eligible cases
identified was 848 and of these, 790 parents (93%) participated in the study. Population-based
controls, individually matched to the sex and age at diagnosis of the cases, were identified from
government registries of all children in the province (1980-1993) and the universal health
insurance files (1994-2000). The parents of 790 (86%) of the 916 eligible controls who were
identified participated in the study. Data were collected by using a structured telephone
interview. Some information (i.e., job title, dates, type of industry, industry name and address)
was obtained for all jobs held since age 18, and additional information (e.g., materials and
machines used, typical activities) was obtained for jobs held by the mother from 2 years before
the pregnancy through the birth of the child. Specialized exposure modules were also used to
collect information about specific jobs (e.g., nurse, waitress, hair dresser, textile dry cleaner).
All of this information was reviewed by chemists and industrial hygienists, blinded to case-
control status, to classify exposure to over 300 chemicals, although the primary focus of the
study was on solvents (21 individual substances, including dichloromethane, and six mixtures).
The exposure assessment included ratings of confidence (possible, probable, and definite),
frequency of exposure during a normal workweek (<5, 5-30, or >30% of the time), and level of
concentration (low = slightly above background, high = highest possible exposure in the study
population, and medium for in-between levels). A weak association was seen between any
dichloromethane exposure during the 2 years before pregnancy up to the birth and risk of
leukemia in the child (OR 1.34 [95% CI 0.54-3.34]), and results were similar when limited to
exposures during pregnancy. Stronger associations were seen with probable or definite exposure
(OR 3.22 [95% CI 0.88-11.7]) compared with possible or no exposure. The estimates for
96 DRAFT - DO NOT CITE OR QUOTE
-------
categories based on concentration and frequency were similar but there was no evidence for an
increasing risk with increasing exposure level.
4.1.3.7. Summary of Cancer Studies by Type of Cancer
The cohort and case-control studies with data relevant to the issue of dichloromethane
exposure and cancer risk are summarized in Tables 4-9 and 4-10, respectively. The strongest of
the cohort studies in terms of design are two of the triacetate film base production cohorts
(Cohort 1 in New York and the United Kingdom cohort, reported in Hearne and Pifer [1999] and
Tomenson et al. [1997], respectively). These are the cohorts with the most extensive exposure
assessment information. The start of eligibility for cohort entrance corresponds with the
beginning of the time when the exposure potential at the work site began, and the follow-up
period is relatively long (mean >25 years). Although Cohort 2 of the New York film base
production study has similar exposure data and follow-up, this cohort was limited to workers
employed between 1964 and 1970 and therefore would have missed anyone leaving (possibly
because of illness or death) before this time. In addition, because of the overlap between
Cohort 1 and Cohort 2, including both cohorts in an evaluation would be double-counting
experiences of some individuals. Several limitations of the triacetate film base production
cohorts should be noted, however. One of these limitations concerns the generalizability of the
results given the relatively low exposure level (mean 8-hour TWA <40 ppm). Exposures in
small, poorly ventilated work areas are also often much higher than those seen in these film base
production cohorts (Estill and Spencer, 1996; Anundi et al., 1993). Other limitations include the
limited power to detect a risk of low-incidence cancers (including brain, liver, leukemia, and
other forms of hematopoetic cancers) and the lack of women and thus lack of data pertaining to
breast cancer. In addition, these cohorts used mortality rather than incidence data, which is of
particular concern for cancers with a relatively high survival rate, such as non-Hodgkin
lymphoma. Although the exposure levels in the cohorts involved in cellulose triacetate fiber
production were much higher than those of the film production cohorts, the duration of exposure
was relatively short in the South Carolina cohort (Lanes et al., 1993), and the majority of
workers were missing job history data. In the Maryland triacetate fiber production plant,
duration of exposure was not reported and the length of follow-up was relatively short (mean
17 years) (Gibbs et al., 1996). Also, the cohort began in 1970, even though production began in
1955, and the missing personnel records made it impossible to recreate an inception cohort. The
exposure assessment in the study of civilian Air Force base workers (Radican et al., 2008; Blair
et al., 1998) allowed for only a dichotomized classification of exposure, and there was
considerable exposure to other solvents among these workers. This Air Force base study was the
largest of the cohort studies that included women and presented data pertaining to breast cancer.
97 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-9. Summary of cohort studies of cancer risk and dichloromethane exposure
Reference and Cohort
Total n, exposure level" and
duration, follow-up period
Inclusion criteria1"
Exposure assessment;
outcome assessment
Results0
Hearne and Pifer( 1999)
Cellulose triacetate film
base production;
New York
Cohort 1
n = 1,311 men; Mean 39 ppm;
mean duration, 17 yr; follow-up
through 1994; mean follow-up,
35 yr
Began working after
1945; worked at least
lyr
Work history (job records) and
personal/air monitoring;
death certificate (underlying cause)
See Table 4-4. Brain cancer SMR
2.16 (95% CI 0.79-4.69); leukemia
SMR 2.04 (95% CI 0.88^.03).
SMRs <1.0 for lung cancer, liver
cancer, and pancreatic cancers
Cohort 2
n = 1,013 men; mean 26 ppm;
mean duration, 24 yr; follow-up
through 1994; mean follow-up,
26 yr
Employed at least 1 yr
between 1964 and
1970 (potential
exposure began 1946)
Work history (job records) and
personal/air monitoring;
death certificate (underlying cause)
See Table 4-4. Results similar to
Cohort 1 except for pancreatic
cancer, SMR 1.55 (95% CI 0.67-
3.06)
Tomenson et al. (1997)
Cellulose triacetate film
base production;
United Kingdom
n = 1,473 men; mean 19 ppm;
mean duration, 9 yr; follow-up
through 1994; mean follow-up,
27 yr
Employed anytime
between 1946 and
1988
Work history (job records) and
personal/air monitoring;
death certificate (underlying cause)
See Table 4-6. Brain cancer SMR
1.45 (95% CI 0.40-3.72). Lung
cancer SMR 0.46 (95% CI 0.29-
0.75)
Lanes etal. (1993)
Cellulose triacetate fiber
production;
South Carolina
n = 551 men and 720 women
(total n = 1,271); median 140,
280, and 475 ppm in low,
moderate, and high, respectively;
56% <5 yr work duration; follow-
up through 1990; meanfollow-
up, -28 yr
Worked at least 3 mo
in the preparation or
extrusion areas from
1954 to 1977
Job history data and personal/air
monitoring of specific areas (but job
history data available for 37%);
death certificate (underlying and
contributing causes)
See Table 4-7. Liver cancer
SMR 2.98 (95% CI 0.81-7.63),
estimate from earlier follow-up
SMR 5.75 (95% CI 1.82-13.8);
lung cancer SMR 0.80 (95% CI
0.43-1.37)
Gibbsetal. (1996)
Cellulose triacetate fiber
production;
Maryland
n = 1,931 men and 978 women
(total n = 2,909); 50-100 ppm in
low and 350-700 ppm in high
exposure; duration not reported;
follow-up through 1989; mean
follow-up 17 yr
Employed on or after
January 1, 1970, for at
least 3 mo (potential
exposure began 1955)
Work history (job records) and
personal/air monitoring;
death certificate (fields used not
stated)
See Table 4-8. Increasing risk
across exposure groups seen for
prostate cancer and cervical cancer.
In men, SMRs -1.0 or <1.0 for lung
cancer and, combining exposure
groups, leukemia. In women,
SMRs-1.0 or<1.0 for breast
cancer and, combining exposure
groups, lung cancer
(Table 4-9 continues on next page)
98
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-9. Summary of cohort studies of cancer risk and dichloromethane exposure
Reference and Cohort
Total n, exposure level" and
duration, follow-up period
Inclusion criteria1"
Exposure assessment;
outcome assessment
Results0
Radican et al., Air Force
Base, Utah (follow-up of
Blair etal., 1998 and
Spirtas et al., 1991)
n= 10,461 men and
3,605 women (total n = 14,066)d;
exposure dichotomized (yes, no);
exposure duration not reported;
follow-up through 2000; mean
follow-up -29 yr
Employed at least 1 yr
from 1952 to 1956
(potential exposure
began 1939)
Work history (job records) and
industrial hygiene assessment based
on work site review (dichotomized
exposure); (underlying and
contributing causes)
See Section 4.1.3.5. In men, non-
Hodgkin's lymphoma RR 2.02
(95% CI 0.76-5.42) and multiple
myeloma RR 2.58 (95% CI 0.86-
7.76). In women, breast cancer RR
2.35 (95% CI 0.98-5.65)
38-hrTWA.
blf dichloromethane was used at the plant before the first date of entrance into the cohort, the yr that potential exposure began is noted.
'Results for all studies are described as SMR and 95% CI except for Blair et al. (1998), in which results are presented as RR (relative risks) (95% CI). Results are
presented cancers based on a minimum of four observed cases. More comprehensive information, when available for other cancers, is shown in the summary
tables for each study.
Includes whites and unknown race.
99
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-10. Summary of case-control studies of cancer risk and dichloromethane exposure
Cancer type,
reference
Location
n cases, n controls (source) and time period,
demographic group
Exposure assessment
Results"
Brain
Heineman et al.
(1994)
Louisiana, New Jersey, Philadelphia;
300 cases, 320 controls (death certificates);
1978-1981; cancer confirmed by hospital
records; white men
Job exposure matrix applied to detailed
information on all jobs held (at least 1 yr)
since age 15, as obtained from next-of-kin
interviews; probability, duration, intensity,
and cumulative exposure scores; six solvents
evaluated
See Section 4.1.3.6.1. OR 1.3 (0.9, 1.8) for
any exposure; increased risk with increased
probability (trends-value < 0.05, OR 2.4
[1.0, 5.9] for high probability), increased
duration, increased intensity; strongest
effects seen in high probability plus high
duration, OR 6.1 (1.1, 43.8) or high intensity
and high duration, OR 6.1 (1.5, 28.3)
combinations; no association with
cumulative exposure score
Brain
Coccoetal. (1999)
24 states (United States); 12,980 cases,
51,920 controls (death certificates); 1984-1992;
women
Job exposure matrix applied to death
certificate occupation; probability, and
intensity scores; 11 exposures evaluated
See Section 4.1.3.6.1. Weak association
overall, OR 1.2 (1.1, 1.3), no trend with
probability or intensity scores
Breast
Cantor etal. (1995)
24 states (United States); 33,509 cases,
117,794 controls (death certificates); 1984-
1989; black and white women
Job exposure matrix applied to death
certificate job data, probability, and exposure
level; 31 substances evaluated
See Section 4.1.3.6.2. Little evidence of
association with exposure probability; weak
association with highest exposure level in
whites, OR 1.17 (1.1, 1.3) and in blacks,
OR 1.46 (1.2, 1.7)
Pancreas
Kernanetal. (1999)
24 states (United States); 63,037 cases,
252,386 controls (death certificates); 1984-
1993; black and white men and women
Job exposure matrix applied to death
certificate occupation, probability, and
intensity scores; 11 chlorinated solvents and
formaldehyde evaluated
See Section 4.1.3.6.3. Little evidence of
associations with intensity or probability
Kidney
Dosemeci et al.
(1999)
Minnesota
438 incident cases (Minnesota cancer registry),
687 controls (random digit dialing and
Medicare records); 1988-1990; cancer
confirmed by histology; men and women
Job exposure matrices applied to most recent
and usual job, as ascertained from
interviews; nine solvents evaluated
See Section 4.1.3.6.4. No evidence of
increased risk associated with
dichloromethane in men, OR 0.85 (0.6, 1.2)
or women, OR 0.95 (0.4, 2.2)
(Table 4-10 continues on next page)
100
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-10. Summary of case-control studies of cancer risk and dichloromethane exposure
Cancer type,
reference
Location
n cases, n controls (source) and time period,
demographic group
Exposure assessment
Results"
Rectum
Dumas et al. (2000)
Montreal, Canada; 257 incident cases,
1,295 other cancer controls from 19 hospitals;
533 population-based controls (electoral rolls
and random digit dialing), 1979-1985; cancer
confirmed by histology; men
Job exposure matrix applied to detailed
information on all jobs held, as ascertained
from interviews; 294 substances evaluated
See Section 4.1.3.6.5. Little evidence of an
association with any exposure, OR 1.2 (0.5,
2.8), but increased risk in a small,
"substantial exposure" group, OR 3.8 (1.1,
12.2) (using cancer controls; analysis of
population controls not given for this
exposure)
Non-hodgkin
lymphoma
Seidler et al. (2007)
Germany (6 areas); 710 incident cases, 710
population-based controls (area population
files), 1999-2003; ages 18-80 years, men and
women.
Job exposure matrix applied to work history
(all jobs held at least 1 yr) ascertained
through interviews, job-specific and
industry-specific questionnaires (for solvent-
and other chemical-related jobs). Probability
and intensity ratings; 8 specific solvents
See Section4.1.3.6.
Cumulative exposure (ppm-yrs):
0 1.0 (referent)
>0to<26.3 0.4(0.7,5.2)
> 26.3 to <175 0.8(0.3, 1.9)
>175 2.2(04,11.6)
Non-hodgkin
lymphoma
Wang et al. (2009);
Barry etal. (2011)
Connecticut, United States; 601 incident cases,
717 population-based controls (random digit
dialing and Medicare files), 1996-2000; ages
21-84 years, women. Barry et al. (2011) is
limited to 518 cases and 597 controls with
blood or buccal cell sample for genotyping
Job exposure matrix applied to work history
(all jobs held at least 1 yr) ascertained
through interviews (job and industry titles,
duties). Probability and intensity ratings; 8
specific solvents
See Section4.1.3.6.
Association seen with dichloromethane
exposure: Ever OR 1.5 (1.0, 2.3)
Little difference in risk by probability or
intensity score. Stronger risk seen in diffuse
large B-cell lymphoma: OR 2.10 (1.15,
3.85), and in TT genotype of CYP2E1
rs20760673: OR 4.42 (2.03, 9.62)
Non-hodgkin
lymphoma
Miligi et al. (2006)
Italy (8 areas)
1,428 incident cases, 1,530 population-based
controls (area population files), 1991-1993;
cancer classification based on NCI protocol;
ages 20-74 years, men and women
Job exposure matrix applied to work history
(all jobs held at least 5 yrs) ascertained
through interviews, job-specific and
industry-specific questionnaires (for solvent-
and other chemical-related jobs). Probability
and intensity ratings; 10 specific solvents
See Section 4.1.3.6.6. Association with
intensity measure:
very low/low OR 0.9 (0.5, 1.6)
medium/high OR 1.7 (0.7, 4.3)
In small lymphocytic subtype, any exposure
OR 3.2 (1.0, 10.1)
Chronic lymphoid
leukemia
Costantini et al.
(2008)
Italy (7 areas)
586 incident cases, 1,278 population-based
controls (area population files), 1991-1993;
cancer classification based on NCI protocol;
ages 20-74 yrs, men and women
Job exposure matrix applied to work history
(all jobs held at least 5 yrs) ascertained
through interviews, job-specific and
industry-specific questionnaires (for solvent-
and other chemical-related jobs). Probability
and intensity ratings; 10 specific solvents
See Section 4.1.3.6.6. Association with
intensity measure:
very low/low OR 0.7 (0.3, 1.7)
medium/high OR 0.5 (0.1, 2.3)
Multiple myeloma
Costantini et al.
(2008)
Italy (6 areas)
263 incident cases, 1,100 population-based
controls (area population files), 1991-1993;
Job exposure matrix applied to work history
(all jobs held at least 5 yrs) ascertained
through interviews, job-specific and
See Section 4.1.3.6.6. Only 4 exposed cases;
association not estimated.
101
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-10. Summary of case-control studies of cancer risk and dichloromethane exposure
Cancer type,
reference
Location
n cases, n controls (source) and time period,
demographic group
Exposure assessment
Results"
cancer classification based on NCI protocol;
ages 20-74 yrs, men and women
industry-specific questionnaires (for solvent-
and other chemical-related jobs). Probability
and intensity ratings; 10 specific solvents
Multiple myeloma
Gold etal. (2011)
Seattle, Washington and Detroit, Michigan (2
SEER sites), United States
180 incident cases; 481 population-based
controls (random digit dialing and Medicare
files), 2000-2002; ages 35-74 yrs, men and
women
Job exposure matrix applied to work history
(all jobs held since age 15) ascertained
through interviews, job-specific
questionnaires (for solvent-related jobs held
at least 2 yrs). Probability, frequency,
intensity and confidence ratings; 6 specific
solvents
See Section4.1.3.6.6. Inanalyses
classifying low-confidence jobs as
unexposed:
Ever exposed OR 2.0 (1.2, 3.2)
Trends with duration (p = 0.01), cumulative
exposure (p = 0.08) and 10-yr lagged
cumulative exposure (p = 0.06)
Childhood leukemia
(acute lymphoblastic
leukemia)
Infante-Rivard et al.
(2005)
Quebec, Canada
790 incident cases (hospitals—all provinces),
790 population-based controls (government
population registries), 1980-2000; cancer based
on oncologist or hematologist diagnosis
ages 0-14 yrs,b boys and girls
Systematic review of detailed information on
all jobs held by the mother from 2 yrs before
pregnancy through birth of the child;
21 individual substances and six mixtures
evaluated (mostly solvents); confidence,
frequency, and concentration of exposure
ratings
See Section 4.1.3.6.6. Little evidence of
association with any exposure, OR 1.34
(0.54, 3.34), but stronger associations with
probable or definite, OR 3.22 (0.88, 11.7)
(referent group = possible/no exposure) and
with combinations of frequency and
concentration
"Results given as OR and (95% CI).
bFrom 1980 to 1993, study was limited to diagnoses of ages 0-9, but this was expanded between 1994 and 2000 to ages 0-14.
102
DRAFT - DO NOT CITE OR QUOTE
-------
Case-control studies offer the potential for increased statistical power for assessing
associations with relatively rare cancers such as brain cancer and leukemia. Case-control studies
are often designed to examine incidence rather than mortality, which is of particular importance
in etiologic research for diseases with relatively high survival rates and diseases in which
survival may be strongly related to factors that are difficult to adjust for without detailed data
collection (e.g., access to health care). There is a considerable range in the detail and quality of
the exposure assessment used in case-control studies, however. Case-control studies rarely
include specific measurements taken at specific work sites of individual study participants.
Although it is more difficult to determine absolute exposure levels without these individual
measurements, the exposure assessment methodology used in case-control studies can result in
useful between-group comparisons of risk if the intra-group variability is less than the intergroup
variability in potential exposure levels. The use of death certificate data to classify disease and
occupational exposures in the three studies using the large 24-state death certificate database
(brain cancer, Cocco et al. [1999]; breast cancer, Cantor et al. [1995]; pancreatic cancer, Kernan
et al. [1999]) is likely to have resulted in nondifferential misclassification of both outcome and
exposure (and thus attenuated associations). Several other case-control studies included
diagnoses based on medical records, and detailed job-specific and industry-specific questionnaire
modules focusing on potential exposure to specific solvents, and included assessment of intensity
and probability of exposure (Gold et al., 2011; Costantini et al., 2008; Seidler et al., 2007; Miligi
et al., 2006). The assessment methodology used in Wang et al. (2009) and Heineman et al.
(1994) also included the assessment of exposure intensity and probability, but were somewhat
more limited in that specific follow-up modules designed to obtain more detailed information for
jobs with potentially relevant exposures were not included. Each of these seven studies did,
however, obtain detailed information about all jobs held (or, in the case of the study of childhood
leukemia, jobs held in the two years before and during the pregnancy), rather than just the usual
or most recent job, and focused on a relatively small number of exposures. In addition,
Heineman et al. (1994) included more detailed coding of specific jobs within the 4-digit industry
and occupation code categories to distinguish those of particular relevance to a specific exposure
(e.g., production of paint removers within the broader category of production of paints,
varnishes, lacquers, enamels, and allied products) (Dosemici et al., 1994). Heineman et al.
(1994) obtained the work history from interviews with next-of-kin, however, which is also likely
to have resulted in nondifferential misclassification of exposure, and thus attenuation in the
observed associations.
Considering the issues described above with respect to the strengths and limitations of the
available epidemiologic studies, a summary of the epidemiologic evidence relating to
dichloromethane exposure and specific types of cancer can be made, as described below. The
available epidemiologic data suggest an association between dichloromethane and brain cancer,
liver cancer, and specific hematopoietic cancers, but not lung cancer.
103 DRAFT - DO NOT CITE OR QUOTE
-------
4.1.3.7.1. Brain and CNS cancer. An increased risk of brain and CNS cancers was seen in the
strongest cohort studies; SMRs were 2.16 in Cohort 1 in New York (Hearne and Pifer, 1999) and
1.45 in the United Kingdom cohort (Tomenson et al., 1997). These estimates are based on a
small number of observations (six cases in New York and four in the United Kingdom) and so
are relatively imprecise. It is only in the latest follow-up of the New York film base production
cohort that an elevated SMR was observed, further suggesting that the statistical power of the
other cohort studies for examining risk of this disease may be quite low. Two case-control
studies of dichloromethane exposure and brain cancer have been conducted (Cocco et al., 1999;
Heineman et al., 1994). The Heineman et al. (1994) study, which is the stronger study in terms
of exposure assessment strategy and confirmation of diagnosis, reported relatively strong trends
(p < 0.05) with increasing probability, duration, and intensity measures of exposure, but a
nonlinear trend was seen with the cumulative exposure metric. This difference could reflect a
more valid measure of relevant exposures in the brain from the intensity measure, as suggested
by the study in rats reported by Savolainen et al. (1981) in which dichloromethane levels in the
brain were much higher with a higher intensity exposure scenario compared with a constant
exposure period with an equivalent TWA (see Section 3.2). The combination of high probability
of exposure and long (> 20 years) duration of employment in exposed jobs was strongly
associated with brain cancer risk (OR 6.1, 95% CI 1.1-43.8) in the Heineman et al. (1994) study;
similar associations were seen with the high intensity in combination with long duration
measures. The available epidemiologic studies provide evidence of an association between
dichloromethane and brain cancer, and this area of research represents a data gap in the
understanding of the carcinogenic potential of dichloromethane.
4.1.3.7.2. Liver and biliary duct cancer. Liver and biliary duct cancer are relatively uncommon
(age-adjusted incidence 7.3 per 100,000 person-years) (SEER website, seer.cancer.gov, accessed
May 2, 2011), so it is difficult to study in most occupational cohorts of limited size. The cohort
study with the higher exposures, the Rock Hill triacetate fiber production plant, suggested an
increased risk of liver cancer (Lanes et al., 1993, 1990). The SMR for liver and cancer was
2.98 (95% CI 0.81-7.63) in the latest update of this cohort. This observation was based on four
cases; three of these cases were biliary duct cancers. As the follow-up period has increased, the
strength of this association has decreased, although it is relatively strong (albeit with wide CIs).
The decrease in the SMR with increasing follow-up reflects the increase in number of expected
cases because the four observed cases were seen earlier in the follow-up period. No other cohort
study has reported an increased risk of liver cancer mortality, although it should be noted that
there is no other inception cohort study of a population with exposure levels similar to those of
the Rock Hill plant, and no data from a case-control study of liver cancer are available pertaining
to dichloromethane exposure. The available epidemiologic studies, with biological plausibility
104 DRAFT - DO NOT CITE OR QUOTE
-------
inferred from the localization of GST-T1 in the nuclei of bile duct epithelial cells in human
samples (Sherratt et al., 2002), provide some evidence of an association between
dichloromethane and liver and biliary duct cancer, although it should be noted that this evidence
is based on limited epidemiologic data in that these observations were based on one study.
4.1.3.7.3. Lung cancer. In the stronger cohort studies (Cohort 1 in the New York Eastman
Kodak Company triacetate film production study reported by Hearne and Pifer [1999] and the
United Kingdom triacetate film production study reported by Tomenson et al. [1997]), the SMRs
for lung cancer were well below 1.0. The New York study had also obtained data on smoking
history that indicated it was unlikely that differences in smoking could be masking an effect of
dichloromethane (Hearne et al., 1987). Lung cancer is a common cancer (age-adjusted incidence
62 per 100,000 person-years) (SEER website, seer.cancer.gov, accessed May 2, 2011) so the
expected rates, even in small cohorts, are based on relatively robust estimates. The only group in
any study that had an increased risk for lung cancer was the high-exposure women in the
triacetate fiber production cohort in Maryland (Gibbs et al., 1996). However, this was based on
only two cases and was a highly imprecise estimate (SMR 2.3 [95% CI 0.28-8.3]). No case-
control study of dichloromethane exposure and lung cancer risk is available. The available
epidemiologic studies do not provide evidence for an association between dichloromethane and
lung cancer, although it should be noted that the studies with the best designs are limited to
relatively low exposure levels.
4.1.3.7.4. Pancreatic cancer. An early study (Hearne et al., 1990) of Cohort 2 of the New York
triacetate film production cohort had reported 8 observed and 4.2 expected pancreatic cancer
deaths, for a twofold increased SMR (p = 0.13). This association was reduced in the subsequent
follow-up (SMR 1.5 [95% CI 0.7-3.0]) (Hearne and Pifer, 1999) but was not seen in the more
methodologically sound Cohort 1 (SMR 0.92) or in any of the other cohorts. A meta-analysis of
the cohort studies (using the data of Hearne et al. [1990]) reported a summary association of
1.42 (95% CI 0.80-2.53) (Ojajarvi et al., 2001). This summary measure would be further
reduced with the updated data for Cohort 2 and the addition of Cohort 1 from Hearne and Pifer
(1999). The only case-control study of pancreatic cancer mortality risk and dichloromethane
exposure (based on death certificate data) did not report consistent patterns with respect to
intensity or exposure among the race-sex groups studied. The available epidemiologic studies do
not provide evidence for an association between dichloromethane and pancreatic cancer.
4.1.3.7.5. Leukemia and lymphoma. Each of the individual hematopoietic cancers is relatively
uncommon, with age-adjusted incidence rates ranging from 1.6 to 4.2 per 100,000 person-years
for specific types of leukemias (SEER website, seer.cancer.gov, accessed May 2, 2011); the
incidence rate for non-Hodgkin lymphoma is 19.8 per 100,000 person-years. The relatively
105 DRAFT - DO NOT CITE OR QUOTE
-------
inconsistent (point estimates ranging from <0.50 to >2.0) and imprecise measures of association
in the cohort studies between dichloromethane exposure and leukemia are thus expected, given
the relatively small size of the available cohorts. Three large population-based case-control
studies of incident non-Hodgkin lymphoma in adults in Germany (Seidler et al., 2007), Italy
(Miligi et al., 2006,) and Connecticut (Barry et al., 2011; Wang et al., 2009) observed ORs
between 1.5 and 2.2 with dichloromethane exposure (ever exposed, or highest category of
exposure). There was also some evidence of higher risk among specific subsets of disease, such
small lymphocytic non-Hodgkin lymphoma (Miligi et al., 2006) or diffuse large B-cell
lymphoma (Barry et al., 2011). An extensive exposure assessment protocol was used in the
studies from Italy and Germany, including job-specific and industry-specific questionnaire
modules focusing on potential exposure to specific solvents. No association was seen in a
related study of chronic lymphoid leukemia, conducted in Italy using the same study protocol
(Costantini et al., 2008). Only four cases of multiple myeloma within this study were exposed to
dichloromethane, and risk estimates were not presented. In a population-based case-control
study of multiple myeloma conducted in two SEER sites in the United States (also using an
extensive exposure assessment protocol focusing on specific solvents), a twofold increased risk
was seen with ever exposure to dichloromethane (OR 2.0, 95% CI 1.2-3.2), with trends seen
with duration, cumulative exposure, and cumulative exposure lagged by 10 years (Gold et al.,
2011). One case-control study of childhood leukemia (acute lymphoblastic leukemia) examining
maternal occupational history and dichloromethane exposure is available (Infante-Rivard et al.,
2005). This is a large, population-based study of incident cases of leukemia, with a detailed
exposure assessment pertaining to the period before and during pregnancy. A threefold
increased risk was seen with probable or definite exposure (OR 3.22 [95% CI 0.88-11.7])
compared with possible or no exposure.
Thus, the available epidemiologic studies show consistent observations of associations
between dichloromethane and non-Hodgkin lymphoma, but do not definitively establish that
dichloromethane exposure is related to an increased cancer risk. These studies are limited by
relatively small number of exposed cases, resulting in imprecise effect estimates.
4.1.3.7.6. Breast cancer. Only one large cohort study included women and reported data
pertaining to breast cancer risk (Radican et al., 2008; Blair et al., 1998), and this is a cohort with
a limited exposure assessment (dichotomized) and multiple exposures. A twofold increased risk
was seen between dichloromethane exposure and breast cancer mortality in this study (rate ratio
2.35 [95% CI 0.98-5.65]). Similar associations were seen with several other chemicals, and the
potential effect of confounding and misclassification of these exposures may have biased the
estimate in either direction. The only case-control study of breast cancer risk and
dichloromethane exposure used the 24-state death certificate data to classify exposure and
disease. The available epidemiologic studies do not provide an adequate basis for the evaluation
106 DRAFT - DO NOT CITE OR QUOTE
-------
of the role of dichloromethane in breast cancer because there are currently no cohort studies with
adequate statistical power and no case-control studies with adequate exposure methodology to
examine this relationship.
4.2. SUBCHRONIC AND CHRONIC STUDIES AND CANCER BIOASSAYS IN
ANIMALS—ORAL AND INHALATION
4.2.1. Oral Exposure: Overview of Noncancer and Cancer Effects
Results from studies of animals exposed by the oral route for short-term, subchronic, and
chronic durations identify the liver and the nervous system as the most sensitive targets for
noncancer toxicity from repeated oral exposure to dichloromethane. In a 90-day exposure study,
nonneoplastic histopathologic changes in the liver were observed in F344 rats exposed to
drinking water doses of >166 mg/kg-day (males) or>209 mg/kg-day (females) (Kirschman et al.,
1986). Similar changes were seen in F344 rats in a 2-year exposure of >50 mg/kg-day (Serota et
al., 1986a).
The 2-year oral exposure study in F344 rats did not produce evidence of increasing
incidence of liver tumors across all of the dose groups in males or females (Serota et al., 1986a).
In females, however, a jagged stepped pattern of increasing incidence was observed. In a
parallel study in B6C3Fi mice (Serota et al., 1986b; Hazleton Laboratories, 1983), a clearer trend
with respect to hepatic cancer was seen in males but not females.
None of the chronic oral exposure studies included a systematic measurement of potential
neurological effects. One 14-day study focusing on neurobehavioral changes is available,
however. Changes in autonomic, neuromuscular, and sensorimotor functions were observed in
F344 rats exposed for 14 days to gavage doses >337 mg/kg-day (Moser et al., 1995) (see
Section 4.4.3 for more details).
No effects on reproductive parameters were observed in Charles River CD rats exposed
for 90 days to gavage doses as high as 225 mg/kg-day (General Electric Company, 1976) or in
pregnant F344 rats exposed to gavage doses of up to 450 mg/kg-day on GDs 6-19 (Narotsky and
Kavlock, 1995). However, no oral exposure studies examining developmental neurobehavioral
effects have been conducted (see Section 4.3 for more details).
4.2.1.1. Toxicity Studies of Subchronic Oral Exposures: Hepatic Effects
Kirschman et al. (1986) examined the toxicity of dichloromethane in a 90-day drinking
water study in F344 rats (20/sex/dose level). The nominal concentration of dichloromethane in
the water was 0.15, 0.45, or 1.5%. Based on BW and water consumption data, average intakes
were reported to be 0, 166, 420, or 1,200 mg/kg-day for males and 0, 209, 607, or 1,469 mg/kg-
day for females. Clinical chemistry tests (hematological and chemical variables in samples of
blood and urine) and tissue histopathology were evaluated in groups of five rats/sex/dose level
107 DRAFT - DO NOT CITE OR QUOTE
-------
after 1 month of treatment. These endpoints were also evaluated in the remaining rats sacrificed
after 90 days of exposure.
Exposure to dichloromethane did not affect mortality or cause adverse clinical signs of
toxicity. Gross necropsy was also unremarkable. Reported changes in mean values for clinical
chemistry variables compared with controls included elevated serum ALT activities for all
treated males at 1 month and for the high-dose females at 3 months, elevated serum AST activity
in high-dose females at 3 months, elevated serum lactate dehydrogenase activities in mid- and
high-dose females at 3 months, and decreases in serum concentrations of fasting glucose,
cholesterol, and triglycerides in all exposed groups of both sexes at 1 and 3 months. Actual
values for clinical chemistry variables or the magnitude of the changes, however, were not
presented in the report.
No histopathologic alterations were seen in tissues after 1 month of treatment (a detailed
description of tissues examined was not presented). In rats exposed for 3 months, exposure-
related histopathologic changes were restricted to the liver. Elevated, statistically significant
incidences of hepatocytic vacuolation were observed in all exposed male and female groups (see
Table 4-11). The most frequently observed vacuolation was described as generalized and
occurring throughout the lobule, and Oil Red-O-staining indicated that most were lipid-
containing vacuoles. The incidences of generalized vacuolation scored as mild or moderate were
higher in all of the female dose groups compared with the controls. The authors stated that the
no-observed-adverse-effect level (NOAEL) based on this study is <200 mg/kg-day and the
lowest-observed-adverse-effect level (LOAEL) for males was 166 mg/kg-day. The authors did
not explicitly provide a LOAEL for females. The results indicate that 166 mg/kg-day and
209 mg/kg-day were the LOAELs for liver effects in male and female rats, respectively.
108 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-11. Incidences of histopathologic changes in livers of male and
female F344 rats exposed to dichloromethane in drinking water for 90 days
Lesion, by sex
Males — n per group3
Estimated mean intake (mg/kg-d)
Number (%) with:
Hepatocyte vacuolation (generalized, centrilobular,
or periportal)
Generalized vacuolation severity:
minimal
mild
moderate
Centrilobular severity:
minimal
mild
moderate
Hepatocyte degeneration
Focal granuloma
Females — n per group3
Estimated mean intake (mg/kg-d)
Number (%) with:
Hepatocyte vacuolation (generalized, centrilobular,
or periportal)
Generalized vacuolation severity:
minimal
mild
moderate
Centrilobular severity:
minimal
mild
moderate
marked
Hepatocyte degeneration
Focal granuloma
Controls
15
0
1(7)
0(0)
0
0
0
0(0)
0
0
0
0(0)
1(7)
15
0
6(40)
5(33)
5
0
0
0(0)
0
0
0
0
0(0)
0(0)
Low dose
15
166
10b (67)
5b (33)
4
0
1
1(7)
1
0
0
0(0)
0(0)
15
209
13b (87)
13b (87)
8
4
1
0(0)
0
0
0
0
0(0)
0(0)
Mid dose
15
420
9b (60)
8b (53)
7
1
0
0(0)
0
0
0
0(0)
0(0)
15
607
15b (100)
15b (100)
6
5
4
1(7)
0
1
0
0
0(0)
4 (27)c
High dose
15
1,200
7b (47)
6b (40)
6
0
0
2(13)
0
2
0
2(13)
1(7)
15
1,469
15b (100)
15b (100)
8
6
1
llb(28)
2
4
o
J
2
12b (80)
6b (40)
320 per group; 5 sacrificed at 1 mo; these endpoints for the remaining 15 per group.
bStatistical significance testing not reported by authors; Fisher's exact test for comparison with control
p-value < 0.05 (two-sided).
Statistical significance testing not reported by authors; Fisher's exact test for comparison with control
p-value < 0.10 (two-sided). Authors stated LOAEL =166 mg/kg-d in males but did not explicitly provide LOAEL
for females; NOAEL is <200 mg/kg-d.
Source: Kirschmanetal. (1986).
Kirschman et al. (1986) conducted a similar 90-day study in B6C3Fi mice (20/sex/dose
level). The estimated average intakes were 0, 226, 587, or 1,911 mg/kg-day for males and 231,
586, or 2,030 mg/kg-day for females. Six mice (two controls, two low dose, and two mid dose)
died during the study from unknown causes. Administration of dichloromethane did not cause
109
DRAFT - DO NOT CITE OR QUOTE
-------
adverse clinical signs of toxicity or affect food consumption, ophthalmology, or serum ALT
activity. Gross necropsy examinations were also unremarkable.
Histopathologic evaluation of tissues from mice killed after 1 month of treatment did not
reveal any compound-related effects. Evaluation at 3 months showed subtle generalized or
centrilobular changes in the liver (characterized as increased vacuolation with fat deposition),
which was evident in all exposed groups and most prominent in mid- and high-dose female
groups (Table 4-12). The most frequently detected change was characterized as a generalized
vacuolation. Some evidence was found for an increase in severity of the generalized vacuolation
with dichloromethane exposure, but the incidence of this lesion in the control mice was
substantial, especially in females (Table 4-12). Incidences for centrilobular vacuolation were
significantly increased only for the mid-dose female group. No other changes were found.
110 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-12. Incidences of histopathologic changes in livers of male and
female B6C3Fi mice exposed to dichloromethane in drinking water for
90 days
Lesion, by sex
Males — n per group3
Estimated mean intake (mg/kg-d)
Number (%) with:
Hepatocyte vacuolation (generalized, centrilobular,
or periportal)
Generalized vacuolation, severity:
minimal
mild
moderate
marked
Centrilobular severity:
minimal
mild
moderate
Females — n per group3
Estimated mean intake (mg/kg-d)
Number (%) with:
Hepatocyte vacuolation (generalized, centrilobular,
or periportal)
Generalized vacuolation severity:
minimal
mild
moderate
marked
Centrilobular severity:
minimal
mild
moderate
marked
Controls
14
0
9(64)
7(50)
4
2
1
0
2(14)
2
0
0
14
0
13 (93)
13 (93)
1
8
4
0
0(0)
0
0
0
0
Low dose
14
226
12 (86)
12b (86)
3
7
2
0
0(0)
0
0
0
11
231
11(100)
11(100)
3
7
1
0
0(0)
0
0
0
0
Mid dose
14
587
13 (93)
13b(93)
9
5
0
0
1(7)
0
0
1
13
586
13 (100)
13 (100)
5
6
2
0
5C (39)
0
2
3
0
High dose
15
1,911
12 (80)
10 (67)
7
3
0
0
5(33)
1
3
1
15
2,030
13 (87)
13 (87)
3
6
1
3
1(7)
0
1
0
0
320 per group; 5 sacrificed at 1 mo.
bStatistical significance testing not reported by authors; Fisher's exact test for comparison with control
p-value = 0.10 for low dose group andp = 0.032 for mid-dose group (two-sided).
Statistical significance testing not reported by authors; Fisher's exact test for comparison with control
^-value = 0.016 (two-sided). Authors say LOAEL = 587 mg/kg-d; NOAEL between 226 and 587 mg/kg-d for
males; not explicitly stated for females.
Source: Kirschmanetal. (1986).
Using the results from this study to select doses for a chronic study, Kirschman et al.
(1986) expressed the opinion that the mid-dose level (587 mg/kg-day) was the LOAEL in this
study. Although incidences for generalized vacuolation were increased in the low- and mid-dose
male groups, the incidences in the high-dose groups were not significantly increased compared
with controls (Table 4-12). The study authors identified a LOAEL of 587 mg/kg-day for
centrilobular vacuolation in male B6C3Fi mice. The NOAEL for males was considered by the
investigators to be between 226 and 587 mg/kg-day.
Ill
DRAFT - DO NOT CITE OR QUOTE
-------
4.2.1.2. Toxicity Studies of Chronic Oral Exposures: Hepatic Effects and Carcinogenicity
Longer-term (up to 2-year) oral exposure studies in mice and rats are summarized in
Table 4-13 and described in more detail below. These studies provide additional information
pertaining to hepatotoxicity and carcinogenicity.
Table 4-13. Studies of chronic oral dichloromethane exposures (up to
2 years)
Reference,
strain/species
Scrota etal. (1986a)
F344 rats
Scrota etal. (1986b);
Hazleton
Laboratories (1983)
B6C3FJ mice
Maltoni etal. (1988)
Sprague-Dawley
rats
Maltoni etal. (1988)
Swiss mice
Number per group
85/sex/dose +
135 controls
Males: 125,200,
100, 100, 125
Females: 100, 100,
50, 50, 50
50/sex/dose
50/sex/dose + 60
controls
Exposure information
Drinking water, 2 yrs, target dose 0,
5, 50, 125, 250 mg/kg-d
Mean intake:
males: 0, 6, 52, 125, 235 mg/kg-d
females: 0, 6, 58, 136,
263 mg/kg-d
Drinking water, 2 yrs, target dose 0,
60, 125, 185, 250 mg/kg-d
Mean intake:
males: 0, 61, 124, 177,
234 mg/kg-d
females: 0,59, 118, 172,
238 mg/kg-d
Gavage, up to 64 wks
0, 100, 500 mg/kg-d, 4-5 d/wk
Gavage, up to 64 wks
0, 100, 500 mg/kg-d, 4-5 d/wk
Comments
Nonneoplastic liver effects
(foci/areas of alteration) in males
and females (see Table 4-14);
jagged stepped pattern of
increasing incidence of neoplastic
nodules or hepatocellular
carcinoma in females (i.e.,
increased in the 50 and 250 mg/kg-
d but not 125 mg/kg-d groups) (see
Table 4-14)
Increasing trend of liver cancer
(hepatocellular adenoma or
carcinoma) in males (see
Table 4-15)
High mortality in high dose group
led to termination of study at
64 wks; statistically nonsignificant
increase in malignant mammary
tumors in female rats
High mortality in high dose group
led to termination of study at
64 wks
4.2.1.2.1. Chronic oral exposure in F344 rats (Scrota et al, 1986a). Treatment with
dichloromethane in drinking water did not induce adverse clinical signs or affect survival in the
F344 rats (Serota et al., 1986a). BWs of rats in the 125 and 250 mg/kg-day groups were
generally lower than in controls throughout the study. The authors stated that the differences,
although small, were statistically significant, but the data were not shown in the published report.
Water consumption was lower throughout the study in both sexes of rats from the 125 and
250 mg/kg-day groups relative to controls; food consumption was also lower in these groups
during the first 13 weeks of treatment. Mean hematocrit, hemoglobin, and red blood cell count
were increased in both sexes at dichloromethane levels of 50, 125, and 250 mg/kg-day for 52 and
78 weeks. Half of these increases were reported to be statistically significant, but the report did
not provide the numerical values or specify which parameters were significant. Clinical
112
DRAFT - DO NOT CITE OR QUOTE
-------
chemistry results showed decreases in alkaline phosphatase (AP), creatinine, blood urea nitrogen,
total protein, and cholesterol in both sexes at 250 mg/kg-day, and most of these changes were
statistically significant at one or both of the intervals evaluated. (Significant parameters not
specified and the mean group values were not presented in the published report.) No significant
deviations in urinary parameters were observed. Organ weights were not significantly affected
by treatment with dichloromethane.
No treatment-related histopathological effects were noted in the tissues examined except
for the liver (Serota et al., 1986a). Examination of liver sections showed a dose-related positive
trend (positive Cochran-Armitage trend test) in the incidences of foci/areas of cellular alteration
in treated F344 rats (Table 4-14). Comparisons of incidences with control incidences indicated
statistically significant elevations at all dose levels except 5 mg/kg-day. These liver changes
were first noted after treatment for 78 weeks and progressed until week 104. Livers of animals
treated with dichloromethane also showed an increased incidence of fatty change, but incidence
data for this lesion were not presented in the published report. The recovery group also showed
an increased incidence of areas of cellular alterations, but the fatty changes were less pronounced
than in the 250 mg/kg-day group dosed for 104 weeks. The authors indicate that 5 mg/kg-day
was a NOAEL and 50 mg/kg-day was a LOAEL for liver changes (foci/areas of cellular
alteration) in male and female F344 rats exposed to dichloromethane in drinking water for
2 years.
113 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-14. Incidences of nonneoplastic liver changes and liver tumors in
male and female F344 rats exposed to dichloromethane in drinking water for
2 years
Target dose (mg/kg-d)
Oa
(Controls)
5
50
125
250
Trend
/7-valueb
250 with
recovery0
Males
Estimated mean intake (mg/kg-d)
total n
n at terminal killd
Number (%) with:
Liver foci/areas of alteration
Neoplastic nodules
Hepatocellular carcinoma
Neoplastic nodules and
hepatocellular carcinoma
0
135
76
52 (70)
9(12)
3(4)
12 (16)
6
85
34
22 (65)
1(3)
0(0)
1(3)
52
85
38
35 (92)e
0(0)
0(0)
0(0)
125
85
35
34 (97)e
2(6)
0(0)
2(6)
235
85
41
40 (98)e
1(2)
1(2)
2(5)
0.0001
Not
reported
Not
reported
Not
reported
232
25
15
15 (100)e
2(13)
0(0)
2(13)
Females
Estimated mean intake (mg/kg-d)
total n
n at terminal killd
Number (%) with:
Liver foci/areas of alteration
Neoplastic nodules
Hepatocellular carcinoma
Neoplastic nodules and
hepatocellular carcinoma
0
135
67
34(51)
0(0)
0(0)
0(0)
6
85
29
12(41)
1(3)
0(0)
1(3)
58
85
41
30 (73)e
2(5)
2(5)
4 (10)f
136
85
38
34 (89)e
1(3)
0(0)
1(3)
263
85
34
31(91)e
3(9)
2(6)
5 (14)f
0.0001
Not
reported
Not
reported
p<0.0\
239
25
20
17 (85)e
2(10)
0(0)
2 (10)f
"Two control groups combined. Sample size (incidence of liver foci) in group 1 and 2, respectively, was 36 (75%)
and 40 (63%) in males and 31 (55%) and 36 (47%) in females.
bCochran-Armitage trend test was used for trend test of liver foci/areas of alteration. For tumor mortality-
unadjusted analysis, a Cochran-Armitage trend test was used, and for tumor mortality-adjusted analyses, tumor
prevalence analytic method by Dinse and Lagakos (1982) was used. Similar results were seen in these two
analyses.
'Recovery group was exposed for 78 wks and then had a 26-wk period without dichloromethane exposure; n = 15
for nonneoplastic lesions and n = 17 for neoplastic lesions.
dExcludes 5, 10, and 20 per group sacrificed at 25, 52, and 78 wks, respectively, and unscheduled deaths, which
ranged from 5 to 19 per group.
eSignificantly (p < 0.05) different from control with Fisher's exact test.
Significantly (p < 0.05) different from controls with Fisher's exact test, mortality-unadjusted and mortality-
adjusted analyses.
Source: Serotaetal. (1986a).
Dichloromethane-exposed male rats showed no statistically significant increased
incidence of liver tumors. In females, there was a positive trend for increasing incidence of
114
DRAFT - DO NOT CITE OR QUOTE
-------
hepatocellular carcinoma or neoplastic nodules with increasing dose (Table 4-14) (Serota et al.,
1986a). Statistically significant increases in tumor incidences were observed in the 50 and
250 mg/kg-day groups (incidence rates of 10 and 14%, respectively) but not in the 125 mg/kg-
day group (incidence rate of 3%). Incidence of neoplastic nodules was also increased (10%) in a
group exposed for 78 weeks followed by a 26-week period of no exposure; however, the
characterization of malignant potential of the nodules was not described. The incidence of
hepatocellular carcinoma or neoplastic nodules in this control group (0%) was lower than that
seen in historical controls from the same laboratory (324 female F344 rats; 4 with carcinoma,
21 with neoplastic nodules; 25/324 = 7.7%).
4.2.1.2.2. Chronic oral exposure in B6C3Fi mice (Serota et al, 1986b; Hazleton Laboratories,
1983). A 2-year drinking water study similar to the previously described study in F344 rats was
also conducted in B6C3Fi mice (Serota et al., 1986b; Hazleton Laboratories, 1983). The mice
received target doses of 0, 60, 125, 185, or 250 mg/kg-day of dichloromethane in deionized
drinking water for 24 months. Treatment groups consisted of 100 female mice in the low-dose
(60 mg/kg-day) group and 50 in the remaining treatment groups; larger sample sizes were used
in the male bioassay, with 200, 100, 100, and 125 male mice in the 60, 125, 185 and 250 mg/kg-
day groups, respectively. One hundred females (in two groups of 50) and 125 males (in two
groups of 60 and 65 mice) served as controls. The authors indicate that this study design
involving two groups of control mice was used because of the high and erratic incidence of liver
tumors in historical control B6C3Fi mice; when the results were similar in the two control
groups, the groups could be combined to provide a more statistically precise estimate for
comparisons with the exposed groups. Based on water consumption and BW measurements,
mean intakes were reported to be 61, 124, 177, and 234 mg/kg-day for males and 59, 118, 172,
and 238 mg/kg-day for females. Endpoints examined included clinical signs, BW and water
consumption, hematology at weeks 52 and 104, and gross and microscopic examinations of
tissues and organs at termination. All tissues from the control and 250 mg/kg-day groups were
examined microscopically, as well as the livers and neoplasms from all groups and the eyes of all
males from all groups.
Throughout the 2-year study, mice from both control and treatment groups exhibited a
high incidence of convulsions (Serota et al., 1986b; Hazleton Laboratories, 1983). The
convulsions were noted only during handling for BW determinations, and efforts to establish a
basis for this response were unsuccessful. The incidence of convulsions did not correlate with an
increased mortality rate. Survival to 104 weeks was high (82% in males and 78% in females),
and no evidence for exposure-related negative effects on survival were found. Exposure had no
significant effect on BW or water consumption. Mean leukocyte count was significantly
elevated in males and females dosed with 250 mg/kg-day dichloromethane for 52 weeks, but the
authors indicated that the mean values were within the normal historical range for the laboratory.
115 DRAFT - DO NOT CITE OR QUOTE
-------
Treatment-related nonproliferative histopathologic effects were restricted to the liver and
consisted of a marginal increase in the amount of Oil Red O-positive material in the liver of
males and females dosed with 250 mg/kg-day (group incidences for this lesion, however, were
not presented in the published report). The results indicate that 185 mg/kg-day was a NOAEL
and 250 mg/kg-day was a LOAEL for marginally increased amounts of fat in livers of male and
female B6C3Fi mice.
Incidences of liver tumors in female mice were not presented in the published reports
(Serota et al., 1986b; Hazleton Laboratories, 1983), but it was reported that exposed female mice
did not show increased incidences of proliferative hepatocellular lesions. In the male B6C3Fi
mice, incidences for hepatic focal hyperplasia showed no evidence of an exposure-related effect
(Table 4-15). The incidence of hepatocellular adenomas or carcinomas was 18 and 20% in each
of the two control groups. Because of the similarity in the results for these groups, the combined
group is presented in this table and used as the comparison group for the analysis. The incidence
of hepatocellular adenomas or carcinomas across exposure groups was 26, 30, 31, and 28% in
the 60, 125, 185 and 250 mg/kg-day groups, respectively. Similar patterns are seen with the
mortality-adjusted incidences (Table 4-15). The trend tests and the tests of the comparisons
between individual exposure groups and the controls were not reported by Serota et al. (1986b)
but were reported in Hazleton Laboratories (1983). Exposed male mice showed a marginally
increased combined incidence of hepatocellular adenomas and carcinomas, with a linear trendy-
value = 0.058; the individuals-values for the 60, 125, 185, and 250 mg/kg-day dose groups were
0.071, 0.023, 0.019, and 0.036, respectively.
116 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-15. Incidences for focal hyperplasia and tumors in the liver of male
B6C3Fi mice exposed to dichloromethane in drinking water for 2 years
n per group0
Estimated mean intake (mg/kg-d)
Number (%) with:
Focal hyperplasiad
Hepatocellular adenoma
( mortality-adjusted percent)
/>-valuee
Hepatocellular carcinoma
(mortality-adjusted percent)
/>-valuee
Hepatocellular adenoma or carcinoma
(mortality-adjusted percent)
/>-valuee
Target dose (mg/kg-d)
Oa
(Controls)
125
0
10(8)
10(8)
(9)
14(11)
(13)
24 (19)
(21)
60
200
61
14(7)
20 (10)
(12)
p = 0.24
33 (17)
(19)
p = 0.082
51 (26)
(29)
;? = 0.071
125
100
124
4(4)
14 (14)
(17)
p = 0.064
18(18)
(21)
p = 0.073
30 (30)
(34)
p = 0.023
185
99
177
10(10)
14(14)
(16)
p = 0.076
17(17)
(19)
p = 0.11
31(31)
(34)
;? = 0.019
250
125
234
13 (10)
15 (12)
(12)
p = 0.l3
23 (18)
(21)
p = 0.044
35 (28)
(32)
0.036
Trend
/7-valueb
Not
reported
0.172
0.147
0.058
aTwo control groups combined. Sample size (incidence of hepatocellular adenoma or carcinoma) in group 1 and 2,
respectively, was 60 (18%) and 65 (20%). Two additional sets of analyses using the individual control groups were
also presented in Hazleton Laboratories (1983).
bCochran-Armitage trend test (source: Hazleton Laboratories [1983]).
°Number at start of treatment.
dSome mice with hyperplasia also had hepatocellular neoplasms, but the exact number was unspecified by Serota et
al. (1986b).
ePercent calculated based on number at risk, using Kaplan-Meier estimation, taking into account mortality losses;
p-value for comparison with control group, using asymptotic normal test (source: Hazleton Laboratories [1983]).
Sources: Serota et al. (1986b); Hazleton Laboratories (1983).
Serota et al. (1986b) summarized these results as showing slight increases in proliferative
hepatocellular lesions in exposed male B6C3Fi mice that were not dose-related and were within
the range of historical controls, with no effect seen in female B6C3Fi mice. Serota et al. (1986b)
concluded that dichloromethane "did not induce a treatment-related carcinogenic response in
B6C3Fi mice" under the conditions of this study. An alternative conclusion, as determined by
EPA based on the results of the analysis shown in Hazleton Laboratories (1983), is that
dichloromethane induced a carcinogenic response in male B6C3Fi mice as evidenced by small
but statistically significant (p < 0.05) increases in hepatocellular adenomas and carcinomas at
dose levels of 125, 185, and 250 mg/kg-day, and by a marginally increased trend test (p = 0.058)
for combined hepatocellular adenomas and carcinomas. The incidence in the control groups was
almost identical to the mean seen in the historical controls (17.8%, based on 354 male B6C3Fi
mice), so there is no indication that the observed trend is being driven by an artificially low
incidence in controls. There is also no indication that the experimental conditions resulted in a
systematic increase in the incidence of hepatocellular adenomas and carcinomas. Given the
information provided regarding the incidence in historical controls (mean 17.8%, range 5 to
117
DRAFT - DO NOT CITE OR QUOTE
-------
40%), the pattern of results (increased incidence in all four dose groups, with three of these
increases significant at a p-value < 0.05) indicates a treatment-related increase.
One reason for the difference between Serota et al. (1986b) and EPA in the interpretation
of these data is the difference in the significance level used to evaluate the between-group
comparisons. EPA used the standard two-tailed significance level ofp = 0.05. Serota et al.
(1986b) state that a two-tailed significance level ofp = 0.05 was used for all tests, but this does
not appear to correctly represent the statistic used by Serota et al. (1986b). As can be seen by the
/7-values in Table 4-15, each of the/?-values for the comparison of the 125, 185, and 250 mg/kg-
day dose groups with the controls wasp < 0.05. (As noted previously, theses-values were found
in the full report of this study, see Hazleton Laboratories, 1983, but were not included in the
Serota et al., 1986b publication). Interpretation of the study findings in Serota et al. (1986b)
appears to be based on a statistical significance level of 0.0125 rather than 0.05; none of the
group comparisons shown in Table 4-15 are statistically significant when ap-va\ue of 0.0125 is
used. Hazleton Laboratories (1983) indicated that a correction factor for multiple comparisons
was used specifically for the liver cancer data, reducing the nominal p-va\ue from 0.05 to 0.0125;
this correction was not identified by Serota et al. (1986b). A multiple comparisons correction is
sometimes advocated in situations examining many different types of effects (e.g., >20
individual causes of death) or many different types of exposures (>20 different chemicals or
several hundred genes) to protect against inappropriately focusing on spurious findings. EPA
concluded that the use of this multiple comparisons correction factor specifically for a primary
hypothesis under investigation in the 2-year mouse oral carcinogenicity assay conducted by
Hazleton Laboratories is not warranted.
4.2.1.2.3. Chronic oral exposure in Sprague-Dawley rats and Swiss mice (Maltoni et al,
1988). Maltoni et al. (1988) conducted gavage carcinogenicity studies in Sprague-Dawley rats
and in Swiss mice. Groups of rats (50/sex/dose level) were gavaged with dichloromethane
(99.9% pure) in olive oil at dose levels of 0 (olive oil), 100, or 500 mg/kg-day 4-5 days/week for
64 weeks. This dosing regime was also used for groups of Swiss mice (50/sex/dose level plus
60/sex as controls). Endpoints monitored included clinical signs, BW, and full necropsy at
sacrifice (when spontaneous death occurred). For each animal sacrificed, histopathologic
examinations were performed on the following organs: brain and cerebellum, zymbal glands,
interscapular brown fat, salivary glands, tongue, thymus and mediastinal lymph nodes, lungs,
liver, kidneys, adrenals, spleen, pancreas, esophagus, stomach, intestine, bladder, uterus, gonads,
and any other organs with gross lesions. High mortality was observed in male and female high-
dose rats (data not shown) and achieved significance (p < 0.01) in males. The increased
mortality became evident after 36 weeks of treatment and led to the termination of treatment at
week 64. Explanation of the mortality was not provided by the study authors. As with the rats,
118 DRAFT - DO NOT CITE OR QUOTE
-------
high mortality occurred in male and female mice from the high-dose group (p < 0.01), and the
exposure was terminated after 64 weeks.
Little information is provided regarding nonneoplastic effects (Maltoni et al., 1988).
Treatment with dichloromethane did not affect BW in the Sprague-Dawley rats. A reduction in
BW was apparent in treated mice after 36-40 weeks of treatment, but no data were shown to
determine the magnitude of the effect. The lack of reporting of nonneoplastic findings from the
histopathologic examinations precludes assigning NOAELs and LOAELs for possible
nonneoplastic effects in these studies.
The Maltoni et al. (1988) studies of Sprague-Dawley rats and Swiss mice did not find
distinct exposure-related carcinogenic responses following gavage exposure to dichloromethane
at dose levels up to 500 mg/kg-day, although the early termination of the study (at 64 weeks)
limits the interpretation of this finding. Dichloromethane exposure was not related to the
percentage of either study animal bearing benign and/or malignant tumors or to the number of
total malignant tumors per 100 animals. High-dose female rats showed an increased incidence in
malignant mammary tumors, mainly due to adenocarcinomas (8, 6, and 18% in the control, 100,
and 500 mg/kg dose groups, respectively; the number of animals examined was not provided),
but the increase was not statistically significant. A dose-related increase, although not
statistically significant, in pulmonary adenomas was observed in male mice (5, 12, and 18% in
control, 100, and 500 mg/kg-day groups, respectively). When mortality was taken into account,
high-dose male mice that died in the period ranging from 52 to 78 weeks were reported to show a
statistically significantly (p < 0.05) elevated incidence for pulmonary tumors (1/14, 4/21, and
7/24 in control, 100, and 500 mg/kg-day groups, respectively). Details of this analysis were not
provided. EPA applied a Fisher's exact test to these incidences and determined a/>-value of
0.11 for the comparison of the 500 mg/kg-day group (7/24) to the controls (1/14).
4.2.2. Inhalation Exposure: Overview of Noncancer and Cancer Effects
Inhalation dichloromethane exposure studies in rats and mice using subchronic and
chronic durations identify the CNS, liver, and lungs as potential toxicity targets. Data from other
studies indicate that hamsters are less susceptible to the nonneoplastic and neoplastic effects of
dichloromethane than are rats and mice.
Increased incidences of nonneoplastic liver lesions were observed in Sprague-Dawley
rats exposed to >500 ppm for 2 years (Nitschke et al., 1988a; Burek et al., 1984), F344 rats
exposed to >1,000 ppm for 2 years (Mennear et al., 1988; NTP, 1986), and B6C3Fi mice
exposed to >2,000 ppm for 2 years (Mennear et al., 1988; NTP, 1986).
Two-year inhalation exposure studies at concentrations of 2,000 or 4,000 ppm
dichloromethane produced increased incidences of lung and liver tumors in B6C3Fi mice
(Mennear et al., 1988; NTP, 1986). Additional studies examining mechanistic issues regarding
this effect are described in Sections 4.5.2 and 4.5.3 (Maronpot et al., 1995; Foley et al., 1993;
119 DRAFT - DO NOT CITE OR QUOTE
-------
Kari et al., 1993). Significantly increased incidences of benign mammary tumors (primarily
fibroadenomas) were also observed in male and female F344/N rats exposed by inhalation to
2,000 or 4,000 ppm for 2 years (Mennear et al., 1988; NTP, 1986). In the male rats, the
incidence of fibromas or sarcomas originating from the subcutaneous tissue around the
mammary gland was also increased, but the difference was not statistically significant. In other
studies in Sprague-Dawley rats with exposures of 50-500 ppm (Nitschke et al., 1988a) and 500-
3,500 ppm (Burek et al., 1984), the incidence of benign mammary tumors was not increased, but
in females, the number of tumors per tumor-bearing rat increased at the higher dose levels.
No obvious clinical signs of neurological impairment were observed in the 2-year
bioassays involving exposure concentrations up to 2,000 ppm in F344 rats (Mennear et al., 1988;
NTP, 1986) or 3,500 ppm in Sprague-Dawley rats (Nitschke et al., 1988a; Burek et al., 1984). In
B6C3Fi mice exposed to 4,000 ppm there was some evidence of hyperactivity during the first
year of the study and lethargy during the second year, with female mice appearing to be more
sensitive (Mennear et al., 1988; NTP, 1986). Studies that evaluated batteries of neurobehavioral
endpoints following subchronic or chronic inhalation exposure are limited to one in F344 rats
exposed to concentrations up to 2,000 ppm for 13 weeks (Mattsson et al., 1990). No effects were
observed >64 hours postexposure in an observational battery, a test of hind-limb grip strength, a
battery of evoked potentials, or histologic examinations of brain, spinal cord, or peripheral
nerves (see Table 4-26 and Section 4.4.3).
No effects on reproductive performance were found in a two-generation reproductive
toxicity study with F344 rats exposed to concentrations up to 1,500 ppm for 14 and 17 weeks
before mating of the FO and Fl generations, respectively (Nitschke et al., 1988b) (described
more completely in Section 4.3). Developmental effects following exposure of Long-Evans rats
to 4,500 ppm for 14 days prior to mating and during gestation (or during gestation alone)
included decreased offspring weight at birth and changed behavioral habituation of the offspring
to novel environments (Bornschein et al., 1980; Hardin and Manson, 1980) (see Section 4.3 for
more details). In standard developmental toxicity studies involving exposure to 1,250 ppm on
GDs 6-15, no adverse effects on fetal development were found in Swiss-Webster mice or
Sprague-Dawley rats (Schwetz et al., 1975) (see Section 4.3).
4.2.2.1. Toxicity Studies of Subchronic Inhalation Exposures: General, Renal, and Hepatic
Effects
Data pertaining to general (e.g., BW, mortality), hepatic, and renal effects from several
inhalation exposure studies in various species with exposure periods of 3-6 months are described
below. (Studies providing detailed neurological data are described separately in Section 4.4.3)
The earliest study involved several different species with exposures of 5,000 ppm for up to
6 months (Heppel et al., 1944). Two 14-week studies in dogs, monkeys, rats, and mice were
conducted with exposures at 0, 1,000, and 5,000 ppm (Haun et al., 1972, 1971; Weinstein et al.,
120 DRAFT - DO NOT CITE OR QUOTE
-------
1972) and at 0, 25, and 100 ppm (Haun et al., 1972). Neurological effects and hepatic
degeneration were seen at the 1,000 ppm dose. In the lower-dose portion of the Haun et al.
(1972) study in mice, decreased CYP levels in liver microsomes and some histopathologic liver
changes (fat stains and cytoplasmic vacuolation) were seen at 100 ppm, but more obvious
adverse effects were not observed. Leuschner et al. (1984) reported data from a high exposure
(10,000 ppm) 90-day study of rats; beagle dogs were also included in this study at an exposure
level of 5,000 ppm. No evidence of toxicity was reported by the authors of this study. In a
13-week exposure study conducted by NTP (1986), decreased BWs and increased incidence of
foreign body pneumonia were seen at 8,400 ppm in F344 rats, and histologic changes in the liver
in B6C3Fi mice were seen at 4,200 ppm.
The first experimental study of dichloromethane exposure included dogs, rabbits, guinea
pigs, and rats with an exposure of approximately 5,000 ppm for 7 hours/day, 5 days/week for up
to 6 months (Heppel et al., 1944). The strains of the animals, the comparability between exposed
and unexposed groups (in terms of sex distribution and other attributes), and the process by
which animals were chosen for histologic examination are not clearly described in the report.
Exposed animals included adult dogs (1 male and 5 females), juvenile dogs (1 male and 1 female
born in the exposure chamber and exposed daily from birth), adult rabbits (2 males and
2 females), guinea pigs (14 males), and rats (15 males and 6 females). The nonexposed control
group included 14 guinea pigs, 28 rats, 4 rabbits, and an unspecified number of dogs. Exposure
produced no significant effects on BWs except in the guinea pigs; after 131 exposures, average
BWs were 0.820 and 1.025 kg for exposed and control guinea pigs, respectively. Three exposed
guinea pigs died after 35, 90, and 96 exposures. No other deaths occurred except for one
exposed female rat that died after 22 exposures and giving birth to a litter. Autopsy showed
thrombi in the renal vessels associated with marked cortical infarction. No adverse clinical signs
of toxicity (such as decreased activity or incoordination) were observed in exposed animals
during the study. Urinalysis, hematology tests, and tests of liver function performed on dogs
during the study showed no treatment-related effects. At termination, gross and microscopic
examination of the major organs showed no pathological changes after exposure to 5,000 ppm
dichloromethane, with the exception that two of the exposed guinea pigs that died showed
extensive pneumonia associated with moderate centrilobular fatty degeneration of the liver. The
results indicate that 5,000 ppm was a NOAEL for systemic effects in dogs, rabbits, and rats
exposed 7 hours/day, 5 days/week for up to 6 months. The findings of three deaths (two with
pulmonary congestion and centrilobular fatty degeneration) and 20% decreased average BW
among the 14 exposed guinea pigs indicates that 5,000 ppm was a LOAEL in this species.
Haun et al. (1972, 1971) and Weinstein et al. (1972) reported results from studies in
which groups of 8 female beagle dogs, 4 female rhesus monkeys, 20 male Sprague-Dawley rats,
and 380 female ICR mice were continuously exposed to 0, 1,000, or 5,000 ppm dichloromethane
for up to 14 weeks in whole-body exposure chambers. Gross and histopathologic examinations
121 DRAFT - DO NOT CITE OR QUOTE
-------
were scheduled to be made on animals that died or were sacrificed during or at termination of the
study. At 5,000 ppm, obvious nervous system effects (e.g., incoordination, lethargy) were
observed in dogs, monkeys, and mice. At 1,000 ppm, these effects were most apparent in dogs
and monkeys (Haun et al., 1971). Food consumption was reduced in all species at 5,000 ppm
and in dogs and monkeys at 1,000 ppm. All exposed animals either lost weight or showed
markedly decreased BW gains compared with controls. For example, rats exposed to 1,000 or
5,000 ppm for 14 weeks showed average BWs that were roughly 10 and 20% lower than control
values. Significant numbers of dogs (4) and mice (123), as well as 1 monkey, died within the
first 3 weeks of exposure to 5,000 ppm. Because of this high mortality, all surviving 5,000 ppm
animals were sacrificed at 4 weeks of exposure, except for one half (10) of the rats that went on
to survive the 14-week exposure period. At 1,000 ppm, 6/8 dogs died by 7 weeks, at which time
the remaining two were sacrificed. Monkeys, rats, and all but a few mice survived exposure to
1,000 ppm for 14 weeks.
Gross examination of tissues showed yellow, fatty livers in dogs that died during
exposure to 1,000 or 5,000 ppm, "borderline" liver changes in 3 monkeys exposed to 5,000 ppm,
and mottled liver changes in 4/10 rats exposed to 5,000 ppm for 14 weeks (Haun et al., 1971).
Comprehensive reporting of the histologic findings from this study were not available, but Haun
et al. (1972) reported that the primary target organ was the liver and that in some exposed
animals, the kidney was also affected. Light and electron microscopy of liver sections from
groups of 4-10 mice sacrificed after 1, 4, 8, and 12 hours and 1, 2, 3, 4, 6, and 7 days of
exposure to 5,000 ppm showed hepatocytes with balloon degeneration (dissociation of
polyribosomes and swelling of rough endoplasmic reticulum) as early as 12 hours of exposure
(Weinstein et al., 1972). The degeneration peaked in severity after 2 days of exposure and,
subsequently, partially reversed in severity. Information on possible histopathologic changes in
mice exposed to 1,000 ppm was not provided.
The results from this study demonstrate that dogs and mice were more sensitive than
were rats and monkeys to lethal effects, nervous system depression, and possibly liver effects
from continuous exposure to 1,000 or 5,000 ppm. The results indicate that continuous exposure
to 1,000 ppm was an adverse effect level for mortality and effects on the nervous system and
liver in dogs (exposed for up to 4 weeks) and for BW changes in rats (exposed for 14 weeks).
The 5,000 ppm level induced mortality in beagle dogs, ICR mice, and rhesus monkeys (but not
in Sprague-Dawley rats); obvious nervous system effects in dogs, mice, monkeys, and rats; and
gross liver changes in dogs, mice, monkeys, and rats.
Haun et al. (1972) also conducted studies with groups of 20 mice, 20 rats, 16 dogs, and
4 monkeys exposed continuously to 0, 25, or 100 ppm dichloromethane for 100 days (14 weeks).
The animals presumably were of the same strains and sexes as those used in the studies involving
exposure to 1,000 or 5,000 ppm dichloromethane (Haun et al., 1972, 1971; Weinstein et al.,
1972). All animals underwent necropsy and histopathologic evaluation at termination of the
122 DRAFT - DO NOT CITE OR QUOTE
-------
exposure, but a list of the tissues examined and incidence or severity data were not presented in
the report. Hematology and clinical chemistry variables (including COHb levels) were measured
in blood samples collected from dogs and monkeys at biweekly or monthly intervals during
exposure. COHb levels were elevated in a dose-related manner in monkeys and peaked at about
5% (approximately 0.8% preexposure) after 6 weeks of exposure. COHb levels in dogs were
unaffected by the 25 ppm exposure level and rose to about 2% (from about 0.6%) from week 4 in
high-dose dogs. Additional groups of mice were included for assessment of hexobarbital sleep
times at monthly intervals; levels of cytochromes P-450, P-420, and bs in liver microsomes at
monthly intervals; and spontaneous physical activity at several intervals during the study.
No clinical signs of toxicity or alterations in weight gain were seen in any of the species
examined. In dogs and monkeys, hematology and clinical chemistry results throughout the study
and at termination were unremarkable, as were the results of the gross and histopathologic
examinations. In mice exposed to 100 ppm, CYP levels in liver microsomes were significantly
decreased (compared with control values) after 30, 60, and 90 days of exposure to 100 ppm,
whereas levels of cytochrome bs and P-420 decreased after 30 days and increased after 90 days
of exposure. At 25 ppm, no significant differences from controls were seen in mouse liver levels
of cytochromes. Mice exposed to 25 ppm showed no histopathologic changes, while histologic
changes in mice at 100 ppm were restricted to positive fat stains and some cytoplasmic
vacuolation in the liver. In rats at both exposure levels, the livers showed positive staining for
increased fat, and the kidneys showed evidence of nonspecific tubular degenerative and
regenerative changes. Haun et al. (1972) indicate that no distinctively adverse effects were
found in monkeys, dogs, rats, or mice continuously exposed to 25 or 100 ppm for up to
14 weeks. Decreased CYP levels in liver microsomes and some histopathologic liver changes
(fat stains and cytoplasmic vacuolation) were seen at the 100 ppm dose.
Leuschner et al. (1984) exposed Sprague-Dawley rats (20/sex/dose level) to 0 or
10,000 ppm and beagle dogs (3/sex/dose level) to 0 or 5,000 ppm dichloromethane in whole-
body exposure chambers. Exposure periods were 6 hours/day for 90 consecutive days.
Endpoints evaluated in both species included clinical signs, food and water consumption, BW,
hematology, clinical chemistry, urinalysis, and gross and microscopic evaluation of 27 organs at
termination. Electrocardiography and blood pressure measurements were also done in dogs.
The only significant effect observed in rats was a slight redness of the conjunctiva
1-10 hours after each exposure. In dogs, compound-related effects were restricted to slight
sedation throughout the exposure period and slight erythema lasting up to 10 hours after
exposure. In this 90-day study involving daily 6-hour exposures, 10,000 and 5,000 ppm were
NOAELs for behavioral, clinical chemistry, hematologic, and histologic signs of toxicity in
Sprague-Dawley rats and beagle dogs, respectively.
NTP (1986) exposed groups of F344 rats and B6C3Fi mice (10/sex/dose level) to target
concentrations of 0, 525, 1,050, 2,100, 4,200, or 8,400 ppm dichloromethane 6 hours/day,
123 DRAFT - DO NOT CITE OR QUOTE
-------
5 days/week for 13 weeks in whole-body exposure chambers. Endpoints monitored included
clinical signs, BW, and necropsy at termination. Comprehensive sets of tissues and organs in
control and high-dose animals were histologically examined; tissues from the lower dose groups
were examined to determine the no-observed-effect level. One male and one female rat from the
8,400 ppm exposure group died before the end of the study, but the cause of death was not
discussed. The final mean BWs of 8,400 ppm male and female rats were reduced by 23 and
11%, respectively, relative to controls. Foreign-body pneumonia was present in 4/10 male and
6/10 female rats exposed to 8,400 ppm and in 1/10 female rats from the 4,200 ppm exposure
group. The liver lipid/liver weight ratios for 8,400 ppm rats of both sexes and 4,200 ppm female
rats were significantly lower than in controls. In mice, 4/10 males and 2/10 females exposed to
8,400 ppm died before the end of the study, and these deaths were considered treatment-related.
Histologic changes in exposed mice consisted of hepatic centrilobular hydropic degeneration (of
minimal to mild severity) in 3/10 males and 8/10 females at 8,400 ppm and in 9/10 females from
the 4,200 ppm exposure group. Histologic changes in the 2,100 ppm mouse group were not
mentioned. The liver lipid/liver weight ratio for the high-dose female mice was significantly
lower than in controls. In this 13-week study involving 6-hour exposure periods for
5 days/week, 4,200 ppm was a NOAEL and 8,400 ppm was a LOAEL for decreased BWs and
increased incidence of foreign-body pneumonia in F344 rats. In B6C3Fi mice, 2,100 ppm was a
NOAEL and 4,200 ppm was a LOAEL for histologic changes in the liver.
4.2.2.2. Toxicity Studies from Chronic Inhalation Exposures
Chronic inhalation exposure studies are summarized in Table 4-16. Details of each study
are described below, with the results pertaining to nonneoplastic and neoplastic effects
summarized in the following sections.
124 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-16. Studies of chronic inhalation dichloromethane exposures
Reference,
strain/species
Mennear et al. (1988);
NTP (1986)
F344 rats
Mennear et al. (1988);
NTP (1986)
B6C3FJ mice
Bureketal. (1984)
Syrian hamsters
Bureketal. (1984)
Sprague-Dawley rats
Nitschke et al. (1988a)
Sprague-Dawley rats
Maltoni et al. (1988)
Sprague-Dawley rats,
female
Number per
group
50/sex/dose
50/sex/dose
95/sex/dose
92-97/sex/dose
90/dose/sex
54-60/dose
Exposure
information
2 yrs, 6 hrs/d, 5 d/wk
0, 1,000, 2,000,
4,000 ppm
2 yrs, 6 hrs/d, 5 d/wk
0, 2,000, 4,000 ppm
2 yrs, 6 hrs/d, 5 d/wk
0, 500, 1,500, 3,500
ppm
2 yrs, 6 hrs/d, 5 d/wk
0, 500, 1,500, 3,500
ppm
2 yrs, 6 hrs/d, 5 d/wk
0, 50, 200, 500 ppm
2 yrs, 4 hrs/d, 5 d/wk
for 7 wks; 7 hrs/d, 5
d/wk for 97 wks
0, 100 ppm
Comments
Nonneoplastic liver effects and hemosiderosis in
males and females (see Table 4-17)
Weak trend for neoplastic nodule or hepatocellular
carcinoma in females, benign mammary tumors in
males and females (see Table 4-18)
Varied nonneoplastic effects (see Table 4-19)
Liver and lung tumors (adenomas or carcinomas) in
males and females (see Table 4-20)
Decreased mortality
Increased COHb at 500 ppm (see Section 4.2.2.2.3)
Nonneoplastic liver effects in males and females
(see Table 4-21)
Increased COHb at 500 ppm
Increased number of benign mammary tumors per
tumor bearing rat (females) (see Table 4-21)
Nonneoplastic liver effects in males and females
(statistically significant in females) (see
Table 4-22)
Increased COHb at 50 ppm
Increased number of benign mammary tumors per
animal in females (see Table 4-23)
No effects seen on total number of benign or
malignant cancers
4.2.2.2.1. Chronic inhalation exposure in F344/N rats (Mennear et al, 1988; NTP, 1986).
NTP conducted a 2-year inhalation exposure study in F344/N rats (Mennear et al., 1988; NTP,
1986). The rats (50/sex/exposure level) were exposed to dichloromethane (>99% pure) by
inhalation in exposure chambers 6 hours/day, 5 days/week for 2 years. Exposure concentrations
were 0, 1,000, 2,000, or 4,000 ppm. Mean daily concentrations never exceeded 110% of target
and were <90% of target in only 23 of 1,476 analyses. Endpoints monitored included clinical
signs, mortality, and gross and microscopic examinations of 32 tissues at study termination.
Clinical examinations were conducted weekly for 3.5 months and biweekly until month 8. After
8 months, the animals were clinically examined and palpated for tumors and masses monthly
until the end of the study.
Dichloromethane exposure did not significantly alter BW gain or terminal BWs
(Mennear et al., 1988; NTP, 1986). Survival of male rats was low in all exposed groups and the
control group, and no significant exposure-related differences were apparent. Most deaths
occurred during the last 16 weeks of the study. Survival at week 86 was 36/50, 39/50, 37/50, and
33/50 for the control, 1,000, 2,000, and 4,000 ppm groups, respectively. In female rats, there
was a trend towards decreased survival, and the survival of high-dose female rats was
125
DRAFT - DO NOT CITE OR QUOTE
-------
significantly reduced, possibly due to leukemia. Survival in the females at 86 weeks was 30/50,
22/50, 22/50, and 15/50 for the control, 1,000, 2,000, and 4,000 ppm groups, respectively.
Nonneoplastic lesions with statistically significantly elevated incidences compared with controls
included hepatocyte cytoplasmic vacuolation and necrosis and liver hemosiderosis in males and
females, renal tubular cell degeneration in males and females, splenic fibrosis in males, and nasal
cavity squamous metaplasia in females (Table 4-17). The results indicate that 1,000 ppm
(6 hours/day, 5 days/week) was a LOAEL for liver changes (hepatocyte cytoplasmic vacuolation
and necrosis, hepatic hemosiderosis) in male and female F344/N rats. A NOAEL was not
established because effects were observed at the lowest dose.
126 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-17. Incidences of nonneoplastic histologic changes in male and
female F344/N rats exposed to dichloromethane by inhalation (6 hours/day,
5 days/week) for 2 years
Lesion, by sex
Males
n per groupb
Number (%)c with:
Liver changes
Hepatocyte cytoplasmic vacuolation
Hepatocyte focal necrosis
Hepatocytomegaly
Hemosiderosis
Bile duct fibrosis
Renal tubular cell degeneration
Splenic fibrosis
Females
n per group6
Number (%)c with:
Liver changes
Hepatocyte cytoplasmic vacuolation
Hepatocyte focal necrosis
Hepatocytomegaly
Hemosiderosis
Bile duct fibrosis
Renal tubular cell degeneration
Splenic fibrosis
Nasal cavity squamous metaplasia
Exposure (ppm)a
Controls
0
50
8(16)
7(14)
2(4)
8(16)
8(16)
11(22)
2(4)
50
10 (20)
2(4)
3(6)
19 (38)
4(8)
14 (28)
0(0)
1(2)
1,000
50
26 (53)d
23 (47)d
10 (20)
29 (59)d
10 (20)
13 (26)
6(12)
50
43 (86)d
32 (64)d
10 (20)d
29 (58)d
3(6)
20 (40)
2(4)
2(4)
2,000
50
22 (44)d
6(12)
6(12)
37 (74)d
17 (34)
23 (46)d
ll(22)d
50
44 (88)d
19 (38)d
18 (36)d
38 (76)d
10 (20)d
22 (44)
4(8)
3(6)
4,000
50
25 (50)d
16 (32)d
5(10)
42 (84)d
23 (46)d
10 (20)d
8 (16)d
50
43 (86)d
9 (18)d
5(10)
45 (90)d
3(6)
25 (51)d
4(8)
9 (18)d
al,000 ppm = 3,474 mg/m3, 2,000 ppm = 6,947 mg/m3, 4,000 ppm = 13,894 mg/m3.
bNumber of male rats necropsied per group; only 49 1,000 ppm livers were examined microscopically.
Percentages were based on the number of tissues examined microscopically per group.
dStatistical significance not reported in publications but significantly (p < 0.05) different from controls as calculated
by Fisher's exact test.
eNumber of females necropsied per group; only 49 4,000 ppm kidneys and spleens were examined microscopically.
Sources: Mennear et al. (1988); NTP (1986); Appendix B, Tables Cl and C2 of the NTP (1986) report.
Incidences of mammary fibroadenomas were significantly increased in 4,000 ppm males
and 2,000 and 4,000 ppm females compared with controls (Table 4-18). Similar patterns were
seen with the combination of fibroadenomas and adenomas (not shown in Table 4-18). In males,
subcutaneous tissue fibroma or sarcoma was seen in 1/50, 1/50, 2/50, and 5/50 rats in the 0,
1,000, 2,000, and 4,000 ppm groups, respectively, but these lesions were not seen in females.
Incidences of female rats with liver neoplastic nodules or carcinomas (combined) showed a
127
DRAFT - DO NOT CITE OR QUOTE
-------
significant trend test after survival adjustment only, but the incidences at the two highest dose
levels were not significantly increased relative to the control (Table 4-18).
Incidences for mononuclear cell leukemias in male rats were 34/50, 26/50, 32/50, and
35/50 in controls, 1,000-, 2,000-, and 4,000-ppm dose groups, respectively); a stronger response
was seen in females (17/50, 17/50, 23/50, and 23/50 in controls, 1,000-, 2,000-, and 4,000-ppm
rats, respectively); the incidence in 2,000 ppm- and 4,000-ppm female rats were statistically
significant using a survival-adjustment analysis. NTP (1986) considered the relationship
between exposure to dichloromethane and mononuclear cell leukemia to be equivocal, noting the
relatively high incidence seen in all exposure groups in the males. The mean incidence rate in
historical controls from the NTP lab was 27% in males and 17% in females, compared with 68%
and 34% in male and female controls in this study. The pattern seen in females was not
considered to be treatment-related (Mennear et al., 1988), although the rationale for this
interpretation was not discussed in detail. Other neoplasms that had increased incidences
included mesotheliomas (predominantly in the tunica vaginalis) in males (0/50, 2/50, 5/50, and
4/50 in controls, 1,000, 2,000, and 4,000 ppm rats, respectively). This lesion was not considered
to be related to dichloromethane exposure because the concurrent control incidence (0/50) for
this neoplasm was low relative to earlier inhalation studies conducted at this laboratory (4/100,
4%) and in other NTP studies with male F344/N rats (44/1,727) (mean historical percentage
across NTP studies = 3 ± 2%).
NTP (1986) concluded that there was "some evidence of carcinogenicity of
dichloromethane" in male F344/N rats as shown by increased incidence of benign mammary
gland tumors and "clear evidence of carcinogenicity" of dichloromethane in female F344/N rats
as shown by increased incidence of benign mammary gland tumors. The summary of the hepatic
effects in rats in the NTP (1986) report also notes the positive trend in the incidence of
hepatocellular neoplastic nodules or carcinomas in females which "may have been due to
dichloromethane exposure."
128 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-18. Incidences of selected neoplastic lesions in male and female F344/N rats exposed to dichloromethane by
inhalation (6 hours/day, 5 days/week) for 2 years
Neoplastic lesion, by sex
Exposure (ppm)a
0 (Controls)
n
(%)b
(%)°
1,000
n
(%)b
(%)c
2,000
n
(%)b
(%)°
4,000
n
(%)b
(%)c
Trend
/7-valued
Males
n per group
Liver — neoplastic nodule or hepatocellular carcinoma
Liver — hepatocellular carcinoma
Lung — bronchoalveolar adenoma or carcinoma
Mammary gland
Adenoma, adenocarcinoma, or carcinoma
Subcutaneous tissue fibroma or sarcoma
Fibroadenoma
Mammary gland or subcutaneous tissue adenoma,
fibroadenoma, fibroma, or sarcoma
Brain (carcinoma, not otherwise specified, invasive)
50
2
2
1
0
1
0
1
0
(4)
(4)
(0)
(2)
(0)
(2)
(0)
(10)
(10)
(6)
(0)
(6)
50
3
1
1
0
1
0
1
1
(6)
(2)
(2)
(0)
(2)
(0)
(2)
(2)
(13)
(4)
(6)
(0)
(6)
50
4
2
2
0
2
2
4
0
(8)
(4)
(4)
(0)
(4)
(4)
(8)
(0)
(19)
(10)
(9)
(12)
(21)
50
1
1
1
1
5
1
9e
0
(2)
(2)
(2)
(2)
(10)
(2)
(18)
(0)
(6)
(6)
(23)
(8)
(49)
0.55
Not reported
0.008
0.001
0.001
Females
n per group
Liver — neoplastic nodule or hepatocellular carcinoma
Liver — hepatocellular carcinoma
Lung — bronchoalveolar adenoma or carcinoma
Mammary gland
Adenocarcinoma or carcinoma
Adenoma, adenocarcinoma, or carcinoma
Fibroadenoma
Mammary gland adenoma, fibroadenoma, or adenocarcinoma
Brain (carcinoma, not otherwise specified, invasive, and
oligodendroglioma/
50
2
0
1
1
1
5
6
1
(4)
(0)
(2)
(2)
(2)
(10)
(12)
(2)
(7)
(0)
(16)
(18)
50
1
0
1
2
2
lle
13
0
(2)
(0)
(2)
(4)
(4)
(22)
(26)
(0)
(2)
(0)
(41)
(44)
50
4
1
0
2
2
13e
14e
2
(8)
(2)
(0)
(4)
(4)
(26)
(28)
(4)
(14)
(4)
(44)
(45)
50
5
0
0
0
1
22e
23e
0
(10)
(0)
(0)
(0)
(2)
(44)
(46)
(0)
(20)
(0)
(79)
(86)
0.08
Not reported
0.001
0.001
al,000 ppm = 3,474 mg/m3, 2,000 ppm = 6,947 mg/m3, 4,000 ppm = 13,894 mg/m3.
Percentages based on the number of tissues examined microscopically per group; for males, 49 livers and lungs were examined microscopically in the 1,000 ppm
groups and only 49 brains were examined microscopically in the 4,000 ppm group. For comparison, incidence in historical controls reported in NTP (1986) were 1%
for female liver tumors and 16% for female mammary fibroadenomas.
'Mortality-adjusted percentage.
dLife-table trend test, as reported by NTP (1986).
eLife-table test comparison dose group with control O.05, as reported by NTP (1986).
fThe oligodendroglioma occurred in the 2,000 ppm group.
Sources: Mennear et al. (1988); NTP (1986); Appendix A and Appendix E, Tables El and E2 of the NTP (1986) report.).
129
DRAFT - DO NOT CITE OR QUOTE
-------
4.2.2.2.2. Chronic inhalation exposure in B6C3F, mice (Mennear et al, 1988; NTP, 1986). A
2-year inhalation exposure study in B6C3Fi mice, similar to that in F344/N rats, was also
conducted by NTP (Mennear et al., 1988; NTP, 1986). The mice (50/sex/exposure level) were
exposed to dichloromethane (>99% pure) by inhalation at concentrations of 0, 2,000, or
4,000 ppm in exposure chambers 6 hours/day, 5 days/week for 2 years. As with the study in rats,
mean daily concentrations in the mice never exceeded 110% of target and were <90% of target in
only 23 of 1,476 analyses. Endpoints monitored included clinical signs, mortality, and gross and
microscopic examinations of 32 tissues at study termination. Clinical examinations were
conducted weekly for 3.5 months and biweekly until month 8. After 8 months, the animals were
clinically examined and palpated monthly for tumors and masses until the end of the study.
The BW of 4,000 ppm males was comparable to controls until week 90 and 8-11% below
controls thereafter. The BW of 4,000 ppm females was 0-8% lower than that of controls from
week 51 to 95 and 17% lower at study termination. No information was provided regarding food
consumption during the study. Male and female mice from the high-dose groups (4,000 ppm)
were hyperactive during the first year of the study; during the second year, high-dose females
appeared lethargic. Exposure was associated with decreased survivability of both male and
female mice (males: 39/50, 24/50, and 11/50 and females: 25/50, 25/50, and 8/50 in controls,
2,000, and 4,000 ppm at 104 weeks, respectively). In 4,000 ppm mice, statistically significant
incidences of nonneoplastic lesions were found in the liver (cytologic degeneration), testes
(atrophy), ovary and uterus (atrophy), kidneys (tubule casts in males only), stomach (dilatation),
and spleen (splenic follicles in males only) (Table 4-19). In 2,000 ppm mice, the only
nonneoplastic lesions showing statistically significantly elevated incidences were ovarian
atrophy, renal tubule casts, and hepatocyte degeneration in female mice (Table 4-19). The
results indicate that 2,000 ppm, the lowest exposure level, was a LOAEL for nonneoplastic
changes in the ovaries, kidneys, and livers of female B6C3Fi mice. A NOAEL was not
established because effects occurred at the lowest exposure level.
130 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-19. Incidences of nonneoplastic histologic changes in B6C3Fi mice
exposed to dichloromethane by inhalation (6 hours/day, 5 days/week) for
2 years
Lesion, by sex
Males: npergroupb
Number (%)c with:
Liver changes
Hepatocyte cytoplasmic vacuolation
Hepatocyte focal necrosis
Cytologic degeneration
Testicular atrophy
Renal tubule casts
Stomach dilatation
Splenic follicular atrophy
Females: n per group6
Number (%)c with:
Liver changes
Hepatocyte cytoplasmic vacuolation
Hepatocyte focal necrosis
Cytologic degeneration
Ovarian atrophy
Uterus atrophy
Renal tubule casts
Glandular stomach dilatation
Splenic follicular atrophy
Exposure (ppm)a
Controls
0
50
Not reported
0(0)
0(0)
0(0)
6(12)
3(6)
0(0)
50
Not reported
Not reported
0(0)
6(12)
0(0)
8(16)
1(2)
0(0)
2,000
50
Not reported
0(0)
0(0)
4(8)
11(22)
7(15)
3(6)
50
Not reported
Not reported
23 (48)d
28 (60)d
1(2)
23 (48)d
2(4)
0(0)
4,000
50
Not reported
2(4)
22 (45)d
31(62)d
20 (40)d
9 (18)d
7(15)d
50
Not reported
Not reported
21 (44)d
32 (74)d
8 (17)d
23 (49)d
10 (20)d
1(2)
a2,000 ppm = 6,947 mg/m3, 4,000 ppm = 13,894 mg/m3.
bNumber of male mice necropsied per group. The number biopsied in the 0, 2,000, and 4,000 ppm dose groups was
50, 49, and 49 for liver; 50, 49, and 50 for renal tubules; 49, 47, and 49 for stomach; and 49, 49, and 48 for spleen.
Percentages were based on the number of tissues examined microscopically per group.
dStatistical significance not reported in publications but significantly different (p < 0.05) from control as calculated
by EPA using Fisher's exact test.
eNumber of females necropsied per group. The number biopsied in the 0, 2,000, and 4,000 ppm dose groups was
50, 48, and 48 for liver; 50, 47, and 43 for ovaries; 50, 48, and 47 for uterus; 49, 48, and 47 for renal tubule; 49, 47,
and 48 for stomach; and 49, 48, and 47 for spleen.
Sources: Mennear et al. (1988); NTP (1986); Appendix C, Tables Dl and D2 of the NTP (1986) report.
At both exposure levels, statistically significantly elevated incidences were found for
hepatocellular adenomas (males only), hepatocellular carcinomas, hepatocellular adenomas and
carcinomas combined, bronchoalveolar adenomas, bronchoalveolar carcinomas, and
bronchoalveolar adenomas and carcinomas combined (Table 4-20). Statistically significant
positive trend tests were found for each of these tumor types in female mice. The trend tests
were significant for the liver tumors in male mice after life-table adjustment for reduced survival.
The only other statistically significant carcinogenic response was for increased incidence of
131
DRAFT - DO NOT CITE OR QUOTE
-------
hemangiosarcomas or combined hemangiomas and hemangiosarcomas in male mice exposed to
4,000 ppm. NTP (1986) concluded that the elevated incidences of liver and lung tumors
provided clear evidence of carcinogenicity in male and female B6C3Fi mice.
132 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-20. Incidences of neoplastic lesions in male and female B6C3Fi mice
exposed to dichloromethane by inhalation (6 hours/day, 5 days/week) for
2 years
Neoplastic lesion, by sex
Exposure (ppm)a
0 (Controls)
n
(%)b
(%)c
2,000
n
(%)b
(%)c
4,000
n
(%)b
(%)c
Trend
/7-valued
Males
Liver
Hepatocellular adenoma
Hepatocellular
Hepatocellular adenoma or carcinoma
10
13
22
(20)
(26)
(44)
(23)
(30)
(48)
14
15
24
(29)
(30)
(49)
(47)
(44)
(67)
14
26e
33e
(29)
(53)
(67)
(68)
(76)
(93)
0.19
0.004
0.013
Lung
Bronchoalveolar adenoma
Bronchoalveolar carcinoma
Bronchoalveolar adenoma or
carcinoma
Mammary adenocarcinomaf
Hemangioma or hemangiosarcoma,
combined
3
2
5
-
2
(6)
(4)
(10)
(4)
(8)
(5)
(12)
(5)
19e
10e
27e
-
2
(38)
(20)
(54)
(4)
(56)
(34)
(74)
(8)
24e
28e
40e
-
6
(48)
(56)
(80)
(12)
(79)
(93)
(100)
(26)
0.001
0.001
0.001
0.08
Females
Liver
Hepatocellular adenoma
Hepatocellular carcinoma
Hepatocellular adenoma or carcinoma
2
1
o
J
(4)
(1)
(6)
(7)
(4)
(10)
6
11
16e
(13)
(23)
(33)
(21)
(34)
(48)
22e
32e
40e
(46)
(67)
(83)
(83)
(97)
(100)
0.001
O.001
O.001
Lung
Bronchoalveolar adenoma
Bronchoalveolar carcinoma
Bronchoalveolar adenoma or
carcinoma
Mammary adenocarcinoma
Hemangioma or hemangiosarcoma,
combinedf
2
1
o
J
2
—
(4)
(1)
(6)
(4)
(7)
(4)
(11)
(8)
23e
13e
30e
3
—
(48)
(27)
(63)
(6)
(58)
(46)
(83)
(10)
28e
29e
41e
0
—
(58)
(60)
(85)
(0)
(91)
(92)
(100)
(0)
O.001
O.001
O.001
0.21
a2,000 ppm = 6,947 mg/m3, 4,000 ppm = 13,894 mg/m3.
Percentages based on the number of tissues examined microscopically per group; for males, 49 livers were
examined in the 2,000 and 4,000 ppm groups; for females, only 48 livers and lungs and 49 mammary glands were
microscopically examined in the 2,000 and 4,000 ppm groups. For comparison, incidence in historical controls
reported in NTP (1986) were 28% for male liver tumors, 31% for male lung tumors, 5% for female liver tumors,
and 10% for female lung tumors.
'Mortality-adjusted percentage.
dLife-table trend test, as reported by NTP (1986).
eLife-table test comparison dose group with control O.05, as reported by NTP (1986).
fData not reported.
Sources: Mennear et al. (1988); NTP (1986); Appendix E, Tables E3 and E4)of the NTP (1986) report.
133
DRAFT - DO NOT CITE OR QUOTE
-------
4.2.2.2.3. Chronic inhalation exposure in Syrian hamsters (Burek et al, 1984). Burek et al.
(1984) conducted a chronic toxicity and carcinogenicity study in rats and hamsters. In the
hamster study, groups of 95 Syrian golden hamsters of each sex were exposed to 0 (filtered air),
500, 1,500, or 3,500 ppm dichloromethane (>99% pure) under dynamic airflow conditions in
whole-body exposure chambers 6 hours/day, 5 days/week for 2 years. Exposure started when the
animals were approximately 8 weeks of age. Interim sacrifices were conducted at 6, 12, and
18 months. The hamsters were observed daily during exposure days and were palpated monthly
for palpable masses starting the third month of the study. BWs were monitored weekly for the
first 8 weeks of the study and monthly thereafter. Hematologic determinations included packed
cell volume, total erythrocyte counts, total red blood cells, differential leukocyte counts, and
hemoglobin concentration. The mean corpuscular volume, mean corpuscular hemoglobin, and
MCHC were also determined. A reticulocyte count was also performed on all animals at the
18-month kill and on 10 animals/sex/dose at 24 months. Clinical chemistry determinations
included serum AP and ALT activities, blood urea nitrogen levels, and total protein and albumin.
Urinary parameters measured were specific gravity, pH, glucose, ketones, bilirubin, occult blood,
protein, and urobilinogen. Hematology, clinical chemistries, and urinalysis were performed at
interim sacrifices and at termination. COHb was measured after a single 6-hour exposure and
following 22 months of exposure. Gross and microscopic examinations were conducted on all
tissues. In addition, the weights of the brain, heart, liver, kidneys, and testes were recorded.
In the study using Syrian hamsters (Burek et al., 1984), hamsters were exposed to
analytical concentrations of dichloromethane of 510 ± 27, 1,510 ± 62, and 3,472 ± 144 ppm for
the target concentrations of 500, 1,500, and 3,500 ppm, respectively. No exposure-related
clinical signs were observed in the hamsters throughout the study. Significantly decreased
mortality was observed in females exposed to 3,500 ppm from the 13* through the 24* month
and from the 20th to the 24th month in females exposed to 1,500 ppm. Exposure to
dichloromethane had no significant effect on BW or on mean organ weights. Regarding
hematology parameters (actual data were not shown), Burek et al. (1984) stated that a few
statistically significant changes occurred, but no obvious pattern could be discerned and most
values were within the expected range for the animals. There were no exposure-related
alterations in clinical chemistry or urinalysis values. Male and female hamsters in all dose
groups had significantly elevated COHb values after a single 6-hour exposure and after
22 months of exposure, but at both time points there was no dose-response relationship above the
first dose level and no apparent significant differences in the magnitude of the changes between
the two time points. For example, mean values (± SD) for percentage COHb in male hamsters
after 22 months of exposure were 3.3 (± 3.5), 28.4 (± 5.9), 27.8 (± 2.9), and 30.2 (± 4.9), for the
control through 3,500 ppm groups, respectively. Similar values were obtained for females at
22 months and for males and females after the first day of exposure. Pathological evaluation of
hamsters showed a lack of evidence of definite target organ toxicity. Specific observations
134 DRAFT - DO NOT CITE OR QUOTE
-------
mentioned by the authors included a trend of increasing hemosiderin in the liver of male
hamsters at 6 and 12 months; decreased amyloid deposit in organs, such as the liver, kidneys,
adrenal, and thyroid glands in exposed animals; and fewer biliary cysts in the liver. Increased
hepatic hemosiderin at the 12-month sacrifice was observed in 1/5, 1/5, 3/5, and 5/5 male
hamsters in the control through 3,500 ppm groups, respectively. No exposure-related increased
incidences of hepatic hemosiderin or other liver effects were reported for the terminal sacrifice.
The exposure-related decreases in geriatric changes (i.e., amyloid deposits and biliary cysts)
were more prominent in females and were associated with the increased survivability in the
exposed female hamsters compared with controls. The results indicate that 3,500 ppm was a
NOAEL for adverse changes in clinical chemistry and hematological variables, as well as for
histologic changes in tissues in male and female Syrian golden hamsters. A LOAEL was not
established based on the lack of adverse changes in clinical chemistry and hematological
variables as well as the absence of histologic changes in tissues in male and female Syrian
golden hamsters.
Evaluation of the total number of hamsters with a tumor, the number with a benign
tumor, or the number with a malignant tumor revealed no exposure-related differences in male
hamsters. In the high-dose female group, there was a statistically significant increase in the total
number of benign tumors at any tissue site (the report did not specify which sites), but this was
considered to be secondary to the increased survival of this group. Incidences of male or female
hamsters with tumors in specific tissues were not statistically significantly elevated in exposed
groups compared with control incidences. The results indicate that no statistically significant,
exposure-related carcinogenic responses occurred in male or female Syrian golden hamsters
exposed (6 hours/day, 5 days/week) to up to 3,500 ppm dichloromethane for 2 years.
4.2.2.2.4. Chronic inhalation exposure in Sprague-Dawley rats (Burek et al, 1984). In the rat
study, groups of 92-97 Sprague-Dawley rats of each sex were exposed (similar to the hamster
study described in the previous section) to 0, 500, 1,500, or 3,500 ppm dichloromethane
6 hours/day, 5 days/week for 2 years (Burek et al., 1984). Rats were approximately 8 weeks old
when exposure started. Interim sacrifices were conducted at 6, 12, 15, and 18 months.
Endpoints monitored in rats were the same as in hamsters except that total protein and albumin in
blood were not determined in rats. In addition to measurement at scheduled sacrifices, serum
ALT activity was also measured after 30 days of exposure. COHb was measured after 6, 11, 18,
and 21 months of exposure. Bone marrow cells were collected for cytogenetic studies from
5 rats/sex/dose after 6 months of exposure. The scope of the pathological examinations of the
rats was the same as in the hamster study.
No significant exposure-related signs of toxicity were observed in the rats during the
study. A significant increase in mortality was seen in high-dose female rats from the 18* to the
24th month of exposure, and this appeared to be exposure-related. Exposure to dichloromethane
13 5 DRAFT - DO NOT CITE OR QUOTE
-------
had no significant effect on BW gain in either males or females. The only exposure-related
alterations in organ weights was a significant increase in both absolute and relative liver weight
in high-dose males at the 18-month interim kill and a significant increase in relative liver weight
in high-dose females also at 18 months. Statistically significant changes in hematologic
parameters were restricted to increased mean corpuscular volume and mean corpuscular
hemoglobin values at 15 months in males. The clinical chemistry tests revealed no significant
exposure-related effects. Male and female rats in all exposed groups had significantly elevated
COHb values at all time points, but no dose-response relationship was apparent. For example,
mean (± SD) values for percentage COHb after 21 months of exposure were 0.4 (± 0.7),
12.8 (± 2.6), 14.8 (± 4.4), and 12.2 (± 5.7) for the control through 3,500 ppm female rat groups,
respectively. Exposure-related statistically significant increases in incidences of nonneoplastic
lesions were restricted to the liver (Table 4-21). The incidences of males or females with
hepatocellular vacuolation consistent with fatty change increased as the exposure concentration
increased. Hepatocellular necrosis occurred at elevated incidences in male rats exposed to
1,500 or 3,500 ppm compared with controls, but this endpoint was not reported in the female
data. Liver lesions were initially observed after 12 months of treatment. There was some
evidence that exposure at the two highest levels provided some inhibition of the age-related
glomerulonephropathy observed in the control rats at termination. The results indicate that the
lowest exposure level, 500 ppm, was a LOAEL for fatty changes in the liver of male and female
Sprague-Dawley rats and that exposure to>l,500 ppm induced hepatocellular necrosis in males.
136 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-21. Incidences of selected nonneoplastic and neoplastic histologic
changes in male and female Sprague-Dawley rats exposed to
dichloromethane by inhalation (6 hours/day, 5 days/week) for 2 years
Lesion, by sex
Males — n per group
Number (%) with:
Liver changes
Hepatocellular necrosis
Coagulation necrosis
Hepatic vacuolation (fatty change)
Foci of altered hepatocytes
Foci of altered hepatocytes, basophilic
Area of altered hepatocytes
Multinucleated hepatocytes
Glomerulonephropathy
Severe
Any degree
Mammary changes
Rats with benign mammary tumors
Total number of benign mammary tumors
Number of tumors per tumor-bearing ratf
Females — n per group
Number (%) with:
Liver changes
Hepatocellular necrosis
Coagulation necrosis
Hepatic vacuolation (fatty change)
Foci of altered hepatocytes
Foci of altered hepatocytes, basophilic
Area of altered hepatocytes
Multinucleated hepatocytes
Glomerulonephropathy
Severe
Any degree
Mammary changes
Rats with benign mammary tumors
Total number of benign mammary tumors
Number of tumors per tumor-bearing ratf
Exposure (ppm)a
0 (Controls)
92
2b(2)
d
16b (17)
-
-
-
-
70b (76)
92b'e (100)
7b(8)
8
1.1
96
lb(l)
33b (34)
35b (37)
ob /o\
3 (3)
19b (20)
7b(7)
5(5)
62b (65)
79 (82)
165
2.1
500
95
8(8)
-
36 (38)c
-
-
-
-
62 (65)
91 (96)
3(3)
6
2.0
95
0(0)
49 (52)c
36 (38)
0(0)
24 (25)
36 (38)c
3(3)
64 (67)
81 (85)
218
2.7
1,500
95
10 (10)c
-
43 (45)c
-
-
-
-
53 (56)c
93 (98)
7(7)
11
1.6
96
2(2)
56 (58)c
27 (28)
4(4)
28 (29)
34 (35)c
4(4)
59 (62)
80 (83)
245
3.1
3,500
97
ll(ll)c
-
52 (54)c
-
-
-
-
39 (40)c
90 (93)
14 (14)
17
1.2
97
7(7)
63 (65)c
50 (52)c
10 (10)
35 (36)c
29 (30)c
5(5)
48 (50)c
83 (86)
287
3.5
a500 ppm = 1,737 mg/m3, 1,500 ppm = 5,210 mg/m3, 3,500 ppm = 12,158 mg/m3.
bSignificant dose-related trend—Cochran-Armitage trend test;? < 0.05.
Significantly higher than control incidence by Fisher's exact test.
d- = Reported as "no exposure effect" by Burek et al. (1984); data not given.
eBurek et al. (1984) reported that 93/92 male mice had glomerulonephropathy in the kidney in the control group;
the incidence was corrected to 92/92.
Calculated by EPA.
Source: Burek etal. (1984).
137
DRAFT - DO NOT CITE OR QUOTE
-------
In females, an increasing trend was seen in the incidence of foci or areas of altered
hepatocytes. Female rats in all exposed groups showed increased incidence of multinucleated
hepatocytes in the centrilobular region compared with controls, but there was no evidence of
increasing incidence or severity with increasing exposure level (Table 4-21). The foci and areas
were apparent after 12 months, and their number and size increased thereafter, but incidences for
neoplastic nodules in the liver or hepatocellular carcinomas were not increased in any exposure
group. A statistically significant increased incidence of salivary gland sarcomas was reported for
male rats exposed to 3,500 ppm. Burek et al. (1984) considered this finding unusual and
inconsistent with other existing data because the primary target organ for dichloromethane seems
to be the liver. Incidences of rats with benign mammary gland tumors were not statistically
significantly higher in exposed male or female groups compared with controls, and exposed male
and female groups showed no significantly increased incidences for malignant mammary gland
tumors. The average number of benign mammary tumors per tumor-bearing rat increased with
increasing exposure level. In females, the values were 2.1, 2.7, 3.1, and 3.5 in the control
through 3,500 ppm groups, respectively; males showed a similar response with increasing
exposure level, albeit to a lesser extent (Table 4-21). Burek et al. (1984) concluded that the
significance of this benign mammary tumor response (i.e., increase in number of tumors per
tumor-bearing rat) was unknown but speculated that the predisposition of this strain of rats
(historical control incidences of females with benign mammary tumors normally exceeded 80%)
plus the high exposure to dichloromethane may have resulted in the response.
4.2.2.2.5. Chronic inhalation exposure in Sprague-Dawley rats (Nitschke et al., 1988a).
Nitschke et al. (1988a) examined the toxicity and carcinogenicity of lower concentrations of
dichloromethane in Sprague-Dawley rats. Groups of 90 male and 90 female rats were exposed to
0, 50, 200, or 500 ppm dichloromethane (>99.5% pure) 6 hours/day, 5 days/week for 2 years.
Interim sacrifices were conducted at 6, 12, 15, and 18 months (five rats/sex/interval). An
additional group of 30 female rats was exposed to 500 ppm for 12 months and then exposed to
room air for up to an additional 12 months, and another group of 30 female rats was exposed to
room air for the first 12 months, followed by exposure to 500 ppm for the last 12 months of the
study. These latter groups were included to examine temporal relationships between exposure
and potential carcinogenic response. All groups of rats were examined daily for signs of toxicity
and all rats were examined for palpable masses prior to the initial exposure and at monthly
intervals after the first 12 months. BW was checked twice a month for the first 3 months and
monthly thereafter. Blood samples were collected at interim sacrifices and analyzed for total
bilirubin, cholesterol, triglycerides, potassium, estradiol, follicle-stimulating hormone, and
luteinizing hormone levels. In addition, COHb was determined at multiple times in blood
collected from the tail vein. DNA synthesis (incorporation of 3H-thymidine as a measure of
cellular proliferation) was measured in the liver of separate groups of female rats after exposure
13 8 DRAFT - DO NOT CITE OR QUOTE
-------
to the various concentrations for 6 and 12 months (four females/exposure group/interval). All
rats were subjected to a complete necropsy, and sections from most tissues were processed for
microscopic examination.
Exposure to dichloromethane at any of the exposure levels did not significantly alter
mortality rates, BWs, organ weights, clinical chemistry values, or plasma hormone levels
(Nitschke et al., 1988a). Blood COHb was elevated in a dose-related manner but not in an
exposure duration-related fashion, suggesting lack of accumulation with repeated exposures. For
example, mean (± SD) values for percentage COHb were 2.2 (± 1.3), 6.5 (± 1.1), 12.5 (± 0.8),
and 13.7 (± 0.6) for male rats in the control through 500 ppm groups, respectively, at the
terminal sacrifice. These values were similarly affected at the 6-month and 12-month intervals
(e.g., respective values for males were 0.3 [± 0.7], 2.8 [± 0.3], 9.6 [± 1.2], and 12.7 [± 1.6] at the
12-month sacrifice).
The results of the thymidine incorporation experiment revealed no detectable alteration in
the rate of liver DNA synthesis in the exposed groups compared with controls. Statistically
significantly increased incidences of nonneoplastic liver lesions (hepatic vacuolation and
multinucleated hepatocytes) occurred only in females in the 500 ppm group (Table 4-22). Male
rat incidence for hepatocyte vacuolation was elevated at 500 ppm but not to a statistically
significant degree. In the group of female rats exposed for only 12 months to 500 ppm,
significantly increased incidences of nonneoplastic lesions compared with controls were
restricted to liver cytoplasmic vacuolization (16/25 = 64%) and multinucleated hepatocytes
(9/25 = 36%) in rats exposed during the first 12 months of the study; rats exposed only during
the last 12 months of the study showed no elevated incidences of the liver lesions.
139 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-22. Incidences of selected nonneoplastic histologic changes in male
and female Sprague-Dawley rats exposed to dichloromethane by inhalation
(6 hours/day, 5 days/week) for 2 years
Lesion, by sex
Males — n per group
Number (%) with:
Hepatic vacuolation (fatty change)
Multinucleated hepatocytes
Females — n per group
Number (%) with:
Hepatic vacuolation (fatty change)
Multinucleated hepatocytes
Exposure (ppm)a
0 (Controls)
70
22(31)
-
70
41 (59)
8(11)
50
70
e
-
70
42 (60)
6(9)
200
70
-
70
41 (59)
12(17)
500
70
28 (40)
-
70
53 (76)f
27 (39)f
Trend
/7-valueb
0.01
0.0001
Late
500C
NAd
25
15 (60)
3(12)
Early
500C
NA
25
16 (64)f
9 (36)f
a50 ppm = 174 mg/m3, 200 ppm = 695 mg/m3, 500 ppm = 1,737 mg/m3.
bCochran-Armitage trend test.
°Late 500 = no exposure for first 12 mo followed by 500 ppm for last 12 mo; early 500 = 500 ppm for first 12 mo
followed by no exposure for last 12 mo.
dNA = there were no male rats in these exposure groups.
e- = Incidences not reported.
Significantly (p < 0.05) higher than control incidence by Fisher's exact test (Nitschke et al, 1988a).
Source: Nitschke et al. (1988a).
A few fibrosarcomas or undifferentiated sarcomas in the mammary gland were seen in
the exposed rats, but these incidences were not statistically significant (Table 4-23).
Significantly increased incidences of rats with neoplastic lesions were restricted to benign
mammary tumors in female rats exposed for 2 years to 200 ppm compared with controls (61/69 =
88%) (Table 4-23). However, significantly elevated incidences of this tumor type were not
observed in 500 ppm females, and the 200 ppm incidence was within the range of historical
control values for benign mammary tumors in female Sprague-Dawley rats (79-82%) from two
other chronic toxicity/carcinogenicity studies from the same laboratory. A slight but statistically
significant increase in the number of palpable masses in subcutaneous or mammary regions (at
23 months) per tumor-bearing rat was observed only in the 500 ppm female group. The numbers
of benign mammary tumors per tumor-bearing rat were slightly elevated in the exposed groups
compared with control groups, but no statistical analysis of this variable was performed. In
female rats exposed to 500 ppm (during the first or second 12 months of the study), slight but
statistically significant elevations were found in the number of palpable masses in subcutaneous
or mammary regions per tumor-bearing rat; the numbers of benign mammary tumors per tumor-
bearing rat were slightly elevated compared with those of controls, but statistical analysis of this
variable was not performed.
140
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-23. Incidences of selected neoplastic histologic changes in male
and female Sprague-Dawley rats exposed to dichloromethane by inhalation
(6 hours/day, 5 days/week) for 2 years
Lesion, by sex
Males — n per group
Number (%)c with:
Liver tumors
Lung tumors
Mammary gland tumors
Adenocarcinoma or carcinoma
Fibroadenoma
Fibroma
Fibrosarcoma
Undifferentiated sarcoma
Fibroma, fibrosarcoma, or undifferentiated
sarcomad
Brain tumors
Astrocytoma or glial cell
Granular cell
Females — n per group
Number (%)c with:
Liver tumors
Neoplastic nodule(s)
Hepatocellular carcinoma
Lung tumors
Mammary gland tumors
Adenocarcinoma or carcinoma
Adenoma
Fibroadenoma
Fibroma
Fibrosarcoma
Number with palpable masses in subcutaneous
or mammary region
Number of palpable masses in subcutaneous or
mammary region per tumor-bearing rat
Number with benign tumors
Number of benign tumors per tumor-bearing
rat
Exposure (ppm)a
0
(Controls)
70
0(0)
0(0)
0(0)
2(4)
6(11)
0(0)
0(0)
6(11)
0(0)
0(0)
70
4(6)
1(1)
0(0)
6(9)
1(1)
51 (74)
0(0)
1(1)
55 (78)
1.8
52 (75)
2.0
50
70
0(0)
0(0)
0(0)
0(0)
1(6)
1(6)
2(4)
4(6)
1(1)
0(0)
70
4(6)
0(0)
0(0)
5(7)
1(1)
57 (83)
1(1)
0(0)
56(81)
2.1
58 (84)
2.3
200
70
0(0)
0(0)
0(0)
2(3)
6(11)
1(6)
0(0)
7(12)
2(3)
0(0)
70
3(4)
2(3)
0(0)
4(6)
2(3)
60 (87)
0(0)
0(0)
60 (87)
2.0
61f(88)
2.2
500
70
0(0)
0(0)
0(0)
2(3)
10 (16)
0(0)
0(0)
10 (16)
1(1)
1(1)
70
4(6)
1(1)
0(0)
4(6)
1(1)
55 (80)
1(1)
0(0)
59 (86)
2.2e
55 (80)
2.7
Late
500b
0
25
0(0)
0(0)
0(0)
3(12)
2(8)
22 (88)
1(4)
0(0)
22 (88)
2.3e
23 (92)
2.2
Early
500b
0
25
1(4)
0(0)
0(0)
2(8)
0(0)
23 (92)
1(1)
0(0)
23 (92)
2.7e
23 (92)
2.6
(Table 4-23 continues on next page)
141
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-23. Incidences of selected neoplastic histologic changes in male
and female Sprague-Dawley rats exposed to dichloromethane by inhalation
(6 hours/day, 5 days/week) for 2 years
Lesion, by sex
Brain tumors
Astrocytoma or glial cell
Granular cell
Exposure (ppm)a
0
(Controls)
0(0)
1(1)
50
0(0)
0(0)
200
0(0)
0(0)
500
2(3)
1(1)
Late
500b
0(0)
0(0)
Early
500b
0(0)
0(0)
a50 ppm = 174 mg/m3, 200 ppm = 695 mg/m3, 500 ppm = 1,737 mg/m3.
bLate 500 = no exposure for first 12 mo followed by 500 ppm for last 12 mo; early 500 = 500 ppm for first 12 mo
followed by no exposure for last 12 mo. No males were included in these exposure groups.
Percentages were based on the number of tissues examined microscopically per group. In males, 69 lungs were
examined microscopically in the 50 ppm groups, and only 57, 65, 59, and 64 mammary glands were examined in
the control, 50, 200, and 500 ppm groups, respectively. In females, 69 mammary glands were examined
microscopically in the control, 50, 200, and 500 ppm groups.
dEPA summed across these three tumors, assuming no overlap.
Significantly (p < 0.05) higher than control by Haseman's test (Nitschke et al., 1988a).
Significantly (p < 0.05) higher than control incidence by Fisher's exact test (Nitschke et al., 1988a).
Source: Nitschke et al. (1988a).
A statistically significant increased incidence of brain or CNS tumors was not observed,
but six astrocytoma or glioma (mixed glial cell) tumors were seen in the exposed groups (four in
males, two in females); no tumors of this type were seen in either male or female control groups.
The authors concluded that there was no distinct exposure-related malignant carcinogenic
response in male or female Sprague-Dawley rats exposed (6 hours/day, 5 days/week) to up to
500 ppm dichloromethane for 2 years (Nitschke et al., 1988a).
4.2.2.2.6. Chronic inhalation exposure in Sprague-Dawley rats (Maltoni et al., 1988). Maltoni
et al. (1988) conducted an inhalation exposure study in Sprague-Dawley rats. Two groups of
female rats (54-60/dose) were exposed to 0 or 100 ppm dichloromethane for 104 weeks. The
exposure period was 4 hours/day, 4 days/week for 7 weeks and then 7 hours/day, 5 days/week
for 97 weeks. Endpoints monitored included clinical signs, BW, and full necropsy at sacrifice
(when spontaneous death occurred). For each animal sacrificed, histopathologic examinations
were performed on the following organs: brain and cerebellum, zymbal glands, interscapular
brown fat, salivary glands, tongue, thymus and mediastinal lymph nodes, lungs, liver, kidneys,
adrenals, spleen, pancreas, esophagus, stomach, intestine, bladder, uterus, gonads, and any other
organs with gross lesions.
There was no evidence of increased mortality in the exposed group, and there was no
effect on BW (Maltoni et al., 1988). Little information was provided regarding nonneoplastic
effects, precluding identification of NOAELs and LOAELs for nonneoplastic effects in this
142
DRAFT - DO NOT CITE OR QUOTE
-------
study. Dichloromethane exposure was not related to the percentage of rats with benign and
malignant tumors, malignant tumors, or the number of total malignant tumors per 100 animals.
The percentage of rats with benign mammary tumors was 40.0% in controls and 64.8% in the
exposed group, and the percentage of malignant mammary tumors was 3.3 and 5.5% in controls
and exposed rats, respectively. Neither of these differences was statistically significant.
4.3. REPRODUCTIVE/DEVELOPMENTAL STUDIES—ORAL AND INHALATION
Reproductive and development studies of dichloromethane exposure are summarized in
Table 4-24 and described in detail below. No effects on reproductive performance were
observed in a 90-day gavage study in Charles River CD rats with doses up to 225 mg/kg-day
(General Electric Company, 1976) or in a two-generation reproductive toxicity study with
F344 rats exposed to concentrations up to 1,500 ppm for 14 or 17 weeks before mating of the
FO and Fl generations, respectively, as well as during the Fl gestational period (GDs 0-21)
(Nitschke et al., 1988b). Reproductive parameters (e.g., number of litters, implants/litter, live
fetuses/litter, percent dead/litter, percent resorbed/litter, or fertility index4) were also examined in
a study of male Swiss-Webster mice administered dichloromethane (250 or 500 mg/kg) by
subcutaneous injection 3 times/week for 4 weeks, and in a similar study involving inhalation
exposure to 0, 100, 150, or 200 ppm dichloromethane; no statistically significant effects were
seen in either protocol, although some evidence of a decrease in fertility index was seen in the
150 and 200 ppm groups (Raje et al., 1988).
fertility index defined as number of females impregnated divided by total number of females mated times 100.
143 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-24. Summary of studies of reproductive and developmental effects of dichloromethane exposure in animals
Species and n
Exposure dose
Exposure period
Results
Reference
Gavage or subcutaneous
Charles River rats (males
and females),
10/sex/dose group
Swiss-Webster mice
(males), 20/group
F344 rats (females), 17-
21/dose group
0, 25, 75, 225 mg/kg
(gavage)
0, 250, 500 mg/kg
(subcutaneous injection),
3 x per wk
0, 337.5, 450 mg/kg-d
(gavage)
90 d before mating (10 d
between last exposure and
mating period)
4 wks prior to mating (1 wk
between last exposure and
mating period)
CDs 6-19
No effects on fertility index, number of pups per litter, pup
survival, orFl BW, hematology, and clinical chemistry tests
(up to 90 d of age)
No effects on fertility index, number of litters, implants per
litter, live fetuses per litter, resorption rate; no testicular
effects
Decreased maternal weight gain; no effect on resorption rate,
number of live litters, implants, live pups, or pup weight
General Electric
Company (1976)
Raje et al.
(1988)
Narotsky and
Kavlock (1995)
Inhalation
F344 rats (males and
females, two generation),
307 sex/dose group (FO
andFl)
Swiss-Webster mice
(males), 20/group
Long-Evans rats
(female), 16-2 I/dose
group
Long-Evans rats
(female), 16-2 I/dose
group
Swiss-Webster mice
(females), 30-40/group
Sprague-Dawley rats
(females), 20-35/group
0, 100,500, 1,500 ppm,
6hrs/d
0, 100, 150, 200 ppm,
2hrs/d
0, 4,500 ppm
0, 4,500 ppm
0, 1,250 ppm, 7 hrs/d
0, 1,250 ppm, 7 hrs/d
14 wks prior to mating (FO),
CDs 0-21, and 17 wks prior
to mating, beginning PND 4,
(Fl)
6 wks, prior to mating (2 d
between last exposure and
mating period)
12-14 d before mating and/or
CDs 1-17
12-14 d before mating and/or
CDs 1-17
CDs 6-15
CDs 6-15
No effect on fertility index, litter size, neonatal survival,
growth rates, or histopathologic lesions
Fertility index decreased in 150 and 200 ppm group
(statistical significance depends on test used); no effects on
number of litters, implants per litter, live fetuses per litter,
resorption rate; no testicular effects
Gestational exposure resulted in increased absolute and
relative maternal liver weight, decreased fetal BW
Altered rate of behavioral habituation to novel environment
(at 4 d of age). No effect on crawling (at 10 d), movement in
photocell cage (15 d), use of running wheel (45-108 d), and
shock avoidance (4 mo)
Increased incidence of extra center of ossification in sternum,
increased (-10%) maternal blood COHb, increased maternal
weight, increased maternal absolute liver weight
Decreased incidence of lumbar ribs or spurs, increased
incidence of delayed ossification of sternebrae, increased
(-10%) maternal blood COHb, increased maternal absolute
liver weight
Nitschke et al.
(1988b)
Raje et al.
(1988)
Hardin and
Manson(1980)
Bornschein et al.
(1980)
Schwetz et al.
(1975)
Schwetz et al.
(1975)
PND = postnatal day
144
DRAFT - DO NOT CITE OR QUOTE
-------
Following exposure of pregnant F344 rats to gavage doses of up to 450 mg/kg-day on
GDs 6-19, maternal weight gain was decreased, but no effects were found on the number of
resorption sites, pup survivability, or pup weights at postnatal days (PNDs) 1 or 6 (Narotsky and
Kavlock, 1995). The developmental effects following exposure of Long-Evans rats to
4,500 ppm for 14 days prior to mating and during gestation (or during gestation alone) were
decreased offspring weight at birth and changed behavioral habituation of the offspring to novel
environments (Bornschein et al., 1980; Hardin and Manson, 1980) (see Section 4.3.2 for more
details). In standard developmental toxicity studies involving exposure to 1,250 ppm on GDs 6-
15, no adverse effects on fetal development were found in Swiss-Webster mice or Sprague-
Dawley rats, but the incidence of minor skeletal variants (e.g., delayed ossification of sternebrae)
was increased. (Schwetz et al., 1975) (see Section 4.3.2).
4.3.1. Reproductive Toxicity Studies
4.3.1.1. Gavage and Subcutaneous Injection Studies
In a study sponsored by the General Electric Company (1976), Charles River CD rats
(10/sex/dose level) were administered 0, 25, 75, or 225 mg/kg-day dichloromethane by gavage in
water for 90 days. The test material was dichloromethane (of unspecified purity) purchased from
Dow Chemical Company. At approximately 100 days of age, the rats were mated 1 to 1 to
produce the Fl generation. Fl rats (15/sex/dose level) received the same treatment as FO for
90 days, at which time they were sacrificed and necropsied. Comprehensive sets of 24 tissues
from 10 male and 10 female Fl rats from the control and 225 mg/kg-day groups were examined
microscopically after embedding, sectioning, and staining. Fl rats were monitored for clinical
signs, BW effects, and food consumption. Reproductive parameters examined were fertility
index, number of pups per litter, and pup survival. Fl rats also underwent hematology and
clinical chemistry tests and urinalysis at 1, 2, and 3 months of the study and ophthalmoscopic
examination at 3 months. There were no significant compound-related alterations in any of the
endpoints monitored.
Raje et al. (1988) administered dichloromethane (250 or 500 mg/kg) by subcutaneous
injection 3 times/week for 4 weeks to male Swiss-Webster mice (20/group). Mating with
unexposed females started 1 week after the last exposure and continued for 2 weeks. After the
mating period, the males were sacrificed and the testes were examined microscopically. On
GD 17, the females were sacrificed and the uterine horns examined for live, dead, or resorbed
fetuses. The authors reported that exposure to dichloromethane had no statistically significant
effects on number of litters, implants/litter, live fetuses/litter, percent dead/litter, percent
resorbed/litter, or fertility index. Examination of the testes showed no significant alterations
compared with controls.
145 DRAFT - DO NOT CITE OR QUOTE
-------
4.3.1.2. Inhalation Studies
Nitschke et al. (1988b) conducted a two-generation reproductive toxicity study in rats.
Groups of F344 rats (30/sex/dose level) were exposed by inhalation in whole-body chambers to
0, 100, 500, or 1,500 ppm dichloromethane (99.86% pure) 6 hours/day, 5 days/week for
14 weeks and then mated to produce the Fl generation. Exposure of dams continued after
mating on GDs 0-21 but was interrupted until PND 4. After weaning, 30 randomly selected
Fl pups/sex/dose level were exposed as the parental generation for 17 weeks and subsequently
mated to produce the F2 generation. The results showed no statistically significant exposure-
related changes in reproductive performance indices (fertility, litter size), neonatal survival,
growth rates, or histopathologic lesions in Fl (Table 4-25) or F2 weanlings sacrificed at time of
weaning. According to the authors, none of the values in Table 4-25 were significantly different
from control values (a = 0.05).
Table 4-25. Reproductive outcomes in F344 rats exposed to
dichloromethane by inhalation for 14 weeks prior to mating and from GDs
0-21
Fertility index3
Gestation index0
Gestation survival indexd
4-d survival index6
28-d survival indexf
Sex ratio on d 1 (M:F)
Litter size
DO
D28
Pup BWs, g
Dl
D4
D 28, male
D 28, female
Exposure (ppm)a
0
77%
100%
99.6%
91.0%
99.4%
48:52
11±2
7±2
5.2 ±0.4
7.4 ±0.7
44.6 ±5.8
43.2 ±4.3
100
77%
100%
100%
95.2%
99.4%
50:50
10 ±2
7±2
5.3 ±0.5
7.5 ±1.1
45.9 ±5.0
43.8 ±4.5
500
63%
100%
100%
98.5%
100%
50:50
10 ±3
7±2
5. 3 ±0.4
7.7 ±0.7
47.0 ±5.4
44.4 ±5.7
1,500
87%
100%
96.6%
98.6%
99.5%
52:48
11±2
8±2
5.2 ±0.4
7.3 ±0.7
45.0 ±5.9
43.0 ±4.8
a!00 ppm = 347 mg/m3, 500 ppm = 1,737 mg/m3, 1,500 ppm = 5,210 mg/m3.
bNumber of females delivering a litter expressed as a percentage of females placed with a male.
°Number of females delivering a live litter expressed as a percentage of the number of females delivering a litter.
Percentage of newborn pups that were alive at birth.
Percentage of pups surviving to d 4.
Percentage of pups alive on d 4 and surviving to d 28.
Source: Nitschke et al. (1988b).
Raje et al. (1988) exposed groups of male Swiss-Webster mice (20/group) to 0, 100, 150,
or 200 ppm dichloromethane (FIPLC grade, JT Baker Chemical Co.) in inhalation chambers
146
DRAFT - DO NOT CITE OR QUOTE
-------
2 hours/day, 5 days/week for 6 weeks. Mating with unexposed females started 2 days after the
last exposure. As in the subcutaneous injection protocol described in the previous section, after
the 2-week mating period, the males were sacrificed and the females were sacrificed on GD 17.
Exposure of the male mice to dichloromethane had no statistically significant effects on number
of litters, implants/litter, live fetuses/litter, percent dead/litter, or percent resorbed/litter, and no
significant alterations in the testes were noted. The fertility index was 95, 95, 80, and 80% in the
control, 100, 150 and 200 ppm groups, respectively. This decrease was not statistically
significant as reported by the authors. Details of the statistical analyses were not provided. The
overall %2p-va\ue was 0.27. Using a Cochran-Armitage exact trend test on these data, EPA
calculated a one-sided p-va\ue of 0.059. Individual ^-values for the comparison of each group
with the control group were 0.97, 0.17, and 0.17 for the 100, 150, and 200 ppm groups,
respectively. The results for the combined 150 and 200 ppm groups were statistically different
from the combined controls and 100 ppm group (Fisher's exact test, one-sided^-value = 0.048),
suggesting a NOAEL of 100 ppm and LOAEL of 150 ppm.
4.3.2. Developmental Toxicity Studies
4.3.2.1. Gavage Studies and Culture Studies
Narotsky and Kavlock (1995) evaluated developmental effects of dichloromethane
(99.9% pure) in F344 rats (17-2I/dose group) treated with 0, 337.5, or 450 mg/kg-day
dichloromethane by gavage in corn oil on GDs 6-19. Dams were weighed on GDs 6, 8, 10, 13,
16, and 20 and allowed to deliver naturally. They were sacrificed on PND 6 to count uterine
implantation sites. Pups were grossly examined for developmental abnormalities and weighed
on PNDs 1, 3, and 6. Dead pups or pups with no gross abnormalities were sacrificed and
examined for soft tissue abnormalities. Maternal weight gain during pregnancy was significantly
reduced in high-dose dams (by 33%, as estimated from Figure 5 of the paper); this group also
exhibited rales and nasal congestion. Treatment with dichloromethane did not induce resorptions
or alter the number of live litters on PND 1 or 6, the number of implants, the number of live pups
on PND 1 or 6, or pup weight per litter. No gross or soft tissue abnormalities were observed.
Rat embryos in culture medium were exposed to 0, 3.46, 6.54, 9.79, or 11.88 umol/mL
dichloromethane for 40 hours. At the end of the exposure, embryos were observed for
development of yolk sac vasculature, crown-rump length, total embryonic protein content, and
number of somite pairs. A concentration of dichloromethane of 6.54 umol/mL of culture
medium resulted in decreased crown-rump length, decreased somite number, and decreased
amount of protein per embryo, whereas no effects were seen at 3.46 umol/mL (Brown-Woodman
et al., 1998). A time-course experiment conducted with a concentration of dichloromethane of
9.22 umol/mL showed that marked differences in growth and development from controls were
not significant until about 8 hours of culture. Brown-Woodman et al. (1998) noted that the
concentrations that caused embryotoxicity in this study were much higher than those found in
147 DRAFT - DO NOT CITE OR QUOTE
-------
individuals studied under controlled exposure conditions and comparable to those found in
postmortem blood after fatal inhalation.
4.3.2.2. Inhalation Studies
Schwetz et al. (1975) exposed pregnant Swiss-Webster mice (30-40/group) and Sprague-
Dawley rats (20-35/group) by inhalation in whole-body chambers to 0 or 1,250 ppm
dichloromethane (97.86% pure) 7 hours/day on GDs 6-15. Maternal BWs were recorded on
GDs 6, 10, and 16 and on the day of sacrifice (GD 18 for mice, GD 21 for rats). At sacrifice,
uterine horns were excised and examined for fetal position and number of live, dead, or absorbed
fetuses. Fetuses were observed for gross, soft tissue, and skeletal abnormalities. The only
effects seen on developing fetuses were changes in the incidence of minor skeletal variants. In
rats, the incidence of lumbar ribs or spurs was significantly decreased compared with controls,
whereas the incidence of delayed ossification of sternebrae was significantly greater than in
controls. In mice, a significant number of litters contained pups with a single extra center of
ossification in the sternum. Exposure to dichloromethane produced significantly elevated blood
COHb content in dams of both species (approximately 9-10% after 10 exposures versus 1-2% in
controls). BWs in exposed mouse dams were significantly increased (11-15%) compared with
those in controls but were not affected in exposed rat dams. Mean absolute liver weights of
exposed dams of both species were significantly elevated compared with controls, but mean
relative liver weights were not affected. The results indicate that 1,250 ppm was a LOAEL for
minimal maternal effects (increased COHb and increased absolute liver weight) and adverse
effects on the fetuses.
Hardin and Manson (1980) conducted a study in female Long-Evans rats to determine
whether exposure before and during gestation is more detrimental to reproductive outcome than
exposure either before or during gestation alone. Four groups of 16-21 rats were formed in
which the rats were exposed by inhalation in whole-body chambers to 4,500 ppm
dichloromethane (technical grade, >97% pure) 6 hours/day for 12-14 days before breeding
and/or on GDs 1-17 or were exposed to filtered air. Maternal BWs were measured every 4 days.
Dams were euthanized on GD 21, and livers and uteri were removed. Livers were weighed, and
uterine horns were examined for fetal position and number of live, dead, or absorbed fetuses.
Fetuses were observed for gross, soft-tissue, and skeletal abnormalities. Exposure during
gestation (with or without pregestation exposure) significantly increased maternal absolute and
relative liver weights by about 10-12 and 9-12%, respectively, and decreased fetal BW by about
9-10% relative to those exposed to filtered air during gestation. None of the groups showed
significant alterations in the incidence of gross, external, skeletal, or soft-tissue anomalies.
Using the same study design and exposure level, Bornschein et al. (1980) observed
behavioral activities at various ages in the offspring. Assessed activities included head
movement/pivoting when placed in a novel environment (4 days of age), limited crawling
148 DRAFT - DO NOT CITE OR QUOTE
-------
(10 days), movement in a photocell cage (15 days), use of running wheel (45-108 days), and
shock avoidance (4 months). Exposure during gestation (with or without pregestation exposure)
caused altered rates of behavioral habituation to novel environments in the pups tested as early as
10 days of age; these altered rates were still present at 150 days of age. Growth, food and water
consumption, wheel running activity, and avoidance learning were not significantly affected by
exposure to dichloromethane. The results indicate that 4,500 ppm was a LOAEL for maternal
effects (10% increased absolute and relative liver weight) and for effects on the fetuses (10%
decreased fetal BW and altered behavioral habituation to novel environments).
In a study of early-life (including gestational) exposures, Maltoni et al. (1988) exposed
54 pregnant Sprague-Dawley rats to 100 ppm dichloromethane via inhalation 4 hours/day,
5 days/week for 7 weeks, followed by 7 hours/day, 5 days/week for 97 weeks. Exposure
apparently started on GD 12. Groups of 60 male and 69 female newborns continued to be
exposed after birth to 60 ppm dichloromethane 4 hours/day, 5 days/week for 7 weeks, followed
by exposure 7 hours/day, 5 days/week for 97 weeks. Additional groups of 60 male and
70 female newborn were exposed after birth to 60 ppm dichloromethane 4 hours/day,
5 days/week for 7 weeks and then for 7 hours/day, 5 days/week for 8 weeks. BWs were
measured every 2 weeks during exposure and every 8 weeks thereafter. At the end of exposure,
animals were sacrificed and histologic examinations were performed on 20 tissue types.
Early life exposures of Sprague-Dawley rats to dichloromethane (Maltoni et al., 1988)
did not affect mortality or BW in any group. Also, there was no significant effect of exposure to
dichloromethane on the percentage of animals with benign and malignant tumors and malignant
tumors, the number of malignant tumors per 100 animals, or the percentage of animals with
benign mammary tumors, malignant mammary tumors, leukemias, pheochromocytomas, and
pheochromoblastomas. The results provide no evidence that gestational exposure to 100 ppm
dichloromethane during early life stages of development increases the susceptibility of Sprague-
Dawley rats to the potential carcinogen!city of dichloromethane, but further conclusions from
these results are precluded because the study included only one exposure level that was below
the maximum tolerated dose for adult Sprague-Dawley rats. Experiments comparing cancer
responses from early-life exposures with adult exposures are not available for F344 rats or
B6C3Fi mice, the strains of animals in which carcinogenic responses to dichloromethane have
been observed.
In summary, the potential for gestational exposure to CO and to dichloromethane
(through its transfer across the placenta) resulting from maternal dichloromethane exposure via
oral and inhalation routes, raises concerns regarding neurodevelopmental effects. Although few
developmental effects were observed at high exposures of dichloromethane (Bornschein et al.,
1980; Schwetz et al., 1975), there are no studies that have adequately evaluated neurobehavioral
and neurochemical changes resulting from gestational dichloromethane exposure. The available
data identify changes of behavior habituation at 4,500 ppm (Bornschein et al., 1980) and
149 DRAFT - DO NOT CITE OR QUOTE
-------
increases in COHb at 1,250 ppm (Schwetz et al., 1975). The behavioral changes observed at
4,500 ppm indicate developmental neurotoxic effects; this is the only dose group used in the
Bornschein et al. (1980) study. No other neurological endpoints have been evaluated in the
available developmental studies of dichloromethane. The potential for developmental
neurotoxicity occurring at lower exposures to dichloromethane represents a data gap.
4.4. OTHER DURATION- OR ENDPOINT-SPECIFIC STUDIES
4.4.1. Short-term (2-Week) Studies of General and Hepatic Effects in Animals
Two short-term (2-week) studies examined hepatic and renal effects of dichloromethane
exposure in F344 rats (Berman et al., 1995) and CD-I mice (Condie et al., 1983). Berman et al.
(1995) administered dichloromethane by gavage in corn oil for up to 14 days to groups of eight
female F344 rats at dose levels of 0, 34, 101, 337, or 1,012 mg/kg-day. Starting at day 4, deaths
occurred in the 1,012 mg/kg-day exposure group, with seven of eight rats dying before the end of
the 14-day exposure period. In the dose groups that did not experience this high mortality,
incidences of increased necrotic hepatocytes were 0/8, 0/8, 0/8, and 3/8 for the 0, 34, 101, and
337 mg/kg-day groups, respectively. The increase in liver lesions was not accompanied by
increases in serum activities of ALT or AST. Kidneys, spleen, and thymus were also
histopathologically examined in this study, but none showed exposure-related lesions. The
results indicate that 101 mg/kg-day was a NOAEL and 337 mg/kg-day was a LOAEL for
increased incidence of degenerative lesions in female rats exposed for 14 days. In a companion
study with groups of eight female F344 rats that were given single doses of 0, 101, 337, 1,012, or
1,889 mg/kg-day, incidences of rats with increased necrotic hepatocytes were 1/8, 0/8, 8/8, 7/8,
and 8/8, respectively (Berman et al., 1995).
Condie et al. (1983) detected exposure-related liver lesions in a 14-day gavage study in
which dichloromethane in corn oil was administered to male CD-I mice at dose levels of 0, 133,
333, or 665 mg/kg-day. Incidences of mice with minimal or slight cytoplasmic vacuolation were
1/16, 0/5, 3/5, and 4/5 for the control through high-dose groups, respectively. The kidneys were
also examined histopathologically in this study but showed no exposure-related lesions. No
other tissues were prepared for histologic examination. Blood urea nitrogen, serum creatinine,
and serum ALT activities were not significantly altered by exposure. All dose levels
significantly reduced to the same extent the active transport of />-aminohippurate into renal
cortical slices in vitro, a measure of proximal tubule function. The results most clearly identify
133 mg/kg-day as a NOAEL and 333 mg/kg-day as a LOAEL for increased incidence of
hepatocyte vacuolation in male mice.
4.4.2. Immunotoxicity Studies in Animals
Aranyi et al. (1986) studied the effects of acute inhalation exposures to 50 or 100 ppm
dichloromethane on two measures of immune response (susceptibility to respiratory infection
150 DRAFT - DO NOT CITE OR QUOTE
-------
and mortality due to Streptoccocus zooepidemicus exposure and ability of pulmonary
macrophages to clear infection with Klebsiellapneumoniae). Female CD1 mice that were 5-
7 weeks of age at the start of the exposure portion of the experiment were used for both assays.
Up to five replicate groups of about 30 mice were challenged with viable S. zooepidemicus
during simultaneous exposure to dichloromethane or to filtered air. Deaths were recorded over a
14-day observation period. Clearance of 35S-labeled K. pneumoniae by pulmonary macrophages
was determined by measuring the ratio of the viable bacterial counts to the radioactive counts in
each animal's lungs 3 hours after infection; 18 animals were used per dose group. A single
3-hour exposure to 100 ppm dichloromethane significantly increased the susceptibility to
respiratory infection and greater mortality following exposure to S. zooepidemicus (p < 0.01).
Twenty-six deaths occurred in 140 (18.6%) mice challenged during a 3-hour exposure to
100 ppm dichloromethane; in contrast, nine deaths occurred in 140 mice (6.4%) exposed to
filtered air. The 3-hour exposure to 100 ppm dichloromethane was associated with a statistically
significant (p < 0.001) 12% decrease in pulmonary bactericidal activity (91.6 and 79.6% of
bacteria killed in controls and 100 ppm group, respectively). No difference was seen in either
mortality rate or bactericidal activity in experiments using a single 3-hour exposure to 50 ppm or
3-hour exposures to 40 ppm dichloromethane repeated daily for 5 days compared with control
animals exposed to filtered air. These results suggest that 3-hour exposure to 50 ppm
dichloromethane was a NOAEL and 100 ppm was a LOAEL for decreased immunological
competence (immunosuppression) in CD-I mice.
Aranyi et al. (1986) also conducted a similar set of experiments with 13 other chemicals
(acetaldehyde, acrolein, propylene oxide, chloroform, methyl chloroform, carbon tetrachloride,
allyl chloride, benzene, phenol, monochlorobenzene, benzyl chloride, perchloroethylene, and
ethylene trichloride). Perchloroethylene and ethylene trichloride were the only chemicals in this
group for which an increased mortality risk from Streptococcal pneumonia was seen (mortality
risk 15.0 and 31.4% in controls and 50 ppm exposure groups, respectively, for perchloroethylene
and 13.4 and 58.1% in controls and 50 ppm exposure groups, respectively, for ethylene
trichloride). Decreased bactericidal activity was also seen with acetaldehyde, acrolein, methyl
chloroform, allyl chloride, benzene, benzyl chloride, perchloroethylene, and ethylene trichloride
at one or more exposures. Results from several chemicals suggest that 5 days of exposure results
in greater decrease in bactericidal activity (i.e., acetaldehyde, acrolein, and benzene), and others
(e.g., perchloroethylene) suggest that 5 days of exposure does not result in greater suppression
than a single exposure period.
There was considerable variation in both measures of immune response among the
controls in the experiments (Aranyi et al., 1986). Among the controls in the experiments with
the 13 chemicals other than dichloromethane, mortality in the Streptococcal infectivity model
151 DRAFT - DO NOT CITE OR QUOTE
-------
ranged from 5.7 to 22.1%, with a mean of 12.7%.5 Bactericidal activity in the klebsiella model
among controls ranged from 67.9 to 94.7%, with a mean of 81.8%. The number of bacteria
deposited in the lung in an inhalation bacterial infectivity model can show considerable variation,
(i.e., between 750 to 1,500 viable Streptococcus or Klebsiella organisms, [Ehrlich, 1980]).
Therefore, concurrent controls are particularly important due to the variation in preparation and
aerosol administration of the bacteria in these assays.
Warbrick et al. (2003) evaluated immunocompetence in male and female Sprague-
Dawley rats by measuring the immunoglobulin M (IgM) antibody responses following
immunization with sheep red blood cells in addition to hematological parameters and
histopathology of the spleen, thymus, lungs, and liver. Groups of rats (8/sex/dose level) were
exposed to 0 or 5,000 ppm dichloromethane 6 hours/day, 5 days/week for 28 days. Rats injected
with cyclophosphamide served as positive controls. Five days before sacrifice (day 23 of
exposure) all rats were injected with sheep red blood cells. IgM levels in response to the sheep
red blood cells were comparable between dichloromethane-exposed and air-exposed rats,
indicating that dichloromethane did not produce immunosuppression in the animals under these
exposure conditions. Cyclophosphamide-treated animals had significantly lower levels of IgM
in the blood serum, indicating immunosuppression. Rats exposed to dichloromethane showed
reduced response to sound, piloerection, and hunched posture during exposures. Neither BW
gain nor the hematological parameters monitored were significantly affected by exposure to
dichloromethane. Relative and absolute liver weights were significantly increased in females but
not in males. Relative spleen weight was reduced in females, and no significant changes were
seen in the weight of the thymus and lungs. Histopathology of the tissues examined was
unremarkable. Exposure to 5,000 ppm dichloromethane did not affect antibody production to the
challenge with sheep red blood cells.
In the 2-year drinking water study (Serota et al., 1986a, b) and 2-year inhalation study
(Nitschke et al., 1988a), histopathologic analyses were conducted on the lymph nodes, thymus,
and spleen among several other organs, and no significant changes were noted.
In summary, one study (Aranyi et al., 1986) demonstrated evidence of immuno-
suppression, including increased risk of Streptococcal-pneumonia-related mortality and
decreased clearance of klebsiella bacteria following a single dichloromethane exposure at
100 ppm for 3 hours in CD-I mice. The Streptococcal and Klebsiella bacterial inhalation assays
are models of respiratory infection that test for local immune effects associated with inhalation
exposure rather than systemic immunosuppression. The NOAEL identified in this study was
50 ppm. In contrast, in a functional immune assay of systemic immunosuppression conducted in
rats, Warbrick et al. (2003) did not observe changes in the antibody response to sheep red blood
cells in a 28-day inhalation exposure to 5,000 ppm dichloromethane. Histopathologic analyses
5EPA did not include the duplicate assay of perchloroethylene in calculating this summary statistic. If this
additional assay is included, the mortality risk ranges from 5.7 to 45.7%, with a mean of 15.0%.
152 DRAFT - DO NOT CITE OR QUOTE
-------
of immune system organs in chronic exposure studies for B6C3Fi mice and F344 rats (Nitschke
et al., 1988a; Serota et al., 1986a, b) revealed no changes from controls. However, no assays of
functional immunity were included in these chronic studies. These two studies for
dichloromethane do not suggest systemic immunosuppression, but the Aranyi et al. (1986) study
provides evidence of route-specific local immunosuppression from acute inhalation exposure in
CD1 mice. Due to the acute exposure duration used in Aryani et al. (1986), the immune effects
of short-term or chronic exposure to dichloromethane are unclear.
4.4.3. Neurotoxicology Studies in Animals
Neurological evaluations in animals during and after exposure to dichloromethane have
resulted in CNS depressant effects similar to other chlorinated solvents (e.g., trichloroethylene,
perchloroethylene) and ethanol. Overall, there are decreased motor activity, impaired memory,
and changes in responses to sensory stimuli. Neurobehavioral, neurophysiological, and
neurochemical/neuropathological studies have been used to characterize the effects of
dichloromethane on the CNS. A brief overview of these types of studies is provided below,
followed by a detailed description of individual studies.
Neurobehavioral studies with dichloromethane used protocols to measure changes in
spontaneous motor activity, a functional observational battery (FOB) test (to evaluate gross
neurobehavioral deficits), and a task developed to assess learning and memory. The FOB
protocol includes various autonomic parameters, neuromuscular parameters, sensorimotor
parameters, excitability measures, and activity. Learning and memory changes with
dichloromethane were studied by using a passive avoidance task. The oral and inhalation studies
that examined neurobehavioral endpoints are summarized in Table 4-26.
153 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-26. Studies of neurobehavioral changes from dichloromethane, by route of exposure and type of effect
Species
Exposure(s)
Duration
Neurobehavioral effect
Reference
Gavage exposure
FOB
F344 rat, female
F344 rat, female
101,337, 1,012,
1,889 mg/kg, gavage
34, 101,337, 1,012
mg/kg-d, gavage
Acute — evaluated 4 and 24 hrs
after dosing
14 d — evaluated on d 4, 9, and 15
FOB neuromuscular and sensorimotor
parameters significantly different from
controls at 1,012 and 1,889 mg/kg
(337 mg/kg = NO AEL)
All FOB parameters (except activity)
significantly affected from d 4 at doses of
337 and 1,012 mg/kg-d
Moseretal. (1995)
Moseretal. (1995)
Inhalation exposure
Spontaneous activity
NMRI mouse, male
Rat, male
Wistar rat, male
ICR mouse, female
Sprague-Dawley rat, male
ICR mouse, female
Beagle dog, female
Rhesus monkey, female
ICR mouse, female
400-2,500 ppm
5,000 ppm
500 ppm
5,000 ppm
1,000, 5,000 ppm
1,000, 5,000 ppm
1,000, 5,000 ppm
1,000, 5,000 ppm
25, 100 ppm
Ihr
1 hr, every other d for 10 d
6 hrs/d, 6 d
Continuous, 7 d
Continuous, 14 wks
Continuous, 14 wks
Continuous, 14 wks
Continuous, 14 wks
Continuous, 14 wks
Initial increase in activity followed by a
pronounced decrease at exposures
>600 ppm
Decreased spontaneous locomotor activity
Increased preening frequency
Increased spontaneous activity in first few
hrs and then decreased activity
No neurobehavioral changes
Incoordination, lethargy
Incoordination, lethargy
Incoordination, lethargy
Increased spontaneous activity at 25 ppm
Kjellstrandetal. (1985)
Heppel and Neal ( 1944)
Savolainenetal. (1977)
Weinsteinetal. (1972)
Haunetal. (1971)
Haunetal. (1971)
Haunetal. (1971)
Haunetal. (1971)
Thomas etal. (1972)
FOB
F344 rat, male and female
50, 200, 2,000 ppm
6 hrs/d, 5 d/wk, 13 wks + 65 hrs
exposure free
No effects observed on FOB, grip strength
Mattssonetal. (1990)
Learning and memory
Swiss-Webster mouse,
male
47,000 ppm
Approximately 20 sec + 1 hr
exposure free before training;
retested at d 1, 2, and 4
Significant decrease in learning and recall
ability
Alexeef and Kilgore
(1983)
154
DRAFT - DO NOT CITE OR QUOTE
-------
Neurophysiological studies with dichloromethane exposure consisted of measuring
evoked responses in response to sensory stimuli. In these studies, animals were implanted with
electrodes over the brain region that responds to the particular stimuli. For example, an electrode
would be implanted over the visual cortex in an animal presented with a visual stimulus. Once
the stimulus is presented to the animal, an evoked response is elicited from the brain region and
transmitted to the implanted electrode. During administration of a chemical, if there is a
significant change in the magnitude, shape, and latency (among other measures) in the evoked
response, then the chemical is considered to produce neurological effects. A summary of studies
examining dichloromethane exposure and neurophysiological changes is shown in Table 4-27.
155 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-27. Studies of neurophysiological changes as measured by evoked potentials resulting from
dichloromethane, by route of exposure
Species
Exposure(s)
Duration
SEPs measured
Effect
Reference
Intraperitoneal
Long-Evans rat, male
57.5, 115,230,
460 mg/kg
Acute; tested at 15 min,
1 hr, and 5 hrs after
dosing
FEP
Significant changes in FEPs were noted in
animals dosed >1 15 mg/kg; FEP changes
time and dose dependent
Herr and Boyes
(1997)
Inhalation exposure
F344 rat, male
F344 rat, male and female
5,000, 10,000,
15,000 ppm
50, 200,
2,000 ppm
Acute, 1 hr; tested during
exposure
Subchronic, 6 hr/d,
5 d/wk, 13 wks; tested
65 hrs after last exposure
Electroencephalogram,
BAER, CAEP, FEP, SEP
FEP, CAEP, BAER, SEP
Significant changes in SEP, FEP, BAER,
and CAEP responses at all exposures;
slight recovery noted at 1 hr after exposure
No significant changes noted in any
evoked potential measurements
Rebert et al.
(1989)
Mattsson et al.
(1990)
BAER = brainstem-auditory-evoked response; CAEP = cortical-auditory-evoked potential; FEP = flash-evoked potential;SEP = somatosensory-evoked potential
156
DRAFT - DO NOT CITE OR QUOTE
-------
In neurochemical/neuropathological studies with dichloromethane, animals were first
exposed to dichlorom ethane (orally or via inhalation or injection), and then the brains were
removed. Changes in excitatory neurotransmitters, such as glutamate and acetylcholine and the
inhibitory neurotransmitter, GABA, were measured. Additionally, dopamine and serotonin
levels, which are associated with addiction and mood, were also measured. Other parameters
that were measured included DNA/protein content and regional brain changes in the cerebellum
and hippocampus. Measurement of neurochemical changes provides mechanistic information,
and neurobehavioral and neurophysiological effects can be correlated to these results.
Table 4-28 summarizes studies of neurochemical changes and dichloromethane.
157 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-28. Studies of neurochemical changes from dichloromethane, by route of exposure
Species and sex
Exposure
Duration
Regions
Effect3
Reference
Oral exposure
Sprague-Dawley rat, male
534 mg/kg
Acute, single dose;
evaluated 2 hrs after dosed
Hippocampus,
medulla, midbrain,
hypothalamus
t acetylcholine in hippocampus
t dopamine and serotonin in medulla
I norepinephrine in midbrain
I norepinephrine and serotonin in
hypothalamus
Kanada et al.
(1994)
Inhalation exposure
Wistar rat, male
Wistar rat, male
Wistar rat, male
Wistar rat, male
Sprague-Dawley rat, male
Mongolian gerbil, male
and female
Mongolian gerbil, male
and female
Mongolian gerbil, male
and female
1, 000 ppm TWA (basal
exposure of 100 ppm +
2,800 ppm, 1 hr peak
exposures at hrs 1 and 4)
1,000 ppm TWA
1,000 ppm
1,000 ppm
70, 300, 1,000 ppm
210, 350 ppm
2 10 ppm
2 10 ppm
6 hrs/d, 5 d/wk, 2 wks
6 hrs/d, 5 d/wk, 2 wks + 7
d exposure free
6 hrs/d, 5 d/wk, 2 wks
6 hrs/d, 5 d/wk, 2 wks + 7
d exposure free
6 hrs/d, 3 d
Continuous (24 hrs/d),
3 mo + 4 mo exposure free
Continuous (24 hrs/d),
3 mo
Continuous (24 hrs/d),
3 mo + 4 mo exposure free
Cerebrum,
cerebellum
Cerebrum,
cerebellum
Cerebrum, cerebellum
Cerebrum
Caudate nucleus —
medial
Hippocampus,
cerebellum
cerebral cortex
Frontal cortex,
cerebellum
Hippocampus, olfactory
bulbs, cerebral cortex
t NADPH diaphorase, succinate
dehydrogenase in cerebrum
t cerebral RNA
I succinate dehydrogenase in cerebellum
J, succinate dehydrogenase in both regions
t acid proteinase
J, succinate dehydrogenase in cerebellum
J, cerebral RNA
t catecholamine levels (70 ppm)
I catecholamine levels (300 and
1,000 ppm)
No effect on luteinizing hormone release
I DNA concentration per wet weight in
hippocampus (210, 350 ppm) and
cerebellar hemispheres (350 ppm)
t astroglial proteins in frontal and sensory
motor cerebral cortex
| glutamate, GAB A,
phosphoethanolamine in frontal cortex
t glutamate, GAB A in posterior
cerebellar vermis
J, DNA concentration per wet weight in
hippocampus only
Savolainen et
al. (1981)
Savolainen et
al. (1981)
Savolainen et
al. (1981)
Savolainen et
al. (1981)
Fuxe et al.
(1984)
Rosengren et
al. (1986)
Briving et al.
(1986)
Karlsson et
al. (1987)
aAll effects shown in this table were statistically significant.
t = increase; J, = decrease; NADPH = nicotinamide adenine dinucleotide phosphate
158
DRAFT - DO NOT CITE OR QUOTE
-------
4.4.3.1. Neurotoxicology Studies—Oral Exposures
Three studies evaluated the neurotoxic potential of dichloromethane by either
administering the solvent orally or by injection; two of these studies (Herr and Boyes, 1997;
Kanada et al., 1994) only evaluated acute effects (2-5 hours) from single-dose exposures.
Observed neurological effects included decreased spontaneous activity (Moser et al., 1995),
changes in flash-evoked potential (FEP) measurements (Herr and Boyes, 1997), and changes in
catecholamine levels in the brain (Kanada et al., 1994).
Moser et al. (1995) conducted neurobehavioral evaluations in female F344 rats following
an acute or 14-day oral administration of dichloromethane. A FOB protocol was utilized to
determine changes in autonomic parameters (lacrimation, salivation, pupil response, urination,
defecation), neuromuscular parameters (gait, righting reflex, forelimb and hind-limb grip
strength, landing foot splay), sensorimotor parameters (tail pinch, click response, touch
response), excitability measures (handling reactivity, arousal, clonic, and/or tonic movements),
and activity (rearing, motor activity). A baseline FOB was performed on all rats prior to initial
dichloromethane administration. After dichloromethane administration, a FOB was conducted at
selected time points followed by a motor activity test in a maze. In the acute study, rats were
dosed with 0, 101, 337, 1,012, or 1,889 mg/kg dichloromethane. At 4 and 24 hours after the
administered dose, rats were tested for the neurological parameters. Significant changes in the
neuromuscular and sensorimotor parameters were observed and occurred mostly in rats
administered with the highest dose. These significant changes were only observed at the 4-hour
time point and not when measured at 24 hours. The NOAEL identified by the authors for this
study was 337 mg/kg based on no observable changes in the FOB. In the 14-day study, rats were
administered 0, 34, 101, 337, or 1,012 mg/kg-day. FOB testing was conducted on days 4 and
9 (before the daily dose) and approximately 24 hours after the last (14*) dose. With the
exception of the activity measurements, all other neurobehavioral parameters (neuromuscular,
sensorimotor, autonomic, excitability) were significantly affected from the 4* day through the
entire 14-day exposure cycle. The NOAEL identified for the 14-day study was 101 mg/kg-day
based on FOB changes associated with the dichloromethane exposure.
A single dose acute neurophysiology study by Herr and Boyes (1997) evaluated the effect
of dichloromethane on FEPs in adult male Long-Evans rats. Rats were implanted with epidural
electrodes over the visual cortex area. After placement in an enclosed rectangular mirror
chamber, FEPs were stimulated with a 10 usec flash. Baseline FEPs were collected and rats
were injected intraperitoneally with 0 (corn oil, n = 16), 57.5 (n = 15), 115 (n = 15), 230 (n = 14),
or 460 (n = 15) mg/kg dichloromethane. Animals were retested at 15 minutes, 1 hour, and
5 hours after injection. Amplitude decreases in the early FEP components were observed. The
FEP amplitude changes were time- and dose-dependent with maximal effects at 15 minutes after
dichloromethane dosage. All of the waveform amplitudes returned to control levels when
measured at the 1-hour time point for all doses tested. Response latencies were still different
159 DRAFT - DO NOT CITE OR QUOTE
-------
from controls when measured 5 hours after dosing, but the effect was less pronounced than at the
15-minute and 1-hour time points. In this study, 57.5 mg/kg did not produce any significant
changes in the FEP measures as compared to controls and was considered this study's NOAEL.
The LOAEL was 115 mg/kg based on changes in the FEP amplitudes.
Kanada et al. (1994) examined the effect of dichloromethane on acetylcholine and
catecholamines (dopamine, norepinephrine, serotonin) and their metabolites in the midbrain,
hypothalamus, hippocampus, and medulla from male Sprague-Dawley rats (4-5/group) in a
neurochemical/neuropathology study. The rats were sacrificed 2 hours after a single gavage dose
of 0 or 534 mg/kg of undiluted dichloromethane. Administration of dichloromethane
significantly increased the concentration of acetylcholine in the hippocampus by approximately
10% and increased dopamine and serotonin levels in the medulla by approximately 75%.
Dichloromethane decreased norepinephrine levels in the midbrain and hypothalamus by 12-
15%, and serotonin levels were decreased in the hypothalamus by approximately 30%. There
was a trend toward decreased dopamine in the hypothalamus, but the variability between the
animals was so high that the effect was not significant. (These values for the percent changes
were estimated by EPA from the figures presented in the paper.) The authors speculated that
increased acetylcholine release associated with exposure to dichloromethane and other solvents
may originate from the nerve terminals.
4.4.3.2. Neurotoxicology Studies—Inhalational Exposure
The database pertaining to neurotoxic effects from inhalation exposure to
dichloromethane is considerably larger than the oral exposure database. Acute (<1 day) and
short-term (1-14 days) exposures resulted in an initial increase in spontaneous activity followed
by a decrease for exposures between 500 and 2,500 ppm (Kjellstrand et al., 1985; Savolainen et
al., 1977). Higher (5,000 ppm) acute and short-term exposures resulted in decreased
spontaneous activity and lethargy (Weinstein et al., 1972; Heppel andNeal, 1944). Longer-term
exposures (up to 14 weeks) produced decreased motor activity and lethargy in several animals at
1,000 and 5,000 ppm (Haun et al., 1971), and exposures at 25 ppm for 14 weeks produced
significant increases in activity in mice, starting at week 9. CNS depression was evidenced by
decreased responses in the auditory, visual, and somatosensory regions of the brain in a study of
sensory-evoked potential effects in 12 adult male F344 rats exposed to 0, 5,000, 10,000, and
15,000 ppm for 1 hour periods (Rebert et al., 1989). Altered learning and memory abilities were
demonstrated in young (3-, 5-, and 8-week-old) male Swiss-Webster mice exposed to 168 mg/L
(-47,000 ppm) dichloromethane for approximately 20 seconds (until there was a loss of the
righting reflex) (Alexeef and Kilgore, 1983).
160 DRAFT - DO NOT CITE OR QUOTE
-------
4.4.3.2.1. Inhalational exposure—neurobehavioral studies.
Spontaneous motor activity—acute and short-term studies. Heppel and Neal (1944)
evaluated the neurological effects of 5,000 ppm dichloromethane in five male rats by measuring
changes in spontaneous activity during and after exposure. The five rats were not randomly
selected, since the investigators indicated they picked the most active animals in the litter.
During the 1-hour testing runs, rats were placed in a rotating drum. Spontaneous activity was
reported as the number of drum revolutions/hour. Twenty control test runs (1 run/day) were
conducted prior to dichloromethane exposure runs. After the preexposure period, rats were
exposed to 5,000 ppm dichloromethane every other day for 1 hour, and activity was measured in
the same manner as in the control runs. Once dichloromethane exposure was stopped, the
animals were allowed to recover for 30 minutes and a second 1-hour test run was performed to
evaluate spontaneous activity during recovery. On nonexposure days, spontaneous activity was
also measured in 1-hour intervals to compare to the preexposure period. A total of five
dichloromethane exposures, five postexposure, and five nonexposure trials were conducted over
10 days. Spontaneous activity significantly declined (p < 0.01, Fisher's t-test) during exposure
to 5,000 ppm dichloromethane in comparison to nonexposure days. The average number of
revolutions for all five rats over the test runs was 576 on nonexposure days and 59 revolutions
during dichloromethane exposure.
Weinstein et al. (1972) continuously exposed female ICR mice to 5,000 ppm
dichloromethane for up to 7 days. Clinical behavioral observations of the mice were made
during dichloromethane exposure. Within the first few hours of exposure, spontaneous activity
increased in comparison to control animals. After 24 hours of continuous exposure, there was a
considerable decrease in spontaneous activity as noted by observation only. The mice also
appeared to be very lethargic and had a hunched posture and a rough hair coat, which are all
signs of CNS depressive effects in rodents. These effects became progressively worse until after
96 hours of exposure, where many mice resumed normal activity. After the 7-day exposure,
mice were nearly as active as the control animals but had a rougher coat and were judged to be
emaciated and dehydrated.
Male Wistar rats exposed to 500 ppm dichloromethane 6 hours/day for 6 days exhibited
an increase in preening frequency and time 1 hour after the last exposure relative to controls
(Savolainen et al., 1977). However, there were no significant changes in other types of
spontaneous activity.
In the study by Kjellstrand et al. (1985), male NMRI mice were exposed to
dichloromethane concentrations ranging from 400 to 2,500 ppm. At concentrations of
>600 ppm, exposures for 1 hour produced a biphasic pattern of activity characterized by an
initial increase in activity (as high as 200% of preexposure motor activity at 2,200 ppm, as
estimated from Figure 6 in Kjellstrand et al. [1985]), during exposure followed by a decrease that
reached the lowest point 1-2 hours after the end of exposure (as low as 40% motor activity at
161 DRAFT - DO NOT CITE OR QUOTE
-------
2,200 ppm in comparison to preexposure). Motor activity returned to normal levels after the
decreased activity observed 1-2 hours after exposure was stopped, indicating that the effect was
reversible in this study design.
Spontaneous motor activity—subchronic (14 week) studies. Haun et al. (1971) reported
results from studies in which female beagle dogs, female rhesus monkeys, male Sprague-Dawley
rats, and female ICR mice were continuously exposed to 0, 1,000, or 5,000 ppm dichloromethane
for up to 14 weeks in whole-body exposure chambers. Gross and histopathologic examinations
were made on animals that died or were sacrificed during or at termination of the study. At
5,000 ppm, obvious nervous system effects (e.g., incoordination, lethargy) were most apparent in
dogs and were also observed in monkeys and mice. Rats did not demonstrate any of these
sedative effects. At 1,000 ppm, these effects were observed to a lesser extent in monkeys and
mice, but dogs still displayed prominent CNS depressive behavior. Histopathologic analysis
revealed edema of the brain in three dogs that died during exposure to 5,000 ppm
dichloromethane. No other gross brain-related changes were reported. The results indicate that
continuous exposure to 1,000 ppm was an adverse effect level for mortality and effects on the
nervous system and liver in dogs (exposed for up to 4 weeks) and for BW changes in rats
(exposed for 14 weeks). The 5,000 ppm level induced mortality in beagle dogs, ICR mice, and
rhesus monkeys (but not Sprague-Dawley rats); obvious nervous system effects in dogs, mice,
monkeys, and rats; and gross liver changes in dogs, mice, monkeys, and rats.
In the study by Thomas et al. (1972), female ICR mice were exposed continuously to 0,
25, or 100 ppm dichloromethane for 14 weeks. Spontaneous activity of mice was evaluated by
using closed circuit television for monitoring. Mice were evaluated in daily 2-hour testing
sessions. The 25 and 100 ppm exposure groups were tested for 2 weeks prior to the onset of
dichloromethane exposure. Starting at week 9, mice exposed to 25 ppm dichloromethane
exhibited increases in spontaneous activity, but no quantitative measurements or statistical
analysis were reported. The authors stated that no significant effect was observed in the group
exposed to 100 ppm.
FOB—subchronic (13 week) study. Only one study, a 13-week inhalation study in F344
rats (Mattsson et al., 1990), has conducted an FOB testing paradigm following a subchronic
exposure to dichloromethane. Groups of rats (12/sex/exposure level) were exposed to 0, 50, 200,
or 2,000 ppm dichloromethane 6 hours/day, 5 days/week for 13 weeks. An additional group of
rats was exposed to 135 ppm CO to induce approximately 10% COHb, approximately the level
produced by saturation of oxidative metabolism of dichloromethane. After the 13 weeks of
exposure (beginning 65 hours after the last exposure), rats were subject to an FOB to evaluate
any neurobehavioral changes from the dichloromethane exposure. Autonomic parameters were
first characterized. Then the rat was placed in a clear plastic box to evaluate locomotor activity
162 DRAFT - DO NOT CITE OR QUOTE
-------
and then responsiveness to touch, sharp noise, and tail pinch. Hind-limb grip strength was also
measured by using a strain gauge. All animals were examined clinically at weekly intervals and
were tested at the end of the exposure period by FOB, grip strength, BW, temperature, and
sensory-evoked potentials. No exposure-related effects were observed on the FOB, grip
strength, or sensory-evoked potentials. No histopathologic changes were noted in brains, spinal
cords, or peripheral nerves from the high-dose dichloromethane group (or the CO exposure
group) compared with control animals. In the absence of changes, lower concentrations were not
examined.
Learning and memory—acute study. In a study by Alexeef and Kilgore (1983), a
learning and memory evaluation was conducted following acute exposure to dichloromethane.
Mice were exposed to 168 mg/L (-47,000 ppm) dichloromethane and were tested for learning
ability by using a passive-avoidance conditioning task. Male Swiss-Webster mice (3, 5, and
8 weeks old) were used in this study. In the passive avoidance task, mice were placed on a metal
platform that extended into a hole. If the mouse went into the hole (a darkened area, which
would be the preferred area for the mouse), it received a foot shock. Prior to the training session,
mice were exposed to either air or -47,000 ppm dichloromethane. Animals were exposed to
dichloromethane until there was a loss of the righting reflex, which would take about 20 seconds
on average, and then placed back in their home cage. One hour after exposure, animals were
trained to learn the passive avoidance task. A mouse was considered to have learned the task
once it remained on the platform for at least 30 seconds without entering the hole. Mice were
then tested for recollection of the task at either 1, 2, or 4 days after the initial training session. In
the learning phase of the task, 74% of the control mice retained the task in comparison to 59% of
the dichloromethane-exposed group, indicating the significant effect of dichloromethane on
learning. There was also an age-related effect since exposed 3-week-old mice were less likely to
recall the task than 5- or 8-week-old mice. There was no difference in task recall between the
5- and 8-week-old mice. Dichloromethane at the exposure used in the study was demonstrated to
be nonanalgesic, since pain-response times were comparable to those in air-exposed animals in
the hot-plate pain test, and therefore, the results of the passive avoidance test were not
confounded by potential analgesic effects. As a result, it is demonstrated that exposure to an
acute and high concentration of dichloromethane alters learning ability in mice.
4.4.3.2.2. Inhalational exposure—neurophysiological studies. The effect of dichloromethane
on sensory stimuli was evaluated by measuring sensory-evoked responses during an acute
exposure (Rebert et al., 1989) and following a subchronic (13-week) exposure (Mattsson et al.,
1990). Rebert et al. (1989) evaluated the effects of dichloromethane on sensory-evoked
potentials (auditory, visual, and somatosensory) in F344 rats exposed to 0, 5,000, 10,000, and
15,000 ppm dichloromethane for 1 hour in a head-only exposure chamber. Twelve adult male
163 DRAFT - DO NOT CITE OR QUOTE
-------
rats were implanted with chronic epidural electrodes placed over the visual and somatosensory
cortices. Each rat served as its own control, with a 1-week recovery period between testing
sessions. During each testing session, spontaneous electroencephalograms were recorded.
Additionally, brainstem-auditory-evoked responses (BAERs) (tone stimulus), cortical-auditory-
evoked potentials (CAEPs) (click stimulus), FEPs (flash stimulus), and somatosensory-evoked
potentials (SEPs) (tail current stimulus) were measured in response to the stimuli.
Dichloromethane decreased the SEP response to the tail current stimulus, and earlier components
of the FEP response were attenuated and eventually eliminated with increasing exposures. The
BAER response profile was also significantly altered. Dichloromethane completely abolished
the CAEP at all concentrations tested. Slight recovery of this response was noted approximately
1 hour after exposure. The collective results strongly suggest a CNS depressive profile for
dichloromethane and indicate that this chemical affects the auditory, visual, and somatosensory
regions of the brain.
In a subchronic exposure study, male and female F344 rats were exposed to
dichloromethane 6 hours/day, 5 days/week for 13 weeks (Mattsson et al., 1990). Twelve animals
of each sex were selected for exposure to 0, 50, 200, or 2,000 ppm dichloromethane or 135 ppm
CO. For electrophysiological measures, rats were surgically implanted with epidural electrodes
10 weeks after the onset of exposure. Electrodes were placed over the somatosensory, visual,
and cerebellar region. Electrophysiological measures that were recorded included FEP
measurements, cortical flick fusion responses, CAEPs, BAERs, and SEPS recorded from the
sensory (SEP-S) and cerebellar (SEP-C) regions. None of these measures were significantly
altered by any dichloromethane or CO treatment in this study. However, it should be noted that
all of the electrophysiological measures were conducted at least 65 hours after the last
dichloromethane exposure. As a result, it can be concluded that a subchronic exposure to
dichloromethane did not result in persistent changes in any of the neurophysiological measures
that were evaluated in this study. It is not known if any neurological compensation occurred,
since SEP measurements were not taken during actual dichloromethane exposure in this
subchronic study.
Based on these two studies, the significant changes noted in several SEP measures during
dichloromethane exposure were not observed after a subchronic exposure where animals were
tested at least 65 hours after the last exposure. As a result, it is difficult to ascertain if tolerance
is developed to dichloromethane-mediated changes in sensory potentials seen during an acute
exposure, or if these effects are still maintained during repeated exposure, since measurements
were not taken during the subchronic exposure.
4.4.3.2.3. Inhalational exposure—neurochemistry and neuropathology studies. The studies
evaluating specific neurochemical changes in relation to dichloromethane exposure include
studies of effects of short-term (3-day to 2-week) exposures (Fuxe et al., 1984; Savolainen et al.,
164 DRAFT - DO NOT CITE OR QUOTE
-------
1981) and subchronic (3-month) exposures (Karlsson et al., 1987; Briving et al., 1986;
Rosengren et al., 1986).
Savolainen et al. (1981) examined three different exposure schemes in male Wistar rats.
The rats were exposed to 500, 1,000, or 1,000 ppm TWA dichloromethane 6 hours/day,
5 days/week for 2 weeks. (Note: The abstract of this paper describes the exposures as 500,
1,000, and 100 ppm TWA, but, based on information in the body of the paper, the abstract
appears to be incorrect.) The 1,000 ppm TWA exposure consisted of a basal 100 ppm exposure
with two 2,800 ppm 1-hour peak concentrations (at 1 and 4 hours) resulting in a time-weighted
exposure of 1,000 ppm. Brains were removed from rats at the end of study and analyzed. The
1,000 ppm TWA group displayed increases in cerebral RNA. Other changes noted for this group
in the cerebrum included significant increases in nicotinamide adenine dinucleotide phosphate
(NADPH) diaphorase and succinate dehydrogenase activity. In the 1,000 ppm constant exposure
group, acid proteinase activity was below the levels observed in control animals in the first week
but increased to levels above control animals in the second week. In the cerebellum, there were
no changes in RNA concentration, and there was a decrease in succinate dehydrogenase activity
in both the 1,000 and 1,000 ppm TWA groups. After a 7-day withdrawal, RNA levels in the
cerebrum were significantly lower in the 1,000 ppm group. Succinate dehydrogenase levels
remained lowered in the 1,000 ppm TWA group after the 7-day exposure-free period. No
significant effects were seen at 500 ppm.
Fuxe et al. (1984) evaluated changes in brain catecholamine levels after a 3-day exposure
to dichloromethane using male Sprague-Dawley rats. Rats were exposed to 70, 300, and
1,000 ppm dichloromethane 6 hours/day for 3 consecutive days. Additional groups of rats were
exposed to the same levels of dichloromethane and given intraperitoneal injections of the
tyrosine hydroxylase inhibitor, a-methyl-dl-p-tyrosine methyl ester (H44/68), 2 hours prior to
sacrifice. Brains were removed, stained, and evaluated for catecholamine changes 16-18 hours
after the last exposure. Catecholamine levels were measured in the hypothalamus, frontal cortex,
and caudate nucleus among other brain regions. At all exposures, there was a significant
decrease by approximately 10-15% of catecholamine concentrations in the posterior
periventricular region of the hypothalamus. In the medial part of the caudate nucleus, which is
involved in memory processes, catecholamine levels were significantly higher (12%) in the
70 ppm group but significantly lower in the 300 ppm (1%) and 1,000 ppm (8%) groups
compared with controls. The impact of dichloromethane was also evaluated on the
hypothalamic-pituitary gonadal axis. The hypothalamus regulates secretion of reproductive
hormones, such as follicle-stimulating hormone and luteinizing hormone. The levels of the
hormone release were not significantly changed with dichloromethane exposure. However,
when rats were dosed concurrently with H44/68 and dichloromethane, statistically significant,
inversely dose-related increases in luteinizing hormone levels were observed (330, 233, and
172% higher than controls in the 70, 300, and 1,000 ppm groups, respectively). The study
165 DRAFT - DO NOT CITE OR QUOTE
-------
overall demonstrates significant changes in catecholamine levels in the hypothalamus and
caudate nucleus. No significant changes in catecholamine levels in the frontal cortex were
reported. Catecholamine level changes in the hypothalamus did not appear to significantly affect
hormone release; however, decreased catecholamine levels in the caudate nucleus at higher
exposures may lead to memory and learning impairment.
A series of studies were conducted in male and female Mongolian gerbils exposed
continuously to 210 (Karlsson et al., 1987; Driving et al., 1986), 350, or 700 ppm (Rosengren et
al., 1986) dichloromethane for 3 months, followed by a 4-month exposure-free period. High
mortality rates occurred at 350 ppm (6/10 males and 3/10 females by 71 days) and 700 ppm
(10/10 males and 9/10 females by 52 days). Rosengren et al. (1986) monitored two astroglial
proteins, S-100 and GFA, as well as DNA concentrations in the brain. Decreased DNA
concentrations were noted in the hippocampus at both the 210 and 350 ppm exposures. At
350 ppm, there was also decreased DNA concentration in the cerebellar hemispheres, indicating
a decreased cell density in these regions, probably due to cell loss. Increased astroglial proteins
were found in the frontal and sensory motor cerebral cortex, which directly correlated to the
astrogliosis that was observed in those areas. Up-regulation of these astroglial proteins is a good
indicator of neuronal injury (Rosengren et al., 1986).
Karlsson et al. (1987) measured DNA concentrations in different regions of the gerbil
brain. After the solvent-free exposure period, brains were removed and the olfactory bulbs and
cerebral cortices were dissected. Brain weights and weights of the dissected brain regions were
the same between control and dichloromethane-exposed animals. The total protein concentration
per wet weight was not significantly different between dichloromethane-exposed and control
animals. However, DNA concentrations per wet weight were significantly decreased in the
hippocampus after dichloromethane exposure. No other examined regions demonstrated
significant changes in DNA concentrations after dichloromethane exposure. This selective DNA
concentration decrease observed in the hippocampus is a sign of neurotoxicity and may possibly
explain why some studies have noted memory and learning deficits with dichloromethane
exposure. In a companion paper, in which only the 210 ppm level was tested, it was found that
exposure to dichloromethane decreased the levels of glutamate, y-aminobutyric acid, and
phosphoethanolamine in the frontal cortex, while glutamine and y-aminobutyric acid were
increased in the posterior cerebellar vermis (Driving et al., 1986). Increased levels of glutamate
in the posterior cerebellar vermis could reflect an activation of astrocytic glia, since glutamine
synthetase is localized exclusively in astrocytes. The gerbils did not have a solvent-free
exposure period as in the other two studies (Karlsson et al., 1987; Rosengren et al., 1986). The
exposure regime in these studies did not affect BW or brain weight. Furthermore, the
neurochemical changes observed in these studies were not attributed to formation of CO.
Neurological changes have been investigated by measuring changes in neurotransmitter
levels and changes in neurotransmitter localization. Changes in catecholamine levels in the
166 DRAFT - DO NOT CITE OR QUOTE
-------
caudate nucleus after an acute exposure (Fuxe et al., 1984) as well as decreased DNA content in
the hippocampus after a subchronic dichloromethane exposure (Rosengren et al., 1986) suggest
that memory functions are altered since both brain regions are associated with learning and
memory. The results from Fuxe et al. (1984) directly correlated with the finding that learning
and memory were impaired in mice after an acute (single) and very high exposure (47,000 ppm)
to dichloromethane (Alexeef and Kilgore, 1983). Additionally, changes in the hippocampus also
suggest memory effects after a long-term, continual exposure to dichloromethane, although no
conclusive evidence has been presented to date. In another subchronic, continuous exposure to
350 ppm dichloromethane for 3 months, decreased DNA concentration was observed in the
cerebellar hemispheres of Mongolian gerbils and is suggestive of cell loss (Rosengren et al.,
1986). However, in a 2-week exposure study in male Wistar rats, RNA changes were not noted
in the cerebellum, although enzyme activity was significantly decreased in this region (but was
increased in the cerebrum) (Savolainen et al., 1981). These results suggest that the cerebellum is
a target for dichloromethane. Noted neurobehavioral effects that may be linked to impaired
cerebellar function include changes in motor activity and impaired neuromuscular function
(Moser et al., 1995).
4.5. MECHANISTIC DATA AND OTHER STUDIES IN SUPPORT OF THE MODE OF
ACTION
4.5.1. Genotoxicity Studies
This section discusses the genotoxic potential of dichloromethane. The application of
genotoxicity data to predict potential carcinogenicity is based on the principle that genetic
alterations are found in all cancers. Genotoxicity is the ability of chemicals to alter the genetic
material in a manner that permits changes to be transmitted during cell division. Although most
tests for mutagenicity detect changes in DNA or chromosomes, some specific modifications of
the epigenome, including proteins associated with DNA or RNA, can also cause transmissible
changes. Genetic alterations can occur through a variety of mechanisms, including gene
mutations, insertions, deletions, translocations, or amplification; evidence of mutagenesis
provides mechanistic support for the inference of potential for carcinogenicity in humans.
Evaluation of genotoxicity data entails a weight of evidence approach that includes
consideration of the various types of genetic damage that can occur. In acknowledging that
genotoxicity tests are by design complementary evaluations of different mechanisms of
genotoxicity, a recent IPCS publication (Eastmond et al., 2009) notes that "multiple negative
results may not be sufficient to remove concern for mutagenicity raised by a clear positive result
in a single mutagenicity assay." These considerations inform the present approach. In addition,
consistent with U.S. EPA's Guidelines for Carcinogenic Risk Assessment and Supplemental
Guidance for Assessing Susceptibility from Early-Life Exposure to Carcinogens) (2005a, b), the
approach does not consider quantitative issues related to the probable production of specific
167 DRAFT - DO NOT CITE OR QUOTE
-------
metabolites in vivo. Instead, the analysis of genetic toxicity data presented here focuses on the
identification of a genotoxic hazard of dichloromethane or its metabolites; a quantitative analysis
of dichloromethane metabolism to reactive intermediates, via PBPK modeling, is considered in
the derivation of the cancer toxicity values. The following section summarizes available data on
genotoxicity for dichloromethane by genotoxic endpoint and organism.
4.5.1.1. In Vitro Genotoxicity Assays
Bacterial, yeast, and fungi mutagenicity assays. Numerous in vitro studies have
demonstrated dichloromethane as being mutagenic in bacterial assays, yeast, and fungi, and
several studies provide evidence that the genotoxic action of dichloromethane in bacterial
systems is enhanced in the presence of GSH (e.g., DeMarini et al., 1997; Pegram et al., 1997;
Oda et al., 1996; Thier et al., 1993; Dillon et al., 1992) (Table 4-29). Considering the results are
primarily dependent on the presence of GSH, activation likely involves the GST-T1 metabolic
pathway, which produces two proposed DNA-reactive metabolites, S-(chloromethyl)glutathione
and formaldehyde.
168 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-29. Results from in vitro genotoxicity assays of dichloromethane with bacteria, yeast, or fungi
Assay
Test system
Concentration(s)
Results
Without metabolic
activation
-S9
With metabolic
activation
+S9
Reference
Bacteria
Reverse mutation
Reverse mutation
Reverse mutation
Reverse mutation
Reverse mutation
Reverse mutation
Reverse mutation
Reverse mutation
Reverse mutation
Reverse mutation
Salmonella typhimurium
TA98a, TA1003
S. typhimurium
TA98, TA100
S. typhimurium
TA1535b, TA1537b,
TA1538b
S. typhimurium
TA100
S. typhimurium
TA100
S. typhimurium
TA100, TA1535,
TA19503,
E. coli WU3610893
S. typhimurium
TA100
S. typhimurium
TA100, NG54C
S. typhimurium
TA100, TA1535 and
TA1538 (+GSTA1-1 and
GSTP1-1)
S. typhimurium
TA1535 (+GST5-5),
TA1535 (wild type)
6-hr exposure to 0, 7,000, and
14,000 ppm
Up to 3,600 ug/plate
Up to 3,600 jig/plate
6-hr exposure to 0, 7,000, and
14,000 ppm
Up to 84,000 ppm, 3-d
exposure
10 uL/plate
2- and 6-hr exposures to 0,
2,500, 5,000, 7,500,
10,000 ppm
0, 50, 100, and 200 uL/plate
0-2.0 mM/plate
+
+
+
+
+ forTA100,TA1950,
WU361089
-forTA1535
+
+
+ forTA100
-forTA1535,TA1538
+ forTA1535(+GST5-5)
- for TA1535 (wild type)
+
++
++
+
Not determined
Not
determined
+
Not determined
Not determined
Jongenetal. (1978)
Gockeetal. (1981)
Gockeetal. (1981)
Jongenetal. (1982)
Green (1983)
Osterman-Golkar et al.
(1983)
Zeiger(1990)
Dillon etal. (1992)
Simula etal. (1993)
Pegrametal. (1997);
Thier etal. (1993)
(Table 4-29 continues on next page)
169
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-29. Results from in vitro genotoxicity assays of dichloromethane with bacteria, yeast, or fungi
Assay
Reverse mutation
Reverse mutation
Forward mutation
Gene mutation
Prophage induction
Reverse mutation
Forward mutation
Forward mutation
Test system
S. typhimurium
TA100,TA100/NG-lld
S. typhimurium TA100,
RSJ100C
S. typhimurium BA13
S. typhimurium
TA1535/pSK1002c,
NM5004C
E. coli K-39 (I)
E. co//WP2uvrapKM101
E. coli K12
E. coli Uvr+, UvrB"
Concentration(s)
0, 30, 60, 130 mM/plate
Up to 24,000 ppm
0-130 umol/plate
0,2.5,5.0, 10, 20 mM
10 uL/plate
2- and 6-hr exposures to 6,300,
12,500, 25,000, and
50,000 ppm
0, 30, 60, 130 mM/plate
20,000 ppm
Results
Without metabolic
activation
-S9
++ for TA100
+ forTA100/NG-ll
+ forTA100
+ forRSJ100
+++
+ NM5004
-TA1535/pSK1002
+
+
-
+
With metabolic
activation
+S9
Not determined
+ for TA100
+ forRSJ100
+
Not determined
Not determined
+
+
Not determined
Reference
Graves etal. (1994a)
DeMarinietal. (1997)
Roldan-Arjona and
Pueyo (1993)
Oda etal. (1996)
Osterman-Golkar et al.
(1983)
Dillon etal. (1992)
Graves etal. (1994a)
Zielenska etal. (1993)
Fungi and yeasts
Mitotic segregation
Gene conversion
and recombination
Aspergillus nidulans
-diploid strain PI
-haploidstrain35
Saccharomyces cerevisiae
-strains D7 and D4
Up to 8,000 ppm
Up to 209 mM
+ only at 4,000 ppm; no
dose-response relationship
established
+
Not determined
Not determined
Crebelli etal. (1988)
Callen etal. (1980)
"Bacterial strains that have GSH (e.g., TA100, TA 98).
bBacterial strains that do not have GSH (e.g., TA1535).
°Bacterial strains engineered to have more GSH activity than wild type.
dBacterial strains engineered to have less GSH activity than wild type.
170
DRAFT - DO NOT CITE OR QUOTE
-------
Dichloromethane induced mutations in Salmonella typhimurium strains containing GSH
(e.g., TA100, TA98). These effects were not markedly influenced by the addition of exogenous
mammalian liver fractions, suggesting that endogenous metabolism in these strains was
sufficient to activate dichloromethane (Green, 1983; Jongen et al., 1982; 1978; Gocke et al.,
1981). In support of this hypothesis, dichloromethane exposure of NG-11, a glutathione-
deficient variant of S. typhimurium strain TA100, produced twofold fewer base-pair mutations
compared with exposure of strain TA100, which produces normal levels of GSH. Furthermore,
this difference was not apparent when the culture medium contained 1 mM GSH (Graves et al.,
1994a).
In contrast to strain TA100, S. typhimurium strains TA1535, TA1537, and TA1538
(strains deficient in GSH) did not develop base-pair mutations in response to dichloromethane
exposure (Pegram et al., 1997; Simula et al., 1993; Thier et al., 1993; Osterman-Golkar et al.,
1983; Gocke et al., 1981). However, when strain TA1535 was transfected with rat GST-T1,
dichloromethane induced base-pair reverse mutations (DeMarini et al., 1997; Pegram et al.,
1997; Thier et al., 1993). A 60-fold higher concentration of dichloromethane was needed to
induce a response (i.e., a sixfold increase over background levels in reverse mutations) in
S. typhimurium strain TA100 than in TA1535 transfected with rat GST-T1 (DeMarini et al.,
1997). This study also included several trihalomethanes; dichloromethane was several-fold less
genotoxic than dibromochloromethane or bromoform, but was similar in potency to
bromodichloromethane (DeMarini et al., 1997; Pegram et al., 1997). The authors suggest that
these results support a role of GST-T1 in the mutagenicity of the trihalomethanes.
The mutagenic effects of dichloromethane have also been examined in fungi and yeast
assays with both systems reporting positive results. Fungi assays were positive for mitotic
segregation in Asperigillus ridulam (Crebelli et al., 1988), but there was not a dose response
relationship as only the 4,000 ppm dichloromethane exposure was positive (exposure up to
8,000 ppm). A yeast assay was positive for gene conversion and recombination in
Saccharomyces cerevisiae for concentrations up to 209 mM (Callen et al., 1980).
Mammalian assays. In the in vitro mammalian system studies conducted with murine
cell lines (Table 4-30), dichloromethane was negative for producing point mutations in the
mouse lymphoma L5178Y cell line (Thilagar et al., 1984) but was positive in producing DNA
single stranded breaks (SSBs) in mouse Clara cells (Graves et al., 1995) and mouse hepatocytes
(Graves et al., 1994b). Given that exposure to dichloromethane results specifically in lung and
liver tumors, this pattern is not surprising. Additionally, GST is localized in the nucleus of
hepatocytes and lung cells in the mouse (Mainwaring et al., 1996), which would also increase
sensitivity of these particular cell fractions to genotoxic effects of dichloromethane. DNA SSBs
were induced at lower concentrations in mouse hepatocytes (0.5 mM) than in rat hepatocytes
(30 mM). The extent of DNA damage was shown to be reduced to the background level seen in
control (no exposure) conditions by pretreating the cells with buthionine sulfoxime to deplete
171 DRAFT - DO NOT CITE OR QUOTE
-------
cellular levels of GSH and thus inhibit dichloromethane metabolism via the GST pathway
(Graves et al., 1995, 1994b). Similar results were seen in mouse lung Clara cells. Freshly
isolated Clara cells from the lungs of B6C3Fi mice also showed significantly increased,
concentration-dependent amounts of DNA SSBs when incubated in vitro for 2 hours in the
presence of 5-60 mM dichloromethane. Pretreatment with buthionine sulphoximine before
Clara-cell isolation or the presence of buthionine sulphoximine in the culture medium decreased
the amount of induced in vitro DNA damage.
172 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-30. Results from in vitro genotoxicity assays of dichloromethane with mammalian systems, by type of test
Assay
Test system
Concentrations
Results
Reference
Mouse
Point mutation
DNASSBsby
alkaline elution
DNASSBsby
alkaline elution
DNA-protein cross-
links
Mouse lymphomaL5178Y
cells (thymidine kinase
locus)
Mouse hepatocytes
(B6C3FO
Mouse Clara cells
(B6C3FO
Mouse hepatocytes
(B6C3FO
Not provided
0, 0.4, 3.0, 5.5 mM
0, 5, 10, 30, 60 mM
0.5-5 mM
Negative
Positive at 0.4 mM
Positive, but DNA damage was reduced by incubating in
the presence of GSH depletory
Positive
Thilagaretal. (1984)
Graves etal. (1994b)
Graves etal. (1995)
Casanova etal. (1997)
Rat
Unscheduled DNA
synthesis
Unscheduled DNA
synthesis
DNASSBsby
alkaline elution
DNA-protein cross-
links
Rat hepatocytes
Rat hepatocytes
Rat hepatocytes
Rat hepatocytes
Up to 16 mM (measured);
30 mM (nominal)
Not provided
0, 30, 90, 90 mM
0.5-5 mM
Negative
Marginally positive
Positive at 30 mM
Negative
Andrae and Wolff (1983)
Thilagaretal. (1984)
Graves etal. (1994b)
Casanova etal. (1997)
Hamster with GST activity from mouse
DNA-protein cross-
links
HPRT mutation
analysis
DNA SSBs and
DNA-protein cross-
links
DNA-protein cross-
links
Chinese hamster ovary
cells
Chinese hamster ovary
cells
Chinese hamster ovary
cells
Syrian golden hamster
hepatocytes
60 mM
2,500 ppm
3,000 ppm (0.3%, volume
per volume [v/v]) and
5,000 ppm (0.5%, v/v)
0.5-5 mM
Positive with mouse liver cytosol (negative without) at
much higher concentrations of dichloromethane (60 mM)
than formaldehyde (0.5-4 mM)
Positive with mouse liver cytosol
Positive at concentration of 0.5% (v/v) for SSBs in
presence of mouse liver cytosol, but increase in DNA-
protein cross-links marginal; formaldehyde (in absence of
mouse liver cytosol) was positive at 0.5 mM for both DNA
SSBs and DNA-protein cross-links; Chinese hamster ovary
cell cultures were suspended
Negative
Graves etal. (1994b)
Graves etal. (1996)
Graves and Green (1996)
Casanova etal. (1997)
173
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-30. Results from in vitro genotoxicity assays of dichloromethane with mammalian systems, by type of test
Assay
Comet assay
Test system
V79 hamster cells
transfected with mouse
GST-T1
Concentrations
2.5,5, 10 mM
Results
A significant, dose-dependent increase in DNA damage
resulting from DNA-protein cross-links in V79 cells
transfected with mouse GST-T1 compared to parental cells
Reference
Hu et al. (2006)
Hamster without GST activity from mouse
Forward mutation
Unscheduled DNA
synthesis
Sister chromatid
exchange
Chromosomal
aberrations
Sister chromatid
exchange
DNA and protein
synthesis
DNASSBsby
alkaline elution
Chinese hamster epithelial
cells (hgprt locus)
Chinese hamster epithelial
cells
Chinese hamster epithelial
cells
Chinese hamster ovary
cells
Chinese hamster ovary
cells
Chinese hamster ovary
cells
Hamster hepatocytes
5,000, 10,000, 30,000,
50,000 ppm
5,000, 10,000, 30,000,
50,000 ppm
5,000, 10,000, 20,000,
30,000, and 40,000 ppm
Not provided
Not provided
Up to l,OOOug/mL
0.4-90 mM
Negative
Negative
Weak positive with or without rat-liver microsomal system
Positive, independent of rat liver S9
Negative with or without rat liver S9
Negative
Negative
Jongenetal. (1981)
Jongenetal. (1981)
Jongenetal. (1981)
Thilagar and Kumaroo
(1983)
Thilagar and Kumaroo
(1983)
Garrett and Lewtas (1983)
Graves etal. (1995)
Calf
DNA Adducts
DNA Adducts
Calf thymus DNA
Calf thymus DNA
50 mM
0-8.0 umol (0-60 mM)
Positive in the presence of bacterial GST DM11 and
dichloromethane dehalogenase; adducts primarily formed
with the guanine residues
Positive in the presence of bacterial GST DM1 1, rat
GST5-5, and human GSTT1 1; adducts primarily formed
with the guanine residues
Kayser and Vuilleumier
(2001)
Marsch et al. (2004)
Human
Unscheduled DNA
synthesis
Unscheduled DNA
synthesis
Sister chromatid
exchange
Chromosomal
aberrations
Human peripheral
lymphocytes
Primary human fibroblast
Human peripheral
lymphocytes
Human peripheral
lymphocytes
250, 500, 1,000 ppm
5,000, 10,000, 30,000,
50,000 ppm
Not provided
Not provided
Negative with or without rat liver S9
Negative
Weak positive
Positive
Perocco and Prodi (1981)
Jongenetal. (1981)
Thilagar etal. (1984)
Thilagar etal. (1984)
174
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-30. Results from in vitro genotoxicity assays of dichloromethane with mammalian systems, by type of test
Assay
DNASSBsby
alkaline elution
Micronucleus test
DNA-protein cross-
links
DNA damage by
comet assay
Sister chromatid
exchange
Test system
Human hepatocytes
Human AHH-1, MCL-5,
h2El cell lines
Human hepatocytes
Primary human lung
epithelial cells
Primary human peripheral
blood mononuclear cells
Concentrations
Up to 120 mM
Up to 10 mM
0.5-5 mM
10, 100, 1,000 uM
0, 15, 30, 60, 125, 250, 500
ppm for 72 hours.
Results
Negative at concentrations between 5 and 120 mM
Positive in all three cell lines
Negative
Weak trend, independent of GST activity (GST enzymatic
activity not present in the cultured cells)
Sister chromatid exchanges significantly increased at
exposures of 60 ppm and higher, most strongly in the high
GST-T1 activity group
Reference
Graves etal. (1995)
Dohertyetal. (1996)
Casanova etal. (1997)
Landi et al. (2003)
Olvera-Bello etal. (2010)
HPRT = hypoxanthine-guanine phosphoribosyl transferase
175
DRAFT - DO NOT CITE OR QUOTE
-------
In a series of experiments with freshly isolated hepatocytes from multiple species
(Table 4-30), DNA-protein cross-links were detected in hepatocytes of B6C3Fi mice but not in
hepatocytes of F344 rats, Syrian golden hamsters, or three human subjects following 2-hour in
vitro exposures to concentrations ranging from 0.5 to 5 mM dichloromethane (Casanova et al.,
1997). Within the range of concentrations tested, DNA-protein cross-links in mouse hepatocytes
appeared to increase with increasing concentration of dichloromethane.
Negative results for dichloromethane were predominantly seen in in vitro test systems
that used rat or hamster cell lines with low or no GST activity (Table 4-30). Several genotoxic
endpoints including DNA and protein synthesis (Garrett and Lewtas, 1983), chromosomal
aberrations or sister chromatid exchanges (Thilagar et al., 1984; Thilagar and Kumaroo, 1983;
Jongen et al., 1981), unscheduled DNA synthesis (Thilagar et al., 1984; Andrae and Wolff, 1983;
Jongen et al., 1981), and mutations (Thilagar et al., 1984; Jongen et al., 1981) were evaluated in
these cell lines. In contrast, positive results (DNA-protein cross-links and DNA SSBs) were
observed when mouse liver cytosol was included in Chinese hamster ovary (CHO) cells (Graves
et al., 1995, 1994b). Dichloromethane also induced hypoxanthine-guanine phosphoribosyl
transferase (HPRT) gene mutations in CHO cells when they were incubated with GST-competent
mouse liver cytosol preparations (Graves et al., 1996).
The instability of the S-(chloromethyl)glutathione-adducts presents considerable
challenges to studies of these products (Hashmi et al., 1994). Kayser and Vuilleumier (2001),
however, demonstrated the formation of DNA adducts with radiolabeled dichloromethane in calf
thymus DNA in the presence of dichloromethane dehalogenase/GST purified from a bacterial
source (Methylophilus sp. strain DM11) and GSH (Table 4-30). The type of adduct could not be
identified because of low yield, but it was determined that guanine was more actively
incorporated than cytosine, adenine, or thymine by at least twofold in the presence of
GST-activated dichloromethane, indicating a base specificity for these adducts. Incubation of
calf thymus DNA with formaldehyde and GSH, however, did not result in detectable DNA
adduct formation. In another study, Marsch et al. (2004) further evaluated the presence of
adducts in calf thymus DNA in the presence of dichloromethane and human (GST-T1), rat
(GST5-5), or bacterial (DM11) GST (Marsch et al., 2004). This study found that all three
enzymes yielded a similar pattern of adduct formation, forming primarily with guanine and to a
lesser extent with cytosine, adenine, and thymine (two- to threefold less than guanine), consistent
with the results reported by Kayser and Vuilleumier (2001). High levels of guanosine-specific
adducts were also seen with S-(l-acetoxymethyl)glutathione, a compound that is structurally
similar but more stable than S-(chloromethyl)glutathione (Marsch et al., 2001). These findings
indicate that the S-(chloromethyl)glutathione intermediate formed by GSH conjugation has
mutagenic potential and is likely responsible, at least in part, for the mutagenic response
observed following dichloromethane exposure.
176 DRAFT - DO NOT CITE OR QUOTE
-------
In studies with human cell lines or isolated cells, positive results were reported for sister
chromatid exchanges and chromosomal aberrations (Thilagar et al., 1984; Olvera-Bello et al.,
2010), for mitotic index (Olvero-Bello et al., 2010) and in the micronucleus test (Doherty et al.,
1996). Negative results with human cells were seen in the unscheduled DNA synthesis assays
(Jongen et al., 1981; Perocco and Prodi, 1981), DNA SSBs, and DNA-protein cross-links
(Casanova et al., 1997; Graves et al., 1995).
Dichloromethane-induced DNA damage (comet assay) was examined in primary cultures
of human lung epithelial cells collected by brush biopsy from four healthy volunteers (Landi et
al., 2003). This study was designed to assess the genotoxicity of four thrihalomethanes
(chloroform, bromodichloromethane, dibromochloromethane, and bromoform), with
dichlorom ethane included because of its known activation by GST-T1. Two of the subjects were
of the GST-T1+ genotype, and two were of the GST-T1" genotype.6 The cells had been frozen,
and GST activity was not detected in the cultured cells. DNA damage was reported to occur in
the combined GST-T1" samples (tail extent moment 7.1, 13.7, and 15.3 in the 10, 100, and
1,000 jiM dichloromethane groups, respectively) but not in the combined GST-T1+ samples (tail
extent moment 8.1, 11.5, and 10.4 in the 10, 100, and 1,000 jiM dichloromethane groups,
respectively). This pattern was not seen across the individual samples, however, as only one
sample exhibited a clear dose-response gradient. Given the absence of GST activity, an analysis
combining the four samples could provide a more informative picture of the dose-response
relationship between dichloromethane (and the other compounds studied) and DNA damage.
For dichloromethane, values of 9.4, 7.6, 12.6, and 12.9 were seen in the 0, 10, 100, and
1,000 jiM groups, respectively. This pattern was similar to that seen with chloroform (9.4, 6.9,
11.4, and 12.7 in the 0, the 10, 100, and 1,000 jiM groups, respectively) but weaker than the
pattern for bromoform (9.4, 12.5, 15.8, and 18.2 in the 0, the 10, 100, and 1,000 |iM groups,
respectively), and much weaker than the pattern for bromodichloromethane (9.4, 25.2, 28.5, and
39.1 in the 0, the 10, 100, and 1,000 jiM groups, respectively).7 No dose-response gradient was
seen with dibromochloromethane (9.4, 6.5, 8.1, and 8.0 in the 0, 10, 100, and 1,000 jiM groups,
respectively). This relative pattern is also seen in the estimated slopes (beta coefficient for the
change in tail extent moment per unit increase in jiM concentration): 0.0, 0.003, 0.004, 0.006,
and 0.02 for dibromochloromethane, dichloromethane, chloroform, bromoform, and
bromodichloromethane, respectively (statistical significance not reported). A stronger and more
consistent response was seen under the same experimental conditions with
6Landi et al. (2003) did not clearly describe their treatment of GST-Tl+/~ heterozygote genotypes; EPA considers it
likely that they were included in the pool from which the GST-T1+ samples were drawn. In addition, there is a
discrepancy in the paper regarding the coding of the GST-T1 genotypes. Samples A and C are noted to be the
GST-Tr samples in one part of the paper, and C and D are described as the GST-XT samples in another part of the
paper.
'These values are based on the mean of the GST-T1+and the GST-Tr samples from Table 1 ofLandietal. (2003).
177 DRAFT - DO NOT CITE OR QUOTE
-------
bromodichloromethane, but dibromochloromethane resulted in no increase in DNA damage in
any of the donor cells at any concentration tested.
Genotoxicity was examined in human peripheral blood mononuclear cells obtained from
a total of 20 healthy, non-smoking, male volunteers; 4 of the 20 had low GST-T1 activity, 10 had
medium GST-T1 activity, and 6 had high GST-T1 activity (Olvero-Bello et al., 2010). Mean
GST-T1 activity was 0.077, 0.325 and 7.365 nmol HCOH/min/mg protein, respectively, in the
low, medium and high activity groups. The cultured cells were exposed to dichloromethane (15,
30, 60, 125, 250, or 500 ppm) for 72 hours and sister chromatid exchanges were evaluated as a
measure of genotoxicity. Frequency of sister chromatid exchanges significantly increased
beginning at 125 ppm. Compared with controls, the strongest increase in sister chromatid
exchanges was seen among the highest GST-T1 activity group, with a linear increase from a 1.5-
fold increase at 30 ppm to a 2.5-fold increase at 500 ppm. The low activity group exhibited a
more modest increase, up to a 1.5-fold increase at 500 ppm. Little change was seen in the
medium activity group. Indices of cytoxicity (mitotic index) and cytostaticity (cell proliferation
kinetics) were also examined. The mitotic index decreased significantly and in a dose-dependent
manner at exposures of 60-500 ppm; this decrease was seen in each of the GST-T1 activity
groups. The mitotic index was approximately 50% lower in the highest activity group compared
to the medium and low activity groups at all concentrations. Similar patterns were seen with the
cell proliferation kinetics measure. This study demonstrates evidence of genotoxic and cytoxic
damage in human in vitro cell experiments at relatively low exposure concentrations, with the
strongest effects seen in samples with the highest GST-T1 activity .
Several studies have examined patterns of mutations or DNA damage with
dichloromethane and formaldehyde to assess the relative role of S-(chloromethyl)glutathione and
formaldehyde in the observed genotoxicity. In a study in CHO cells incubated with
dichoromethane (0.3% plus mouse liver cytosol), 2.5-fold increases in DNA-protein cross-links
that are indicative of formaldehyde exposure were observed, compared with a 25-fold increase
when 1 mM formaldehyde was added directly to cultures. Both treatments induced a comparable
degree of DNA SSBs (Graves and Green, 1996). In a subsequent study, Graves et al. (1996)
compared the mutational spectra induced by dichloromethane to that induced by direct addition
of formaldehyde or 1,2-dibromoethane (a chemical known to act through a glutathionyl
conjugate metabolite) at the HPRT locus in CHO cells. The mutations induced by
dichloromethane and 1,2-dibromoethane were predominantly GC to AT transitions, while all six
formaldehyde-induced mutants sequenced were single base transversions. This provided further
evidence that the S-(chloromethyl)glutathione intermediate may be primarily responsible for
dichloromethane genotoxicity. In contrast, Hu et al. (2006) found evidence of significant
amounts of formaldehyde formation following dichloromethane exposure in the cytosol of
V79 (hamster) cells transfected with the murine GST-T1 gene compared to the parent cell line.
In accordance with this, they observed concentration-dependent increases in DNA-protein cross-
178 DRAFT - DO NOT CITE OR QUOTE
-------
links in the GST-T1 transfected cells using the comet assay with and without proteinase K
treatment that frees DNA from cross-links and allows DNA migration. These findings are
consistent with those by Casanova et al. (1997), who performed a comparison of the amounts of
DNA-protein and RNA-formaldehyde cross-links formed following dichloromethane exposure in
hepatocytes isolated from mice, rats, hamsters, and human GST-T1 genetic variants. Only
DNA-protein cross-links were observed in mouse hepatocytes, but RNA-formaldehyde cross-
links were found in all species and were highest in the mouse hepatocytes (4-, 7-, and 14-fold
higher than rats, humans, and hamsters, respectively). These results showed that human
hepatocytes can metabolize dichloromethane to formaldehyde, resulting in RNA-formaldehyde
cross-links. In addition, the results indicate that there is considerable variation among species
and that the human variation in the GST-T1 gene can also affect the amount of formaldehyde
produced. The authors also noted that comparing results following ectopic addition of
formaldehyde directly to cells with results following dichloromethane metabolism in situ can be
misleading, as the formaldehyde produced internally may reside in different locations
intracellularly, potentially affecting the capability of interacting with DNA. These results show
that, while most studies indicate the importance of the S-(chloromethyl)glutathione intermediate
in mediating genotoxic damage following dichloromethane exposure, DNA damage resulting
from formaldehyde formation should also be considered.
4.5.1.2. In Vivo Genotoxicity Assays
Genotoxicity findings in Drosophila melanogaster assays are mixed (Table 4-31). A
study of gene mutation in D. melanogaster showed a marginal increase in sex-linked recessive
deaths following oral exposure (Gocke et al., 1981). An additional feeding study (Rodriguez-
Arnaiz, 1998) reported a positive response in the somatic w/w+ assay. A third study of
D. melanogaster (Kramers et al., 1991) found no evidence of increased sex-linked recessive
deaths, somatic mutations, or recombinations following exposure to airborne dichloromethane.
Table 4-31. Results from in vivo genotoxicity assays of dichloromethane in
insects
Assay
Gene mutation (sex-
linked recessive lethal)
Gene mutation (sex-
linked recessive lethal,
somatic mutation and
recombination)
Somatic w/w+ assay
Test system
Drosophila
Drosophila
Drosophila
Doses
125, 620 mM
6 hrs— 1,850, 5,500 ppm
1 wk— 2,360, 4,660 ppm
2 wks— 1,370, 2,360 ppm
(all approximate)
50, 100, 250, 500 mM
Result
Positive (feeding
exposure)
Negative (inhalation
exposure)
Positive (feeding
exposure)
Reference
Gocke etal. (1981)
Kramers etal. (1991)
Rodriguez-Arnaiz (1998)
179
DRAFT - DO NOT CITE OR QUOTE
-------
Some in vivo studies investigating certain genotoxic endpoints in mice exposed to
dichloromethane produced negative results (Table 4-32). Unscheduled DNA synthesis was not
induced in hepatocytes from mice (and rats) after 2- or 6-hour inhalation exposures to
concentrations that were carcinogenic in the NTP (1986) mouse bioassay (Trueman and Ashby,
1987) or after other exposure routes (Lefevre and Ashby, 1989). Although positive results were
not observed in the unscheduled DNA synthesis studies, it is generally recognized that this assay
is not sensitive for detecting genotoxic chemicals (Eastmond et al., 2009; Madle et al., 1994).
Distinct, unequivocal cytogenetic effects (e.g., induction of micronuclei, sister chromatid
exchanges, or chromosome aberrations) were not consistently found in bone marrow or
erythrocytes in several studies of mice after acute oral exposures (Sheldon et al., 1987) or
parenteral exposures (Westbrook-Collins et al., 1990; Gocke et al., 1981). However,
tumorigenic effects in mice are generally localized to the liver and lung (due to high GST
activity) and therefore, it is not surprising that genotoxic effects were for the most part not
observed in the bone marrow or erythrocytes (cell types with minimal GST activity). Crebelli et
al. (1999) stated that genotoxic effects induced by halogenated hydrocarbons (such as
dichloromethane) are not very effective in inducing micronucleus formation in mouse bone
marrow, and a negative bone marrow micronucleus assay should not offset the consistently
positive in vitro results.
180 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-32. Results from in vivo genotoxicity assays of dichloromethane in mice
Assay
Micronucleus test
Micronucleus test
Micronucleus test
Micronucleus test
DNA synthesis
Unscheduled DNA synthesis
Sister chromatid exchange
Sister chromatid exchange
Sister chromatid exchange
Sister chromatid exchange
Chromosome aberrations
Chromosome aberrations
Chromosome aberrations
DNA-protein cross-links
Test system
Mouse bone marrow
(NMRI)
Mouse bone marrow
(C57BL/6J/Alpk)
Mouse peripheral red
blood cells (B6C3FO
Mouse peripheral red
blood cells (B6C3FO
Mouse liver (B6C3FO
Mouse hepatocytes
(B6C3FO
Mouse bone marrow
(C57BL/6J)
Mouse bone marrow
(B6C3FO
Mouse lung cells and
peripheral lymphocytes
(B6C3FO
Mouse lung cells
(B6C3FO
Mouse bone marrow
(C57BL/6J)
Mouse bone marrow
(B6C3FO
Mouse lung and bone
marrow cells (B6C3Fi)
Mouse liver and lung
cells (B6C3FO
Route and dose
425, 850, or 1,700 mg/kg
Gavage,
1,250, 2,500, and
4,000 mg/kg
Inhalation 6 hr/d, 5 d/wk, 0,
4,000, 8,000 ppm
Inhalation, 6 hr/d, 5 d/wk, 0,
2,000 ppm
Gavage, 1,000 mg/kg;
inhalation, 4,000 ppm
Inhalation, 2,000 and
4,000 ppm.
Intraperitoneal, 100, 1,000,
1,500, 2,000 mg/kg
Subcutaneous, 0, 2,500,
5,000 mg/kg
Inhalation 6 hr/d, 5 d/wk,
0, 4,000, 8,000 ppm
Inhalation 6 hr/d, 5 d/wk,
0, 2,000 ppm
Intraperitoneal, 100, 1,000,
1,500, 2,000 mg/kg
Subcutaneous, 0, 2,500,
5,000 mg/kg
Inhalation, 6 hr/d, 5 d/wk,
0, 4,000, 8,000 ppm
Inhalation, 6 hr/d, 3 d,
4,000 ppm
Duration
Two doses
Single dose
2wk
12wks
Single dose;
2hrs
2 or 6 hrs
Single dose
Single dose
2 wks
12wks
Single dose
Single dose
2 wks
3d
Results
Negative at all doses
Negative at all doses
Positive at 4,000 and
8,000 ppm
Positive at 2,000 ppm
Negative in both oral and
inhalation studies
Negative
Negative
Negative at all doses
Positive at 8,000 ppm for both
mouse lung cells and peripheral
lympocytes
Positive at 2,000 ppm
Negative
Negative
Positive at 8,000 ppm for both
mouse lung and bone marrow
cells.
Positive in mouse liver cells at
4,000 ppm; negative in mouse
lung cells
Reference
Gockeetal. (1981)
Sheldon et al.
(1987)
Allen etal. (1990)
Allen etal. (1990)
Lefevre and Ashby
(1989)
Trueman and Ashby
(1987)
Westbrook-Collins
etal. (1990)
Allen etal. (1990)
Allen etal. (1990)
Allen etal. (1990)
Westbrook-Collins
etal. (1990)
Allen etal. (1990)
Allen etal. (1990)
Casanova et al.
(1992)
(Table 4-32 continues on next page)
181
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-32. Results from in vivo genotoxicity assays of dichloromethane in mice
Assay
DNA-protein cross-links
DNA SSBs by alkaline elution
DNA SSBs by alkaline elution
DNA damage by comet assay
DNA damage by comet assay
DNA adducts
Kras and Hras oncogenes
p53 tumor suppressor gene
Test system
Mouse liver and lung
cells (B6C3FO
Mouse hepatocytes
(B6C3FO
Mouse liver and lung
homogenate (B6C3FO
Mouse liver and lung
cells (CD-I)
Mouse stomach, urinary
bladder, kidney, brain,
bone marrow (CD-I)
Mouse liver and kidney
cells (B6C3FO
Mouse liver and lung
tumors (B6C3FO
Mouse liver and lung
tumors (B6C3FO
Route and dose
Inhalation, 6 hr/d, 150, 500,
1,500, 3,000, 4,000 ppm
Inhalation, 2,000 and
4,000 ppm
Liver: inhalation, 2,000,
4,000, 6,000, 8,000 ppm
Lung: inhalation, 1,000,
2,000, 4,000, 6,000 ppm
Gavage, 1,720 mg/kg;
organs harvested at 0
(control), 3, and 24 hrs
Gavage, 1,720 mg/kg;
organs harvested at 0
(control), 3, and 24 hrs
Intraperitoneal, 5 mg/kg
0, 2,000 ppm
0, 2,000 ppm
Duration
3d
3 or 6 hrs
3hrs
3hrs
Single dose
Single dose
Single dose
Up to 104 wks
Up to 104 wks
Results
Positive in mouse liver cells at
500-4,000 ppm; negative in
mouse lung cells
Positive at 4,000 ppm at 3 and
6 hrs
Liver: positive at 4,000-8,000
ppm
Lung: positive at 2,000-
4,000 ppm
Positive only at 24 hrs after
dosing
Negative 3 or 24 hr after
dosing
Negative
No difference in mutation
profile between control and
dichloromethane-induced liver
tumors; number of spontaneous
lung tumors (n = 4) limits
comparison at this site
Loss of heterozygocity
infrequently seen
Reference
Casanova et al.
(1996)
Graves et al.
(1994b)
Graves etal. (1995)
Sasaki etal. (1998)
Sasaki etal. (1998)
Watanabe et al.
(2007)
Devereux et al.
(1993)
Hegi etal. (1993)
182
DRAFT - DO NOT CITE OR QUOTE
-------
When genotoxic endpoints were examined in the cancer target tissues (liver and lung) in
mice exposed to dichloromethane, positive results were consistently reported (Table 4-32).
Increased sister chromatid exchanges were found in lung cells and peripheral lymphocytes from
mice exposed by inhalation for 2 weeks to 8,000 ppm or for 12 weeks to 2,000 ppm (Allen et al.,
1990). Under the same exposure conditions, increased chromosomal aberrations in lung and
bone cells and micronuclei in peripheral red blood cells also were found (Allen et al., 1990).
DNA-protein cross-links were detected in mouse hepatocytes but not in lung cells after a 3-day
inhalation exposure to 4,000 ppm (Casanova et al., 1992) and between 500 and 4,000 ppm
(Casanova et al., 1996). DNA damage, detected as increased DNA SSBs, was observed in liver
and lung tissue of B6C3Fi mice immediately following 3-hour exposures (Graves et al., 1995).
The DNA damage was not detectable 2 hours after in vivo exposure, indicating that DNA repair
occurs rapidly. Pretreatment of mice with buthionine sulphoximine, a GSH depletor, caused a
decrease to levels seen in controls in the amount of DNA damage detected immediately after in
vivo exposure in liver and lung tissue, indicating GSH involvement in the genotoxic process.
DNA damage (detected by the comet assay) was also reported in liver and lung tissues from male
CD-I mice sacrificed 24 hours after administration of a single oral dose of 1,720 mg/kg of
dichloromethane (Sasaki et al., 1998). In this study, DNA damage in lung and liver was not
detected 3 hours after dose administration, and no DNA damage occurred at either time point in
several other tissues in which a carcinogenic response was not seen in chronic animal cancer
bioassays (e.g., stomach, kidney, bone marrow).
Formation of DNA adducts was evaluated in male and female B6C3Fi mice as well as in
male F344 rats (Watanabe et al., 2007). Animals were administered 5 mg/kg intraperitoneally of
radiolabeled dichloromethane and sacrificed at 1 or 8 hours after administration. The kidneys
and livers were removed and the DNA was isolated from these tissues to evaluate formation of
DNA adducts. At the administered dose, DNA adducts were not detected.
Other studies in mice have looked for mutations in specific oncogenes (K-ras or H-ras)
(Devereux et al., 1993) or in a tumor suppressor gene (p53) (Hegi et al., 1993) in liver or lung
tumors from dichloromethane-exposed mice. These studies have not demonstrated exposure-
related patterns of mutations in these genes, although it should be noted that the statistical power
of this analysis for the lung tumors is limited (discussed further in Sections 4.5.2 and 4.5.3).
Results from in vivo studies in other mammals (i.e., rats and hamsters) of hepatocyte
sensitivity to dichloromethane induction of DNA SSBs (Table 4-33) are consistent with
interspecies differences in the induction of liver tumors in the inhalation cancer bioassays. A
gavage study in rats reported the presence of DNA SSBs with a dose of 1,275 mg/kg (Kitchin
and Brown, 1989). The other available studies, however, did not find any genotoxicity following
dichloromethane exposure. No increase in unscheduled DNA synthesis in rat hepatocytes was
seen following inhalation of dichloromethane for 2-6 hours at 2,000 or 4,000 ppm (Trueman and
Ashby, 1987), exposure by gavage up to 1,000 mg/kg (Trueman and Ashby, 1987), or
183 DRAFT - DO NOT CITE OR QUOTE
-------
intraperitoneal exposure of 400 mg/kg (Mirsalis et al., 1989). DNA adducts were not detected in
the livers and kidneys of male F344 rats dosed with 5 mg/kg dichloromethane intraperitoneally
(Watanabe et al., 2007). DNA SSBs were significantly increased in hepatocytes isolated from
B6C3Fi mice exposed to 4,831 ppm (4,000 ppm nominal) for 6 hours but were not increased in
hepatocytes from Sprague-Dawley rats exposed to 4,527 ppm (4,000 ppm nominal) for 6 hours
(Graves et al., 1994b). Results from in vivo interspecies comparisons of dichloromethane
induction of DNA-protein cross-links in hepatocytes (expected products of the GSH pathway)
are also consistent with the hypothesis that the greater activity of the GST pathway in the mouse
contributes to the increased sensitivity to genotoxic effects. DNA-protein cross-links were
formed in the liver of mice but not hamsters following in vivo exposure to air concentrations
ranging from 500 to 4,000 ppm, 6 hours/day for 3 days (Casanova et al., 1996). The absence of a
genotoxic response in the rat and hamster is consistent with considerably lower GST activity and
therefore, these mammalian systems would be expected to be less sensitive at detecting
genotoxic effects than the studies conducted in mice.
184 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-33. Results from in vivo genotoxicity assays of dichloromethane in rats and hamsters
Assay
Unscheduled DNA synthesis
Unscheduled DNA synthesis
Unscheduled DNA synthesis
DNA SSBs by alkaline elution
DNA SSBs by alkaline elution
DNA SSBs by alkaline elution
DNA-protein cross-links
DNA adducts
Test system
Rat hepatocytes
Rat hepatocytes
Rat hepatocytes
Rat hepatocytes
Rat liver homogenate
Rat liver and lung
homogenate
Hamster liver and lung
cells
Rat liver and kidney
cells
Route and dose
Gavage, 100, 500,
1,000 mg/kg
Inhalation, 2 or 6 hrs,
2,000 and 4,000 ppm
Intraperitoneal, single dose,
400 mg/kg
Inhalation, 3 or 6 hrs,
2,000 and 4,000 ppm
Gavage, 2 doses, 425 mg/kg
and 1,275 mg/kg,
administered 4 and 21 hrs
before liver harvesting
Liver: inhalation, 4,000,
5,000 ppm
Lung: inhalation, 4,000 ppm
Inhalation, 6 hr/d, 500,
1,500, 4,000 ppm
Intraperitoneal, 5 mg/kg
Duration
Liver harvested 4 and
12 hrs after dosing
2 or 6 hrs
Single dose
3 or 6 hrs
4 or 2 1 hrs (time
between dosing and
liver harvesting)
3hrs
3hrs
3d
Single dose
Results
Negative 4 or 12 hrs after
dosing
Negative at both
concentrations and exposure
durations
Negative 48 hrs after dosing
Negative at all
concentrations and time
points
Positive at 1,275 mg/kg
Negative for both liver and
lung at all concentrations
Negative at all
concentrations
Negative
Reference
Trueman and
Ashby (1987)
Trueman and
Ashby (1987)
Mirsalis et al.
(1989)
Graves et al.
(1994b)
Kitchin and
Brown (1989)
Graves et al.
(1995)
Casanova et al.
(1996)
Watanabe et al.
(2007)
185
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-34 compares results from studies of mice and rats in which comparable tissue-
specific endpoints were examined in in vivo genotoxicity assays. Several of the endpoints that
were positive in mice (e.g., sister chromatid exchange, DNA-protein cross-links, comet assay)
have not been examined in the rat. Unscheduled DNA synthesis has been demonstrated in
mouse but not in rat hepatocytes. In contrast to the positive results seen in mouse inhalation
exposure studies, DNA SSB induction was not seen in rat inhalation studies but was seen in an
gavage study.
186 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-34. Comparison of in vivo dichloromethane genotoxicity assays targeted to lung or liver cells, by species
Assay
DNA
synthesis
Unscheduled
DNA
synthesis
Unscheduled
DNA
synthesis
Sister
chromatid
exchange
Chromosome
aberrations
DNA-protein
cross-links
DNA SSBs
by alkaline
elution
DNA SSBs
by alkaline
elution
Studies in BeCSF^ice
Test system
Liver
Hepatocytes
Lung cells
Lung cells
Liver and
lung cells
Hepatocytes
Liver and
lung
homogenate
Route, dose (duration)
Gavage, 1,000 mg/kg;
inhalation, 4,000 ppm
(2hrs)
Inhalation, 2,000 and
4,000 ppm
(2 or 6 hrs)
Inhalation 6 hr/d, 5 d/wk,
0, 4,000, 8,000 ppm
(2 wks)
Inhalation 6 hr/d, 5 d/wk,
0,2,000 ppm (12 wks)
Inhalation, 6 hr/d, 5 d/wk,
0, 4,000, 8,000 ppm
(2 wks)
Inhalation, 6 hr/d, 3 d,
4,000 ppm (3 d)
Inhalation, 6 hr/d, 150,
500, 1,500, 3,000,
4,000 ppm (3 d)
Inhalation, 2,000 and
4,000 ppm (3 or 6 hrs)
Liver: inhalation, 2,000,
4,000, 6,000, 8,000 ppm
(3 hrs)
Lung: inhalation, 1,000,
2,000, 4,000, 6,000 ppm
(3 hrs)
Results
Negative in oral and
inhalation studies
Negative
Positive at
8,000 ppm
Positive at
2,000 ppm
Positive at
8,000 ppm
Positive in liver
4,000 ppm
Positive in liver at
500-4,000 ppm;
both studies negative
in lung
Positive at
4,000 ppm
Liver: Positive at
4,000-8,000 ppm
Lung: Positive at
2,000-4,000 ppm
Reference
Lefevre and
Ashby
(1989)
Trueman and
Ashby
(1987)
Allen et al.
(1990)
Allen et al.
(1990)
Casanova et
al. (1992)
Graves et al.
(1994b)
Graves et al.
(1995)
Studies in rats
Test system
Hepatocytes
Hepatocytes
Hepatocytes
Liver and
lung
homogenate
Route, dose (duration)
Inhalation, 2,000 and
4,000 ppm (2 or 6 hrs)
Intraperitoneal,
400 mg/kg
Inhalation, 3 or 6 hrs,
2,000 and 4,000 ppm
Liver: inhalation,
4,000, 5,000 ppm
Lung: inhalation,
4,000 ppm
Results
Negative
Negative
Negative at all
concentrations
and time points
Negative in
liver and lung at
all
concentrations
and time points
Reference
No studies
Trueman and
Ashby (1987)
Mirsalis et al.
(1989)
No studies
No studies
No studies
Graves et al.
(1994b)
Graves et al.
(1995)
(Table 4-34 continues on next page)
187
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-34. Comparison of in vivo dichloromethane genotoxicity assays targeted to lung or liver cells, by species
Assay
DNA SSBs
by alkaline
elution
DNA damage
by comet
assay
DNA adducts
Studies in BeCSF^ice
Test system
Liver and
lung cells
Liver and
kidney cells
Route, dose (duration)
Gavage, 1,720 mg/kg;
organs harvested at 0
(control), 3, and 24 hrs
Intraperitoneal, 5 mg/kg
Results
Positive only at
24 hrs after dosing
Negative
Reference
Sasaki et al.
(1998)
Watanabe et
al. (2007)
Studies in rats
Test system
Liver
homogenate
Liver and
kidney cells
Route, dose (duration)
Gavage, 425 mg/kg and
1,275 mg/kg
Intraperitoneal, 5 mg/kg
Results
Positive at
1,275 mg/kg
Negative
Reference
Kitchin and
Brown (1989)
No studies
Watanabe et
al. (2007)
188
DRAFT - DO NOT CITE OR QUOTE
-------
In summary, the available data provide evidence for mutagenicity of dichloromethane.
Most of the in vitro bacterial assays showed positive results when there was GST activity.
Nonpositive results were reported only in bacterial assays with low GST activity; in experiments
where GST was added, positive results were then observed. Evaluation of the in vitro
mammalian studies also demonstrates the influence of GST activity on the observation of
genotoxic effects. In rat and hamster cell lines where GST activity is significantly less than
mouse, primarily negative results were reported following dichloromethane exposure. However,
when mouse liver cytosol or transfected mouse GST were included in these same cell lines,
mutagenic effects were reported after dichloromethane exposure. In mouse cell lines, positive
results were obtained in Clara cells, but no effects were observed in a mouse lymphoma cell line,
which is consistent with the absence of tumors in this site for mice. In vitro studies using human
cells reported effects of dichloromethane on frequency of sister chromatid exchange,
chromosomal aberrations and DNA damage (comet assay), but no effects on unscheduled DNA
synthesis, DNA SSBs, or DNA-protein cross-link. The results of in vivo mutagenicity in mice
also provide support for the site-specificity of the observed tumors. Assays using mouse bone
marrow were all negative. However, micronuclei and sister chromatid exchange tests in
peripheral blood produced a positive response at high doses. With the exception of one study of
unscheduled DNA synthesis in hepatocytes, numerous site-specific studies in either the liver or
lung were also positive at various doses. These liver and lung studies included chromosomal
aberrations, SSBs, sister chromatid exchanges, and DNA-protein cross-links and correspond to
genotoxic and mutagenic effects associated with metabolites from the GST pathway.
4.5.2. Mechanistic Studies of Liver Effects
One of the major target organs from dichloromethane exposure is the liver, and several
studies have focused on examining the potential mechanisms producing liver tumors. This
section summarizes the primary mechanistic studies that were conducted in order to examine the
hepatic tumors produced by dichloromethane in mice. A parallel set of studies, discussed in the
next section, focus on potential mechanisms that produce lung tumors. Briefly,
dichloromethane-induced liver tumors first appeared in mice after 52 weeks of exposure
(Maronpot et al., 1995; Kari et al., 1993), which was when tumors began to appear in control
mice, indicating a similar time course in tumor formation between treated and untreated groups.
Onset of liver tumor formation is not preceded by liver cell proliferation (Casanova et al., 1996;
Foley et al., 1993). Further mechanistic studies were conducted to assay the tumor for
significant changes in proto-oncogene activation and tumor suppressor gene inactivation
(Maronpot et al., 1995; Devereux et al., 1993; Hegi et al., 1993). A second subset of mechanistic
studies was conducted to elucidate the reason that mice are the most sensitive species to liver
tumors and if other species exhibited changes in liver function (Thier et al., 1998; Reitz et al.,
1989). It was found that mice have the highest level of GST-T1 catalytic activity but that
189 DRAFT - DO NOT CITE OR QUOTE
-------
humans, rats, and hamsters, among other species, also metabolize dichloromethane in the liver to
a GST conjugate. In contrast, there has been little research focusing on the mechanisms through
which noncancer hepatic effects (seen most strongly in the rat) are produced, and the role of the
parent material, metabolites of the CYP2E1 pathway, metabolites of the GST pathway, or some
combination of parent material and metabolites is not known.
Liver tumor characterization studies. Several studies have examined the time course of
appearance of liver tumors in B6C3Fi mice exposed to 2,000 or 4,000 ppm and possible links
between hepatic nonneoplastic cytotoxicity, enhanced hepatic cell proliferation, and the
development of liver tumors (Casanova et al., 1996; Maronpot et al., 1995; Foley et al., 1993;
Kari et al., 1993). The studies provide no clear evidence for a sustained liver cell proliferation
response to dichloromethane that can be linked to the development of dichloromethane-induced
liver tumors. Additionally, a few studies have examined if dichloromethane-induced liver
tumors are the result of proto-oncogene activation and tumor suppressor gene inactivation
(Maronpot et al., 1995; Devereux et al., 1993; Hegi et al., 1993).
Kari et al. (1993) (also summarized by Maronpot et al. [1995]) reported data from six
groups of 68 female B6C3Fi mice exposed to six "stop-exposure" protocols of differing
durations and sequences, with each exposure concentration standardized at 2,000 ppm for
6 hours/day, 5 days/week. The six stop-exposure protocols were 26 weeks of exposure followed
by 78 weeks without exposure, 78 weeks without exposure followed by 26 weeks of exposure,
52 weeks without exposure followed by 52 weeks with exposure, 52 weeks of exposure followed
by 52 weeks without exposure, 78 weeks of exposure followed by 26 weeks without exposure,
and 26 weeks without exposure followed by 78 weeks of exposure. A control group (no
exposure, 104 weeks duration) and a maximum exposure (104 weeks duration) group were also
included. Exposure for 26 weeks did not result in an increased incidence of liver tumors
(adenomas or carcinomas). Respective percentages of animals with liver tumors were
27 (18/67), 40 (27/67), and 34% (23/67) for the controls, early 26-week exposure, and late
26-week exposure groups, respectively. Exposure to 2,000 ppm for 52 weeks (followed by no
exposure until 104 weeks), 78 weeks (either early or late exposure periods), or 104 weeks
produced increased incidences of mice with liver tumors (p < 0.05), but this increase was not
seen in the 52-week late exposure group. Respective percentages of animals with liver tumors
(adenomas and carcinomas combined) were 44 (28/64), 31 (21/67), 62 (42/68), 48 (32/67), and
69% (47/68) for the 52 (early exposure), 52 (late exposure), 78 (early exposure), 78 (late
exposure), and 104 week exposure periods, respectively. With the 78 week exposures, the
difference in the liver tumor incidence between the early and late exposure periods was
statistically significant (p < 0.01). A greater increase in multiplicity of liver tumors was also
seen with the early 78-week exposure period. These data suggest that 52 weeks of exposure was
190 DRAFT - DO NOT CITE OR QUOTE
-------
required to increase the incidence of liver tumors in mice, that early exposure was more effective
than late exposure, and that the increased risk continued after cessation of exposure.
Histopathologic examination of liver tissue at interim killings at eight time periods (13,
26, 52, 68, 75, 78, 83, or 91 weeks) of exposure to 2,000 ppm (n = 20 mice per killing) found no
evidence of nonneoplastic cytotoxicity that preceded the appearance of proliferative neoplastic
liver lesions. Incidences of mice with liver adenomas or carcinomas were elevated (between
40 and 60%) at five of the six interim killings after 52 weeks. The incidence rates at each time
period were 0/20 (0%) at 13 weeks, 1/20 (5%) at 26 weeks, 8/20 (40%) at 52 weeks, 4/26 (15%)
at 68 weeks, 13/20 (65%) at 75 weeks, 12/19 (63%) at 78 weeks, 8/20 (40%) at 83 weeks, and
20/30 (66%) at 91 weeks. The collected liver lesion data identify no exposure-related increased
incidence of nonneoplastic liver lesions that could be temporally linked to liver tumor
development. Liver tumors first appeared at about the same time in control and exposed animals
(52 weeks).
Foley et al. (1993) examined indices of cell proliferation in livers of female B6C3Fi mice
exposed to 1,000, 2,000, 4,000, or 8,000 ppm dichloromethane (6 hours/day, 5 days/week) for 1,
2, 3, or 4 weeks or to 2,000 ppm for 13, 26, 52, or 78 weeks but found no evidence for sustained
cell proliferation with prolonged exposure to dichloromethane. To label liver cells in S phase,
tritiated thymidine (1- to 4-week exposure protocols) or bromodeoxyuridine (13- to 78-week
protocols) was administered subcutaneously via an osmotic mini-pump for 6 days prior to
killing. Labeled hepatocytes in liver sections (from 10 mice in each exposure/duration group)
were counted to assess the number of cells in S-phase per 1,000 cells. S-phase labeling indices
in livers of exposed mice at most killings were equivalent to or less than those in control mice.
A transient increase in S-phase labeling index of about two- to fivefold over controls was
observed at the 2-week killing of mice exposed to 1,000, 4,000, or 8,000 ppm. Because of the
transient nature and small magnitude of the response, it is not expected to be of significance to
the promotion of liver tumors in chronically exposed mice. Foley et al. (1993) also compared
cell proliferation labeling indices in foci of cellular alteration and nonaffected liver regions in
control and exposed mice but found no significant difference between control and exposed mice.
S-phase labeling was accomplished by immunohistologic staining for proliferating cell nuclear
antigen in liver sections from 24 control mice and 15 exposed mice, with livers showing foci of
cellular alteration. In both control and exposed livers, the labeling index was about four- to
fivefold higher in foci of cellular alteration than in surrounding unaffected liver tissue.
In mice exposed to 2,000 ppm for 13-78 weeks, relative liver weights were statistically
significantly elevated compared with controls; about 10% increased at 13 and 26 weeks and
about 30-40% increased at 52 and 78 weeks. Histologic changes in liver sections of 2,000 ppm
mice exposed for 13-78 weeks were restricted to hepatocellular hypertrophy (observed at all
killing intervals) and preneoplastic (foci of cellular alteration) and neoplastic (adenoma and
carcinoma) lesions. No signs of liver tissue degeneration were found. Adenoma and focus of
191 DRAFT - DO NOT CITE OR QUOTE
-------
alteration were first detected at 26 weeks (2/10 versus 0/10 in controls). At 52 weeks,
4/10 exposed mice had proliferative lesions (one focus, one adenoma, and two carcinomas),
compared with 1/10 in controls (one adenoma). At 78 weeks, 7/10 exposed mice had
proliferative lesions (two foci, three adenomas, six carcinomas) compared with 1/10 in controls
(one adenoma). In summary, the results indicate that inhalation exposure to 2,000 ppm
dichloromethane produced an increased incidence of liver tumors in female B6C3Fi mice. No
evidence was found for sustained cell proliferation or liver tissue degeneration in response to
dichloromethane exposure, but exposure was associated with relative liver weight increases and
hepatocellular hypertrophy.
Casanova et al. (1996) found no clear evidence of increased cell proliferation in the livers
of male B6C3Fi mice exposed to dichloromethane concentrations >1,500 ppm 6 hours/day for
3 days. Three or four groups of three mice were exposed to 146, 498, 1,553, or 3,923 ppm
unlabeled dichloromethane for 2 days and then exposed to [14C]-labeled dichloromethane for
6 hours on the third day. Radiolabel incorporated into liver DNA deoxyribonucleosides was
measured as an index of cell proliferation. Radiolabel incorporated into liver DNA
deoxyribonucleosides increased approximately fivefold from 146 to 1,553 ppm, but further
increases were not apparent at 3,923 ppm. (In contrast, as described in Section 4.5.3, radiolabel
incorporation into lung DNA deoxyribonucleosides displayed a 27-fold increase over this
concentration range.) The small magnitude of the increase in radiolabel incorporation into liver
DNA deoxyribonucleosides with increasing exposure concentration suggests that little if any
enhanced cell proliferation occurred in the liver in response to dichloromethane exposure.
Devereux et al. (1993) (also reported in Maronpot et al. [1995]) analyzed liver tumors in
female B6C3Fi mice for the presence of activated H-ras oncogenes. Fifty dichloromethane-
induced and 49 spontaneous liver tumors were screened for H-ras mutations. There was a
relatively high frequency of activated H-ras among the nonexposed B6C3Fi mice: 67% of the
spontaneous tumors and 76% of the dichloromethane-induced tumors contained mutations in the
H-ras gene. Overall, the mutation profile of the dichlorom ethane-induced tumors did not
significantly differ from the spontaneous tumors.
Similarly, Hegi et al. (1993) analyzed the liver tumors from female B6C3Fi mice for
inactivation of the tumor suppressor genes, p53 and Rb-1. Half of the liver tumors used for this
study had an activated H-ras oncogene. Twenty liver tumors (15 carcinomas and 5 adenomas)
were screened for loss of heterozygosity (LOH) on chromosome 11 and 14, which is associated
with malignant conversion of thep53 gene (chromosome 11) and the Rb-1 gene
(chromosome 14). Only one tumor out of 20 contained a LOH at chromosome 14, and no
dichloromethane-induced liver tumors contained a LOH at chromosome 11.
Liver metabolic studies. As described in detail in Section 3.3, GST-T1 enzymatic activity
and distribution is variable among species, and there is also considerable intraspecies variability
192 DRAFT - DO NOT CITE OR QUOTE
-------
among humans. In summary, Reitz et al. (1989) demonstrated a greater metabolic activity with
respect to dichloromethane in livers of B6C3Fi mice compared with F344 rats, Syrian golden
hamsters, and humans. The rates of in vitro metabolism by the GST pathway were about 4-, 12-,
and 20-fold greater in B6C3Fi mouse liver samples than in F344 rat, human, and Syrian golden
hamster samples, respectively (Reitz et al., 1989). A more recent study characterized the
dichloromethane metabolic capacity specifically of hepatic GST-T1 (Thier et al., 1998).
Enzymatic activities of GST-T1 in liver from F344 rats, B6C3Fi mice, Syrian golden hamsters,
and humans with three different GST-T1 phenotypes (nonconjugators, low conjugators, high
conjugators) showed the following order with dichloromethane as a substrate: mouse » rat >
human high conjugators > human low conjugators > hamster > human nonconjugators.
4.5.3. Mechanistic Studies of Lung Effects
The finding of increased lung tumors in B6C3Fi mice exposed to dichloromethane
(Mennear et al., 1988; NTP, 1986) has stimulated a number of studies designed to examine the
mechanism for dichloromethane-induced lung tumors in this animal. The lung tumor mechanism
studies were conducted with B6C3Fi mice, and the frequency of lung tumors in control animals
was very low. Time-course studies for lung tumor development were conducted, and it was
found that the onset of lung tumor development was much shorter than liver tumors (Kari et al.,
1993) (reported in Maronpot et al., 1995). As a result, it is hypothesized that a potential common
mechanism independent of liver metabolism is producing tumors in the lung. As with the liver
tumors, there were no significant increases in mutations for the K-ras oncogene (Devereux et al.,
1993) or thep53 and Rb-1 tumor suppressor genes (Hegi et al., 1993). Additionally, the Clara
cells, which are nonciliary secretory cells found in the primary bronchioles of the lung, are
selectively targeted after dichloromethane exposure. Acute dichloromethane exposure produces
Clara cell vacuolization, which is not sustained with long-term exposure (Foster et al., 1992).
There is a correlation between the acute effects on the Clara cell and the lung tumors from
chronic exposure to dichloromethane (Kari et al., 1993). However, the exact mechanism for
producing these lung effects is not completely understood. Provided below is a summary of the
studies examining the potential mechanisms for producing lung tumors resulting from
dichloromethane exposure.
Lung tumor characterization studies. Kari et al. (1993) (also summarized in Maronpot et
al. [1995]) demonstrated that only 26 weeks of exposure to 2,000 ppm was necessary to produce
significantly increased incidences of lung tumors in female B6C3Fi mice. In the six "stop-
exposure" protocol experiments (26 weeks exposure followed by 78 weeks without exposure,
78 weeks without exposure followed by 26 weeks exposure, 52 weeks without exposure
followed by 52 weeks with exposure, 52 weeks exposure followed by 52 weeks without
exposure, 78 weeks exposure followed by 26 weeks without exposure, and 26 weeks without
193 DRAFT - DO NOT CITE OR QUOTE
-------
exposure followed by 78 weeks exposure), early but not late exposure for 26 or 52 weeks
resulted in an increased incidence of animals with lung tumors (adenoma or carcinomas).
Respective percentages of animals with lung tumors were 7.5 (5/67), 31 (21/68), 4 (3/67),
63 (40/63), and 15% (10/67) for the controls, early 26-, late 26-, early 52-, and late 52-week
exposure groups, respectively. With the 78-week exposures, both the early and late exposure
regimens produced an increased incidence of lung tumors compared with controls (56 [38/68]
and 19% [13/68], respectively), compared with the incidence of 63% (42/67) seen in the group
exposed for the full 104 weeks. Thus a plateauing of risk was seen, with similar incidence rates
seen with the early 52-week, early 78-week, and 104-week exposure periods. The difference in
the lung tumor incidence between the early and late exposure periods of similar duration was
statistically significant (p < 0.01) for the 26-, 52-, and 78-week duration protocols. A greater
increase in multiplicity of lung tumors was also seen with the early 78-week exposure period.
As with the liver tumor data from the same series of experiments, these data suggest that early
exposure was more effective than late exposure and that the increased risk continued after
cessation of exposure.
Histopathologic examination of lung tissue from mice killed at 13, 26, 52, 68, 75, 78, 83,
or 91 weeks of exposure to 2,000 ppm (n = 20 mice per killing) found no evidence of
nonneoplastic cytotoxicity that preceded the appearance of proliferative neoplastic lung lesions.
In contrast, incidences of mice with lung adenomas or carcinomas (combined) were elevated at
interim killings >52 weeks; incidences for the interim killings of mice exposed to 2,000 ppm
(6 hours/day, 5 days/week) between 13 and 91 weeks were 0/20 (0%) at 13 weeks, 0/20 (0%) at
26 weeks, 6/20 (30%) at 52 weeks, 6/26 (23%) at 68 weeks, 8/20 (40%) at 75 weeks, 9/19 (47%)
at 78 weeks, 10/20 (50%) at 83 weeks, and 14/30 (47%) at 91 weeks. Lung hyperplasia was
found at an increased incidence only at 91 weeks, well after the 26- and 52-week periods that
induced increased incidences of mice with lung tumors.
Kanno et al. (1993) found no evidence for histologic changes or increased cell
proliferation in lung tissue of female B6C3Fi mice exposed to 2,000 or 8,000 ppm
dichloromethane for 1,2, 3, or 4 weeks compared with control mice, or in mice exposed to
2,000 ppm for 13 or 26 weeks. Osmotic mini-pumps were used to deliver tritiated thymidine and
label cells undergoing replicative DNA synthesis over 6-day periods before killing. Labeled
cells undergoing rapid DNA synthesis and cell proliferation were assessed in sections of
proximal and terminal bronchioles and alveoli of lungs from groups of 5 mice exposed for 1-
4 weeks or 10 mice exposed for 13 or 26 weeks. There were no exposure-related histopathologic
or labeling index changes in the alveoli, but lower labeling indices were found in the bronchiolar
epithelium of exposed mice compared with controls.
The combined results from the Kari et al. (1993) and Kanno et al. (1993) studies are
consistent with the hypothesis that dichloromethane-induced lung tumors in B6C3Fi mice are not
preceded by overt cytotoxicity, enhanced and sustained cell proliferation, or hyperplasia in the
194 DRAFT - DO NOT CITE OR QUOTE
-------
lung. Two other studies (Casanova et al., 1996; Foster et al., 1992), however, have reported
evidence for enhanced cell proliferation in lungs of B6C3Fi mice exposed for acute durations to
airborne dichlorom ethane. Only one of these studies (Foster et al., 1992), however, looked for
sustained cell proliferation in the lung with prolonged exposure. In agreement with the results
from Kanno et al. (1993), no evidence was found for sustained cell proliferation in lungs with
prolonged exposure to dichloromethane at concentrations demonstrated to induce lung tumors in
mice.
Casanova et al. (1996) detected evidence of increased cell proliferation in the lungs of
male B6C3Fi mice exposed to dichloromethane concentrations >1,500 ppm 6 hours/day for
3 days. Three or four groups of three mice were exposed to 146, 498, 1,553, or 3,923 ppm
unlabeled dichloromethane for 2 days and then exposed to [14C]-labeled dichloromethane for
6 hours on the third day. Radiolabel incorporated into lung DNA deoxyribonucleosides (after
removal of DNA-protein cross-links containing radiolabeled formaldehyde) was measured as an
index of cell proliferation. Radiolabel incorporation into lung DNA deoxyribonucleosides
increased with increasing exposure concentration, with the amount increasing by about 27-fold
between 146 and 3,923 ppm. In hamsters that did not develop tumors in response to chronic
inhalation exposure to 3,500 ppm dichloromethane (Burek et al., 1984), no evidence for
enhanced radiolabel incorporation into lung DNA deoxyribonucleosides was found following
acute exposure (Casanova et al., 1996).
Devereux et al. (1993) (also summarized in Maronpot et al. [1995]) analyzed lung tumors
in female B6C3Fi mice for the presence of activated K-ras oncogenes. Fifty-four
dichlorom ethane-induced and 17 spontaneous lung tumors (7 from the NTP [1986] study and
10 from a study in C57BL/6 x C34F1 mice reported by Candrian et al. [1991]) were screened for
K-ras mutations. Twenty percent of the dichloromethane-induced tumors and 24% of the
spontaneous tumors contained mutations in the K-ras gene. Devereux et al. (1993) stated that
there may be a significant difference in the incidence of K-ras activation between spontaneous
and dichloromethane-induced tumors. However, the small number of the spontaneous tumors
that were available for the study limits the conclusions that can be made from the results.
Hegi et al. (1993) analyzed the lung tumors from female B6C3Fi mice for inactivation of
the tumor suppressor genes, p53 and Rb-1. Forty-nine dichloromethane-induced lung
carcinomas, five lung adenomas, and seven spontaneous lung carcinomas were screened for
LOH on mouse chromosome 11 and 14, which is associated with malignant conversion of the
p53 gene (chromosome 11) and the Rb-1 gene (chromosome 14). Fourteen percent (n = 7) of the
dichloromethane-induced lung carcinomas exhibited LOH at chromosome 11. No/?53 mutations
were detected in the seven spontaneous lung tumors or the five dichloromethane-induced lung
adenomas. Only three dichloromethane-induced tumors exhibited LOH at chromosome 14. The
authors noted that inactivation of thep53 and Rb-1 tumor suppressor genes infrequently occur in
lung and liver tumors.
195 DRAFT - DO NOT CITE OR QUOTE
-------
Clara cell studies. Foster et al. (1992) found enhanced cell proliferation in bronchiolar
cells and, to a lesser degree, alveolar cells in the lungs of male B6C3Fi mice exposed for acute
durations (2, 5, 8, or 9 days) to 4,000 ppm dichloromethane (6 hours/day, 5 days/week), but the
response was less distinct after subchronic durations of exposure (89, 92, or 93 days). To
measure cell proliferation, mice (n = 5 per exposure-duration group) were given subcutaneous
doses of tritiated thymidine for 5 consecutive days prior to killing. Labeled cells in bronchiolar
or alveolar epithelium in lung sections were counted to assess the number of cells in S phase per
1,000 cells. Counts of bronchiolar epithelium cells in S phase in exposed mice sacrificed on
days 2, 5, 8, and 9 were approximately 2-, 15-, 3-, and 5-fold higher, respectively, than those of
unexposed mice at day 0 of the experiment. In exposed mice sacrificed on days 89, 92, and 93,
less than twofold increases in bronchiolar epithelium cell labeling were observed. Increased cell
labeling was found in alveolar epithelium only on day 8 (about seven- to eightfold increase) and
day 9 (about fourfold increase). Vacuolation of the Clara cells of the bronchiolar epithelium was
observed on day 2 (scored as ++, majority of cells affected), day 9 (+++, virtually all the cells
affected), and day 44 (+, moderate effect to cells) but was not apparent on days 5, 8, 40, 43, 89,
92, or 93. No hyperplasia of the bronchiolar epithelium or changes to Type II alveolar epithelial
cells were observed in the lungs of any of the exposed mice at any time point. The appearance
and disappearance of the Clara cell vacuolation was generally correlated with the appearance and
disappearance of enhanced cell proliferation in the bronchiolar epithelium; enhanced cell
proliferation was observed on days 2, 5, 8, and 9 (along with appearance of Clara cell
vacuolation on days 2 and 9) but was not observed on days 89, 92, and 93 when Clara cell
lesions also were not observed. This suggests that cell proliferation was enhanced in response to
Clara cell damage but was not sustained with repeated exposure to dichloromethane.
Currently, a mechanistic connection has not been established between the acute effects of
dichloromethane on Clara cells in the bronchiolar epithelium and the development of lung
tumors in B6C3Fi mice exposed by inhalation to concentrations >2,000 ppm dichloromethane
for 2 years (NTP, 1986) or for 26 weeks followed by no exposure through 2 years (Maronpot et
al., 1995; Kari et al., 1993). There is a concordance between exposure concentrations inducing
acute Clara cell vacuolation (>2,000 ppm) and those inducing lung tumors (>2,000 ppm).
However, transient acute Clara cell vacuolation does not appear to progress to necrosis or lead to
sustained cell proliferation (which could promote the growth of tumor-initiated cells) and
appears to be dependent on CYP metabolism of dichloromethane (see the following paragraphs
discussing pertinent findings reported by Foster et al. [1994, 1992]). In contrast, there is
consistent and specific evidence for an association between the formation of DNA-reactive GST-
pathway metabolites and the formation of lung and liver tumors from inhalation exposure (see
Sections 4.5.2 and 4.7.3).
196 DRAFT - DO NOT CITE OR QUOTE
-------
Foster et al. (1992) noted that the appearance and disappearance of Clara cell vacuolation
in mouse lungs showed concordance with temporal patterns for immunologic staining for
CYP2B1 and CYP2B2 levels in lung sections. A similar temporal pattern was reported for
CYP2B1 and CYP2B2 monooxygenase activities (ethoxycoumarin O-dealkylation or aldrin
epoxidation) assayed in lung microsomes. When there was a marked decrease in CYP2B1 and
CYP2B2 staining (e.g., on day 5) or monooxygenase activities, the lesion was not present.
Similarly, the appearance of the lesion was preceded (the day before) by the recovery of
monooxygenase activities or immunologic staining close to control levels. These patterns
suggested to Foster et al. (1992) that Clara cells may have developed tolerance to
dichloromethane due to inactivation of a CYP isozyme.
In subsequent studies, increased percentages of vacuolated bronchiolar epithelium cells
were noted in mice exposed to 2,000 ppm (26.3 ± 6.7%) or 4,000 ppm (64.8 ± 12.8%), but
vacuolated cells were not observed in bronchi olar epithelium of lung sections from control mice
or mice exposed to 125, 250, 500, or 1,000 ppm (Foster et al., 1994). Pretreatment with the CYP
inhibitor, piperonyl butoxide, counteracted the 2,000 ppm effect (2.4 ± 3.6% vacuolated cells),
whereas GSH-depleted mice showed no statistically significant change in percentage of
vacuolated cells (32.7 ± 16.9%) compared with the mean percentage in mice exposed to
2,000 ppm without pretreatment. No consistent, statistically significant, exposure-related
changes were found in cytosolic GST metabolic activities (with dichloromethane as substrate) or
microsomal CYP monooxygenase activities (ethoxycoumarin O-dealkylation), but mean
cytosolic levels of nonprotein sulfhydryl compounds were elevated in lungs of mice exposed to
1,000 and 2,000 ppm (134.6 ± 17.1 and 146.4 ± 6.7 nmol/mg protein, respectively) compared
with control levels (109.5 ± 7.6 nmol/mg protein). Increased cell proliferation was found in
cultured Clara cells isolated from 4,000 ppm mice compared with nonexposed mice; respective
values for percentage of cell sin S phase were 18.97± 1.18 and 2.02 ± 0.86% (Foster et al.,
1994).
Results from the studies by Foster et al. (1994, 1992) indicate that 6-hour exposures of
B6C3Fi mice to dichloromethane concentrations >2,000 ppm caused transient Clara cell
vacuolation in the bronchiolar epithelium, which was not consistently observed following
repeated exposures. With repeated exposure to 4,000 ppm, the Clara cell vacuolation did not
progress to necrosis, and no hyperplasia of the bronchiolar epithelium was found after up to
13 weeks of exposure. The transient Clara cell vacuolation was decreased by CYP inhibition
with piperonyl butoxide and was unaffected by GSH depletion, indicating that a CYP metabolite
was involved. Clara cell vacuolation was not found after five consecutive, daily 6-hour
exposures to 4,000 ppm but reappeared after 2 days without exposure followed by two additional
consecutive, daily exposures (day 9). With repeated exposure, the lesion was detected at a
diminished severity on day 44 (but was not found on day 40 or 43) and on day 93 (but was not
found on day 89 or 92). The temporal pattern of Clara cell vacuolation with repeated exposure
197 DRAFT - DO NOT CITE OR QUOTE
-------
was reflected in the occurrence of transiently decreased CYP metabolic activity after the
appearance of vacuolation. Foster et al. (1994, 1992) proposed that the diminishment of severity
or the disappearance of the Clara cell vacuolation with repeated exposure was due to the
development of a tolerance to dichloromethane, linked with a decrease of CYP metabolism of
di chl or om ethane.
4.5.4. Mechanistic Studies of Neurological Effects
Several neurobehavioral studies (see Section 4.4.3 for a complete summary) have
demonstrated that dichloromethane exposure results in decreased spontaneous motor activity
with pronounced lethargy at high concentrations (> 1,000 ppm). These effects, combined with
the observation that dichloromethane impairs learning and memory (Alexeef and Kilgore, 1983)
and affects production of evoked responses to sensory stimuli (Herr and Boyes, 1997; Rebert et
al., 1989), indicate that dichloromethane produces significant neurological effects. The
mechanisms behind producing these changes have been examined by measuring changes in
neurotransmitter levels and changes in neurotransmitter localization. Specific brain regions (e.g.,
hippocampus, caudate nucleus, cerebellum) were analyzed to determine if dichloromethane-
induced behavioral effects, such as learning and memory (hippocampus, caudate nucleus) and
movement (cerebellum), are resulting from pathological changes in these regions. Changes in
neurotransmitter levels were also monitored to see if there was any correlation in behavior and
neurochemical changes. Summaries of these studies are provided below. It is not yet known if
dichloromethane directly interacts with neuronal receptors, as has been demonstrated for toluene
and ethanol, two other solvents with neurobehavioral and neurophysiological profiles that are
similar to those of dichloromethane (for a review see Bowen et al. [2006]).
Kanada et al. (1994) examined the effect of dichloromethane on acetylcholine and
catecholamines (dopamine, norepinephrine, serotonin) and their metabolites in the midbrain,
hypothalamus, hippocampus, and medulla from male Sprague-Dawley rats (four to five per
group). The rats were sacrificed 2 hours after a single gavage dose of 0 or 534 mg/kg of
undiluted dichloromethane. Administration of dichloromethane significantly increased the
concentration of acetylcholine in the hippocampus and increased dopamine and serotonin levels
in the medulla. Dichloromethane decreased norepinephrine levels in the midbrain, and
hypothalamus and serotonin levels were decreased in the hypothalamus. There was a trend
toward decreased dopamine in the hypothalamus, but the variability between the animals was so
high that the effect was not significant. The authors speculated that increased acetylcholine
release from dichloromethane administration may be due to decreased acetylcholine release from
the nerve terminals. It is unclear as to how these neurochemical changes could be correlated to
the neurobehavioral changes observed after dichloromethane exposure.
In a 2-week exposure study, male Wistar rats were exposed to dichloromethane at 500 or
1,000 ppm (6 hours/day, 5 days/week for 1 or 2 weeks) or 1,000 ppm TWA (1 hour at 100 ppm,
198 DRAFT - DO NOT CITE OR QUOTE
-------
1 hour peak at 2,800 ppm, 1 hour at 100 ppm, repeated immediately, 5 days/week for 1 or
2 weeks) (Savolainen et al., 1981). Brains were removed from rats at the end of the study and
analyzed. The 1,000 ppm TWA group displayed increases in cerebral RNA. Other changes
noted for this group in the cerebrum included significant increases in NADPH-diaphorase and
succinate dehydrogenase activity. These changes suggest increased neural activity to possibly
offset the overall inhibitory effect of dichloromethane in the CNS. It could also possibly explain
why lethargic effects decrease with continued dichloromethane exposure, and this result
demonstrates a neuroprotective mechanism resulting from dichloromethane exposure. After a
7-day withdrawal, RNA levels in the cerebrum were significantly lower in the 1,000 ppm group.
Succinate dehydrogenase levels remained lowered in the 1,000 ppm TWA group after the 7-day
exposure-free period.
Changes in brain catecholamine levels after a subacute exposure to dichloromethane were
evaluated using male Sprague-Dawley rats (Fuxe et al., 1984). Rats were exposed to 70, 300,
and 1,000 ppm dichloromethane 6 hours/day for 3 consecutive days. At all exposures, there was
a significant decrease of catecholamine concentrations in the posterior periventricular region of
the hypothalamus. The impact of dichloromethane was also evaluated on the hypothalamic-
pituitary gonadal axis. The hypothalamus regulates secretion of reproductive hormones such as
follicle-stimulating hormone and luteinizing hormone. The levels of the hormone release were
not significantly changed with dichloromethane exposure. In the caudate nucleus, which is
involved in memory processes, the catecholamine level initially increased (at 70 ppm) and then
was lower (1,000 ppm) in comparison to the control. The study demonstrates significant changes
in catecholamine levels in the hypothalamus and caudate nucleus. Catecholamine level changes
in the hypothalamus did not have any significant effect on hormonal release, but decreased
catecholamine levels in the caudate nucleus at higher exposures may lead to memory and
learning impairment.
A series of studies were conducted in male and female Mongolian gerbils exposed
continuously to >210 ppm dichloromethane for 3 months, followed by a 4-month exposure-free
period (Karlsson et al., 1987; Briving et al., 1986; Rosengren et al., 1986). Decreased DNA
concentrations were noted in the hippocampus at both the 210 and 350 ppm exposures. At
350 ppm, there was also decreased DNA concentration in the cerebellar hemispheres, indicating
a decreased cell density in these regions probably due to cell loss (Rosengren et al., 1986).
These findings indicate that the cerebellum, which is the section of the brain that regulates motor
control, is a target for dichloromethane. In the same study, increased astroglial proteins were
found in the frontal and sensory motor cerebral cortex, which directly correlated to the
astrogliosis that was observed in those areas. Up-regulation of these astroglial proteins is a good
indicator of neuronal injury (Rosengren et al., 1986).
Karlsson et al. (1987) measured DNA concentrations in different regions of the gerbil
brain. The total brain protein concentration per wet weight was not significantly different
199 DRAFT - DO NOT CITE OR QUOTE
-------
between dichloromethane-exposed and control animals. However, DNA concentrations per wet
weight were significantly decreased in the hippocampus after dichloromethane exposure. No
other examined regions demonstrated significant changes in DNA concentrations after
dichloromethane exposure. Therefore, this result indicates that the hippocampus, which plays a
role in the formation of new memories, is another target for dichloromethane in the CNS. This
selective DNA concentration decrease observed in the hippocampus is a sign of neurotoxicity as
noted by the authors and may possibly explain why some studies have noted memory and
learning deficits with dichloromethane exposure.
At a 210 ppm exposure, Driving et al. (1986) observed that dichloromethane decreased
glutamate, y-aminobutyric acid, and phosphoethanolamine levels in the frontal cortex, while
glutamate and y-aminobutyric acid were increased in the posterior cerebellar vermis. Increased
levels of glutamate in the posterior cerebellar vermis could reflect an activation of astrocytic glia,
since glutamine synthetase is localized exclusively in astrocytes.
4.6. SYNTHESIS OF MAJOR NONCANCER EFFECTS
4.6.1. Oral
4.6.1.1. Summary of Human Data
Human studies involving oral exposure to dichloromethane are limited to case reports of
neurological impairment, liver and kidney effects (as severe as organ failure), and
gastrointestinal irritation in individuals who ingested amounts ranging from about 25 to 300 mL
(Chang et al., 1999; Hughes and Tracey, 1993). Neurological effects with these individuals
consisted of general CNS depressive symptoms, such as drowsiness, confusion, headache, and
dizziness. Hemoglobinuria has been noted as a kidney effect associated with ingestions.
4.6.1.2. Summary of Animal Data
Acute oral or intraperitoneal administration of dichloromethane in animals has resulted in
several significant effects. General activity and function were affected as evidenced by
decreased neuromuscular activity (Moser et al., 1995). Additionally, decreased sensorimotor
function was detected through measurement of evoked potentials (Herr and Boyes, 1997) and by
using the FOB (Moser et al., 1995). Neurochemical changes (e.g., acetylcholine, dopamine,
norepinephrine, serotonin) were detected 2 hours after oral dosage of dichloromethane within
specific parts of the brain. It should be noted that all the acute effects that were observed after
oral or intraperitoneal administration occurred within 5 hours after dosage. No other significant
organ effects were noted after a single acute oral exposure, but in oral pharmacokinetic studies, it
is known that dichloromethane is primarily distributed to the liver, lungs, and kidneys (Angelo et
al., 1986a).
Results from short-term, subchronic, and chronic oral toxicity studies in laboratory
animals are summarized in Table 4-35. The data indicate that rats may be more sensitive than
200 DRAFT - DO NOT CITE OR QUOTE
-------
mice to nonneoplastic liver effects from orally administered dichloromethane, as evidenced by
lower NOAELs and LOAELs with more severe liver effects in rats. The most frequently
observed liver effect was hepatocyte vacuolation, seen with drinking water exposure (90 days) in
F344 rats at >166 mg/kg-day and B6C3Fi mice at 586 mg/kg-day (Kirschman et al., 1986) and
with gavage exposure (14 days) in CD-I mice at 333 mg/kg-day (Condie et al., 1983).
Hepatocyte degeneration or necrosis was observed in female F344 rats exposed by drinking
water for 90 days to 1,469 mg/kg-day (Kirschman et al., 1986) and in female F344 rats exposed
by gavage for 14 days to 337 mg/kg-day (Berman et al., 1995) but was not seen in a 90-day
drinking water study in B6C3Fi mice exposed to doses as high as 2,030 mg/kg-day (Kirschman
et al., 1986). In the chronic-duration (2-year) study, liver effects were described as foci and
areas of alteration in F344 rats exposed to drinking water doses between 50 and 250 mg/kg-day;
an increased incidence of fatty changes in the liver was also noted but the incidence was not
provided (Serota et al., 1986a). These effects were considered to be nonneoplastic for several
reasons. Serota et al. (1986b) observed a dose-related increased incidence of 0, 65, 92, 97, 98,
and 100% in male rats and 51, 41, 73, 89, 91, and 85% in female rats for the 0, 5, 50, 125, 250,
and 250 mg/kg-day with recovery groups, respectively. Evidence for liver tumors has been
reported in female rats only. Specifically, evidence for liver tumors in rats includes a small
number of hepatocellular carcinomas observed in female rats at 50 and 250 mg/kg-day, which
reached statistical significance (for trend and for individual pairwise comparisons) only with the
combined grouping of neoplastic nodules and hepatocellular carcinomas. In male rats, only one
hepatocellular carcinoma was observed in all of the exposure groups (compared to 4 in the
controls), and the incidence of neoplastic nodules and hepatocellular carcinomas was higher in
controls (16%) than in any exposure group (16, 3, 0, 6, 5, and 13% for the 0, 5, 50,125, 250
mg/kg-day, and 250 mg/kg-day with recovery groups, respectively). The authors (Serota et al.,
1986a) did not elaborate on the characterization of the altered foci. However, the
characterization of altered foci could range from a focal change in fat distribution (nonneoplastic
effect) to enzyme altered foci which are generally considered a precursor to tumor formation
(Goodman et al., 1994). Serota et al. (1986a) reported an increased incidence of fatty change in
the liver at doses of >50 mg/kg-day, but the incidence was not reported. In addition, a 90-day
study (Kirschman et al., 1986) demonstrated that increased fatty deposits were present in the
hepatocyte vacuoles. Therefore, the altered foci (i.e., change in fat distribution) observed by
Serota et al. (1986b) may represent a precursor to fatty liver changes which is considered a
nonneoplastic effect. Taken together, the data support the conclusion that the altered foci were
nonneoplastic.
201 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-35. NOAELs and LOAELs in selected animal studies involving oral exposure to dichloromethane for
short-term, subchronic, or chronic durations
Type of effect and
exposure, reference
Hepatic, 14-d gavage
Herman etal. (1995)
Condieetal. (1983)
Species and exposure details
F344 rat, female, 8/dose group
0, 34, 101, 337, 1,012 mg/kg-d
CD-I mouse, male, 5/group for histological
examinations; 8/group for blood urea
nitrogen, serum creatinine, and serum
glutamate-pyruvate transaminase; 0, 133,
333, 665 mg/kg-d
Results
Hepatocyte necrosis
Hepatocyte vacuolation (minimal to mild in 3/5)
NOAEL
LOAEL
(mg/kg-d)
101
133
337
333
Hepatic, 90-d drinking water
Kirschman et al.
(1986)
Kirschman et al.
(1986)
F344 rat, male and female; 15/sex/group;
males 0, 166, 420, 1,200 mg/kg-d
females 0, 209, 607, 1,469 mg/kg-d
B6C3FJ mouse, male and female,
males 0, 226, 587, 1,911 mg/kg-d
females 0, 231, 586, 2,030 mg/kg-d
Hepatic vacuolation (generalized, centrilobular, or periportal, at
lowest dose, in 10/15 males and 13/15 females compared with
1/15 males and 6/15 females in controls)
Hepatic vacuolation (increased severity of centrilobular fatty
change in mid- and high-dose groups compared with controls)
Not
identified
231
166
586
Hepatic, 104-wk drinking water
Serotaetal. (1986a)
Serotaetal. (1986b);
Hazleton Laboratories
(1983)
F344 rat, male and female,
0, 5, 50, 125, 250 mg/kg-d
B6C3FJ mouse, male and female,
0, 60, 125, 185, 250 mg/kg-d
Liver foci/areas of alteration (considered nonneoplastic histologic
changes); fatty liver changes also seen at same doses but
incidence data not reported; no evidence that increased altered
foci progresses to liver tumor formation
Some evidence of fatty liver; marginal increase in the Oil Red-O-
positive material in the liver
5
185
50
250
Neurologic, 14 d
Moseretal. (1995)
F344 rat, female,
0, 34, 101, 337, 1,012 mg/kg-d
FOB 24 hrs postexposure: altered autonomic, neuromuscular, and
sensorimotor and excitability measures
101
337
(Table 4-35 continues on next page)
202
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-35. NOAELs and LOAELs in selected animal studies involving oral exposure to dichloromethane for
short-term, subchronic, or chronic durations
Type of effect and
exposure, reference
Species and exposure details
Results
NOAEL LOAEL
(mg/kg-d)
Reproductive
General Electric
Company (1976)
Rajeetal. (1988)
Charles River CD rat, male and female,
gavage for 90 d before mating (10 d between
last exposure and mating period); 0, 25, 75,
225 mg/kg-d; Fl offspring received same
treatment as parents for 90 d
Swiss-Webster mouse, male, 0, 250,
500 mg/kg (subcutaneous injection), 3 x per
wk, 4 wks prior to mating with nonexposed
females (1 wk between last exposure and
mating period)
Reproductive performance of FO and histologic examination of
tissues fromFl offspring
No statistically significant effects ontestes, number of litters, live
fetuses/litter, percent dead fetuses/litter, percent resorbed/litter, or
fertility index
225
500
Not
identified
Not
identified
Developmental
Narotsky and Kavlock
(1995)
F344 rat, pregnant female, gavage on GDs 6-
19; 0, 338, 450 mg/kg-d
Maternal: weight gain depression
Fetal: no effects on pup survival, resorptions, pup weight
338
450
450
Not
identified
203
DRAFT - DO NOT CITE OR QUOTE
-------
The NOAEL and LOAEL, 101 and 337 mg/kg-day, respectively, for altered neurological
functions in female F344 rats (as reported by Moser et al. [1995]) were identical to those
reported by Berman et al. (1995) for hepatocyte necrosis in female F344 rats. In the 90-day
(Kirschman et al., 1986) and 104-week (Serota et al., 1986a, b) drinking water studies, no
obvious clinical signs of neurological impairment were observed in rats or mice at exposure
levels that induced liver effects (see Table 4-35), but these studies did not include standardized
neurological testing batteries.
Results from the available studies do not provide evidence for effects on reproductive or
developmental endpoints (Table 4-35). No effects on pup survival, resorptions, or pup weight
were found following exposure of pregnant F344 rats to doses as high as 450 mg/kg-day on GDs
6-19, a dose that depressed maternal weight gain (Narotsky and Kavlock, 1995), and no effects
on reproductive performance endpoints (fertility index, number of pups per litter, pup survival)
were found in studies in male and female Charles River CD rats (General Electric Company,
1976) and in male Swiss-Webster mice (Raje et al., 1988). There are no oral exposure studies
focusing on neurobehavioral effects or other developmental outcomes.
4.6.2. Inhalation
4.6.2.1. Summary of Human Data
As discussed in Section 4.1.2, acute inhalation exposure of humans to dichloromethane
has been associated with cardiovascular impairments due to decreased oxygen availability from
COFIb formation and neurological impairment from interaction of dichloromethane with nervous
system membranes. Results from studies of acutely exposed human subjects indicate that acute
neurobehavioral deficits measured, for example, by psychomotor tasks, tests of hand-eye
coordination, visual evoked response changes, and auditory vigilance, may occur at
concentrations >200 ppm with 4-8 hours of exposure (Bos et al., 2006; ACGIH, 2001; ATSDR,
2000; Cherry et al., 1983; Putz et al., 1979; Gamberale et al., 1975; Winneke, 1974).
The clinical and workplace studies of noncancer health effects of chronic
dichloromethane exposure have examined markers of disease and specific clinical endpoints
relating to cardiac disease, neurological disease, hepatic function, and reproductive health. As
summarized in Section 4.1.2.9, the limited available data do not provide evidence of cardiac
damage related to dichloromethane exposure in occupationally exposed workers (Hearne and
Pifer, 1999; Tomenson et al., 1997; Gibbs et al., 1996; Lanes et al., 1993; Ott et al., 1983d;
Cherry et al., 1981). Limitations in these studies that would result in a reduced ability to detect
cardiovascular effects include the presence of the healthy worker effect in these worker cohorts,
and the absence of data pertaining to workers who died before the establishment of the analytic
cohort (Gibbs et al., 1996; Gibbs, 1992).
Relatively little is known about the long-term neurological effects of chronic exposures,
although there are studies that provide some evidence of an increased prevalence of neurological
204 DRAFT - DO NOT CITE OR QUOTE
-------
symptoms among workers with average exposures of 75-100 ppm (Cherry et al., 1981), long-
term effects on some neurological measures (i.e., possible detriments in attention and reaction
time in complex tasks) in workers whose past exposures were in the 100-200 ppm range (Lash et
al., 1991), and an increased risk of suicide in worker cohort studies (Hearne and Pifer, 1999;
Gibbs, 1992). Given the suggestions from these studies and their limitations (particularly with
respect to sample size and power considerations), the statement that there are no long-term
neurological effects of chronic exposures to dichloromethane cannot be made with confidence.
With respect to markers of hepatic damage, three studies measured serum enzyme and
bilirubin levels in workers exposed to dichloromethane (Soden, 1993; Kolodner et al., 1990; Ott
et al., 1983c). There is some evidence of increasing levels of serum bilirubin with increasing
dichloromethane exposure (Kolodner et al., 1990; Ott et al., 1983c), but there are no consistent
patterns with respect to the other hepatic enzymes examined (serum y-glutamyl transferase,
serum AST, serum ALT). Thus, to the extent that this damage could be detected by these
serologic measures, these studies do not provide clear evidence of hepatic damage in
dichloromethane exposed workers.
Only limited and somewhat indirect evidence pertaining to immune-related and infectious
disease effects of dichloromethane in humans is available. No risk of the broad category of
infection- and parasite-related mortality was reported by Hearne and Pifer (1999), but there was
some evidence of an increased risk of influenza and pneumonia-related mortality at two cellulose
triacetate fiber production work sites in Maryland and South Carolina (Gibbs, 1992), and an
increased risk of bronchitis-related mortality, based on only four exposed cases, was seen in
another cohort study (Radican et al., 2008).
Few studies have been conducted pertaining to reproductive effects (i.e., spontaneous
abortion, low birth weight, or oligospermia) of dichloromethane exposure from workplace
settings (Wells et al., 1989; Kelly, 1988; Taskinen et al., 1986) or environmental settings (Bell et
al., 1991). Of these, the data pertaining to spontaneous abortion provide the strongest evidence
of an adverse effect of dichloromethane exposure. The limitations of the only study pertaining to
this outcome (Taskinen et al., 1986), however, do not allow firm conclusions to be made
regarding dichloromethane and risk of spontaneous abortion in humans.
4.6.2.2. Summary of Animal Studies
Acute and short-term (up to 7 days) inhalational exposure to dichloromethane in animals
has resulted in neurological and hepatocellular changes. Several neurological-mediated
parameters were reported, including decreased activity (Kjellstrand et al., 1985; Weinstein et al.,
1972; Heppel and Neal, 1944), impairment of learning and memory (Alexeef and Kilgore, 1983),
and changes in responses to sensory stimuli (Rebert et al., 1989). Although learning and
memory properties were impaired in one acute exposure (47,000 ppm until loss of righting
reflex), it should be noted that this effect has not been characterized by using other learning and
205 DRAFT - DO NOT CITE OR QUOTE
-------
memory tasks nor any other exposure paradigms. In a 3-day exposure to dichloromethane (70,
300, or 1,000 ppm 6 hours/day), it was found that in the rat brain, there were changes in
catecholamine (dopamine, serotonin, norepinephrine) in the hypothalamus and caudate nucleus
(Fuxe et al., 1984). The catecholamine level changes did not affect hormonal release which is a
primary function of the hypothalamus.
Another acute exposure study examined immunological response reported an increase in
Streptoccocal pneumonia-related mortality and decrease in bactericidal activity of pulmonary
macrophages in CD-I mice following a single 3-hour exposure to dichlorom ethane at 100 ppm
(Aranyi et al., 1986). No effects were seen at 50 ppm. A 4-week inhalation exposure to 5,000
ppm dichloromethane in rats did not result in changes in immune response as measured by the
sheep red blood cell assay (Warbrick et al., 2003). These studies suggest a localized, portal-of-
entry effect within the lung without evidence of systemic immunosuppression.
Mouse hepatocytes showed balloon degeneration (dissociation of polyribosomes and
swelling of rough endoplasmic reticulum) within 12 hours of exposure to 5,000 ppm (Weinstein
et al., 1972). A subacute exposure in Wistar rats to 500 ppm dichloromethane 6 hours/day for
6 days resulted in increased hemochrome content in liver microsomal CYP (Savolainen et al.,
1977).
Results pertaining to liver, lung, and neurological effects from longer (>7 days)
subchronic and chronic inhalation toxicity studies in laboratory animals are summarized in
Table 4-36; reproductive and developmental studies are summarized in Table 4-37.
206 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-36. NOAELs and LOAELs in animal studies involving inhalation exposure to dichloromethane for
subchronic or chronic durations, hepatic, pulmonary, and neurologic effects
Type of effect and
exposure period,
reference
Species and exposure details
Results
NOAEL
LOAEL
ppm
Hepatic, subchronic (13-14 wks)
Haunetal. (1971)
Beagle, female (n = 8);
Rhesus monkeys, female (n = 4);
Sprague-Dawley rat, male (n = 20);
ICR mouse, female (n = 380)
0, 1,000, 5,000 ppm (continuous exposure;
14 wks)
Fatty liver at 1,000 ppm in dogs: "borderline" liver
changes in monkey at 5,000 ppm; mottled liver
changes in rats at 5,000 ppm; hepatocytes degeneration
at 5,000 ppm in mice, no information about liver
effects in mice at 1,000 ppm; decreased movement and
lethargy at 1,000 ppm in dogs, mice, and monkey
Not identified
(dog)
Not identified
(monkey)
1,000 (rat)
Not identified
Mouse)
1,000 (dog)
5,000 (monkey)
5,000 (rat)
5,000 (mouse)
Haunetal. (1972)
Beagle (n = 16);
Rhesus monkey (n = 4);
Sprague-Dawley rat (n = 20),
ICR mouse (n = 20)
0, 25, 100 ppm (continuous exposure;
14 wks)
Increased hepatic cytoplasmic vacuolation and
decreased CYP levels in liver microsomes in mice at
100 ppm; increased fatty liver content at 25, 100 ppm
in rats
100 (dog)
100 (monkey)
Not identified
(rat)
25 (mouse)
Not identified
(dog)
Not identified
(monkey)
25 (rat)
100 (mouse)
Leuscheretal. (1984)
Sprague-Dawley rat, male and female,
(20/sex/group) - 0, 1,000 ppm (6 hrs/d;
90 d);
Beagle, male and female (3/sex/group) •
0, 5,000 ppm
No liver effects noted
1,000 (rat)
5,000 (dog)
Not identified (rat)
Not identified
(dog)
NTP (1986)
F344/N rat, male and female (10/sex/group)
0, 525, 1,050, 2,100, 4,200, 8,400 ppm
(6 hrs/d, 5 d/wk, 13 wks)
Decreased lipid:liver weight ratios at 4,200 (females);
8,400 (males); decreased BW by 23 and 11% in males
and females at 8,400 ppm compared with controls; one
male and one female died at 8,400 ppm before the end
of the study
4,200
8,400
NTP (1986)
B6C3FJ mouse, male and female
(10/sex/group)
0, 525, 1,050, 2,100, 4,200, 8,400 ppm
(6 hrs/d, 5 d/wk, 13 wks)
Hepatocyte centrilobular degeneration at
4,200 females) and 8,400 (males); decreased lipid:liver
weight ratios at 8,400 (females); at 8,400 ppm,
4/10 males and 2/10 females died before end of study
2,100
4,200
(Table 4-36 continues on next page)
207
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-36. NOAELs and LOAELs in animal studies involving inhalation exposure to dichloromethane for
subchronic or chronic durations, hepatic, pulmonary, and neurologic effects
Type of effect and
exposure period,
reference
Species and exposure details
Results
NOAEL
LOAEL
ppm
Hepatic, 2 yrs (6 hrs/d, 5 d/wk)
Mennearetal. (1988);
NTP (1986)
Mennearetal. (1988);
NTP (1986)
Bureketal. (1984)
Bureketal. (1984)
Nitschkeetal. (1988a)
F344/N rat, male and female
0, 1,000, 2,000, 4,000 ppm
B6C3F! mouse, male and female
0, 2,000, 4,000 ppm
Syrian golden hamster, male and female
0, 500, 1,500, 3,500 ppm
Sprague-Dawley rat, male and female
0, 500, 1,500, 3,500 ppm
Sprague-Dawley rat, male and female
0, 50, 200, 500 ppm
Hepatocyte vacuolation and necrosis
Hemosiderosis in liver
Renal tubular degeneration
Hepatocyte degeneration
Renal tubule casts
No effects on histologic, clinical chemistry, urinalytic,
and hematologic variables no obvious clinical signs of
toxicity
Hepatocyte vacuolation (males and females)
Hepatocyte necrosis (males only), no obvious clinical
signs of toxicity)
Hepatocyte vacuolation significantly increased in
females; nonsignificant increase in males at 500 ppm
(3 1% in controls and 40% in 500 ppm group)
Not identified
Not identified
1,000
Not identified
Not identified
3,500
Not identified
500
200
1,000
1,000
2,000
2,000
2,000
Not identified
500
1,500
500
Pulmonary, 13 wks (6 hrs/d, 5 d/wk)
NTP (1986)
NTP (1986)
Foster etal. (1992)
F344 rat, male and female
0, 525, 1,050, 2,100, 4,200, 8,400 ppm
B6C3F! mouse, male and female
(10/sex/group)
0, 525, 1,050, 2,100, 4,200, 8,400 ppm
B6C3FJ mouse, male and female
0, 4,000 ppm
Foreign body pneumonia
No nonneoplastic pulmonary lesions
Clara cell vacuolation
4,200
8,400
Not identified
8,400
Not identified
4,000
Neurological, 14 d
Savolainen et al.
(1981)
Wistar rat, male
500, 1,000, 1,000 TWA (100 + 2,800 1-hr
peaks3) ppm (6 hrs/d, 5 d/wk, 2 wks)
Increased RNA in cerebrum at 1,000 ppm; increased
enzymatic activities'3 in cerebrum and cerebellum at
1,000 ppm TWA
500
1,000 for brain
RNA
concentration;
1,000 TWA for
brain enzymatic
activity
(Table 4-36 continues on next page)
208
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-36. NOAELs and LOAELs in animal studies involving inhalation exposure to dichloromethane for
subchronic or chronic durations, hepatic, pulmonary, and neurologic effects
Type of effect and
exposure period,
reference
Species and exposure details
Results
NOAEL
LOAEL
ppm
Neurological, 13-14 wks
Mattssonetal. (1990)
Haunetal. (1971)
Karlssonetal. (1987)
Brivingetal. (1986)
Rosengren et al.
(1986)
Thomas etal. (1972)
F344 rat, male and female
0, 50, 200, 2,000 ppm
(6 hrs/d, 5 d/wk)
Beagle dog (female);
Rhesus monkey (female);
Sprague-Dawley rat (male);
ICR mouse (female)
0, 1,000, 5,000 ppm
(continuous exposure)
Mongolian gerbil, male and female
210, 350, 700 ppm (continuous exposure,
followed by 4 mo exposure-free period)
ICR mouse, female
0, 25, 100 ppm, continuous
No exposure-related effects on an observational
battery, hind-limb grip strength, a battery of evoked
potentials, or histology of brain, spinal cord, peripheral
nerves; measured 64 hrs postexposure
Clinical signs (incoordination, lethargy) of CNS
depression most evident in dogs
Astrogliosis in frontal and sensory motor cerebral
cortex suggested by increases in astroglial proteins;
cell loss in cerebellar regions; decreased DNA in
hippocampus; neurochemical changes observed at all
exposures
Increased spontaneous activity observed at 25 ppm but
not 100 ppm
2,000
Not identified
(dog)
Not identified
(monkey)
1,000 (rat)
Not identified
(mouse)
Not identified
Not identified
Not identified
1,000 (dog)
1,000 (monkey)
5,000 (rat)
1,000 (mouse)
210
25
CoHb, 13-14 wks
Haunetal. (1972)
Beagle (n = 16);
Rhesus monkey (n = 4);
Sprague-Dawley rat (n = 20),
ICR mouse (n = 20)
0, 25, 100 ppm (continuous exposure;
14 wks)
CoHb levels significantly higher at 25, 100 ppm for
monkeys and 100 ppm for beagles
Not identified
25
(Table 4-36 continues on next page)
209
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-36. NOAELs and LOAELs in animal studies involving inhalation exposure to dichloromethane for
subchronic or chronic durations, hepatic, pulmonary, and neurologic effects
Type of effect and
exposure period,
reference
Species and exposure details
Results
NOAEL
LOAEL
ppm
COHb, 2 yrs (6 hrs/d, 5 d/wk)
Bureketal. (1984)
Bureketal. (1984)
Nitschkeetal. (1988a)
Syrian golden hamster, male and female
0, 500, 1,500, 3,500 ppm
Sprague-Dawley rat, male and female
0, 500, 1,500, 3,500 ppm
Sprague-Dawley rat, male and female
0, 50, 200, 500 ppm
About 30% COHb in each exposed group
About 12-14% COHb in each exposed group
COHb values at 2 yrs: about 2, 7, 13, 14%
Not identified
Not identified
Not identified
500
500
500
Equivalent to 1,000 ppm TWA.
'Decreased GSH, y-aminobutyric acid, and phosphoethanolamine in frontal cortex; GSH and y-aminobutyric acid increased in posterior cerebellar vermis.
210
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-37. NOAELs and LOAELs in selected animal studies involving inhalation exposure to dichloromethane,
reproductive and developmental effects
Type of effect and exposure
period, reference
Species and exposure details
Results
NOAEL
LOAEL
ppm
Reproductive
Nitschke et al. (1988b)
Mennear et al. (1988); NTP
(1986)
Rajeetal. (1988)
F344 rat, male and female, FO: 6 hr/d,
5 d/wk for 14 wk before mating and GDs 0 to
21; Fl: 6 hr/d, 5 d/wk, beginning PND 4 for
17 wk before mating; 0, 100, 500, 1,500 ppm
B6C3FJ mouse; 0, 2,000 or 4,000 ppm,
6 hrs/d, 5 d/wk for 2 yrs
Swiss-Webster mouse, male, 2 hr/d, 5 d/wk
for 6 wk before mating with nonexposed
females; 0, 100, 150, 200 ppm
No statistically significant effects on fertility or
litter size, neonatal survival, growth rates, or
histopathologic lesions inFl orF2 weanlings
Testicular atrophy
Ovarian atrophy (considered secondary to hepatic
effects)
No statistically significant effects on testes,
number of litters, live fetuses/litter, percent dead
fetuses/litter, percent resorbed/litter
Fertility index was lower in 150 and 200 ppm
groups (80%) compared with controls and
100 ppm groups (95%) (statistical significance
depends on test used)
1,500
2,000
Not identified
200
100
Not identified
4,000
2,000
Not identified
150
Developmental
Schwetzetal. (1975)
Schwetzetal. (1975)
Swiss-Webster mouse, pregnant female,
7 hr/d, GDs 6-15; 0, 1,250 ppm
Sprague-Dawley rat, pregnant female, 7 hr/d,
GDs 6-15; 0, 1,250 ppm
Maternal effects: 9-10% COHb; increased
absolute, not relative, liver weight, increased
maternal weight (11-15%).
Fetal effects: increased litters with extra center of
ossification in sternum
Maternal: 9-10% COHb; increased absolute, not
relative, liver weight
Fetal: increased incidence of delayed ossification
of sternebrae
Not identified
1,250
Not identified
1,250
1,250
Not identified
1,250
Not identified
Other developmental
Bornschein et al. (1980);
Hardin and Manson (1980)
Long-Evans rat, female, 6 hr/d for 12-14 d
before breeding and GDs 1-17; 6 hr/d on
GDs 1-15; 0, 4,500 ppm
Maternal (both protocols): increased absolute and
relative liver weight (-10%)
Fetal/offspring: decreased fetal BW (-10%);
changed behavioral habituation to novel
environments; no changes in gross, skeletal, or
soft-tissue anomalies
Not identified
Not identified
4,500
4,500
211
DRAFT - DO NOT CITE OR QUOTE
-------
Hepatic centrilobular degeneration was observed in several studies containing different
species and inhalational exposures. This effect was observed in guinea pigs exposed to
5,000 ppm (7 hours/day) for 6 months (Heppel et al., 1944). Monkeys, rats, and mice
continuously exposed (24 hours/day) to 5,000 ppm dichloromethane for 14 weeks also had
increased centrilobular degeneration (Haun et al., 1972, 1971). This effect was also observed at
lower exposures when mice were exposed to 4,200 ppm (6 hours/day) for 13 weeks (NTP, 1986)
and in dogs exposed to 1,000 ppm (24 hours/day) for up to 14 weeks (Haun et al., 1972, 1971).
Increased incidences of histologic hepatic lesions were not found in F344 rats exposed to
4,200 or 8,400 ppm (6 hours/day) for 13 weeks (NTP, 1986) or in Sprague-Dawley rats exposed
to 10,000 ppm (6 hours/day) for 90 days (Leuschner et al., 1984). Hepatic lesions were also not
observed in beagle dogs exposed to 5,000 ppm (6 hours/day) for 90 days (Leuschner et al., 1984)
or in dogs, monkeys, rats, and mice exposed to 25 or 100 ppm (24 hours/day) for up to 14 weeks
(Haun et al., 1972). Heppel et al. (1944) also demonstrated absence of hepatic lesions in
unspecified strains of monkeys, rabbits, and rats exposed to 10,000 ppm (4 hours/day) for up to
8 weeks and in unspecified strains of dogs, rabbits, and rats exposed to 5,000 ppm (7 hours/day)
for up to 6 months.
Gross neurological impairments were observed in several laboratory species with
repeated exposure to 10,000 ppm 4 hours/day for 8 weeks (Heppel et al., 1944) or to 1,000 or
5,000 ppm 24 hours/day for 14 weeks (Haun et al., 1972, 1971). Dogs exposed to 5,000 ppm
6 hours/day for 90 days showed slight sedation during exposures, but Sprague-Dawley rats
exposed to 10,000 ppm for 90 days did not (Leuschner et al., 1984). In F344 rats exposed to
concentrations up to 2,000 ppm 6 hours/day for 13 weeks, no effects were observed on an
observational battery, hind-limb grip strength, a battery of evoked potentials, or histology of the
brain, spinal cord, or peripheral nerves; these tests were conducted beginning >65 hours after the
last exposure (Mattsson et al., 1990).
Exposure-related nonneoplastic effects on the lungs reported in the subchronic studies
were restricted to foreign body pneumonia in rats exposed to 8,400 ppm 6 hours/day for
13 weeks (NTP, 1986), Clara cell vacuolation in mice exposed to 4,000 ppm 6 hours/day for
13 weeks (Foster et al., 1992), and pulmonary congestion in guinea pigs exposed to 5,000 ppm
7 hours/day for 6 months (Heppel et al., 1944).
The chronic duration inhalation studies were conducted at lower exposure levels than the
short-term and subchronic studies and provide results indicating that the liver is the most
sensitive target for noncancer toxicity in rats and mice (Table 4-36). Life-time exposure was
associated with hepatocyte vacuolation and necrosis in F344 rats exposed to 1,000 ppm
6 hours/day (Mennear et al., 1988; NTP, 1986), hepatocyte vacuolation in Sprague-Dawley rats
exposed to 500 ppm 6 hours/day (Nitschke et al., 1988a; Burek et al., 1984), and hepatocyte
degeneration in B6C3Fi mice exposed to 2,000 ppm 6 hours/day (lower concentrations were not
tested in mice) (Mennear et al., 1988; NTP, 1986). As shown in Tables 4-36 and 4-37, other
212 DRAFT - DO NOT CITE OR QUOTE
-------
effects observed include renal tubular degenerations in F344 rats and B6C3Fi mice at 2,000 ppm,
testicular atrophy in B6C3Fi mice at 4,000 ppm, and ovarian atrophy in B6C3Fi mice at
2,000 ppm (considered secondary to hepatic effects). No exposure-related increased incidences
of nonneoplastic lung lesions were found in any of the chronic studies (Table 4-36).
In comparison to rats and mice, Syrian golden hamsters are less sensitive to the chronic
inhalation toxicity of dichloromethane. No exposure-related changes were found in
comprehensive sets of histologic, clinical chemistry, urinalytic, and hematologic variables
measured in hamsters exposed for 2 years to 500, 1,500, or 3,500 ppm for 6 hours/day, with the
exception that mean COHb percentages were about 30% in each of these groups compared with
a mean value of about 3% for the controls (Burek et al., 1984).
The reproductive and developmental studies are limiting in terms of the exposure
regimen used. Nitschke et al. (1988b) used a noncontinuous exposure period (i.e., exposure of
dams before mating and on GDs 0-21, and beginning again on PND 4), and two of the
developmental studies using only a single, relatively high daily exposure over the gestational
period (1,250 ppm, GDs 6-15 in Schwetz et al. [1975] and 4,500 ppm, GDs 1-17 in Bornschein
[1980] and Hardin and Manson [1980]). No significant effects on reproductive performance
variables were found in a two-generation reproduction assay with F344 rats exposed to
concentrations as high as 1,500 ppm (Nitschke et al., 1988b). No effects on most of the
measures of reproductive performance were observed in male mice exposed to 200 ppm
2 hours/day for 6 weeks before mating to nonexposed females. In the study by Raje et al.
(1988), fertility index was reduced in the 150 and 200 ppm groups, but the statistical significance
of this effect varied considerably depending on the statistical test used in this analysis. No
adverse effects on fetal development were found following exposure of pregnant Swiss-Webster
mice or Sprague-Dawley rats to 1,250 ppm 6 hours/day on GDs 6-15 (Schwetz et al., 1975).
Following exposure of female Long-Evans rats to 4,500 ppm (6 hours/day) for 14 days before
breeding plus during gestation or during gestation alone, a 10% decrease in fetal BW and
changed behavioral habituation of the offspring to novel environments were seen (Bornschein et
al., 1980; Hardin and Manson, 1980). No exposure-related changes in gross, skeletal, or soft-
tissue anomalies were found.
4.6.3. Mode of Action Information
4.6.3.1. Mode of Action for Nonneoplastic Liver Effects
Studies of chronically exposed rats, both by the oral route and the inhalation route,
identified liver changes as the most sensitive exposure-related noncancer effect associated with
exposure to dichloromethane (Tables 4-35 to 4-37). The liver changes included increased
incidence of liver foci/areas of alteration and hepatocyte vacuolation in rats and degenerative
liver effects in rats, guinea pigs, monkeys, and mice.
213 DRAFT - DO NOT CITE OR QUOTE
-------
The mode of action by which dichloromethane induces these nonneoplastic hepatic
effects is not known. The determination of whether these effects are due to the parent material,
metabolites of the CYP2E1 pathway, metabolites of the GST pathway, or some combination of
parent material and metabolites has not been elucidated. The available data indicate that rats
may be more sensitive than mice to the noncancer hepatotoxicity, but a mechanistic explanation
of this possible interspecies difference is not currently available.
4.6.3.2. Mode of Action for Nonneoplastic Lung Effects
Single 6-hour inhalation exposures to concentrations >2,000 ppm dichloromethane
produced a transient vacuolation of Clara cells in the bronchiolar epithelium of B6C3Fi mice.
Vacuolization of the Clara cells disappeared or was diminished with repeated exposure and was
correlated with subsequent transient diminishment of CYP metabolic activity. CYP inhibition
with piperonyl butoxide counteracted the vacuolation observed in the Clara cells (Foster et al.,
1994, 1992). With repeated exposure to 4,000 ppm (up to 13 weeks), the Clara cell vacuolation
did not appear to progress to necrosis, and no hyperplasia of the bronchiolar epithelium was
found. Foster et al. (1994, 1992) proposed that the diminished severity or disappearance of Clara
cell vacuolation with repeated exposure was due to the development of tolerance to
dichloromethane, linked with a transient decrease of CYP metabolism of dichloromethane. The
available data suggests that CYP metabolism of dichloromethane may be involved in the mode
of action for the acute effects of dichloromethane on the bronchiolar epithelium of mice.
Mode of action research attention on lung effects from chronic exposure to
dichloromethane has focused on neoplastic effect; nonneoplastic lung effects have received
relatively little attention. No exposure-related increased incidences of nonneoplastic lung lesions
(including epithelial hyperplasia) were found in any of the chronic studies listed in Table 4-36,
but chronic inhalation exposure of B6C3Fi mice to concentrations >2,000 ppm has consistently
been shown to induce lung tumors in several studies (Kari et al., 1993; NTP, 1986). In a study
that included interim sacrifices at 13, 26, 52, 68, 75, 78, 83, and 91 weeks of B6C3Fi mice
exposed to 2,000 ppm, hyperplasia of lung epithelium (the only nonneoplastic lung lesion found)
was found in only three of the eight interim sacrifices (68, 78, and 91 weeks) and was only
statistically significantly elevated at 91 weeks (5/30 versus 0/15 in controls) (Kari et al., 1993).
4.6.3.3. Mode of Action for Neurological Effects
Results from studies of acutely exposed human subjects indicate that mild
neurobehavioral deficits may occur at air concentrations >200 ppm with 4-8 hours of exposure
(Bos et al., 2006; ACGffl, 2001; ATSDR, 2000; Cherry et al., 1983; Putz et al., 1979; Gamberale
et al., 1975; Winneke, 1974). Acute high-dose exposures also resulted in gross neurological
impairments in several laboratory species (Haun et al., 1972, 1971; Heppel et al., 1944).
Exposure of F344 rats to concentrations up to 2,000 ppm 6 hours/day for 13 weeks produced no
214 DRAFT - DO NOT CITE OR QUOTE
-------
effects on an observational battery, hind-limb grip strength, a battery of evoked potentials, or
histology of the brain, spinal cord, or peripheral nerves (Mattsson et al., 1990). However, oral
exposures have been shown to alter autonomic, neuromuscular, and sensorimotor functions in
F344 rats exposed to gavage doses >337 mg/kg-day for 14 days (Moser et al., 1995).
Dichloromethane may be metabolized by the CYP2E1 enzyme to CO (Guengerich, 1997;
Hashmi et al., 1994; Gargas et al., 1986). Many of the acute human exposure studies evaluated if
CO was the primary metabolite responsible for producing the CNS depressant effects observed
during dichloromethane exposure. Overall, at lower exposures and acute durations, it appears
that CO may be a major mediator of the neurobehavioral effects, although unmetabolized
dichloromethane may contribute to these effects. Putz et al. (1979) demonstrated that similar
neurobehavioral deficits were present in humans when an equivalent COHb blood level (and CO
exposure) was achieved between CO and dichloromethane exposures; however after a longer
duration, neurobehavioral deficits are more pronounced with dichloromethane exposure in
comparison to CO exposure alone. This additional increase in the CNS depressive effects is
most likely due to the saturation of the CYP2E1 metabolic pathway. CYP2E1 pathway
saturation with dichloromethane has also been noted in hamsters (Burek et al., 1984) and rats
(Nitschke et al., 1988a; McKenna et al., 1982). It is probable that initially CYP2E1 is
metabolizing dichloromethane to CO, which results in neurological effects. At higher
concentrations and for longer durations, however, the CYP2E1 pathway is most likely saturated.
The resulting effects would thus be due to either the remaining dichloromethane compound or to
metabolites generated through the GST pathway.
Neurological changes associated with COHb levels have primarily represented
neurobehavioral endpoints. Neurophysiological and neurochemical changes could also be
related to dichloromethane, CO, or potentially other metabolites. Based on the available
literature on other solvents, such as toluene and perchloroethylene (for a review see Bowen et al.
[2006]), it can be hypothesized that dichloromethane or a metabolite may interact directly with
excitatory and inhibitory receptors such as the NMD A, GAB A, dopamine, and serotonin
receptors, among other targets, to produce the resulting neurobehavioral effects. This hypothesis
is supported by the evidence that changes in relation to dichloromethane exposures in glutamate,
GAB A, dopamine, serotonin, acetylcholine, and other neurotransmitters are found in the brain
(Kanada et al., 1994; Briving et al., 1986; Fuxe et al., 1984). Additionally, several
neurobehavioral effects such as decreased spontaneous motor activity, deficits in learning and
memory, and deficits in FOB parameters are similar to other more characterized solvents such as
toluene. However, more comprehensive studies specifically designed to determine the mode of
action for dichloromethane-induced impairment of neurological functions have not been
conducted.
215 DRAFT - DO NOT CITE OR QUOTE
-------
4.6.3.4. Mode of Action for Neurodevelopmental Effects
The mode of action for neurodevelopmental effects observed in Bornschein et al. (1980)
and Hardin and Manson (1980) can be hypothesized to involve dichloromethane or the
production of CO from the CYP2E1 pathway. The placental transfer of dichloromethane has
been demonstrated with inhalation exposure (Withey and Karpinski, 1985; Anders and Sunram,
1982). CO is a known developmental neurotoxicant. Demonstrated effects include
neurobehavioral deficits and neurochemical changes (Giustino et al., 1999; Cagiano et al., 1998;
De Salvia et al., 1995; Fechter, 1987). In humans, CYP2E1 activity in the brain occurs earlier in
gestation than it does in the liver, with activity in the brain seen in the first trimester (Johnsrud et
al., 2003; Brzezinski et al., 1999). Thus, the direct effects of dichloromethane in fetal
circulation, as well as the effects of CO and the effects of the CYP2E1-related metabolism in the
fetal liver and the fetal brain, may be relevant to the risk of developmental effects in humans.
4.6.3.5. Mode of Action for Immunotoxicity
Evidence of a localized immunosuppressive effect in the lung resulting from inhalation
dichloromethane exposure was seen in an acute exposure (3 hours, 100 ppm) study in CD-I mice
(Aranyi et al., 1986). The lung infectivity assay used in this study examined response to
bacterial challenges (i.e., risk of Streptococcal-pneumonia-related mortality and clearance of
Klebsiella bacteria). The innate immune response plays an important role in limiting the initial
lung burden of bacteria through the activity of macrophages, neutrophils, and dendritic cells, and
alveolar macrophages are particularly important in the response to respiratory infections
(Marriott and Dockrell, 2007). The adaptive response develops from several days up to several
weeks following infection so that an effective immune response in a lung infectivity assay
requires multiple immune mechanisms and, in particular, cooperation of macrophages,
neutrophils, and T cells along with the appropriate cytokines (Belgrade and Gilmour, 2006).
Although immunosuppression in the Streptococcal and Klebsiella infectivity models has been
reported in the acute exposure scenarios tested in Aranyi et al. (1986), mechanistic studies of
dichloromethane or its metabolites that would provide mode of action information on the
immune system cells or function have not been performed.
4.7. EVALUATION OF CARCINOGENICITY
4.7.1. Summary of Overall Weight of Evidence
Following U.S. EPA (2005a) Guidelines for Carcinogen Risk Assessment,
dichloromethane is "likely to be carcinogenic in humans," based predominantly on evidence of
carcinogenicity at two sites in 2-year bioassays in male and female B6C3Fi mice (liver and lung
tumors) with inhalation exposure (NTP, 1986) and at one site in male B6C3Fi mice (liver
tumors) with drinking water exposure (Serota et al., 1986b; Hazleton Laboratories, 1983). The
incidence rates for liver tumors in female mice were not presented (Serota et al., 1986b; Hazleton
Laboratories, 1983), but it was reported that exposed female mice did not show increased
216 DRAFT - DO NOT CITE OR QUOTE
-------
incidences of proliferative hepatocellular lesions. Evidence of a trend for increased risk of liver
tumors (described as neoplastic nodule or hepatocellular carcinoma) was seen in female F344
rats exposed via drinking water (p < 0.01) (Serota et al., 1986a) or inhalation (p = 0.08) (NTP,
1986). However, the potential malignant characterization of the nodules was not described, and
the data for hepatocellular carcinomas are much more limited. Additional evidence of the
tumorigenic potential of dichloromethane in rats comes from the observation of an increase in
benign mammary tumors following inhalation exposure (Nitschke et al., 1988a; Burek et al.,
1984; NTP, 1986). A gavage study in female Sprague-Dawley rats reported an increased
incidence of malignant mammary tumors, mainly adenocarcinomas (8, 6, and 18% in the control,
100, and 500 mg/kg dose groups, respectively), but the increase was not statistically significant;
data were not provided to allow an analysis that accounts for differing mortality rates (Maltoni et
al., 1988). An inhalation study (exposures of 0, 50, 200, and 500 ppm) also reported the
presence of another relatively rare tumor in rats, astrocytoma or glioma (mixed glial cell) tumors
(Nitschke et al., 1988a). Taken together, the rat data provide supporting evidence of
carcinogenicity. Studies in humans also observed evidence linking occupational exposure to
dichloromethane and increased risk for some specific cancers, including brain cancer (Hearne
and Pifer, 1999; Cocco et al., 1999; Tomenson et al., 1997; Heineman et al., 1994), liver cancer
(Lanes et al., 1993, 1990), non-Hodgkin lymphoma (Barry et al., 2011; Wang et al., 2009;
Seidler et al., 2007, Miligi et al., 2006), and multiple myeloma (Gold et al., 2011).
The proposed mode of action for dichloromethane-induced tumors is through a
mutagenic mode of carcinogenic action. Mode of action data indicate that dichloromethane-
induced DNA damage in cancer target tissues of mice (i.e., liver and lung) involves DNA-
reactive metabolites produced via a metabolic pathway initially catalyzed by GST-T1. Evidence
of mutagenicity includes in vitro bacterial assays in several strains (for example, Jongen et al.,
1982, 1978; Gocke et al., 1981; Green, 1983; Their et al., 1993), and in vitro mutagenicity tests
in mammalian systems, including the HPRT gene mutation assay in CHO cells with added GST
activity (Graves et al., 1996) and the micronucleus test in human AHH-1, MCL-5, and h2El cell
lines (Doherty et al., 1996). In in vivo studies using mouse red blood cells, the micronucleus test
and assays for chromosome aberrations were also positive at inhalation doses consistent with the
doses that induced mouse tumors (Allen et al., 1990). Additional in vivo evidence of
genotoxicity as evidenced by sister chromatid exchanges and DNA damage (comet assay) has
also been seen in mouse liver and lung cells (Sasaki et al., 1998; Graves et al., 1995. 1994b;
Casanova et al., 1992; Allen et al.,1990), although mutational events in critical genes (tumor
suppressor genes, oncogenes) leading to tumor initiation and tumor promotion have not been
established (Devereux et al., 1993; Hegi et al., 1993). This metabolic pathway has been found in
human tissues, albeit at lower activities than in mouse tissues; therefore, the cancer results in
animals are considered relevant to humans.
217 DRAFT - DO NOT CITE OR QUOTE
-------
EPA's Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a) indicate that for
tumors occurring at a site other than the initial point of contact the weight of evidence for
carcinogenic potential may apply to all routes of exposure that have not been adequately tested at
sufficient doses. An exception occurs when there is convincing toxicokinetic data that
absorption does not occur by other routes. For dichloromethane, systemic tumors were observed
in mice following inhalation and oral exposure. No animal cancer bioassay data following
dermal exposure to dichloroemthane are available. Based on the observance of systemic tumors
following oral exposure and inhalation exposure, and in the absence of information to indicate
otherwise, it is assumed that an internal dose will be achieved regardless of the route of
exposure. Therefore, dichloromethane is "likely to be carcinogenic to humans" by all routes of
exposure.
4.7.2. Synthesis of Human, Animal, and Other Supporting Evidence
Section 4.1.3 reviewed the results, strengths, and limitations of epidemiological research
of dichloromethane and cancer, including cohort and case-control studies. The available
epidemiologic studies provide evidence of an association between dichloromethane and brain
cancer, liver cancer, and some hematopoieitic cancers (specifically non-Hodgkin lymphoma and
multiple myeloma).
Two small cohort studies with relatively good exposure metrics and relatively long
follow-up periods (mean over 25 years) reported an increased risk of brain cancer, with SMRs of
1.45 (95% CI 0.40-3.72) in Tomenson et al. (1997) and 2.2 (95% CI 0.79-4.69) in Cohort 1 of
Hearne and Pifer (1999). Cohort 1 is an inception cohort, following workers from the beginning
of employment, which is methodologically more robust than Cohort 2, which only included
workers who were working between 1964 and 1970. These observations are supported by the
data from a case-control study of brain cancer using lifetime job history data that reported
relatively strong trends (p < 0.05) with increasing probability, duration, and intensity measures of
exposure but not with a cumulative exposure measure (Heineman et al., 1994). This difference
in results between different exposure measures could reflect a relatively more valid measure of
relevant exposures in the brain from the intensity measure, as suggested by the study in rats
reported by Savolainen et al. (1981) in which dichloromethane levels in the brain were much
higher with a higher intensity exposure scenario compared with a constant exposure period with
an equivalent TWA (see Section 3.2). The combination of high probability of exposure and long
(> 20 years) duration of employment in exposed jobs was strongly associated with brain cancer
risk (OR 6.1, 95% CI 1.1-43.8) in the Heineman et al. (1994) study; similar associations were
seen with the measure combining high intensity with long duration. In a case-control study of
female brain cancer cases, Cocco et al. (1999), using more limited occupation data obtained from
death certificates, observed a weak overall association with dichloromethane exposure, and no
trends with probability or intensity. A statistically significant increased incidence of brain or
218 DRAFT - DO NOT CITE OR QUOTE
-------
CNS tumors has not been observed in any of the animal cancer bioassays, but a 2-year study
using relatively low exposure levels (0, 50, 200, and 500 ppm) in Sprague-Dawley rats observed
a total of six astrocytoma or glioma (mixed glial cell) tumors in the exposed groups (in females,
the incidence was 0, 0, 0, and 2 in the 0, 50, 200, and 500 ppm exposure groups, respectively; in
males, the incidence was 0, 1,2, and 1 in the 0, 50, 200, and 500 ppm exposure groups,
respectively; sample size of each group was 70 rats). These tumors are exceedingly rare in rats,
and there are few examples of statistically significant trends in animal bioassays (Sills et al.,
1999). These cancers were not seen in two other studies in rats, both involving higher doses
(1,000-4,000 ppm) (NTP, 1986; Burek et al., 1984), or in a high dose (2,000-4,000 ppm) study
inmice(NTP, 1986).
With respect to epidemiologic studies of liver and biliary duct cancer, the highest
exposure cohort, based in the Rock Hill, South Carolina, triacetate fiber production plant,
suggested an increased risk of liver cancer with an SMR of 2.98 (95% CI 0.81-7.63) in the latest
study update (Lanes et al., 1993). This observation was based on four cases; an earlier analysis
in this cohort reported an SMR of 5.75 (95% CI 1.82-13.8), based on these same four cases but
with a shorter follow-up period (and thus a lower number of expected cases) (Lanes et al., 1990).
No other cohort study has reported an increased risk of liver cancer mortality, although it should
be noted that there is no other inception cohort study of a population with exposure levels similar
to those of the Rock Hill plant, and no data from a case-control study of liver cancer are
available pertaining to dichloromethane exposure.
The primary limitation of all of the available dichloromethane cohort studies is the
limited statistical power for the estimation of effects relating to relatively rare cancers (such as
brain cancer, liver cancer, and leukemia). Limitations with respect to studies of other cancers
can also be noted. With respect to breast cancer, the only cohort that included a significant
percentage of women had limited exposure information (analysis was based on a dichotomous
exposure variable) and co-exposure to other solvents that also exhibited associations of similar
magnitude to that seen with dichloromethane (Radican et al., 2008). Thus, in this situation,
potential confounding by these other exposures should be considered. The only breast cancer
case-control study available used death certificate data to classify disease and occupational
exposure (Cantor et al., 1995), which is likely to result in significant misclassification; exposure
misclassification in particular would be expected to result in an attenuated measure of
association (Rothman and Greenland, 1998). The available epidemiologic studies do not provide
a definitive evaluation of non-Hodgkin lymphoma, but the consistent observations of
associations seen in three large case-control studies in Germany (Seidler et al., 2007), Italy
(Miligi et al., 2006), and the United States (Barry et al., 2011; Wang et al., 2009) provide
evidence of an increased risk of specific types of hematopoietic cancers in humans. These
studies are limited by relatively small number of exposed cases, resulting in imprecise effect
estimates.
219 DRAFT - DO NOT CITE OR QUOTE
-------
In addition to the epidemiologic studies, several dichloromethane cancer bioassays in
animals are available. In the only oral exposure cancer bioassay involving lifetime exposure,
increases in incidence of liver adenomas and carcinomas were observed in male but not female
B6C3Fi mice exposed for 2 years (Table 4-38 for males; female data not presented in the
summary reports) (Serota et al., 1986b; Hazleton Laboratories, 1983). The authors concluded
that these increases were "within the normal fluctuation of this type of tumor incidence," noting
that there was no dose-related trend and that there were no significant differences comparing the
o
individual dose groups with the combined control group. Serota et al. (1986b) state that a two-
tailed significance level ofp = 0.05 was used for all tests, but that statement does not appear to
accurately represent top-value used in their statistical analysis. Each of the/>-values for the
comparison of the 125, 185, and 250 mg/kg-day dose groups with the controls wasp < 0.05.
(Theses-values were found in the full report of this study, see Hazleton Laboratories, 1983, but
were not included in the Serota et al., 1986b publication). Hazleton Laboratories (1983)
indicated that a correction factor for multiple comparisons was used specifically for the liver
cancer data, reducing the nominal p-va\ue to 0.0125 rather than the value of 0.05 that was
reported to be used by Serota et al. (1986b); none of these individual group comparisons are
statistically significant when ap-va\ue of 0.0125 is used. EPA did not consider the use of a
multiple comparisons correction factor for the evaluation of the liver tumor data (a primary a
priori hypothesis) to be warranted. Thus, based on the Hazleton Laboratories (1983) statistical
analysis, EPA concluded that dichloromethane induced a carcinogenic response in male B6C3Fi
mice as evidenced by a marginally increased trend test (p = 0.058) for combined hepatocellular
adenomas and carcinomas, and by small but statistically significant (p < 0.05) increases in
hepatocellular adenomas and carcinomas at dose levels of 125 (p = 0.021), 185 (p = 0.019), and
250 mg/kg-day (p = 0.036).
With respect to comparisons with historical controls, the incidence in the control groups
was almost identical to the mean seen in the historical controls from this laboratory (17.8% based
on 354 male B6C3Fi mice), so there is no indication that the observed trend is being driven by
an artificially low rate in controls and no indication that the experimental conditions resulted in a
systematic increase in the incidence of hepatocellular adenomas and carcinomas. Although the
occurrence of one elevated rate in an exposed group may be within the normal fluctuations of
this type of tumor incidence (described for this laboratory as 5-40%, with a mean of 17.8%,
based on 354 male controls), the pattern of incidence rates (increased incidence in all four dose
groups, with three of these increases significant at a p-value < 0.05) suggest a treatment-related
increase.
Two control groups were used because of the potential for high and erratic liver tumor incidence in B6C3F1 mice.
The incidence of hepatocellular adenomas or carcinomas was 18 and 20% in the two control groups, and the
combined group is used for the subsequent analysis because of the improved statistical precision of estimates based
on the larger sample size (n=125 compared with n=60 and 65 for the individual control groups).
220 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-38. Incidence of liver tumors in male B6C3Fi mice exposed to
dichloromethane in a 2-year oral exposure (drinking water) study"
Estimated mean intake
(mg/kg-d)a
Number of male mice
Hepatocellular adenoma or carcinoma
Mortality-adjusted percent"1
Mortality-adjusted />-valued
0 (Controls)
125b
61
200
124
100
111
99
234
125
Trend
/7-valuec
Number of cancers (%)
24 (19)
(22)
51 (26)
(29)
;? = 0.071
30 (30)
(34)
p = 0.023
31(31)
(35)
;? = 0.019
35 (28)
(32)
;? = 0.036
0.058
"Target doses were 60, 125, 185, and 250 mg/kg-d from the lowest dose group (excluding controls) to the highest
dose group, respectively.
bTwo control groups combined (n=60 and 65 in the individual groups). The mortality-adjusted incidence in control
groups 1 and 2 were 20 and 23%, respectively. Two additional sets of analyses using the individual control groups
were also presented in Hazleton Laboratories (1983).
°Cochran-Armitage trend test (Hazleton Laboratories [1983]).
dMortality-adjusted percent calculated based on number at risk, using Kaplan-Meier estimation, taking into account
mortality losses; />-value for comparison with control group using asymptotic normal test (source: Hazleton
Laboratories [1983]).
Sources: Serota et al. (1986b); Hazleton Laboratories (1983).
In a similar study in F344 rats (Serota et al., 1986a), no increased incidence of liver
tumors was seen in male rats, and the pattern in female rats was characterized by a jagged
stepped pattern of increasing incidence of hepatocellular carcinoma or neoplastic nodule
(Table 4-39). Information was not provided which would allow characterization of the nodules
as benign or malignant. Statistically significant increases in incidences were observed in the
50 and 250 mg/kg-day groups (incidence rates of 0, 3, 10, 3, and 14%, respectively, for the 0, 5,
50, 125, and 250 mg/kg-day groups) and in the group exposed to 250 mg/kg-day for 78 weeks
followed by a 26-week period of no exposure (incidence rate 10%). A similar pattern, but with
more sparse data, was seen for hepatocellular carcinomas, with two incidences in the 50 mg/kg-
day and two in the 250 mg/kg-day groups. The authors concluded that dichloromethane
exposure did not result in an increased incidence of liver tumors because the increase was based
on a low rate (0%) in the controls and because of a lack of monotonicity.
221
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-39. Incidences of liver tumors in male and female F344 rats exposed
to dichloromethane in drinking water for 2 years
Males
Estimated mean intake (mg/kg-d)
total n
n at terminal killd
Number (%) with neoplastic nodules
Number (%) with hepatocellular
carcinoma
Number (%) with neoplastic nodules
and hepatocellular carcinoma
Females
Estimated mean intake (mg/kg-d)
total n
n at terminal killd
Number (%) with neoplastic nodules
Number (%) with hepatocellular
carcinoma
Number (%) with neoplastic nodules
and hepatocellular carcinoma
Target dose (mg/kg-d)
Oa
(Controls)
0
135
76
9(12)
3(4)
12(16)
0
135
67
0(0)
0(0)
0(0)
5
6
85
34
1(3)
0(0)
1(3)
6
85
29
1(3)
0(0)
1(3)
50
52
85
38
0(0)
0(0)
0(0)
58
85
41
2(5)
2(5)
4 (10)e
125
125
85
35
2(6)
0(0)
2(6)
136
85
38
1(3)
0(0)
1(3)
250
235
85
41
1(2)
1(2)
2(5)
263
85
34
3(9)
2(6)
5 (14)e
Trend
/7-valueb
Not
reported
Not
reported
Not
reported
Not
reported
p<0.0l
250 with
recovery0
232
25
15
2(13)
0(0)
2(13)
239
25
20
2 (10)e
0(0)
2 (10)e
aTwo control groups combined.
bCochran-Armitage trend test was used for trend test of liver foci/areas of alteration. For tumor mortality-
unadjusted analyses, a Cochran-Armitage trend test was used, and for tumor mortality-adjusted analyses, a tumor
prevalence analytic method by Dinse and Lagakos (1982) was used. Similar results were seen in these two
analyses.
'Recovery group was exposed for 78 wks and then had a 26-wk period without dichloromethane exposure; n = 17
for neoplastic lesions.
dExcludes 5, 10, and 20 per group sacrificed at 25, 52, and 78 wks, respectively, and unscheduled deaths, which
ranged from 5 to 19 per group.
Significantly (p < 0.05) different from controls with Fisher's exact test, mortality-unadjusted and mortality-
adjusted analyses.
Source: Serotaetal. (1986a).
Gavage exposure studies in Sprague-Dawley rats and in Swiss mice provides limited data
concerning cancer incidence because the study was terminated early (at 64 weeks) due to high
treatment-related mortality (Maltoni et al., 1988). Exposure groups included controls (olive oil),
100, or 500 mg/kg-day 4-5 days/week. High-dose female rats showed an increased incidence of
malignant mammary tumors, mainly adenocarcinomas (8, 6, and 18% in the control, 100, and
500 mg/kg dose groups, respectively), but the increase was not statistically significant. Data
were not provided to allow an analysis accounting for differing mortality rates. A dose-related
increase, although not statistically significant, in pulmonary adenomas was observed in male
222
DRAFT - DO NOT CITE OR QUOTE
-------
mice (5, 12, and 18% in control, 100, and 500 mg/kg-day groups, respectively). When mortality
was taken into account, high-dose male mice that died in the period ranging from 52 to 78 weeks
were reported to show a statistically significantly (p < 0.05) elevated incidence for pulmonary
tumors (1/14, 4/21, and 7/24 in control, 100, and 500 mg/kg-day groups, respectively). Details
of this analysis were not provided. EPA applied a Fisher's exact test to these incidences and
determined ap-va\ue of 0.11 for the comparison of the 500 mg/kg-day group (7/24) to the
controls (1/14).
As discussed in Section 4.2, repeated inhalation exposure to concentrations of 2,000 or
4,000 ppm dichloromethane produced increased incidences of lung and liver tumors in B6C3Fi
mice (Mennear et al., 1988; NTP, 1986). The incidence of mortality-adjusted liver tumors across
dose groups of 0, 2,000, and 4,000 ppm increased from 48 to 67 and 93%, respectively, in male
mice (trends-value = 0.013) and from 10 to 48 and 100% in female mice (trend
^-values < 0.001) (Table 4-40). For lung tumors, the mortality-adjusted incidence was 12, 74,
and 100% in males and 11, 83, and 100% in females in the 0, 2,000, and 4,000 ppm groups,
respectively (trends-values < 0.001). Elevated incidences of lung and liver tumors in B6C3Fi
mice were observed with 52 weeks of exposure to 2,000 ppm, and lung tumors were also
elevated by week 104 in mice exposed for only 26 weeks to 2,000 ppm, followed by 78 weeks
without exposure (Maronpot et al., 1995; Kari et al., 1993).
223 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-40. Incidences of selected neoplastic lesions in B6C3Fi mice exposed
to dichloromethane by inhalation (6 hours/day, 5 days/week) for 2 years
Sex and neoplastic lesion
Exposure (ppm)a
0 (Controls)
n
(%)b
(%)c
2,000
n
(%)b
(%)c
4,000
n
(%)b
(%)°
Trend
/7-valued
Males
Liver — hepatocellular adenoma or
carcinoma
Lung — bronchoalveolar adenoma or
carcinoma
22
5
(44)
(10)
(48)
(12)
24
27e
(49)
(54)
(67)
(74)
33e
40e
(67)
(80)
(93)
(100)
0.013
0.001
Females
Liver — hepatocellular adenoma or
carcinoma
Lung — bronchoalveolar adenoma or
carcinoma
3
3
(6)
(6)
(10)
(11)
16e
30e
(33)
(63)
(48)
(83)
40e
41e
(83)
(85)
(100)
(100)
0.001
0.001
a2,000 ppm = 6,947 mg/m3, 4,000 ppm = 13,894 mg/m3.
bTotal sample size was 50 per sex and dose group. Percentages based on the number of tissues examined
microscopically per group; for male mice, 49 livers were examined in the 2,000 and 4,000 ppm groups; for female
mice, 48 liver and lungs were examined. For comparison, incidences in historical controls reported in NTP (1986)
were 28% for male liver tumors, 31% for male lung tumors, 5% for female liver tumors, and 10% for female lung
tumors.
'Mortality-adjusted percentage.
dLife-table trend test, as reported by NTP (1986).
eLife-table test comparison dose group with control O.05, as reported by NTP (1986).
Sources: Mennear et al. (1988); NTP (1986).
A moderate trend of increasing incidence of what was described as neoplastic nodules or
hepatocellular carcinoma was seen in female F344 rats (trends-value = 0.08) but not males in
the NTP (1986) study (Table 4-41). Liver tumors are relatively rare in F344 rats. As with the rat
oral exposure study by Serota et al. (1986a), the nodules were not characterized as benign or
malignant. There was no evidence of an increasing trend in incidence when hepatocellular
carcinomas only were considered.
224
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-41. Incidences of selected neoplastic lesions in F344/N rats exposed to dichloromethane by inhalation
(6 hours/day, 5 days/week) for 2 years
Sex and neoplastic lesion
Exposure (ppm)a
0 (Controls)
n
(%)b
(%)°
1,000
n
(%)b
(%)c
2,000
n
(%)b
(%)°
4,000
n
(%)b
(%)°
Trend
/7-valued
Males
Liver — Neoplastic nodule or hepatocellular carcinoma
Liver — hepatocellular carcinoma
Lung — bronchoalveolar adenoma or carcinoma
Mammary gland
Adenoma, adenocarcinoma, or carcinoma
Subcutaneous tissue fibroma or sarcoma
Fibroadenoma
Mammary gland or subcutaneous tissue adenoma,
fibroadenoma, fibroma, or sarcoma
2
2
1
0
1
0
1
(4)
(4)
(0)
(2)
(0)
(2)
(10)
(10)
(6)
(0)
(6)
3
1
1
0
1
0
1
(6)
(2)
(2)
(0)
(2)
(0)
(2)
(13)
(4)
(6)
(0)
(6)
4
2
2
0
2
2
4
(8)
(4)
(4)
(0)
(4)
(4)
(8)
(19)
(10)
(9)
(12)
(21)
1
1
1
1
5
1
9d
(2)
(2)
(2)
(2)
(10)
(2)
(18)
(6)
(6)
(23)
(8)
(49)
0.55
nr
0.008
0.001
0.001
Females
Liver — neoplastic nodule or hepatocellular carcinoma
Liver — hepatocellular carcinoma
Lung — bronchoalveolar adenoma or carcinoma
Mammary gland
Adenocarcinoma or carcinoma
Adenoma, adenocarcinoma, or carcinoma
Fibroadenoma
Mammary gland adenoma, fibroadenoma, or
adenocarcinoma
2
0
1
1
1
5
6
(4)
(0)
(2)
(2)
(2)
(10)
(12)
(7)
(0)
(16)
(18)
1
0
1
2
2
lld
13
(2)
(0)
(2)
(4)
(4)
(22)
(26)
(2)
(0)
(41)
(44)
4
1
0
2
2
13d
14d
(8)
(2)
(0)
(4)
(4)
(26)
(28)
(14)
(4)
(44)
(45)
5
0
0
0
1
22d
23e
(10)
(0)
(0)
(0)
(2)
(44)
(46)
(20)
(0)
(79)
(86)
0.08
nr
0.001
0.001
al,000 ppm = 3,474 mg/m3, 2,000 ppm = 6,947 mg/m3, 4,000 ppm = 13,894 mg/m3.
bTotal sample size was 50 per sex and dose group. Percentages based on the number of tissues examined microscopically per group; for male rats, 49 livers
were examined in the 2,000 and 4,000 ppm groups; for females, only 48 liver and lungs and 49 mammary glands were microscopically examined in the
2,000 and 4,000 ppm groups. For comparison, incidence in historical controls reported in NTP (1986) were 1% for female liver tumors and 16% for female
mammary fibroadenomas.
'Mortality-adjusted percentage.
dLife-table trend test, as reported by NTP (1986); nr = not reported.
eLife-table test comparison dose group with control < 0.05, as reported by NTP (1986).
Sources: Mennear et al. (1988); NTP (1986).
225
DRAFT - DO NOT CITE OR QUOTE
-------
Female F344 rats exposed by inhalation to 2,000 or 4,000 ppm showed significantly
increased incidences of benign mammary tumors (adenomas or fibroadenomas) (Table 4-41); the
number of benign mammary tumors per animal also increased with dichloromethane exposure in
studies in Sprague-Dawley rats at levels of 50-500 ppm (Nitschke et al., 1988a) and 500-
3,500 ppm (Burek et al., 1984) (Table 4-42). Male rats in two of these studies (Nitscke et al.,
1988a; NTP, 1986) also exhibited a low rate of sarcoma or fibrosarcoma in mammary gland or
subcutaneous tissue around the mammary gland.
226 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-42. Incidences of mammary gland tumors in two studies of male
and female Sprague-Dawley rats exposed to dichloromethane by inhalation
(6 hours/day, 5 days/week) for 2 years
Study, lesion
0
(Controls)
Exposure (ppm)a
50
200
500
Late
500b
Early
500b
1,500
3,500
Nitschke et al. (1988a)
Males — n per group
Number (%) with:
Mammary gland tumors
Adenocarcinoma or carcinoma
Fibroadenoma
Fibroma
Fibrosarcoma
Undifferentiated sarcoma
Fibroma, fibrosarcoma, or
undifferentiated sarcomad
Females — n per group
Number (%) with:
Mammary gland tumors
Adenocarcinoma or carcinoma
Adenoma
Fibroadenoma
Fibroma
Fibrosarcoma
Number with benign tumors6
Number of benign tumors per
tumor-bearing rate
57
0(0)
2(4)
6(11)
0(0)
0(0)
6(11)
69
6(9)
1(1)
51 (74)
0(0)
1(1)
52 (74)
2.0
65
0(0)
0(0)
1(6)
1(6)
2(4)
4(6)
69
5(7)
1(1)
57 (83)
1(1)
0(0)
58 (83)
2.3
59
0(0)
2(3)
6(11)
1(6)
0(0)
7(12)
69
4(6)
2(3)
60 (87)
0(0)
0(0)
61(87)f
2.2
64
0(0)
2(3)
10 (16)
0(0)
0(0)
10 (16)
69
4(6)
1(1)
55 (80)
1(1)
0(0)
55 (79)
2.7
c
25
3(12)
2(8)
22 (88)
1(4)
0(0)
23 (92)
2.2
C
25
2(8)
0(0)
23 (92)
1(1)
0(0)
23 (92)
2.6
C
C
C
c
Bureketal. (1984)
Males — n per group
Number (%) with benign tumors
Total number of benign tumors
Number of tumors per tumor-
bearing rat8
Females — n per group
Number (%) with benign tumors
Total number of benign tumors
Number of tumors per tumor-
bearing ratf
92
7(8)
8
1.1
96
79 (82)
165
2.1
c
c
c
c
95
3(3)
6
2.0
95
81 (85)
218
2.7
c
c
c
c
96
7(7)
11
1.6
96
80 (83)
245
3.1
97
14 (14)
17
1.2
97
83 (86)
287
3.5
a50 ppm = 174 mg/m3, 200 ppm = 695 mg/m3, 500 ppm = 1,737 mg/m3, 1,500 ppm = 5,210 mg/m3, 3,500 ppm =
12,158 mg/m3.
bLate 500 = no exposure for first 12 mo followed by 500 ppm for last 12 mo; early 500 = 500 ppm for first 12 mo
followed by no exposure for last 12 mo.
°No data for this exposure level in this study.
dEPA summed across these tumor types, assuming no overlap.
eln historical controls, percent with benign tumors reported was 79-82% and number per tumor-bearing rat was 2.1.
Significantly (p < 0.05) higher than control incidence by Fisher's exact test (Nitschke et al., 1988a).
Calculated by EPA.
Sources: Nitschke et al. (1988a); Burek et al. (1984).
227
DRAFT - DO NOT CITE OR QUOTE
-------
In Syrian golden hamsters exposed to 500, 1,500, or 3,500 ppm for 2 years, no
statistically significantly increased incidences of tumors were found in any tissues (Burek et al.,
1984).
Supporting evidence for the carcinogenicity of dichloromethane comes from the results
of genotoxicity and mode of action studies discussed in Section 4.5. A mutagenic mode of
carcinogenic action for dichloromethane involves metabolic activation by GST, as evidenced by
several observations, including the enhancement of dichloromethane mutagenic activity in
normally unresponsive S. typhimurium strain TA1535 after it is transfected with the gene for rat
GST-T1 (DeMarini et al., 1997; Thier et al., 1993); increased HPRT gene mutations and DNA
damage (DNA SSBs) in CHO cells when they are incubated with dichloromethane in the
presence of mouse liver cytosol preparations rich in GST enzymatic activities (Graves and
Green, 1996; Graves et al., 1996, 1994b); the detection of DNA damage (DNA SSBs) in liver
and lung tissue of B6C3Fi mice immediately following 6-hour inhalation exposure to
dichloromethane (2,000-8,000 ppm); and a suppression of the DNA damage to levels seen in
controls when mice were pretreated with buthionine sulphoximine, a GSH depletor (Graves et
al., 1995).
Additional data from several studies indicate that dichloromethane genotoxicity is
expressed in cancer target tissues in mice following in vivo exposure. Increased sister chromatid
exchanges were observed in lung cells of B6C3Fi mice after 90 days of inhalation exposure to
2,000 ppm or 10 days of exposure to 4,000 or 8,000 ppm (Allen et al., 1990). DNA damage
(comet assay) was detected in liver and lung tissue (but not stomach, kidney, brain, or bone
marrow) 24 hours after oral administration of 1,720 mg/kg dichloromethane to CD-I mice
(Sasaki et al., 1998). DNA-protein cross-links were observed in the liver of B6C3Fi mice but
not hamsters following inhalation exposure to concentrations ranging from 500 to 4,000 ppm
6 hours/day for 3 days (Casanova et al., 1996, 1992). Much less is known about genotoxicity in
the liver in rats. Studies of DNA SSBs in rat hepatocytes or liver homogenate were negative
with inhalation exposures up to 5,000 ppm for 3 hours (Graves et al., 1995, 1994b), but positive
results were seen in a high-dose gavage study (1,275 mg/kg) (Kitchin and Brown, 1989). Few
other specific types of genotoxicity endpoints (e.g., sister chromatid exchange, DNA-protein
cross-links) have been studied in the rat liver.
Since there are limited data on mutagenic events following oral exposure, EPA conducted
a pharmacokinetic analysis to evaluate how comparable the internal doses to the liver in the oral
bioassay (Serota et al., 1986b; Hazleton Laboratories, 1983) were to the internal doses to the
liver in the inhalation bioassay (Mennear et al., 1988; NTP, 1986). The PBPK model of Marino
et al. (2006) predicted that the average daily amount of dichloromethane metabolized via GST
per liter of liver was about 14-fold lower in mice exposed to the highest dose of 234 mg/kg-day
in the drinking water bioassay than in mice exposed to the lowest inhalation exposure of 2,000
ppm inducing liver tumors (Table 4-43). Thus, the lower incidence of liver tumors induced by
228 DRAFT - DO NOT CITE OR QUOTE
-------
oral doses of 234 mg/kg-day compared with the higher incidence induced by inhalation exposure
to 2,000 ppm is consistent with the predicted lower liver dose of GST metabolites (and hence
lower probability of DNA modification) with oral exposure.
While the amount metabolized by the GST pathway for inhalation exposure shown in
Table 4-43 is lower in the rat versus the mouse, as one would expect based on the enzyme
expression level, for oral exposure a higher amount of GST metabolism is predicted for the rat
than the mouse. This difference occurs because for oral exposure 100% of the dose is absorbed,
rather than absorption being limited by metabolism as it is for inhalation, and because the ratio of
GST to CYP activity is higher in the rat than in the mouse. Specifically kfc/Vmaxc is 0.626 for
the rat and is 0.152 for the mouse, so for the rat the fraction of the absorbed dose going to GST is
roughly four times that in the mouse, hence for the same oral dose per kg body weight per day
(with 100% absorbed), approximately four times more is metabolized by GST in the rat than in
the mouse.
Table 4-43. Comparison of internal dose metrics in inhalation and
oral exposure scenarios in male mice and rats
External dose
Inhalation (ppm)
2,000
4,000
Oral (mg/kg-d)b
61
124
177
234
Internal exposure in liver (mg metabolized through GST pathway/L
liver tissue/d)a
Male
Mouse
Rat
2,364
4,972
1,509
3,124
17.5
63.3
112.0
169.5
77.1
233.6
385.7
559.0
aMouse values derived by EPA from the PBPK model of Marino et al. (2006); rat values derived from
EPA based on the modified PBPK model of Andersen et al. (1991) (see Appendix C for model details).
bActual doses administered to mice (Serota et al., 1986a); BWs not given for males and females, so
simulation results only provided for one gender.
Another interspecies difference is the localization of GST-T1 in the nuclei of hepatocytes
and bile duct epithelial cells in the mouse, while the rat liver does not show preferential nuclear
localization of GST-T1. In human liver tissue, some hepatocytes show nuclear localization of
GST-T1 and others show localization in cytoplasm, as well as in nuclei of bile duct epithelial
cells (Sherratt et al., 2002; Mainwaring et al., 1996).
229
DRAFT - DO NOT CITE OR QUOTE
-------
4.7.3. Mode of Action Information
4.7.3.1. Hypothesized Mode of Action
The hypothesized mode of action for dichloromethane-induced tumors is through a
mutagenic mode of carcinogenic action. Much of the experimental mode of action research has
focused on the liver and lung, the sites of tumor formation in chronic bioassays in mice
(Mennear et al., 1988; NTP, 1986; Serotaetal., 1986b, Hazleton Laboratories, 1983). The
studies examining measures of mutagenicity (e.g., HPRT mutations, micronucleus tests,
chromosomal aberrations) and the positive genotoxicity data in the database of available studies
of dichloromethane, specifically in in vivo studies focusing on tissues that are the site of
carcinogenic response, support the conclusion that a mutagenic MOA is operative in
dichloromethane-induced tumors. The mode of action is potentially relevant to other sites,
particularly those in which GST-T1 is expressed (i.e., GSH conjugation), such as mammary
tissue (Lehmann and Wagner, 2008) and the brain (Juronen et al., 1996) in humans, and rat
olfactory epithelium (Banger et al., 1994). While the extent of GSH conjugation in these other
tissues may not be significant to the overall dosimetry of dichloromethane, they can be
significant in understanding the sites of action of dichloromethane and differences in tumor sites
observed between species. The role of specific mutations in mouse or human cancers has not
been established.
Support for the importance of GST in the hypothesized mutagenic mode of action has
been demonstrated in in vitro bacterial and mammalian assays as well as in in vivo mammalian
system assays. Dichloromethane is consistently mutagenic in S. typhimurium strains with GST
capability but did not produce mutagenic effects in non-GST S. typhimurium strains
(summarized in Section 4.5.1.1 and Table 4-29). In vitro mammalian cell studies have
demonstrated mutagenicity (i.e., HPRT gene mutations) in CHO cell lines when a mouse liver
cytosol fraction was exogenously added (Graves et al., 1996) and in the micronucleus test in
human AHH-1, MCL-5 and h2El cell lines (Doherty et al., 1996); positive responses were seen
in studies measuring DNA-protein cross-links, and DNA SSBs (see Table 4-30). Other studies
have demonstrated DNA adducts with dichloromethane exposure in calf thymus DNA in the
presence of bacterial GST DM11. Negative results were seen in most of the other in vitro cell
studies using rat hepatocytes or CHO cells without mouse liver cytosol incubation. These
studies were conducted in cell lines where GST activity is considerably lower than in mouse cell
lines and therefore, these results are consistent with the importance of GST in the carcinogenicity
of dichloromethane.
In studies with human cell lines or isolated cells, positive results were reported for sister
chromatid exchanges, chromosomal aberrations, and in the micronucleus test. Negative results
in human cells were reported in unscheduled DNA synthesis, DNA SSBs, and DNA-protein
crosslink assays (Section 4.5.1.1 and Table 4-30).
230 DRAFT - DO NOT CITE OR QUOTE
-------
In vivo studies in mice (Section 4.5.1.2 and Table 4-32) consistently showed
chromosomal aberrations and genotoxic effects following dichloromethane exposure in the liver
and lung where tumors are observed. Other organs in the mouse were evaluated and effects were
not consistently observed. The specificity of the observed effects support the hypothesized mode
of action since these positive mutagenic responses are seen in organs where tumor formation
occurs (i.e., liver and lung) rather than in areas that were not the site of tumors in the mouse
bioassays (e.g., stomach, bladder, kidney). As noted previously, however, GSH conjugation in
other tissues may be relevant with respect to differences in tumor sites observed between species.
Rats and hamsters, as well as humans, have considerably lower GST activity than the
mouse and may be less sensitive to dichloromethane-induced genotoxic effects. In vivo
genotoxicity studies in rats and hamsters were predominantly non-positive (see Table 4-33). In
studies using human cells (peripheral blood mononuclear cells obtained from healthy, non-
smoking, male volunteers), however, increased incidences of sister chromatid exchanges
following exposure to dichloromethane were seen (Olvera-Bello et al., 2010). An additional
increase in sister chromatid exchanges was reported in blood cells that had the highest GST
activity following dichloromethane treatment.
A limitation of the hypothesized mutagenic mode of carcinogenic action for
dichloromethane is the lack of data demonstrating binding of the reactive GST metabolite, S-
(chloromethyl)glutathione, to DNA. In vivo binding of S-(chloromethyl)glutathione to DNA
was not demonstrated in one study (Watanabe et al., 2007) in rats and mice using a relatively low
dose (5 mg/kg). The reactivity of the postulated DNA-reactive species and the instability of the
derived adducts presents considerable challenges to the ability to provide direct evidence of
adduct formation. Thus, this lack of in vivo evidence of S-(chloromethyl)glutathione binding to
DNA does not in itself represent a basis for invalidating the proposed mode of action.
4.7.3.1.1. Experimental support for the hypothesized mode of action
Strength, consistency, and specificity of association. It is hypothesized that mutagenic
events lead to the development of liver and lung tumors following dichloromethane exposure.
Several observations from experimental studies support the mutagenicity of dichloromethane and
the key role of GST metabolism and the formation of DNA-reactive GST-pathway metabolites
(Table 4-44). The GST pathway produces two metabolites of dichloromethane,
S-(chloromethyl)glutathione and formaldehyde, which are potentially reactive with DNA and
other cell macromolecules. Enhanced dichloromethane mutagenicity in bacterial and
mammalian (i.e., CHO) in vitro assays with the introduction of GST metabolic capacity provides
support that GST metabolism and metabolites are involved (DeMarini et al., 1997; Graves and
Green, 1996; Graves etal., 1996, 1995, 1994b; Thieretal., 1993).
In bacterial strains where GST activity was not present (e.g., TA1535, TA1538),
mutagenic effects were not reported following dichloromethane exposure (Oda et al., 1996;
231 DRAFT - DO NOT CITE OR QUOTE
-------
Simula et al., 1993; Osterman-Golkar et al., 1983; Gocke et al., 1981). Further tests of
GST-dependent mutagenicity were evaluated by transfecting GST into non-GST bacterial strains
or decreasing GST activity in GST bacterial strains (e.g., TA100). When GST-T1 was cloned
into bacterial strain TA1535, dichloromethane treatment resulted in reverse mutations in this new
GST+ TA1535 strain, and these mutations were independent of rat S9 metabolic activation
(DeMarini et al., 1997; Pegram et al., 1997; Thier et al., 1993). Similarly, TA100/NG-11, a
bacterial strain with decreased GST activity in comparison to the wild-type TA100 strain,
showed significantly decreased mutagenicity (reverse mutations) following dichloromethane
treatment (Graves et al., 1994a).
In vitro mammalian genotoxicity studies also support the importance of the GST pathway
in relation to the positive effects observed following dichloromethane exposure. Positive results
in the in vitro assays were limited to experiments with the presence of GST in the cell system.
When mouse liver cytosol was added to hamster cell lines, dichloromethane induced HPRT gene
mutations as well as DNA-protein cross-links and DNA SSBs (Graves and Green, 1996; Graves
et al., 1996, 1994b). Additionally, in mouse Clara cells (GST is localized in the lung cells of
mice), DNA SSBs were reported following dichloromethane treatment, and the extent of DNA
damage was significantly decreased when the cells were pretreated with a glutathione depletor
(Graves et al., 1995). Other studies evaluating similar genotoxic endpoints in rat or CHO cells
without modification of the low GST activity in the test system generally reported no evidence of
genotoxic events (Graves et al., 1995; Andrae and Wolff, 1983; Garrett and Lewtas, 1983;
Thilagar and Kumaroo, 1983; Jongen et al., 1981). A study evaluating the genotoxic effects of
dichloromethane (up to 6 mM) in freshly isolated mouse, rat, hamster, and human hepatocytes
provides additional supporting evidence of the influence of GST activity on genotoxicity (DNA-
protein cross-links) (Casanova et al., 1997). Positive results were only observed in hepatocytes
from B6C3Fi mice, the species with the highest GST metabolic capacity (Reitz et al., 1989). In
studies with human cell lines or isolated cells, positive results were reported for chromosomal
aberrations and the micronucleus test (Doherty et al., 1996; Thilagar et al., 1984), as well as for
DNA damage (comet assay) and sister chromatid exchange (Olvera-Bello et al, 2010; Landi et
al., 2003). Negative results were obtained with human cells in unscheduled DNA synthesis
assays (Jongen et al., 1981; Perocco and Prodi, 1981) and dichloromethane was not demonstrated
to be genotoxic in studies of human hepatocytes (Casanova et al., 1997; Graves et al., 1995).
Two of three in vivo genotoxicity studies in insects reported positive results.
Genotoxicity was observed in Drosophila for the gene mutation assay (Gocke et al., 1981) and
the somatic assay (Rodriguez-Arnaiz, 1998) when dichloromethane was administered through
the food. When Drosophila were exposed to dichloromethane via inhalation, genotoxic effects
were negative as measured through gene mutation assays (sex-linked recessive lethal, somatic
mutation and recombination) (Kramers et al., 1991).
232 DRAFT - DO NOT CITE OR QUOTE
-------
In vivo genotoxicity studies reported chromosomal aberrations, DNA-protein cross-links,
DNA SSBs, and sister chromatid exchanges in liver cells of B6C3Fi mice following acute
inhalation exposure to concentrations producing liver tumors with chronic exposure (Casanova et
al., 1996, 1992; Graves et al., 1995, 1994b). The formation of DNA SSBs was suppressed when
the mice were pretreated with a GSH depletor (Graves et al., 1995), providing additional support
for the involvement of GST metabolism. Increased sister chromatid exchanges and
chromosomal aberrations were found in the lungs of mice exposed to dichloromethane for
2 weeks to 8,000 ppm or for 12 weeks to 2,000 ppm. In this study, however, there was evidence
of damage at other sites, too: sister chromatid exchanges were also seen in peripheral
lymphocytes, chromosomal aberrations were seen in bone marrow, and micronuclei were seen in
peripheral red blood cells under the same exposure protocol (Allen et al., 1990). As was seen in
the liver, DNA SSBs were seen in lungs of B6C3Fi mice following acute inhalation exposure to
concentrations producing lung tumors with chronic exposure, and this effect was suppressed with
pretreatment with a GSH depletor, buthionine sulfoximine (Graves et al., 1995). Other studies of
sister chromatid exchange (Allen et al., 1990) or DNA damage detected by the comet assay
(Sasaki et al., 1998) also provide evidence of genotoxic effects specifically in lung cells of mice.
These in vivo mammalian genotoxicity studies demonstrate site-specific effects correlating to the
dichloromethane-induced tumors in animals. Additional evidence for site specificity comes from
a study in which DNA damage (detected by the comet assay) was enhanced in liver tissue but not
stomach, kidney, brain, or bone marrow 24 hours after oral administration of 1,720 mg/kg
dichloromethane to CD-I mice (Sasaki et al., 1998).
DNA reaction products (e.g., DNA adducts) produced by GST metabolites, such as
S-(chloromethyl)glutathione, have not been identified in in vivo studies (Watanabe et al., 2007).
The authors speculated that these results are due to the instability of the reaction products
(Hashmi et al., 1994). DNA adducts, however, have been observed in in vitro studies in which
calf thymus DNA was incubated with dichloromethane and GST or was incubated with
S-(l-acetoxymethyl)glutathione, a compound structurally similar to S-(chloromethyl)glutathione
(Marsch et al., 2004; Kayser and Vuilleumier, 2001). These findings indicate that the S-
(chloromethyl)glutathione intermediate formed by GSH conjugation has mutagenic potential and
is likely responsible, at least in part, for the mutagenic response observed following
dichloromethane exposure. However, other studies (Hu et al., 2006; Casanova et al., 1996)
provide evidence of formaldehyde-related DNA-protein cross-links in relation to
dichloromethane exposure. These results show that, while most studies indicate the importance
of the S-(chloromethyl)glutathione intermediate in mediating genotoxic damage following
dichloromethane exposure, DNA damage resulting from formaldehyde formation should also be
considered.
Mutagenic data in critical genes leading to the initiation of dichloromethane-induced liver
or lung tumors are not available. In vivo assays evaluating mutations in tumor suppressor genes
233 DRAFT - DO NOT CITE OR QUOTE
-------
and oncogenes reported similar frequencies of activated H-ras genes and inactivation of the
tumor suppressor genes, p53 and Rb-1, in the liver tumors seen in the nonexposed and
dichloromethane-exposed B6C3Fi mice (Devereaux et al., 1993; Hegi et al., 1993). There were
too few lung tumors (n = 4) in controls to provide a conclusive comparison of mutation patterns
between exposed and nonexposed tumors.
Dose-response concordance. Statistically significant increases in liver tumor incidences
in male and female (2,000 and 4,000 ppm) mice were observed in the inhalation bioassay in
B6C3Fi mice (NTP, 1986). Several studies provide evidence of an association between
mutagenicity (as determined by the micronucleus test or presence of chromosome aberrations)
and genotoxicity (as determined by the presence of DNA-protein cross links or DNA SSBs) and
the exposure levels inducing liver tumors in B6C3Fi in this study (Table 4-44).
Concentration-dependent increases in genotoxicity have been observed in in vitro assays
of DNA-protein cross-links, DNA SSBs, and DNA damage as measured by the comet assay at
concentrations ranging from 2.5 to 60 mM when mouse liver cytosol was added or if mouse
GST-T1 was transfected into hamster cell lines (Hu et al., 2006; Graves et al., 1996, 1994b). In
mouse hepatocytes, DNA-protein cross-links were observed following dichloromethane
exposures ranging between 0.5 and 6.0 mM (Casanova et al., 1997). DNA-protein cross-links
were detected in mouse hepatocytes incubated with 1.9 mM dichloromethane (Casanova et al.,
1997), a concentration chosen based on its correspondence to the TWA liver concentration of
dichloromethane that was predicted by the Andersen et al. (1987) PBPK model for mice exposed
by inhalation to 4,000 ppm for 6 hours (a dose that resulted in increased liver tumor incidence in
the 2-year bioassay reported by NTP, 1986).
DNA-protein cross-links were not detected in livers of mice exposed to 146 ppm
6 hours/day for 3 days, but a concentration-dependent increase in DNA-protein cross-links was
observed in DNA from livers of mice exposed to several concentrations between 500 and
4,000 ppm (Casanova et al., 1996). Increased DNA SSBs were detected in liver tissue of
B6C3Fi mice immediately following a 6-hour inhalation exposure to dichloromethane at
concentrations ranging from 2,000 to 8,000 ppm (Graves et al., 1995), and in mouse hepatocytes
after a 3-hour exposure to 4,000 (but not 2,000) ppm (Graves et al., 1994b).
Statistically significant increases in the incidence of lung tumors were observed in the
inhalation chronic bioassay in male and female B6C3Fi mice exposed to 2,000 or 4,000 ppm
dichloromethane (Mennear et al., 1988; NTP, 1986). Evidence of mutagenicity and genotoxicity
at these exposure levels comes from two inhalation studies (Graves et al., 1995; Allen et al.,
1990). In the study by Allen et al. (1990), increased presence of sister chromatid exchanges was
observed in mouse lung cells following a 12-week exposure at 2,000 ppm; shorter durations of
exposure (2 weeks) were positive for measures of sister chromatid exchange and chromosome
aberrations at 8,000 ppm, but not at 2,000 or 4,000 ppm. Increased DNA SSBs were detected in
234 DRAFT - DO NOT CITE OR QUOTE
-------
lung tissue of B6C3Fi mice immediately following a 6-hour inhalation exposure to
dichloromethane at concentrations ranging from 2,000 to 8,000 ppm (Graves et al., 1995).
DNA adducts were observed and increased with dose in an in vitro preparation of calf
thymus DNA when treated with dichloromethane (5-60 mM) and bacterial, rat, or human GST
(Marsch et al., 2004).
Temporal relationship. Dichloromethane-induced liver and lung tumors first appeared in
mice after 52 weeks of exposure (Maronpot et al., 1995; Kari et al., 1993). Tumor formation was
seen with early exposure (i.e., exposure only during the first 26 weeks for lung tumors, or 52
weeks for liver tumors, of a 2-year bioassay), but not with exposure that began at week 52 or 78
(Kari et al., 1993). The detection of DNA-protein cross-links in the livers of B6C3Fi mice
following short-term inhalation exposures to dichloromethane concentrations that induced
tumors with chronic exposure (Casanova et al., 1996, 1992) provides temporal support for the
proposed mutagenic mode of action (Table 4-44). Additional supporting evidence comes from
observations that increased levels of DNA SSBs were detected in the liver and lungs of B6C3Fi
mice immediately following 3-hour inhalation exposure to 2,000-8,000 ppm dichloromethane
(Graves et al., 1995; 1994b). Single dose and inhalation exposure studies of <6 hours did not
detect an effect on DNA synthesis (Lefevre and Ashby, 1989) or unscheduled DNA synthesis
(Trueman and Ashby, 1987) in mouse liver cells.
Biological plausibility and coherence. Bioactivation of a parent compound into a
mutagenic metabolite resulting in cancer is a plausible mode of action of carcinogenicity in
humans and is a generally accepted mode of action. Dichloromethane-induced carcinogenicity is
hypothesized to be due to metabolism of the parent compound by the GST pathway (GST-T1) to
a metabolite that is tumorigenic. The GST metabolite, S-(chloromethyl)glutathione, formed
from dichloromethane, has been characterized as labile and highly reactive through in vitro
evaluation of dichloromethane metabolism in hepatocytes using [13C]-NMR techniques (Hashmi
et al., 1994) and through an enzyme digestion assay using calf thymus DNA and GST-T1
enzyme (Marsch et al., 2004). The hypothesis that the formation of a mutagenic metabolite is a
preliminary step resulting in carcinogenicity is based on evidence that malignant tumors are
primarily located in areas where dichloromethane is highly metabolized by GST-T1, such as the
liver and the lung, and on mutagenicity studies indicating the importance of the GST pathway
and that the lung and liver are more prone to mutagenic effects of dichloromethane (Sasaki et al.,
1998; Casanova et al., 1996, 1992; Graves et al., 1995, 1994b). The site selectivity of the
mutagenicity in liver and lung tissue as evidenced by several studies suggests that the GST
reactive metabolite remains in the tissue where it is formed. Collectively, the studies support the
hypothesis that dichloromethane-mediated carcinogenicity results from a GST metabolite that
produces selective DNA damage in the tissues where the metabolite is formed, but this
23 5 DRAFT - DO NOT CITE OR QUOTE
-------
hypothesis is based in part on assumptions regarding metabolite clearance and reactivity. DNA
damage in the liver and lung, as well as the increased incidence of tumor formation resulting
from dichloromethane exposure, indicates coherence of the mutagenic and carcinogenic effects
and is evidence supporting a mutagenic mode of action.
Differences in GST activity in mice compared with other species, and the interspecies
variability in genotoxic effects corresponding to interspecies variability in tumor response,
support the mode of action hypothesis. DNA SSBs were not detected in liver or lung cells in rats
exposed to similar inhalation exposures that induce strand breaks in mice (Graves et al., 1995;
Graves et al., 1994b) and were detected at much lower in vitro concentrations in isolated
hepatocytes from B6C3Fi mice (0.4 mM) than in hepatocytes from Alpk:APfSD rats (30 mM)
(Graves et al., 1995, Figure 3). The difference in susceptibility to carcinogenic response between
mice and rats likely reflects differences in GST metabolism. Toxicokinetic studies indicate that
with increasing exposure levels, increasing amounts of dichloromethane are metabolized via
GST metabolism.
236 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-44. Experimental support for mutagenic mode of action for
dichloromethane
Criteria
Evidence for dichloromethane
Selected references
Strength,
consistency, and
specificity of
association
[Is there a
statistically
significant
association
between the key
events and tumor
response?]
Presence or activation of GST is necessary for genotoxicity and
mutagenicity in bacterial and mammalian cell systems
Positive in vitro mutagenicity tests in mammalian systems
including the HPRT gene mutation assay in CHO cells with
added GST activity and the micronucleus test in three human
cell lines.
In vivo mammalian studies demonstrate site-specific effects
correlating to the dichloromethane-induced tumors in animals:
DNA-protein cross-links, DNA SSBs, chromosomal
aberrations, and sister chromatid exchanges in liver and lung
cells of B6C3F! mice following acute inhalation exposure to
concentrations producing liver and lung tumors with chronic
exposure
Formation of DNA SSBs was suppressed when the mice were
pretreated with a GSH depletor
DNA damage (detected by the comet assay) after
dichloromethane exposure enhanced in liver tissue but not
stomach, kidney, brain, or bone marrow in CD-I mice
Binding of the reactive GST metabolite, S-
(chloromethyl)glutathione to DNA was not seen in one study in
rats and mice using a relatively low dose (5 mg/kg); postulated
DNA-reactive species considered to be highly reactive and
unstable
In human cell lines (peripheral blood cells), positive results for
sister chromatid exchanges, chromosomal aberrations, DNA
damage (comet assay), and in the micronucleus test (Olvero-
Belloetal., 2010)
Allen etal., 1990;
Casanova et al., 1996,
1992;
Graves etal., 1995,
1994b, 1996;
Dohertyetal., 1996;
Sasaki etal., 1998;
Watanabe et al., 2007;
Olvera-Bello etal., 2010
Dose-response
concordance
[Do the key
events increase
with dose?]
Dose-dependent increase in formation of DNA adducts in vitro
preparation of calf thymus DNA when treated with
dichloromethane (5-60 mM) and bacterial, rat, or human GST
Dose-dependent increase in DNA strand breaks in B6C3F!
mouse hepatocytes
Dose-dependent increased presence of sister chromatid
exchanges in mouse lung cells
Thilagaretal., 1984;
Allen etal., 1990;
Graves etal., 1995;
Dohertyetal., 1996;
Casanova 1997, 1996;
Landietal., 2003;
Marsch et al., 2004
Temporal
relationship
[Does the key
event precede
tumor
appearance?]
Formation of DNA-protein cross-links and SSBs in B6C3F1
mouse liver and lung following short (3 or 6 hour) inhalation
exposure to 2,000-8,000 ppm dichloromethane
DNA damage (detected by the comet assay) in liver in CD-
1 mice, 24 hours after oral administration of 1,720 mg/kg
dichloromethane
Tumor formation seen only with early exposure (i.e., exposure
only during the first 26 weeks for lung tumors, or 52 weeks for
liver tumors, of a 2-year bioassay)
Graves etal., 1994b,
1995;
Casanova et al., 1996,
1992;
Sasaki etal., 1998;
Karietal., 1993
237
DRAFT - DO NOT CITE OR QUOTE
-------
4.7.3.1.2. Other possible modes of action for liver or lung tumors in rodents. Data are not
available to support other possible modes of action for the liver and lung tumors in rodents.
Efforts to observe sustained cell proliferation in liver following dichloromethane exposure of
B6C3Fi mice have been unsuccessful. Groups of female B6C3Fi mice that were exposed to 0 or
2,000 ppm dichloromethane 6 hours/day, 5 days/week for up to 78 weeks did not exhibit
enhanced cell proliferation in the liver when assessed at various intervals during exposure (Foley
etal., 1993).
Indices of enhanced cell proliferation have been measured in the lungs of male B6C3Fi
mice following acute duration exposure at concentrations of about 1,500, 2,500, or 4,000 ppm
dichloromethane (6 hours/day for 2 days) but not at exposure concentrations of 150 or 500 ppm
and not in lungs of Syrian golden hamsters exposed to concentrations up to 4,000 ppm
(Casanova et al., 1996). Earlier studies showed somewhat consistent findings in that the
numbers of bronchiolar cells undergoing DNA synthesis (thymidine incorporation labeling) were
markedly increased (about 6- to 15-fold) in bronchiolar cells of B6C3Fi mice exposed to
4,000 ppm dichloromethane 6 hours/day on days 5, 8, and 9 of exposure, but no evidence for
increased cell proliferation was found after 89, 92, or 93 days of exposure (Foster et al., 1992).
The results suggest that enhanced cell proliferation is not sustained in the lung with longer-term
exposure to dichloromethane concentrations associated with lung tumor development in mice,
and that this mode of tumor promotion is not important in the development of dichloromethane-
induced lung tumors.
4.7.3.2. General Conclusions About the Mode of Action for Tumors in Rodents and
Relevance to Humans
The mode of action for dichloromethane is hypothesized to involve mutagenicity via
reactive metabolites. Mechanistic evidence indicates that dichloromethane-induced DNA
damage in cancer target tissues of mice involves DNA-reactive metabolites produced via a
metabolic pathway initially catalyzed by GST. Although mutational events in critical genes
leading to tumor initiation have not been established, evidence supporting a mutagenic mode of
action includes the identification of mutagenic response (reverse mutations) in short-term
bacterial assays (with microsomal activation) and induced DNA-protein cross-links and DNA
SSBs in mammalian cell assays. There are numerous positive in vivo genotoxicity studies
specifically examining responses in the liver and/or lung; these studies included evidence of
chromosomal aberrations, SSBs, sister chromatid exchanges, and DNA-protein cross-links. The
negative in vivo genotoxicity assays are generally those that were based on a micronucleus test
using mouse bone marrow, which is expected, as halogenated hydrocarbons (such as
dichloromethane) are not very effective in this type of assay (Dearfield and Moore, 2005;
Crebelli et al., 1999).
23 8 DRAFT - DO NOT CITE OR QUOTE
-------
Is the hypothesized mode of action sufficiently supported in test animalsl Consistent and
specific evidence for the association between the formation of DNA-reactive GST-pathway
metabolites and the formation of liver and lung tumors from inhalation includes: (1) enhanced
GST metabolic capacity in the liver and lung and enhanced localization of GST-T1 in hepatic
cell nuclei in B6C3Fi mice compared with rats and hamsters, which do not show strong tumor
responses to chronic inhalation exposure, (2) the detection of DNA-protein cross-links, or DNA
SSBs in livers and lungs of B6C3Fi mice following acute inhalation exposure to concentrations
that produce tumors with chronic exposure, (3) suppression of the formation of DNA SSBs in
livers and lungs of B6C3Fi mice pretreated with a GSH depletory, (4) the inability to detect
DNA-protein cross-links or DNA SSBs in livers or lungs of similarly exposed rats or hamsters,
(5) detection of DNA SSBs at much lower in vitro concentrations in isolated hepatocytes from
B6C3Fi mice than in hepatocytes from Alpk: APfSD rats, (6) dose-response concordance and a
temporal relationship for the formation of DNA-protein cross-links and DNA SSBs with the
formation of liver and lung tumors in B6C3Fi mice exposed to dichloromethane, (7) the
detection of increased sister chromatid exchanges in lung cells from CD-I mice exposed by
inhalation to dichloromethane, and (8) enhancement of dichloromethane genotoxicity in bacterial
and mammalian in vitro assays with the introduction of GST metabolic capacity. However,
mutations in critical genes linked to initiation of tumor cells have not been identified.
The much weaker carcinogenic response in the liver of rats and mice to chronic drinking
water exposure (Serota et al., 1986a, b) than that noted in mice exposed by inhalation (Kari et al.,
1993; NTP, 1986) is correlated with much smaller amounts of GST metabolites produced in the
liver under the exposure conditions of the oral bioassay than in the inhalation bioassay (Andersen
etal., 1987).
In conclusion, there is sufficient evidence supporting a mutagenic mode of action and
indicating the involvement of GST metabolism in the lung and liver carcinogenicity of
dichloromethane in mice.
Is the hypothesized mode of action relevant to humans! The postulated mode of action
that dichloromethane is metabolized by GST to reactive metabolites that induce mutations in
DNA leading to carcinogenicity is possible in humans. Mutagenicity as a mode of action for
carcinogenicity in humans is generally accepted and is a biologically plausible mechanism for
tumor induction. The toxicokinetic and toxicodynamic processes that would enable reactive
metabolites to produce mutations in animal models are biologically plausible in humans.
Furthermore, the detection of the GST pathway in human tissues indicates that the hypothesized
mode of action involving reactive metabolites from this pathway, S-(chloromethyl)glutathione
and formaldehyde, is relevant to humans.
Some investigators question the relevance of the proposed mode of action to humans in
low-exposure scenarios given the high exposure conditions of the genotoxicity and bioassay
239 DRAFT - DO NOT CITE OR QUOTE
-------
studies in mice and the relatively high GST activity in this species (Green et al., 1997).
Comparisons in mice, rats, humans, and hamsters of GST enzyme activity in liver and lung
tissues have indicated the following rank order: mice > rats > or ~ humans > hamsters (Thier et
al., 1998;Reitzetal., 1989).
Underlying questions of human relevance of dichloromethane-induced mouse tumors is
the assumption that, at very low exposures, the amount metabolized through the GST activity in
humans is effectively zero, and so any risk to humans would thus effectively be zero. EPA
considered this line of reasoning, but found that it was not supported by several pieces of
evidence. As discussed in Section 3.3, based on enzyme kinetics, the rate of reaction below
-20% of the Km becomes indistinguishable from a first-order reaction, as it depends on the
probability that a substrate molecule collides with an unoccupied active site on the enzyme.
Thus, at low concentrations (i.e., [substrate] « Km) the rate of enzyme-catalyzed reactions
becomes proportional to the concentration of the substrate(s) and enzyme, and the reaction will
proceed at a non-zero rate as long as GSH, GST, and dichloromethane are present at non-zero
concentrations. The linearity of this metabolism at very low concentrations is discussed in the
section on uncertainties in low-dose extrapolation (Section 5.4.5). At very low exposures, the
amount of dichloromethane metabolized in humans through the GST pathway, while very low, is
not zero.
Another factor noted by Green (1997) that may play a role in the apparent species
differences in carcinogenicity resulting from dichloromethane exposure is species differences in
intracellular localization of GST-T1. Nuclear production of S-(chloromethyl)glutathione
catalyzed by GST-T1 in the nucleus is more likely than cytoplasmic production to lead to DNA
alkylation. Using immunostaining techniques, Mainwaring et al. (1996) demonstrated that in
mouse liver tissue, GST-T1 was localized in the nuclei of hepatocytes and bile-duct epithelium,
whereas the rat and human liver did not show preferential nuclear localization of GST-T1. A
later study by Sherratt et al. (2002) reported that in human tissue samples, bile duct epithelial
cells and some hepatocytes showed nuclear localization of GST-T1, and other hepatocytes
showed localization in cytoplasm. Although the degree of GST-T1 localization in the mouse is
greater than in humans, the finding of some nuclear localization of GST-T1 in human liver tissue
and in the nuclei of bile duct epithelial cells, and the observation of three biliary tract cancers, a
very rare cancer, in a small cohort of dichloromethane exposed workers (Lanes et al., 1993;
1990) support the relevance of the hypothesized mode of action to humans.
Which populations or lifestages can be particularly susceptible to the hypothesized mode
ofactionl As discussed in Section 3.3, a polymorphism of the GST-T1 gene is present in
humans. People with two functional copies of the gene (+/+) readily conjugate GSH to
dichloromethane. Individuals having only one working copy of the gene (+/-) display relatively
decreased conjugation ability. Individuals with no functional copy of the gene (-/-) do not
240 DRAFT - DO NOT CITE OR QUOTE
-------
express active GST-T1 protein and do not metabolize dichloromethane via a GST-related
pathway (Thier et al., 1998). Thus, the GST-T1+ + (wild-type) genotype would be considered to
be the more "at risk" population; this subgroup represents approximately 30% of the U.S.
population (Haber et al., 2002) but would be expected to be more common among Caucasians
and African-Americans than among Asians (Raimondi et al., 2006; Garte et al., 2001; Nelson et
al., 1995) (see Table 3-3).
According to the Supplemental Guidance for Assessing Susceptibility from Early-Life
Exposure to Carcinogens (U.S. EPA, 2005b), children exposed to carcinogens with a mutagenic
mode of action are assumed to have increased early-life susceptibility. The Supplemental
Guidance (U.S. EPA, 2005b) recommends the application of age-dependent adjustment factors
(ADAFs) for carcinogens that act through a mutagenic mode of action. Although the database is
lacking in vivo evidence of specific mutagenic events following chronic exposure to
dichloromethane, the weight of the available evidence indicates that dichloromethane is acting
through a mutagenic mode of carcinogenic action. Application of ADAFs is recommended for
both the oral and inhalation routes of exposure when risks are assessed that are associated with
early-life exposure.
241 DRAFT - DO NOT CITE OR QUOTE
-------
4.8. SUSCEPTIBLE POPULATIONS AND LIFE STAGES
4.8.1. Possible Childhood Susceptibility
In humans, hepatic CYP2E1 begins to be expressed in the second trimester. In a study of
73 fetal liver samples from 7 to 37 weeks gestation, CYP2E1 was not detected in the first
trimester, ranged from 0 to 3 pmol per mg microsomal protein (median 0.35 pmol per mg
microsomal protein) during the second trimester, and increased to a median of 6.7 pmol per mg
microsomal protein during the third trimester (Johnsrud et al., 2003). CYP2E1 continues to
increase during the first year of life, with adult levels reached after age 3 months (Hines, 2007;
Johnsrud et al., 2003; Treluyer et al., 1996; Vieira et al., 1996). In a study using 10 fetal brain
samples collected between 7 and 12 weeks gestation, CYP2E1 activity is seen as early as GD 53,
with increasing levels (from 2.2 to 4.5 pmol produced per mg microsomal protein per hr) seen
through at least GD 113 (Brzezinski et al., 1999). The relatively high activity of CYP2E1 in the
brain compared to the liver of the developing human fetus raises the potential for
neurodevelopmental effects from dichloromethane exposure. Results from a developmental
toxicity study in rats also raise concern for possible neurodevelopmental effects. Decreased
offspring weight at birth and changed behavioral habituation of the offspring to novel
environments were seen following exposure of adult Long-Evans rats to 4,500 ppm for 14 days
prior to mating and during gestation (or during gestation alone) (Bornschein et al., 1980; Hardin
and Manson, 1980). In the only other animal study examining possible early-life susceptibility
to dichloromethane toxicity, Alexeef and Kilgore (1983) found that exposure of young male
mice to approximately 47,000 ppm for about 20 seconds significantly impaired the ability to
learn using a passive-avoidance conditioning task. Three-week-old mice were more affected
than 5- or 8-week-old mice. The broad issue of childhood susceptibility to chronic
neurobehavioral effects of early life exposure represents a data gap in the understanding of the
health effects of dichloromethane.
The relatively low CYP2E1 activity in the liver of young infants (<3 months of age)
would tend to shift metabolism of dichloromethane to the GST pathway. This shift could affect
cancer risk, given the evidence of genotoxicity through this metabolic pathway. However, the
available data in humans are not sufficient to address the question of whether in utero or early
life exposures represent a period of increased susceptibility to potential carcinogenic effects of
dichloromethane. McCarver and Hines (2002) reviewed data pertaining to fetal expression of
GST-M, GST-P, and other classes of enzymes, but information pertaining to GST-T1 was not
included. GST-T1 was not detected in any tissue examined in an 8-week embryo or a 13-week
fetus (Raijmakers et al., 2001), but several studies have reported an interaction between maternal
smoking and fetal GST-T1 null genotype on risk of oral cleft (Shi et al., 2007; Lammer et al.,
2005; van Rooij et al., 2001). These studies indicate the potential relevance of fetal GST-T1
genotype in relation to in utero exposures. A threefold increased risk of childhood leukemia
(acute lymphoblastic leukemia) was seen in relation to maternal occupational exposure to
242 DRAFT - DO NOT CITE OR QUOTE
-------
dichloromethane in the year before and during pregnancy in one population-based case-control
study (OR 3.22 [95% CI 0.88-11.7]) for ratings of "probable or definite" exposure compared
with possible or no exposure (Infante-Rivard et al., 2005). The estimates for categories based on
concentration and frequency were similar, but there was no evidence for an increasing risk with
increasing exposure level. This study did not examine interactions with GST-T1 genotype.
Experiments comparing cancer responses from early-life exposures with those from adult
exposures are not available for F344 rats or B6C3Fi mice, the strains of animals in which
carcinogenic responses to dichloromethane have been observed. Animal data evaluating the
effect of age on the susceptibility to dichloromethane carcinogenicity are restricted to a bioassay
in which 54 pregnant Sprague-Dawley rats were exposed starting on GD 12 to 100 ppm
dichloromethane 4 hours/day, 5 days/week for 7 weeks, followed by 7 hours/day, 5 days/week
for 97 weeks (Maltoni et al., 1988). Groups of 60 male and 69 female newborns continued to be
exposed after birth to 60 ppm dichloromethane 4 hours/day, 5 days/week for 7 weeks, followed
by exposure 7 hours/day, 5 days/week for 97 weeks. Additional groups of 60 male and
70 female newborns were exposed after birth to 60 ppm dichloromethane 4 hours/day,
5 days/week for 7 weeks and then for 7 hours/day, 5 days/week for 8 weeks. Endpoints
monitored included clinical signs, BW, and full necropsy at sacrifice (when spontaneous death
occurred). For each animal sacrificed, histopathologic examinations were performed on the
following organs: brain and cerebellum, zymbal glands, interscapular brown fat, salivary glands,
tongue, thymus and mediastinal lymph nodes, lungs, liver, kidneys, adrenals, spleen, pancreas,
esophagus, stomach, intestine, bladder, uterus, gonads, and any other organs with gross lesions.
There was no significant effect of exposure to dichloromethane on the incidence of benign or
malignant tumors among adults or the progeny. The results provide no evidence that Sprague-
Dawley rats would be more sensitive to potential carcinogenic activity of dichloromethane
during early life stages. Further conclusions from these results are precluded because the study
included only one exposure level, which was below the maximum tolerated dose for adult
Sprague-Dawley rats.
4.8.2. Possible Gender Differences
The limited data available from studies in humans do not indicate that there are large
differences by gender in sensitivity to cardiovascular, neurologic, cancer, or other effects; studies
have not been conducted specifically to examine this question and so do not provide information
pertaining to smaller or more subtle differences. There was no evidence of variation by gender
in hepatic CYP2E1 activity and protein levels in humans based on 238 (136 male, 83 female) samples
(Johnsrud et al., 2003), and the prevalence of the GST-T1 polymorphisms does not vary by gende (Garte
et al., 2001). The available animal studies similarly do not establish whether either gender may be
more susceptible to the toxic effects of dichloromethane. Studies of the carcinogenic effects of
243 DRAFT - DO NOT CITE OR QUOTE
-------
dichloromethane, either by inhalation or by the oral route, have not suggested an increased
susceptibility of either male or female animals.
4.8.3. Other
As discussed in Section 3.3, a polymorphism exists within the GST-T1 gene in humans,
resulting in individuals with diminished or a lack of ability to conjugate GSH to
dichloromethane. While the possible effects of this polymorphism on the toxicity of
dichloromethane have not been directly demonstrated, it can be inferred from the proposed mode
of action that a decrease in the GST-T1 metabolic pathway would result in a decreased
generation of reactive metabolites and a decrease in any chronic effects mediated through those
metabolites (Jonsson and Johanson, 2001; El-Masri et al., 1999).
Interindividual variation in the ability to metabolize dichloromethane via GST-T1 is
associated with genetic polymorphisms in humans. Estimated U.S. population prevalence of
nonconjugators (-/- at the GST-T1 locus) is about 20%, but higher prevalences (47-64%) have
been reported for Asians (Raimondi et al., 2006; Haber et al., 2002; Garte et al., 2001; Nelson et
al., 1995). Although nonconjugators are expected to have negligible extra risk for
dichloromethane-induced cancer, the U.S. prevalences for low (+/- at the GST-T1 locus) and
high (+/+) conjugators have been estimated at 48 and 32%, respectively (Haber et al., 2002).
The liver and kidney are the most enriched tissues in GST-T1, but evidence is available for the
presence of GST-Tlat lower levels in other tissues including the brain and lung (Sherratt et al.,
2002, 1997).
Individuals may vary in their ability to metabolize dichloromethane through the CYP2E1
pathway. Individuals with decreased CYP2E1 activity may experience decreased generation of
CO and an increased level of GST-related metabolites following exposure to dichloromethane,
which may result in increased susceptibility to the chronic effects of dichloromethane from
GST-related metabolites. Conversely, individuals with higher CYP2E1 activity may experience
relatively increased generation of CO at a given dichloromethane exposure level and therefore,
may be more susceptible to the acute CO-related toxicity or other chronic effects of
dichloromethane resulting from this metabolic pathway. These individuals would be expected to
be at a lower risk of chronic effects of dichloromethane from GST-related metabolites. Several
studies indicate a three- to sevenfold variability in CYP2E1 activity among humans, as assessed
by various types of measurements among "healthy" volunteers (Sweeney et al., 2004; Haufroid
et al., 2003; Lipscomb et al., 2003; Lucas et al., 2001, 1999; Bernauer et al., 2000; Kim et al.,
1995; Shimada et al., 1994). This variability is incorporated into the PBPK models for
dichloromethane. Factors that may induce or inhibit CYP2E1 activity (e.g., obesity, alcohol use,
diabetes) or co-exposures (i.e., to various solvents or medications) (Lucas et al., 1999) may result
in greater variation within segments of the population. This variation in CYP2E1 activity may
244 DRAFT - DO NOT CITE OR QUOTE
-------
result in earlier saturation of this pathway and greater exposure to the parent compound, which
would be of particular relevance to neurological effects.
245 DRAFT - DO NOT CITE OR QUOTE
-------
5. DOSE-RESPONSE ASSESSMENTS
5.1. ORAL REFERENCE DOSE (RfD)
5.1.1. Choice of Principal Study and Critical Effect—with Rationale and Justification
As discussed in Section 4.6.1, human data for oral exposures to dichloromethane are
limited to case reports involving intentional (i.e., suicidal) or accidental, acute ingestion
exposures (Chang et al., 1999; Hughes and Tracey, 1993). Reported effects reflect frank toxicity
from very high doses such as marked CNS depression, injury to the gastrointestinal tract, liver
and kidney failure, coma, and death. No studies of human chronic oral exposures are available.
In the absence of adequate studies evaluating possible health effects in humans repeatedly
exposed to dichloromethane via the oral route, the results from the chronic laboratory animal
studies are assumed to be relevant to humans.
The database of laboratory animal oral exposure studies includes 90-day (Kirschman et
al., 1986) and 2-year drinking water toxicity studies in F344 rats (Serota et al., 1986a) and
B6C3Fi mice (Serota et al., 1986b; Hazleton Laboratories, 1983). A reproductive study exposed
Charles River CD rats via gavage before mating (General Electric Company, 1976), and a
developmental study exposed F344 rats via gavage during GDs 6-19 (Narotsky and Kavlock,
1995). A 14-day gavage study examined neurotoxicity in F344 rats (Moser et al., 1995).
Hepatic effects (hepatic vacuolation, liver foci) are the primary dose-dependent
noncancer effects associated with oral exposure to dichloromethane (see Table 4-35). The 90-
day drinking water toxicity study in F344 rats (Kirschman et al., 1986) reported significant
increases in hepatocyte vacuolation and necrosis in animals dosed between 166 and 1,200
mg/kg-day (males) or 200 and 1,469 mg/kg-day (females). These doses were used to develop
dosing levels for the 104-week drinking water study (Serota et al., 1986a). The 104-week
drinking water study of F344 rats (Serota et al., 1986a) provides adequate data to describe dose-
response relationships for liver lesions from chronic oral exposure to dichloromethane (e.g.,
includes four exposure levels and a control group). In this study, rats dosed at >50 mg/kg-day in
both sexes had increased fatty livers, but quantitative data were not provided by the authors.
Liver lesions, described as foci or areas of cellular alteration, were also seen in this study in the
same dose groups in which the fatty changes had occurred. A limitation of this study is that
Serota et al. (1986a) did not describe the evaluation of the altered foci in detail. However,
increases in altered foci did not correspond to tumor rate incidences in either male or female rats.
Instead, the altered foci correlated more closely to fatty liver incidence changes for both sexes in
the rats. Altered foci could range from a focal fatty change (nonneoplastic) to an enzymatic
altered foci change (neoplastic) (Goodman et al., 1994). Several lines of evidence were
considered in determining whether the lesions should be characterized as nonneoplastic or
neoplastic: (1) there is a congruence between the incidence of this lesion and the incidence of
-------
the fatty liver in the study by Serota et al. (1986a), (2) at higher doses, hepatocyte vacuolation
and hepatocyte necrosis were seen (Berman et al., 1995; Kirschman et al., 1986), and (3) there is
no clear indication that these altered foci progress to liver tumors since the rate of increased foci
did not correlate with liver tumor increases in either male or female rats. Based on these
observations, EPA concluded that the altered foci were more likely to be representative of a focal
fatty change (nonneoplastic) than a neoplastic event.
The LOAELs for the liver lesions in rodents following repeated oral exposure (50-586
mg/kg-day) (Table 4-35) are in the same range or below the NOAELs of 225 mg/kg-day for
reproductive performance in Charles River CD rats exposed for 90 days before mating (General
Electric Company, 1976) and 450 mg/kg-day for developmental toxicity in pregnant F344 rats
exposed during gestation (Narotsky and Kavlock, 1995). The LOAEL (337 mg/kg-day) and
NOAEL (101 mg/kg-day) for mild neurological impairment in a 14-day gavage exposure study
of F344 rats (Moser et al., 1995) indicates that the threshold for neurological effects may be
similar to the threshold for liver effects. A limitation of the Moser et al. (1995) study, however,
is that the observed effects were limited to measures taken within 4 hours of exposure.
The subchronic (i.e., <90-day study) data were not considered in the selection of a
principal study for deriving the chronic RfD because the database contains reliable dose-response
data from a chronic study at lower doses than the 90-day study (Kirschman et al., 1986)
(conducted to provide data pertaining to relevant doses to use in the chronic study). The data
from the subchronic studies are, however, used to corroborate the findings in the chronic studies
with respect to relevant endpoints (i.e., hepatic and neurological effects). The neurotoxicity
study was not selected as the principal study due to the limited measurements to inform the
chronic exposure to dichloromethane. The rat rather than the mouse chronic bioassay (Serota et
al., 1986a) was selected as the principal study for the RfD because of the consistent evidence that
rats may be more sensitive than mice to noncancer liver effects from orally administered
dichloromethane; available rat LOAELs for liver lesions are lower than mouse LOAELs (see
Table 4-35). Figure 5-1 is an exposure-response array that presents NOAELs, LOAELs, and the
dose range tested, corresponding to selected health effects from the short-term
(neurotoxicological) and subchronic studies, and from the chronic, reproductive, and
developmental toxicity studies that were evaluated for use in the derivation of the RfD.
-------
zuuu
1400
1200
«* 1000
7
M 8°°
E
£ 600
o
Q
400-
200-
n
t ¥
1 if
Fatty liver
Nonneoplastic (B6C3F1
liver foci (F344 mice, male
rat, male and and
female) - Serota female) -
etal. (1986a) Serota et
al. (1986b)
CHRONIC HEPATIC
i
<
i
1
•
i
1
•
*
Hepatocyte
(f ^ vacuolation
Hepatocyte (B6C3F1 male
vacuolation (F344 mice) -
rat) - Kirschman et Kirschman et
al. (1986) al- (I986)
SUBCHRONIC HEPATIC
1
ft
Neurologic,
Functional
Observational
Battery (F344
rat, female) -
Moser et al.
(1995)
NEUROTOX
ONOAEL
• LOAEL
Th^ vertical lin^s =
range of exposures in
study.
Closed dots ( •) =
used in study
Tit
1
A •
1 1
Reproductive Reproductive Maternal weight
Performance organs; gain (F344 rat, Fetal
(CD rat, male performance pregnant female) Toxicity
and female)- (male Swiss -Narotskyand (F344 rat)
General Webster)- Kavlock(1995) Narotsky
Electric Co. Raje et al. and Kavlo
(1976) (1988) (1995)
REPRODUCTIVE AND DEVELOPMENTAL
Figure 5-1. Exposure response array for oral exposure to dichloromethane.
248
DRAFT - DO NOT CITE OR QUOTE
-------
5.1.2. Derivation Process for Noncancer Reference Values
The toxicity values (oral RfD and inhalation RfC) for noncancer endpoints were derived
by using rat and human PBPK models to calculate internal doses in rats from experimental
exposures and extrapolate points of departure to human equivalent exposures. Figure 5-2
illustrates the process of using the PBPK models for toxicity value derivation. The process for
the RfD and RfC is summarized below, using the example of a noncancer liver effect.
249 DRAFT - DO NOT CITE OR QUOTE
-------
Benchmark Dose Analysis
PBPK Model
Response Data
Estimates of Rodent
Internal Dose
Monte Carlo Sampling from
Distributions of Human Model
Parameters
Probabilistic
Human PBPK Model
(What administered doses will
produce a BMDL10 in a
population?)
Distribution of Human Equivalent Doses (mg/kg) or
Inhalation Concentrations (mg/m3) (Points of
Departure)
95% Lower Bound Estimate of Internal
Dose Associated with a 10% response
Divide by Uncertainty Factors
for Interspecies Toxicodynamic
Variability, Human
Toxicodynamic Variability and
Database Deficiencies)
Oral Reference Doses or
Inhalation References
Concentrations
Recommend lower percentile (e.g.,
1st) to protect sensitive individuals
Figure 5-2. Process for deriving noncancer oral RfDs and inhalation RfCs using rodent and human PBPK
models.
250
DRAFT - DO NOT CITE OR QUOTE
-------
A deterministic PBPK model for dichloromethane in rats was first used to convert rat
drinking water or inhalation exposures to values of an internal liver dose metric (see Appendix C
for details of the rat PBPK model). Available models in EPA benchmark dose (BMD) software
(BMDS) version 2.0 were then fit to the liver lesion incidence data, and internal liver dose data
for rats and BMDio values and their lower 95% confidence limits associated with a 10% extra
risk (BMDLio) were calculated from each of the models. Adequacy of model fit was assessed by
r\
overall % goodness of fit (p > 0.10) and examination of residuals, particularly in the region of the
benchmark response (BMR). The choice of best-fitting model was based on the lowest Akaike's
Information Criterion (AIC) among models with adequate fits (U.S. EPA, 2000b).9
In some situations, the use of a PBPK model can replace the use of the BW075 scaling
factor to account for interspecies differences in toxicokinetics. Whether or not to use a scaling
factor depends on the dose metric that is used. Where PBPK models predict the concentration
(in particular, the AUC) of the proximate causative agent, a scaling factor to account for
interspecies differences is not typically used. That is, it is assumed that if the time-averaged (or
steady-state) concentration of the proximate causative agent predicted by the PBPK model in the
target tissue is the same in the test species as in humans, and the test species was exposed for an
equivalent portion of its lifetime (2 years in rats and mice being equivalent to a 70-year lifetime
in humans), then the resulting risks in the two species are the same. However, when the PBPK
model predicts the rate of production of the agent rather than its concentration, then a BW° 75
scaling factor may be appropriate, depending on what is known or expected regarding the rate of
clearance of the agent or metabolite of interest. Two different scenarios can be considered. If
the metabolite formed is considered to be highly reactive and is unlikely to involve processes or
cofactors for which the rate or availability can be expected to scale allometrically, then it can be
assumed that the rate of clearance (i.e., disappearance due to local reactivity) for this metabolite
per volume tissue is equal in rodents and humans. Thus, in that situation as with the AUC dose
metric, no BW° 75 scaling factor is necessary, although differences in tissue volume fraction in
humans versus rats (as occurs for liver) should be and are accounted for by the PBPK model.
However, if the metabolite is removed by processes that scale allometrically (including
enzymatic reactions or reactions with cofactors whose supply is limited by overall metabolism)
then it is expected that interspecies differences in clearance or removal of the toxic metabolite
follow the generally assumed BW075 scaling for rates of metabolism and blood circulation. In
this case, or in situations in which the reactivity or rate of removal of the metabolite has not been
established, it is appropriate to use a scaling factor based on BW ratios to account for this
difference. In the case of the noncancer liver effects of dichloromethane, very limited
9 If more than one model shares the lowest AIC value, BMDL10 values from these models may be averaged to obtain
a POD. However, this average is not a well-defined lower bound, and should be referred to only as averages of
BMDL10s. U.S. EPA does not support averaging BMDLs in situations in which AIC values are similar, but not
identical, because the level of statistical confidence is lost and because there is no consensus regarding a cut-off
between similar and dissimilar AIC values.
251 DRAFT - DO NOT CITE OR QUOTE
-------
information is available on the mechanism(s) involved in creating the type of hepatic damage
seen. The dose metric used in the PBPK modeling is a rate of metabolism rather than the
concentration of putative toxic metabolites, and the clearance of these metabolites may be slower
per volume tissue in the human compared with the rat. Thus, the rat internal dose metric for
noncancer effects was adjusted by dividing by a pharmacokinetic scaling factor to obtain a
human-equivalent internal BMDLio.
A probabilistic PBPK model for dichloromethane in humans, adapted from the model of
David et al. (2006) as described in Appendix B, was then used to calculate distributions of
chronic exposures associated with the human equivalent internal BMDLio, based on the
responses in rats. Parameters in the human PBPK model are described by distributions that
incorporate information about dichloromethane toxicokinetic and physiological variability and
uncertainty among humans, with additional information on human variability for both the
CYP2E1 and GST-T1 metabolic pathways (see Table 3-9 and Appendix B). Monte Carlo
sampling was performed in which each human model parameter was defined by a value
randomly drawn from its respective parameter distribution. The model was then executed by
using the human internal BMDLio as input, and the resulting human equivalent dose or human
equivalent concentration (HEC) was recorded. This process was repeated for 10 to
20,000 iterations to generate a distribution of human equivalent doses or concentrations.
As discussed in Section 3.5.2, the statistics reported for the fitted metabolic parameters
by David et al. (2006; Table 4 in that publication) only represent the population mean and
uncertainty in that mean for each parameter. For the parameters other than Vmaxc and kfc, EPA
considers it reasonable to assume that the there is little true interindividual variability in the
values, so the distributions were used as published in David et al. (2006). For the physiological
parameters, the distributions presented by David et al. (2006) were supposed to represent a
known range of interindividual variability, but EPA found that these did not adequately describe
the full population, so many of the distributions were changed as discussed in Appendix B. For
Vmaxc, an independent data set (Lipscomb et al., 2003) where CYP2E1 levels were measured in
vitro using liver samples from 75 human donors was used to estimate the degree of
interindividual variability, and a "two-dimensional" sampling routine was used to incorporate the
uncertainty as estimated by David et al. (2006) from those in vitro data. Finally, EPA concluded
that the trivariate distribution (based on GST-T1 genotype), which David et al. (2006) used in
place of the observed parameter uncertainty (based on ex vivo data from Warholm et al., 1994),
adequately represented interindividual variability but neglected the uncertainty in the population
mean. Therefore, a two-dimensional sampling routine was used: first a specific value for the
population mean was sampled from the mean and variance (uncertainty) indicated for kfc in
Table 4 of David et al. (2006); second, given that value for the population mean, an individual
value was sampled using the trivariate distribution as indicated in Table 2 of David et al. (2006),
but re-scaled to the (sampled) population mean.
252 DRAFT - DO NOT CITE OR QUOTE
-------
The sampling routine used by EPA effectively assumes that the parameters are distributed
independently, ignoring the covariance that was likely represented in the actual posterior chains,
hence will tend to over-estimate the overall range of parameters and distribution of dose metrics
in the population compared to what one would obtain if the covariance were explicitly included.
(This is offset to some extent by the assumption that there is no interindividual variability in the
metabolic parameters other than Vmaxc and kfc.) Thus if the covariance (i.e., the variance-
covariance matrix) for the set of parameters had been reported by David et al. (2006) it could
have been used to narrow the predicted distribution of internal doses or equivalent applied doses.
Lacking such information, the approach used will not underestimate risk or overestimate lower
bounds on human equivalent exposure levels.
From these distributions of human equivalent doses (or concentrations), candidate RfDs
or RfCs were derived by dividing the first percentile value (point of departure) by uncertainty
factors (UFs) to account for uncertainty about potential interspecies toxicodynamic variability,
human toxicodynamic variability, and database deficiencies. The first percentile was chosen
because it allowed generation of a stable estimate for the lower end of the distribution while
being protective of the overall human population, including sensitive individuals. Choosing this
lower point replaces the use of an additional UF to account for human toxicokinetic variability.
In summary, the PBPK model that is used only allows one to estimate the dose rate at
which the metabolites are introduced to the target tissues, i.e., the oral or inhalation exposure rate
(mg/kg-day inhaled or ingested), and not the rate of clearance of these metabolites from the
tissue. The parent dichloromethane PBPK model is used to estimate the relationship between
dichloromethane exposure and the metabolite exposure rate that occurs in the animal bioassay,
and to back-estimate the human dichloromethane exposure (oral dose rate or inhalation
concentration) distribution that gives rise to the internal human equivalent metabolite dose rate.
Because the dose metric used in this model is an exposure rate, a scaling factor or "dosimetric
adjustment factor" using BW° 75 scaling is applied for animal-to-human extrapolation.
Accordingly, only the toxicodynamic component of the UF for animal-to-human extrapolation
(UFA = 3) is applied in deriving the RfD (see Section 5.1.5). Use of the 1st percentile of the
distribution of the human equivalent dose (or concentration) replaces the toxicokinetic
component of the UF for human variability, and only the toxicodynamic component of this UF is
applied (UFn = 3). As discussed further in Section 5.1.5, the UFs for animal-to-human
extrapolation and human variability in response do not include adjustment for toxicokinetic
uncertainty.
5.1.3. Evaluation of Dose Metrics for Use in Noncancer Reference Value Derivations
There are no data to support the role of a specific metabolite in the development of the
noncancer liver lesions seen in oral and inhalation exposure studies. Four dose metrics were
examined as potential metrics for the internal dose of interest: rate of hepatic metabolism
253 DRAFT - DO NOT CITE OR QUOTE
-------
through the CYP pathway, rate of hepatic metabolism through the GST pathway, the combined
rate of hepatic metabolism through the CYP and GST pathways, and the concentration (AUC) of
dichloromethane (the parent compound) in the liver. The dose-response patterns for each of
these metrics in the oral study in rats (Serota et al., 1986a) and in two inhalation studies in rats
(Nitscke et al., 1988a; Burek et al., 1984) were examined for fit and congruence.
Using the oral exposure data, only one of the seven models, the log-logistic model,
produced an adequate fit (p > 0.10) for the GST metabolism metric and the dichloromethane
AUC metrics. Adequate model fit was seen in all of the models using the CYP dose metric with
the oral data and using the GST, CYP, and AUC dose metrics for the inhalation data.
A limitation in using the GST metric can be observed when comparing the oral and
inhalation responses at various exposure levels. At 200 ppm, where the GST metric is predicted
by the PBPK model to be 93 mg metabolism per liter liver per day, no liver effects were seen. In
contrast, liver responses were elevated at an oral dose of 50 mg/kg-day, where the GST metric is
predicted to be approximately 85 mg metabolism per liter liver per day (see Tables 5-1 and 5-5,
respectively, for the oral and inhalation internal metrics). Thus the liver GST metric produces an
inconsistency in the dose-response relationship with different responses observed depending on
the route of exposure. A similar inconsistency occurs with the AUC metric. These differences
are not observed, however, when using the CYP metric. At the 200 ppm inhalation exposure,
where no hepatoxicity was observed, the CYP metric is predicted to be 665 mg per liter liver per
day. This internal CYP metabolism metric is less than that predicted for the oral dose for the 50
mg/kg-day group (i.e., 724 and 801 mg metabolism per liter liver per day in males and females,
respectively) in which liver effects were observed. Thus, the CYP internal metric is consistent
with the observed responses seen in the oral and inhalation exposure studies.
The GST metabolism and the AUC dose metrics did not present reasonable choices based
on model fit and consistency of response across studies at comparable dose levels. Given these
results, the combination of hepatic metabolism through the GST and the CYP pathways would
not be expected to result in an improvement to a metric based only on CYP metabolism. Thus
the CYP-metabolism dose metric is the most consistent with the data and this metric was
selected for the subsequent RfD and RfC derivations. The lack of information on mechanisms
with respect to noncancer health effects represents data gaps in the understanding of the health
effects of dichloromethane.
254 DRAFT - DO NOT CITE OR QUOTE
-------
5.1.4. Methods of Analysis—Including Models (PBPK, BMD, etc.)
PBPK models for dichloromethane in rats were described previously in Section 3.5.
From the evaluation described in Appendix C, a modified model of Andersen et al. (1991) was
selected for the calculation of internal dosimetry of ingested dichloromethane in the rats in the
principal study (Serota et al., 1986a).
PBPK model simulations of the drinking water study of Serota et al. (1986a) (Table 5-1)
were performed to calculate average lifetime daily internal liver doses in male and female
F344 rats. In the absence of data for group- and sex-specific BWs, reference values were used
for F344 rats in chronic studies (U.S. EPA, 1988a). The mode of action by which
dichloromethane induces noncancer liver effects in rodents has not received research attention to
determine the role of the parent material, metabolites of the CYP2E1 pathway, metabolites of the
GST pathway, or some combination of parent material and metabolites. In the absence of this
kind of knowledge, and considering the pattern of response seen in the oral and inhalation studies
(as described in Section 5.1.3), an internal dose metric based on the amount of dichloromethane
metabolized via the CYP pathway in the liver (mg dichloromethane metabolized via CYP
pathway per liter liver per day) was used. Figure 5-3 shows the comparison between oral
external and internal doses using this dose metric for the rat and for the human.
255 DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-1. Incidence data for liver lesions and internal liver doses based on
various metrics in male and female F344 rats exposed to dichloromethane in
drinking water for 2 years
Sex
Male
(BW =
380 g)
Female
(BW =
229 g)
Nominal (actual)
daily intake
(mg/kg-d)
0(0)
5(6)
50 (52)
125 (125)
250 (235)
0(0)
5(6)
50 (58)
125 (136)
250 (263)
Rat liver
lesion incidence3
52/76 (68%)
22/34 (65%)
35/38 (92%)c
34/35 (97%)c
40/41 (98%)c
34/67(51%)
12/29 (41%)
30/41 (73%)c
34/38 (89%)c
31/34(91%)c
Rat internal liver doseb
CYP
0
131.6
723.9
1,170.5
1,548.0
0
132.8
801.4
1,261.5
1,672.4
GST
0
2.60
81.5
276.5
612.1
0
2.47
93.3
307.8
705.6
GST and
CYP
0
134.2
805.3
1,446.9
2,160.1
0
135.3
894.7
1,569.3
2.378.0
Parent
AUC
0
0.58
18.1
61.3
135.7
0
0.47
17.8
58.6
134.4
"Liver foci/areas of cellular alteration; number affected divided by total sample size.
blnternal doses were estimated using a rat PBPK model from simulations of actual daily doses reported by the study
authors. CYP dose is in units of mg dichloromethane metabolized via CYP pathway/L tissue/d; GST dose is in
units of mg dichloromethane metabolized via GST pathway/L tissue/d; GST and CYP dose is in units of mg
dichloromethane metabolized via CYP and GST pathways/L tissue/d; and parent AUC dose is in units of mg
dichloromethane x hrs/L tissue.
Significantly (p < 0.05) different from control with Fisher's exact test.
Source: Serotaetal. (1986a).
256
DRAFT - DO NOT CITE OR QUOTE
-------
~o >
^3
OJ^
=s
p"g
N N
= = 1,000
-Q-S
roS
*J Qj
8
yu
U°
QOl
D>C.
E 100
.-.-*.-.„ L H
— Rat
• Human mean
* Human 5th %
+ Human 95th %
c—-H
"T T T T"
*r
10
100
Oral dose (mg/kg/d)
Six simulated daily drinking water episodes are described by Reitz et al. (1997).
The human metabolism rates were estimated using a computational sample of
1,000 individuals per dose, including random samples of the three GST-T1
polymorphisms (+/+, +/-, -/-) in the current U.S. population based on data from
Haber et al. (2002). Since a different set of samples was used for each dose, some
stochasticity is evident as the human points (values) do not fall on smooth curves.
Figure 5-3. PBPK model-derived internal doses (mg dichloromethane
metabolized via the CYP pathway per liter liver per day) in rats and humans
and their associated external exposures (mg/kg-day), used for the derivation
of RfDs.
The seven dichotomous dose-response models in BMDS version 2.0 were fit to the rat
liver lesion incidence data and PBPK model-derived internal dose data to derive a rat internal
BMD10 and corresponding BMDL10 associated with 10% extra risk (Table 5-2). The quantal
model is identical to the one-stage multistage model and so is not included in this set of models.
A BMR of 10% was selected because, in the absence of information regarding the magnitude of
change in a response that is thought to be minimally biologically significant, a BMR of 10% is
generally recommended since it provides a consistent basis of comparison across assessments.
There are no additional data to suggest that the critical response has a greater sensitivity that
would warrant a lower BMR. The male rats exhibited a greater sensitivity compared to the
female rats (based on lower BMDLio values for all of the models examined), and thus the male
data are used as the basis for the RfD derivation. The logistic model was the best fitting model
for the male incidence data based on AIC value among models with adequate fit (U.S. EPA,
2000b). Modeling results are shown in detail in Appendix D-l.
257
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-2. BMD modeling results for incidence of liver lesions in male and
female F344 rats exposed to dichloromethane in drinking water for 2 years,
based on liver-specific CYP metabolism dose metric (mg dichloromethane
metabolism via CYP pathway per liter liver tissue per day)
Sex and model"
BMD10
BMDL10
x2
goodness of fit
/7-value
AIC
Males
Gamma3
Logistic1"
Log-logistic3
Multistage (I)3
Probit
Log-probit3
Weibull3
138.12
70.69
188.81
57.08
81.83
176.71
105.85
41.28
51.42
38.74
39.68
62.57
66.89
40.69
0.58
0.69
0.80
0.64
0.64
0.77
0.52
185.46
183.85
184.85
184.05
184.01
184.93
185.67
Females
Gamma3
Logistic
Log-logistic3
Multistage (I)3
Probit
Log-probit3
Weibull3
287.41
137.58
340.15
100.41
145.33
336.41
240.71
81.95
109.45
95.35
74.33
118.47
143.31
80.40
0.49
0.53
0.57
0.40
0.53
0.57
0.43
233.19
231.99
232.88
232.72
231.97
232.87
233.43
"These models in EPA BMDS version 2.0 were fit to the rat dose-response data shown in Table 5-1 by using
internal dose metrics calculated with the rat PBPK model. Details of the models are as follows: Gamma and
Weibull models restrict power >1; Log-logistic and Log-probit models restrict to slope >1, multistage model
restrict betas >0; lowest degree polynomial with an adequate fit is reported (degree of polynomial noted in
parentheses).
bBolded model is the best-fitting model in the most sensitive sex (males), which is used in the RfD derivation.
Source: Serotaetal. (1986a).
The BMDLio from the logistic model was used as the point of departure (POD) for the
RfD calculations (Table 5-3). This rat internal dose metric for noncancer effects was adjusted to
obtain a human-equivalent internal BMDLio by dividing by a pharmacokinetic scaling factor
n 9s
based on a ratio of BWs (BWhUman/BWrat) ' = 4.09). This scaling factor was used because the
metric is a rate of metabolism rather than the concentration of putative toxic metabolites, and the
clearance of these metabolites may be slower per volume tissue in the human compared with the
rat (that is, total rate of removal may scale as BW°'75, while tissue volume scales as BW1).
258
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-3. RfD for dichloromethane based on PBPK model-derived
probability distributions of human drinking water exposures extrapolated
from liver lesion incidence data for male rats exposed via drinking water for
2 years, based on liver-specific CYP metabolism dose metric (mg
dichloromethane metabolized via CYP pathway per liter liver tissue per
day)
Model3
Logistic
Rat
internal
BMDL10b
51.42
Human
internal
BMDL10C
13.31
Human equivalent dose
(mg/kg-d)d
1st
percentile
0.189
5th
percentile
0.225
Mean
0.350
Human
RfD
(mg/kg-d)e
6 x 10'3
aBased on the best-fitting model from Table 5-2.
bRat dichloromethane PBPK model-derived internal liver dose associated with the lower bound on 10% extra risk
for developing liver foci/areas of cellular alteration.
°Human dichloromethane internal liver dose, derived by dividing the rat internal BMDL10 by a scaling factor of
4.09 [(B Whuman/B Wrat)°25] to account for potential interspecies pharmacokinetic differences in the clearance of
metabolites.
dPBPK model-derived distributions of daily average dichloromethane drinking water doses predicted by the PBPK
model to yield an internal dose in humans equal to the dichloromethane internal BMDL10.
eHuman RfD, based on male rat data, derived by dividing the 1st percentile of human equivalent dose value by a
total UF of 30: 3 (10°5) for possible toxicodynamic differences between species, 3 (10°5) for variability in human
toxicodynamic response, and 3 (1005) for database deficiencies. The 1st percentile POD is a stable estimate of the
lower end of the distribution. Use of this value in the lower tail replaces use of a UF for human toxicokinetic
variability. See Section 5.1.5 for discussion of UFs.
Source: Serotaetal. (1986a).
The human PBPK model (adapted from David et al. [2006], as described in Appendix B),
using Monte Carlo sampling techniques, was used to calculate quantiles of human equivalent
administered oral daily doses (in mg/kg-day) associated with the internal BMDLio values
(Table 5-3), as described above in Section 5.1.2. The human model used parameter values
derived from Monte Carlo sampling of probability distributions for each parameter, including
MCMC-derived distributions for the metabolic parameters for the metabolism through the
CYP2E1 pathway (Vmax and Km), and a distribution of GST metabolic rate constants that is
weighted to reflect the estimated frequency of GST-T1 genotypes (20% GST-T1", 48%
GST-T1+/", and 32% GST-T1+/+) in the current U.S. population based on data from Haber et al.
(2002). All simulations also included a distribution of CYP activity based on data from
Lipscomb et al. (2003). The drinking water exposures comprised six discrete drinking water
episodes for specified times and percentages of total daily intake (Reitz et al., 1997). The mean
and two lower points on the distributions of human equivalent administered daily doses derived
from the Serota et al. (1986a) data for male rats, using the BMDLio from the logistic model, are
shown in Table 5-3. Although a lower value in this distribution could be calculated, this would
require proportionately greater iterations (i.e., up to 10,000) to achieve numerical stability.
259
DRAFT - DO NOT CITE OR QUOTE
-------
5.1.5. RfD Derivation—Including Application of Uncertainty Factors (UFs)
The 1st percentile POD is a numerically stable estimate of the lower end of the
distribution. Use of this value associated with a sensitive human population addresses the
uncertainty associated with human toxicokinetic variability. To derive the candidate RfD based
on data from male rats, the first percentile value of the distribution of human equivalent dose
associated with the male rat-derived BMDLio was divided by a composite UF of 30 (3 [10°5] to
account for uncertainty about interspecies toxicodynamic equivalence, 3 [10°5] to account for
uncertainty about toxicodynamic variability in humans, and 3 [10°5] for database deficiencies)
(Table 5-3). The resulting RfD recommended for dichloromethane is 6 x 10~3 mg/kg-day.
In deriving this RfD, factors for the following areas of uncertainty were considered:
• Uncertainty in extrapolating from laboratory animals to humans (UFA). The use of
PBPK models to extrapolate internal doses from rats to humans reduces toxicokinetic
uncertainty in extrapolating from the rat liver lesion data but does not account for the
possibility that humans may be more sensitive than rats to dichloromethane due to
toxicodynamic differences. A UF of 3 (10°5) to account for this toxicodynamic
uncertainty was applied, as shown in Table 5-3.
• Uncertainty about variation from average humans to sensitive humans (UFn). The
probabilistic human PBPK model used in this assessment incorporates the best
available information about variability in toxicokinetic disposition of
dichloromethane in humans but does not account for humans who may be sensitive
due to toxicodynamic factors. Thus, a UF of 3 (10°5) was applied to account for
possible toxicodynamic differences in sensitive humans.
• Uncertainty in extrapolating from LOAELs to NOAELs (UFL,). A UF for
extrapolation from a LOAEL to a NOAEL was not applied because BMD modeling
was used to determine the POD, and this factor was addressed as one of the
considerations in selecting the BMR. The BMR was selected based on the
assumption that it represents a minimum biologically significant change.
• Uncertainty in extrapolating from subchronic to chronic durations (UFs). The derived
RfD is based on results from a chronic-duration drinking water toxicity study. No
cross-duration UF is necessary.
• Uncertainty reflecting database deficiencies (UFD). The oral database for
dichloromethane includes well-conducted chronic drinking water studies in rats
(Serota et al., 1986a) and mice (Serota et al., 1986b) and a supporting subchronic
study in rats and mice (Kirschman et al., 1986). These studies provided dose-
response data for the hepatic effects of dichloromethane. The database also includes
one-generation oral reproductive toxicity (General Electric Company, 1976) and
developmental toxicity (Narotsky and Kavlock, 1995) studies that found no
reproductive or developmental effects at dose levels in the range of doses associated
with liver lesions. A two-generation oral exposure study is not available; however, a
two-generation inhalation exposure study by Nitschke et al. (1988a) reported no
260 DRAFT - DO NOT CITE OR QUOTE
-------
effect on fertility index, litter size, neonatal survival, growth rates, or histopathologic
lesions at exposures of >100 ppm. This study is limited in its ability to fully evaluate
reproductive and developmental toxicity because exposure was not continued
throughout the gestation and nursing periods.
No oral exposure studies that evaluated neurobehavioral effects in offspring were
identified. This is a relevant endpoint given the neurotoxicity associated with
dichloromethane exposure after oral and inhalation exposures (see Section 4.4.3 for
references), and the observed behavioral changes following inhalation developmental
exposure to dichloromethane (Bornschein et al., 1980; Hardin and Manson, 1980).
Dichloromethane is capable of crossing the placental barrier and entering fetal
circulation (Withey and Karpinski, 1985; Anders and Sunram, 1982), and activity of
CYP2E1 in the brain is relatively high compared to the liver of the developing human
fetus (Hines, 2007; Johnsrud et al., 2003; Brzezinski et al., 1999). Dichloromethane,
as well as the metabolite CO, has been implicated in neurological effects (reviewed in
Section 4.6.3.3, Mode of Action for Neurological Effects). However, PBPK
modeling predicts that CO levels from exposures around the RfD would not be high
enough to result in neurodevelopmental toxicity.
There are no oral exposure studies that include functional immune assays;
however, there is a 4-week inhalation study of potential systemic immunotoxicity that
found no effect of dichloromethane exposure at concentrations up to 5,000 ppm on
the antibody response to sheep red blood cells (Warbrick et al., 2003). The Warbrick
et al. (2003) data suggest that systemic immunosuppression is not a concern for
dichloromethane exposure. Because of concerns regarding the lack of an oral two-
generation reproductive study, limitations in the available inhalation two-generation
reproductive study, and the adequacy of available data pertaining to possible
neurodevelopmental toxicity, a UFo of 3 was applied.
5.1.6. Previous RfD Assessment
The previous IRIS assessment derived an RfD of 0.06 mg/kg-day based on the NOAELs
of 5.85 and 6.47 mg/kg-day for liver toxicity (foci/areas of cellular alteration) in male and female
rats, respectively, in a 2-year drinking water study (Serota et al., 1986a). The LOAELs
associated with these NOAELs were 52.58 and 58.32 mg/kg-day for males and females,
respectively. The RfD of 0.06 mg/kg-day was derived by dividing the average NOAEL of 6
mg/kg-day (for male and female rats) by a UF of 100 (10 for intraspecies variability and 10 for
interspecies variability).
5.1.7. RfD Comparison Information
Use of the mean value (3.50 x 10"1 mg/kg-day) of the human equivalent dose distribution
instead of the 1st percentile, with an additional UF of 3 (10°5) to account for human toxicokinetic
variability, would yield an RfD of 4 x 10"3 mg/kg-day.
Additional comparisons between the derived RfD and values developed from other
endpoints or data sets using NOAEL/LOAEL methods are shown in Table 5-4 and Figure 5-4.
NOAELs were used as comparison points of departure and were not scaled allometrically.
261 DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-4. Potential points of departure with applied UFs and resulting candidate RfDs
Endpoint
Alterations in
liver foci,
male ratsb
Neurological
changes
(FOB), female
rats
Maternal
weight gain,
female rats
POD
(mg/kg-d)
0.189
101
338
POD type and description
1s* percentile human
equivalent dose based on
BMDLio for increase in
incidence of liver lesion
NOAEL; No effect at POD,
approximate doubling of
severity score of
neuromuscular and
sensorimotor domains
NOAEL; No effect at POD,
approximate 33% decrease
in weight gain seen at next
dose
UFs applied3
Total UF
30
3,000
300
UFA
3
10
10
UFH
3
10
10
UFL
1
1
1
UFS
1
10
1
UFD
3
3
o
J
Candidate RfD
(mg/kg-d)
6 x 10 3
3.4 x 10"2
1.1
Reference
Scrota et al.
(1986a)
Moser et al.
(1995)
Narotsky and
Kavlock (1995)
aA UF for extrapolation from a LOAEL to NOAEL (UFL) was not used for any of these studies. For the Serota et al. (1986a) study, the use of the first percentile of
the human equivalent dose distribution as the POD replaces the use of a UFH for human toxicokinetic variability.
folded value is the basis for the RfD of 6 x io~3 mg/kg-d.
POD = point of departure; UFA = uncertainty in extrapolating from laboratory animals to humans; UFH = uncertainty about variation from average humans to
sensitive humans; UFL = uncertainty about extrapolating from LOAEL to NOAEL; UFS = uncertainty in extrapolating from subchronic to chronic durations; UFD =
uncertainty reflecting database deficiencies
262
DRAFT - DO NOT CITE OR QUOTE
-------
Point of Departure
UFA - Interspecies;
animal to human
UFH - Intraspecies;
human variability
UFs - Subchronic to
chronic exposure
duration
UFD - Database
Reference Dose
10
Nonneoplastic liver foci
- first percentile Human
Equivalent Dose from
rat;
Serotaetal. (1986a)
Neurologic, Functional
Observational Battery -
NOAEL from rats; Moser
etal. (1995)
Maternal weight gain
- NOAEL from rats;
Narotsky and
Kavlock(1995)
Figure 5-4. Comparison of candidate RfDs derived from selected points of departure for endpoints presented in
Table 5-4.
263
DRAFT - DO NOT CITE OR QUOTE
-------
5.2. INHALATION REFERENCE CONCENTRATION (RfC)
5.2.1. Choice of Principal Study and Critical Effect—with Rationale and Justification
Figure 5-5 includes exposure-response arrays from some of the human studies that were
evaluated for use in the derivation of the RfC. Several acute-duration controlled exposure
studies (Section 4.1.2.2) and cross-sectional occupational studies (Sections 4.1.2.3 and 4.1.2.4) in
humans are available that show neurological effects from dichloromethane exposure. These
effects include an increase in prevalence of neurological symptoms among workers (Cherry et
al., 1981) and possible detriments in attention and reaction time in complex tasks among retired
workers (Lash et al., 1991). However, these studies have inadequate power for the detection of
effects with an acceptable level of precision. In addition, the Cherry et al. (1981) study is limited
by the definition and documentation of neurological symptom history. The Lash et al. (1991)
study has exposure measurements from 1974 to 1986, but the work histories of exposed workers
go back to the 1940s. In addition, the outcome assessment was conducted a mean of 5 years
after leaving the workplace, and so would underestimate health effects that diminish post-
exposure. Ott et al. (1983c) reported an increase in serum bilirubin among exposed workers, but
there was no association seen with respect to the other hepatic enzymes examined (serum y-
glutamyl transferase, serum AST, serum ALT), and no evidence of hepatic effects was seen in a
later study of the same cohort (Soden, 1993). Because of these limitations, these human studies
of chronic exposures do not serve as an adequate basis for RfC derivation. As discussed in
Section 5.2.6, however, the quantitative measures of neurological function from Cherry et al.
(1983) and Lash et al. (1991) were used to derive a comparative RfC.
264 DRAFT - DO NOT CITE OR QUOTE
-------
10000
fl
? T
"1
> <
1 <
1 I \
• V
, 1
1 1
1 1 Hepatocyte Hepatocyte Renal tubular
3 $ S $ vacuolation, vacuolation, degeneration,
Hepatocyte Hepatocyte male and necrosis, male and
vacuolation, necrosis, female hemosiderosis female F344
Sprague- Sprague- Sprague- in liver, male rat(Mennear
Dawleyrat Dawleyrat Dawleyrat andfemale etal., 1988;
(Nitschkeet (Bureket (Bureket F344 rat NTP, 1986)
al., 1988a) al., 1984) al., 1984) (Mennear et al.,
1988; NTP,
1986)
I I
Hepatocyte Renal
degeneration, tubule casts,
male and mlae and
female female
B6C3F1 mice B6C3F1
(Mennear et mice
al., 1988; (Mennear et
NTP, 1986) al., 1988;
NTP, 1986)
ONOAEL BLOAEL
The vertical lines = range of
exposures in study.
Closed dots (0) = exposure
concentrations used in study
I T 1
I [
Changes Chronic Cardiac -
inCNS neurological ST segment
measures, effects, depression,
males males males (Ott
(Lash et (Cherry et et al.,
al., 1991) al., 1983) 1983c)
1000
100
0)
u
o
10
RAT
MOUSE
HUMAN
Figure 5-5. Exposure response array for chronic (animal) or occupational (human) inhalation exposure to
dichloromethane (log Y axis).
265
DRAFT - DO NOT CITE OR QUOTE
-------
The database of experimental animal dichloromethane inhalation studies includes
numerous 90-day and 2-year studies, with data on hepatic, pulmonary, and neurological effects,
(see Table 4-36) and reproductive and developmental studies (Table 4-37) (see summary in
Section 4.6.2). NOAELs, LOAELs, and the dose range tested corresponding to selected health
effects from the chronic studies are shown in Figure 5-5, and effects seen in subchronic,
reproductive, and developmental studies are shown in Figure 5-6. The subchronic (i.e., <90-day
study) data were not considered in the selection of a principal study for deriving the RfC because
the database contains reliable dose-response data from the chronic study at lower doses than the
90-day study. The data from the subchronic studies are, however, used to corroborate the
findings with respect to relevant endpoints (i.e., hepatic and neurological effects).
266 DRAFT - DO NOT CITE OR QUOTE
-------
10,000
1 000
Jl ., \J\J\J
-— v
E
a.
-5? 100
^^ L\J\J
i
1
w 10-
C
O
J
1-
ii
X
^ i i
• i
|
|
|
Ilipidliver Hepatocyte
weight centrilobular
ratios (F334 degeneratioi
rat, (J and (B6C3F1
J) - NTP, mice, (J and
1986 5) - NTP,
1986
HEPATIC
|
|
(
(
f .
i
)
1
Foreign Clara cell
body vacuolation
pneumonia (B6C3F1
(F344rat, mice, (J and
6" and?)- ¥)- Foster
NTP, 1986 etal., 1992
PULMONARY
O
•
I
Increased Increased IgM
Infection response
Susceptibility, (Sprague-
(CD1 mice, Dawleyrat,cT
$) - Aranyi et and ^) —
al, 1986 Waibrick et «1,
2003
IMMUNO
y\
1
1
s
\
I
FOB, Grip
Strength,
SEPs, (F344
rat, 3 and?)
- Mattsson
etal., 1990
NEURO
ONOAEL
• • LOAEL
The vertical lines =
* range of exposures in
• • Y study-
Closed dots (•) =
exposure concentrations
< > used in study
0 < I
s/ * •
Adverse T Maternal T Maternal Fetal body Reproductive Reproductive
fetal effects liver weight liver weight; weight and performance; organs;
| + i fetal histopathology growth rates; performance
1 . 1 bw/altered (Sprague- organ (
-------
Hepatic effects (hepatic vacuolation and necrosis, hemosiderosis, hepatocyte
degeneration) are the primary dose-dependent noncancer effects associated with inhalation
exposure to dichloromethane. These effects were seen in mice (Mennear et al., 1988; NTP,
1986) and rats (Mennear et al., 1988; Nitschke et al., 1988a; NTP, 1986; Burek et al., 1984) but
not in Syrian golden hamsters (Burek et al., 1984). Inhalation bioassays with Sprague-Dawley
rats identified the lowest inhalation LOAEL for liver lesions in the database: 500 ppm
(6 hours/day, 5 days/week for 2 years) (Nitschke et al., 1988a; Burek et al., 1984); Nitschke et al.
(1988a) identified a NOAEL of 200 ppm in female rats. Based on the results reviewed above,
liver lesions (specifically, hepatic vacuolation) in rats are identified as the critical noncancer
effect from chronic dichloromethane inhalation in animals. Hepatocyte vacuolation was
considered a lexicologically relevant effect since the effect was characterized as correlating with
fatty change (Burek et al., 1984) or as a vacuolation of lipids in the hepatocyte (Nitschke et al.,
1988a). Accumulation of lipids in the hepatocyte may lead to more serious effects such as
hepatic steatosis (fatty liver) and was observed in guinea pigs (Heppel et al., 1944), dogs (Haun,
1971) as well as in rats from a drinking water exposure (Serota et al., 1986a). Because Nitschke
et al. (1988a) examined a range of exposures that included doses at the low end of the range
compared with the range examined in Burek et al. (1984), the former study was selected as the
principal study for derivation of a chronic inhalation RfC.
Reproductive performance (e.g., as assessed by number of litters, resorption rate, fetal
survival, and growth) was not affected in two generations of F344 rats exposed to up to
1,500 ppm for 14 or 17 weeks before mating of the FO and Fl generations, respectively
(Nitschke et al., 1988b). Exposure in this study also continued from GD 0 to 21 and began again
at PND 4. In addition, reproductive performance was not affected in a study of Swiss-Webster
mice or Sprague-Dawley rats exposed to 1,250 ppm on GDs 6-15 (Schwetz et al., 1975). A
decrease in fertility index was seen in the 150 and 200 ppm groups in a study of male Swiss-
Webster mice exposed via inhalation for 6 weeks prior to mating (Raje et al., 1988), but the
statistical significance of this effect varied considerably depending on the statistical test used in
this analysis. Two types of developmental effects (decreased offspring weight at birth and
changed behavioral habituation of the offspring to novel environments) were seen in Long-Evans
rats following exposure to 4,500 ppm for 14 days prior to mating and during gestation (or during
gestation alone) (Bornschein et al., 1980; Hardin and Manson, 1980). This dose was the only
exposure dose used in this study. Schwetz et al. (1975) did not observe an adverse effect on
gross development or soft tissue abnormalities in a study involving exposure to 1,250 ppm on
GD 6 in Swiss-Webster mice or Sprague-Dawley rats, but an increase in delayed ossification of
the sternebrae was seen.
Neurological impairment was not seen in lifetime rodent bioassays involving exposure to
airborne dichloromethane concentrations of <2,000 ppm in F344 rats (Mennear et al., 1988;
NTP, 1986), <3,500 ppm in Sprague-Dawley rats (Nitschke et al., 1988a; Burek et al., 1984), or
268 DRAFT - DO NOT CITE OR QUOTE
-------
<4,000 ppm in B6C3Fi mice (Mennear et al., 1988; NTP, 1986). It should be noted, however,
that these studies did not include standardized neurological or neurobehavioral testing. The only
subchronic or chronic study in which neurobehavioral batteries were utilized found no effects in
an observational battery, a test of hind-limb grip strength, a battery of evoked potentials, or
brain, spinal cord, or peripheral nerve histology in F344 rats exposed to concentrations up to
2,000 ppm for 13 weeks, with the tests performed beginning 65 hours after the last exposure
(Mattsson et al., 1990).
Other effects associated with lifetime inhalation exposure to dichloromethane include
renal tubular degeneration in F344 rats exposed to >2,000 ppm, testicular atrophy in male
B6C3Fi mice exposed to 4,000 ppm, and ovarian atrophy in female B6C3Fi mice exposed to
>2,000 ppm (Mennear et al., 1988; NTP, 1986). No effects on histologic, clinical chemistry,
urinalysis, or hematologic variables were found in Syrian golden hamsters exposed to
concentrations up to 3,500 ppm for 2 years, with the exception that the mean COHb percentage
of exposed hamsters was about 30% compared with values of about 3% in controls (Burek et al.,
1984).
5.2.2. Derivation Process for RfC Values
The derivation process used for the RfC parallels the process described in Section 5.1.2
for the RfD derivation; consideration of dose metrics was described in Section 5.1.3. As was
noted in the RfD discussion, the mechanistic issues with respect to noncancer health effects
represent data gaps in the understanding of the health effects of dichloromethane.
5.2.3. Methods of Analysis—Including Models (PBPK, BMD, etc.)
The modified rat PBPK model of Andersen et al. (1991), described in Appendix C and
also used in the derivation of the RfD (Figure 5-2), was used for calculating internal dosimetry of
inhaled dichloromethane in Sprague-Dawley rats. Simulations of 6 hours/day, 5 days/week
inhalation exposures used in the Nitschke et al. (1988a) study were performed to calculate
average daily internal liver doses (Table 5-5). In the absence of data for group- and sex-specific
BWs, reference values for male and female Sprague-Dawley rats in chronic studies were used
(U.S. EPA, 1988a).
269 DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-5. Incidence data for liver lesions (hepatic vacuolation) and internal
liver doses based on various metrics in female Sprague-Dawley rats exposed
to dichloromethane via inhalation for 2 years
Sex
Male
Female
(BW =
229 g)
Exposure
(ppm)
0
50
200
500
0
50
200
500
Liver lesion
incidence"
22/70(31)
Not reported
Not reported
28/70 (40)
41/70 (59%)
42/70 (60%)
41/70 (58%)
53/70 (76%)c
Rat internal liver doseb
CYP
GST
GST and
CYP
Parent
AUC
Not modeled because results from male rats were not provided for the
50 and 200 ppm groups
0
285.3
665.3
782.1
0
6.17
93.2
360.0
0
291.4
758.5
1,142.1
0
1.18
17.8
68.6
"Number affected divided by total sample size.
blnternal doses were estimated using a rat PBPK model using exposures reported by study authors (50 ppm =
174 mg/m3, 200 ppm = 695 mg/m3, and 500 ppm = 1,737 mg/m3) and are weighted-average daily values for 1 wk of
exposure at 6 hr/d, 5 d/wk. CYP dose is in units of mg dichloromethane metabolized via CYP pathway/L tissue/d;
GST dose is in units of mg dichloromethane metabolized via GST pathway/L tissue/d.; GST and CYP dose is in
units of mg dichloromethane metabolized via CYP and GST pathways/L tissue/d; and Parent AUC dose is in units
of mg dichloromethane x hrs/L tissue.
Significantly (p < 0.05) different from control with Fisher's exact test.
Source: Nitschke et al. (1988a).
As described in Section 5.1.2, the internal dose metric used was based on total hepatic
metabolism through the CYP2E1 pathway (mg dichloromethane metabolized via CYP pathway
per liter liver per day). Figure 5-7 shows the comparison between inhalation external and
internal doses, using this dose metric for the rat and the human.
270
DRAFT - DO NOT CITE OR QUOTE
-------
> 10,000
-O >
"•v to
1,000
o o
o
E
100
10
+—.|—i—H___ i—i-i-+-
- - — H -t i r- i [ - --
Rat
Human mean
Human 5th %
Human 95th %
-:.
10 100 1,000
Inhalation concentration (ppm)
Average daily doses were calculated from simulated rat exposures of 6 hours/day,
5 days/week, while simulated human exposures were continuous. The human
metabolism rates were estimated using a computational sample of
1,000 individuals per dose, including random samples of the three GST-T1
polymorphisms (+/+, +/-, -/-) in the current U.S. population based on data from
Haber et al. (2002). Since a different set of samples was used for each dose, some
stochasticity is evident as the human points (values) do not fall on smooth curves.
Figure 5-7. PBPK model-derived internal doses (mg dichloromethane
metabolized via the CYP pathway per liter liver per day) in rats and humans
versus external exposures (ppm).
The seven dichotomous dose-response models available in EPA BMDS version 2.0 were
fit to the female rat liver lesion incidence of Nitschke et al. (1988a) and PBPK model-derived
internal dose data to derive rat internal BMDio and the associated BMDLio values (Table 5-6).
The quantal model is identical to the one-stage multistage model; therefore, it is not included in
this set of models. A BMR of 10% was selected because, in the absence of information
regarding the magnitude of change in a response that is thought to be minimally biologically
significant, a BMR of 10% is generally recommended, as it provides a consistent basis of
comparison across assessments. There are no additional data to suggest that the critical response
has a greater sensitivity that would warrant a lower BMR. The log-probit model was the best
fitting model for the female incidence data based on AIC value among models with adequate fit.
Modeling results are shown in detail in Appendix D-2.
271
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-6. BMD modeling results for incidence of noncancer liver lesions in
female Sprague-Dawley rats exposed to dichloromethane by inhalation for
2 years, based on liver specific CYP metabolism metric (mg
dichloromethane metabolized via CYP pathway per liter liver tissue per
day)
Model3
Gamma3
Logistic
Log-logistic3
Multistage (3)a
Probit
Log-probita'b
Weibulf
BMD10
622.10
278.31
706.50
513.50
279.23
737.93
715.15
BMDL10
227.29
152.41
506.84
155.06
154.52
531.82
494.87
x2
goodness of fit
/7-value
0.48
0.14
0.94
0.25
0.14
0.98
0.95
AIC
367.24
369.77
365.90
368.54
369.76
365.82
365.88
"These models in EPA BMDS version 2.0 were fit to the rat dose-response data shown in Table 5-5 by using
internal dose metrics calculated with the rat PBPK model. Gamma and Weibull models restrict power >1; Log-
logistic and Log-probit models restrict to slope >1, multistage model restrict betas >0; lowest degree polynomial
with an adequate fit reported (degree of polynomial in parentheses).
bBolded model is the best-fitting model in the most sensitive sex (females), which is used in the RfC derivation.
Source: Nitschke et al. (1988a).
As with the RfD derivation, the human-equivalent internal BMDLio was obtained by
dividing this rat internal dose metric by a pharmacokinetic scaling factor based on a ratio of BWs
(scaling factor = 4.09) (Table 5-7). This scaling factor was used because the metric is a rate of
metabolism rather than the concentration of putative toxic metabolites, and the clearance of these
metabolites may be slower per volume tissue in the human compared with the rat. (See Section
5.1.2. for a fuller discussion of this issue.) A probabilistic PBPK model for dichloromethane in
humans, adapted from the model of David et al. (2006) as described in Appendix B, was then
used with Monte Carlo sampling to calculate distributions of chronic HECs (in units of mg/m3)
associated with the internal BMDLio based on the responses in female Sprague-Dawley rats.
Estimated mean, first, and fifth percentiles of this distribution are shown in Table 5-7.
272
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-7. Inhalation RfC for dichloromethane based on PBPK model-
derived probability distributions of human inhalation exposure extrapolated
from liver lesion data for female rats exposed via inhalation for 2 years,
based on liver-specific CYP metabolism dose metric (mg dichloromethane
metabolized via CYP pathway per liter liver tissue per day)
Model3
Log-probit
Rat internal
BMDL10b
531.82
Human
internal
BMDL10C
130.03
HEC (mg/m3)d
1st
percentile
17.2
5th
percentile
21.3
Mean
48.5
Human RfC
(mg/m3)'
0.6
aBased on the best-fitting model from Table 5-6.
bRat dichloromethane PBPK model-derived internal liver dose associated with lower bound on 10% extra risk for
developing hepatocyte vacuolation.
°Human dichloromethane internal liver dose, derived by dividing the rat internal BMDL10 by a scaling factor of
4.09 [(B Whuman/B Wrat)°25] to account for potential interspecies pharmacokinetic differences in the clearance of
metabolites.
dPBPK model-derived distributions of long-term, daily average airborne dichloromethane concentrations predicted
by the PBPK model to yield an internal dose in humans equal to the dichloromethane internal BMDL10.
eHuman candidate RfC, based on female rat data, derived by dividing the 1st percentile of HEC values by a total UF
of 30: 3 (10°5) for possible toxicodynamic differences between species, 3 (10°5) for variability in human
toxicodynamic response, and 3 (1005) for database deficiencies. The 1st percentile POD is a stable estimate of the
lower end of the distribution. Use of this value in the lower tail replaces use of a UF for human toxicokinetic
variability. See Section 5.2.4 for discussion of UFs.
5.2.4. RfC Derivation—Including Application of Uncertainty Factors (UFs)
The 1st percentile POD is a numerically stable estimate of the lower end of the
distribution. Use of this value associated with a sensitive human population addresses the
uncertainty associated with human toxicokinetic variability. The RfC was calculated by dividing
the first percentile of the HEC distribution in Table 5-7 by a composite UF of 30 (3 [10°5] to
account for uncertainty about interspecies toxicodynamic equivalence, 3 [1005] to account for
uncertainty about toxicodynamic variability in humans, and 3 [10°5] for database deficiencies).
The resulting RfC was 0.6 mg/m3 based on liver lesions in female Sprague-Dawley rats in
Nitschke et al. (1988a). In deriving this RfC, factors for the following areas of uncertainty were
considered:
• Uncertainty in extrapolating from laboratory animals to humans (UFA). The use of
PBPK models to extrapolate internal doses from rats to humans reduces toxicokinetic
uncertainty in extrapolating from the rat liver lesion data but does not account for the
possibility that humans may be more sensitive than rats to dichloromethane due to
toxicodynamic differences. A UF of 3 (10°5) to account for this toxicodynamic
uncertainty was applied, as shown previously in Table 5-7.
• Uncertainty about variation in human toxicokinetics (UFn). The probabilistic human
PBPK model used in this assessment incorporates the best available information
about variability in toxicokinetic disposition of dichloromethane in humans but does
-------
not account for humans who may be sensitive due to toxicodynamic factors. Thus, a
UFof3 (10°'5)wa
sensitive humans.
UF of 3 (10°5) was applied to account for possible toxicodynamic differences in
• Uncertainty in extrapolating from LOAELs to NOAELs (UPi). A UF for
extrapolation from a LOAEL to a NOAEL was not applied because BMD modeling
was used to determine the POD, and this factor was addressed as one of the
considerations in selecting the BMR. The BMR was selected based on the
assumption that it represents a minimum biologically significant change.
• Uncertainty in extrapolating from subchronic to chronic durations (UFS). The derived
RfD is based on results from a chronic-duration drinking water toxicity study. No
cross-duration UF is necessary.
• Uncertainty reflecting database deficiencies (UFo). A UF of 3 was selected to
address the deficiencies in the dichloromethane toxicity database. The inhalation
database for dichloromethane includes several well-conducted chronic inhalation
studies. In these chronic exposure studies, the liver was identified as the most
sensitive noncancer target organ in rats (Nitschke et al., 1988a; NTP, 1986; Burek et
al., 1984). The critical effect of hepatocyte vacuolation was corroborated in the two
principal studies (Nitschke et al., 1988a; Burek et al., 1984), which identified 500
ppm as the lowest inhalation LOAEL for noncancer liver lesions. Gross signs of
neurologic impairment were not seen in lifetime rodent inhalation bioassays for
dichloromethane at exposure levels up to 4,000 ppm (see Section 4.2.2.2 for
references), and no exposure-related effects were observed in an observational
battery, a test of hind-limb grip strength, a battery of evoked potentials, or histologic
examinations of nervous tissues in F344 rats exposed to dichloromethane
concentrations as high as 2,000 ppm (Mattson et al., 1990).
A two-generation reproductive study in F344 rats reported no effect on fertility
index, litter size, neonatal survival, growth rates, or histopathologic lesions at
exposures >100 ppm dichloromethane (Nitschke et al., 1988b). Since exposure was
not continuous throughout the gestation and nursing periods, however, it may not be
representative of a typical human exposure and would not completely characterize
reproductive and developmental toxicity associated with dichloromethane. Fertility
index (measured by number of unexposed females impregnated by exposed males per
total number of unexposed females mated) was reduced following inhalation
exposure of male mice to 150 and 200 ppm dichloromethane 2 hours/day for 6 weeks,
but the statistical significance of this effect varied considerably depending on the
statistical test used in this analysis (Raje et al., 1988).
The available developmental studies are all single-dose studies that use relatively
high exposure concentrations (1,250 ppm in Schwetz et al. [1975]; 4,500 ppm in
Hardin and Manson [1980]; and 4,500 ppm in Bornschein et al. [1980]). In one of the
single-dose studies, decreased offspring weight at birth and changed behavioral
habituation of the offspring to novel environments were seen following exposure of
adult Long-Evans rats to 4,500 ppm for 14 days prior to mating and during gestation
(or during gestation alone) (Bornschein et al., 1980; Hardin and Manson, 1980). The
results from these single-dose developmental toxicity studies, the placental transfer of
dichloromethane (Withey and Karpinski, 1985; Anders and Sunram, 1982), and the
relatively high activity of CYP2E1 in the brain compared to the liver of the
-------
developing human fetus (Hines, 2007; Johnsrud et al., 2003; Brzezinski et al., 1999),
raise concerns regarding possible neurodevelopmental toxicity from gestational
exposure to inhaled dichloromethane. Dichloromethane, as well as the metabolite
CO, has been implicated in neurological effects (reviewed in Section 4.6.3.3, Mode of
Action for Neurological Effects); however, at exposures around the RfC, PBPK
models have predicted that CO levels would not be high enough to result in
neurodevelopmental toxicity.
In addition, Aranyi et al. (1986) demonstrated evidence of immunosuppression
following a single 100 ppm dichloromethane exposure for 3 hours in CD-I mice.
This exposure is lower than the POD for the liver effects that serve as the critical
effect for the RfC. This study used a functional immune assay that is directly relevant
to humans (i.e., increased risk of Streptococcal pneumonia-related mortality and
decreased clearance of Klebsiella bacteria). A recent study used a similar approach
for the evaluation of immunosuppression from acute exposures to trichloroethylene
and chloroform (Belgrade and Gilmour, 2010). Although dichloromethane was not
included in this study, Belgrade and Gilmour (2010) provide support for the
methodological approach used by Aranyi et al. (1986). Increases of some viral and
bacterial diseases, particularly bronchitis-related mortality, is also suggested by some
of the cohort studies of exposed workers (Radican et al., 2008; Gibbs et al., 1996;
Gibbs, 1992; Lanes et al., 1993, 1990). Systemic immunosuppression was not seen in
a 4-week, 5,000-ppm inhalation exposure study measuring the antibody response to
sheep red blood cells in Sprague-Dawley rats (Warbrick et al., 2003). These studies
suggest a localized, portal-of-entry effect within the lung rather than a systemic
immunosuppression. Because the Aranyi et al. (1986) study involved a single acute
inhalation exposure, interpretation of the findings from this study in the context of
chronic inhalation exposure is unclear.
In consideration of the entire database for dichloromethane, a database UF of 3
was selected. This UF accounts for limitations in the two-generation reproductive
toxicity study (i.e., incontinuous exposure throughout the lifecycle) and limitations in
the design of the available developmental studies (including a lack of
neurodevelopmental endpoints). There is an additional potential concern for
immunological effects as suggested by a single acute inhalation study, specifically
immunosuppressive effects that may be relevant for infectious diseases spread
through inhalation.
5.2.5. Previous RfC Assessment
No RfC was derived in the previous IRIS assessment.
5.2.6. RfC Comparison Information
A candidate RfC, based on a different approach to accounting for human toxicokinetic
variability is similar to the derived RfC of 0.6 mg/m3. Use of the mean value on the HEC
distribution (48.5) with an additional UF of 3 (10°5) to account for human toxicokinetic
variability would yield an RfC of 0.5 mg/m3.
For an additional comparison, an RfC was derived based on neurological endpoints from
human occupational exposures. Cherry et al. (1983) compared 56 exposed and 36 unexposed
workers at an acetate film manufacturing plant for dichloromethane inhalation exposure, blood
-------
levels of dichloromethane, subjective self-reporting of general health, and two objective,
quantitative measurements of neurological function (digit symbol substitution and simple
reaction time). The exposed and unexposed individuals were matched to within 3 years of age.
The measured dichloromethane concentrations from personal breathing zone sampling of the
exposed workers ranged from 28 to 173 ppm. No information on exposure duration was given,
and Cherry et al. (1983) did not indicate if the exposure measurements were indicative of
historical exposure levels. There were no significant differences between exposed and
unexposed workers in subjective or objective measurements collected at the beginning of the
work shift on a Monday (after 2 nonworking days). Exposed workers showed a slightly slower
(but not statistically significant) score than the control workers on a reaction time test, but the
scores did not deteriorate during the shift. These findings suggest that repeated inhalation
exposures in the range of 28-173 ppm do not result in significant effects, but the actual duration
of exposure of the workers is uncertain. In the absence of data for the mean exposure levels, the
exposure range midpoint of 101 ppm serves as a NOAEL for chronic neurological effects from
dichloromethane exposure. Thus, a candidate RfC of 3.5 mg/m3 was derived by dividing the
NOAEL of 351 mg/m3 (101 ppm) by a composite UF of 100. A UF of 10 was applied to account
for potentially susceptible individuals in the absence of quantitative information on the
variability of neurological response to dichloromethane in the human population. A UF of 10
was applied for database deficiencies. The duration of exposures of acetate film workers (Cherry
et al., 1983) was not reported, and a limited number of endpoints was evaluated. Further,
definitive neurological batteries were not administered in chronic-duration animal bioassays.
Another candidate RfC was developed by using the neurological data from the study by
potential long-term CNS effects in a study of retired aircraft maintenance workers (Lash et al.,
1991). Retired aircraft maintenance workers, ages 55-75 years, employed in at least 1 of
14 targeted jobs (e.g., paint strippers) with dichloromethane exposure for >6 years between
1970 and 1984 (n = 25) were compared to a like group of workers without dichloromethane
exposure (n = 21); the mean duration of retirement was 5 years in both groups. From 1974 to
1986, when 155 measurements for dichloromethane exposure were made, mean breathing zone
TWAs ranged from 82 to 236 ppm and averaged 225 ppm for painters and 100 ppm for
mechanics; information on exposure levels prior to this time was not provided although the work
histories of exposed workers goes back to the 1940s. The evaluation included several standard
neurological tests, including physiological measurement of odor and color vision senses,
auditory response potential, hand grip strength, measures of reaction time (simple, choice, and
complex), short-term visual memory and visual retention, attention, and spatial ability. The
exposed group had a higher score on verbal memory tasks (effect size approximately 0.45,/> =
0.11) and lower score on attention tasks (effect size approximately -0.55, p = 0.08) and complex
reaction time (effect size approximately -0.40,/? = 0.18) compared with the control group. None
of these differences were statistically significant. Given the sample size, however, the power to
-------
detect a statistically significant difference between the groups was very low (i.e., approximately
0.30 for an effect size of 0.40 using a two-tailed alpha of 0.05), and these results cannot be taken
as evidence of no effect. An estimated exposure level from the study can be generated from the
midpoint value from the exposure range (82-236 ppm; mean =159 ppm), converted to
552 mg/m3. If these results are viewed as a LOAEL and this estimated mean exposure level of
552 mg/m3 was used, a composite UF of 1,000 would be applied for interspecies toxicodynamics
(10), extrapolation from a LOAEL to a NOAEL (10), and database uncertainties (10), resulting
in an RfC of 0.55 mg/m3.
The value of the candidate RfC based on the data from Cherry et al. (1983), 3.5 mg/m3, is
approximately 6-fold higher, and the value of the candidate RfC based on the data from Lash et
al. (1991), 0.55 mg/m3, is very similar to the derived RfC of 0.6 mg/m3 based on liver lesions in
rats. The animal-derived RfC is preferable to the human-derived RfC because of the
uncertainties about the exposure durations, statistical analysis, and statistical power in Cherry et
al. (1983) and the uncertainties regarding the exposure levels, effect sizes (and how this may
differ depending on time since last exposure), and statistical power in Lash et al. (1991), and
because, in comparison with the value based on Cherry et al. (1983), the RfC based on the rat
data is more sensitive.
Additional comparisons among the RfC and candidate values developed from other
endpoints or data sets using NOAEL/LOAEL methods are shown in Table 5-8 and Figure 5-8.
-------
Table 5-8. Potential points of departure with applied UFs and resulting candidate RfCs
Endpoint
Hepatocyte vacuolation,
female ratd
Renal tubular degeneration;
NOAEL, male rat
Reproductive - fertility
index; NOAEL, male
mouse
Increased infection
susceptibility (mortality
risk), female mouse
Increased IgM production,
male and female rat
Chronic CNS effects,
human male
CNS changes, human male
POD
(mg/m3)3
17.2
620
20.7
15.5
17,366
351
552
POD Type and
Description1"
1s* percentile human
equivalent
concentration based on
BMDL10 for increase in
incidence of liver lesion
NOAEL
No effect at POD, 16%
decrease in fertility index
seen at LOAEL dose
NOAEL
NOAEL
NOAEL
LOAEL
UFsc
Total
UF
30
100
100
1,000
1,000
30
300
UFA
3
3
3
3
3
1
1
UFH
3
10
10
10
10
10
10
UFL
1
1
1
1
1
1
10
UFS
1
1
1
10
10
1
1
UFD
3
3
3
3
3
3
3
RfC (mg/m3)
0.6
6.02
0.21
0.015
17.4
11.7
1.84
Reference
Nitschke et al.
(1988a)
Mennear et al.
(1988); NTP (1986)
Rajeetal. (1988)
Aranyietal. (1986)
Warbrick et al.
(2003)
Cherry etal. (1983)
Lash etal. (1991)
aPOD = point of departure. For Nitschke et al. (1988a), this is based on BMD modeling of a 10% increase in liver lesions using internal liver dose metric (mg
dichloromethane metabolism via CYP pathway per liter liver tissue per d) derived from a rat PBPK model. After an allometric scaling factor of 4.09 was applied,
the human internal BMDL10 was 130 mg dichloromethane metabolism via CYP pathway per liter liver tissue per d. A probabilistic human PBPK model adapted
from David et al. (2006) was used to generate a distribution of HECs (in units of mg/m3) from the human internal BMDLj 0 and the first percentile of this distribution
was used as the POD. For other rodent studies, the NOAEL or LOAEL concentration, in mg/m3, was adjusted to a continuous exposure taking into account hrs/d and
d/wk of exposure. This adjusted exposure was then converted to an HEC by multiplying the value by a dosimetric adjustment factor (DAF). PBs were 8.24 for
humans, 19.8 for rats, and 23 for mice. Since the PBs for both the mice and rats were greater than for humans, a DAF of 1 is recommended and was used. NOAELs
or LOAELs were used as points of departure in human studies since the concentrations were already human exposures.
''Extra risk defined for incidence data as (Incidence: - Incidence0)/(l-Incidence0), where 1 = dose at observed increased and 0 = background incidence.
°UFA = uncertainty in extrapolating from laboratory animals to humans, UFH = uncertainty about variation from average humans to sensitive humans, UFL =
uncertainty about extrapolating from LOAEL to NOAEL, and UFD = uncertainty reflecting incompleteness of the overall database. A UF extrapolating from
subchronic to chronic durations (UFS) was not used for any of these studies.
dBolded value is the basis of the RfC of 0.6 mg/m3.
278
DRAFT - DO NOT CITE OR QUOTE
-------
f>
g
o
• M
-M
«
'-
•4-»
c
o
u
c
o
U
103
102
101
10°
10-'
io-2
io-3
—
ft
\/
1
A
*~ l~
I
I
^
J
Hepalocyte Renal tubular Reproductive Increased Increased Chronic CNS
vacuolation: degeneration: Performance Infection IgM adj. CNS effects; changes:
lslpercentile adj. HEC -Fertility Susceptibility: HECrat- NOAEL LOAEL
HEC from from rat- Index:; adj. adj. HEC Warbricket from human from human
female rat- Mcnnearet HEC mouse- mouse- al. (2003) males- males -Last
Nitschkeet al.(1988): Rajeetal. Aranvi et al. Cherryetal. etal.(1991)
al.(1988a) NTP(1986) (1988) (1986) (1983')
0 Point of Departure
| UFA-Interspecies;
animal to human
Q UFH - Intraspecies;
human variability
] UF, - LOAEL to
NOAEL
| UFS - Subchronic to
Chronic
,—I UFD-Database
. Reference Dose
Figure 5-8. Comparison of candidate RfCs derived from selected points of departure for endpoints presented in
Table 5-8.
-------
5.3. UNCERTAINTIES IN THE ORAL REFERENCE DOSE AND INHALATION
REFERENCE CONCENTRATION
Risk assessments need to include a discussion of uncertainties associated with the derived
toxicity values. For dichloromethane, uncertainties related to inter- and intraspecies differences
in toxicodynamics and database deficiencies are treated quantitatively via the UF approach (U.S.
EPA, 1994b). Uncertainties in the toxicokinetic differences of dichloromethane between species
and within humans are reduced by application of the PBPK models for rats and humans. These
and other areas of uncertainty of the derived RfD and RfC are discussed below.
Adequacy of database for derivation of RfD and RfC. As summarized in Sections 4.6.1.1
and 4.6.2.1, data from the available human studies on the health effects from occupational
inhalation exposures provide some but not conclusive evidence of long-term health
consequences of chronic dichloromethane exposure, specifically with respect to neurologic and
hepatic damage. These data are not adequate for derivation of an RfD or RfC. However, a broad
range of animal toxicology data is available for the hazard assessment of dichloromethane, as
described in Section 4. The database of oral (Table 4-35) and inhalation (Tables 4-36 and 4-37)
toxicity studies includes numerous chronic, subchronic, acute, reproductive, and developmental
studies. Liver toxicity in multiple rodent species is consistently identified as the most sensitive
noncancer effect from oral and inhalation exposure to dichloromethane. In addition to the oral
and inhalation toxicity data, there are numerous studies describing the toxicokinetics of
dichloromethane. Consideration of the available dose-response data to determine an estimate of
oral exposure that is likely to be without an appreciable risk of adverse noncancer health effects
over a lifetime has led to the selection of noncancer liver lesions in the 2-year drinking water
study in F344 rats (Serota et al., 1986a) as the critical effect and principal study for deriving the
RfD for dichloromethane. The critical effect selected for the derivation of the chronic RfC is
also hepatic lesions; two different studies in Sprague-Dawley rats (Nitschke et al., 1988a; Burek
et al., 1984) spanning overlapping exposures reported data on hepatic vacuolation, and the lower
exposure study was chosen as the principal study (Nitschke et al., 1988a).
The lack of an adequate two-generation reproductive toxicity study, with continuous
exposure throughout the lifecycle, has been identified as a data gap. Another critical data
uncertainty was identified for neurodevelopmental effects. Animal bioassays have not identified
gross or microscopic effects on neural tissues from long-term exposures or single (Schwetz et al.,
1975) or multigenerational (Nitschke et al., 1988b) developmental toxicity studies. However,
behavioral changes were observed in pups born to rats exposed to high levels (4,500 ppm) of
dichloromethane (Bornschein et al., 1980; Hardin and Manson, 1980); lower exposures were not
examined in this study. Uncertainty exists as to the development of neurological effects from
lower gestational exposures in animals or humans. In addition, a critical data uncertainty has
been identified that relates to potential immunotoxicity, specifically immunosuppression seen as
280 DRAFT - DO NOT CITE OR QUOTE
-------
a localized portal-of-entry effect within the lung with an acute inhalation exposure (Aranyi et al.,
1986). The lack of data on immune effects from longer-term exposure represents a significant
data gap and is of particular importance because of the potential importance of
immunosuppression with respect to response to infections and tumor surveillance. The weight of
evidence for noncancer effects in humans and animals suggests that the development of liver
lesions is the most sensitive effect, with a UF applied because of deficiencies in the reproductive
and developmental studies and the lack of neurodevelopmental studies for the RfD and, for the
RfC, the additional uncertainty regarding immune system toxicity (specifically, a portal-of-entry
immune suppression effect) at low exposures.
Dose-response modeling. The selection of the BMD model(s) for the quantitation of the
RfD and RfC does not lead to significant uncertainty in estimating the POD. It should be noted,
however, that a level of uncertainty is inherent given the lack of data in the region of the BMR.
Interspecies extrapolation ofdosimetry and risk. The extrapolation of internal
dichloromethane dosimetry from liver lesions in rats to human risk was accomplished using
PBPK models for dichloromethane in rats and humans. Uncertainties in rat and human
dosimetry used for RfD and RfC derivation can arise from uncertainties in the PBPK models
with regard to accurately simulating the toxicokinetics of dichloromethane for animals under
bioassay conditions and humans experiencing relatively low, chronic environmental exposures.
Specific uncertainties regarding the model structure are described in detail in Section 3.5.5. A
structural uncertainty previously discussed arises from the indication by various data that the
standard Michaelis-Menten equation used in the existing model may not accurately describe the
CYP2E1-catalyzed oxidation of dichloromethane. An alternate equation described by Korzekwa
et al. (1998) may better represent CYP2E1-induced oxidation of dichloromethane, which would
lead to a higher fraction of total dichloromethane predicted to be metabolized by CYP2E1 at
higher dichloromethane doses (or exposures). Since this shift in predicted metabolism would
occur for both the human and rodent PBPK models, if the alternate equation was applied, it is
difficult to estimate the net impact of using this equation on risk predictions. As described in
Section 3.5.5 and Appendix C, the error in the ratio of GST:CYP metabolism at low
concentrations appears to be less than 13% based on comparison of model predictions to CO
metabolism data. Further, analysis of the GST-mediated metabolism of dichloromethane
measured by Reitz et al. (1989) shows that those results are within a factor of three of the GST
kinetic parameters used in the current PBPK model, indicating that any error in the GST:CYP
balance is no greater than that.
Also as discussed in detail in Section 3.5.2, there appears to be inconsistency in the
numerical results of David et al. (2006) for the liver GST activity (coefficient), kfc, between that
obtained for each published data set when analyzed separately and that obtained for the
281 DRAFT - DO NOT CITE OR QUOTE
-------
combined data set. Since the numerical average of the mean kfc values for the four data sets
included in the combined data set was 12.4 and the upper bound was 12, the impact of using an
intermediate value of kfc, specifically the DiVincenzo and Kaplan value of 5.87 kg0 3/hour was
explored. Changing only the k^ is not realistic since the dichloromethane data effectively define
total metabolism (sum of CYP and GST pathways), and there is naturally a negative correlation
between the predicted CYP metabolic rate and the GST metabolic rate required to describe this
total. It would be inconsistent with the dichloromethane data to increase kfc without adjusting
the CYP metabolic rate downward and likewise all other parameters. Therefore, for consistency,
the distributions for all of the fitted parameters were rescaled by the ratio of the mean for
DiVincenzo and Kaplan (1981) to the mean for the combined data set (e.g., the distribution for
Vmaxc was multiplied by 10.2/9.42 and the distribution for kfc was multiplied by 5.87/0.852, the
respective ratios of the two posterior means for each parameter). The HEC and human
equivalent dose calculations (to which the RfC and RfD are proportional) increased by only 25
and 12%, respectively, for the mixed GST-T1 population. Thus the impact of this model
uncertainty appears to be modest for the noncancer assessment.
The dose metric used in the models is the rate of metabolism to a putative toxic
metabolite rather than an average or AUC of the metabolite concentration, so the model
specifically fails to account for rodent-human differences in clearance or removal of the toxic
metabolite. Therefore, a scaling factor based on BW ratios was used to account for this
difference.
Sensitivity analysis of rat model parameters. The rat model was modified and utilized in
a deterministic manner. Data were not available to perform a hierarchical Bayesian calibration
in the rat. Thus, uncertainties in the rat model predictions had to be assessed qualitatively. To
address these uncertainties, a sensitivity analysis was conducted to determine which model
parameters most influence the predictions for a given dose metric and exposure scenario.
Sensitivity is a measure of the degree to which a given model output variable (i.e., dose
metric) is influenced by perturbation in the value of model parameters. The approach
implemented was a univariate analysis in which the value of an individual model parameter was
perturbed by an amount (A) in the forward and reverse direction (i.e., an increase and decrease
from the nominal value), and the change in the output variable was determined. Sensitivity
coefficients were calculated as follows:
. , x
Ax f(x)
282 DRAFT - DO NOT CITE OR QUOTE
-------
where x is the model parameter,/^ is the output variable, Ax is the perturbation of the
parameter from the nominal value, andf'(x) is the sensitivity coefficient. In equation 5-1, the
sensitivity coefficients are scaled to the nominal value of x andf(x) to eliminate the potential
effect of units of expression. Therefore, the sensitivity coefficient is a measure of the
proportional (unitless) change in the output variable produced by proportional change in the
parameter value. Parameters that have higher sensitivity coefficients have greater influence on
the output variable. They are considered more sensitive than parameters with lower values. The
results of the sensitivity analysis are useful for assessing uncertainty in model predictions, based
on the level of confidence or uncertainty in the model parameter(s) to which the dose metric is
most sensitive.
Sensitivity coefficients for the noncancer dose metric (mg dichloromethane metabolized
via CYP-mediated pathway per liter liver per day) were determined for each of the model
parameters. Sensitivity analyses for both oral and inhalation exposures were performed. The
exposure conditions were set to be near or just below the lowest bioassay exposure resulting in
significant increases in the critical effect.
For the CYP-mediated metabolism from oral exposure, the liver volume (VLC) and
slowly perfused tissue volume (VSC) parameters exert the largest influence (Figure 5-9). The
high influence of these two parameters was due to the fact that the dose metric is a tissue-specific
rate of metabolism, the majority of CYP metabolism is attributed to the liver, and changes in
liver volume have a greater impact on the total CYP metabolism than the individual Vmaxc value.
For inhalation exposures Vmaxc, VLC, and VSC have the highest sensitivity coefficients
(Figure 5-10). The physiological parameters (VLC and VSC) are known with a high degree of
confidence (Brown et al., 1997). Vmaxc for the rat was estimated by fitting to the
pharmacokinetic data as described in Chapter 3 and Appendix C, subject to model
structure/equation uncertainties as detailed above, and hence is known with less certainty than
the physiological parameters. That total exhaled CO, which is specific to the CYP pathway, is
within 50% of measured levels (Figure C-8, panel C), however, provides a similar level of
confidence in the balance between CYP and GST pathways predicted by the rat PBPK model.
283 DRAFT - DO NOT CITE OR QUOTE
-------
KFC
A2
KA
i_
,2 VMAXC
o>
E PB
(0
<5 VSC
Q.
VLC
VPR
QCC
u
c
T-
Or
•al e>
\fff>
:posi
jre
r?
rfffffffff
C
E
E
i
T?7i
3
fff*
_r
rffffff
i
H
BCYP
DGST
:>T-LOLOLOOLOLOLO-<-
vj ' r-~ Q CN CN Q r-~
- O ' O O O
Normalized sensitivity coefficient
iO
CN
Figure 5-9. Sensitivity coefficients for long-term mass CYP- and GST-mediated
metabolites per liver volume from a daily drinking water concentration of 10
mg/L in rats.
\
KFC
A2
£ VMAXC
| PB
5 VSC
(0
Q- VLC
VPR
QCC
u
r
T
Inh
alati<
\fff,
an e>
fffi
epos
c
Tfff
n
ure
^
fff
\
G
E
rff^pffffffffji
771
=
i
i
•
BCYP
DGST
TT-LOLOLOOLOLOLOT-
M ' r-~ Q CN CN Q r-~
^ 0 ' 0 0 0
Normalized sensitivity coefficient
CN
(KA is not included since it has no impact on inhalation dosimetry.)
Figure 5-10. Sensitivity coefficients for long-term mass CYP- and GST-mediated
metabolites per liver volume from a long-term average daily inhalation
concentration of 500 ppm in rats.
In summary, the uncertainties associated with use of the rat PBPK model should not
markedly affect the values (i.e., an effect of no more than 30%) of the RfD and RfC based on the
metrics considered. An additional uncertainty results from the lack of knowledge concerning the
most relevant dose metric (e.g., a specific metabolite) for the noncancer endpoints considered.
284
DRAFT - DO NOT CITE OR QUOTE
-------
This basic research question represents a data gap. This uncertainty was addressed by
considering different dose metrics (CYP metabolism alone, GST metabolism alone, sum of GST
and CYP, and the AUC of the parent compound). The GST metabolism and the AUC dose
metrics did not present reasonable choices based on model fit and consistency of response across
studies at comparable dose levels. Given these results, the combination of hepatic metabolism
through the GST and the CYP pathways would not be expected to result in an improvement to a
metric based only on CYP metabolism. The CYP-metabolism dose metric seems to be most
consistent with the data, and so is the metric chosen for the RfD and RfC derivations.
Sensitive human populations. The potential for sensitivity to dichloromethane in a
portion of the human population due to pharmacokinetic differences was addressed
quantitatively by using a human probabilistic PBPK model, as modified by EPA, to generate
distributions of human exposures likely to result in a specified internal BMDLio. The model and
resulting distributions take into account the known nonchemical-specific variability in human
physiology as well as total variability and uncertainty in dichloromethane-specific metabolic
capability. The first percentile values of the distributions of human equivalent doses (Table 5-3)
and HECs (Tables 5-7) served as points of departure for candidate RfDs and RfCs, respectively,
to protect toxicokinetically sensitive individuals. Selection of the first percentile allows
generation of a numerically stable estimate for the lower end of the distribution. The mean value
of the human equivalent oral dose in Table 5-3 was about twofold higher than the corresponding
first percentile values, and the mean value of human equivalent inhalation concentration in
Table 5-7 was approximately threefold higher than the first percentile value. The internal dose
metric in the analyses described in these tables was the mg dichloromethane metabolized via the
CYP pathway per liter liver per day, and thus the comparisons of the first percentile and mean
values give estimates of the amount of variability in the population to metabolize
dichloromethane by the CYP metabolic pathways on a liver-specific basis. The mean:
1st percentile ratios for these distributions is attributed to the dependence of the dose metric on
hepatic blood flow rate (metabolism being flow-limited). This blood flow is expected to be
highly and tightly correlated with liver volume, resulting in very similar delivery of
dichloromethane per volume liver across the population. The mean: 1st percentile ratio for the
oral distribution is 1.85, which is less than the default intra-human toxicokinetic UF of 3. The
population-structured distributions for physiological parameters and broadened distributions for
metabolic parameters used here provide a good degree of confidence that the population
variability has not been underestimated.
The internal dose metric used in the RfD and RfC derivations was based on the rate of
CYP metabolism. GST-T1 polymorphisms could affect this rate, as the GST-T1 null genotype
would be expected to result in an increase in the metabolism through the CYP pathway, resulting
in a greater sensitivity to a CYP-related effect. The effect of GST variability on the RfD and
285 DRAFT - DO NOT CITE OR QUOTE
-------
RfC values was examined by comparing results obtained specifically for the GST-T1 null
genotype to those obtained for the population of mixed genotypes. The values for human
equivalent doses and HECs were very similar for these two groups (e.g., mean HEC 47.36 and
47.49 for the mixed and the GST-IT7" null genotypes, respectively; 1st percentile HEC 16.63 and
16.69 for the mixed and the GST-T1" null genotypes, respectively), and use of this population
would not result in a change in the recommended RfD or RfC.
As a further level of sensitivity analysis, model predictions of the human equivalent dose
for the general population, as listed in Table 5-3 (estimates covered 0.5- to 80-year-old male and
female individuals), were compared to three subpopulations: 1-year-old children (males and
females), 70-year-old men, and 70-year-old women. For the general population and each
subpopulation, a Monte Carlo simulation representing 10,000 individuals was conducted, and
histograms of the resulting distribution of human equivalent doses are shown in Figure 5-11,
with corresponding statistics in Table 5-9. All groups used in these comparisons were limited to
the GST-T1V-.
12
0)
Human equivalent
dose distributions
General
70 yo Male
••70 yo Female
1 yo Child
0.2
0.9
0.3 0.4 0.5 0.6 0.7 0.8
Human equivalent applied dose (mg/kg-day)
All groups were restricted to the GST-TT ~ population.
Figure 5-11. Frequency density of human equivalent doses in specific
populations in comparison to a general population (0.5- to 80-year-old males
and females) estimate for an internal dose of 12.57 mg dichloromethane
metabolized by CYP per liter liver per day.
286
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-9. Statistical characteristics of human equivalent doses in specific
populations of the GST-T1 ~'~ group
Population
All agesb
1-yr-old children
70-yr-old men
70-yr-old women
Human equivalent dose
(mg/kg-d)a
Mean
3.31 x lO'1
5.25 x 10'1
2.64 x 10'1
2.20 x 10'1
5th percentile
2.11 x ID'1
4.06 x 10'1
2.59 x 10'1
1.68 x 10'1
1s* percentile
1.79 x ID'1
3.74 x 10'1
2.06 x 10'1
1.55 x 10'1
""Exposure levels predicted to result in 15.1 mg dichloromethane metabolized via CYP pathway per liter liver per d
(based on BMDL10 from the best-fitting model from Table 5-2; human dichloromethane internal liver dose, derived
by dividing the rat internal BMDL10 by a scaling factor of 4.09 [(B Whuman/B Wrat)°25] to account for potential
interspecies pharmacokinetic differences in the clearance of metabolites).
b0.5- to 80-yr-old males and females.
The results shown above for differences in human equivalent dose values in different
populations are qualitatively what would be expected: a relatively broad distribution for the
general population with specific populations representing narrower components of that
distribution. There are some differences between men and women at 70 years of age, but neither
of these would be greatly misrepresented by the general population estimate. While 1-year-old
children represent more of a distinct tail in the general population, in this case, the distribution of
HECs in the general estimate is lower than that seen in what would otherwise be considered a
more sensitive population. This difference most likely results from the higher specific
respiration rate in children versus adults, which allows them to eliminate more of orally ingested
dichloromethane by exhalation, leading to lower internal metabolized doses.
A similar comparison was made for inhalation HEC values, as shown in Figure 5-12 and
Table 5-10. For FtEC values, the distributions for 70-year-old men and women are both virtually
indistinguishable from the general population, and while 1-year-old children are clearly distinct,
they are less different than in the human equivalent dose comparison and, in this case, are more
sensitive than the population in general. As described in detail in Appendix B, the allometric
alveolar ventilation constant (QAlvC) is about 28 L/hour-kg0'75 in a 1-year-old child but averages
around 14 L/hour-kg0'75 in an adult. Combining this with the difference between a BW of 10 kg
in that child and 70 kg in an "average" adult, the respiration rate per kg BW is about threefold
higher in the child versus adult. As noted above, for oral exposures, this leads to faster
elimination by respiration in children, while for inhalation exposures it leads to higher uptake for
a given air concentration.
287
DRAFT - DO NOT CITE OR QUOTE
-------
Human equivalent
concentration distributions*
General population
1 year old
-70 yo male
-70 yo female
20 40 60 80 100 120
Human equivalent concentration (mg/m3)
140
*A11 groups restricted to the GST-T1"" population.
Figure 5-12. Frequency density of HECs in specific populations in
comparison to a general population (0.5- to 80-year-old males and females)
estimate for an internal dose of 130.0 mg dichloromethane metabolized by
CYP per liter liver per day.
288
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-10. Statistical characteristics of HECs in specific populations of the
GST-T1 -'- group
Population
All agesb
1-yr-old children
70-yr-old men
70-yr-old women
HEC (mg/m3)a
Mean
48.7
25.2
46.9
51.0
5th percentile
21.2
14.5
21.8
22.3
1s* percentile
16.7
12.2
17.9
18.2
""Exposure levels predicted to result in 128.1 mg dichloromethane metabolized via CYP pathway per liter liver per d
(based on BMDL10 from the best-fitting model from Table 5-6; Human dichloromethane internal liver dose, derived
by dividing the rat internal BMDL10 by a scaling factor of 4.09 [(B Whuman/B Wrat)°25] to account for potential
interspecies pharmacokinetic differences in the clearance of metabolites).
b0.5- to 80-yr-old males and females.
The lack of difference in elderly adults versus the general population in HEC values is
likely due to the fact that the rate of exposure and rates of metabolism (the latter being the key
dose metric) both scale as BW0'75, with the scaling coefficients being either similar (respiration)
or identical (metabolism) among adults who comprise the majority of the population. Hepatic
CYP-mediated metabolism is relatively stable with increasing age (Bebia et al., 2004;
Schmucker et al., 2001). For oral exposures, the exposure rate is normalized to total BW and
scales as BW1, while elimination routes increase as BW°75. Moreover, oral exposures are
simulated as occurring in a series of bolus exposures (drinking episodes) during the day, and the
higher body-fat content occurring in the elderly (see Appendix B) means that such a dose that
might saturate metabolism and therefore have a higher fraction exhaled in a leaner individual
will tend to be more sequestered in fat and slowly released, resulting in a higher fraction
metabolized (less saturation of metabolism) in a more obese individual. The difference among
adults of different ages for dosimetry from oral ingestion (bolus exposure) will be greater than
the difference for inhalation exposures. More careful examination of Figure 5-12 shows that the
distribution for 70-year-old women, for whom the fat fraction is estimated to be greatest, has a
lower peak and higher upper tail than for the general population. Thus, the physiological
differences have some impact that is qualitatively consistent with what is seen from oral
exposure, given the mechanistic considerations described here. But the impact of those
differences is far less for inhalation exposure.
No data are available regarding toxicodynamic differences within a human population.
Therefore, a UF of 3 for possible differences in human toxicodynamic responses is intended to
be protective for sensitive individuals.
289
DRAFT - DO NOT CITE OR QUOTE
-------
5.4. CANCER ASSESSMENT
5.4.1. Cancer OSF
5.4.1.1. Choice of Study/Data—with Rationale and Justification
No human data are available for the quantification of potential neoplastic effects from
oral exposures to dichloromethane. In the only chronic (2-year) oral exposure cancer bioassay,
significant increases in the incidence of liver adenomas and carcinomas were observed in male
B6C3Fi mice exposed by drinking water, with incidence rates of 19, 26, 30, 31, and 28% in
groups with estimated mean intakes of 0, 61, 124, 177, and 234 mg/kg-day, respectively (trendy-
value = 0.058) (Table 4-38) (Serota et al., 1986b; Hazleton Laboratories, 1983). Incidences of
liver tumors in female mice were not presented in the summary reports, but it was reported that
exposed female mice did not show increased incidences of proliferative hepatocellular lesions
(Serota et al., 1986b; Hazleton Laboratories, 1983). Evidence of a trend for increased risk of
liver tumors (described as neoplastic nodule or hepatocellular carcinoma) was seen in female
F344 rats but not males exposed via drinking water (p < 0.01) (Serota et al., 1986a). However,
the potential malignant characterization of the nodules was not described, and no trend was seen
in the data limited to hepatocellular carcinomas.
The derivation of the cancer OSF is based on the male mouse data (Serota et al., 1986b;
Hazleton Laboratories, 1983) because of their greater sensitivity compared to female mice and to
male and female rats. The study authors concluded that there was no dose-related trend and that
there were no significant differences comparing the individual dose groups with the combined
control group, and that the observed incidences were "within the normal fluctuation of this type
of tumor incidence." Although Serota et al. (1986b) state that a two-tailed significance level ofp
= 0.05 was used for all tests, this does not appear to correctly represent the statistic used by
Serota et al. Each of the/>-values for the comparison of the 125, 185, and 250 mg/kg-day dose
groups with the controls wasp < 0.05. (Theses-values were found in the full report of this
study, see Hazleton Laboratories, 1983, but were not included in the Serota et al., 1986b
publication). Hazleton Laboratories (1983) indicated that a correction factor for multiple
comparisons was used specifically for the liver cancer data, reducing the nominal /7-value from
0.05 to 0.0125; none of these individual group comparisons are statistically significant when ap-
value of 0.0125 is used. EPA did not consider the use of a multiple comparisons correction
factor for the evaluation of the liver tumor data (a primary a priori hypothesis) to be warranted.
Thus, based on the Hazleton Laboratories (1983) statistical analysis, EPA concluded that
dichloromethane induced a carcinogenic response in male B6C3Fi mice as evidenced by a
marginally increased trend test (p = 0.058) for combined hepatocellular adenomas and
carcinomas, and by small but statistically significant (p < 0.05) increases in hepatocellular
adenomas and carcinomas at dose levels of 125 (p = 0.023), 185 (p = 0.019), and 250 mg/kg-day
(p = 0.036).
290 DRAFT - DO NOT CITE OR QUOTE
-------
With respect to comparisons with historical controls, the incidence in the control groups
(19%) was almost identical to the mean seen in the historical controls from this laboratory
(17.8% based on 354 male B6C3Fi mice), so there is no indication that the observed trend is
being driven by an artificially low rate in controls and no indication that the experimental
conditions resulted in a systematic increase in the incidence of hepatocellular adenomas and
carcinomas. Although the occurrence of one elevated rate in an exposed group may reflect
normal fluctuations in the incidence of these tumors (described for this laboratory as 5-40%,
with a mean of 17.8%, based on 354 male controls), the pattern of incidence rates (increased
incidence in all four dose groups, with three of these increases significant at ap-va\ue of < 0.05)
suggest a treatment-related increase.
The development of liver tumors in B6C3Fi mice is associated with metabolite
production in this tissue via the GST metabolic pathway (Section 4.7.3), a pathway that also
exists in humans. Modeling intake, metabolism, and elimination of dichloromethane in mice and
humans is feasible. Thus, it is reasonable to apply the best available PBPK models to estimate
equivalent internal doses in mice and humans.
5.4.1.2. Derivation ofOSF
In a manner similar to the derivation of the noncancer toxicity values, PBPK models for
dichloromethane in mice and humans were used in the derivation of toxicity values (cancer OSF
and IUR) for cancer endpoints based on lung (for inhalation) and liver (for oral and inhalation)
tumor data in the mouse (Figure 5-13). A deterministic PBPK model for dichloromethane in
mice was first used to convert mouse drinking water or inhalation exposures to long-term daily
average values of internal lung-specific GST metabolism (GST metabolism in lung/lung volume)
or liver-specific GST metabolism (GST metabolism in liver/liver volume). The choice of these
dose metrics was made based on data pertaining to the mechanism(s) involved in the
carcinogenic response, specifically data supporting the involvement of a GST metabolite(s). The
evidence pertaining to the GST pathway is discussed in Section 4.7 and includes the enhanced
genotoxicity seen in bacterial and mammalian in vitro assays with the introduction of GST
metabolic capacity (Graves et al., 1994a) and the suppression of the production of DNA SSBs by
pretreatment with a GSH depletory seen in acute inhalation exposure to dichloromethane in mice
(Graves et al., 1995). Although the GST metabolic pathway takes on a greater role as the CYP
pathway is saturated, both the GST and CYP pathways are operating even at low exposures. The
PBPK model incorporates the metabolic shift and expected nonlinearity (GST dose attenuation
with low exposures) in the exposure-dose relationship across exposure levels.
291 DRAFT - DO NOT CITE OR QUOTE
-------
Benchmark Dose Analysis
Rodent Dose
Response Data
Koaeni
PBPK Model
Estimates of Rodent
Internal Dose
BMD
Modeling
0.5
0.4
0.2
Multistage-
•' ................. BMDL
.''. ................. BMD
Human Tumor Risk Factor
(internal dose)-1
Scaling
Factor
J-
Rodent Tumor Risk Factor
(internal dose)"1
(0.1/RodentBMDL10)
10
20 30
Dose
40 50
60
Rodent Internal BMDU
95% Lower Bound Estimate of Internal
Dose Associated with a 10% response
Multiply Human Tumor Risk Factor
By Distribution of Human Internal
Unit Doses
95th 99th
Distribution of Human Cancer
Oral Slope Factors or
Inhalation Unit Risks
Recommend mean value
Apply Age-Dependent Adjustment Factors
(ADAFs) for early life exposure
Probabilistic
Human PBPK
Model
Distribution of Human Internal
Doses from Unit Oral Doses
(1mg/kg) or Inhalation
Concentrations (1 ug/m3)
Monte Carlo
Sampling from
Distributions of
Human PBPK
Model Parameters
Figure 5-13. Process for deriving cancer OSFs and lURs by using rodent and human PBPK models.
292
DRAFT - DO NOT CITE OR QUOTE
-------
The multistage cancer model (using BMDS version 2.0) was fit to the tumor incidence
data and internal dose data for rodents, and BMDio and associated BMDLio values (for a BMR
of 10% extra risk) were estimated. A probabilistic PBPK model for dichloromethane in humans,
adapted from David et al. (2006) (see Appendix B), was used with Monte Carlo sampling to
calculate distributions of internal lung or liver doses associated with chronic unit oral (1 mg/kg-
day) or inhalation (1 ug/m3) exposures. The resulting distribution of human internal doses was
multiplied by a human internal dose tumor risk factor (in units of reciprocal internal dose) to
generate a distribution of OSFs or lURs associated with a chronic unit oral or inhalation
exposure, respectively.
As discussed in Section 3.5.2, the statistics reported for the fitted metabolic parameters
by David et al. (2006; Table 4 in that publication) only represent the population mean and
uncertainty in that mean for each parameter. EPA's revision of the model parameter
distributions are generally described in Section 5.1.2, with details provided in Appendix B.
Subsequent to this revision, the human PBPK parameter distributions are expected to
appropriately account for both parametric uncertainty and interindividual variability, with
sampling weighted to represent the full population from 6 months to 80 years of age. The model
code also allows estimation of risk for subpopulations defined by a specific age in that range,
gender, and/or GST-T1 genotype (e.g., the GST-T1 +/+ subpopulation).
5.4.1.3. Dose-Response Data
Data for liver tumors in male B6C3Fi mice following exposure to dichloromethane in
drinking water were used to develop oral cancer slope factors (Serota et al., 1986b; Hazleton
Laboratories, 1983). Significant increases in incidence of liver adenomas and carcinomas were
observed in male but not female B6C3Fi mice exposed for 2 years (Table 5-11). No significant
decreases in survival were observed in the treated groups of either sex compared with controls.
The at-risk study populations (represented by the denominators in the incidence data) were
determined by excluding all animals dying prior to 52 weeks.
293 DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-11. Incidence data for liver tumors and internal liver doses, based
on GST metabolism dose metrics in male B6C3Fi mice exposed to
dichloromethane in drinking water for 2 years
Sex
Male
(BW =
37.3 g)
Nominal (actual) daily
intake (mg/kg-d)
0(0)
60 (61)
125 (124)
185 (177)
250 (234)
Mouse liver
tumor incidence"
24/125 (19%)
51/199(26%)
30/99 (30%)
31/98(32%)
35/123 (28%)
Mouse internal liver
metabolism doseb
0
17.5
63.3
112.0
169.5
Mouse whole body
metabolism dosec
0
0.73
2.65
4.68
7.1
"Hepatocellular carcinoma or adenoma, combined. Mice dying prior to 52 wks, as estimated from the survival data
shown in Figure 1 of Hazleton Laboratories (1983), were excluded from the denominators. Cochran-Armitage
trends-value = 0.058. P-values for comparisons with the control group were 0.071, 0.023, 0.019, and 0.036 in the
60, 125, 185, and 250 mg/kg-d groups, respectively, based on statistical analyses reported by Hazleton Laboratories
(1983).
bmg dichloromethane metabolized via GST pathway/L liver/d. Internal doses were estimated from simulations of
actual daily doses reported by the study authors.
°Based on the sum of dichloromethane metabolized via the GST pathway in the lung plus the liver, normalized to
total BW (i.e., [lung GST metabolism (mg/d) + liver GST metabolism (mg/d)]/kg BW). Units = mg
dichloromethane metabolized via GST pathway in lung and liver/kg-d.
Sources: Serota et al. (1986b); Hazleton Laboratories (1983).
5.4.1.4. Dose Conversion and Extrapolation Methods: Cancer OSF
Dose conversion. The mouse PBPK model of Marino et al. (2006) was based on the
PBPK model for dichloromethane by Andersen et al. (1987), which was modified to include
dichloromethane metabolism in the lung compartment and kinetics of CO and COHb (Andersen
et al., 1991). For the mouse, physiological parameters and partition coefficients were adjusted to
match those reported in Andersen et al. (1991, 1987) and Clewell et al. (1993), respectively,
while QCC, VPR, and metabolic parameter distribution mean values were derived via MCMC
model calibration reported by Marino et al. (2006) (Appendix B). The model of Marino et al.
(2006) was used to simulate daily drinking water exposures comprising six discrete drinking
water episodes for specified times and percentage of total daily intake (Reitz et al., 1997), and to
calculate average lifetime daily internal doses for the male mouse data shown in Table 5-11. A
first-order oral absorption rate constant (ka) of 5 hours"1 was taken from Reitz et al. (1997) to
describe the uptake of dichloromethane from the gastrointestinal tract to the liver. Study-specific
BWs were not available, so reference BWs for male B6C3Fi mice in chronic studies (U.S. EPA,
1988a) were used. Based on evidence that metabolites of dichloromethane produced via the
GST pathway are primarily responsible for dichloromethane carcinogenicity in mouse liver
(summarized in Section 4.7.3) and the assumption that these metabolites are sufficiently reactive
that they do not have substantial distribution outside the liver, the recommended selected internal
dose metric for liver tumors was daily mass of dichloromethane metabolized via the GST
pathway per unit volume of liver (Table 5-11). Figure 5-14 shows the comparison between
294
DRAFT - DO NOT CITE OR QUOTE
-------
internal and external doses in the liver in mice and humans. The whole-body metabolism metric
was also examined; however, this metric would be more relevant under a scenario of slowly
cleared metabolites that undergo general circulation.
IUUU
re
0)
^~
^100
0)
_N
"5
.0
E
| 10
oT
2
o
o
SI
8 1
Liver GST dose |
for oral exposure
T | S ^
i*
\ 1
; i'*'
I y
, s
>>* 1 ^^
xx ^i '"
J^^ ^^
^^^ ^ f
T ^^ \
•^
^ ^ ,
,^
'
r^^f ^>
s^*^
' 1
-n ^-V
zfs
ff
i
H
* •
r
i
s ' ^
— - - Mouse
- - - - Human mixed
Human +/-
Human +/+
0) 1
S 10 100
Dose (mg/kg/day)
1000
Six simulated drinking water episodes are described by Reitz et al. (1997). The
human metabolism rates were estimated using a computational sample of
1,000 individuals per dose, including random samples of the three GST-T1
polymorphisms (+/+, +/-, -/-; "Human mixed" curve) or samples restricted to the
GST +/+ or +/- populations in the current U.S. population based on data from
Haber et al. (2002). Since a different set of samples was used for each dose, some
stochasticity is evident as the human points (values) do not fall on smooth curves.
Error bars indicate the range of ^ -95* percentile for the subpopulations sampled
at select concentrations.
Figure 5-14. PBPK model-derived internal doses (mg dichloromethane
metabolized via the GST pathway per liter liver per day) in mice and
humans and their associated external exposures (mg/kg-day) used for the
derivation of cancer OSFs based on liver tumors in mice.
Dose-response modeling and extrapolation. The multistage dose-response model was fit
to the mouse liver tumor incidence and PBPK model-derived internal dose data to derive a
mouse internal BMDio and BMDLio associated with 10% extra risk (Table 5-12). Different
r\
polynomial models were compared based on adequacy of model fit as assessed by overall %
goodness of fit (p-value > 0.10) and examination of residuals at the 0 dose exposure (controls)
295
DRAFT - DO NOT CITE OR QUOTE
-------
and in the region of the BMR. Due to the lack of a monotonic increase in tumor response at the
high dose, the model did not adequately fit the data. In consideration of the region of interest
(i.e., low-dose risk estimation), the highest dose group was excluded. The modeling of the
remaining four dose groups exhibited an adequate fit to the data. Appendix E-l provides details
of the BMD modeling results. The mouse liver tumor risk factor (extra risk per unit internal
dose) was calculated by dividing 0.1 by the mouse BMDLio for liver tumors.
Table 5-12. BMD modeling results and tumor risk factors for internal dose
metric associated with 10% extra risk for liver tumors in male B6C3Fi mice
exposed to dichloromethane in drinking water for 2 years, based on liver-
specific GST metabolism and whole body GST metabolism dose metrics
Internal
dose metric
Liver-
specific
Whole-body
BMDS
modelb
MS (1,1)
MS (1,1)
x2
goodness of
fit/7-value
0.56
0.56
Mouse
BMD10C
73.0
3.05
Mouse
BMDL10C
39.6
1.65
Allometric-
scaled human
BMDL10d
5.66
0.24
Tumor risk factor6
Scaling = 1.0
2.53 x 1Q-3
-
Allometric-
scaled
1.77 x 1Q-2
4.24 x 1Q-1
"Liver specific dose units = mg dichloromethane metabolized via GST pathway per liter tissue per d; whole-body
dose units = mg dichloromethane metabolized via GST pathway in lung and liver/kg-d).
bThe multistage (MS) model in EPA BMDS version 2.0 was fit to the mouse dose-response data shown in
Table 5-11 using internal dose metrics calculated with the mouse PBPK model. Numbers in parentheses indicate
(1) the number of dose groups dropped in order to obtain an adequate fit; and (2) the degree polynomial of the
model.
°BMD10 and BMDL10 refer to the BMD-model-predicted mouse internal and its 95% lower confidence limit,
associated with a 10% extra risk for the incidence of tumors.
dMouse BMDL10 divided by (BWhuman/BWmouse)025 = 7.
eDichloromethane tumor risk factor (extra risk per unit internal dose) derived by dividing the BMR (0.1) by the
mouse BMDL10 and by the allometric-scaled human BMDL10 for the scaling =1.0 and allometric-scaled risk
factors, respectively.
Linear extrapolation from the internal human BMDLio values (0.1/BMDLio) was used to
derive oral risk factors for liver tumors based on tumor responses in male mice. Proposed key
events for dichloromethane carcinogenesis are discussed in Sections 4.7 and 5.4.1.1. The linear
low-dose extrapolation approach for agents with a mutagenic mode of action was selected.
Application of allometric scaling factor. As discussed in Section 4.7 and summarized in
5.4.1.2, several lines of evidence point to the involvement of the GST metabolic pathway in the
carcinogenic response seen in dichloromethane. The role of specific metabolites has not been
firmly established, however. S-(chloromethyl)-glutathione is an intermediate to the production
of formaldehyde through this pathway (Hashmi et al., 1994). Formation of the free hydrogen ion
is also hypothesized, although no direct evidence supporting this has been presented. The pattern
of HPRT gene mutations seen in CHO cells incubated with GST-complete mouse liver cytosol
preparations suggest that S-(chloromethyl)glutathione, rather than formaldehyde, may be
296
DRAFT - DO NOT CITE OR QUOTE
-------
responsible for the mutagenic effects associated with dichloromethane (Graves et al., 1996).
DNA reaction products (e.g., DNA adducts) produced by S-(chloromethyl)glutathione have not
been quantified, possibly due to potential instability of these compounds (Watanabe et al., 2007;
Hashmi et al., 1994). Other studies indicate that DNA damage resulting from formaldehyde
formation, in addition to the S-(chloromethyl)glutathione intermediate, should also be considered
(Hu et al. 2006; Casanova et al., 1997).
The question of the role of specific metabolites and particularly how these metabolites
are transformed or removed is a key question affecting the choice of a scaling factor to be used in
conjunction with the internal dose metric based on rate of GST metabolism. If the key
metabolite is established and is known to be sufficiently reactive to not spread in systemic
circulation, then it can be assumed that: (1) the level of reactivity and rate of clearance (i.e.,
disappearance due to local reactivity) for this metabolite per volume tissue is equal in rodents
and humans, and (2) risk is proportional to the long-term daily average concentration of the
metabolite. Under these assumptions, rodent internal BMDLio values based on tissue-specific
dichloromethane metabolism require no allometric scaling to account for toxicodynamic
differences and predict the corresponding level of human risk as a function of the metric (i.e., the
scaling factor in Figure 5-13 was equal to 1.0). (A single metabolite is referenced, but the same
argument holds in general for more than one metabolite). Under this scenario and assumptions,
humans and rodents with the same long-term daily average metabolite formation per volume
tissue (e.g., equal internal BMDLio) should both experience the same long-term average
concentration of the metabolite when the metabolite is highly reactive and hence experience the
same extra risk.
Although the evidence points to a specific metabolic pathway and to site-specific actions
resulting from a reactive metabolite that does not escape the tissue in which it is formed, some
assumptions remain concerning this hypothesis. Specifically, the active metabolite(s) have not
been established, and data pertaining to the reactivity or clearance rate of these metabolite(s) are
lacking. Quantitative measurements of adducts of interest or of the half life of relevant
compounds in humans and in mice are not available. It is not known that the rate of reaction is
proportional to the liver perfusion rate, cardiac output, or body surface area, and it is not known
that the rate of reaction is not proportional to these factors. To address the uncertainties in the
available data, it is appropriate to use a scaling factor that addresses the possibility that the rate
of clearance for the metabolite is limited by processes that are known to scale allometrically,
such as blood perfusion, enzyme activity, or availability of reaction cofactors that is limited by
overall metabolism. This case would result in use of a mouse:human dose-rate scaling factor of
(BWhuman/BWmouse)0'25 = 7 to adjust the mouse-based BMDLio values downward. Using this
internal dose metric (liver-specific metabolism with allometric scaling), equivalent rodent and
human internal BMDLio values result in a human liver tumor risk factor (0.1/BMDLio) that is
assumed equal to that for the mouse, given a 70-year lifetime exposure.
297 DRAFT - DO NOT CITE OR QUOTE
-------
Another alternative that can be used is based on an allometrically-scaled whole-body
metabolism metric. In this case, less weight is given to the evidence of site-specificity, as this
metric allows for systemic circulation of the relevant metabolites.
The cancer toxicity values derived using each of these metrics and scaling factors (i.e.,
liver-specific metabolism with and without allometric-scaling and the whole-body metabolism
metric) are presented in the following tables. Considering the lack of data pertaining to
clearance rates or the actual AUC of the active carcinogenic metabolite(s) in mice and humans,
the OSF recommended by EPA is based on the allometrically-scaled tissue-specific GST
metabolism rate dose metric.
Calculation ofOSFs. The human PBPK model adapted from David et al. (2006) (see
Appendix B), using Monte Carlo sampling techniques, was used to calculate distributions of
human internal dose metrics of daily mass of dichloromethane metabolized via the liver-specific
GST pathway per unit volume of liver resulting from a long-term average daily drinking water
dose of 1 mg/kg dichloromethane. In another analysis of whole body metabolism, a dose metric
based on the total metabolites formed in liver and lungs via GST metabolism per BW was used.
The human model used parameter values derived from Monte Carlo sampling of probability
distributions for each parameter, including MCMC-derived distributions for the metabolic
parameters (David et al., 2006). The drinking water exposures comprised six discrete drinking-
water episodes for specified times and percentage of total daily intake (Reitz et al., 1997)
(Appendix B).
The distribution of cancer OSFs shown in Table 5-13 was derived by multiplying the
human oral liver tumor risk factors by the respective distributions of human average daily
internal doses resulting from chronic, unit oral exposures of 1 mg/kg-day dichloromethane. The
mean slope factor was selected as the recommended value; other values at the upper end of the
distribution are also presented.
298 DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-13. Cancer OSFs for dichloromethane based on PBPK model-derived internal liver doses in B6C3Fi mice
exposed via drinking water for 2 years, based on liver-specific GST metabolism and whole body metabolism dose
metrics, by population genotype
Internal dose metric
and scaling factor"
Liver-specific,
allometric -scaled
Liver-specific,
scaling = 1.0
Whole-body,
allometric -scaled
Population
genotype1"
GST-T1+/+
Mixed
GST-T1+/+
Mixed
GST-T1+/+
Mixed
Human tumor
risk factor0
1.77 x ID'2
1.77 x ID'2
2.53 x ID'3
2.53 x ID'3
4.24 x ID'1
4.24 x ID'1
Distribution of human internal dichloromethane
doses from 1 mg/kg-d exposure"1
Mean
0.94 x ID'1
0.53 x ID'1
0.94 x 1Q-1
0.53 x 1Q-1
2.20 x 1Q-3
1.27 x 1Q-3
95th
percentile
2.98 x 1Q-1
1.96 x ID'1
2.98 x ID'1
1.96 x ID'1
7.20 x ID'3
4.66 x ID'3
99th
percentile
5.43 x ID'1
3.78 x ID'1
5.43 x ID'1
3.78 x ID'1
1.30 x ID'2
9.41 x ID'3
Resulting candidate human
OSFe (mg/kg-d) ^
Mean
1.7 x ID'3
9.4 x ID'4
2.4 x ID'4
1.3 x ID'4
9.3 x lO'4
5.4 x ID'4
95th
percentile
5.3 x ID'3
3.5 x ID'3
7.5 x ID'4
5.0 x ID'4
3.1 x ID'3
2.0 x ID'3
99th
percentile
9.6 x ID'3
6.7 x ID'3
1.4 x ID'3
9.6 x ID'4
5.5 x ID'3
4.0 x ID'3
aLiver specific dose units = mg dichloromethane metabolized via GST pathway per liter tissue per d; Whole-body dose units = mg dichloromethane metabolized via
GST pathway in lung and liver/kg-d.
bGST-Tl+/+ = homozygous, full enzyme activity; mixed = population reflecting estimated frequency of genotypes in current U.S. population: 20% GST-T"'", 48%
GST-T1+A, and 32% GST-T1+/+ (Haber et al., 2002).
°Dichloromethane tumor risk factor (extra risk per unit internal dose per d) derived by dividing the BMR (0.1) by the allometric-scaled human BMDL10 and the
mouse BMDL10 for the allometric-scaled and scaling =1.0 risk factors, respectively (from Table 5-12).
dMean, 95th, and 99th percentile of the human PBPK model-derived probability distribution of daily average internal dichloromethane dose resulting from chronic
oral exposure of 1 mg/kg-d.
Derived by multiplying the dichloromethane tumor risk factor by the PBPK model-derived probabilistic internal doses from daily exposure to 1 mg/kg-d.
299
DRAFT - DO NOT CITE OR QUOTE
-------
Consideration of Sensitive Human Subpopulations. An important issue in the derivation
process used by EPA, pertaining to the use of the human PBPK model, stems from the
assumption regarding the population for which the derivation should be applied. The inclusion
of the GST-T1 null subpopulation in effect dilutes the risk that would be experienced by those
who carry a GST-T1 allele by averaging in nonresponders (i.e., the GST-T1"" genotype). Thus,
the cancer OSF was derived specifically for carriers of the GST-T1 homozygous positive (+/+)
genotype, the population that would be expected to be most sensitive to the carcinogenic effects
of dichloromethane given the GST-related dose metric under consideration. In addition, cancer
values derived for a population reflecting the estimated frequency of GST-T1 genotypes in the
current U.S. population (20% GST-T1V', 48% GST-T1+/', and 32% GST-T1+/+, i.e., the "mixed"
population) are also presented. All simulations also included a distribution of CYP activity
based on data from Lipscomb et al. (2003).
5.4.1.5. Oral Cancer Slope Factor
The recommended cancer OSF for dichloromethane is 2 x 10"3 (mg/kg-day)"1 (rounded
from 1.7 x 10"3) and is based on liver tumor responses in male B6C3Fi mice exposed to
dichloromethane in drinking water for 2 years (Serota et al., 1986b; Hazleton Laboratories,
1983). The OSF was derived by using a tissue-specific GST metabolism dose metric with
allometric scaling for the population that is presumed to have the greatest sensitivity (the
GST-T1+/+ genotype). The application of ADAFs to the cancer OSF is recommended and is
described in Section 5.4.4.
5.4.1.6. Alternative Derivation Based on Route-to-Route Extrapolation
For comparison, alternative cancer OSFs were derived via route-to-route extrapolations
from the data for liver tumors in male and female B6C3Fi mice exposed by inhalation for 2 years
(Mennear et al., 1988; NTP, 1986). This derivation, shown in Table 5-14, uses the cancer IUR
derived in Section 5.4.2.4 (see Table 5-19 for these IUR values) and the distribution of human
internal dichloromethane exposures from 1 mg/kg-day exposure using the tissue-specific GST
metabolism dose metric (mg dichloromethane metabolized via the GST pathway per liter liver
per day). The cancer OSFs based on the route-to-route extrapolations from liver tumors in mice
exposed by inhalation (Table 5-14) are about one order of magnitude lower than those based on
the liver tumor responses in mice exposed via drinking water.
300 DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-14. Alternative route-to-route cancer OSFs for dichloromethane extrapolated from male B6C3Fi mouse
inhalation liver tumor incidence data using a tissue-specific GST metabolism dose metric, by population genotype
Internal dose metric
and scaling factor
Liver-specific,
allometric -scaled
Liver-specific, scaling =
1.0
Whole-body metabolism
Population
genotype3
GST-T1+/+
Mixed
GST-T1+/+
Mixed
GST-T1+/+
Mixed
Human
tumor risk
factorb
1.29 x ID'3
1.29 x ID'3
1.84x lO'4
1.84x lO'4
3.03 x ID'2
3.03 x ID'2
Distribution of human internal dichloromethane
doses from 1 mg/kg-d exposure0
Mean
0.94 x ID'1
0.53 x ID'1
0.94 x ID'1
0.53 x ID'1
2.20 x ID'3
1.27 x ID'3
95th
percentile
2.98 x ID'1
1.96 x ID'1
2.98 x ID'1
1.96 x ID'1
7.20 x ID'3
4.66 x ID'3
99th
percentile
5.43 x ID'1
3.78 x ID'1
5.43 x ID'1
3.78 x ID'1
1.30 x ID'2
9.41 x ID'3
Resulting candidate human
OSFd (mg/kg-d) ^
Mean
1.2 x ID'4
6.8 x ID'5
1.7 x ID'5
9.7 x ID'6
6.7 x ID'5
3.9 x ID'5
95th
percentile
3.8 x ID'4
2.5 x ID'4
5.5 x ID'5
3.6 x ID'5
2.2 x ID'4
1.4 x ID'4
99th
percentile
7.0 x ID'4
4.9 x ID'4
1.0 x ID'4
6.9 x ID'5
3.9 x ID'4
2.9 x ID'4
aGST-Tl+/+ = homozygous, full enzyme activity; mixed = population reflecting estimated frequency of genotypes in current U.S. population: 20% GST-T"'", 48%
GST-T1+A, and 32% GST-T1+/+ (Haber et al., 2002).
bDichloromethane tumor risk factor (extra risk per milligram dichloromethane metabolized via GST pathway per liter tissue per d) derived by dividing the BMR
(0.1) by the allometric-scaled human BMDL10 and the mouse BMDL10 for the allometric-scaled and scaling = 1.0 risk factors, respectively (from IUR data, Table
5-19).
°Mean, 95th, and 99th percentile of the human PBPK model-derived probability distribution of daily average internal dichloromethane dose (mg dichloromethane
metabolized via GST pathway per liter tissue per d) resulting from chronic oral exposure of 1 mg/kg-d.
dDerived by multiplying the dichloromethane tumor risk factor by the PBPK model-derived probabilistic internal doses from daily exposure to 1 mg/kg-d.
301
DRAFT - DO NOT CITE OR QUOTE
-------
5.4.1.7. Alternative Based On Administered Dose
One comparison that can be made is with an alternative OSF based on liver tumors in
mice, using the external concentrations of dichloromethane in the mouse as converted to human
equivalent doses and then applying this by using BMD modeling to obtain the BMDLio and
resulting oral cancer risk. Mouse bioassay exposures were adjusted to human equivalent doses
as follows:
human equivalent dose = (nominal daily intake/BW scaling factor) x daily exposure
adjustment factor
where BW scaling factor = (BWhuman/BWn
and
daily exposure adjustment factor = 5/7
= 7
'
The human equivalent doses for the 0, 60, 125, 185, and 250 mg/kg-day dose groups used
in the liver tumor analysis (Table 5-11) (from Serota et al. [1986b]) were 0, 6.12, 12.75, 18.87,
and 25.51 mg/kg-day, respectively. The BMD modeling and OSF derived from these values are
shown in Table 5-15. The resulting OSF based on the liver tumors in the mouse is
approximately one order of magnitude higher than the current recommended value obtained by
using the mouse and human PBPK models.
Table 5-15. Cancer OSF based on a human BMDLio using administered
dose for liver tumors in male B6C3Fi mice exposed to dichloromethane in
drinking water for 2 years
Sex,
tumor type
Male, liver
BMDS model3
MS (0,1)
x2
goodness of fit
/7-value
0.55
Human
BMD10C
19.4
Human
BMDL10C
10.4
Cancer
OSFd
(mg/kg-d)1
1.0 x 10'2
aThe multistage (MS) model in EPA BMDS version 2.0 was fit to the mouse liver tumor data shown in Table 5-11.
The human equivalent doses for the 0, 60, 125, 185, and 250 mg/kg-d dose groups used in the liver tumor analysis
were 0, 6.12, 12.75, 18.87, and 25.51 mg/kg-d, respectively, based on application of BW scaling factor =
uman/B Wmouse) = 7 and adjusting for daily exposure by multiplying by 5/7 d. Numbers in parentheses
indicate: (1) the number of dose groups dropped in order to obtain an adequate fit, starting with the highest dose
group, and (2) the degree polynomial of the model.
°BMD10 and BMDL10 refer to the BMD-model-predicted human equivalent dose (mg/kg-d) and its 95% lower
confidence limit, associated with a 10% extra risk for the incidence of tumors.
dCancer OSF (risk per mg/kg-d) = 0. I/human BMDL10.
The administered dose methodology can be considered equivalent to using a single-
compartment, whole-body model of dichloromethane where the internal dose metric is the AUC
of dichloromethane itself, and clearance of dichloromethane scales from mice to humans as
BW075. The estimates based on the PBPK model, in contrast, use the rate of GST metabolism of
dichloromethane as the metric. Another difference is that the administered dose methodology
302
DRAFT - DO NOT CITE OR QUOTE
-------
does not account in any way for the GST polymorphism and so might be considered as
representing the general/mixed-GST-genotype population rather than the +/+ subpopulation.
5.4.1.8. Previous IBIS Assessment: Cancer OSF
The previous IRIS assessment derived a cancer OSF of 7.5 x 10~3 (mg/kg-day)"1 by the
application of the multistage model to combined incidence of hepatocellular adenomas and
carcinomas from two studies; a 2-year drinking water study in B6C3Fi mice (Hazleton
Laboratories, 1983) and a 2-year inhalation study in B6C3Fi mice (NTP, 1986). The slope factor
was the arithmetic mean of two candidate slope factors, 1.2 x io~2 (mg/kg-day)"1 (Hazleton
Laboratories, 1983) and 2.6 x 10'3 (mg/kg-day)"1 (NTP, 1986). Since the NTP (1986) animal
data were from inhalation exposures, the estimated inhaled doses were calculated for mice and
humans (assuming near complete uptake into lung tissues and blood) and converted to
administered doses in units of mg/kg-day. Assumed inhalation rates of 0.0407 and 20 m3/day
were used for mice and humans, respectively. No adjustments were made for species differences
in metabolism or toxicokinetics.
5.4.1.9. Comparison of Cancer OSFs Using Different Methodologies
Cancer OSFs derived using different dose metrics and assumptions are summarized in
Table 5-16. The recommended OSF of 2 x 10~3 per mg/kg-day (rounded to one significant digit)
is based on a tissue-specific GST-internal dose metric with allometric scaling (=7) because of
some uncertainty regarding the rate of clearance of the relevant metabolite(s) formed via the
GST pathway. The value derived specifically for the GST-T1+/+ population is recommended to
provide protection for the population that is hypothesized to be most sensitive to the carcinogenic
effect. The values based on the GST-T1+ + group are approximately twofold higher than those
for the full population for the dose metrics used in this assessment (Table 5-16). Within a
genotype population, the values of the OSF among most of the various dose metrics vary by
about one to two orders of magnitude.
303 DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-16. Comparison of OSFs derived using various assumptions and metrics, based on tumors in male mice
Population"
GST-Tl+/+b
Mixed
Dose metric
Tissue-specific GST-metabolism rate
Tissue-specific GST-metabolism rate
Whole-body GST metabolism rate
Route-to-route extrapolation, tissue-specific metabolism
Route-to-route extrapolation, tissue-specific metabolism
Route-to-route extrapolation, whole-body metabolism
Tissue-specific GST-metabolism rateb
Tissue-specific GST-metabolism rate
Whole-body GST metabolism rate
Route-to-route extrapolation, tissue-specific metabolism
Route-to-route extrapolation, tissue-specific metabolism
Route-to-route extrapolation, whole-body metabolism
Applied dose (human equivalent dose)
1995 IRIS assessment
Species,
sex
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Tumor
Liver
Liver
Liver
Liver
Liver
Liver
Liver
Liver
Liver
Liver
Liver
Liver
Liver
Liver
Scaling
factor
7.0
1.0
7.0
7.0
1.0
7.0
7.0
1.0
7.0
7.0
1.0
7.0
Mean OSF
(mg/kg-d)1
1.7 x 10 3
2.4 x ID'4
9.3 x ID'4
1.2 x ID'4
1.7 x ID'5
6.7 x ID'5
9.4 x ID'4
1.3 x ID'4
5.4 x ID'4
6.8 x ID'5
9.7 x 1Q-6
3.9 x ID'5
1.0 x 10'2
7.5 x ID'3
Source
(table)
Table 5-13
Table 5-13
Table 5-13
Table 5-14
Table 5-14
Table 5-14
Table 5-13
Table 5-13
Table 5-13
Table 5-14
Table 5-14
Table 5-14
Table 5-15
aGST-Tl+/+ = homozygous, Ml enzyme activity; Mixed = genotypes based on a population reflecting the estimated frequency of genotypes in the current U.S.
population: 20% GST-Tr7', 48% GST-T1+A, and 32% GST-T1+/+ (Haber et al., 2002).
folded value is the basis for the recommended OSF of 2 x 10"3 per mg/kg-d.
304
DRAFT - DO NOT CITE OR QUOTE
-------
5.4.2. Cancer IUR
5.4.2.1. Choice of Study/Data—with Rationale and Justification
As discussed in Section 4.7, results from several cohort mortality studies of workers
repeatedly exposed to dichloromethane and several case-control studies provide supporting
evidence of carcinogenicity in humans, specifically with respect to liver and brain cancer, and
some hematopoietic cancers. However, the epidemiologic studies do not provide adequate data
to estimate exposure-response relationships for dichloromethane exposure and these cancers.
Results from several bioassays provide sufficient evidence of the carcinogenicity of
dichloromethane in mice and rats exposed by inhalation, as well as adequate data to describe
dose-response relationships. As discussed in Section 4.7.2, repeated inhalation exposure to
concentrations of 2,000 or 4,000 ppm dichloromethane produced increased incidences of lung
and liver tumors in male and female B6C3Fi mice (Maronpot et al., 1995; Foley et al., 1993;
Kari et al., 1993; Mennear et al., 1988; NTP, 1986). A weaker trend (p = 0.08) was seen with
respect to liver tumor incidence (described as neoplastic nodules or hepatic carcinomas) in
female rats, but this trend was not seen when limited to hepatic carcinomas (NTP, 1986). A
statistically significant increased incidence of brain tumors has not been observed in any of the
animal cancer bioassays, but a 2-year study using relatively low exposure levels (0, 50, 200, and
500 ppm) in Sprague-Dawley rats observed a total of six astrocytoma or glioma (mixed glial
cell) tumors (combining males and females) in the exposed groups (Nitschke et al., 1988a).
These tumors are exceedingly rare in rats, and there are few examples of statistically significant
trends in animal bioassays (Sills et al., 1999). Male and female F344 rats exposed by inhalation
to 2,000 or 4,000 ppm showed significantly increased incidences of benign mammary tumors
(adenomas or fibroadenomas) and the male rats also exhibited a low rate of sarcoma or
fibrosarcoma in mammary gland or subcutaneous tissue around the mammary gland (NTP,
1986).
The NTP inhalation study in B6C3Fi mice (NTP, 1986) was used to derive an IUR for
dichloromethane because of the completeness of the data, adequate sample size, and clear dose
response with respect to liver and lung tumors. The liver tumor incidence in male mice
increased from 44% in controls to 66% in the highest dose group; in females, the incidence of
this tumor rose from 6 to 83%. For lung tumors, the incidence rose from 10 to 80% in males and
from 6 to 85% in females. Compelling evidence exists for the role of GST-mediated metabolism
of dichloromethane in carcinogenicity in mice (Section 4.7.3), and both mice and humans
possess this metabolic pathway. Modeling intake, metabolism, and elimination of
dichloromethane in mice and humans is feasible. Thus, it is reasonable to apply the best
available PBPK models to estimate equivalent internal doses in mice and humans.
The mammary tumor data from the NTP (1986) study was also used to derive a
comparative IUR. However, the toxicokinetic or mechanistic events that might lead to
mammary gland tumor development in rats are unknown, and so a clear choice of the optimal
305 DRAFT - DO NOT CITE OR QUOTE
-------
internal dose metric could not be made. Thus, this derivation is based on the average daily AUC
for dichloromethane in blood. The role of CYP- or GST-mediated metabolism in the mammary
gland is uncertain, although both GST-T1 (Lehmann and Wagner, 2008) and CYP2E1 (El-Rayes
et al., 2003; Hellmold et al., 1998) expressions have been detected in human mammary tissue. It
is also possible that some metabolites enter systemic circulation from the liver and lung where
they are formed.
The female rat liver cancer data from the NTP (1986) inhalation study was not used to
derive an IUR because the trend was weaker than that seen in the mouse (incidence increased
from 4% in controls to 10% in the highest dose group, trend/? = 0.08), and because the effect
categorization included neoplastic nodule or hepatocellular carcinoma. The brain tumor data
seen in the Nitschke et al. (1988a) study in Sprague-Dawley rats were not used to develop an
IUR because of the low incidence of this rare tumor (a total of four astrocytoma or glioma
tumors in exposed males and two in exposed females). The mechanistic issues with respect to
mammary tumors and health effects issues with respect to brain tumors represent data gaps in the
understanding of the health effects of dichloromethane and relevance of the rat data to humans.
5.4.2.2. Derivation of the Cancer IUR
The derivation of the IUR parallels the process described in Section 5.4.1.2 for the cancer
OSF. Since modeling metabolism and elimination kinetics of dichloromethane in mice and
humans is feasible, it is reasonable to apply the best available PBPK models to determine
equivalent target organ doses in mice and humans. Although the GST metabolic pathway takes
on a greater role as the CYP pathway is saturated, both the GST and CYP pathways are operating
even at low exposures. The PBPK model incorporates the metabolic shift and expected
nonlinearity (GST dose attenuation with low exposures) in the exposure-dose relationship across
exposure levels.
5.4.2.3. Dose-Response Data
Data for liver and lung tumors in male and female B6C3Fi mice following exposure to
airborne dichloromethane were used to develop lURs for dichloromethane (Mennear et al., 1988;
NTP, 1986). As discussed in Section 5.4.1.8, the liver tumor dose-response data were also the
basis of an OSF derived by route-to-route extrapolation using the PBPK models to compare with
an OSF based on liver tumor data in mice exposed to dichloromethane in drinking water (Serota
et al., 1986b). In the NTP (1986) study, significant increases in incidence of liver and lung
adenomas and carcinomas were observed in both sexes of B6C3Fi mice exposed 6 hours/day,
5 days/week for 2 years (Table 5-17). Since significant decreases in survival were observed in
the treated groups of both sexes, the at-risk study populations (represented by the denominators
in the incidence data) were determined by excluding all animals dying prior 52 weeks.
306 DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-17. Incidence data for liver and lung tumors and internal doses
based on GST metabolism dose metrics in male and female B6C3Fi mice
exposed to dichloromethane via inhalation for 2 years
Sex,
tumor type
Male, liver0
Male, lunge
Female, liver0
Female, lunge
BW(g)
-
34.0
32.0
-
34.0
32.0
-
30.0
29.0
-
30.0
29.0
External
dichloromethane
concentration
(ppm)
0
2,000
4,000
0
2,000
4,000
0
2,000
4,000
0
2,000
4,000
Mouse
tumor incidence
22/50 (44%)d
24/47(51%)
33/47 (70%)
5/50 (10%)d
27/47 (55%)
40/47 (85%)
3/47 (6%)d
16/46 (35%)
40/46 (87%)
3/45 (6%)d
30/46 (65%)
41/46 (89%)
Mouse internal
tissue dose"
0
2,363.7
4,972.2
0
475.0
992.2
0
2,453.2
5,120.0
0
493.0
1,021.8
Mouse whole body
metabolism doseb
0
100.2
210.7
0
100.2
210.7
0
104.0
217.0
0
104.0
217.0
aFor liver tumors: mg dichloromethane metabolized via GST pathway/L liver tissue/d from 6 hrs/d, 5 d/wk
exposure; for lung tumors: mg dichloromethane metabolized via GST pathway/L lung tissue/d from 6 hrs/d, 5 d/wk
exposure.
bBased on the sum of dichloromethane metabolized via the GST pathway in the lung plus the liver, normalized to
total BW (i.e., [lung GST metabolism (mg/d) + liver GST metabolism (mg/d)]/kg BW). Units = mg
dichloromethane metabolized via GST pathway in lung and liver/kg-d.
°Hepatocellular carcinoma or adenoma. Mice dying prior to 52 wks were excluded from the denominators.
dStatistically significant increasing trend (by incidental and life-table tests; p < 0.01).
eBronchoalveolar carcinoma or adenoma. Mice dying prior to 52 wks were excluded from the denominators.
Sources: Mennear et al. (1988); NTP (1986).
5.4.2.4. Dose Conversion and Extrapolation Methods: Cancer IVR
Dose conversion. The PBPK model of Marino et al. (2006) for dichloromethane in the
mouse was used to simulate inhalation exposures of 6 hours/day, 5 days/week (Mennear et al.,
1988; NTP, 1986) and to calculate long-term daily average internal doses. Study-, group-, and
sex-specific mean BWs were used. Based on evidence that metabolites of dichloromethane
produced via the GST pathway are primarily responsible for dichloromethane carcinogenicity in
mouse liver and lung (summarized in Section 4.7.3) and the assumption that these metabolites
are sufficiently reactive that they do not have substantial distribution outside these tissues, the
recommended selected internal dose metrics for liver tumors and lung tumors were long-term
average daily mass of dichloromethane metabolized via the GST pathway per unit volume of
liver and lung, respectively (Table 5-17). Figure 5-15 shows the comparison between inhalation
external and internal doses in the liver and lung using this dose metric for the mouse and for the
307
DRAFT - DO NOT CITE OR QUOTE
-------
human. A whole-body metabolism metric was also examined; however, this metric would be
more relevant under a scenario of slowly cleared metabolites that undergo general circulation.
308 DRAFT - DO NOT CITE OR QUOTE
-------
w
o
TJ
10,000
1,000
100
^ _ ^ _ J
T
" ; ..... ; ..... ; [[[ * [[[ ;^j>^v '•" ......................... ^ ......... ^ .................. '
~-j^: :::Sfe:::c:5
"'" -:-±-A--—
r
-
'
Mouse
Human mixed GST
Human GST +/-
Human GST +/+
............ ; ••;"_••; .......... y ....................................
; ::;:::::::
100 1,000
Inhalation concentration (ppm)
10,000
B.
1,000 ;
QJ
(A
O
TJ
C
0.01
-J ...... I ........... I ........ |-
Lung
------
m
-"i' .................. ! .................................. .
— ..--. .-...J ---------- — _.-
i .......... i ...... j ...... I....A...I...J...I. j .................... A .......... 4 ...... ••
.
;EEEEEEEEEEE3EEEEEEEEEEiEEJEEEEE
::j::::::|:::H:::!::+:!::{:H:::::
4 i.-4-44444.j
^U^UJ^UijUUI'iii
:EEEEE=;EE;E:3=E:E
-+ I—H~4
~f~rt:i
—i—-iji-q—i—i--i-t-H—
EEEEEiEEEEEIJjeEEiEEEEIE3EIE
—|.-4.4.444j.
^^4 = ^^t^i^4^Ui^4:
EEEEEIEEEEE£EEEEEEEIEEEEiE3Ej;
4 t___+__t__h.t.H_(.
:::::i:::::i:::Ti:i::[:i:r~
—t t—4—1-:
!:. ...... -\-
^^L I J 1 L .................. i ........ i ...... |.....A....{...J.
-
:Hs3ivdGt«*k8afea»la^^fci-«sass»
::::::::::^:::^::H:::^+^:t:H:::::::::i
IEEEEEEEEEEEEEEEEEEEEIEEEE3EEEI
::::::::::+::::+:::}::+H:+:!
4.—i—4_.i4+4+i
=i^S
^1
^:^::::
_^S-4 +—t—M-+-I
.......... I- ----- — — — --t-K'
— t— >
i
»~-T—V-+-H-H * t—
Mouse
Human mixed GST
Human GST +/-
Human GST +/ +
100 1,000
Inhalation concentration (ppm)
10,000
Average daily doses were calculated from simulated mouse exposures of
6 hours/day, 5 days/week, while simulated human exposures were continuous.
The GST metabolism rate in each simulated human population was obtained by
generating 1,000 random samples from each population (ages 0.5-80 years, males
and females) for each exposure level and calculating the average GST metabolic
rate for each sample.
Figure 5-15. PBPK model-derived internal doses (mg dichloromethane
metabolized via the GST pathways per liter tissue per day) for liver (A) and
lung (B) in mice and humans and their associated external exposures (ppm)
used for the derivation of cancer lURs.
-------
Dose-response modeling and extrapolation. The multistage dose-response model was fit
to the mouse tumor incidence and PBPK model-derived internal dose data to derive mouse
internal BMDio and BMDLio values associated with 10% extra risk (Table 5-18). Different
polynomial models were compared based on adequacy of model fit as assessed by overall ^
goodness of fit (p-value > 0.10)) and examination of residuals at the 0 dose exposure (controls)
and in the region of the BMR (U.S. EPA, 2000c). Appendix E-2 provides details of the BMD
modeling results for the male. The mouse liver and lung tumor risk factors (extra risk per unit
internal dose) were calculated by dividing 0.1 by the mouse BMDLio for liver and lung tumors,
respectively.
310 DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-18. BMD modeling results and tumor risk factors associated with 10% extra risk for liver and lung
tumors in male and female B6C3Fi mice exposed by inhalation to dichloromethane for 2 years, based on liver-
specific GST metabolism and whole body GST metabolism dose metrics
Internal dose
metric"
Tissue-specific
Whole body
Male, liver
Male, lung
Female, liver
Female, lung
Male, liver
Male, lung
Female, liver
Female, lung
BMDS
modelb
MS(1)
MS(1)
MS (2)
MS(1)
MS(1)
MS(1)
MS (2)
MS(1)
x2
goodness of fit
/7-value
0.40
0.64
0.53
0.87
0.40
0.66
0.53
0.88
Mouse BMD10C
913.9
61.7
1,224.1
51.2
38.7
13.1
51.9
10.8
Mouse BMDL10C
544.4
48.6
659.7
40.7
23.1
10.3
28.0
8.6
Allometric-scaled
human BMDL10d
77.8
7.0
94.2
5.8
3.3
1.5
4.0
1.2
Tumor risk factor6
Scaling = 1.0
1.84 x 10'4
2.06 x 10'3
1.52 x 10'4
2.46 x 10'3
-
-
-
-
Allometric-scaled
1.29 x 10'3
1.44 x 10'2
1.06 x 10'3
1.72 x 10'2
3.03 x 10'2
6.80 x 10'2
2.50 x 10'2
8.14 x 10'2
""Tissue-specific dose units = mg dichloromethane metabolized via GST pathway per liter tissue (liver or lung) per d; Whole-body dose units = mg dichloromethane
metabolized via GST pathway in lung and liver/kg-d).
bThe multistage (MS) model in EPA BMDS version 2.0 was fit to the mouse dose-response data shown in Table 5-17 using internal dose metrics calculated with the
mouse PBPK model. Numbers in parentheses indicate: (1) the degree polynomial of the model.
°BMD10 and BMDL10 refer to the BMD-model-predicted mouse internal dose and its 95% lower confidence limit, associated with a 10% extra risk for the incidence
of tumors.
dMouse BMDL10 divided by (BWhuman/BWmouse)025 = 7.
"Dichloromethane tumor risk factor (extra risk per unit internal dose) derived by dividing the BMR (0.1) by the mouse BMDL10 and by the allometric-scaled human
BMDL10, for the scaling =1.0 and allometric-scaled risk factors, respectively.
311
DRAFT - DO NOT CITE OR QUOTE
-------
Linear extrapolation from the internal BMDLio (0.1/BMDLio) was used to derive
inhalation risk factors for lung and liver tumors in male and female mice (Table 5-18). As
discussed in Section 4.7, the linear low-dose extrapolation approach is applied for agents with a
mutagenic mode of action.
Application of allometric scaling factor. As discussed in Section 5.4.1.4, the choice of a
scaling factor is based on the question of the role of specific metabolites and particularly how
these metabolites are transformed or removed. If the key metabolite is established and is known
to be sufficiently reactive to not spread in systemic circulation, then it can be assumed that:
(1) the level of reactivity and rate of clearance (i.e., disappearance due to local reactivity) for this
metabolite per volume tissue is equal in rodents and humans, and (2) risk is proportional to the
long-term daily average concentration of the metabolite. Under these assumptions, rodent
internal BMDLio values based on tissue-specific dichloromethane metabolism require no
allometric scaling to account for toxicodynamic differences and predict the corresponding level
of human risk as a function of the metric (i.e., the scaling factor in Figure 5-13 was equal to 1.0).
(A single metabolite is referenced, but the same argument holds in general for more than one
metabolite). Under this scenario and assumptions, humans and rodents with the same long-term
daily average metabolite formation per volume tissue (e.g., equal internal BMDLio) should both
experience the same long-term average concentration of the metabolite when the metabolite is
highly reactive and, hence, experience the same extra risk. However, some uncertainties remain
concerning the hypothesized role of a highly reactive metabolite in the carcinogenic effects seen
with dichloromethane. The active metabolite(s) have not been established, and data pertaining to
the reactivity or removal (clearance) rate of these metabolite(s) are lacking. For example,
quantitative measurements of adducts of interest or of the half life of relevant compounds in
humans and in mice are not available. It is not known that the rate of reaction is proportional to
the liver perfusion rate, cardiac output, or body surface area, and it is not known that the rate of
reaction is not proportional to these factors. To address these uncertainties, use of a scaling
factor that addresses the possibility that the rate of clearance for the metabolite is limited by
processes that scale allometrically, such as blood perfusion, reaction cofactor supply (e.g.,
antioxidant supply), or enzyme activity, may be appropriate. This case would result in use of a
f\ r\c
mouse:human dose-rate scaling factor of (BWhuman/BWmouse) ' = 7 to adjust the mouse-based
BMDLio values downward. Using this internal dose metric (liver-specific metabolism with
allometric scaling), equivalent rodent and human internal BMDLio values result in a human liver
tumor risk factor (0.1/BMDLio) that is assumed equal to that for the mouse, given a 70-year
lifetime exposure. Another alternative that can be used is based on an allometrically-scaled,
whole-body metabolism metric. In this case, less weight is given to the evidence of site-
specificity of the effects. As with the OSF derivations, the cancer toxicity values derived using
each of these metrics and scaling factors (i.e., liver-specific metabolism with and without
312 DRAFT - DO NOT CITE OR QUOTE
-------
allometric-scaling and the whole-body metabolism metric) are presented in the following tables.
Considering the lack of data pertaining to clearance rates or the actual AUC of the active
carcinogenic metabolite(s) in mice and humans, the lURs recommended by EPA are based on
the allometrically-scaled tissue-specific GST metabolism rate dose metric.
Calculation oflURs. A probabilistic PBPK model for dichloromethane in humans,
adapted from David et al. (2006) (see Appendix B), was used with Monte Carlo sampling to
calculate distributions of internal lung, liver, or blood doses associated with chronic unit
inhalation (1 ug/m3) exposures. The data on which the model is based indicate that the
relationship between exposure and internal dose is linear at low doses. Parameters in the human
PBPK model developed by David et al. (2006) are distributions that incorporate information
about dichloromethane toxicokinetic variability and uncertainty among humans. Monte Carlo
sampling was performed in which each human model parameter was defined by a value
randomly drawn from each respective parameter distribution. The model was then executed by
using the external unit exposure as input, and the resulting human equivalent internal dose was
recorded. This process was repeated for 10,000 iterations to generate a distribution of human
internal doses.
The resulting distribution of lURs shown in Table 5-19 was derived by multiplying the
human internal dose tumor risk factor (in units of reciprocal internal dose) by the respective
distributions of human average daily internal dose resulting from a chronic unit inhalation
exposure of 1 ug/m3 dichloromethane. Table 5-19 presents the analysis using the male data.
Analyses based on the female data produced very similar results, and are summarized in
Appendix F. The mean slope factor was selected as the recommended value; other values at the
upper end of the distribution are also presented. As with the cancer OSF derivation, the cancer
IUR is derived for a population composed entirely of carriers of the GST-T1 homozygous
positive genotype (the group that would be expected to be most sensitive to the carcinogenic
effects of dichloromethane), and a population reflecting the estimated frequency of GST-T1
genotypes in the current U.S. population (20% GST-Tl^, 48% GST-T1+/', and 32% GST-T1+/+,
the "mixed" population). All simulations also included a distribution of CYP activity, based on
data from Lipscomb et al. (2003).
313 DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-19. lURs for dichloromethane based on PBPK model-derived internal liver and lung doses in B6C3Fi
male mice exposed via inhalation for 2 years, based on liver-specific GST metabolism and whole body
metabolism dose metrics, by population genotype
Internal dose
metric and
scaling factor"
Tissue-specific,
allometric -scaled
Tissue-specific,
scaling = 1.0
Whole-body,
allometric -scaled
Population
genotype1"
GST-T1+/+
GST-T1+/+
Mixed
Mixed
GST-T1+/+
GST-T1+/+
Mixed
Mixed
GST-T1+/+
GST-T1+/+
Mixed
Mixed
Tumor
type
Liver
Lung
Liver
Lung
Liver
Lung
Liver
Lung
Liver
Lung
Liver
Lung
Human tumor
risk factor0
1.29 x ID'3
1.44 x ID'2
1.29 x ID'3
1.44 x ID'2
1.84 x ID'4
2.06 x ID'3
1.84 x ID'4
2.06 x ID'3
3.03 x 10'2
6.80 x 10'2
3.03 x 10'2
6.80 x 10'2
Distribution of human internal
dichloromethane doses from 1 jig/m3
exposure"1
Mean
6.61 x 1Q-6
3.89 x 1Q-7
3.71 x 1Q-6
2.20 x 1Q-7
6.61 x 1Q-6
3.89 x 1Q-7
3.71 x 1Q-6
2.20 x 1Q-7
1.80 x 10'7
1.80 x 10'7
1.01 x 10'7
1.01 x 10'7
95th
percentile
2.21 x 1Q-5
1.24 x 1Q-6
1.43 x 1Q-5
8.06 x 1Q-7
2.21 x 1Q-5
1.24 x 1Q-6
1.43 x 1Q-5
8.06 x 1Q-7
6.38 x 10'7
6.38 x 10'7
4.00 x 10'7
4.00 x 10'7
99th
percentile
4.47 x 1Q-5
2.42 x 1Q-6
3.03 x 1Q-5
1.69 x 1Q-6
4.47 x 1Q-5
2.42 x 1Q-6
3.03 x 1Q-5
1.69 x 1Q-6
1.41 x 10'6
1.41 x 10'6
9.43 x 10'7
9.43 x 10'7
Resulting candidate human
lUR'Cug/m3)1
Mean
8.5 x 1Q-9
5.6 x 1Q-9
4.8 x 1Q-9
3.2 x 1Q-9
1.2 x 1Q-9
8.0 x 1Q-10
6.8 x 1Q-10
4.5 x IQ-10
5.5 x 10'9
1.2 x 10'8
3.1 x 10'9
6.9 x 10'9
95th
percentile
2.8 x 1Q-8
1.8 x 1Q-8
1.8 x KT8
1.2 x 1Q-8
4.1 x 1Q-9
2.6 x 1Q-9
2.6 x ID'9
1.7 x ID'9
1.9 x 10'8
4.3 x 10'8
1.2 x 1(T8
2.7 x 10'8
99th
percentile
5.8 x ID'8
3.5 x ID'8
3.9 x ID'8
2.4 x ID'8
8.2 x ID'9
5.0 x ID'9
5.6 x ID'9
3.5 x ID'9
4.3 x 10'8
9.6 x 10'8
2.9 x 10'8
6.4 x 10'8
aTissue specific dose units = mg dichloromethane metabolized via GST pathway per liter tissue (liver or lung, respectively, for liver and lung tumors) per d;
whole-body dose units = mg dichloromethane metabolized via GST pathway in lung and liver/kg-d.
bGST-Tl+/+ = homozygous, full enzyme activity;); mixed = population reflecting estimated frequency of genotypes in current U.S. population: 20% GST-T"'",
48% GST-T1+A, and 32% GST-T1+/+ (Haber et al., 2002).
°Dichloromethane tumor risk factor (extra risk per unit internal dose) derived by dividing the BMR (0.1) by the allometric-scaled human BMDL10 or by the mouse
BMDL10 (from Table 5-18) for the allometric-scaled and scaling =1.0 risk factors, respectively.
dMean, 95th, and 99th percentile of the human PBPK model-derived probability distribution of daily average internal dichloromethane dose resulting from chronic
exposure to 1 ug/m3 (0.00029 ppm).
Derived by multiplying the dichloromethane tumor risk factor by the PBPK model-derived probabilistic internal doses from daily exposure to 1 ug/m3.
314
DRAFT - DO NOT CITE OR QUOTE
-------
5.4.2.5. Cancer IUR
The recommended cancer lURs are 9 x 10~9 (ug/m3)'1 and 6 x 10"9 (ug/m3)'1 for the
development of liver and lung cancer, respectively, based on the mean for the GST-T1+/+
population (the group with the greatest presumed sensitivity). These values are based on male
B6C3Fi mice, using a tissue-specific GST metabolism dose metric with allometric scaling
(Table 5-19). Risk estimates were similar to the values based on female mice in the NTP (1986)
inhalation study: 7 x 10~9 (ug/m3)'1 and 7 x 10~9 (ug/m3)'1 for the development of liver and lung
cancer, respectively, in the GST-T1++ population (see Appendix F).
Consideration of combined risk (summing risk across tumors). With two significant
tumor sites, focusing on the more sensitive response may underestimate the overall cancer risk
associated with exposure to this chemical. Following the recommendations of the NRC (1994)
and the Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), an upper bound on total
risk was estimated in order to gain some understanding of the total risk from multiple tumor sites
in the selected data set. Note that this estimate of overall risk describes the risk of developing
either tumor type, not just the risk of developing both simultaneously.
NRC (1994) stated that an approach based on counts of animals with one or more tumors
(or tumor-bearing animals) would tend to underestimate overall risk when tumor types occur
independently and that an approach based on combining the risk estimates from each separate
tumor type should be used. For dichloromethane, there is no reason to expect that the occurrence
of one tumor type depends on the incidence of the other, given the association of different dose
metrics with each tumor response. Therefore, it appears reasonable to assume that the two tumor
types occur independently. However, simply summing upper limit risks may result in an
overestimation of overall combined risk because of the statistical issues with respect to summing
variances of distributions. An additional challenge results from the use of different internal dose
metrics for different tumors, as is the case with the dose metrics based on tissue-specific
metabolism. Statistical methods based on a common metric cannot be used with the tissue-
specific metabolism metric used in these derivations.
An alternative approach is to derive an upper bound on the combined risk estimates by
summing central tendency risks and calculating a pooled SD by using BMDio and BMDLio
values for liver and lung tumors. The SD associated with the IUR for each tumor site is
calculated as the difference between 95* percentiles of the distribution for upper bound and
maximum likelihood estimate lURs (based on either female or male mouse tumor risk factors),
divided by 1.645 (the relevant t statistic, assuming normal distributions of summed quantities).
Variances for each tumor site are the squares of the SDs. Pooled variance and SD are defined as
the sum of variances for lung and liver tumors and the square root of that sum, respectively.
Finally, the upper bound on the combined lung and liver cancer risk is determined by multiplying
the cumulative SD by 1.645 and adding it to the summed central tendency lURs. The
315 DRAFT - DO NOT CITE OR QUOTE
-------
calculations of these upper bound estimates for combined liver and lung tumor risks are shown in
Table 5-20.
316 DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-20. Upper bound estimates of combined human lURs for liver and lung tumors resulting from lifetime
exposure to 1 ug/m3 dichloromethane based on liver-specific GST metabolism and whole body metabolism dose
metrics, by population genotype
Internal dose
metric and scaling
factor"
Tissue-specific,
allometric-scaled
Tissue-specific,
scaling = 1.0
Whole-body,
allometric-scaled
Population
genotype1"
GST-T1+/+
Mixed
GST-T1+/+
Mixed
GST-T1+/+
Mixed
Tumor site
Liver
Lung
Liver or lung
Liver
Lung
Liver or lung
Liver
Lung
Liver or lung
Liver
Lung
Liver or lung
Liver
Lung
Liver or lung
Liver
Lung
Liver or lung
Upper bound
IURC
8.5 x ID'9
5.6 x ID'9
4.8 x ID'9
3.2 x ID'9
1.2 x ID'9
8.0 x ID'10
6.8 x ID'10
4.5 x ID'10
5.5 x 10'9
1.2 x 10'8
3.1 x 10'9
6.9 x 10'9
Central
tendency IURd
5.1 x ID'9
4.4 x ID'9
9.5 x ID'9
2.8 x ID'9
2.5 x ID'9
5.3 x ID'9
7.2 x ID'10
6.3 x ID'10
1.4 x ID'9
4.1 x ID'10
3.6 x ID'10
7.6 x ID'10
3.3 x 10'9
9.6 x 1(T9
1.3 x 10'8
1.8 x 1(T9
5.4 x 10'9
7.2 x 1(T9
Variance of
tissue-specific
tumor risk6
4.36 x 1Q-18
5.18 x ID'19
1.37 x ID'18
1.66 x ID'19
8.91 x ID'20
1.07 x ID'20
2.81 x ID'20
3.41 x ID'21
1.79x 10"18
2.55 x 10'18
5.62 x 10'19
8.03 x 10'19
Combined
tumor risk SDf
2.2 x ID'9
1.2 x ID'9
3.2 x ID'10
1.7 x ID'10
2.1 x 10'9
1.2 x 10'9
Upper bound on
combined tumor risk8
(jig/m3)-1
1.3 x ID'8
7.4 x ID'9
1.9 x ID'9
1.1 x ID'9
1.6 x 10'8
9.2 x 10'9
aTissue specific dose units = mg dichloromethane metabolized via GST pathway per liter tissue (liver or lung, respectively, for liver and lung tumors) per d;
whole-body dose units = mg dichloromethane metabolized via GST pathway in lung and liver/kg-d.
bGST-Tl+/+ = homozygous, full enzyme activity); mixed = population reflecting estimated frequency of genotypes in current U.S. population: 20% GST-T"7", 48%
GST-Tl+/~, and 32% GST-T1+/+ (Haber et al., 2002).
"Estimated at the human equivalent BMDL10 (0.1/BMDL10) (see Table 5-18).
Estimated at the human equivalent BMD10 (0.1/BMD) (see Table 5-18).
Calculated as the square of the difference of the upper bound and central tendency lURs divided by the / statistic, 1.645.
Calculated as the square root of the sum of the variances for liver and lung tumors.
Calculated as the product of the cumulative tumor risk SD and the t statistic, 1.645, added to the sum of central tendency lURs.
317
DRAFT - DO NOT CITE OR QUOTE
-------
Using this approach and the male mouse-derived risk factors, the combined human
O Q 1 O
equivalent IUR values for both tumor types is 1 x 10" (ug/m )" (rounded from 1.3 x 10" ) in the
most sensitive (GST-T1+/+) population. This is the recommended inhalation cancer unit risk
value to be used for chronic exposure to dichloromethane. The corresponding value for a
population with the frequency distribution of GST-T1 genotypes currently found in the U.S.
population is 7 x 10"9 (ug/m3)"1.
5.4.2.6. Comparative Derivation Based on Rat Mammary Tumor Data
Mammary gland tumor data from male and female F344 rats following an inhalation
exposure to dichloromethane were considered in development of a comparative IUR for
dichloromethane (Mennear et al., 1988; NTP, 1986). In both the male and female rats, there
were significant increases in the incidence of adenomas, fibroadenomas, or fibromas in or near
the mammary gland. These were characterized as benign tumors in the NTP report (NTP, 1986).
Increased numbers of benign mammary tumors per animal in exposed groups were also seen in
two studies of Sprague-Dawley rats (Nitschke et al., 1988a; Burek et al., 1984). A gavage study
in Sprague-Dawley rats reported an increased incidence of malignant mammary tumors, mainly
adenocarcinomas (8, 6, and 18% in the control, 100, and 500 mg/kg dose groups, respectively),
but the increase was not statistically significant. Data were not provided to allow an analysis that
accounts for differing mortality rates (Maltoni et al., 1988). There are considerably more
uncertainties regarding the interpretation of these data with respect to carcinogenic risk
compared with the data pertaining to liver and lung tumors. The trends were driven in large part
by benign tumors; adenocarcinomas and carcinomas were seen only in the females with
incidences of 1, 2, 2, and 0 in the 0, 1,000, 2,000, and 4,000 ppm exposure groups, respectively.
There are little data to guide the choice of relevant dose metric, and the genotoxicity and
mechanistic studies have not included mammary tissue. For these reasons, the analysis and the
calculation of the comparative IUR based on rat mammary tumor data are presented in
Appendix G. The IUR based on the female rat data was 1 x 10"7 (ug/m3)"1.
5.4.2.7. Alternative Based on Administered Concentration
Another comparison that can be made is with an alternative IUR based on liver and lung
tumors in mice using the external concentrations of dichloromethane in the mouse studies as
converted to HECs, and then applying this using BMD modeling to obtain the BMDLio and
resulting IUR. Mouse bioassay exposures were adjusted to HECs as follows:
• Adjusting to continuous exposure: External concentrationADJ = External
concentration x (6 hours/24 hours) x (5 days/7 days);
• Concentrations in mg/m3 = concentrations in ppm x 84.93/24.45; and
318 DRAFT - DO NOT CITE OR QUOTE
-------
• [Hb/g]A/[Hb/g]n = the ratio of blood:gas (air) partition coefficients in animals and
humans. Because the partition coefficient for mice (23.0) is higher than for humans
(9.7), a value of 1.0 was used, as per U.S. EPA (1994b) guidance.
Thus, HECs = External concentrationADJ x [Hb/g]A/[Hb/g]n = External concentrationADJ x 1 •
The HECs (mg/m3) for the 0, 2,000, and 4,000 ppm exposure groups were 0, 1,241, and
2,481 mg/m3, respectively. The BMD modeling and lURs derived from these values, in
conjunction with the liver and lung tumor data from Table 5-17 (NTP, 1986), are shown in
Table 5-21. The resulting lURs based on the liver or lung tumors in the mouse are
approximately one order of magnitude higher than the currently recommended value obtained by
using the mouse and human PBPK models.
Table 5-21. Inhalation units risks based on human BMDL10 values using
administered concentration for liver and lung tumors in B6C3Fi mice
exposed by inhalation to dichloromethane for 2 years
Sex,
tumor type
Male, liver
Male, lung
Female, liver
Female, lung
BMDS model3
MS(1)
MS(1)
MS (2)
MS(1)
x2
goodness of fit
/7-value
0.37
0.54
0.38
0.77
BMD10b
463.89
157.23
601.84
126.40
BMDL10b
276.15
124.10
342.83
100.61
Inhalation
unit risk0
(jig/m3)-1
3.6 x 10'7
8.1 x 10"7
2.9 x 10'7
9.9 x 10"7
aThe multistage (MS) model in EPA BMDS version 2.0 was fit to each of the four sets of mouse dose-response
data shown in Table 5-17. The HEC used in these models for the 0, 2,000, and 4,000 ppm exposure groups were
0, 1,241, and 2,481 mg/m3, respectively. Numbers in parentheses indicate: (1) the lowest degree polynomial of
the model showing an adequate fit.
°BMD10 and BMDL10 refer to the BMD-model-predicted HECs (mg dichloromethane per cubic meter), and its
95% lower confidence limit associated with a 10% extra risk for the incidence of tumors.
dIUR (risk/ug-m3) = 0. I/human BMDL10.
Sources: Mennear et al. (1988); NTP (1986).
The difference between the administered concentration methodology and PBPK-based
approaches depends on two key differences: the use of a dichloromethane-metabolite dose-
metric rather than dichloromethane AUC, and the fact that the rate of dichloromethane
conversion to that metabolite is estimated in humans by using human data rather than default
allometric scaling (BW075). In addition, the administered concentration methodology does not
account in any way for the GST polymorphism and so might be considered as representing the
general/mixed-GST-genotype population rather than the +/+ subpopulation.
5.4.2.8. Previous IBIS Assessment: Cancer IVR
The IUR in the previous IRIS assessment was determined from the combined incidence
of liver and lung adenomas and carcinomas in B6C3Fi mice exposed to dichloromethane for
319
DRAFT - DO NOT CITE OR QUOTE
-------
2 years by NTP (1986). A value of 4.7 x 10~7 (ug/m3)"1 was derived by the application of a
modified version of the PBPK model of Andersen et al. (1987), which incorporated the
pharmacokinetics and metabolism of dichloromethane. Internal dose estimates based on
dichloromethane metabolism via the GST pathway were used and corrected for differences in
interspecies sensitivity by applying to the human risks an interspecies scaling factor of 12.7,
which was based on dose equivalence adjusted to BW to the 2/3 power (Rhomberg, 1995; U.S.
EPA, 1987a).
5.4.2.9. Comparison of Cancer IUR Using Different Methodologies
In this assessment, cancer lURs derived by using different dose metrics and assumptions
were examined, as summarized in Table 5-22. The recommended IUR value of 1 x 10~8 (ug/m3)'1
is based on a tissue-specific, GST-internal dose metric with allometric scaling because of the
evidence for the involvement of highly reactive metabolites formed via the GST pathway. The
value derived specifically for the GST-T1+/+ population is recommended to provide protection for
the population that is hypothesized to be most sensitive to the carcinogenic effect. The values
based on the GST-T1+ + group are approximately two- to fivefold higher than those for the full
population for all dose metrics used in this assessment. Within a genotype population, the values
of the IUR among the various dose metrics vary by about one to two orders of magnitude.
320 DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-22. Comparison of lURs derived by using various assumptions and metrics
Population"
GST-T1+/+
GST-T1+/+
GST-T1+/+
GST-T1+/+
GST-T1+/+
GST-T1+/+
GST-T1+/+
GST-T1+/+
GST-T1+/+
Mixed
Mixed
Mixed
Mixed
Mixed
Mixed
Mixed
Mixed
Mixed
Dose metric
Tissue-specific GST-metabolism rate0
Tissue-specific GST-metabolism rate
Tissue-specific GST-metabolism rate
Tissue-specific GST-metabolism rate
Tissue-specific GST-metabolism rate
Tissue-specific GST-metabolism rate
Whole-body GST metabolism rate
Whole-body GST metabolism rate
Whole-body GST metabolism rate
Tissue-specific GST-metabolism rate
Tissue-specific GST-metabolism rate
Tissue-specific GST-metabolism rate
Tissue-specific GST-metabolism rate
Tissue-specific GST-metabolism rate
Tissue-specific GST-metabolism rate
Whole-body GST metabolism rate
Whole-body GST metabolism rate
Whole-body GST metabolism rate
Administered concentration (HEC)
Administered concentration (HEC)
1995 IRIS assessment"
Species, sex
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Tumor type
Liver and lung
Liver
Lung
Liver and lung
Liver
Lung
Liver and lung
Liver
Lung
Liver and lung
Liver
Lung
Liver and lung
Liver
Lung
Liver and lung
Liver
Lung
Liver
Lung
Liver and lung
Scaling
factor
7.0
7.0
7.0
1.0
1.0
1.0
7.0
7.0
7.0
7.0
7.0
7.0
1.0
1.0
1.0
7.0
7.0
7.0
12.7
IURb
(jig/m3)-1
1.3 x 10 8
8.5 x 10'9
5.6 x 10'9
1.9 x KT9
1.2 x 10'9
8.0 x 10'10
1.6 x 10'8
5.5 x 1Q-9
1.2 x 1Q-8
7.4 x 1Q-9
4.8 x 1Q-9
3.2 x 1Q-9
1.1 x KT9
6.8 x 1Q-10
4.5 x IQ-10
9.2 x 1Q-9
3.1 x KT9
6.9 x 1Q-9
3.6 x 1Q-7
8.1 x KT7
4.7 x 10'7
Source
(Table)
Table 5-20
Table 5-19
Table 5-19
Table 5-20
Table 5-19
Table 5-19
Table 5-20
Table 5-19
Table 5-19
Table 5-20
Table 5-19
Table 5-19
Table 5-20
Table 5-19
Table 5-19
Table 5-20
Table 5-19
Table 5-19
Table 5 -21
Table 5 -21
aGST-Tl+/+ = homozygous, full enzyme activity; mixed = genotypes based on a population reflecting the estimated frequency of genotypes in the current U.S.
population: 20% GST-Tr7', 48% GST-T1+A, and 32% GST-T1+/+ (Haber et al., 2002).
cased on mean value of the derived distributions.
°Bolded value is the basis for the recommended IUR of 1 x 10"8 ug/m3 per mg/kg-d.
321
DRAFT - DO NOT CITE OR QUOTE
-------
5.4.3. Differences Between Current Assessment and Previous IRIS PBPK-based
Assessment
To better understand the changes in assessment risk predictions between previous EPA
evaluations and the current assessment, the differences in PBPK model parameters for the
B6C3Fi mouse were evaluated. Values that differed significantly between the model version
used previously and that of Marino et al. (2006), along with derived group parameters that lend
further insight, are shown in Table 5-23.
Table 5-23. Comparison of key B6C3Fi mouse parameters differing
between prior and current PBPK model application
Parameter"
Partition coefficients
PB blood/air
PF fattolood
PF-PB (fat/blood)- (blood/air) = fat/air
PL liveiyblood
PL-PB (liveiyblood)- (blood/air) = liver/air
PLu lung (tissue)/blood
PLu- PB (lung/blood) • (blood/air) = lung/air
PR rapidly perfused^lood
PR-PB rapidly perfused/air
PS slowly perfused/blood
PS-PB slowly perfused/air
Flow rates
QCC cardiac output (L/hr/kg1174)
VPR ventilation:perfusion ratio
Metabolism parameters
VmaxC maximum CYP metabolic rate (mg/hr/kg0 7)
Km CYP affinity (mg/L)
VmaxC/Km (L/hr/kg07)
Al ratio of lung VmaxC to liver VmaxC
Total lung + liver VmaxC/Km
kfc first-order GST metabolic rate constant (kg0 3/hr)
A2 ratio of lung kfc to liver kfc
Total lung + liver kfc
Marino et al. (2006); mean
values as applied (posterior)
23
5.1
117.3
1.6
36.8
0.46
10.6
0.52
12.0
0.44
10.1
24.2
1.45
9.27
0.574
16.1
0.207
19.5
1.41
0.196
1.69
U.S. EPA
(1988b, 1987a, b)
8.29
14.5
120.2
1.71
14.2
1.71
14.2
1.71
14.2
0.96
7.96
14.3
1.0
11.1
0.396
28
0.416
39.7
1.46
0.137
1.66
""Parameters not listed differed by <10% between versions. See Table 3-5 and associated text for details.
While a number of the tissue:blood partition coefficients in Table 5-23 differ significantly
between the two models (e.g., PF, PLu, and PR), the corresponding tissue:air coefficients (e.g.
the products PF-PB, Plu-PB, and PR-PB) generally do not. Since the latter tend to determine the
long-term equilibration between the tissue (tissue group) and air, the differences in the
tissue:blood coefficients are not expected to significantly impact long-term risk predictions.
Thus, the partition coefficients that most significantly differ (the blood:air and liverair partition
coefficients) are, respectively, 2.8- and 2.6-fold higher in the current version. The increased PB
results in a tendency for simulated blood concentrations to rise more quickly and reach higher
322
DRAFT - DO NOT CITE OR QUOTE
-------
values with other parameters being equal. The significantly increased QCC and VPR contribute
even more to this difference, resulting in an even faster rise to steady state during inhalation
exposure simulations but also more rapid delivery to the liver (decreasing blood-flow limitation
of hepatic metabolism) and more rapid exhalation. The increased liverair partition coefficient
leads to higher predicted liver concentrations (again, other parameters being equal) and hence,
higher rates of metabolism.
For metabolism, a much reduced oxidative metabolism is seen, which at low doses
depends on Vmaxc/Km. The revised hepatic metabolism is over 40% lower and the total of lung
plus liver metabolism is 50% lower than previously used. This lower rate of metabolism means
that far less of parent dichloromethane will be removed through metabolism and hence, predicted
blood concentrations will be higher still relative to the impact of changes in partition coefficient,
QCC, and VPR, as noted above.
The result of having higher predicted blood and liver dichloromethane concentrations is
that, while the GSH-pathway metabolic constant, kfc, is virtually the same for the mouse liver in
both cases, the much higher concentration of dichloromethane available will lead to a much
higher predicted rate of metabolism via this pathway. For the lung, since the A2 is 43% higher
in the model of Marino et al. (2006), the relative increase will be even greater.
Because the revised rate of GST metabolism in mice was estimated by using data with
CYP2E1 inhibited by a suicide inhibitor, there is considerable confidence in the relative rate of
metabolism by these two pathways and the GST pathway in particular. The partition coefficients
used by Marino et al. (2006) are as measured by Clewell et al. (1993) and expected to be at least
as reliable as those used in the 1995 assessment. Considering that the revised PBPK model does
an excellent job of reproducing closed-chamber gas uptake data that were not available for
calibration of the 1987 model, as well as blood concentrations after intravenous injection, there is
fairly high confidence in its predictions.
The net result of these model changes is that, under mouse bioassay conditions, the
predicted dose metrics for liver and lung cancer (i.e., GST-mediated metabolism) are higher than
those obtained with the previous model, resulting in a lower risk estimated per unit of dose.
The other piece of the PBPK-based risk estimation is the human model. In updating the
parameter estimates for the human model (see Appendix B for details), the oxidative metabolism
Vmaxc/Km approximately doubled, which leads to lower predicted blood concentrations of
dichloromethane available for metabolism by GST. In addition, the liver GST was reduced by
almost 60%, and the lung:liver GST ratio decreased by almost fivefold, for a net change in lung
GST of over 90%. Given the larger human data set available to David et al. (2006) and the
sophisticated Bayesian analysis used to recalibrate the model, the expectation is that these values
are more reliable than the values used in the 1995 IRIS assessment.
Since actual rates of metabolism at a given exposure level also depend on respiration rate
and blood flows, these changes in metabolic parameters do not completely determine the relative
323 DRAFT - DO NOT CITE OR QUOTE
-------
(predicted) dosimetry. But the difference in cancer risk predictions between the current and
previous assessments is primarily explained by the overall prediction of higher GST-mediated
dosimetry in the mouse (at bioassay conditions) and lower human GST metabolism (due in part
to greater predicted clearance of dichloromethane by oxidative metabolism). In addition to these
changes in PBPK parameters, the reduction of scaling factor from 12.7 to 7 is a significant factor
in the overall change from the previous assessments.
5.4.4. Application of Age-Dependent Adjustment Factors (ADAFs)
The available dichloromethane studies do not include an evaluation of early-life
susceptibility to dichloromethane cancer risk. In the absence of this type of data, and if a
chemical follows a mutagenic mode of action for carcinogenicity like dichloromethane, the
Supplemental Guidance for Assessing Susceptibility from Early-Life Exposure to Carcinogens
(U.S. EPA, 2005b) recommends that ADAFs be applied to the cancer values. Since the OSF of
1 x 10'3 (mg/kg-day)"1 and the IUR of 1 x 10'8 (ug/m3)'1 were calculated from adult
dichloromethane exposures, early-life cancer susceptibility has not been accounted for in these
values, and ADAFs need to be applied in combination with exposure information when
estimating cancer risks that include early-life exposures. Sample calculations that incorporate
ADAFs into the cancer risks are presented in subsequent sections. Additional examples of
evaluations of cancer risks incorporating early-life exposure are provided in Section 6 of the
Supplemental Guidance (U.S. EPA, 2005b).
In the Supplemental Guidance (U.S. EPA, 2005b), ADAFs are established for three age
groups. An ADAF of 10 is applied for age groups <2 years, 3 is applied for ages 2-<16 years,
and 1 is applied for >16 years (U.S. EPA, 2005b). The 10- and 3-fold adjustments in cancer
values are combined with age-specific exposure estimates when early-life exposure
considerations need to be included in cancer risk estimates. The most current information on
usage of ADAFs can be found at http://www.epa.gov/cancerguidelines. For estimation of risk,
the Supplemental Guidance (U.S. EPA, 2005b) recommends obtaining and using age-specific
values for exposure and cancer potency. In the absence of age-specific cancer potency values, as
is true for dichloromethane, age-specific cancer values are estimated by using the appropriate
ADAFs. Using this process, a cancer risk is derived for each age group. The risks are summed
across the age groups to get the total cancer risk for the age-exposure period of interest.
5.4.4.1. Application of ADAFs in Oral Exposure Scenarios
Sample calculations incorporating the use of ADAFs are presented for three exposure
duration scenarios. These scenarios include full lifetime exposure (assuming a 70-year lifespan),
and two 30-year exposures at ages 0-30 and ages 20-50. A constant dichloromethane exposure
of 1 mg/kg-day was assumed for each scenario.
324 DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-24 lists the four factors (ADAFs, OSF, assumed exposure, and duration
adjustment) that are needed to calculate the partial cancer risk based on the early age-specific
group. The partial cancer risk for each age group is the product of the four factors in columns 2-
5. Therefore, the partial cancer risk following daily dichloromethane oral exposure in the age
group 0 to <2 years is the product of the values in columns 2-5 or 10 x (2 x 10~3) x 1 x 2/70 =
5.7 x io~4. The partial risks that are listed in the last column of Table 5-24 are added together to
get the total risk. Thus, a 70-year (lifetime) risk estimate for continuous exposure to 1 mg/kg-
day dichloromethane is 3.3 x 10"3, which is adjusted for early-life susceptibility and assumes a
70-year lifetime and constant exposure across age groups.
Table 5-24. Application of ADAFs to dichloromethane cancer risk following
a lifetime (70-year) oral exposure
Age group (yrs)
0-<2
2-<16
>16
ADAF
10
3
1
Unit risk
(per mg/kg-d)
2 x ID'3
2 x ID'3
2 x ID'3
Exposure concentration
(mg/kg-d)
1
1
1
Duration
adjustment
2 yrs/
70 yrs
14 yrs/
70 yrs
54 yrs/
70 yrs
Total risk
Partial risk
5.7 x 1Q-4
1.2 x 1Q-3
1.5 x KT3
3.3 x 10 3
In calculating the cancer risk for a 30-year constant exposure to dichloromethane at an
exposure level of 1 mg/kg-day from ages 0-30, the duration adjustments would be 2/70, 14/70,
and 14/70, and the partial risks for the three age groups would be 5.7 x 10"4, 1.2 x 10"3, and 4.0 x
10"4, which would result in a total risk estimate of 2.2 x 10"3.
In calculating the cancer risk for a 30-year constant exposure to dichloromethane at an
exposure level of 1 mg/kg-day from ages 20-50, the duration adjustments would be 0/70, 0/70,
and 30/70. The partial risks for the three groups are 0, 0, and 8.6 x 10"4, which would result in a
total risk estimate of 8.6 x 10"4.
5.4.4.2. Application of ADAFs in Inhalation Exposure Scenarios
Sample calculations incorporating the use of ADAFs are presented for three exposure
duration scenarios involving inhalation exposure. These scenarios include full lifetime exposure
(assuming a 70-year lifespan) and two 30-year exposures from ages 0-30 and ages 20-50. A
constant dichloromethane inhalation exposure of 1 ug/m3 was assumed for each scenario.
Similar to the oral exposure scenarios presented in Section 5.4.4.1, Table 5-25 lists the
four factors (ADAFs, unit risk, assumed exposure, and duration adjustment) that are needed to
calculate the partial cancer risk based on the early age-specific group. The partial cancer risk for
each age group is the product of the four factors in columns 2-5. Therefore, the partial cancer
325
DRAFT - DO NOT CITE OR QUOTE
-------
risk following daily dichloromethane inhalation exposure in the age group 0 to <2 years is the
product of the values in columns 2-5 or 10 x (1 x icr8) x 1 x 2/70 = 2.9 x 10~9. The partial risks
that are listed in the last column of Table 5-25 are added together to get the total risk. Thus, a
70-year (lifetime) risk estimate for continuous exposure to 1 ug/m3 dichloromethane is 1.7 x
o
10" , which is adjusted for early-life susceptibility and assumes a 70-year lifetime and constant
exposure across age groups.
Table 5-25. Application of ADAFs to dichloromethane cancer risk following
a lifetime (70-year) inhalation exposure
Age group (yrs)
0-<2
2-<16
>16
ADAF
10
3
1
Unit risk
(jig/m3)
1 x ID'8
1 x ID'8
1 x ID'8
Exposure concentration
(jig/m3)
1
1
1
Duration
adjustment
2 yrs/
70 yrs
14 yrs/
70 yrs
54 yrs/
70 yrs
Total risk
Partial risk
2.9 x 1Q-9
6.0 x 1Q-9
7.7 x 10'9
1.7 x 10 8
In calculating the cancer risk for a 30-year constant exposure to dichloromethane at a
concentration of 1 ug/m3 from ages 0-30, the duration adjustments would be 2/70, 14/70, and
14/70, and the partial risks for the three age groups are 2.9 x 10"9, 6.0 x 10"9, and 2.0 x 10"9.
These partial risks result in a total risk estimate of 1.1 x 10"8.
In calculating the cancer risk for a 30-year constant exposure to dichloromethane at a
concentration of 1 ug/m3 from ages 20-50, the duration adjustments would be 0/70, 0/70, and
30/70, and the partial risks for the three age groups are 0, 0, and 4.3 x 10"9, resulting in a total
risk estimate of 4.3 x 10"9.
5.4.5. Uncertainties in Cancer Risk Values
The derivation of cancer risk values often involves a number of uncertainties in the
extrapolation of dose-response data in animals to cancer risks in human populations. Several
types of uncertainty have been quantitatively integrated into the derivation of the recommended
OSFs and lURs for dichloromethane, while others are qualitatively considered. Table 5-26 and
the ensuing discussion summarize the principal uncertainties identified, their possible effects on
the cancer risk values, and decisions made in the derivations.
326
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-26. Summary of uncertainty in the derivation of cancer risk
values for dichloromethane
Consideration and
Impact On Cancer Risk Value
Decision
Justification and Discussion
Selection of data set
(Selection of an alternative data
set could change the
recommended cancer risk values.)
Select NTP( 1986) as
principal inhalation study
and Serota et al. (1986b) as
principal oral (drinking
water) study to derive
cancer risks for humans
NTP (1986) inhalation mouse bioassay
provides the strongest cancer responses
(liver and lung tumors) and the best dose-
response data in the animal database. The
oral mouse study (Serota et al. (1986b;
Hazleton Laboratories, 1983) provide an
adequate basis for dose-response modeling
Selection of target organ
(Selection of a target organ could
change the recommended cancer
risk values.)
Liver, and for inhalation,
lung selected as target
organ. Cancer risk values
based on mammary gland
tumors in rats also
examined; potential brain
cancer risk and
hematopoetic cancer risk
were identified as data gaps.
The evidence for mammary gland tumors
from dichloromethane exposure is less
consistent than evidence for liver and lung
tumors. Inhalation cancer risk values based
on mammary tumors in rats are about one
order of magnitude higher than risk values
based on liver or lung tumors in mice. No
data are available to allow derivation of unit
risks based on brain or hematopoietic
cancers.
Selection of extrapolation
approach
(Selection of extrapolation
approach could change the
recommended cancer risk values.)
Oral data used for OSF and
inhalation data used for
IUR. Oral cancer risk
values based on route-to-
route extrapolation from
inhalation study also
examined.
Uncertainty associated with an OSF derived
from oral exposure data was considered
lower than with an OSF derived by route-to-
route extrapolation when oral data from a
well-conducted study were available (in this
case, a 2-year drinking water study with
complete histopathology). Oral cancer risk
values based on route-to-route extrapolation
from the NTP (1986) inhalation mouse study
were about one order of magnitude lower
than values based on oral exposure study.
Selection of dose metric
(Selection of dose metric could
change the recommended cancer
risk values.)
Use tissue-specific GST-
metabolism dose metric.
Cancer risk estimates based
on alternative (whole-body)
metrics also examined.
Contribution of CYP pathway to cancer risk
unknown, but strong evidence of GST role in
carcinogenesis supports focus on this
pathway. Values based on tissue-specific
GST metabolism recommended based on
evidence of site locality of effects.
Dose-response modeling
(Human risk values could increase
or decrease, depending on fits of
alternative models)
Use multistage dose-
response model to derive a
BMD
The multistage model has biological support
and is the model most consistently used in
EPA cancer assessments.
Low-dose extrapolation
(Human risk values would be
expected to decrease with the
application of nonlinear tumor
responses in low-dose regions of
dose-response curves.)
Use linear extrapolation of
risk in low-dose region
PBPK model incorporates the metabolic shift
and expected nonlinearity (GST dose
attenuation) in the exposure-dose
relationship across exposure levels. Linear
low-dose extrapolation for agents with a
mutagenic mode of action is supported.
327
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-26. Summary of uncertainty in the derivation of cancer risk
values for dichloromethane
Consideration and
Impact On Cancer Risk Value
Decision
Justification and Discussion
Interspecies extrapolation of
dosimetry and risk
(Alternative values for PBPK
model parameters and cross-
species scaling factor could
increase or decrease human
cancer risk values.)
Use PBPK model and
allometric scaling factor for
the primary dose metric
Use of rodent and human PBPK models
reduced uncertainty due to interspecies
differences in toxicokinetics. Examination
of impact of different values for key
parameters in human model, and sensitivity
analysis of rodent PBPK model parameters
identified influential metabolic parameters
for which limited experimental data exist
Sensitive subpopulations
(Differences in CYP and GST
metabolic rates could change
cancer risk values.)
Risk estimates generated for
presumed most sensitive
(GST-T1+/+) genotype; CYP
variability incorporated into
PBPK model
No data are available to determine the range
of human toxicodynamic variability or
sensitivity, including whether children are
more sensitive than adults.
Data selections for derivation oflUR and OSF. The database of animal bioassays
identifies the liver and lung as the most sensitive target organs for dichloromethane-induced
tumor development. These effects demonstrate a dose-response relationship in mice exposed
orally (liver, males only) or by inhalation (liver and lung, males and females). The liver cancer
effects seen in the oral exposure study (Serota et al., 1986b; Hazleton Laboratories, 1983) were
not strong (increasing from approximately 20% in controls to 30% in the highest 3 dose groups),
but rather were characterized as a marginally increased trend for combined hepatocellular
adenomas and carcinomas (trend test/? = 0.058) and by statistically significantly increases (p <
0.05) at dose levels of 125, 185, and 250 mg/kg-day. These data are considered adequate for
dose-response modeling. Although EPA's interpretation of these data differs from that of the
study authors, the reasons for this difference were described in Section 4.2.1.2.2. In addition, as
shown in Table 4-43, the lower incidence of liver tumors induced by the highest doses used in
the oral exposure study (Serota et al., 1986b; Hazleton Laboratories, 1983) compared with the
higher incidence induced by inhalation exposure to 2,000 ppm (NTP, 1986) is consistent with the
predicted lower liver dose of GST metabolites (and hence lower probability of DNA
modification) with oral exposure.
Statistically significant increases in benign mammary gland tumors were observed in one
study of F344 rats exposed by inhalation to 2,000 or 4,000 ppm (Mennear et al., 1988; NTP,
1986), and evidence for a turnorigenie mammary gland response in Sprague-Dawley rats was
limited to increased numbers of benign mammary tumors per animal at levels of 50-500 ppm
(Nitschke et al., 1988a) or 500-3,500 ppm (Burek et al., 1984). A gavage study in female
Sprague-Dawley rats reported an increased incidence of malignant mammary tumors, mainly
adenocarcinomas (8, 6, and 18% in the 0, 100, and 500 mg/kg dose groups, respectively), but the
increase was not statistically significant. Data were not provided to allow an analysis that
328
DRAFT - DO NOT CITE OR QUOTE
-------
accounts for differing mortality rates (Maltoni et al., 1988). The toxicokinetic or mechanistic
events that might lead to mammary gland tumor development in rats are unknown, although
CYP2E1 (El-Rayes et al., 2003; Hellmold et al., 1998) and GST-T1 expression have been
detected in human mammary tissue (Lehmann and Wagner, 2008).
Rare CNS tumors were observed in one study in rats spanning a relatively low range of
exposures (0-500 ppm). These cancers were not seen in two other studies (NTP, 1986; Burek et
al., 1984) in rats, both involving higher doses (1,000-4,000 ppm), or in a similar high-dose study
(NTP, 1986) in mice. The relative rarity of the tumors precludes the use of the low-dose
exposure study in a quantitative dose-response assessment.
The in vivo genotoxicity and mechanistic data in rodents provide a detailed sequence of
steps from generation of reactive metabolites to mutagenic effects, such as DNA-protein cross-
links and DNA strand breaks. Further, the toxicokinetic pathways implicated in production of
the putative carcinogenic metabolites in animals also exist in humans. Thus, there is high
confidence that the dose-response data for liver and lung cancer in mice represents the best data
currently available for derivation of human cancer risks. A more complete understanding of the
carcinogenic potential of dichloromethane would be achieved by addressing data gaps identified
with respect to issues regarding potential risk and mechanisms relating to brain cancer,
mammary tumors, and hematopoetic cancers.
Target organ. The liver and lung tumor incidence from chronic exposure bioassays
provide clear evidence of the carcinogenic potential of dichloromethane exposure. The
bioassays are supported by a substantial database of genotoxicity and mechanistic studies
(summarized in Section 4.5). It is hypothesized that a highly reactive GST metabolite of
dichloromethane, S-(chloromethyl)-glutathione, produces mutations and genetic changes in liver
and lung cells that progress to tumors in these organs. This mode of action is potentially relevant
to other sites, particularly those in which GST-T1 is expressed, such as mammary tissue
(Lehmann and Wagner, 2008) and the brain (Juronen et al., 1996) in humans, and rat olfactory
epithelium (Banger et al., 1994). The extent of GSH conjugation in these other tissues does not
contribute substantially to the overall dosimetry of dichloromethane, but it may be significant in
understanding the sites of action of dichloromethane, and differences in tumor sites observed
between species. The evidence for mammary gland tumors from dichloromethane exposure is
based primarily on observations of benign tumors in rats with inhalation exposure (NTP, 1986).
Derivation of cancer potency values based on these data are presented in Appendix G. The
potential brain cancer risk, suggested by a limited number of these relatively rare tumors in both
animal and human studies, is identified as a data gap that would benefit from additional research.
In addition, the available epidemiologic studies do not provide a definitive evaluation of non-
Hodgkin lymphoma, but the consistent observations of associations seen in three large case-
control studies in Germany (Seidler et al., 2007), Italy (Miligi et al., 2006), and the United States
329 DRAFT - DO NOT CITE OR QUOTE
-------
(Barry et al., 2011; Wang et al. 2009) and the associations seen in the only large study of
multiple myeloma (Gold et al., 2011) provide evidence of an increased risk of specific types of
hematopoietic cancers in humans. The potential contribution of dichloromethane to non-
Hodgkin lymphoma (or specific subtypes thereof) and multiple myeloma, is another key data
gap.
Extrapolation approach. A route-to-route extrapolation from the NTP (1986) inhalation
mouse bioassay was used to develop an OSF for purposes of comparison. In this case, the
uncertainty associated with an OSF derived using oral exposure data from a well-conducted
study (i.e., a 2-year drinking water study with complete histopathology) was considered lower
than with an OSF derived by route-to-route extrapolation. Therefore, although the exposure-
response effect seen in the oral exposure study (Serota et al., 1986b) is not strong, direct
derivation from oral exposure is recommended over route-to-route extrapolation as the basis for
the OSF.
The comparison of the OSF derived from the oral exposure data and from the route-to-
route extrapolation from the inhalation data provides a crude measure of the uncertainty in
recommending a human OSF based on the available rodent bioassay data. The cancer OSF
based on route-to-route extrapolations from liver tumors in mice exposed by inhalation are about
an order of magnitude lower than those based on the liver tumor responses in mice exposed via
drinking water. This difference may be explained, at least partially, by the heterogeneity of
hepatic cell types within the sinusoid, resulting in regio-specific toxicity. Oral exposure may
result in a higher internal exposure of hepatocytes in the periportal region (particularly those
lining the portal vein, through which all gastrointestinal-absorbed dichloromethane passes) than
in the centrilobular region (SRC, 1989). Further, the metabolic capacity of hepatic cells is also
region specific, with higher CYP activity found in the centrilobular region compared to the
periportal region. Thus, liver perfusion via the systemic arterial circulation or portal drainage of
the gastrointestinal tract, through which inhaled dichloromethane would be introduced, may
influence region specific hepatotoxicity, resulting in the route-of-exposure effects on toxicity.
The available PBPK models do not have the capability to predict region specific disposition of
dichloromethane in the liver.
Dose metric. There is considerable data supporting the role of GST-related metabolism
of dichloromethane in carcinogenicity, as described in Sections 4.5.1 and 4.7. Pretreatment of
mice with buthionine sulphoximine, a GSH depletor, caused a decrease to levels seen in controls
in the amount of DNA damage detected immediately after in vivo exposure in liver and lung
tissue (Graves et al., 1995). The results of Olvera-Bello et al. (2010) indicate, however, that
increased sister chromatid exchanges can be seen with in vitro dichloromethane exposure even in
330 DRAFT - DO NOT CITE OR QUOTE
-------
human lymphocyte cells from individuals with low GST activity., Additional research pertaining
to GST-T1 polymporphisms, other metabolic variations, and genotoxicity is warranted.
Dose-response modeling. The multistage model is used because of its biological
relevance and because of the lack of information supporting a biologically-based or other
particular model instead of the multistage model. The multistage model is the most commonly
used model for cancer derivations and its use maintains comparability with existing assessments.
Because of the adequacy of the fit of the multistage model to the data, little modeling
uncertainty would be expected to be introduced by the choice of this model. Application of the
multistage model allowed for estimation of a POD in the lower region of exposure for observable
cancer effects.
For human oral exposure, ingestion is assumed to occur as six discrete boluses during the
course of the day: 25% of the daily dose consumed at 7 am, noon, and 6 pm; 10% at 10 am and 3
pm; and 5% at 10 pm. When exposure occurs as a bolus, the short-term (peak) concentration of
dichloromethane will be higher, leading to a higher degree of CYP saturation and hence a higher
fraction metabolized by GST, as compared to more continuous exposure such as occurs by
inhalation. Thus if actual ingestion is in fewer/larger boluses than those assumed, the cancer risk
will be somewhat under-predicted. On the other hand if ingestion is in more/smaller boluses, the
opposite will occur. However, when ingestion is fairly small, such that the peak concentration is
well below the saturation constant (Km) for CYP, the difference in metabolic outcome will be
negligible. The pattern used here assumes in effect that the amount of food and liquid ingested
is divided neatly into meals and snacks or breaks as indicated and that the concentration of
dichloromethane in the food and beverages ingested is constant. Thus if one meal or drink
happens to include the bulk of that ingested for a day, total ingestion will be more like a single
daily bolus. But to the extent that people sip beverages or ingest foods over longer periods of
time, actual ingestion will be more continual. Given that both of these are likely to occur to
some extent, the population ingestion pattern is expected to be a distribution that includes the one
used for simulation purposes. While it cannot be said that this pattern is an exact average, given
that the differences in saturation at low total exposure levels will be small, it is considered
sufficiently representative of the population and the uncertainty resulting from inexact
knowledge of actual ingestion is unlikely to be significant.
Low-dose extrapolation. The mode of action is a key consideration in determining how
risks should be estimated for low-dose exposure. The in vitro and in vivo genotoxicity data
suggest that mutagenicity is the most plausible mode of action. Because it was concluded that
dichloromethane acts through a mutagenic mode of action, a linear-low-dose extrapolation
approach was used to estimate OSFs and lURs.
331 DRAFT - DO NOT CITE OR QUOTE
-------
While the rate equation for GST metabolism in the PBPK model is first-order, consistent
with metabolism increasing in direct proportion to the concentration of dichloromethane, because
the GST and CYP pathways compete for dichloromethane and the CYP kinetics are nonlinear
(saturable), the interaction of the two pathways within the whole-body system results in a
nonlinear exposure-dose relationship for both pathways. These nonlinearities are demonstrated
for a simulated group of 30-year-old women (population mean kinetics for continuous inhalation
exposure) in Figure 5-16. The upper panel (A) of Figure 5-16 provides a full-scale plot of CYP
liver metabolism (mg/L liver/d) up to 2000 ppm exposure while the lower panel expands the
CYP metabolism curve up to 400 ppm exposure (with CYP metabolism still included, but not far
off the the y-axis with that scale). Because both CYP and GST metabolism are linear at very low
concentrations (below 10-30 ppm), the exposure-response relationship at low exposures is linear
for both pathway metrics, initially increasing from zero dose at zero concentration in direct
proportion to the exposure level until CYP saturation begins. As CYP becomes saturated,
starting around 50 ppm and reaching half saturation at around 200 ppm, a lower fraction of
dichloromethane is eliminated by CYP metabolism. As a lower fraction is metabolized by CYP
the blood concentration increases faster than directly proportional to exposure concentration with
the result that GST metabolism also increases faster than direct proportionality: the upward
curvature seen in the lower panel (B) of Figure 5-16. While GST metabolism remains less than
CYP over the entire exposure range shown here for humans, in mice, where the GST pathway
has relatively higher activity compared to CYP, GST metabolism increases above CYP
metabolism in the range of bioassay exposures. This transition from CYP-dominated (or vastly-
dominated) clearance at low exposures to a higher fraction of GST metabolism at high exposures
has at times been referred to as a "switch," but as shown in Figure 5-16 the transition is smooth
and continuous, and there is some GST metabolism at all exposure levels with the exposure-
response approaching linearity, without a threshold, at low exposures.
One other important note is that in calculating IUR for humans the relationship between
external exposure and internal dose was determined using the PBPK model at a very low level of
exposure (i.e.,1 ug/m3 or 0.00029 ppm), where the relationship is effectively linear:
the difference between the actual exposure-dose curve and a straight line is less than 1%.
However as one goes to higher concentrations the relationship becomes significantly nonlinear,
and hence application of the cancer toxicity values (IUR) will not accurately represent the risk.
Because GST metabolism increases faster than proportional to exposure level with concentration,
in fact the IUR will under-predict risk at those higher exposure levels. Analysis of the PBPK
model versus the low-exposure linear estimate shows that the extent of nonlinearity is less than
20% for oral exposures at low doses and for inhalation exposures at less than 30 ppm (100
ug/m3). The dose used for calculating the internal dose:exposure ratio for oral exposures, 1
mg/kg-d, was above the transition to nonlinear dosimetry, but only to a small extent. For oral
exposures the linear approximation used differed from the full model by less than 30% for
332 DRAFT - DO NOT CITE OR QUOTE
-------
exposures less than 2 mg/kg-d, but at doses below 1 mg/kg-d the error would be in the direction
of an over-prediction of risk (i.e., actual cancer risks may be somewhat lower, but no more than
1.3-fold lower), than indicated by the linear model.
Panel A
§"
"S>
E
E
~o
n)
"5
>
3,600-
3,200-
2,800-
2,400-
2,000-
1,600-
1 200-
800 -
400-
n -
ss^~~~
/^
/ Average Metabolic Rates in Liver vs. Inhalation
/ Concentration for 30-year-old, GST +/+ Women
/
/
/ f GST metabolism "1
/ ^ CYP mctoboliorn J
/ „
200 400 600 600 1,000 1,200 1,400 1,600 1,800 2,000
Exposure concentration (ppm)
40
35
!j 25-
"ra
•£• 20
E
en
1 ^
(0
* 10
5
Average Metabolic Rates in Liver vs. Inhalation
Concentration for 30-year-old, GST +/+ Women
Panel B
GST metabolism
— CYP metabolism
50 100 150 200 250
Exposure concentration (ppm)
300
350
400
The curves represent average results for a simulated population of 1000 women
with the GST-T1 +/+ genotype. A) Relationships scaled to show full range of
CYP metabolism up to 2,000 ppm inhalation exposure. B) Relationships scaled to
show low-dose linearity (below 50 ppm) and curvature (transition) in GST
metabolism (above 50 ppm), as CYP metabolism saturates.
Figure 5-16. PBPK-model-predicted exposure-response relationships for
hepatic CYP and GST metabolism for continuous inhalation exposure to
dichloromethane in 30-year-old GST +/+ women.
333
DRAFT - DO NOT CITE OR QUOTE
-------
Interspecies extrapolation ofdosimetry and risk. Target organ dosimetry for neoplastic
mouse responses and estimation of equivalent internal human doses were accomplished using
PBPK models for dichloromethane in mice and humans. Uncertainty in the ability of the PBPK
models to estimate animal and human internal doses from lifetime bioassay low-level
environmental exposures may affect the confidence in the cancer risk extrapolated from animal
data. Uncertainties in the mouse and human model parameter values were integrated
quantitatively into parameter estimation by utilizing hierarchical Bayesian methods to calibrate
the models at the population level (David et al., 2006; Marino et al., 2006). The use of Monte
Carlo sampling to define human model parameter distributions allowed for derivation of human
distributions ofdosimetry and cancer risk, providing for bounds on the recommended risk
values.
A detailed discussion of PBPK model structure (CYP rate equation) and parameter
uncertainties is provided in Sections 3.5.2 and 3.5.5, respectively. While the structure and
equations used in the existing model have been described in multiple peer-reviewed publications
over the past two decades, there are discrepancies between dichloromethane kinetics observed in
vitro and the model parameters obtained from in vivo data, and the model poorly fits some of the
in vivo data (e.g., fraction of dose exhaled as CO at higher exposure levels in mice). The
discrepancies are significant enough that simply re-estimating model parameters is unlikely to
resolve them, but based on a limited analysis, it appears that an alternative (dual-binding-site)
CYP metabolic equation (Korzekwa et al., 1998) may provide the resolution. At present, the
suggestion of this alternate equation is a hypothesis which should be tested experimentally.
Further, integration of the alternate rate equation into the PBPK modeling and then quantitative
risk assessment will likely require several years of further research and, hence, is beyond the
scope of the current assessment. Since the GST activity in the current model is within a factor of
three of that measured in vitro (when both are evaluated per gram of liver), the impact of that
model uncertainty is also expected to be no more than a factor of three.
Also as detailed in Section 3.5.2, the results of David et al. (2006) for the GST-T1
activity parameter, kfc, for the combined human data set appear to be discrepant with the range
of results for each of the individual data sets. Therefore, sensitivity to the human PBPK
parameter distributions was evaluated by reseating the parameters to the mean values obtained
by David et al. (2006) for a specific data set (DiVincenzo and Kaplan, 1981) for which the GST
activity was intermediate among those obtained across individual data sets. Specifically, the
hepatic GST internal dose for a 1 |ig/m3 inhalation exposure or a 1 mg/kg-day oral exposure was
simulated in the GST-T1+ + population, using the alternate parameters. The upper bounds on
internal dose for both exposure routes increased by just over an order of magnitude and the mean
values increased by approximately 20-fold. Thus, it is possible that revision or refinement of the
PBPK model could have a larger impact on the cancer risk estimates. The ultimate impact will
depend on how revisions effect model predictions for both the animal and the human. If the
334 DRAFT - DO NOT CITE OR QUOTE
-------
predicted GST metabolism per unit exposure increases in both mice and humans by a similar
factor, there will be little impact on the risk estimate. But if the GST activity predicted in the
mouse is decreased by a factor of 3, while that in the human is increased by a factor of 3, for
example, then the net impact would be an increase of ninefold in human risk estimates.
Sensitivity analysis of the mouse PBPK parameters. The mouse and rat PBPK models
were utilized deterministically; i.e., the single-value parameter estimates for the rat PBPK model
were used for rat dosimetry simulations, and the mean parameter estimates from the Bayesian
analysis of Marino et al. (2006) were used for the mouse dosimetry simulations. To assess the
effect of using point estimates of parameter values for calculation of rodent dosimetry, a
sensitivity analysis was performed to identify model parameters most influential on the
predictions of dose metrics used to estimate oral and inhalation cancer risks. As was described
in the RfD and RfC sensitivity analysis calculation, this procedure used a univariate analysis in
which the value of an individual model parameter was perturbed by an amount (A) in the forward
and reverse direction (i.e., an increase and decrease from the nominal value), and the change in
the output variable was determined. Results are for the effects of a perturbation of ±1% from the
nominal value of each parameter on the output values at the end of a minimum of 10,000
simulated hours. This time was chosen to achieve a stable daily value of the dose metric, given
that the simulated bioassay exposures did not include weekend exposures. The exposure
conditions represented the lowest bioassay exposure resulting in significant increases in the
critical effect. For inhalation exposures in mice, the PB, followed closely by the first-order GST-
mediated metabolism rate (kfc), had the greatest impact on the dose metric for liver cancer (mg
dichloromethane metabolized via GST pathway per liter liver per day) (Figure 5-17). For
drinking water exposures in mice, the kfc, followed by the CYP-mediated maximum reaction
velocity (Vmaxc), affected the liver cancer dose metric to the greatest extent (Figure 5-18). For
mice inhaling dichloromethane, the lung cancer dose metric (mg dichloromethane metabolized
via GST pathways per liter lung per day), like the liver cancer metric, was highly affected by the
kfc and the PB (Figure 5-19). However, since GST-mediated lung metabolism is calculated as a
constant fraction of the liver metabolism rate (A2 x kfc), the lung cancer dose metric was most
sensitive to the proportional yield of liver GST-mediated metabolic activity attributed to the
lung. The PB was experimentally determined, lending high confidence to its value. Values for
the three metabolic parameters were determined by computational optimization against data sets
not directly measuring dichloromethane or its metabolites in the target/metabolizing tissues. It is
uncertain how alternative values for these three parameters would affect the estimation of animal
BMDL10 values and, ultimately, the OSFs and lURs.
335 DRAFT - DO NOT CITE OR QUOTE
-------
Inhalation exposure: liver GST
KFC
A2
£ VMAXC..
| PB
5 VSC
re
°- VLC-
VPR
QCC
-0.
r
1
i i i
-
1
' 1 1
1 ' '
i i i i
75 -0.5 -0.25 0 0.25 0.5 0.75
Normalized sensitivity coefficient
Figure 5-17. Sensitivity coefficients for long-term mass GST-mediated
metabolites per liver volume from a long-term average daily inhalation
concentration of 2,000 ppm in mice.
336
DRAFT - DO NOT CITE OR QUOTE
-------
Oral exposure: liver GST
KFC
A2
K^
VMAXC
PB
v&e-
VLC
VPR
QCC
-0
-
1
!
r
i
i
-
I
-
•...
~
J
75 -0.5 -0.25 0 0.25 0.5 0.75
Normalized sensitivity coefficient
1
Figure 5-18. Sensitivity coefficients for long-term mass GST-mediated
metabolites per liver volume from a long-term average daily drinking water
concentration of 500 mg/L in mice.
o>
+j
o>
£
(0
a.
KFC
A2
VMAXC
PB
VSC
VI -P
VPR
QCC
-0.
Inhalation exposure: lung GST
1
1 :
-+-
I
i ;
3-
\
-4-
1 1
II III
75 -0.5 -0.25 0 0.25 0.5 0.75 1
Normalized sensitivity coefficient
Figure 5-19. Sensitivity coefficients for long-term mass GST-mediated
metabolites per lung volume from a long-term average daily inhalation
concentration of 500 ppm in mice.
There is uncertainty as to whether the reactivity of the toxic dichloromethane metabolites
is sufficiently high enough to preclude systemic distribution. Therefore, alternative derivations
of cancer risk values were performed under the assumption that high reactivity leads to complete
clearance from the tissue in which the active metabolite is formed (scaling factor = 1.0). The
difference in scaling factor (7.0 for allometric scaling versus 1.0) results in a sevenfold decrease
in estimated cancer toxicity values. Using a whole-body GST metabolism dose metric, the
337
DRAFT - DO NOT CITE OR QUOTE
-------
resulting OSF and IUR for liver cancer was approximately fivefold lower than when tissue-
specific dose metrics were used (Table 5-16 and Table 5-22); however, the lURs for lung cancer
and for the combined liver and lung cancer risk were higher with the whole-body compared with
the tissue-specific metric (Table 5-22). This difference reflects the lower metabolism that occurs
in human versus mouse lung (relative to total); lung-specific metabolism is lower in humans than
mice, so the predicted risk in the lung is lower when based on that metabolism versus when
whole-body metabolism is used. The mechanistic data support the hypothesis that reactive
metabolites produced in the target tissues do not distribute significantly beyond those tissues and
cause deleterious effects in the metabolizing tissues soon after generation. Thus, there is less
uncertainty in the cancer risk values derived by using a tissue-specific GST metabolism dose
metric compared with those derived using a whole-body GST metabolism dose metric.
Sensitive human populations. Possible sensitive populations include persons with altered
CYP (e.g., obese individuals, alcoholics, diabetics, and the very young) and GST (e.g., GST-T1
homozygous conjugators) metabolic capacity. The PBPK model includes an estimate of the
variability of CYP metabolism (sixfold variation) within the general population but does not
specifically address what could be greater variation in these other groups. However, the known
polymorphisms for GST-T1 expression were integrated into the human model. The distributions
of human IUR values (from which the recommended [i.e., mean] values were taken) show that
the 99th percentiles are approximately seven- and sixfold higher than means for liver and lung
cancer, respectively. For the distribution of OSFs, the 99* percentile is approximately twofold
higher than the mean for liver cancer.
To further characterize the potential sensitivity of specific subpopulations, internal dose
distributions for oral exposure to 1 mg/kg-day or inhalation exposure to 1 mg/m3 were estimated
for 1-year-old children and 70-year-old men and women to compare with the broader population
results used to estimate cancer risks above. Since the recommended cancer risk estimate is based
on the GST-T1+/+ subpopulation, this analysis was also restricted to that subpopulation so that
only the factors of age and gender would differ. The impact of considering other GST-T1 groups
can be seen where risk estimates for the GST-T1+" and entire population mix are given above.
Specification of age- and gender-specific parameters are as described in Appendix B. This
sensitivity analysis is qualitatively similar to that described previously for the noncancer
assessments of dichloromethane, where the variability in human equivalent dose and HEC values
was estimated.
For this analysis, however, consideration of exclusively GST-T1++ individuals will
clearly narrow any estimate of variability. This analysis will also differ from that for noncancer
effects in that the inverse of the former relationship is being considered (i.e., the variation in a
specific internal dose for a fixed exposure is being computed, whereas for the human equivalent
dose and HEC, the variability in exposure levels corresponding to a fixed internal dose are
338 DRAFT - DO NOT CITE OR QUOTE
-------
estimated). The results of this analysis are shown in Figure 5-20 and Table 5-27 for oral
exposures and in Figure 5-20 and Table 5-28 for inhalation exposures.
Internal dose distribution,
GST-T1 ++ population,
oral ingestion
General
1 yo child
70 yo Male
— - -70yo Female
0 0.05 0.1 0.15 0.2 0.25 0.3
Internal dose (mg GST metabolites/L liver/day)
Figure 5-20. Histograms for a liver-specific dose of GST metabolism (mg
GST metabolites per liter liver per day) for the general population (0.5- to
80-year-old males and females), and specific age/gender groups within the
population of GST-T1+/+ genotypes, given a daily oral dose-rate of 1 mg/kg-
day dichloromethane.
Table 5-27. Statistical characteristics of human internal doses for 1 mg/kg-
day oral exposures in specific populations
Population
All agesb
1-yr-old children
70-yr-old men
70-yr-old women
Internal dose (mg/L liver per d)a
Mean
9.43 x ID'2
7.82 x ID'2
9.71 x ID'2
1.01 x ID'1
95th percentile
2.98 x ID'1
2.41 x ID'1
2.99 x ID'1
5.33 x ID'1
99th percentile
5.43 x ID'1
4.00 x 1Q-1
5.51 x ID'1
9.84 x ID'1
aLiver-specific GST-T1 metabolism in GST-T1++ individuals exposed orally to 1 mg/kg-d dichloromethane.
b0.5- to 80-yr-old males and females.
339
DRAFT - DO NOT CITE OR QUOTE
-------
0.25
Internal dose distribution,
GST-T1 ++ population,
1 mg/m3 inhalation
All ages, M&F
— - - 1 yo Child
70 yo Male
70 yo Female
10 20 30
ng GST metabolites/L liver/day
40
Figure 5-21. Histograms for liver-specific dose of GST metabolism (mg GST
metabolites per liter liver per day) for the general population (0.5- to
80-year-old males and females), and specific age/gender groups within the
population of GST-T1+/+ genotypes, given a continuous inhalation exposure
to 1 mg/m3 dichloromethane.
Table 5-28. Statistical characteristics of human internal doses for 1 mg/m3
inhalation exposures in specific subpopulations
Population
All agesb
1-yr-old children
70-yr-old men
70-yr-old women
Internal dose (mg/L liver per d)a
Mean
6.61 x 1(T6
1.65 x 1(T5
5.09 x 1(T6
4.14 x 1(T6
95th percentile
2.21 x 10'5
5.11 x 10'5
1.68 x 10'5
1.37 x 10'5
99th percentile
4.47 x 10'5
9.04 x 10'5
3.12 x 10'5
2.56 x 10'5
aLiver-specific GST-T1 metabolism in GST-T1+/+ individuals exposed continuously by inhalation to 1 mg/m3
dichloromethane.
b0.5- to 80-yr-old males and females.
For the oral exposure analysis, the distribution of internal doses shows little variation
among the different age/gender groups (Figure 5-21, Table 5-27). The cancer analysis is based
on a very low internal dose where little enzymatic saturation is expected to occur, allowing for
efficient first-pass metabolism, which is independent of differences in respiration; differences
will be more significant at the higher doses analyzed for the noncancer human equivalent applied
340
DRAFT - DO NOT CITE OR QUOTE
-------
dose. Thus, the consideration of only GST metabolism and the narrower range of metabolic rate
for that pathway in the +/+ population at low oral exposure rates results in minimal age/gender
sensitivity differences (the 7-year-old female is only 5% more sensitive from pharmacokinetic
factors than the general population).
For inhalation, an internal liver GST dose (mean value) about 2.5 times higher in the
child than the general population is predicted due to the higher inhalation rate. The results for
the liver GST dose for inhalation (Figure 5-21 and Table 5-28) indicate that the 70-year-old male
and female populations are only slightly shifted from the general population, while the
population for the 1-year-old child is a distinct upper tail of the general distribution.
341 DRAFT - DO NOT CITE OR QUOTE
-------
6. MAJOR CONCLUSIONS IN THE CHARACTERIZATION OF HAZARD AND
DOSE RESPONSE
6.1. HUMAN HAZARD POTENTIAL
Dichloromethane (CASRN 75-09-2), also known as methylene chloride, is a colorless
liquid with a penetrating, ether-like odor. It is produced by the direct reaction of methane with
chlorine at either high temperatures or low temperatures under catalytic or photolytic conditions.
The principal uses for dichloromethane have been in paint strippers and removers, as a propellant
in aerosols, in the manufacture of drugs, pharmaceuticals, film coatings, electronics, and
polyurethane foam, and as a metal-cleaning solvent.
Dichloromethane is rapidly absorbed through both oral administration and inhalation
exposure with a near steady-state saturation occurring with inhalation. Results from studies of
animals show that following absorption, dichloromethane is rapidly distributed throughout the
body and has been detected in all tissues that have been evaluated. Metabolism of
dichloromethane involves two primary pathways. Dichloromethane is metabolized to CO in a
CYP-dependent oxidative pathway (CYP2E1) that is predominant at low exposure levels. The
other major pathway for dichloromethane metabolism involves the conjugation of
dichloromethane to GSH, catalyzed by GST (GST-T1). This results in the formation of a GSH
conjugate that is eventually metabolized to CC>2. The conjugation of dichloromethane to GSH
results in the formation of two reactive intermediates that have been hypothesized to be involved
in dichloromethane carcinogenicity, S-(chloromethyl)glutathione and formaldehyde. Formation
of formaldehyde leads to several covalent modifications of cellular macromolecules, including
DNA-protein cross-links (Casanova et al., 1996) and RNA-formaldehyde adducts (Casanova et
al., 1997). Evidence is also available that S-(chloromethyl)glutathione can result in both DNA
SSBs and DNA mutations, presumably through DNA alkylation (Green, 1997; Graves and
Green, 1996; Graves et al., 1996, 1994a; Hashmi et al., 1994). However, DNA reaction products
(e.g., DNA adducts) produced by S-(chloromethyl)glutathione have not been found in vivo,
possibly due to potential instability of these compounds or due to the limited doses used in
attempts to detect them (Watanabe et al., 2007; Hashmi et al., 1994). DNA adducts, however,
have been observed in in vitro studies in which calf thymus DNA was incubated with
dichloromethane and GSH or was incubated with S-(l-acetoxymethyl)glutathione, a compound
structurally similar to S-(chloromethyl)glutathione (Marsch et al., 2004; Kayser and Vuilleumier,
2001).
Information on noncancer effects in humans exposed orally to dichloromethane are
restricted to case reports of neurological impairment (general CNS depression), liver and kidney
effects (as severe as organ failure), and gastrointestinal irritation in individuals who ingested
amounts ranging from about 25 to 300 mL (Chang et al., 1999; Hughes and Tracey, 1993). The
-------
animal toxicity database identifies hepatic effects (hepatic vacuolation, liver foci) as the critical
dose-dependent noncancer endpoint associated with oral exposure to dichloromethane. The most
frequently observed liver effect was hepatocyte vacuolation, seen with drinking water exposure
(90 days) in F344 rats at>166 mg/kg-day and B6C3Fi mice at 586 mg/kg-day (Kirschman et al.,
1986) and with gavage exposure (14 days) in CD-I mice at 333 mg/kg-day (Condie et al., 1983).
Hepatocyte degeneration or necrosis was observed in female F344 rats exposed in drinking water
for 90 days to 1,469 mg/kg-day (Kirschman et al., 1986) and in female F344 rats exposed by
gavage for 14 days to 337 mg/kg-day (Berman et al., 1995). In the chronic-duration (104-week)
study, liver effects (areas of foci alteration) were observed in F344 rats exposed to drinking
water doses between 50 and 250 mg/kg-day (Serota et al., 1986a). In the reproductive oral
administration studies, no significant effect on reproductive function or parameters was observed
in rats up to 225 mg/kg-day (General Electric Company, 1976) or in mice up to 500 mg/kg-day
(Raje et al., 1988). The NOAEL and LOAEL for altered neurological functions in female F344
rats were 101 and 337 mg/kg-day (as reported by Moser et al., 1995).
Acute inhalation exposure of humans to dichloromethane has been associated with
cardiovascular impairments due to decreased oxygen availability from COHb formation and
neurological impairment from interaction of dichloromethane with nervous system membranes
(Bos et al., 2006; ACGffl, 2001; ATSDR, 2000; Cherry et al., 1983; Putz et al., 1979; Gamberale
et al., 1975; Winneke, 1974). Relatively little is known about the long-term neurological effects
of chronic exposures, although there are studies that provide some evidence of an increased
prevalence of neurological symptoms among workers with average exposures of 75-100 ppm
(Cherry et al., 1981) and long-term effects on some neurological measures (i.e., possible
detriments in attention and reaction time in complex tasks) in retired workers whose past
exposures were in the 100-200 ppm range (Lash et al., 1991). These studies are limited by the
relatively small sample sizes and low power for detecting statistically significant results for these
endpoints.
Following repeated inhalation to dichloromethane, the liver is the most sensitive target
for noncancer toxicity in rats and mice. Lifetime exposure was associated with hepatocyte
vacuolation and necrosis in F344 rats exposed to 1,000 ppm 6 hours/day (Mennear et al., 1988;
NTP, 1986), hepatocyte vacuolation in Sprague-Dawley rats exposed to 500 ppm 6 hours/day
(Nitschke et al., 1988a; Burek et al., 1984), and hepatocyte degeneration in B6C3Fi mice
exposed to 2,000 ppm 6 hours/day (lower concentrations were not tested in mice) (Mennear et
al., 1988; NTP, 1986). Other effects observed include renal tubular degenerations in F344 rats
and B6C3Fi mice at 2,000 ppm, testicular atrophy in B6C3Fi mice at 4,000 ppm, and ovarian
atrophy in B6C3Fi mice at 2,000 ppm.
Other studies with inhalation exposure to dichloromethane revealed no significant effects
on reproductive performance in rats (up to 1,500 ppm) (Nitschke et al., 1988b). This study is
limited in its ability to fully evaluate reproductive and developmental toxicity, however, since
-------
exposure was not continued through the gestation and nursing periods. Some evidence of a
decrease in fertility index was seen in male mice exposed to 150 and 200 ppm (Raje et al., 1988),
and no adverse effects on fetal development of mice or rats exposed up to 1,250 ppm were seen
by Schwetz et al. (1975). Decreases in fetal BW and changes in behavioral habituation were
observed in offspring of Long-Evans rats exposed to 4,500 ppm during the gestational period
(Bornschein et al., 1980; Hardin and Manson, 1980). Exposure-related noncancer effects on the
lungs consisted of foreign-body pneumonia in rats exposed to 8,400 ppm 6 hours/day for
13 weeks (NTP, 1986), Clara cell vacuolation in mice exposed to 4,000 ppm 6 hours/day for
13 weeks (Foster et al., 1992), and pulmonary congestion in guinea pigs exposed to 5,000 ppm
7 hours/day for 6 months (Heppel et al., 1944). Several neurological mediated parameters
including decreased activity (Kjellstrand et al., 1985; Weinstein et al., 1972; Heppel and Neal,
1944), impairment of learning and memory (Alexeef and Kilgore, 1983), and changes in
responses to sensory stimuli (Rebert et al., 1989) are reported from acute and short-term
dichloromethane exposure. Evidence of a localized immunosuppressive effect in the lung
resulting from inhalation dichloromethane exposure was seen in an acute exposure (3 hours,
100 ppm) study in CD-I mice (Aranyi et al., 1986).
Numerous in vitro studies have demonstrated mutagenic and genotoxic effects associated
with dichloromethane exposure. For example, bacterial assays, yeast, and fungi provide
evidence that the mutagenic action of dichloromethane in bacterial systems is enhanced by
metabolic activation (e.g., Dillon et al., 1992; Jongen et al., 1982; Gocke et al., 1981). Positive
results from assays of DNA damage with in vitro mammalian systems provide support that
dichloromethane genotoxicity is linked to metabolism by GST enzymes (Graves et al., 1996,
1995, 1994b). Consistent evidence for several genotoxic endpoints in target tissues (liver and
lung) in mice following in vivo exposure to dichloromethane provides supporting evidence that
GST-pathway metabolites are key actors in the mutagenic and carcinogenic mode of action for
dichloromethane. Pretreatment of mice with buthionine sulphoximine, a GSH depletor, caused a
decrease to levels seen in controls in the amount of DNA damage detected immediately after in
vivo exposure in liver and lung tissue, indicating GSH involvement in the genotoxic process
(Graves et al., 1995). DNA damage (detected by the comet assay) was also reported in liver and
lung tissues from male CD-I mice sacrificed 24 hours after administration of a single oral dose
of 1,720 mg/kg of dichloromethane (Sasaki et al., 1998). In this study, DNA damage in lung and
liver was not detected 3 hours after dose administration, and no DNA damage occurred at either
time point in several other tissues in which a carcinogenic response was not seen in chronic
animal cancer bioassays (e.g., stomach, kidney, bone marrow). The weight of evidence from
these studies suggests that dichloromethane is carcinogenic by a mutagenic mode of action.
Dichloromethane is "likely to be carcinogenic in humans" under the Guidelines for
Carcinogen Risk Assessment (U.S. EPA, 2005a). Results from 2-year bioassays provide
adequate evidence of the carcinogenicity of dichloromethane in mice and rats exposed by
-------
inhalation, as well as adequate data to describe dose-response relationships. Oral exposure to
dichloromethane produced statistically significant increases in hepatocellular adenomas and
carcinomas in male B6C3Fi mice (Serota et al., 1986b; Hazleton Laboratories, 1983). Inhalation
exposure to concentrations of 2,000 or 4,000 ppm dichloromethane produced increased
incidences of lung and liver tumors in B6C3Fi mice (Maronpot et al., 1995; Foley et al., 1993;
Kari et al., 1993; Mennear et al., 1988; NTP, 1986). Significantly increased incidences of benign
mammary tumors (adenomas or fibroadenomas) were observed in male and female F344/N rats
exposed by inhalation to 2,000 or 4,000 ppm (Mennear et al., 1988; NTP, 1986). A statistically
significant increased incidence of brain or CNS tumors has not been observed in any of the
animal cancer bioassays, but a 2-year study using relatively low exposure levels (0, 50, 200, and
500 ppm) in Sprague-Dawley rats observed a total of six astrocytoma or glioma (mixed glial
cell) tumors in the exposed groups (Nitschke et al., 1988a). These tumors are exceedingly rare in
rats, and there are few examples of statistically significant trends in animal bioassays (Sills et al.,
1999). Studies in humans also provide some evidence for an association between occupational
exposure to dichloromethane and increased risk for some specific cancers, including brain cancer
(Hearne and Pifer, 1999; Tomenson et al., 1997; Heineman et al., 1994), liver cancer (Lanes et
al., 1993, 1990), non-Hodgkin lymphoma (Barry et al., 2011; Wang et al., 2009; Seidler et al.,
2007; Miligi et al., 2006), and multiple myeloma (Gold et al., 2011).
6.2. DOSE RESPONSE
6.2.1. OralRfD
The available oral toxicity data for animals identify hepatic effects (hepatic vacuolation,
liver foci) as the most sensitive noncancer endpoint associated with chronic oral exposure to
dichloromethane. The 104-week drinking-water study in F344 rats (Serota et al., 1986a) was
selected as the principal study for RfD derivation because the study provided a sensitive endpoint
(liver foci) and used lower doses in comparison to other chronic oral administration studies. In
this study, four doses (6, 52, 125, and 235 mg/kg-day in males; 6, 58, 136, and 263 mg/kg-day in
females) were used. A NOAEL of 6 mg/kg-day in males and females and a LOAEL of 52
(male) and 58 (female) mg/kg-day for alterations of liver foci was identified.
An RfD of 6 x 10"3 mg/kg-day is recommended for use in humans. The RfD derivation
process involved first fitting all available dichotomous models in BMDS version 2.0 to the
incidence data for male and female rats; the male data were used because a greater sensitivity
was seen in males compared with females in this study. A dose metric of average daily mass of
dichloromethane metabolized via the CYP pathway per unit volume of liver was derived from a
EPA-modified rat PBPK model (see Appendix C). This metric was chosen because there are no
data to support the role of a specific metabolite in the development of the noncancer liver lesions
seen in oral and inhalation exposure studies and the CYP-metabolism dose metric was
determined to be most consistent with the data. Then, the BMDLio for liver lesions was derived
-------
based on the best fitting model (in terms of the value of the AIC and examination of model fit
and residuals). Because the metric is a rate of metabolism rather than the concentration of
putative toxic metabolites and the clearance of these metabolites may be slower per volume
tissue in the human compared with the rat, this rodent internal dose metric for noncancer effects
was adjusted by dividing by a pharmacokinetic allometric BW°75 scaling factor (operationalized
as [BWhuman/BWrat]0'25 ~ 4.09) to obtain a human equivalent internal BMDLio. This BMDLio
was then converted to the human equivalent dose by using a human PBPK model (adapted from
David et al., 2006; see Appendix B) that utilizes Monte Carlo sampling techniques to provide a
distribution of human equivalent doses. The first percentile of the distribution of human
equivalent doses (0.189 mg/kg-day) was chosen to include the most sensitive population while
staying within bounds of what is considered computationally stable. The first percentile human
equivalent dose was used as a POD and was divided by a composite UF of 30 (3 [10°5] to
account for uncertainty about interspecies toxicodynamic equivalence, 3 [1005] to account for
uncertainty about toxicodynamic variability in humans, and 3 [10°5] for database deficiencies) to
arrive at an RfD of 6 x 10"3 mg/kg-day.
Use of the mean value (3.50 x 10"1 mg/kg-day) of the human equivalent dose distribution
instead of the 1st percentile, with an additional UF of 3 (10°5) to account for human toxicokinetic
variability, would yield a candidate RfD of 4 x 10"3, which is relatively similar to the
recommended RfD of 6 x 10"3.
Confidence in the principal study, Serota et al. (1986a), is high. The 2-year drinking
water study in rats is a well-conducted, peer-reviewed study that used four dose groups plus a
control. Confidence in the oral database is medium-high. The oral database includes a 2-year
drinking water study in rats (Serota et al., 1986a) and mice (Serota et al., 1986b) as well as a
supporting subchronic exposure study (Kirschman et al., 1986) that reports similar liver effects
to those observed in the chronic oral exposure studies. The toxicity of orally-administered
dichloromethane has also been investigated in an oral administration immunotoxicity study
(Warbrick et al., 2003), a one-generation oral reproductive toxicity study (General Electric
Company, 1976), and an orally dosed developmental toxicity study (Narotsky and Kavlock,
1995). Several studies have also evaluated neurotoxicity associated with oral exposure to
dichloromethane. The oral database lacks a two-generation reproductive study and a
developmental neurotoxicity study; neurodevelopmental outcomes are relevant endpoints given
that dichloromethane is capable of crossing the placental barrier and entering fetal circulation
(Withey and Karpinski, 1985; Anders and Sunram, 1982) and has neurotoxic effects. Overall
confidence in the RfD is high.
6.2.2. Inhalation RfC
The liver is the most sensitive target for noncancer toxicity in rats and mice following
repeated inhalation exposure to dichloromethane. Liver lesions (specifically, hepatic
-------
vacuolation) in rats are the critical noncancer effect from chronic dichloromethane inhalation in
animals. Inhalation bioassays with Sprague-Dawley rats identified the lowest inhalation LOAEL
for liver lesions in the database: 500 ppm (6 hours/day, 5 days/week for 2 years) (Nitschke et al.,
1988a; Burek et al., 1984). Nitscke et al. (1988a) identified a NOAEL of 200 ppm for
hepatocyte vacuolation in female rats. Because the Nitschke et al. (1988a) study more
adequately covers the range spanning the BMR compared with the study by Burek et al. (1984),
the former study was selected as the principal study for derivation of a chronic inhalation RfC.
An RfC of 0.6 mg/m3 is derived based on the observed critical effect in the principal
study. As was described above for the RfD, the RfC derivation process was based on a dose
metric of average daily mass of dichloromethane metabolized via the CYP pathway per unit
volume of liver. This metric was derived from a EPA-modified rat PBPK model (see Appendix
C). Then, the BMDLio risk for liver lesions was derived based on the best fitting model in terms
of the value of the AIC and examination of model fit and residuals. Because the metric is a rate
of metabolism rather than the concentration of putative toxic metabolites and the clearance of
these metabolites may be slower per volume tissue in the human compared with the rat, this
rodent internal dose metric for noncancer effects was adjusted by dividing by a pharmacokinetic
allometric BW°'75 scaling factor (operationalized as [BWhUman/BWrat]0'25 ~ 4.09) to obtain a
human-equivalent internal BMDLio. This BMDLio was then converted to the HEC by using a
human PBPK model (adapted from David et al., 2006; see Appendix B) that utilizes Monte Carlo
sampling techniques to provide a distribution of HECs.
The first percentile HEC (17.2 mg/m3) was used as a POD. This percentile was chosen
because it included the most sensitive population while staying within bounds of what is
considered computationally stable. This POD was divided by a composite UF of 30 (3 [10°5] to
account for uncertainty about interspecies toxicodynamic equivalence, 3 [1005] to account for
uncertainty about toxicodynamic variability in humans, and 3 [10°5] for database deficiencies) to
arrive at an RfC of 0.6 mg/m3.
Use of the mean value (48.5 mg/m3) of the HEC distribution instead of the 1st percentile
with an additional UF of 3 (10°5) to account for human toxicokinetic variability would yield a
candidate RfC of 0.5, similar to the recommended value of 0.6 mg/m3. In addition, two
comparison values derived from occupational studies produced values of 3.5 mg/m3 (Cherry et
al., 1983) and 0.55 mg/m3 (Lash et al., 1991). The animal-derived candidate RfC is preferable to
the human-derived candidate RfC because of the uncertainties about the exposure durations for
the workers in the Cherry et al. (1983) study and uncertainties regarding the exposures and effect
sizes in Lash et al. (1991), and because, in comparison with the value based on Cherry et al.
(1983), the RfC based on the rat data is more sensitive.
Confidence in the principal study, Nitschke et al. (1988a), is high. The 2-year inhalation
study in mice is a well-conducted, peer-reviewed study that used three concentration groups plus
a control. Confidence in the inhalation database is medium. The inhalation database includes
-------
several well-conducted chronic inhalation studies that consistently identified the liver as the most
sensitive noncancer target organ in rats (Nitschke et al., 1988a; NTP, 1986; Burek et al., 1984).
A two-generation reproductive toxicity study (Nitschke et al., 1988b), developmental studies at
relatively high exposures (>1,250 ppm), several neurotoxicity studies, and an immunotoxicity
study have been conducted in animals following inhalational exposures to dichloromethane.
However, the two-generation study is limited in its ability to fully evaluate reproductive and
developmental toxicity, since exposure was not continued through the gestation and nursing
periods. The results from the single dose developmental toxicity study in rats (Bornschein et al.,
1980; Hardin and Manson, 1980), the placental transfer of dichloromethane, and the relatively
high activity of CYP2E1 in the brain compared to the liver of the developing human fetus
(Hines, 2007; Johnsrud et al., 2003; Brzezinski et al., 1999), raise uncertainty regarding possible
neurodevelopmental toxicity from gestational exposure to inhaled dichloromethane. An acute, 3-
hour exposure to 100 ppm dichloromethane demonstrated evidence of immunosuppression in
CD-I mice (Aranyi et al., 1986). This study used a functional immune assay that is relevant to
humans (i.e., increased risk of Streptococcal pneumonia-related mortality and decreased
clearance of Klebsiella bacteria). Chronic and/or repeated exposure studies evaluating functional
immunity are not available and represent a data gap. The inhalation database lacks adequate
developmental neurotoxicity and immunotoxicity studies at chronic low exposures. Overall
confidence in the RfC is medium.
6.2.3. Uncertainties in RfD and RfC Values
One data uncertainty identified is the potential for neurodevelopmental effects. Animal
bioassays have not identified gross or microscopic effects on neural tissues from long-term
exposures or single (Schwetz et al., 1975) or multigenerational (Nitschke et al., 1988b)
developmental toxicity studies. However, behavioral changes were observed in pups born to rats
exposed to high levels (4,500 ppm) of dichloromethane (Bornschein et al., 1980; Hardin and
Manson, 1980); 4,500 ppm was the only dose used in this study. Thus, uncertainty exists as to
the development of neurological effects from lower gestational exposures in animals or in
humans. Immunotoxicity data revealed an additional area of data uncertainty specifically with
respect to inhalation exposure. Data from Aranyi et al. (1986) demonstrated evidence of
immunosuppression following a single 100 ppm dichloromethane exposure for 3 hours in
CD-I mice. The weight of evidence for nonneoplastic effects in humans and animals suggests
that the development of liver lesions is the most sensitive effect, with a UF applied because of
deficiencies in the reproductive and developmental studies for the RfD and, for the RfC, the
additional uncertainty regarding immune system toxicity (specifically, a portal-of-entry immune
suppression effect) at low exposures.
The extrapolation of internal dichloromethane dosimetry from rat liver responses to
human risk was accomplished by using PBPK models for dichloromethane in rats and humans.
-------
Uncertainties in rat and human dosimetry used for RfD and RfC derivation can arise from
uncertainties in the PBPK models to accurately simulate the toxicokinetics of dichloromethane
for animals under bioassay conditions and humans experiencing relatively low, chronic
environmental exposures. Specific uncertainties identified with the PBPK models used here are
described in detail in Sections 3.5.2 and 3.5.5. In brief, there is a structural uncertainty in that
the equation used to describe the CYP2E1-mediated metabolism that could be the source of
discrepancies between the model and some of the data (both in vitro and in vivo), although the
impact of that uncertainty does not appear to be large. There is also a parametric uncertainty in
the GST-T1 parameter, kfc, evident from an inconsistency in the values obtained by David et al.
(2006). The impact of this uncertainty was evaluated by re-estimating human dosimetry with the
mean values for the fitted metabolic parameter reset to match those obtained by David et al.
(2006) from analysis of one data set (DiVincenzo and Kaplan, 1981) rather than the combined
data. The impact of this change on the RfC and RfD was modest (HEC and human equivalent
dose values increased by 25 and 12%, respectively), but cancer risk estimates were increased by
10 to 20-fold.
Given the level of uncertainty identified by this sensitivity analysis, EPA obtained the
MCSim model code and data sets from the authors and re-ran the Markov Chain Monte Carlo
analysis in order to validate the results of David et al. (2006). As was done by David et al.
(2006), EPA ran the chains for 50,000 iterations, saved the output from every 5th iteration (i.e., a
total of 10,000 entries), then analyzed the final 4,000 of these entries. The resulting population-
parameter mean values and CVs were virtually identical to those reported for the combined data
set by David et al. (2006), providing assurance that the computational result was reproducible.
When the output was analyzed by current methods for convergence of the Markov Chain,
however, not all of those measures were satisfied. Visual inspection of plots of the chains did
not reveal any observable trend towards higher or lower values for any of the parameters; i.e., it
appeared that continuing the chains would not significantly alter the estimated population mean
values. There was a high degree of auto-correlation in the chains, however, indicating that the
statistical procedure had not yet obtained a good measure of the covariance among the
parameters.
Autocorrelation in the Markov Chains used to estimate the population parameters
indicates that the assumed degree of independence among the parameters is over-predicted; i.e.,
some sets of population means for all parameters predicted as "likely" given the current results
are probably not likely. If some combinations of parameters are less likely than other
combinations (because the combination does not reflect the true correlation), and the current
estimate treats those combinations as equally likely, then the level of uncertainty that is reflected
in the width of the predicted confidence bounds (distribution percentiles) will be over-estimated.
If the chains are run longer, to reach convergence, the correlation among parameters should be
better identified and the resulting prediction uncertainty (e.g., difference between the predicted
-------
mean and 95th percentile for a dose metric) should be decreased. In short, these results indicate a
lack of convergence in the chains that leads to an over-estimation of parameter uncertainty, such
that the first percentiles of human equivalent concentrations or doses are lower than would be
predicted after full convergence was obtained. Hence, these results likely lead to values of the
RfC and RfD that are more sensitive than would be obtained if the chains are continued to
convergence. Since population mean dose estimates (for the GST-T1+/+ sub-population) were
used to determine cancer risks, the direction in which risk estimates would change with
convergence cannot be predicted for that endpoint. As indicated by the sensitivity analysis,
estimated risks are sensitive to possible changes in the population mean values. But given the
variance in the current estimates of those means, the estimate is not expected to change by more
than a factor of 3 after full convergence. Accordingly, EPA chose to continue using those values
while providing an analysis of the uncertainty.
The dose metric used in the models is the rate of metabolism to a putative toxic
metabolite rather than the concentration (average or area under the concentration curve of the
metabolite), so the model specifically fails to account for rodent-human differences in clearance
or removal of the toxic metabolite. A scaling factor based on BW ratios was used to account for
this difference.
Uncertainties in the human population model parameters and variability in that
population were quantitatively accounted for by utilizing hierarchical Bayesian calibration
methods during model development (David et al., 2006), as modified here by EPA. The rat
model was modified, recalibrated, and utilized in a deterministic manner (Appendix C). Data
were not available to perform a hierarchical Bayesian calibration in the rat, but uncertainties in
the rat model predictions were assessed qualitatively. For both oral and inhalation exposures, the
liver volume, followed closely by the volume of slowly perfused tissues, had the greatest impact
on the internal dose of mg dichloromethane metabolized via CYP pathway per liter tissue per
day. This was due to the fact that the dose metric is a tissue-specific measure, the majority of
CYP metabolism is attributed to the liver, and changes in liver volume have a greater impact on
the total CYP metabolism than either of the individual Vmax values. There is high confidence in
the values used for volume of liver and slowly perfused tissues in the rat, as these are well
studied (Brown et al., 1997). Therefore, except as described in the preceding paragraph, the
uncertainties associated with use of the rat PBPK model should not markedly affect the values of
the RfD and RfC.
An additional uncertainty inherent in this process, however, is the lack of knowledge
concerning the most relevant dose metric (e.g., a specific metabolite) within the context of the
development of the noncancer liver effects. This basic research question represents a data gap,
and this uncertainty is not addressed quantitatively or qualitatively in the assessment.
The effect of dichloromethane on human populations that are sensitive due to
pharmacokinetic differences was addressed quantitatively by using a human probabilistic PBPK
-------
model to generate distributions of human exposures likely to occur given a specified internal
BMDLio. The model and resulting distributions take into account the known differences in
human physiology and metabolic capability with regard to dichloromethane dosimetry. The first
percentile values of the distributions of human equivalent doses (Table 5-3) and HECs
(Table 5-7) served as points of departure for candidate RfDs and RfCs, respectively, to protect
toxicokinetically sensitive individuals. No data are available regarding toxicodynamic
differences within a human population. Therefore, a UF of 3 for possible differences in human
toxicodynamic responses is intended to be protective for sensitive individuals.
6.2.4. Oral Cancer Slope Factor
The recommended cancer OSF for dichloromethane is 2 x 10~3 (mg/kg-day)"1, which is
based on liver tumor responses in male B6C3Fi mice exposed to dichloromethane in drinking
water for 2 years (Serota et al., 1986b; Hazleton Laboratories, 1983). This value was derived by
using a tissue-specific GST metabolism dose metric with allometric scaling to account for
uncertainty regarding the reactivity and clearance of the metabolite(s) involved in the
carcinogenic response.
There was only one adequate oral exposure cancer bioassay evaluating the carcinogenic
potential of orally administered dichloromethane in F344 rats and B6C3Fi mice (Serota et al.,
1986a, b; Hazleton Laboratories, 1983). Significant increases in incidence of liver adenomas and
carcinomas were observed in male but not female B6C3Fi mice (female data were not presented
in the summary reports) (Serota et al., 1986b; Hazleton Laboratories, 1983). The study authors
concluded that in the male bioassay, there was no dose-related trend and that there were no
significant differences comparing the individual dose groups with the combined control group,
and that the observed incidences were "within the normal fluctuation of this type of tumor
incidence." Although Serota et al. (1986b) state that a two-tailed significance level ofp = 0.05
was used for all tests, this does not appear to correctly represent the statistic used by Serota et al.
Each of the/>-values for the comparison of the 125, 185, and 250 mg/kg-day dose groups with
the controls wasp < 0.05. (Theses-values were found in the full report of this study, see
Hazleton Laboratories, 1983, but were not included in the Serota et al., 1986b publication).
Hazleton Laboratories (1983) indicated that a correction factor for multiple comparisons was
used specifically for the liver cancer data, reducing the nominal /7-value from 0.05 to 0.0125;
none of these individual group comparisons are statistically significant when a/?-value of 0.0125
is used. EPA did not consider the use of a multiple comparisons correction factor for the
evaluation of the liver tumor data (a primary a priori hypothesis) to be warranted. Thus, based
on the Hazleton Laboratories (1983) statistical analysis, EPA concluded that dichloromethane
induced a carcinogenic response in male B6C3Fi mice as evidenced by a marginally increased
trend test (p = 0.058) for combined hepatocellular adenomas and carcinomas, and by small but
-------
statistically significant (p < 0.05) increases in hepatocellular adenomas and carcinomas at dose
levels of 125 (p = 0.023), 185 (p = 0.019), and 250 mg/kg-day (p = 0.036).
With respect to the issue of the comparison to historical controls, the incidence in the
control groups (19%) was almost identical to the mean seen in the historical controls from this
laboratory (17.8% based on 354 male B6C3Fi mice), so there is no indication that the observed
trend is being driven by an artificially low rate in controls, no indication that the experimental
conditions resulted in a systematic increase in the incidence of hepatocellular adenomas and
carcinomas, and it is unlikely that the pattern of incidence rates observed in this study (increased
incidence in all four dose groups, with three of these increases significant at ap-va\ue < 0.05)
reflect normal fluctuations in the incidence of these tumors. In F344 rats (Serota et al., 1986a),
no increased incidence of liver tumors was seen in male rats, and the pattern in female rats was
characterized by a jagged stepped pattern of increasing incidence of hepatocellular carcinoma or
neoplastic nodules. However, the potential malignant characterization of the nodules was not
described, and the data for hepatocellular carcinomas are much more limited. The derivation of
the oral cancer slope factor is based on the male mice data because of their greater sensitivity to
liver cancer compared with female mice and to male and female rats.
A modified mouse PBPK model of Marino et al. (2006) was used to approximate the
internal dose of daily dichloromethane (mg) metabolized via the GST pathway per unit volume
of liver from the daily oral administered doses. This approach was taken based on evidence that
GST-pathway metabolites produced from dichloromethane are primarily responsible for
dichloromethane carcinogenicity in mouse liver. The multistage dose-response model (BMDS
version 2.0) was used to fit the mouse liver tumor incidence and PBPK model-derived internal
dose data and to derive a mouse internal BMD and BMDLio. Because the metric is a rate of
metabolism rather than the concentration of putative toxic metabolites and data pertaining to the
reactivity or clearance rate of the relevant metabolite(s) are lacking, the human BMDLio was
derived by multiplying the mouse BMDLio by a BW° 75 allometric scaling factor
(operationalized as [BWhuman/BWmouse]0'25 ~ 7) to account for the potential slower clearance per
volume tissue in the human compared with the mouse. Linear extrapolation from the internal
human BMDLio (0.1/BMDLio) was used to derive oral risk factors for liver tumors based on
tumor responses in male mice. The linear low-dose extrapolation approach for agents with a
mutagenic mode of action was selected because GST-metabolism of dichloromethane is
expected to occur at and below exposures producing the mouse BMDLio, even though CYP2E1
metabolism is expected to be unsaturated and to represent the predominant metabolic pathway in
the liver. Currently, there are no data from chronic oral cancer bioassays in mice providing
support for a nonlinear dose-response relationship.
Probability distributions of human oral cancer slope factors were derived by using a
human PBPK model (adapted from David et al. [2006]; see Appendix B). The cancer reference
values (OSF and IUR) were derived for a sensitive population: a population composed entirely
-------
of carriers of the GST-T1+/+ homozygous genotype (that is, the group that would be expected to
be most sensitive to the carcinogenic effects of dichloromethane). In addition, cancer values
derived for a population reflecting the estimated frequency of GST-T1 genotypes in the current
U.S. population (20% GST-Tl^, 48% GST-T1+/; and 32% GST-T1+/+) were presented. All
simulations also included a distribution of CYP activity based on data from Lipscomb et al.
(2003). The mean OSF based on liver tumors in mice exposed to dichloromethane in drinking
water, 2 x 10"3 (mg/kg-day)"1, based on what is assumed to be the most sensitive of the
populations (the GST-T1+ + group) is the recommended OSF to be for chronic oral exposures to
di chl or om ethane.
An OSF derived from the liver tumor data in the Serota et al. (1986b) study using
administered dose dosimetry rather than PBPK modeling is approximately one order of
magnitude higher than the current recommended value of 2 x 10"3 (per mg/kg-day). There is
approximately one to two orders of magnitude difference among the values based on different
dose metrics, scaling factors, and populations (Table 6-1).
-------
Table 6-1. Comparison of OSFs derived by using various assumptions and metrics, based on liver tumors in male
mice
Population"
GST-Tl+/+b
Mixed
Dose metric
Tissue-specific GST-metabolism rateb
Tissue-specific GST-metabolism rate
Whole-body GST metabolism rate
Route-to-route extrapolation, tissue-specific metabolism
Route-to-route extrapolation, tissue-specific metabolism
Route-to-route extrapolation, whole-body metabolism
Tissue-specific GST-metabolism rateb
Tissue-specific GST-metabolism rate
Whole-body GST metabolism rate
Route-to-route extrapolation, tissue-specific metabolism
Route-to-route extrapolation, tissue-specific metabolism
Route-to-route extrapolation, whole-body metabolism
Applied dose (human equivalent dose)
1995 IRIS assessment
Species, sex
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Tumor
Liver
Liver
Liver
Liver
Liver
Liver
Liver
Liver
Liver
Liver
Liver
Liver
Liver
Liver
Scaling
factor
7.0
1.0
7.0
7.0
1.0
7.0
7.0
1.0
7.0
7.0
1.0
7.0
Mean OSF
(mg/kg-d)1
1.7 x 10 3
2.4 x 1Q-4
9.3 x ID'4
1.2 x 1Q-4
1.7 x ID'5
6.7 x ID'5
9.4 x ID'4
1.3 x 1Q-4
5.4 x 1(T4
6.8 x 1(T5
9.7 x 1(T6
3.9 x 1(T5
1.0 x 10'2
7.5 x ID'3
Source
(table)
Table 5-13
Table 5-13
Table 5-13
Table 5-14
Table 5-14
Table 5-14
Table 5-13
Table 5-13
Table 5-13
Table 5-14
Table 5-14
Table 5-14
Table 5-15
aGST-Tl++ = homozygous, full enzyme activity; Mixed = genotypes based on a population reflecting the estimated frequency of genotypes in the current U. S.
population: 20% GST-Tl'7', 48% GST-T1+A, and 32% GST-T1+/+ (Haber et al., 2002).
bBolded value is the basis for the recommended OSF of 2 x 10"3 per mg/kg-d.
354
DRAFT - DO NOT CITE OR QUOTE
-------
The recommended OSF of 2 x io~3 (per mg/kg-day) is based on a tissue-specific GST
internal dose metric with allometric scaling. Although the involvement of the GST pathway in
carcinogenic response has been established, some uncertainty remains as to the metabolite(s)
involved and the rate at which those metabolites are cleared. The value derived specifically for
the GST-T1+ + population is recommended to provide protection for the population that is
hypothesized to be most sensitive to the carcinogenic effect. Application of ADAFs to the
cancer OSF is recommended in combination with appropriate exposure data when assessing risks
associated with early-life exposure (see Section 5.4.4 for more details).
6.2.5. Cancer IUR
The recommended cancer IUR is 1 x 10~8 (ug/m3)'1 for the development of liver and lung
cancers based on data from male B6C3Fi mice using a tissue-specific GST metabolism dose
metric. Data for liver and lung tumors in male and female B6C3Fi mice following exposure to
airborne dichloromethane were used to develop ITJRs for dichloromethane (Mennear et al., 1988;
NTP, 1986). This study was selected as the principal study to derive an IUR for
dichloromethane because of the completeness of the data, adequate sample size, and clear dose
response. In the NTP (1986) study, significant increases in incidence of liver and lung adenomas
and carcinomas were observed in both sexes of B6C3Fi mice exposed 6 hours/day, 5 days/week
for 2 years.
The PBPK model of Marino et al. (2006) for dichloromethane in the mouse was used to
calculate long-term daily average internal liver doses. The selected internal dose metrics for
liver tumors and lung tumors were long-term average daily mass of dichloromethane
metabolized via the GST pathway per unit volume of liver and lung, respectively. This approach
was taken based on evidence that GST-pathway metabolites produced from dichloromethane are
primarily responsible for dichloromethane carcinogenicity in mouse liver. The multistage dose-
response model (BMDS version 2.0) was used to fit the mouse liver tumor incidence and PBPK
model-derived internal dose data and to derive a mouse internal BMD and BMDLio. Because
the metric is a rate of metabolism rather than the concentration of putative toxic metabolites and
data pertaining to the reactivity or clearance rate of the relevant metabolite(s) are lacking, the
human BMDLio was derived by multiplying the mouse BMDLio by a BW° 75 allometric scaling
n 9s
factor (operationalized as [BWhUman/BWm0use] ~ 7) to account for the potential slower
clearance per volume tissue in the human compared with the mouse. A linear extrapolation
approach using the internal human BMDLio for liver and lung tumors was used to calculate
human tumor risk factors by dividing the BMR of 0.1 by the human BMDL for each tumor type.
Currently, there are no data from chronic inhalation cancer bioassays in mice or rats providing
support for a nonlinear dose-response relationship.
The human PBPK model (adapted from David et al. [2006]; see Appendix B) provided
distributions of human internal dose metrics of daily mass of dichloromethane metabolized via
355 DRAFT - DO NOT CITE OR QUOTE
-------
the GST pathway per unit volume of liver and lung resulting from chronic inhalation exposure to
a unit concentration of 1 ug/m3 dichloromethane (0.00029 ppm). As with the OSF, the cancer
IUR was derived for a sensitive population: a population composed entirely of carriers of the
GST-T1 homozygous positive genotype (that is, the group that would be expected to be most
sensitive to the carcinogenic effects of dichloromethane). In addition, cancer values derived for
a population reflecting the estimated frequency of GST-T1 genotypes in the current U.S.
population (20% GST-T1"'", 48% GST-T1+/", and 32% GST-T1+/+) were also presented. The
distributions of lURs for liver or lung tumors were generated by multiplying the human tumor
risk factor for each tumor type and sex by the distribution of internal doses from chronic
exposure to 1 ug/m3 dichloromethane. A procedure to combine risks for liver and lung tumors
using different dose metrics for the different tumors (i.e., liver-specific and lung-specific
metabolism for liver and lung tumors, respectively) was used to derive the recommended IUR of
O Q 1
1X10" (ug/m )" based on what is assumed to be the most sensitive of the populations, the
GST-T1+/+group.
o 'I 1
The current recommended IUR value of 1 x 10" (ug/m )" is approximately 47-fold lower
than the previous IRIS value of 4.7 x 10"7 (ug/m3)" *. An IUR derived from the liver tumor data
of the NTP (1986) study using applied concentration dosimetry rather than PBPK modeling, 3.6
x 10"7 (ug/m3)"1, is approximately one order of magnitude higher than the currently
O Q 1
recommended value of 1 x 10" (ug/m )" (Table 6-2). There is approximately one to two orders
of magnitude difference among the values based on different dose metrics, scaling factors, and
populations.
O Q 1
The recommended IUR value of 1 x 10" (ug/m )" is based on a tissue-specific GST-
internal dose metric with allometric scaling. Although the involvement of the GST pathway in
carcinogenic response has been established, some uncertainty remains as to the metabolite(s)
involved and the rate at which those metabolites are cleared. The value derived specifically for
the GST-T1+/+ population is recommended to provide protection for the population that is
hypothesized to be most sensitive to the carcinogenic effect. Application of ADAFs to the
cancer IUR is recommended when assessing risks associated with early-life exposure (see
Section 5.4.4 for more details).
356 DRAFT - DO NOT CITE OR QUOTE
-------
Table 6-2. Comparison of lURs derived by using various assumptions and metrics
Population"
GST-T1+/+
GST-T1+/+
GST-T1+/+
GST-T1+/+
GST-T1+/+
GST-T1+/+
GST-T1+/+
GST-T1+/+
GST-T1+/+
Mixed
Mixed
Mixed
Mixed
Mixed
Mixed
Mixed
Mixed
Mixed
Dose metric
Tissue-specific GST-metabolism rate0
Tissue-specific GST-metabolism rate
Tissue-specific GST-metabolism rate
Tissue-specific GST-metabolism rate
Tissue-specific GST -metabolism rate
Tissue-specific GST-metabolism rate
Whole-body GST metabolism rate
Whole-body GST metabolism rate
Whole-body GST metabolism rate
Tissue-specific GST-metabolism rate
Tissue-specific GST-metabolism rate
Tissue-specific GST-metabolism rate
Tissue-specific GST-metabolism rate
Tissue-specific GST-metabolism rate
Tissue-specific GST-metabolism rate
Whole-body GST metabolism rate
Whole-body GST metabolism rate
Whole-body GST metabolism rate
Administered concentration (HEC)
Administered concentration (HEC)
1995 IRIS assessment"
Species, sex
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Mouse, male
Tumor type
Liver and lung
Liver
Lung
Liver and lung
Liver
Lung
Liver and lung
Liver
Lung
Liver and lung
Liver
Lung
Liver and lung
Liver
Lung
Liver and lung
Liver
Lung
Liver
Lung
Liver and lung
Scaling
factor
7.0
7.0
7.0
1.0
1.0
1.0
7.0
7.0
7.0
7.0
7.0
7.0
1.0
1.0
1.0
7.0
7.0
7.0
12.7
IURb
(jig/m3)-1
1.3 x 10 8
8.5 x 1Q-9
5.6 x 1Q-9
1.9 x 1Q-9
1.2 x 1Q-9
8.0 x 1Q-10
1.6 x 1Q-8
5.5 x 1Q-9
1.2 x 1Q-8
7.4 x 10'9
4.8 x 10'9
3.2 x 10'9
1.1 x 10'9
6.8 x KT10
4.5 x 10"10
9.2 x 10'9
3.1 x 10'9
6.9 x 10'9
3.6 x 10'7
8.1 x 10'7
4.7 x 10'7
Source
(table)
Table 5-20
Table 5-19
Table 5-19
Table 5-20
Table 5-19
Table 5-19
Table 5-20
Table 5-19
Table 5-19
Table 5-20
Table 5-19
Table 5-19
Table 5-20
Table 5-19
Table 5-19
Table 5-20
Table 5-19
Table 5-19
Table 5-21
Table 5-21
aGST-Tl+ + = homozygous, Ml enzyme activity; Mixed = genotypes based on a population reflecting the estimated frequency of genotypes in the current U.S.
population: 20% GST-Tr7', 48% GST-T1+/; and 32% GST-T1+/+ (Haber et al., 2002).
TSased on mean value of the derived distributions.
°Bolded value is the basis for the recommended IUR of 1 x 10"8 ug/m3 per mg/kg-d.
357
DRAFT - DO NOT CITE OR QUOTE
-------
6.2.6. Uncertainties in Cancer Risk Values
The database of animal bioassays identifies the liver and lung as the most sensitive target
organs for dichloromethane-induced tumor development, and there is high confidence that the
dose-response data for liver and lung cancer in mice represent the best available data for
derivation of human cancer risks. A dose-response relationship was seen with respect to liver
cancer in mice exposed orally and with respect to liver and lung cancer in mice exposed by
inhalation. Statistically significant increases in benign mammary gland tumors were observed in
one study of F344 rats exposed by inhalation to 2,000 or 4,000 ppm (Mennear et al., 1988; NTP,
1986); evidence for a tumorigenic mammary gland response in Sprague-Dawley rats was limited
to increased numbers of benign mammary tumors per animal at levels of 50-500 ppm (Nitschke
et al., 1988a) or 500-3,500 ppm (Burek et al., 1984). A gavage study in female Sprague-Dawley
rats reported an increased incidence of malignant mammary tumors, mainly adenocarcinomas (8,
6, and 18% in the control, 100, and 500 mg/kg dose groups, respectively), but the increase was
not statistically significant. Data were not provided to allow an analysis that accounts for
differing mortality rates (Maltoni et al., 1988). The toxicokinetic or mechanistic events that
might lead to mammary gland tumor development in rats are unknown, although CYP2E1 (El-
Rayes et al., 2003; Hellmold et al., 1998) and GST-T1 expression have been detected in human
mammary tissue (Lehmann and Wagner, 2008). Rare CNS tumors were observed in one study in
rats spanning a relatively low range of exposures (0-500 ppm) (Nitschke et al., 1988a). These
cancers were not seen in two other studies in rats, both involving higher doses (1,000-
4,000 ppm) (NTP, 1986; Burek et al., 1984), or in a similar high-dose study in mice (NTP,
1986). The relative rarity of the tumors precludes the use of the low-dose exposure study
(Nitschke et al., 1988a) in a quantitative dose-response assessment. The available epidemiologic
studies provide evidence of an association between dichloromethane and liver cancer, brain
cancer, and some hematopoietic cancers. The available epidemiologic studies do not provide an
adequate basis for the evaluation of the role of dichloromethane in breast cancer because there
are currently no cohort studies with adequate statistical power and no case-control studies with
adequate exposure methodology to examine this relationship.
There is uncertainty as to whether the reactivity of the toxic dichloromethane metabolites
is sufficiently high enough to preclude systemic distribution. Therefore, alternative derivations
of cancer risk values were performed under the assumption that high reactivity leads to complete
clearance from the tissue in which the active metabolite is formed (scaling factor = 1.0). The
difference in scaling factor (7.0 for allometric scaling versus 1.0) results in a sevenfold decrease
in estimated cancer toxicity values. Using a whole-body GST metabolism dose metric, the
resulting OSF and IUR for liver cancer was approximately fivefold lower than when tissue-
specific dose metrics were used; however, the lURs for lung cancer and for the combined liver
and lung cancer risk were higher with the whole-body compared with the tissue-specific metric
358 DRAFT - DO NOT CITE OR QUOTE
-------
(Table 6-1 and Table 6-2). This difference reflects the lower metabolism that occurs in human
versus mouse lung (relative to total): lung-specific metabolism is lower in humans than mice, so
the predicted risk in the lung is lower when based on that metabolism versus whole-body
metabolism. The mechanistic data support the hypothesis that reactive metabolites produced in
the target tissues do not distribute significantly beyond those tissues and cause deleterious effects
in the metabolizing tissues soon after generation. Thus, there is less uncertainty in the cancer
risk values derived by using a tissue-specific GST metabolism dose metric compared with those
derived using a whole-body GST metabolism dose metric.
Uncertainty in the ability of the PBPK models to estimate animal and human internal
doses from lifetime bioassay low-level environmental exposures may affect the confidence in the
cancer risk extrapolated from animal data. Uncertainties in the mouse and human model
parameter values were integrated quantitatively into parameter estimation by utilizing
hierarchical Bayesian methods to calibrate the models at the population level (David et al., 2006;
Marino et al., 2006). However, with the subsequent deterministic application of the mouse
model (using only the mean value for each parameter distribution), the information contained in
the mouse parameter uncertainties reported by Marino et al. (2006) is not integrated into the final
risk estimates described here.
The use of Monte Carlo sampling to define human model parameter distributions allowed
for derivation of human distributions of dosimetry and cancer risk, providing for bounds on the
recommended risk values. A sensitivity analysis was performed to identify model parameters
most influential on the predictions of dose metrics used to estimate oral and inhalation cancer
risks. For inhalation exposures in mice, the PB, followed closely by the first-order
GST-mediated metabolism rate, had the greatest impact on the dose metric for liver cancer (mg
dichloromethane metabolized via GST pathway per liter liver per day). For drinking water
exposures in mice, the first-order GST-mediated metabolism rate, followed by the
CYP-mediated maximum reaction velocity (Vmaxc) affected the liver cancer dose metric to the
greatest extent. For mice inhaling dichloromethane, the lung cancer dose metric (mg dichloro-
methane metabolized via GST pathways per liter lung per day), like the liver cancer metric, was
highly affected by the first-order GST-mediated metabolism rate and the PB. However, the lung
cancer dose metric was most sensitive to the proportional yield of liver GST-mediated metabolic
activity attributed to the lung. The PB was experimentally determined, lending high confidence
to its value. In contrast, values for the three metabolic parameters were determined by
computational optimization against data sets not directly measuring dichloromethane or its
metabolites in the target/metabolizing tissues. It is uncertain how alternative values for these
three parameters would affect the estimation of animal BMDL10 values and, ultimately, lURs and
OSFs. In addition, specific uncertainty remains concerning the human PBPK parameter
distributions (see discussion on kfc in Section 3.5.5).
359 DRAFT - DO NOT CITE OR QUOTE
-------
Finally, while the structure and equations used in the existing model have been described
extensively in peer-reviewed publications, uncertainty remains concerning the model structure.
Specifically an alternative (dual-binding-site) CYP metabolic rate equation (Korzekwa et al.,
1998) for dichloromethane may better describe CYP2E1-mediated GST metabolism. However,
this hypothesis requires further testing in the laboratory and integration of the alternate rate
equation into the PBPK modeling.
360 DRAFT - DO NOT CITE OR QUOTE
-------
7. REFERENCES
* denotes reference added after External Peer Review
ACGIH (American Conference of Governmental Industrial Hygienists). (2001) Dichloromethane. In:
Documentation of the threshold limit values and biological exposure indices. 7th edition. Cincinnati, OH: American
Conference of Governmental Industrial Hygienists.
Alexeef, GV; Kilgore, WW. (1983) Learning impairment in mice following acute exposure to dichloromethane and
carbon tetrachloride. J Toxicol Environ Health 11:569-581.
Allen, J; Kligerman, A; Campbell, J; et al. (1990) Cytogenetic analyses of mice exposed to dichloromethane.
Environ Mol Mutagen 15:221-228.
Anders, MW; Sunram, JM. (1982) Transplacental passage of dichloromethane and carbon monoxide. Toxicol Lett
12:231-234.
Andersen, ME; Clewell, HJ, III; Gargas, ML; et al. (1987) Physiologically based pharmacokinetics and the risk
assessment process for methylene chloride. Toxicol Appl Pharmacol 87:185-205.
Andersen, ME; Clewell, HJ, III; Gargas, ML; et al. (1991) Physiologically based pharmacokinetic modeling with
dichloromethane, its metabolite, carbon monoxide, and blood carboxyhemoglobin in rats and humans. Toxicol Appl
Pharmacol 108:14-27.
Andrae, U; Wolff, T. (1983) Dichloromethane is not genotoxic in isolated rat hepatocytes. Arch Toxicol 52:287-
290.
Angelo, MJ; Pritchard, AB; Hawkins, DR; et al. (1986a) The pharmacokinetics of dichloromethane. I. Disposition in
B6C3F! mice following intravenous and oral administration. Food Chem Toxicol 24(9):965-974.
Angelo, MJ; Pritchard, AB; Hawkins, DR; et al. (1986b) The pharmacokinetics of dichloromethane. II. Disposition
in Fischer 344 rats following intravenous and oral administration. Food Chem Toxicol 24(9):975-980.
Anthony, T. (1979) Methylene chloride. In: Kirk, RE; Othmer, DF; eds. Kirk-Othmer encyclopedia of chemical
technology. 3rd edition. New York, NY: John Wiley & Sons, pp. 686-693.
Anundi, H; Lind, ML; Friis, L; et al. (1993) High exposures to organic solvents among graffiti removers. Int Arch
Occup Environ Health 65:247-251.
Aranyi, C; O'Shea, WJ; Graham, JA; et al. (1986) The effects of inhalation of organic chemical air contaminants on
murine lung host defenses. Fundam Appl Toxicol 6:713-720.
Arcus-Arth, A; Blaisdell, RJ. (2007) Statistical distributions of daily breathing rates for narrow age groups of infants
and children. Risk Anal 27:97-110.
Astrand, I; Ovrum, P; Carlsson, A. (1975) Exposure to methylene chloride. I. Its concentration in alveolar air and
blood during rest and exercise and its metabolism. Scand J Work Environ Health 1:78-94.
Atkinson, R. (1989) Kinetics and mechanisms of the gas-phase reactions of the hydroxyl radical with organic
compounds. J Phys Chem Ref Data 1:63.
ATSDR (Agency for Toxic Substances and Disease Registry). (2000) Toxicological profile for methylene chloride.
Public Health Service, U.S. Department of Health and Human Services, Atlanta, GA. Available online at
http://www.atsdr.cdc.gov/toxpro2.html (accessed March 9, 2010).
Bakinson, MA; Jones, RD. (1985) Gassings due to methylene chloride, xylene, toluene, and styrene reported to Her
Majesty's factory inspectorate 1961-80. Br J Ind Med 42:184-190.
361 DRAFT - DO NOT CITE OR QUOTE
-------
*Banger, KK; Foster, JR; Lock, EA; Reed, CJ. (1994) Immunohistochemical localisation of six glutathione S-
transferases within the nasal cavity of the rat. Arch Toxicol 69:91-98.
*Barry, KH; Zhang, Y; Lan, Q; et al. (2011) Genetic variation in metabolic genes, occupational solvent exposure,
and risk of non-hodgkin lymphoma. Am J Epidemiol 173:404-413.
*Bebia, Z; Buch, SC; Wilson, JW; et al. (2004) Bioequivalence revisited: Influence of age and sex on CYP
enzymes. Clin Pharmacol Therap 76: 618-627.
Bell, BP; Franks, P; Hildreth, N; et al. (1991) Methylene chloride exposure and birth weight in Monroe County,
New York. Environ Res 55:31-39.
*Benowitz, NL; Peng, M; Jacob, P 3rd. (2003) Effects of cigarette smoking and carbon monoxide on chlorzoxazone
and caffeine metabolism. Clin Pharmacol Ther 74:468-474.
Berman, E; Schlicht, M; Moser, VC; et al. (1995) A multidisciplinary approach to lexicological screening: I.
Systemic toxicity. J Toxicol Environ Health 45:27-143.
Bernauer, U; Vieth, B; Ellrich, R; et al. (2000) CYP2E1 expression in bone marrow and its intra- and interspecies
variability: approaches for a more reliable extrapolation from one species to another in the risk assessment of
chemicals. Arch Toxicol 73:618-624.
Blair, A; Hartge, P; Stewart, P; et al. (1998) Mortality and cancer incidence of aircraft maintenance workers exposed
to trichloroethylene and other organic solvents and chemicals: extended follow up. Occup Environ Med 55:161-
171.
Blocki, FA; Logan, MSP; Baoli, C; et al. (1994) Reaction of rat liver glutathione s-transferase and bacterial
dichloromethane dehalogenase with dihalomethanes. J Biol Chem 269:8826-8830.
Bogaards, JJP; vanOmmen, B; vanBladeren, PJ. (1993) Interindividual differences in the in vitro conjugation of
methylene chloride with glutathione by cytosolic glutathione S-transferase in 22 human liver samples. Biochem
Pharmacol 45(10):2166-2169.
*Bogen, KT. (1990) Uncertainty in Environmental Health Risk Assessment. London: Taylor & Francis [Chapter
IV].
Bornschein, RL; Hastings, L; Manson, JM. (1980) Behavioral toxicity in the offspring of rats following maternal
exposure to dichloromethane. Toxicol Appl Pharmacol 52:29-37.
Bos, PMJ; Zeilmaker, MJ; van Eijkeren, JCH. (2006) Application of physiologically based pharmacokinetic
modeling in setting acute exposure guideline levels for methylene chloride. Toxicol Sci 91:576-585.
Boublik, T; Fried, V; Hala, E. (1984) The vapor pressures of pure substances: selected values of the temperature
dependence of the vapour pressures of some pure substan
revised edition. Amsterdam, Netherlands: Elsevier, p. 42.
dependence of the vapour pressures of some pure substances in the normal and low pressure region. Vol. 17. 2nd
Bowen, SE; Batis, JC; Paez-Martinez, N; et al. (2006) The last decade of solvent research in animal models of
abuse: mechanistic and behavioral studies. Neurotoxicol Teratol 28:636-647.
Breslow, NE; Day, NE. (1987) Statistical methods in cancer research. Volume II. The design and analysis of cohort
studies. IARC Sci Publ 82:1-406.
Briving, C; Hambergerm, A; Kjellstrand, P; et al. (1986) Chronic effects of dichloromethane on amino acids,
glutathione and phosphoethanolamine in gerbil brain. Scand J Work Environ Health 12:216-220.
Brown, RP; Delp, MD; Lindstedt, SL; et al. (1997) Physiological parameter values for physiologically-based
pharmacokinetic models. Toxicol Ind Health 13:407-484.
362 DRAFT - DO NOT CITE OR QUOTE
-------
Brown-Woodman, PDC; Hayes, LC; Huq, F; et al. (1998) In vitro assessment of the effect of halogenated
hydrocarbons: chloroform, dichloromethane, and dibromoethane on embryonic development of the rat. Teratology
57:321-333.
Bruhn, C; Brockmoller, J; Kerb, R; et al. (1998) Concordance between enzyme activity and genotype of glutathione
S-transferase theta (GST-T1). BiochemPharmacol 56:1189-1193.
Brzezinski, MR; Boutelet-Bochan, H; Person, RE; et al. (1999) Catalytic activity and quantitation of cytochrome P-
450 2E1 in prenatal human brain. J Pharmacol Exp Ther 289:1648-1653.
Burek, JD; Nitschke, KD; Bell, TJ; et al. (1984) Methylene chloride: a two-year inhalation toxicity and oncogenicity
study in rats and hamsters. Fundam Appl Toxicol 4:30-47.
Cagiano, R; Ancona, D; Cassano, T; et al. (1998) Effects of prenatal exposure to low concentrations of carbon
monoxide on sexual behaviour and mesolimbic dopaminergic function in rat offspring. Br J Pharmacol 125(4):909-
915.
Callen, DF; Wolf, CR; Philpot, RM. (1980) Cytochrome P-450 mediated genetic activity and cytotoxicity of seven
halogenated aliphatic hydrocarbons in Saccharomyces cerevisiae. Mutat Res 77:55-63.
Candrian, U; You, M; Goodrow, T; et al. (1991) Activation of protooncogenes in spontaneously occurring non-liver
tumors from C57BL/6 x C3H Fl mice. Cancer Res 51:1148-1153.
Cantor, KP; Stewart, PA; Brinton, LA; et al. (1995) Occupational exposures and female breast cancer mortality in
the United States. J Occup Environ Med 37:336-348.
Carlsson, A; Hultengren, M. (1975) Exposure to methylene chloride. III. Metabolism of 14C-labelled methylene
chloride in rat. Scand J Work Environ Health 1:104-108.
Casanova, M; Deyo, DF; Heck, Hd'A. (1992) Dichloromethane (methylene chloride): metabolism to formaldehyde
and formation of DNA-protein cross links in B6C3Fi mice and Syrian golden hamsters. Toxicol Appl Pharmacol
114:162-165.
Casanova, M; Conolly, RB; Heck, Hd'A. (1996) DNA-protein cross-links (DPX) and cell proliferation in B6C3FJ
mice but not Syrian golden hamsters exposed to dichloromethane: pharmacokinetics and risk assessment with DPX
as dosimeter. Fundam Appl Toxicol 31:103-106.
Casanova, M; Bell, DA; Heck, Hd'A. (1997) Dichloromethane metabolism to formaldehyde and reaction of
formaldehyde with nucleic acids in hepatocytes of rodents and humans with and without glutathione S-transferase
Tl and Ml genes. Fundam Appl Toxicol 37:168-180.
Chang, Y; Yang, CC; Deng, JF; et al. (1999) Diverse manifestations of oral methylene dichloride poisoning: report
of 6 cases. Clin Toxicol 37:497-504.
Cherry, N; Venables, H; Waldron, HA; et al. (1981) Some observations on workers exposed to methylene chloride.
BrJInd Med 38:351-355.
Cherry, N; Venables, H; Waldron, HA. (1983) The acute behavioural effects of solvent exposure. J Soc Occup Med
33:13-18.
*Cho, CS; McLean, AJ; Rivory, LP; et al. (2001) Carbon monoxide wash-in method to determine gas transfer in
vascular beds: application to rat hindlimb. Am J Physiol Heart Circ Physiol. 280:H1802-806.
Clewell, HJ, III; Gearhart, JM; Andersen, ME. (1993) Analysis of the metabolism of methylene chloride in the
B6C3F! mouse and its implications for human carcinogenic risk. Submitted to Occupational Safety and Health
Administration, U.S. Department of Labor, Washington, DC. Docket # H-071, Exhibit #96.
Clewell, HJ; Gentry, PR; Covington, TR; et al. (2004) Evaluation of the potential impact of age- and gender-specific
pharmacokinetic differences on tissue dosimetry. Toxicol Sci 79:381-393.
363 DRAFT - DO NOT CITE OR QUOTE
-------
CMR (Chemical Marketing Reporter). (1973) Chemical profile on methylene chloride. Chemical Marketing
Reporter, October 22, 1973.
CMR (Chemical Marketing Reporter). (1979) Chemical profile on methylene chloride. Chemical Marketing
Reporter, August 6, 1979.
CMR (Chemical Marketing Reporter). (1982) Chemical profile on methylene chloride. Chemical Marketing
Reporter, July 12, 1982.
CMR (Chemical Marketing Reporter). (1997) Chemical profile on methylene chloride. Chemical Marketing
Reporter, November 24, 1997.
CMR (Chemical Marketing Reporter). (2000) Chemical profile on methylene chloride. Chemical Marketing
Reporter, October 9, 2000.
Cocco, P; Heineman, EF; Dosemeci, M. (1999) Occupational risk factors for cancer of the central nervous system
(CNS) among US women. Am J Ind Med 36:70-74.
Cohen J. (1987) Statistical power analysis for the behavioral sciences. Hillsdale, NJ: Lawrence Erlbaum Associates,
p. 36-37.
Condie, LW; Smallwood, CL; Laurie, RD. (1983) Comparative renal and hepatotoxicity of halomethanes:
bromodichloromethane, bromoform, chloroform, dibromochloromethane, and methylene chloride. Drug Chem
Toxicol 6:563-578.
*Costantini, AS; Miligi, L; Kriebel, D; et al. (2001) A multicenter case-control study in Italy on
hematolymphopoietic neoplasms and occupation. Epidemiology 12:78-87.
*Costantini, AS; Benvenuti, A; Vineis, P; et al. (2008) Risk of leukemia and multiple myeloma associated with
exposure to benzene and other organic solvents: evidence from the Italian Multicenter Case-control study. Am J Ind
Med 51:803-811.
Crebelli, R; Benigni, R; Franekic, J; et al. (1988) Induction of chromosome malsegregation by halogenated organic
solvents mAspergillus nidulans: unspecific or specific mechanism? Mutat Res 201:401-411.
Crebelli, R; Carere, A; Leopardi, P; et al. (1999) Evaluation of 10 aliphatic halogenated hydrocarbons in the mouse
bone marrow micronucleus test. Mutagenesis 14:207-215.
Dankovic, DA; Bailer, AJ. (1994) The impact of exercise and intersubject variability on dose estimates for
dichloromethane derived from a physiologically based pharmacokinetic model. Fundam Appl Toxicol 22:20-25.
David, RM; Clewell, HJ; Gentry, PR; et al. (2006) Revised assessment of cancer risk to dichloromethane II.
Application of probabilistic methods to cancer risk determinations. Regul Toxicol Pharmacol 45:55-65.
De Salvia, MA; Cagiano, R; Carratu, MR; et al. (1995) Irreversible impairment of active avoidance behavior in rats
prenatally exposed to mild concentrations of carbon monoxide. Psychopharmacology (Berl) 122:66-71.
Dearfield, KL; Moore, MM. (2005) Use of genetic toxicology information for risk assessment. Environ Mol
Mutagen 46:236-245.
*Dell, LD; Mundt, KA; McDonald, M; Tritschler, JP 2nd; Mundt, DJ. (1999) Critical review of the epidemiology
literature on the potential cancer risks of methylene chloride. Int Arch Occup Environ Health. 72:429^142.
DeMarini, DM; Shelton, ML; Warren, SH; etal. (1997) Glutathione S-transferase-mediated induction of GC^AT
transitions by halomethanes in Salmonella. Environ Mol Mutagen 30:440-447.
Devereux, TR; Foley, JF; Maronpot, RR; et al. (1993) Ras proto-oncogene activation in liver and lung tumors from
B6C3F! mice exposed chronically to methylene chloride. Carcinogenesis 14:795-801.
364 DRAFT - DO NOT CITE OR QUOTE
-------
Dillon, D; Edwards, I; Combes, R; et al. (1992) The role of glutathione in the bacterial mutagenicity of vapour phase
dichloromethane. Environ Mol Mutagen 20:211-217.
Dinse GE; Lagakos SW. (1982) Nonparametric estimation of lifetime and disease onset distributions from
incomplete observations. Biometrics 38:921-932.
DiVincenzo, GD; Kaplan, CJ. (1981) Uptake, metabolism, and elimination of methylene chloride vapor by humans.
Toxicol Appl Pharmacol 59:130-140.
DiVincenzo, GD; Yanno, FJ; Astill, BD. (1971) The gas chromatographic analysis of methylene chloride in breath,
blood, and urine. Am Ind Hyg Assoc J 32:387-391.
DiVincenzo, GD; Yanno, FJ; Astill, BD. (1972) Human and canine exposures to methylene chloride vapor. Am Ind
Hyg Assoc 133:125-134.
Doherty, AT; Ellard, S; Parry, EM; et al. (1996) An investigation into the activation and deactivation of chlorinated
hydrocarbons to genotoxins in metabolically competent human cells. Mutagenesis 11:247-274.
*Dosemeci, M; Cocco, P; Gomez, M; Stewart, PA; Heineman, EF. (1994) Effects of three features of a job-exposure
matrix on risk estimates. Epidemiology 5:124-127.
Dosemeci, M; Cocco, P; Chow, WH. (1999) Gender differences in risk of renal cell carcinoma and occupational
exposures to chlorinated aliphatic hydrocarbons. Am J Ind Med 36:54-59.
Dumas, S; Parent, ME; Siemiatychi, J. (2000) Rectal cancer and occupational risk factors: a hypothesis-generating,
exposure-based case-control study. Int J Cancer 87:874-879.
Eastmond, DA; Hartwig, A; Anderson, D; et al. (2009) Mutagenicity testing for chemical risk assessment: update of
WHO/IPCS harmonized scheme. Mutagenesis 24:341-349.
Ehrlich, R. (1980) Interaction between environmental pollutants and respiratory infections. Environ Health Perspect
35:89-100
El-Masri, HA; Bell, DA; Portier, CJ. (1999) Effects of glutathione transferase theta polymorphism on the risk
estimates of dichloromethane to humans. Toxicol Appl Pharmacol 158:221-230.
El-Rayes, BF; Ali, S; Heilbrun, LK; et al. (2003) Cytochrome P450 and glutathione transferase expression in human
breast cancer. Clin Cancer Res 9:1705-1709.
Engstrom, J; Bjurstrom, R. (1977) Exposure to methylene chloride. Content in subcutaneous adipose tissue. Scand J
Work EnvironHealth 3:215-224.
Estill, CF; Spencer, AB. (1996) Case study: control of methylene chloride exposures during furniture stripping. Am
Ind Hyg Assoc J 57:43-49.
Evans, MV; Caldwell, JC. (2010) Evaluation of two different metabolic hypotheses for dichloromethane toxicity
using physiologically based pharmacokinetic modeling for in vivo inhalation gas uptake data exposure in female
B6C3F1 mice. Toxicol Appl Pharmacol 244:280-290.
Fechner, G; Ortmann, C; Du Chesne, A; et al. (2001) Fatal intoxication due to excessive dichloromethane inhalation.
Forensic Sci Int 122:69-72.
Fechter, LD. (1987) Neurotoxicity of prenatal carbon monoxide exposure. Res Rep Health Eff Inst 12:3-22.
Fisher, JW; Whittaker, TA; Taylor, DH; et al. (1989) Physiologically based pharmacokinetic modeling of the
pregnant rat: a multiroute exposure model for trichloroethylene and its metabolite, trichloroacetic acid. Toxicol Appl
Pharmacol 99:395-414.
365 DRAFT - DO NOT CITE OR QUOTE
-------
Fodor, GG; Prajsnar, D; Schlipkoter, H. (1973) Endogenous CO formation by incorporated halogenated
hydrocarbons of the methane series. Staub Reinhalt Luft 33:260-261.
Foley, JF; Tuck, PD; Ton, T; et al. (1993) Inhalation exposure to a hepatocarcinogenic concentration of methylene
chloride does not induce sustained replicative DNA synthesis in hepatocytes of female B6C3F! mice.
Carcinogenesis 14811-817.
Forster, HV; Graff, S; Hake, CL; et al. (1974) Pulmonary-hematologic studies on humans during exposure to
methylene chloride. Prepared by the Department of Environmental Medicine, Medical College of Wisconsin,
Milwaukee, WI, for the National Institute for Occupational Safety and Health, Centers for Disease Control and
Prevention, U.S. Department of Health and Human Services, Cincinnati, OH. Available from the National Technical
Information Service, Springfield, VA; PB82-151697.
Foster, JR; Green, T; Smith, LL; et al. (1992) Methylene chloride-an inhalation study to investigate pathological and
biochemical events occurring in the lungs of mice over an exposure period of 90 days. Fundam Appl Toxicol
18:376-388.
Foster, JR; Green, T; Smith, LL; et al. (1994) Methylene chloride: an inhalation study to investigate toxicity in the
mouse lung using morphological, biochemical, and Clara cell culture techniques. Toxicology 91:221-234.
Friedlander, BR; Hearne, T; Hall, S. (1978) Epidemiologic investigation of employees chronically exposed to
methylene chloride. J Occup Med 20:657-666.
Fujimoto, K; Arakawa, S; Watanabe, T; et al. (2007) Generation and functional characterization of mice with a
disrupted glutathione S-transferase, theta 1 gene. Drug Metab Dispos 35:2196-2202.
Fuxe, K; Andersson, K; Hansson, T; et al. (1984) Central catecholamine neurons and exposure to dichloromethane.
Selective changes in amine levels and turnover in tel- and diencephalic and Na nerve terminal systems and in the
secretion of anterior pituitary hormone in the male rat. Toxicol 29:293-305.
Gamberale, F; Annwall, G; Hultengren, M (1975) Exposure to methylene chloride. II. Psychological function.
Scand J Work Environ Health 1:95-103.
Gargas, ML; Clewell, HJ; Andersen, ME. (1986) Metabolism of inhaled dihalomethanes in vivo: differentiation of
kinetic constants for two independent pathways. Toxicol Appl Pharmacol 82:211-223.
Garrett, NE; Lewtas, J. (1983) Cellular toxicity in Chinese hamster ovary cells culture. I. Analysis of cytotoxicity
endpoints for twenty-nine priority pollutants. Environ Res 32:455^4-65.
Garte, S; Gaspari, L; Alexandrie, AK; et al. (2001) Metabolic gene polymorphism frequencies in control
populations. Cancer Epidemiol Biomarkers Prev 10:1239-1248.
General Electric Company. (1976) Dichloromethane and ninety day oral toxicity study in rats. Prepared by the
International Research and Development Corporation, Mattawan, MI for the Plastics Tech Department, General
Electric Co., Pittsfield, MA. Submitted under TSCA Section 8D; EPA Document No. 86-878210707; NTIS No.
OTS0205887.
Gibbs, GW. (1992) The mortality of workers employed at a cellulose acetate and triacetate fibers plant in
Cumberland, Maryland: a "1970" cohort followed 1970-1989 [final report]. Prepared by Safety Health
Environment International Consultants Corporation, Winterburn, Alberta, Canada, for the Hoechst Celanese
Corporation, Somerville, NJ.
Gibbs, GW; Amsel, J; Soden, K. (1996) A cohort mortality study of cellulose triacetate-fiber workers exposed to
methylene chloride. J Occup Environ Med 38:693-697.
Giustino, A; Cagiano, R; Carratu, MR; et al. (1999) Prenatal exposure to low concentrations of carbon monoxide
alters habitation and non-spatial working memory in rat offspring. Brain Res 844:201-205.
366 DRAFT - DO NOT CITE OR QUOTE
-------
Glatzel, W; Tietze, K; Gutewort, R; et al. (1987) Interaction of dichloromethane and ethanol in rats: toxicokinetics
and nerve conduction velocity. Alcoholism: ClinExp Res 11:450-455.
Gocke, E; King, MT; Eckhardt, K; et al. (1981) Mutagenicity of cosmetic ingredients licensed by the European
communities. Mutat Res 90:91-109.
*Gold, LS; Stewart, PA; Milliken, K; et al. (2011) The relationship between multiple myeloma and occupational
exposure to six chlorinated solvents. Occup Environ Med 68:391-9.
Gomez, MR. (1996) Exposure determinants needed to improve the assessment of exposure. Am J Ind Med 29:569-
570.
Gomez, MR; Cocco, P; Dosemeci, M; et al. (1994) Occupational exposure to chlorinated aliphatic hydrocarbons: job
exposure matrix. Am J Ind Med 26:171-183.
Goodman, DG; Maronpot, RR; Newbene, PM, et al. (1994) Proliferative and selected other lesions in the liver of
rats. Gl-5. In: Guides fortoxologic pathology. STP/ARP/AFIP, Washington DC: 1-24.
Goulle, JP; Lacroix, C; Vaz, E; et al. (1999) Fatal case of dichloromethane poisoning. J Anal Toxicol 23:380-383.
Graves, RJ; Green, T. (1996) Mouse liver glutathione S-transferase mediated metabolism of methylene chloride to a
mutagen in the CHO/HPRT assay. Mutat Res 367:143-150.
Graves, RJ; Callander, RD; Green, T. (1994a) The role of formaldehyde and S-chloromethylglutathione in the
bacterial mutagenicity of methylene chloride. Mutat Res 320:235-243.
Graves, RJ; Coutts, C; Eyton-Jones, H; et al. (1994b) Relationship between hepatic DNA damage and methylene
chloride-induced hepatocarcinogenicity inB6C3F! mice. Carcinogenesis 15:991-996.
Graves, RJ; Coutts, C; Green, T. (1995) Methylene chloride-induced DNA damage: an interspecies comparison.
Carcinogenesis 16:1919-1926.
Graves, RJ; Trueman, P; Jones, S; et al. (1996) DNA sequence analysis of methylene chloride-induced HPRT
mutations in Chinese hamster ovary cells: comparison with the mutation spectrum obtained for 1,2-dibromoethane
and formaldehyde. Mutagenesis 11:229-233.
Green, T. (1983) The metabolic activation of dichloromethane and chlorofluoromethane in a bacterial mutation
assay using Salmonella typhimurium. Mutat Res 118:277-288.
Green, T. (1989) A biological data base for methylene chloride risk assessment. In: Travis, CC; ed. Biologically
based methods for cancer risk assessment. New York, NY: Plenum Press, p. 289-300.
Green, T. (1997) Methylene chloride induced mouse liver and lung tumours: an overview of the role of mechanistic
studies in human safety assessment. Hum Exp Toxicol 16:3-13.
Guengerich, FP. (1997) Mechanisms of mutagenicity of DNA adducts derived from alkyl and vinyl halides. Jpn J
Toxicol Environ Health 43:69-82.
Haber, LT; Maier, A; Gentry, PR; et al. (2002) Genetic polymorphisms in assessing interindividual variability in
delivered dose. Reg Toxicol Pharmacol 35:177-197.
Hall, AH; Rumack, BH. (1990) Methylene chloride exposure in furniture-stripping shops: ventilation and respirator
use practices. J Occup Med 32:33-37.
Hairier, E; Schroder, KR; Asmuth, K; et al. (1994) Metabolism of dichloromethane (methylene chloride) to
formaldehyde in human erythrocytes: influence of polymorphism of glutathione transferase theta (GST Tl-1). Arch
Toxicol 68:423-427.
367 DRAFT - DO NOT CITE OR QUOTE
-------
Hansch, C; Leo, A; Hoekman, D. (1995) Exploring QSAR. Hydrophobic, electronic, and steric constants. ACS
professional reference book. Washington, DC: American Chemical Society; p. 3.
Hardie, DWF. (1964) Methylene chloride. In: Kirk-Othmer encyclopedia of chemical technology. 2nd edition. New
York, NY: John Wiley & Sons; pp. 111-119.
Hardin, BD; Manson, J. (1980) Absence of dichloromethane teratogenicity with inhalation exposure in rats. Toxicol
Appl Pharmacol 52:22-28.
Hashmi, M; Dechert, S; Dekant, W; et al. (1994) Bioactivation of [13C]dichloromethane in mouse, rat, and human
liver cytosol: 13C nuclear magnetic resonance spectroscopic studies. Chem Res Toxicol 7:291-296.
Haufroid, V; Ligocka, D; Buysschaert, M; et al. (2003) Cytochrome P4502E1 (CYP2E1) expression in peripheral
blood lymphocytes: evaluation in hepatitis C and diabetes. Eur J Clin Pharmacol 59:29-33.
Haun, CC; Harris, ES; Darmer, KI, Jr. (1971) Continuous animal exposure to methylene chloride. In: Proceedings
of the 2nd conference of environmental toxicology; August 31-September2; Fairborn, OH; Paper No. 10; AMRL-
TR-71-120. Aerospace Medical Research Laboratory, Wright-Patterson Air Force Base, OH, pp. 125-135.
Available from the National Technical Information Service, Springfield, VA; AD751432 (individual paper);
AD746660 (entire proceedings).
Haun, CC; Vernot, EH; Darmer, KI, Jr; et al. (1972) Continuous animal exposure to low levels of dichloromethane.
Proceedings of the 3rd conference of environmental toxicology; October 25-27; Fairborn, OH; Paper No. 12;
AMRL-TR-72-120. Aerospace Medical Research Laboratory, Wright-Patterson Air Force Base, OH, pp. 199-208.
Available from the National Technical Information Service, Springfield, VA; AD773766.
Hazlelton Laboratories. (1983) 24-Month oncogenicity study of methylene chloride in mice [final report].
Performed by Hazlelton Laboratories America, Inc., Vienna, VA for the National Coffee Association, New York,
NY (Project No. 2112-106).
Hearne, FT; Pifer, JW. (1999) Mortality study of two overlapping cohorts of photographic film base manufacturing
employees exposed to methylene chloride. J Occup Environ Med 41:1154-1169.
Hearne, FT; Grose, F; Pifer, JW; et al. (1987) Methylene chloride mortality study: dose-response characterization
and animal model comparison. J Occup Med 29:217-228.
Hearne, FT; Pifer, JW; Grose, F. (1990) Absence of adverse mortality effects in workers exposed to methylene
chloride: an update. J Occup Med 32:234-240.
Hegi, ME; Soderkvist, P; Foley, JF; et al. (1993) Characterization of p53 mutations in methylene chloride-induced
lung tumors from B6C3F! mice. Carcinogenesis 14:803-810.
Heineman, EF; Cocco, P; Gomez, MR; et al. (1994) Occupational exposure to chlorinated aliphatic hydrocarbons
and risk of astrocytic brain cancer. Am J Ind Med 26:155-169.
Heineman, EF; Gomez, MR; Dosemeci, M; et al. (1996) Methylene chloride and brain cancer: interpreting a new
study in light of existing literature. Am J Ind Med 30:506-507.
Hellmold, H; Rylander, T; Magnusson, M; et al. (1998) Characterization of cytochrome P450 enzymes in human
breast tissue from reduction mammaplasties. J Clin Endocrinol Metab 83:886-895.
Heppel, LA; Neal, PA. (1944) Toxicology of dichloromethane (methylene chloride). II. Its effect upon running
activity in the male rat. J Ind Hyg Toxicol 26:17-21.
Heppel, LA; Neal, PA; Perrin, TL; et al. (1944) Toxicology of dichloromethane (methylene chloride). J Ind Hyg
Toxicol 26:8-16.
368 DRAFT - DO NOT CITE OR QUOTE
-------
Herr, DW; Boyes, WK. (1997) A comparison of the acute neuroactive effects of dichloromethane,
1,3-dichloropropane, and 1,2-dichlorobenzene on rat flash evokes potentials (FEPs). Fundam Appl Toxicol 35:31-
48.
Hines, RN. (2007) Ontogeny of human hepatic cytochromes P450. J BiochemMol Toxicol 21:169-175.
Holbrook, MT. (2003) Methylene chloride. In: Kirk-Othmer encyclopedia of chemical technology. New York, NY:
John Wiley & Sons. Available online at http://www.mrw.interscience.wiley.com/kirk/kirk_search_fs.html (accessed
March 9, 2010). (subscription required).
Horvath, AL. (1982) Halogenated hydrocarbons: solubility-miscibility with water. New York, NY: Marcel Dekker,
Inc., p. 543.
Hu, Y; Kabbler, SL; Tennant, AH et al. (2006) Induction of DNA-protein crosslinks by dichloromethane in a V79
cell line transfected with the murine glutathione-S-transferase theta 1 gene. Mutation Res 607:231-239.
Hughes, NJ; Tracey, JA. (1993) A case of methylene chloride (nitromors) poisoning, effects on
carboxyhaemoglobin levels. Hum Exp Toxicol 12:159-160.
IARC (International Agency for Research on Cancer). (1999) Dichloromethane. IARC monographs on the
evaluation of carcinogenic risk of chemicals to humans. Vol. 71. Re-evaluation of some organic chemicals,
hydrazine and hydrogen peroxide. Lyon, France: International Agency for Research on Cancer, pp. 251-315.
Available online at http://inchem.org/documents/iarc/vol71/004-dichloromethane.html (accessed March 9, 2010).
Infante-Rivard, C; Siemiatycki, J; Lakhani, R; et al. (2005) Maternal exposure to occupational solvents and
childhood leukemia. Environ Health Perspect 113:787-792.
Ingelman-Sundberg, M. (2004) Human drug metabolising cytochrome P450 enzymes: properties and
polymorphisms. Naunyn-Schmiedeberg's Arch Pharmacol 369:89-104.
Jacubovich, RM; Landau, D; Dayan, YB; et al. (2005) Facial nerve palsy after acute exposure to dichloromethane.
Am J Ind Med 48:389-392.
Johnsrud, EK; Koukouritaki, SB; Divakaran, K; et al. (2003) Human hepatic CYP2E1 expression during
development. J Pharmacol Exp Ther 307:402-407.
Jongen, WMF; Alink, GM; Koeman, JH. (1978) Mutagenic effect of dichloromethane on Salmonella typhimurium.
Mutat Res 56:245-248.
Jongen, WMF; Lohman, PHM; Kottenhagen, MJ; et al. (1981) Mutagenicity testing of dichloromethane in short-
term mammalian test systems. Mutat Res 81:203-213.
Jongen, WMF; Harmsen, EGM; Alink, GM; et al. (1982) The effect of glutathione conjugation and microsomal
oxidation on the mutagenicity of dichloromethane inS. typhimurium. Mutat Res 95:183-189.
Jonsson, F; Johanson, G. (2001) A Bayesian analysis of the influence of GST-T1 polymorphism on the cancer risk
estimate for dichloromethane. Toxicol Appl Pharmacol 174:99-112.
Jonsson, F; Johanson, G. (2003) The Bayesian population approach to physiological toxicokinetic-toxicodynamic
models—an example using the MCSim software. Toxicol Lett 138:143-150.
Jonsson, F; Bois, F; Johanson, G. (2001) Physiologically based pharmacokinetic modeling of inhalation exposure of
humans to dichloromethane during moderate to heavy exercise. Toxicol Sci 59:209-218.
Juronen, E; Tasa, G; Uusktila, M; et al. (1996) Purification, characterization and tissue distribution of human class
theta glutathione S-transferase Tl-1. Biochem Mol Biol Int 39:21-9
369 DRAFT - DO NOT CITE OR QUOTE
-------
Kanada, M; Miyagawa, M; Sato, M; et al. (1994) Neurochemical profile of effects of 28 neurotoxic chemicals on the
central nervous system in rats (1). Effects of oral administration on brain contents of biogenic amines and
metabolites. Ind Health 32:145-164.
Kanno, J; Foley, JF; Kari, F; et al. (1993) Effect of methylene chloride inhalation of replicative DNA synthesis in
the lungs of female B6C3FJ mice. Environ Health Perspect 101(Suppl. 5):271-276.
Kari, FW; Foley, JF; Seilkop, SK; et al. (1993) Effect of varying exposure regimens on methylene chloride-induced
lung and liver tumors in female B6C3Fi mice. Carcinogenesis 14:819-826.
Karlsson, JE; Rosengren, LE; Kjellstrand, P; et al. (1987) Effects of low-dose inhalation of three chlorinated
aliphatic organic solvents on deoxyribonucleic acid in gerbil brain. Scand J Work Environ Health 13:453-458.
Kayser, MF; Vuilleumier, S. (2001) Dehalogenation of dichloromethane by die hloro me thane
dehalogenase/glutathione s-transferase leads to formation of DNA adducts. J Bacteriol 183:5209-5212.
Kelly, M. (1988) Case reports of individuals with oligospermia and methylene chloride exposures. Reprod Toxicol
2:13-17.
Kernan, GJ; Ji, BT; Dosemeci, M; et al. (1999) Occupational risk factors for pancreatic cancer: a case-control study
based on death certificates from 24 US states. Am J Ind Med 36:260-270.
Kim, SK; Kim YC. (1996) Effect of a single administration of benzene, toluene or/w-xylene on
carboxyhaemoglobin elevation and metabolism of dichloromethane in rats. J Appl Toxicol 16:437-444.
Kim, RB; O'Shea, D; Wilkinson, GR. (1995) Interindividual variability of chlorzoxazone 6-hydroxylation in men
and women and its relationship to CYP2E1 genetic polymorphisms. Clin Pharmacol Ther 57:645-655.
Kim, NY; Park, SW; Suh, JK. (1996) Two fatal cases of dichloromethane or chloroform poisoning. J Forensic Sci
41:527-529.
*King-Herbert, A; Thayer, K. (2006) NTP workshop: animal models for the NTP rodent cancer bioassay: stock and
strains - should we switch? Toxicol Pathol 34:1533-1601.
Kirschman, JC; Brown, NM; Coots, RH; et al. (1986) Review of investigations of dichloromethane metabolism and
subchronic oral toxicity as the basis for the design of chronic oral studies in rats and mice. Food Chem Toxicol
24(9):943-949.
Kitchin, KT; Brown, JL. (1989) Biochemical effects of three carcinogenic chlorinated methanes in rat liver. Teratog
Carcinog Mutagen 9:61-69.
Kjellstrand, P; Holmquist, B; Jonsson, I; et al. (1985) Effects of organic solvents on motor activity in mice. Toxicol
35:35^6.
Kolodner, K; Cameron, L; Gittlesohn, A; et al. (1990) Morbidity study of occupational exposure to methylene
chloride using a computerized surveillance system. Center of Occupational Health and Environmental Health, Johns
Hopkins School of Hygiene and Public Health, Baltimore, MD. Submitted to TSCA under Section 8D; EPA
Document No. 86-900000421; NTIS No. OTS0522984.
Korzekwa, KR; Krishnamachary, N; Shou, M; et al. (1998) Evaluation of atypical cytochrome P450 kinetics with
two-substrate models: evidence that multiple substrates can simultaneously bind to cytochrome P450 active sites.
Biochemistry 37:4137-4147.
Kramers, PGN; Mout, HCA; Bissumbhar, B; et al. (1991) Inhalation exposure in Drosophila mutagenesis assays:
experiments with aliphatic halogenated hydrocarbons, with emphasis on the genetic activity profile of 1,2-
dichloroethane. Mutat Res 252:17-33.
*Lammer, EJ; Shaw, GM; lovannisci, DM; Finnell, RH. (2005) Maternal smoking, genetic variation of glutathione
s-transferases, and risk for orofacial clefts. Epidemiology 16:698-701.
370 DRAFT - DO NOT CITE OR QUOTE
-------
Landi, S; Naccarati, A; Ross, MK; et al. (2003) Induction of DNA strand breaks by trihalomethanes in primary
human lung epithelial cells. Mutat Res 538:41-50.
Landry, TD; Ramsey, JC; McKenna, MJ. (1983) Pulmonary physiology and inhalation dosimetry in rats:
development of a method and two examples. Toxicol Appl Pharmacol 71:72-83.
Lanes, SF; Cohen, A; Rothman, KJ; et al. (1990) Mortality of cellulose fiber production workers. Scand J Work
EnvironHealth 16:247-251.
Lanes, SF; Rothman, KJ; Dreyer, NA; et al. (1993) Mortality update of cellulose fiber production workers. Scand J
Work Environ Health 19:426-428.
Lash, AA; Becker, CE; So, Y; et al. (1991) Neurotoxic effects of methylene chloride: are they long lasting in
humans? BrJIndMed 48:418-426.
Lefevre, PA; Ashby, J. (1989) Evaluation of dichloromethane as aninducerof DNA synthesis in the B6C3F! mouse
liver. Carcinogenesis 10:1067-1072.
Lehmann, L; Wagner, J. (2008) Gene expression of 17beta-estradiol-metabolizing isozymes: comparison of normal
human mammary gland to normal human liver and to cultured human breast adenocarcinoma cells. Adv Exp Med
Biol 617:617-624.
Lehnebach, A; Kuhn, C; Pankow, D. (1995) Dichloromethane as an indicator of cytochrome c oxidase in different
tissues of rats. Arch Toxicol 69:180-184.
Leighton, DT, Jr; Calo, JM. (1981) Distribution coefficients of chlorinated hydrocarbons in dilute air-water systems
for groundwater contamination applications. J Chem Eng 26:3 82-3 85.
Leikin, JB; Kaufman, D; Lipscomb, JW; et al. (1990) Methylene chloride: report of five exposures and two deaths.
Am J Emerg Med 8:534-537.
Leitao, MM, Jr; White, P; Cracchiolo, B. (2008) Cervical cancer in patients infected with the human
immunodeficiency virus. Cancer 112:2683-2689.
Leuschner, F; Neumann, B; Htibscher, F. (1984) Report on subacute lexicological studies with dichloromethane in
rats and dogs by inhalation. Arzneim Forsch/Drug Res 34:1772-1774.
Lewis, RJ, Sr; ed. (1997) Hawley's condensed chemical dictionary. 13th edition. New York, NY: John Wiley &
Sons, Inc., p. 736.
Lide, DR; ed. (2000) CRC handbook of chemistry and physics. 81st edition. Boca Raton, FL: CRC Press, pp. 3-206.
*Lindstedt, SL; Schaeffer, PJ. (2002) Use of allometry in predicting anatomical and physiological parameters of
mammals. Laboratory Animals 36:1-19.
Lipscomb, JC; Garrett, CM; Snawder, JE. (1997) Cytochrome P450-dependent metabolism of trichloroethylene:
interindividual differences in humans. Toxicol Appl Pharmacol 142:311-318.
Lipscomb, JC; Teuschler, LK; Swartout, J; et al. (2003) The impact of cytochrome P450 2El-dependent metabolic
variance on a risk-relevant pharmacokinetic outcome in humans. Risk Anal 23:1221-1238.
Lorenz, J; Glatt, HR; Fleischmann, R; et al. (1984) Drug metabolism in man and its relationship to that in three
rodent species: monooxygenase, epoxide hydrolase, and glutathione S-transferase activities in subcellular fractions
of lung and liver. Biochem Med 32:43-56
Lowenfels, AB; Maisonneuve, P. (2005) Risk factors for pancreatic cancer. J Cell Biochem 95:649-656.
371 DRAFT - DO NOT CITE OR QUOTE
-------
Lucas, D; Ferrara, R; Gonzalez, E; et al. (1999) Chlorzoxazone, a selective probe for phenotyping CYP2E1 in
humans. Pharmacogenetics 9:377-388.
Lucas, D; Ferrara, R; Gonzales, E; et al. (2001) Cytochrome CYP2E1 phenotyping and genotyping in the evaluation
of health risks from exposure to polluted environments. Toxicol Lett 124:71-81.
Madle, S; Dean, SW; Andrae, U; et al. (1994) Recommendations forthe performance of UDS tests in vitro and in
vivo. Mutat Res 312:263-285.
Mahmud, M; Kales, SN. (1999) Methylene chloride poisoning in a cabinet worker. Environ Health Perspect
107:769-772.
Mainwaring, GW; Williams, SM; Foster, JR; et al. (1996) The distribution of Theta-class glutathione S-transferase
in the liver and lung of mouse, rat and human. Biochem J 318:297-303.
Maltoni, C; Cotti, G; Perino, G. (1988) Long-term carcinogenicity bioassays on methylene chloride administered by
ingestion to Sprague-Dawley rats and Swiss mice and by inhalation to Sprague-Dawley rats. Ann NY Acad Sci
534:352-366.
Manno, M; Rugge, M; Cocheo, V. (1992) Double fatal inhalation of dichloromethane. Hum Exp Toxicol 11:540-
545.
Marino, DJ; Clewell, HJ; Gentry, PR; et al. (2006) Revised assessment of cancer risk to dichloromethane: part I
Bayesian PBPK and dose-response modeling in mice. Regul Toxicol Pharmacol 45:44-54.
Maronpot, RR; Devereux, TR; Hegi, M; et al. (1995) Hepatic and pulmonary carcinogenicity of methylene chloride
in mice: a search for mechanisms. Toxicology 102:73-81.
Marriott, HM; Dockrell, DH. (2007) The role of the macrophage in lung disease mediated by bacteria. Exp Lung
Res 33:493-505.
Marsch, GA; Mundkowski, R; Morris, BKJ; et al. (2001) Characterization of nucleoside and DNA adducts formed
by s-(l-acetoxymethyl)glutathione and implications for dihalomethane-glutathione conjugates. Chem Res Toxicol
14:600-608.
Marsch, GA; Botta, S; Martin, MV; et al. (2004) Formation and mass spectrometric analysis of DNA and nucleoside
adducts by S-(l-acetoxymethyl)glutathione and by glutathione S-transferase-mediated activation of dihalomethanes.
Chem Res Toxicol 17:45-54.
Mathews, JM; Raymer, JH; Etheridge, AS; et al. (1997) Do endogenous volatile organic chemicals measured in
breath reflect and maintain CYP2E1 levels in vivo? Toxicol Appl Pharmacol 146:255-260.
Mattsson, JL; Albee, RR; Eisenbrandt, DL. (1990) Neurotoxicologic evaluation of rats after 13 weeks of inhalation
exposure to dichloromethane or carbon monoxide. Pharmacol Biochem Behav 36:671-681.
*McCarver, DG; Hines, RN. (2002) The ontogeny of human drug-metabolizing enzymes: Phase II conjugation
enzymes and regulatory mechanisms. J Pharmacol Exp Therap 300:361-366.
McKenna, MJ; Zempel, JA. (1981) The dose-dependent metabolism of [14C] methylene chloride following oral
administration to rats. Food Cosmet Toxicol 19:73-78.
McKenna, MJ; Zempel, JA; Braun, WH. (1982) The pharmacokinetics of inhaled methylene chloride in rats.
Toxicol Appl Pharmacol 65:1-10.
Mendoza-Cantu, A; Castorena-Torres, F; Bermudez, M; et al. (2004) Genotype and allele frequencies of
polymorphic cytochromes P450 CYP1A2 and CYP2E1 in Mexicans. Cell Biochem Funct 22(l):29-34.
Mennear, JH; McConnell, EE; Huff, JE; et al. (1988) Inhalation and carcinogenesis studies of methylene chloride
(dichloromethane) inF344/nrats andB6C3Fl mice. Ann NY Acad Sci 534:343-351.
372 DRAFT - DO NOT CITE OR QUOTE
-------
Meyer, DJ; Coles, B; Pemble, SE; et al. (1991) Theta, a new class of glutathione transferases purified from rat and
man. Biochem 1274:409^14.
Miksys, S; Tyndale, RF. (2004) The unique regulation of brain cytochrome P450 2 (CYP2) family enzymes by
drugs and genetics. Drug Metab Rev 36(2):313-333.
*Miligi, L; Costantini, AS; Benvenuti, A; et al. (2006) Occupational exposure to solvents and the risk of
lymphomas. Epidemiology 17:552-561.
Mirsalis, JC; Tyson, CK; Sleinmelz, KL; et al. (1989) Measurement of unscheduled DNA synthesis and s-phase
synthesis in rodent hepatocytes following in vivo treatment: testing of 24 compounds. Environ Mol Mutagen
14:155-164.
Moser, VC; Cheek, BM; MacPhail, RC. (1995) A multidisciplinary approach to lexicological screening: III.
Neurobehavioral toxicity. J Toxicol Environ Health 45:173-210.
Narotsky, MG; Kavlock, RJ. (1995) A multidisciplinary approach to lexicological screening: II. Developmental
loxicily. J Toxicol Environ Heallh 45:145-171.
Nelson, HH; Wiencke, JK; Chrisliani, DC; el al. (1995) Ethnic differences in the prevalence of the homozygous
deleted genotype of glulalhione S-lransferase Ihela. Carcinogenesis 16:1243-1245.
NIOSH (National Instilule of Occupational Safely and Heallh). (1986) Melhylene chloride. Currenl intelligence
bulletin 46. National Institute of Occupational Safely and Heallh, Centers for Disease Control and Prevention, U.S.
Departmenl of Heallh and Human Services, Cincinnati, OH; DHHS (NIOSH) Publication No. 86-114. Available
online al: http://www.cdc.gov/niosh/86114_46.hlml (accessed June 29, 2006).
Nishimura, M; Yaguli, H; Yoshilsugu, H. (2003) Tissue dislribution of mRNA expression of human cytochrome
P450 isoforms assessed by high-sensitivity real-time reverse Iranscription PCR. Yakugaku Zasshi 123(5):369-375.
Nilschke, KD; Burek, JD; Bell, TJ; el al. (1988a) Melhylene chloride: a 2-year inhalation loxicily and oncogenicity
study in rals. Fundam Appl Toxicol 11:48-59.
Nilschke, KD; Eisenbrandl, DL; Lomax, LG; el al. (1988b) Melhylene chloride: two-generation inhalation
reproductive study in rals. Fundam Appl Toxicol 11:60-67.
NLM (National Library of Medicine). (2003) Dichloromelhane. HSDB (Hazardous Subslances Dala Bank).
National Library of Medicine, National Instilules of Heallh, U.S. Departmenl of Heallh and Human Services,
Belhesda, Maryland. Available online al hltp://loxnel.nlm.nih.gov/ (accessed March 9, 2010).
Norman, WC, III; Boggs, P. (1996) Flawed estimates of melhylene chloride exposures. Am J Ind Med 30:504-509.
NRC (National Research Council). (1983) Risk assessmenl in the federal government: managing the process.
Washington, DC: National Academy Press.
NRC. (1994) Science and judgmenl in risk assessmenl. Washington, DC: National Academy Press.
NTP (National Toxicology Program). (1986) Toxicology and carcinogenesis studies of dichloromelhane (melhylene
chloride) (CAS No. 75-09-2) inF344/N rals and B6C3F1 mice (inhalation studies). Public Heallh Service, U.S.
Departmenl of Heallh and Human Services; NTP TR 306. Available from Ihe National Instilule of Environmental
Heallh Sciences, Research Triangle Park, NC. Available online al
hltp://ntp.niehs.nih.gov/ntp/htdocs/LT_rpls/lr306.pdf (accessed March 9, 2010).
Oda, Y; Yamazaki, H; Thier, R; el al. (1996) A new Salmonella typhimurium NM5004 slrain expressing ral
glulalhione S-lransferase 5-5: use in detection of genoloxicity of dihaloalkanes using an SOS/umu lesl system.
Carcinogenesis 17:297-302.
373 DRAFT - DO NOT CITE OR QUOTE
-------
Ognenovski, VM; Marder, W; Somers EC; et al. (2004) Increased incidence of cervical intraepithelial neoplasia in
women with systemic lupus erythematosus treated with intravenous cyclophosphamide. J Rheumatol 31:1763-
1767.
Oh, SJ; Kim, SK; Kim, YC. (2002) Role of glutathione in metabolic degradation of dichloromethane in rats.
Toxicol Lett 129:107-114.
Ojajarvi, A; Partanen, T; Ahlbom, A; et al. (2001) Risk of pancreatic cancer in workers exposed to chlorinated
hydrocarbon solvents and related compounds: a meta-analysis. Am J Epidemiol 153:841-850.
O'Neil, MJ; Smith, A; Heckelman, PE; et al. (2001) Methylene chloride. In: The Merck index: an encyclopedia of
chemicals, drugs, and biologicals. Whitehouse Station, NJ: Merck & Co., Inc., p. 1082.
Osterman-Golkar, S; Hussain, S; Walles, S; et al. (1983) Chemical reactivity and mutagenicity of some
dihalomethanes. ChemBiolInteract46:121-130.
OSHA (Occupational Safety and Health Administration). (1997) Occupational exposure to methylene chloride.
Federal Register 62:1494-1611.
Ott, MG; Skory, LK; Holder, BB; et al. (1983a) Health evaluation of employees occupationally exposed to
7methylene chloride: general study design and environmental considerations. Scand J Work Environ Health
9(Suppl. l):l-7.
Ott, MG; Skory, LK; Holder, BB; et al. (1983b) Health evaluation of employees occupationally exposed to
methylene chloride: mortality. Scand J Work Environ Health 9(Suppl. 1):8-16.
Ott, MG; Skory, LK; Holder, BB; et al. (1983c) Health evaluation of employees occupationally exposed to
methylene chloride: clinical laboratory evaluation. Scand J Work Environ Health 9(Suppl. 1): 17-25.
Ott, MG; Skory, LK; Holder, BB; et al. (1983d) Health evaluation of employees occupationally exposed to
methylene chloride: twenty-four hour electrocardiographic monitoring. Scand J Work Environ Health 9(Suppl.
1):26-30.
Ott, MG; Skory, LK; Holder, BB; et al. (1983e) Health evaluation of employees occupationally exposed to
methylene chloride: metabolism data and oxygen half-saturation pressures. Scand J Work Environ Health 9(Suppl.
l):31-38.
Ott, MG; Carlo, GL; Steinberg, S; et al. (1985) Mortality among employees engaged in chemical manufacturing and
related activities. Am J Epidemiol 122:311-322.
Pankow, D. (1988) Enhancement of dichloromethane-induced carboxyhemoglobinemia by isoniazid pretreatment.
Biomed Biochem Act 3:293-295.
Pankow, D; Hoffmann, P. (1989) Dichloromethane metabolism to carbon monoxide can be induced by isoniazid,
acetone and fasting. Arch Toxicol Suppl 13:302-303.
Pankow, D; Jagielki, S. (1993) Effect of methanol or modifications of the hepatic glutathione concentration on the
metabolism of dichloromethane to carbon monoxide in rats. HumExp Toxicol 12:227-231.
Pankow, D; Kretschmer, S; Weise, M. (1991a) Effect of pyrazole on dichloromethane metabolism to carbon
monoxide. Arch Toxicol Suppl 14:246-248.
Pankow, D; Matschiner, F; Weigmann, H. (1991b) Influence of aromatic hydrocarbons on the metabolism of
dichloromethane to carbon monoxide in rats. Toxicol 68:89-100.
Pegram, RA; Andersen, ME; Warren, SH; et al. (1997) Glutathione S-transferase-mediated mutagenicity of
trihalomethanes in Salmonella typhimurium: contrasting results with bromodichloromethane and chloroform.
Toxicol Appl Pharmacol 144:183-188.
374 DRAFT - DO NOT CITE OR QUOTE
-------
Perocco, P; Prodi, G. (1981) DNA damage by haloalkanes in human lymphocytes cultured in vitro. Cancer Lett
13:213-218.
Portier, K; Tolson, JK; Roberts, SM. (2007) Body weight distributions for risk assessment. Risk Anal 27(1): 11-26.
Putz, VR; JohnsonBL; Setzer, JV. (1979) A comparative study of the effects of carbon monoxide and methylene
chloride on human performance. J Environ Pathol Toxicol 2:97-112.
Quondamatteo, F; Schulz, TG; Bunzel, N; et al. (1998) Immunohistochemical localization of glutathione
S-transferase-Tl in murine kidney, liver, and lung. Histochem Cell Biol 110:417^4-23.
*Radican, L; Blair, A; Stewart, P; Wartenberg, D. (2008) Mortality of aircraft maintenance workers exposed to
trichloroethylene and other hydrocarbons and chemicals: extended follow-up. J Occup Environ Med 50:1306-1319.
*Raijmakers, MT; Steegers, EA; Peters, WH. (2001) Glutathione S-transferases andthiol concentrations in
embryonic and early fetal tissues. Hum Reprod 16:2445-2450.
Raimondi, S; Paracchini, V; Autrup, H; et al. (2006) Human genome epidemiology (HuGE) review. Meta- and
pooled analysis of GST-T1 and lung cancer: a HuGE-GSEC review. Am J Epidemiol 164:1027-1042.
Raje, R; Basso, M; Tolen, T; et al. (1988) Evaluation of in vivo mutagenicity of low-dose methylene chloride in
mice. J Am Coll Toxicol 7:699-703.
Ramsey, JR; Andersen, ME. (1984) A physiologically-based description of the inhalation pharmacokinetics of
styrene in rats and humans. Toxicol Appl Pharmacol 73:159-175.
Raphael, M; Nadiras, P; Flacke-Vordos, N. (2002) Acute methylene chloride intoxication—a case report on
domestic poisoning. Eur J Emerg Med 9:57-59.
Rebert, CS; Matteucci, MJ; Pryor, GT. (1989) Acute effects of inhaled dichloromethane on the EEG and sensory-
evoked potentials of Fischer-344 rats. Pharmacol Biochem Behav 34:619-629.
Reitz, RH. (1991) Estimating the risk of human cancer associated with exposure to methylene chloride. Ann 1st
Super Sanita 27:609-614.
Reitz, RH; McDougal, JN; Himmelstein, MW; et al. (1988a) In vitro metabolism of methylene chloride in human
and animal tissues: use in physiologically based pharmacokinetic models. Toxicol Appl Pharmacol 97:230-246.
Reitz, RH; Mendrala, AL; Park, CN; et al. (1988b) Incorporation of in vitro enzyme data into the physiologically-
based pharmacokinetic (PB-PK) model for methylene chloride: implications for risk assessment. Toxicol Lett
43:97-116.
*Reitz, RH; Smith, FA; Andersen, ME; et al. (1988c) Use of physiological pharmacokinetics in cancer risk
assessments: A study of methylene chloride, in "The Risk Assessment of Environmental and Human Health
Hazards," D.J. Paustenbach (ed.) John Wiley & Sons, Inc. New York, NY, pp. 238-267.
Reitz, RH; Mendrala, AL; Guengerich, FP. (1989) In vitro metabolism of methylene chloride in human and animal
tissues: use in physiologically based pharmacokinetic models. Toxicol Appl Pharmacol 97:230-246.
Reitz, RH; Hays, SM; Gargas, ML. (1997) Addressing priority data needs for methylene chloride with
physiologically based pharmacokinetic modeling. Prepared for the Agency for Toxic Substances and Disease
Registry, Public Health Service, U.S. Department of Health and Human Services, Atlanta, GA, on behalf of the
Halogenated Solvents Industry Alliance (HSIA), Arlington, VA.
Rhomberg, L. (1995) Use of quantitative modeling in methylene chloride risk assessment. Toxicology 102:95-114.
Riley, EC; Fassett, DW; Sutton, WL. (1966) Methylene chloride vapor in expired air of human subjects. Am Ind
HygAssoc 127:341-348.
375 DRAFT - DO NOT CITE OR QUOTE
-------
Rioux, JP; Myers, RA. (1988) Methylene chloride poisoning: a paradigmatic review. J Emerg Med 6:227-238.
Rodkey, FL; Collison, HA. (1977) Effect of dihalogenated methanes on the in vivo production of carbon monoxide
and methane by rats. Toxicol Appl Pharmacol 40:39-47.
Rodriguez-Arnaiz, R. (1998) Biotransformation of several structurally related 2B compounds to reactive metabolites
in the somatic w/w+ assay ofDrosophila melanogaster. Environ Mol Mutagen 31(4):390-401.
Roldan-Arjona, T; Pueyo, C. (1993) Mutagenic and lethal effects of halogenated methanes in the Aratestof
Salmonella typhimurium: quantitative relationship with chemical reactivity. Mutagenesis 8(2): 127-131.
Rosengren, LE; Kjellstrand, P; Aurell, A; et al. (1986) Irreversible effects of dichloromethane on the brain after long
term exposure: a quantitative study of DNA and the glial cell marker proteins S-100 and GFA. Br J Ind Med
43:291-299.
Roth, RP; Drew, RT; Lo, RJ; et al. (1975) Dichloromethane inhalation, carboxyhemoglobin concentrations, and
drug metabolizing enzymes in rabbits. Toxicol Appl Pharmacol 33:427^37.
Rothman, KJ; Greenland, S. (1998) Precision and validity in epidemiologic studies. In: Modern epidemiology. 2nd
edition. Philadelphia, PA: Lippincott-Raven Publishers, pp. 115-134.
Sakai, T; Morita, Y; Wakui, C. (2002) Biological monitoring of workers exposed to dichloromethane using head-
space gas chromatography. J Chromat B 778:245-250.
*Salmon, AG; Roth, LA. (2010). Cancer risk based on an individual tumor type or summing of tumors. In: "Cancer
Risk Assessment: Chemical Carcinogenesis from Biology to Standards Quantification": Ching-Hung Hsu and Todd
Stedeford, Eds. John Wiley & Sons, Inc., Hoboken, NJ, 2010.
Sasaki, YF; Saga, A; Akasaka, M; et al. (1998) Detection of in vivo genotoxicity of haloalkanes and haloalkenes
carcinogenic to rodents by the alkaline single cell gel electrophoresis (comet) assay in multiple mouse organs.
Mutat Res 419:13-20.
Savolainen, H; Pfaffli, P; Tengen M; et al. (1977) Biochemical and behavioural effects of inhalation exposure to
tetrachloroethylene and dichloromethane. J Neuropathol Exp Neurol 36:941-949.
Savolainen, H; Kurppa, K; Pfaffli, P; etal. (1981) Dose-related effects of dichloromethane in rat brain in short-term
inhalations exposure. Chem-Biol Interact 34:315-322.
*Schmucker, DL. (2001) Liver function and phase I drug metabolism in the elderly. A paradox. Drug Aging 18:
837-851.
Schwetz, BA; Leong, BKJ; Gehring, PJ. (1975) The effect of maternally inhaled trichloroethylene,
perchloroethylene, methyl chloroform, and methylene chloride on embryonal and fetal development in mice and
rats. Toxicol Appl Pharmacol 32:84-96.
Searles, J; McPhail, HA. (1949) Methylene chloride, CH2C12. In: Kirk, RE; Othmer, DF; eds. Encyclopedia of
chemical technology. New York, NY: Interscience Encyclopedia, Inc., pp. 747-751.
*Seidler, A; Mohner, M; Berger, J; et al. (2007) Solvent exposure and malignant lymphoma: a population-based
case-control study in Germany. J Occup Med Toxicol 2:2.
*Selgrade, MK; Gimour, MI. (2010) Suppression of pulmonary host defenses and enhanced susceptibility to
respiratory bacterial infection in mice following inhalation exposure to trichloroethylene and chloroform. J
Immunotoxicol 7:350-356.
Selgrade, MK; Gilmour, MI. (2006) Immunotoxicology of inhaled compounds-assessing risks of local immune
suppression and hypersensitivity. J Toxicol Environ Health A 69:827-844.
376 DRAFT - DO NOT CITE OR QUOTE
-------
Scrota, DG; Thakur, AK; Ulland, BM; et al. (1986a) A two-year drinking water study of dichloromethane in
rodents. I. Rats. Food Chem Toxicol 24:951-958.
Serota, DG; Thakur, AK; Ulland, BM; et al. (1986b) A two-year drinking water study of dichloromethane in
rodents. II. Mice. Food Chem Toxicol 24:959-963.
Sheldon, T; Richardson, CR; Elliott, BM. (1987) Inactivity of methylene chloride in the mouse bone marrow
micronucleus assay. Mutagenesis 2:57-59.
Sherratt, PJ; Pulford, DJ; Harrison, DJ; et al. (1997) Evidence that human class Theta glutathione S-transferase
Tl-1 can catalyse the activation of dichloromethane, a liver and lung carcinogen in the mouse. Biochem J 326:837-
846.
Sherratt, PJ; Williams, S; Foster, J; et al. (2002) Direct comparison of the nature of mouse and human GST Tl-1 and
the implications on dichloromethane carcinogenicity. Toxicol Appl Pharmacol 179:89-97.
*Shi, M; Kristensen, C; Weinberg, CR; et al. (2007) Orofacial cleft risk is increased with maternal smoking and
specific detoxification-gene variants. Am J Hum Genet 80:76-90.
Shimada, T; Yamazaki, H; Mimura, M; et al. (1994) Interindividual variations in human liver cytochrome
P-450 enzymes involved in the oxidation of drugs, carcinogens and toxic chemicals: studies with liver microsomes
of 30 Japanese and 30 Caucasians. J Pharmacol Exp Ther 270:414-423.
Shusterman, D; Quinlan, P; Lowengart, R; et al. (1990) Methylene chloride intoxication in a furniture refinisher. A
comparison of exposure estimates utilizing workplace air sampling and blood carboxyhemoglobin measurements. J
OccupMed 32:451-454.
Sills, RC; Hailey, JR; Neal, J; et al. (1999) Examination of low-incidence brain tumor responses in F344 rats
following chemical exposures in National Toxicology Program carcinogenicity studies. Toxicol Pathol 27:589-599.
Simula, TP; Glancey, MJ; Wolf, CR. (1993) Human glutathione S-transferase-expressing.Sa/ffjo«e//a typhimurium
tester strains to study the activation/detoxification of mutagenic compounds: studies with halogenated compounds,
aromatic amines and aflatoxinBl. Carcinogenesis 14:1371-1376.
Soden, KJ. (1993) An evaluation of chronic methylene chloride exposure. J Occup Med 35:282-286.
Soden, KJ; Marras, G; Amsel, J. (1996) Carboxyhemoglobin levels in methylene chloride-exposed employees. J
Occup Environ Med 38:367-371.
Spirtas, R; Stewart, PA; Lee, JS; et al. (1991) Retrospective cohort mortality study of workers at an aircraft
maintenance facility. I. Epidemiological results. Br J Ind Med 48:515-530.
SRC (Syracuse Research Corporation). (1989) A review of in vitro test methodology for assessment of
hepatotoxicity with a view to application to chemical mixtures. Prepared for U.S. EPA, Environmental Criteria and
Assessment Office, U.S. Environmental Protection Agency; SRC TR-89-205.
Stephens, EA; Taylor, JA; Kaplan, N; et al. (1994) Ethnic variation in the CYP2E1 gene: polymorphism analysis of
695 African-Americans, European-Americans and Taiwanese. Pharmacogenetics 4:185-192.
Stewart, RD; Fisher, TN; Hosko, MJ; et al. (1972a) Experimental human exposure to methylene chloride. Arch
Environ Health 25:342-348.
Stewart, RD; Fisher, TN; Hosko, MJ; et al. (1972b) Carboxyhemoglobin elevation after exposure to
dichloromethane. Science 176:295-296.
Stewart, PA; Lee, JS; Marano, DE; et al. (1991) Retrospective cohort mortality study of workers at an aircraft
maintenance facility. II Exposures and their assessment. Br J Ind Med 48:531-537.
377 DRAFT - DO NOT CITE OR QUOTE
-------
Stott, WT; McKenna, MJ. (1984) The comparative absorption and excretion of chemical vapors by the upper, lower,
and intact respiratory tract of rats. Fundam Appl Toxicol 4:594-602.
Strange, RC; Howie, AF; Hume, R; et al. (1989) The development expression of alpha-, mu- and pi-class
glutathione S-transferases in human liver. Biochim Biophys Acta 993:186-190.
Sweeney, LM; Gargas, ML; Strother, DE; et al. (2003) Physiologically based pharmacokinetic model parameter
estimation and sensitivity and variability analyses for acrylonitrile disposition in humans. Toxicol Sci 71:27-40.
Sweeney, LM; Kirman, CR; Morgott, DA; et al. (2004) Estimation of interindividual variation in oxidative
metabolism of dichloromethane in human volunteers. ToxicolLett 154:201-216.
Takeshita, H; Mogi, K; Yasuda, T; et al. (2000) Postmortem absorption of dichloromethane: a case study and animal
experiments. Int J Legal Med 114:96-100.
Taskinen, H; Lindbohm, ML; Hemminki, K. (1986) Spontaneous abortions among women working in the
pharmaceutical industry. Br J Ind Med 43:199-205.
Tay, P; Tan, KT; Sam, CT. (1995) Fatal gassing due to methylene chloride—a case report. Singapore Med J
36:444-445.
Teschke, K; Olshan, AF; Daniels, JL; et al. (2002) Occupational exposure assessment in case-control studies:
opportunities for improvement. Occup Environ Med 59:575-593.
Thier, R; Taylor, JB; Pemble, SE; et al. (1993) Expression of mammalian glutathione S-transferase 5-5 in
Salmonella typhimurium TA1535 leads to base-pair mutations upon exposure to dihalomethanes. Proc Natl Acad
Sci USA 90:8576-8580.
Thier, R; Wiebel, FA; Hinkel, A; et al. (1998) Species differences in the glutathione transferase GST-T1-1 activity
towards the model substrates methyl chloride and dichloromethane in the liver and kidney. Arch Toxicol 72:622-
629.
Thilagar, AK; Kumaroo, V. (1983) Induction of chromosome damage by methylene chloride in CHO cells. Mutat
Res 116:361-367.
Thilagar, AK; Back, AM; Kirby, PE; et al. (1984) Evaluation of dichloromethane in short term in vitro genetic
toxicity assays. Environ Mutagen 6:418-419.
Thomas, AA; Pinkerton, MK; Warden, JA. (1972) Effects of low level dichloromethane exposure on the
spontaneous activity of mice. Proceedings of the 3rd conference of environmental toxicology; October 25-27;
Fairborn, OH; Paper No. 14; AMRL-TR-72-130. Aerospace Medical Research Laboratory, Wright-Patterson Air
Force Base, OH; pp. 223-238. Available from the National Technical Information Service, Springfield, VA;
AD773766.
Tomenson, JA; Bonner, SM; Heijne, CG; et al. (1997) Mortality of workers exposed to methylene chloride
employed at a plant producing cellulose triacetate film base. Occup Environ Med 54:470^176.
Treluyer, JM; Cheron, G; Sonnier M; et al. (1996) Cytochrome P-450 expression in sudden infant death syndrome.
Biochem Pharmacol 52:497-504.
Trueman, RW; Ashby, J. (1987) Lack of UDS activity in the livers of mice and rats exposed to dichloromethane.
Environ Mol Mutagen 10:189-195.
U.S. Coast Guard. (1999) Dichloromethane. CHRIS: hazardous chemical data. U.S. Coast Guard, Department of
Transportation, Washington, DC. Available online at http://www.chrismanual.com/fmdform.htm.
U.S. EPA (Environmental Protection Agency). (1986a) Guidelines for the health risk assessment of chemical
mixtures. Federal Register 51(185):34014-34025.
378 DRAFT - DO NOT CITE OR QUOTE
-------
U.S. EPA. (1986b) Guidelines for mutagenicity risk assessment. Federal Register 51(185):34006-34012. Available
online at http://www.epa.gov/iris/backgrd.html (accessed March 9, 2010).
U.S. EPA. (1987a) Update to the health assessment document and addendum for dichloromethane (methylene
chloride): pharmacokinetics, mechanism of action and epidemiology [review draft]. Office of Health and
Environmental Assessment, Office of Research and Development, Washington, DC; EPA/600/8-87/030A. Available
from the National Technical Information Service, Springfield, VA, PB87228565, and online at
http://nepis.epa.gov/Exe/ZyPURL.cgi?Dockey=30001GFH.txt (accessed March 9, 2010).
U.S. EPA. (1987b) Technical analysis of new methods and data regarding dichloro methane hazard assessments
[review draft]. Office of Health and Environmental Assessment, Office of Research and Development, Washington,
DC; EPA/600/8-87/029A. Available from the National Technical Information Service, Springfield, VA, PB87-
228557, and online at http://nepis.epa.gov/Exe/ZyPURL.cgi?Dockey=30001GAN.txt (accessed March 9, 2010).
U.S. EPA. (1988a) Recommendations for and documentation of biological values for use in risk assessment.
Environmental Criteria and Assessment Office, Office of Health and Environmental Assessment, Cincinnati, OH;
EPA/600/6-87/008. Available from the National Technical Information Service, Springfield, VA, PB88-179874/AS,
and online at http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=34855 (accessed March 9, 2010).
U.S. EPA. (1988b) The impact of pharmacokinetics on the risk assessment of dichloromethane. Office of Health
and Environmental Assessment, Office of Research and Development, Washington, DC; EPA/600/D-88/219.
Available from National Technical Information Service, Springfield, VA; PB89-173249.
U.S. EPA. (1991) Guidelines for developmental toxicity risk assessment. Federal Register 56(234):63798-63826.
Available online at http://www.epa.gov/iris/backgrd.html (accessed March 9, 2010).
*U.S. EPA (1992) Draft report: a cross-species scaling factor for carcinogen risk assessment based on equivalence
of mg/kg3/4/day. Federal Register 57(109):24152-24173.
U.S. EPA. (1994a) Interim policy for particle size and limit concentration issues in inhalation toxicity: notice of
availability. Federal Register 59(206):53799. Available online at http://www.epa.gov/EPA-
PEST/1994/October/Day-26/pr-ll.html (accessed March 9, 2010).
U.S. EPA. (1994b) Methods for derivation of inhalation reference concentrations and application of inhalation
dosimetry. Environmental Criteria and Assessment Office, Office of Health and Environmental Assessment,
Cincinnati, OH; EPA/600/8-90/066F. Available from the National Technical Information Service, Springfield, VA,
PB2000-500023, and online at http://cfpub.epa.gov/ncea/raf/recordisplay.cfm?deid=71993 (accessed March 9,
2010).
U.S. EPA. (1995) Use of the benchmark dose approach in health risk assessment. Risk Assessment Forum,
Washington, DC; EPA/630/R-94/007. Available from the National Technical Information Service, Springfield, VA,
PB95-213765, and online at http://cfpub.epa.gov/ncea/raf/raf_pubtitles.cfm?detype=document&excCol=archive
(accessed March 9, 2010).
U.S. EPA. (1996) Guidelines for reproductive toxicity risk assessment. Federal Register 61(212):56274-56322.
Available online at http://www.epa.gov/iris/backgrd.html (accessed March 9, 2010).
U.S. EPA. (1998) Guidelines for neurotoxicity risk assessment. Federal Register 63(93):26926-26954. Available
online at http://www.epa.gov/iris/backgrd.html (accessed March 9, 2010).
U.S. EPA. (2000a) Science policy council handbook: risk characterization. Office of Science Policy, Office of
Research and Development, Washington, DC; EPA/100-B-00-002. Available online at
http://www.epa.gov/OSA/spc/pdfs/prhandbk.pdf (accessed March 9, 2010).
U.S. EPA. (2000b) Benchmark dose technical guidance document [external review draft]. Risk Assessment Forum,
Washington, DC; EPA/630/R-00/001. Available online at http://www.epa.gov/iris/backgrd.html (accessed March 9,
2010).
379 DRAFT - DO NOT CITE OR QUOTE
-------
U.S. EPA. (2000c) Supplementary guidance for conducting health risk assessment of chemical mixtures. Risk
Assessment Forum, Washington, DC; EPA/630/R-00/002. Available online at
http://cfpub.epa.gov/ncea/raf/chem_mix.cfm (accessed March 9, 2010).
U.S. EPA. (2000d) Toxicological review of vinyl chloride. Integrated Risk Information System (IRIS), National
Center for Environmental Assessment, Washington, DC; EPA/635/R-00/004. Available online at
http://www.epa.gov/iris (accessed March 9, 2010).
U.S. EPA. (2002) A review of the reference dose concentration and reference concentration processes. Risk
Assessment Forum, Washington, DC; EPA/630/P-02/002F. Available online at
http://cfpub.epa.gov/ncea/raf/raf_pubtitles.cfm?detype=document&excCol=archive (accessed March 9, 2010).
U.S. EPA. (2005a) Guidelines for carcinogen risk assessment. Federal Register 70(66): 17765-18717. Available
online at http://www.epa.gov/cancerguidelines (accessed March 9, 2010).
U.S. EPA. (2005b) Supplemental guidance for assessing susceptibility from early-life exposure to carcinogens. Risk
Assessment Forum, Washington, DC; EPA/630/R-03/003F. Available online at
http://www.epa.gov/cancerguidelines (accessed March 9, 2010).
U.S. EPA. (2006a) Science policy council handbook: peer review. 3rd edition. Office of Science Policy, Office of
Research and Development, Washington, DC; EPA/100/B-06/002. Available online at
http://www.epa.gov/OSA/spc/2peerrev.htm (accessed March 9, 2010).
U.S. EPA. (2006b) A framework for assessing health risk of environmental exposures to children. National Center
for Environmental Assessment, Washington, DC; EPA/600/R-05/093F. Available online at
http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=158363 (accessed March 9, 2010).
*vanRooij, IA; Wegerif, MJ; Roelofs, HM; et al. (2001) Smoking, genetic polymorphisms inbiotransformation
enzymes, and nonsyndromic oral clefting: a gene-environment interaction. Epidemiology 12:502- 507.
Vieira, I; Sonnier, M; Cresteil, T. (1996) Developmental expression of CYP2E1 in human liver. Hypermethylation
control of gene expression during the neonatal period. Eur J Biochem 238:476-483.
*Wang, R; Zhang, Y; Lan, Q; et al. (2009) Occupational exposure to solvents and risk of non-Hodgkin lymphoma in
Connecticut women. Am J Epidemiol 169:176-185.
Warbrick, EV; Kilgour, JD; Dearman, RJ; et al. (2003) Inhalation exposure to methylene chloride does not induce
systemic immunotoxicity in rats. J Toxicol Environ Health A 66:1207-1219.
Warholm, M; Alexandrie, AK; Hogberg, J; et al. (1994) Polymorphic distribution of glutathione transferase activity
with methyl chloride in human blood. Pharmacogenetics 4:307-311.
Watanabe, K; Guengerich, FP. (2006) Limited reactivity of formyl chloride with glutathione and relevance to
metabolism and toxicity of dichloromethane. Chem Res Toxicol 19:1091-1096.
Watanabe, K; Liberman, RG; Skipper, PL; et al. (2007) Analysis of DNA adducts formed in vivo in rats and mice
from 1,2-dibromoethane, 1,2-dichloroethane, dibromomethane, and dichloromethane using HPLC/accelerator mass
spectrometry and relevance to risk estimates. Chem Res Toxicol 20:1594-600.
Weinstein, RS; Boyd, DD; Back, KC. (1972) Effects of continuous inhalation of dichloromethane in the mouse:
morphologic and functional observations. Toxicol Appl Pharmacol 23:660-679.
Wells, VE; Schrader, SM; McCamon, CS; et al. (1989) Letter to the editor. Reprod Toxicol 3:281-282.
Westbrook-Collins, B; Allen, JW; Sharief, Y; et al. (1990) Further evidence that dichloromethane does not induce
chromosome damage. J Appl Toxicol 10:79-81.
*Wiester, MJ; Winsett, DW; Richards, JH; et al. (2002) Partitioning of benzene in blood: Influence of hemoglobin
type in humans and animals. Environ Health Perspect 110:255-261.
380 DRAFT - DO NOT CITE OR QUOTE
-------
Winneke, G. (1974) Behavioral effects of methylene chloride and carbon monoxide as assessed by sensory and
psychomotor performance. In: Xintaras, C; Johnson, BL; de Groot, I; eds. Behavioral toxicology: early detection of
occupational hazards. Washington, DC: National Institute for Occupational Safety and Health, Center for Disease
Control, Public Health Service, U.S. Department of Health, Education and Welfare; pp. 130-144.
Wirkner, K; Damme, B; Peolchen, W; et al. (1997) Effect of long-term ethanol pretreatment on the metabolism of
dichloromethane to carbon monoxide in rats. Toxicol Appl Pharmacol 143:83-88.
Withey, JR; Karpinski, K. (1985) The fetal distribution of some aliphatic chlorinated hydrocarbons in the rat after
vapor phase exposure. Biol Res Pregnancy Perinatal 6:79-88.
Zarrabeitia, MT; Ortega, C; Alruzarra, E; et al. (2001) Accidental dichloromethane fatality: a case report. J Forensic
Sci 46:726-727
Zeiger; E. (1990) Mutagenicity of 42 chemicals in Salmonella. Environ Mol Mutagen 16:32-54.
Zielenska, M; Ahmed, A; Pienkowska, M; et al. (1993) Mutational specificities of environmental carcinogens in the
lad gene of Eschehchia coli. VI. Analysis of methylene chloride-induced mutational distribution in Uvr+ and
UvrB- strains. Carcinogenesis 14:789-794.
3 81 DRAFT - DO NOT CITE OR QUOTE
-------
APPENDIX A: SUMMARY OF EXTERNAL PEER REVIEW AND PUBLIC
COMMENTS AND DISPOSITION
The Toxicological Review of Dichloromethane (dated March 2010) has undergone a
formal external peer review performed by scientists in accordance with the U.S. EPA guidance
on peer review (2006a, 2000a). An external peer-review workshop was held September 23,
2010. The external peer reviewers were tasked with providing written answers to general
questions on the overall assessment and on chemical-specific questions in areas of scientific
controversy or uncertainty. A summary of significant comments made by the external reviewers
and the EPA's responses to these comments arranged by charge question follow. In many cases
the comments of the individual reviewers have been synthesized and paraphrased in development
of Appendix A. When the external peer reviewers commented on decisions and analyses in the
Toxicological Review under multiple charge questions, these comments were organized under
the most appropriate charge question. The EPA also received scientific comments from the
public. These comments and the EPA's responses are included in a separate section of this
appendix.
EXTERNAL PEER REVIEWER COMMENTS
The reviewers made several editorial suggestions to clarify specific portions of the text.
These changes were incorporated in the document as appropriate and are not discussed further.
General Charge Questions
Gl. Is the Toxicological Review logical, clear and concise? Has EPA clearly and
objectively represented and synthesized the scientific evidence for noncancer and cancer
hazard?
Comment: Most reviewers considered the Toxicological Review to be comprehensive, clear,
concise, and well written. One reviewer considered the presentation of material to be repetitive.
Other comments provided in response to General Charge Question 1 were repeated by the peer
reviewers in response to other charge questions, and are summarized and discussed under the
relevant question.
Response: EPA recognizes that some material is repeated, but has sought to limit these
situations to those that are necessary to allow individual sections (e.g., the RfD and the RfC
sections) to be read independently.
A-1 DRAFT - DO NOT CITE OR QUOTE
-------
G2. Please identify any additional studies that would make a significant impact on the
conclusions of the Toxicological Review.
Comment: One reviewer identified 15 references, most of which related to toxicokinetics and to
factors affecting variability in CYP2E1 expression. None of the references specifically
addressed dichloromethane.
Response: Relevant references provided by the reviewer were added to appropriate sections in
the Toxicological Review; however, citations to secondary sources of material already covered
by a primary source were not added.
A. PBPK Modeling
Ala. A rat PBPK model was used for calculating the internal dosimetry for the RfD and
RfC. EPA evaluated several versions of previously published rat PBPK models and
modified the Andersen et al. (1991) model for use in the reference value calculations. Does
the chosen model with EPA's modifications adequately represent the toxicokinetics? Was
the model applied properly? Are the model assumptions and parameters clearly presented
and scientifically supported? Are the uncertainties in the model structure appropriately
considered and discussed?
Comments: Three reviewers supported the chosen model for rat PBPK toxicokinetics, and noted
the clear presentation and discussion of the model assumptions, parameters, and uncertainties.
One of these reviewers suggested two additions to the PBPK modeling section of the
Toxicological Review: a table describing the changes from the rat model compared with the
model used in the previous assessment, and information regarding how CYP2E1 variability was
incorporated into the model (information that was included in Appendix C). Two reviewers did
not comment on this question because it was outside their area of expertise.
Response: A PBPK model for the rat was not used in the previous assessment. The RfD was
derived through application of UFs to a NOAEL for liver lesions observed in a 2-year oral
exposure study (i.e., the same study used in the current IRIS assessment); an RfC was not
developed in the previous assessment. Table 3-10 presents the parameters of the rat PBPK
model used in the current assessment, and Table C-l presents these parameters for all of the
models evaluated in the process of developing the rat PBPK model. A summary of EPA's
consideration of CYP2E1 variability is included in Section 3.5.2 (Probabilistic Human PBPK
Dichloromethane Model (David et al., 2006)).
A-2 DRAFT - DO NOT CITE OR QUOTE
-------
Comments: One review questioned the choices of models used in the derivation of the rat PBPK
model, noting that model A was developed using the flux via GST-pathway in lung tissue in the
mouse. This reviewer also suggested that the goal of the analysis should be the evaluation of a
published model to determine its usefulness for risk assessment; if it was deficient in some
aspect, than modification or alternative models could be evaluated. The comparison between
models would ideally be done using a statistical test, rather than the more commonly used
method of visual inspection of the fit of a model to various data sets.
Response: EPA agreed with these suggestions. The Toxicological Review was revised to start
with a model that is essentially identical to the published model of Andersen et al. (1991)
(formerly "Model B," changed to "Model A"), with the exception that CYP- and GST-mediated
metabolism were added to the lung compartment. This change was made to make the model
consistent with the mouse and human PBPK models and to correctly apportion metabolism
between the two tissues, while holding the total CYP and GST activity constant at the value
reported by Andersen et al. (1991). Also, a clear inadequacy of the Andersen et al. (1991) model
is that it does not have an oral exposure (GI tract) sub-model that is needed to describe dosimetry
for oral exposures. Andersen et al. appear to have only used inhalation, or perhaps inhalation
and intravenous (i.v.) data, to estimate their metabolic constants (Vmaxc and Km for the CYP
pathway, kfc for the GST pathway, and PI, the CO yield factor). While it would be possible to
keep these metabolic parameters unchanged and only fit an oral kinetic constant to appropriate
data, fitting the oral data was challenging and not entirely satisfactory, suggesting that
adjustments in the metabolic parameters should also be considered. The potential for improving
the fit of the Andersen et al. (1991) model by numerical optimization was tested by performing a
statistical comparison between the fit to the inhalation and i.v. data using the original parameters
versus numerical fits, and the latter were found to provide a significant improvement in the
goodness of fit. Therefore the metabolic parameters (Vmaxc, Km, kfc, and PI) along with an
absorption constant (ka) for uptake from the GI tract were globally fit to a larger data set that
included oral toxicokinetic data as well as the inhalation and i.v. data used for initial model
testing.
Comments: One reviewer asked about saturation of the CYP2E1 pathway in humans at 400 ppm
dichloromethane in the study by Ott et al. (1983e), and how this finding compares to that of the
rat based on McKenna and Zembel (1981), with inhibition of CYP2E1 pathway when
dichloromethane exposure was increased from 1 to 50 mg/kg in rats.
Response: These findings are not directly comparable. Ott et al. (1983e) examined the oxygen-
half-saturation pressure (Pso) for ©2 binding to hemoglobin, which is reduced due to COHb
resulting from dichloromethane exposure. The saturation described by McKenna and Zempel
A-3 DRAFT - DO NOT CITE OR QUOTE
-------
(1981) is that of CYP-metabolism of dichloromethane to CO under precisely-controlled
laboratory conditions. The level of CO formed, hence impact on O2-Hb saturation, should
saturate as dichloromethane exposure increases, due to the saturation of CYP metabolism. But
the data from Ott et al. are summary statistics from groups of workers exposed to fairly broad
dichloromethane ranges (0, <100, 100-299, and >300 ppm). Given that in non-smoking men the
reduction was only from PSO = 26.4 ± 0.8 to 22.7 ± 1 mm Hg between 0 ppm and > 300 ppm
dichloromethane (501 ppm time-weighted average), it would be difficult to estimate the
nonlinearity due to CYP-metabolism saturation in those data (and hence the concentration at
which half-saturation is reached), or to compare it to that observed in the McKenna and Zempel
rat studies. In addition, the human exposures of Ott et al. were by inhalation and those of
McKenna and Zembel were oral gavage. Since gavage exposures give a brief but high body
burden for the same total dose compared to an exposure regimen that spreads out the exposure
over a period of hours (inhalation specifically), even at the same total dose in the same species
one may see saturation from the oral exposure but not in the time-distributed exposure.
Comments: One reviewer postulated that CO binding to CYP2E1 would prevent further
metabolism of dichloromethane, leading to increased metabolism through the GST pathway.
The reviewer noted the strong binding of CO to cytochrome P450s. The reviewer was concerned
that any spillover to the GST pathway that resulted from interference of CO:CYP2E1 binding
would result in errors in the estimation of the contribution of each pathway to total metabolism,
and suggested that the impact of this potential binding be discussed in the Toxicological Review.
Response: In examining this issue, limited data relevant to CO and CYP2E1 activity were
identified. In a controlled experimental study, Benowitz et al. (2003) found that cigarette
smoking resulted in an induction of CYP2E1 activity (measured by chlorzoxazone clearance)
while exposure to pure CO at a level that yields the same blood COHb as the cigarettes had no
effect on that activity. Given that CYP2E1 metabolic saturation has been observed for many
volatile organic compounds in the concentration range at which it appears to occur for
dichloromethane, and the observation of a negligible impact on CYP2E1 activity by direct CO
exposure of Benowitz et al. (2003), it seems unlikely that the effect of any CO:CYP2E1
inhibition would be significant. A more quantitative analysis could be performed if the binding
affinity (Kd) for CO-CYP2E1 were available; however, no data pertaining to this value were
identified. However, it should be noted that after a fairly high bolus of dichloromethane (526
mg/kg), the peak COHb seen by Pankow et al. (1991a) was about 9%. Model simulations
indicate that much higher COHb levels are not likely because the CYP-mediated metabolism to
CO becomes saturated. If one then assumes that CO binds to CYP2E1 with an affinity similar to
COHb, this suggests that the maximum level of CYP2E1 inhibition one would then see is around
10. In contrast, when increasing the dose of dichloromethane from 1 to 50 mg/kg
A-4 DRAFT - DO NOT CITE OR QUOTE
-------
dichloromethane, more than an order of magnitude lower than the dose used by Pankow et al.
(1991), McKenna and Zempel (1981) observed a 62% reduction in fraction of total dose exhaled
as CO, and in increasing the dose from 50 to 200 mg/kg, Angelo et al. (1986b) observed a
reduction of over 50% in that fraction. The changes in dichloromethane gas uptake rates
observed by Gargas et al. (1986) are similarly much larger than 10%. Thus, any error introduced
by ignoring a relatively small level of CYP2E1 inhibition (expected to be less than 10%) is not
expected to be significant.
If a metabolite had an effect on an enzyme activity, one would expect that effect to be
time-dependent: as more metabolite is produced (over the minutes and hours after the beginning
of an exposure), metabolite-induced inhibition would lead to a decrease in enzyme activity, and
hence a decrease in the rate constant (k) or Vmax. The model assumes that these constants are
not time-dependent, however, and for the most part appears to be consistent with the
toxicokinetic data. In short, if there was a strong time-dependence (due to inhibition) in the rate
constants, the model would not fit as well as it does. This is particularly true of the fits to the
gas-uptake data of Gargas et al. (1986) shown in Figure C-3. That the models can describe those
data well without explicitly including time-dependent inhibition suggests that the impact of such
inhibition is not significant.
An additional point to consider is that as long as the model accurately describes the shift
in metabolism between the two pathways, the specific mechanism by which the shift occurs (i.e.,
saturation versus inhibition) is not consequential. For the above reasons, and in the absence of
supporting data for the postulated inhibitory effect of CO on CYP2E1 metabolism of
dichloromethane, the Toxicological Review was not revised to include a discussion of this issue.
Comments: One reviewer noted a twofold difference in Km value for dichloromethane oxidative
metabolism to CO, and suggested that the potential reasons for the discrepancy (i.e., twofold
difference) and the overall impact on the dichloromethane PBPK model should be discussed.
This reviewer asked about the influence of CO-CYP2E1 formation on the high affinity and low
affinity Kms, on the overall Km, and noted that both in vitro and in vivo estimates involve some
unrealistic dichloromethane exposure concentrations.
Response: One correction to the reviewer's comment should be noted: the difference in Km
between that estimated by Reitz et al. (1989) and that obtained from the in vivo analysis is not
twofold, but rather, is over two orders of magnitude. Reitz et al. (1989) did not observe
dichloromethane oxidation to CO. CYP2E1 converts dichloromethane to formyl chloride, from
which CO may be produced by subsequent reactions. Reitz et al. observed the appearance of
aqueous-soluble radiolabeled products, likely a combination of multiple products, after
extracting the un-reacted radiolabeled substrate. At the in vivo dichloromethane concentrations,
A-5 DRAFT - DO NOT CITE OR QUOTE
-------
the high Km observed by Reitz et al. would be essentially indistinguishable from a first-order
pathway, since the in vivo dichloromethane concentrations remain well below that Km.
The concentration range studied by Reitz et al. (1989) was 1-10 mM, while the Km
estimated from in vivo toxicokinetic data is less than 10 uM. Thus Reitz et al. were not able to
detect a saturation constant in the range that is consistent with the in vivo data. Reitz et al. did
observe body weight saturation in the millimolar range, consistent with a Km in that range.
Evans and Caldwell (2010) discuss the possibility that CYP2E1 may have two active binding
sites, which could then be consistent with both a high and low Km. The model that they consider,
however, has only been tested (i.e., using in vitro data where the role of other enzymes can be
ruled out) against data for CYP isozymes other than CYP2E1, and not using dichloromethane.
If the reviewer is addressing another difference in Km values, e.g., from two different
studies or between two species, then such a difference is expected given the methods and data
used to estimate Km and inter-species differences in CYPs. In either case, as discussed in detail
in response to similar issues raised in PBPK Charge Question Ala, the largest expected impact
from CO-CYP binding is about 10%, whether two different CYPs or two binding sites on the
same CYP are involved. An effect of this magnitude was not considered significant.
Comments: One reviewer asked about the impact of formaldehyde produced during the
metabolism of dichloromethane on the oxidative metabolism of dichloromethane, and whether
formaldehyde affects the GST metabolic pathway.
Response: As depicted in Figure 3-1, formaldehyde is a product of GST-mediated metabolism,
not the CYP2E1 oxidative pathway. However, EPA is not aware of any quantitative data on the
amount of formaldehyde so produced, either in vitro or in vivo, or of any quantitative data on the
effect of formaldehyde on either CYP2E1 or GST-T1. Therefore, the potential impact of
formaldehyde on either pathway cannot be evaluated.
Alb. The internal dose metric used in the RfD and RfC derivations was based on total
hepatic metabolism via the CYP2E1 pathway. Because the metric is a rate of metabolism,
and the clearance of metabolites is generally expected to be slower in the human compared
with the rat (assuming clearance scales as body weight075), the rat internal dose metric is
adjusted by dividing by a toxicokinetic scaling factor to obtain a human-equivalent internal
dose. Are the choices of dose metric and toxicokinetic scaling factor appropriate and
scientifically supported? Is the rationale for these choices clearly described? Are the
uncertainties in the dose metric selection and calculations appropriately considered and
discussed?
A-6 DRAFT - DO NOT CITE OR QUOTE
-------
Comments: Four reviewers noted agreement with the choice of the dose metric, and one
reviewer did not comment directly on these questions. Two reviewers did not comment on this
question because it was outside their area of expertise.
Response: The dose metric was retained.
Comments: Three reviewers raised questions regarding the use of the toxicokinetic scaling factor
to account for potential differences between the rat and the human in the rate of clearance of
metabolites. The reviewers agreed on the underlying foundation for the use of the scaling factor
(i.e., the uncertainty that arises from the need to use a rate of metabolism rather than the effective
concentration in the PBPK modeling, and the unknown interspecies differences that preclude
prediction of tissue metabolite concentrations). The reviewers disagreed, however, on the
optimal approach for addressing this uncertainty, as described below.
Recommendations 1: One reviewer suggested that an UF rather than a scaling factor should be
used to address the uncertainty regarding the effect of interspecies differences in metabolite
clearance. The reviewer suggested that an alternative derivation using an interspecies
toxicokinetic UF = 3 should be presented.
Response: This issue concerns a specific mechanism—metabolic clearance—for which there is
information on animal-human differences; i.e., clearance is expected to vary as body weight0'75
based on data from various compounds. Thus, a scaling factor is used instead of a more general
UF because some information is available to support the magnitude of interspecies toxicokinetic
difference, rather than general uncertainty. An alternative derivation using an UF = 3 instead of
the scaling factor is not presented because it is not a procedure that is supported by the available
data. In contrast, a UF of 3 was applied to address toxicodynamic extrapolation in the absence of
information to characterize the magnitude of interspecies differences in toxicodynamics.
Recommendations 2: One reviewer suggested that the application of the scaling factor would
have the effect of introducing an additional source of uncertainty into the PBPK modeling. This
reviewer noted the metric used, i.e., the rate of metabolism, was a highly feasible metric given
the available data, but was one with "low desirability" based on relevance and closeness to the
postulated mode of action. In contrast, the more highly "desirable" metric, tissue metabolite
concentration, is not a metric that can be feasibly estimated and used. The use of the body
weight scaling factor addresses the residual uncertainty that comes from using the rate of
metabolism metric, but is itself a source of uncertainty. The reviewer suggested that EPA should
evaluate two approaches, one based on the PBPK-derived metric without the scaling factor and
one that includes the use of the scaling factor. From these two approaches, EPA would then
A-7 DRAFT - DO NOT CITE OR QUOTE
-------
choose based on residual uncertainty (level of confidence) and relevance to mode of action (i.e.,
closeness to the "desired" measure of internal dose).
Response: EPA's approach is consistent with this reviewer's suggestion. In individuals with
lower metabolism, the parent dichloromethane concentration will be higher at a given level of
exposure compared with those with higher metabolism. Since the MOA is that dichloromethane
is metabolically activated, higher risk is expected for populations with higher metabolism; i.e., in
populations with lower parent dichloromethane concentration. Hence parent concentration is
inversely correlated with risk, so use of the parent concentration has poor relevance to the MOA.
Allometric scaling is commonly used because it represents the expected, or most likely, variation
in metabolic clearance across species. Therefore the use of the rate of metabolism metric with
the scaling factor is believed to have the lowest uncertainty (i.e., represents the most likely
relationship for the concentration of the active metabolite) while using a metric that is relevant to
and consistent with the MOA.
Recommendations 3: One reviewer agreed with the use of the scaling factor, but noted that
further clarification regarding its use, consideration of other approaches, and overlap between
use of this factor any of the UFs that were also used should be included in the Toxicological
Review.
Response: A discussion of these points was added to Section 5.1.2 (Derivation Process for
Noncancer Reference Values).
Recommendation 4: One reviewer noted that "Clearance of unidentified CYP metabolites in
rodents versus humans remains an unknown," but questioned the use of the scaling factor for the
dose metric, suggesting that the use of "experimentally-derived or model predicted, species-
specific metabolic rate constants" would be preferable. This reviewer noted that use of this
scaling factor would yield a human equivalent dose that is excessive (i.e., a reference dose that is
lower than necessary).
Response: It is because of the absence of experimentally-derived, species-specific data
pertaining to the relationship between clearance of CYP metabolites in rodents and humans that
the scaling factor based on body weight0'75 scaling is warranted. EPA agrees that if these types
of data were available, it would be preferable to use them to evaluate and potentially modify or
eliminate the scaling factor.
Comments: One reviewer asked about the scientific rationale for using body weight0'75 rather
than a scaling factor based on a different value (e.g., body weight0'9).
A-8 DRAFT - DO NOT CITE OR QUOTE
-------
Response: Lindstedt and Schaeffer (2002) reviewed the use of allometry for anatomical and
physiological parameters. Overall metabolic rates were assessed by the metric of resting oxygen
uptake and found to vary with a coefficient of 0.725 (r2 = 0.997) when the allometric relationship
was regressed against data for 11 species, from a 4 g shrew to a 4,000 kg elephant. When the
regression was performed with data from only mice, rats, dogs, and humans, the coefficient
9
obtained was 0.766 (r = 0.997). Thus, a coefficient of 0.75 is consistent with this measure of
overall metabolism in multiple species, including those most commonly used for toxicity testing
and health assessments. That exogenous metabolism will scale in parallel with endogenous
metabolism is an assumption; in practice other PBPK modelers have used 0.7 or 2/3, but
especially in recent years 0.75 is most commonly used. EPA is not aware of data that would
support a value that is substantially higher or lower than 0.75. Lindstedt and Schaeffer (2002)
similarly found that resting ventilation scales with a coefficient of 0.745 across the 11 species
they evaluated, and cardiac output scaled with a coefficient of 0.750. Since ventilation and
cardiac output (hepatic blood flow) are also key determining factors in overall metabolism and
clearance of xenobiotic compounds, use of 0.75 in the absence of chemical-specific data is a
reasonable choice.
A2a. The mouse PBPK model used in deriving the cancer risk estimates was based on the
published work of Marino et al. (2006). Does the chosen model adequately represent the
toxicokinetics? Was the model applied properly? Are the model assumptions and
parameters clearly presented and scientifically supported? Are the uncertainties in the
model structure appropriately considered and discussed?
Comments: Four reviewers supported the chosen mouse PBPK model. One of these reviewers
suggested two additions to the Toxicological Review: a table describing the changes in the
mouse model (compared with the model used in the previous IRIS assessment), and information
regarding how CYP2E1 variability was incorporated into the model (information that was
included in the appendix). Two reviewers did not comment on this question because it was
outside their area of expertise.
Response: Table 3-5 provides a comparison of parameters used in the previous assessment and
those used in the current mouse model. A summary of EPA's consideration of CYP2E1
variability is included in Section 3.5.2 (Probabilistic Human PBPK Dichloromethane Model
[David et al., 2006]).
Comments: One reviewer noted that since tumor formation is thought to result from the GST
metabolic pathway, dichloromethane levels below the level of CYP saturation should result in no
A-9 DRAFT - DO NOT CITE OR QUOTE
-------
cancer risk or a highly attenuated risk relative to that seen above the saturation of the CYP
pathway.
Response: The predicted risk is highly attenuated below the exposure levels where CYP is
saturated, compared to those occurring above it, because CYP metabolism predominates over
GST metabolism at the lower levels. This attenuation is built into the PBPK modeling. EPA
does not consider, however, the assumption of zero risk to be supported.
A2b. The internal dose metric used in the cancer quantitation was based on tissue-specific
GST metabolism. To account for potential clearance rate differences, the mouse internal
dose metric was adjusted by dividing by a toxicokinetic scaling factor to obtain a human-
equivalent internal dose. Are the choices of dose metric and toxicokinetic scaling factor
appropriate and scientifically supported? Is the rationale for these choices clearly
described? Are the uncertainties in the dose metric selection and calculations
appropriately considered and discussed?
Comments: Two reviewers specifically noted support for the use of the tissue-specific GST
metabolism dose metric, and the other reviewers did not specifically address this question.
Response: The dose metric was retained.
Comments: One reviewer stated that the use of the scaling factor was appropriate and clearly
explained. Four reviewers did not provide comments in response to this charge question (two of
these noting that it was outside their area of expertise). Two reviewers indicated their response
to the scaling factor question was similar to that of the scaling factor addressed in PBPK
Modeling Charge Question Alb, specifically raising questions about the justification of this
factor. One reviewer noted that "the application of the toxicokinetic scaling as done would be
appropriate when the chemical entity (metabolite) itself is the active moiety (which is the case),
further metabolism/reaction renders it inactive (which is likely the case), and the rate of the
metabolism/reaction process is proportional to the liver perfusion rate, cardiac output or to the
body surface (which is not known to be the case)." This reviewer stated that additional support
for the use of the scaling factor addressing these points should be provided in the Toxicological
Review.
Response: As noted by the reviewer, the first two elements justifying the use of the scaling
factor are met. With respect to the third element, it is not known that the rate of reaction is
proportional to the liver perfusion rate, cardiac output, or body surface area. It is also important
to note that it is not known that the rate of reaction is not proportional to these factors. Hence no
A-10 DRAFT - DO NOT CITE OR QUOTE
-------
information is available to inform the choice for this element; i.e., whether the clearance of the
metabolite is proportional to liver perfusion (allometric scaling) or proportional to liver volume
(no scaling factor, or scaling factor =1). Given this uncertainty, use of the scaling factor is
warranted. The discussion of these points was expanded in Sections 5.1.2 (Derivation Process
for Noncancer Reference Values), 5.4.1.4 (Dose Conversion and Extrapolation Methods: Cancer
OSF) and 5.4.2.4 (Dose Conversion and Extrapolation Methods: Cancer IUR).
Comments: One reviewer raised two questions about the extrapolation of the animal results to
humans. One issue concerns the potential contribution of CYP2E1 metabolism, which is much
greater in humans than in mice, to genotoxicity. The other issue concerns target tissue
concordance and the potential relevance to the observation of leukemia and other types of
cancers that were not observed in mice. This reviewer noted the uncertainties arising from these
issues as another justification for the use of the scaling factor, and suggested that additional
discussion of the potential underestimation of exposure to reactive metabolites should be added.
This reviewer reiterated and expanded on this issue in response to PBPK Charge Question A3 a
and Carcinogenesis Charge Questions Cl, C4, and C5.
Response: Responses to this issue are discussed in more detail in response to PBPK Charge
Question A3a and Carcinogenesis Charge Questions Cl.
A3a. A probabilistic human PBPK model (David et al., 2006) was used to estimate a
distribution of human equivalent doses and concentrations for the points of departure
(PODs) for the RfD and RfC, respectively. The 1st percentile of these distributions was
selected to represent the most sensitive portion of the population. For the derivation of the
oral and inhalation cancer risk estimates, the probabilistic human PBPK model was used
to calculate the distribution of human internal doses (mg dichloromethane metabolized via
the tissue-specific GST pathway per unit volume of tissue) that would be expected from a 1
mg/kg-day oral dose or a 1 ug/m3 inhalation concentration. This distribution of human
internal doses was used with the tumor risk factor to generate a distribution of oral slope
factors or inhalation unit risks. Does the chosen model adequately represent the
toxicokinetics? Was the model applied properly? Are the model assumptions clearly
presented and scientifically supported? Are the uncertainties in the model appropriately
considered and discussed?
Comment: Three reviewers supported the selection of the model. Two reviewers did not
comment on this question because it was outside their area of expertise.
A-l 1 DRAFT - DO NOT CITE OR QUOTE
-------
Comment: One reviewer asked for the scientific evidence supporting the assumption that some
dichloromethane is metabolized by the GST pathway at all levels of exposure.
Response: Below -20% of the Km (which has units of concentration), the rate of reaction
becomes indistinguishable from a first-order reaction, as it depends on the probability that a
substrate molecule collides with an unoccupied active site on the enzyme. Thus, at low
concentrations ([Substrate] « Km) the rate of enzyme-catalyzed reactions becomes proportional
to the concentration of the substrate(s) and enzyme. As long as GSH, GST, and dichloromethane
are present at non-zero concentrations, the reaction will proceed at a non-zero rate; that is, the
probability of collision with the active site is non-zero, and the rate of reaction will be non-zero.
EPA is not aware of any data for single-enzyme/single-substrate reactions that are not consistent
with this fundamental theory of enzyme kinetics.
Comments: One reviewer, in response to PBPK Modeling Charge Question A3b, noted that the
choice of the 1st percentile to account for population variability was supported. Another
reviewer, in response to RfD Question B4, noted that the 1st percentile was chosen to be
protective of sensitive individuals, and that this decision, in conjunction with the use of the
toxicodynamic variability UF, was adequate. Another reviewer did not agree with use of the 1st
percentile of distribution of the human internal doses, observing that the PBPK model already
incorporated measures of variability, and that use of the 1st percentile in addition to use of the
most sensitive GST genotype and the toxicokinetic scaling factor in the cancer potency
derivations "over adequately accounts for human variability."
Response: Cancer risk values were based on the mean of the most sensitive GST genotype, not
on the upper or lower tail of the distribution. With respect to the noncancer values (RfD and
RfC), however, the 1st percentile of the distribution of human equivalent doses or concentrations
for the full population (i.e., all genotypes combined) was used. Thus the reviewer's
interpretation of the procedure used in the RfD and RfC derivation was mistaken. The reviewer
was correct that the human PBPK model incorporates variability in the estimations through
statistical sampling of metabolic and physiological parameters from various distributions. This
procedure explicitly accounts for variability that results from these known factors that influence
toxicokinetics and dosimetry and so allows for the generation of dosimetric distributions based
on population variability. However, one must then select a point on the dosimetric distribution
corresponding to a portion or percentile of the population one effectively wishes to protect. The
1st percentile of the distribution was selected for derivation of the RfD and RfC as a low but non-
zero population percentile that can be estimated by computational statistical sampling in a
reasonable amount of time (i.e., with a reasonably stable estimate achieved through a finite
number of iterations).
A-12 DRAFT - DO NOT CITE OR QUOTE
-------
Comments: One reviewer noted that GST activity is greater in mice than rats, and greater in rats
than the most sensitive human (i.e., those who are GST-TI+/+), and noted the relative lack of
carcinogenic response seen in the rat bioassays. This reviewer then asked for clarification
regarding the assumption of higher human responsiveness as discussed in the justification of the
use of the scaling factor in the 1987 dichloromethane assessment, and differences between this
assumption and the use of the scaling factor in the current assessment.
Response: The explanation of what was done in the 1987 dichloromethane assessment was
clarified in Section 3.5 (Physiologically Based Pharmacokinetic Models). Overall cancer risk
associated with exposure to DCM can be viewed as dependent on two toxicokinetic terms: rate
of GST metabolism and rate of GST-metabolite clearance. For cancer it is assumed that given
the same average tissue concentration of active metabolite, humans would have the same average
cancer risk as rodents; i.e., that humans are as "responsive" per unit of active metabolite tissue
concentration as the most sensitive animal species/sex/strain. The concentration of active
metabolite in humans relative to the mouse or rat is approximated as:
[relative human concentration] = [relative GST metabolism]/[relative metabolite clearance].
Because metabolism and other clearance mechanisms (blood perfusion, respiration, renal
filtration) are all expected to be about sevenfold slower in a 70-kg human than a 30-g mouse, the
second term (relative metabolite clearance), which is in the denominator, is assumed to be 1/7
(one over the toxicokinetic scaling factor).
The reason that rats do not exhibit a tumor response similar to that of mice, given the
similarity in GST kinetics (and expected similarity in clearance of GST-metabolites), is not
known; however, such differences in response among species exist for many chemicals.
Consistent with EPA's Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), the
cancer assessment for dichloromethane is based on tumor data from the most sensitive species.
In the 1987 assessment, the scaling factor was applied to adjust for both interspecies
differences in processes that lead to differences in internal doses (e.g., clearance of a given daily
amount of dichloromethane metabolically activated per liter of tissue) and differences in
toxicodynamics or response, whereas the scaling factor used in the current assessment
specifically accounts for expected differences in clearance between species. The factor in the
current assessment is based on scaling as [body weight]3 4, reflecting current practices for
interspecies extrapolation (U.S. EPA, 2005a, 1992); the 1987 assessment applied the scaling
factor of [body weight]213 that was current practice at that time. These differences in scaling
resulted in the different applied factors of 7.0 (current assessment) versus 12.7 (1987
assessment)."
A-13 DRAFT - DO NOT CITE OR QUOTE
-------
Comments: One reviewer supported the use of the Bayesian MCM analysis to address human
variability, but raised one question regarding the blood:air partition coefficient (PB) for the
mouse based on experimentally-derived values from Clewell et al. (1993). This value (23) is
higher than the value (8.3) used in previous models. The value for the human (9.7) is closer to
the previous value, however, and is based on the values from Andersen et al. (1987). The
reviewer raised the question of whether the PB value for humans could be in error, reflecting an
underlying difference in the methods of Clewell et al. (1983) and Andersen et al. (1987).
Response: As noted in Section 3.5.1 (Probabilistic Mouse PBPK Dichloromethane Model), the
value for the mouse PB used in the current assessment is consistent with that measured in rats
(19.4) and hamsters (22.5). The difference between the (new) values for rat, hamster, and mouse
and human PB value in this case is typical of rodent-human differences seen for other
compounds (e.g., see Wiester et al., 2002). The difference between rodent and human values of
PB for VOCs is believed to result from differences in lipid content. The difference between the
earlier measured value (8.3) and the more recent value (23) for mice is likely the result of
improvements in the method for measuring PB in the seven years that passed between the two
studies.
Comments: One reviewer suggested that non-Michaelis-Menten kinetics of CYP2E1 metabolism
of dichloromethane and uncertainty regarding site concordance and dose metrics be added to the
discussion of the model. (A similar comment was made in response to PBPK Modeling Charge
Question A2b).
Response: A discussion of the alternate CYP kinetics is included in Section 3.5.5 (Uncertainties
in PBPK Model Structure for the Mouse, Rat and Human), Section 5.3 (Uncertainties in the Oral
Reference Dose and Inhalation Reference Concentration), and Section 5.4.5 (Uncertainties in
Cancer Risk Values. GST-T1 is expressed in tissues other than the liver and lung, including the
human brain (Juronen et al., 1996), breast tissue (Lehmann and Wagner, 2008), and rat olfactory
epithelium (Banger et al., 1994). While the extent of GSH conjugation in these other tissues may
not be significant to the overall dosimetry of dichloromethane, they can be significant in
understanding the sites of action of dichloromethane as these vary among species.
A3b. EPA modified the parameter distributions in the published David et al. model. Does
the set of model parameter distributions adequately account for population variability and
parameter uncertainty in estimating human equivalent doses? Are the human parameter
values and distributions clearly presented and scientifically supported?
A-14 DRAFT - DO NOT CITE OR QUOTE
-------
Comments: One reviewer considered the development and inclusion of the parameter
distributions that reflected both parameter uncertainty and interindividual variation to be an
important addition to the earlier published version of this model, and that these distributions and
their development were clearly explained and scientifically justified. This reviewer specifically
noted support for the use of the data from Lipscomb et al. (1997) to characterize the distribution
of CYP enzymes activities in humans. Two reviewers did not comment on this charge question
because it was outside their area of expertise.
Response: No response required.
Comments: One reviewer asked why the geometric standard deviations were based on the
trichloroethylene model rather than the distribution of CYP protein.
Response: The geometric standard deviations for CYP were not based on the trichloroethylene
model per se, but on measures of metabolic activity using trichloroethylene as the substrate.
Protein activity is not necessarily proportional to protein levels, and protein levels are more
difficult to quantify accurately. CYP2E1 activity is often measured using p-nitrophenol
hydroxylation activity; in the dichloromethane assessment an alternate CYP2E1-specific
substrate was used to quantify the range in activity.
Comments: One reviewer questioned the body weight adjustment for CYP2E1 activity, and
suggested an adjustment based on liver volume as an alternative.
Response: CYP activity was scaled by body weight because that is the scaling relationship used
by David et al. (2006). Further, EPA does not consider scaling by liver weight to be better
supported. In addition, body weights, but not liver weights, are available for the individual
subjects from which human toxicokinetic data were taken. (Liver weight data would not be
obtainable in this type of study, since it involved healthy individuals.) An assumed relationship
between liver weight and body weight could be used, but then one would implicitly be using
body weight.
Comments: One reviewer asked if the distribution of KfC for the U.S. population derived from
the Swedish study took into account the ethnic make-up of the respective populations.
Response: The proportion of the three genotypes in the U.S. population, as reported by Haber et
al. (2002) and used here, represents the ethnicity distribution of the U.S. population, based on
U.S. census data. This point is provided in Section 3.3 (Metabolism).
A-l 5 DRAFT - DO NOT CITE OR QUOTE
-------
Comments: One reviewer asked how the mass balance of the flows and volumes was ensured
during the MC iterations.
Response: As indicated in the last column of Table B-3, after each set of MC samples for
fractional blood flows, for example, the sampled values were divided by the sum of the sampled
values, so that the sum of the resulting fractions equal one. Volume fractions were normalized to
sum to 0.9215, allowing for 7.85% of the body weight as bone, teeth, nails, and hair.
Comments: One reviewer asked about the basis upon which the normal distribution for GST was
set.
Response: A normal (i.e., Gaussian) distribution, rather than some other distribution form, was
used because that is what was used by David et al. (2006) and it is consistent with the data of
Warholm et al. (1994). Warholm et al. (1994) state that their data are consistent with Hardy-
Weinberg equilibrium: the expected long-term population distribution for a binary polymorphism
in a population assuming no selection advantage for one of the three genotypes.
B. Noncancer Toxicity of Dichloromethane
Oral reference dose (RfD)for dichloromethane
Bl. A chronic RfD for dichloromethane has been derived from a 2-year oral (drinking
water) study in the rat (Scrota et al., 1986a). Please comment on whether the selection of
this study as the principal study is scientifically supported and clearly described. Please
identify and provide the rationale for any other studies that should be selected as the
principal study.
Comments: Six reviewers supported the selection of the Serota et al. (1986a) as the principal
study. One reviewer also suggested that the choice of the principal study would be strengthened
by inclusion of a graphical presentation of the different endpoints based on internal dose metrics.
One reviewer did not comment on this question because it was outside this reviewer's area of
expertise.
Response: EPA agrees that an exposure-response array such as that described by the reviewer
could be an effective way to compare study findings across studies and endpoints; however, a
figure based on internal dose metrics was not developed for the following reasons: 1) equivalent
types of dose metrics (i.e., tissue-specific concentrations or rates of metabolism) cannot be
generated for all of the endpoints that were considered (specifically neurotoxicity and
A-16 DRAFT - DO NOT CITE OR QUOTE
-------
reproductive/developmental toxicity) and 2) if the surrogate blood AUC concentration was used
from the PBPK model, the internal doses would be proportional to the administered dose unless
the endpoints were from different species.
Comment: One reviewer suggested that EPA confirm whether the Agency had access to the
original Hazleton report in the rat to check for any discrepancies between the original report and
the published study (Serota et al., 1986a).
Response: EPA was unable to obtain the original Hazleton report.
B2. An increase in the incidence of liver lesions (foci/areas of alteration) was selected as the
critical effect for the RfD. Please comment on whether the selection of this critical effect is
scientifically supported and clearly described. Please identify and provide the rationale for
any other endpoints that should be selected as the critical effect.
Comments: Six reviewers supported the selection of liver lesions (foci/areas of alteration) as the
critical effect for the RfD. One of these reviewers reiterated the idea of presenting an exposure
response array based on internal dose metrics to strengthen the selection of the critical effect.
One reviewer did not comment on this question because it was outside this reviewer's area of
expertise.
Response: A response that addresses the recommendation for an exposure response array based
on internal dose metrics is provided under RfD Charge Question Bl.
B3. Benchmark dose (BMD) modeling was applied to the incidence data for liver lesions to
derive the POD for the RfD. Has the BMD modeling been appropriately conducted and
clearly described? Is the benchmark response (BMR) selected for use in deriving the POD
(i.e., a 10% increase in incidence of liver lesions) scientifically supported and clearly
described?
Comments: Six reviewers generally agreed that the application of BMD analysis to derive the
POD for the RfD was appropriate and clearly described. One of these reviewers specifically
stated that a BMR of 10% increase in the incidence of liver lesions is scientifically supported and
consistent with EPA guidelines. One reviewer commented that selection of a 10% BMR is not
appropriate if the UF that would be applied to a NOAEL (i.e., UFL = 1) is used. This reviewer
indicated that a 5% BMR more closely corresponds to a NOAEL for this type of data (animal
study and quantal response). One reviewer did not comment on this question because it was
outside this reviewer's area of expertise.
A-17 DRAFT - DO NOT CITE OR QUOTE
-------
One reviewer indicated that using a 1st percentile human equivalent dose as well as a
subset of the GST-T1+ + population for the RfD derivation was not justified. One reviewer
disagreed with applying a toxicokinetic scaling factor to the internal dose. This reviewer also
asked if the dose metric shown in Appendix Figures D-l and D-2 was the same as the dose
metric described in the text, and suggested that it would be useful to show alternative RfDs based
on other dose metrics.
Response: An UF for LOAEL-to-NOAEL extrapolation was not used because the Agency's
current approach is to address this factor as one of the considerations in selecting a BMR for
BMD modeling. A 10% extra risk of increased foci/areas of alterations was applied under the
assumption that it represents a minimal biologically significant degree of change. Therefore, a
UF for extrapolation from a LOAEL to a NOAEL was not applied. There are no additional data
to suggest that the critical response has a greater sensitivity that would warrant a lower BMR.
(See Sections 5.1.4 and 5.2.3 for rationale).
The GST-T1+ + population subset was not selected as a sensitive group for the RfD (or
RfC) derivation. The reasoning behind the use of the 1st percentile human equivalent dose is
discussed in response to comments under PBPK Modeling Charge Question A3a. The use of the
toxicokinetic scaling factor is addressed in the response to comments under PBPK Modeling
Charge Questions Alb and A2b. The dose metric in Appendix Figures D-l and D-2 was
clarified. The rationale for the use of the chosen dose metric is explained in Section 5.1.3
(Evaluation of Dose Metrics for Use in Noncancer Reference Value Derivations). The GST
metabolism, AUC, and combined GST and the CYP dose metrics did not present reasonable
choices based on model fit and consistency of response across studies at comparable dose levels,
and as such development of RfDs based on these dose metrics was not supported.
Comments: One reviewer asked for the rationale for selection of the model, among the suite of
BMD models available, used to derive the RfD and RfC. This reviewer asked how these models
relate to the available scientific data.
Response: There are currently no mechanistic data to support or derive a biologically-based
model or to inform the model selection from among the available empirical models. The BMD
models selected for the reference value derivations were based on the best statistical fit to the
data as described in Sections 5.1.4, 5.2.3 and Appendix D.
A-l 8 DRAFT - DO NOT CITE OR QUOTE
-------
B4. Please comment on the rationale for the selection of the uncertainty factors (UFs)
applied to the POD for the derivation of the RfD. Are the UFs scientifically supported and
clearly described in the document? Please provide a detailed explanation. If changes to an
UF are proposed, please identify and provide a rationale.
Comments: Reviewers generally agreed with application of the UFs for animal to human
extrapolation (i.e., UFA of 3 to account for the toxicodynamics portion of the interspecies UF)
and for human variability in sensitivity (i.e., intraspecies UFn of 3). One reviewer reiterated a
comment on the LOAEL to NOAEL UF, noting that use of an UF that would be applied with a
NOAEL (i.e., UFL = 1) is not appropriate with a BMR of 10%. Two reviewers did not offer
comments on the UFs used to derive the RfD, one noting that the calculation of RfDs was
outside this reviewer's area of expertise.
Response: The issues of the choice of a 10% BMR and related UF pertaining to LOAEL to
NOAEL extrapolation, are discussed in response to RfD Charge Question B3.
Comments: Two reviewers agreed with the selection of a database UF of 3, one of the reviewers
noting the importance of reflecting data deficiencies in the area of neurodevelopmental effects.
Three reviewers did not think that a database UF of 3 was necessary (one noting that there is a
great deal of research on a wide variety of endpoints in several species) or well justified (i.e., that
the case had not been made for uncertainty associated with lack of a neurodevelopmental
assessment). One of these reviewers suggested running a PBPK model to determine the CO
levels expected at low dichloromethane doses and to compare these levels to CO levels that are
known to produce CNS effects in animals and humans.
Response: As recommended by a peer reviewer to better address uncertainty associated with
lack neurodevelopmental toxicity data for dichloromethane, EPA compared CO levels associated
with neurotoxicity and predicted CO levels (via PBPK modeling) from the human equivalent
dose used to derive the RfD using the human dichloromethane PBPK model (David et al., 2006).
The human PBPK model predicted relatively low CO exposures at dichloromethane levels at the
RfD, levels that would not be expected to result in a substantial health risk. This analysis
assumes that CO is the only agent responsible for the observed CNS and potential developmental
neurotoxicity effects. This assumption is not supported by the available data, however, in that
mechanistic studies and studies comparing CO and dichloromethane exposures in rats (Karlsson
et al., 1987; Rosengren et al., 1986; Putz et al., 1979; Stewart et al., 1972a, b), have also
indicated the parent compound can pass through the placenta. The potential involvement of
other metabolites on neurological effects has not been adequately examined. The available data
are consistent with neurodevelopmental toxicity potential for dichloromethane, but it is not
A-19 DRAFT - DO NOT CITE OR QUOTE
-------
known if neurodevelopmental toxicity would be a more or less sensitive endpoint than the
critical endpoint (liver lesions). The database UF is also applied because a two-generation oral
exposure reproductive toxicity study is not available. As mentioned in Section 5.1.5, there is a
two-generation inhalation exposure study by Nitschke et al. (1988a) that reported no
reproductive toxicities at exposures of >100 ppm; however, Nitschke et al. (1988a) did not dose
the animals continuously during gestation and lactation. This represents a limitation of the study.
In consideration of these deficiencies in the database, a database UF of 3 is supported. The
discussion of the database UF (Section 5.1.5.) was revised to more clearly describe the rationale
for application of a database UF of 3.
Comments: One reviewer suggested that an additional UF be included to account for
uncertainties in dichloromethane metabolism.
Response: As discussed in EPA's A Review of the Reference Dose and Reference Concentration
Process (U.S. EPA, 2002), UFs are applied to account for uncertainties in animal to human
extrapolation (UFA), human variability (UFn), subchronic to chronic extrapolation (UFs),
LOAEL to NOAEL extrapolation (UFL), and database deficiencies (UFD). The interspecies
scaling factor accounts for some of the uncertainty in overall dichloromethane metabolism.
Using the 1st percentile internal human equivalent dose rather than the mean of the population
distribution, as predicted by the human PBPK model, also explicitly accounts for uncertainty and
variability in the metabolic activity among humans. Thus, EPA concluded that using these two
factors in combination sufficiently accounts for toxicokinetic uncertainties, which include those
from metabolism.
Comments: One reviewer indicated that where the metric is the rate of metabolism (production)
rather than the more relevant metric of metabolite concentration as is the case with
dichloromethane, both the toxicokinetic component of the interspecies UF (= 3) and the scaling
factor were in effect accounting for interspecies differences in clearance of the parent compound.
The reviewer raised concerns about the "repetitive application" of the scaling factor within this
context, and how this compares to a situation where no PBPK modeling is applied (the "default"
approach) in which either a toxicokinetic UF (= 3) or a body surface area scaling factor (but not
both) is used to account for toxicokinetic uncertainty.
Response: The conceptual difficulty with the dichloromethane PBPK model is that it is an
incomplete model regarding both the parent dichloromethane and the presumed toxic
metabolite(s). Ideally there would be two (or more) linked PBPK models, or sub-models, the
first being the existing model that describes parent dichloromethane absorption, distribution,
metabolism, and elimination (ADME) processes, and the downstream sub-models describing the
A-20 DRAFT - DO NOT CITE OR QUOTE
-------
ADME processes for each toxic metabolite. A linked sub-model would track a metabolite's
distribution throughout the body, its metabolism and elimination, and therefore a linked sub-
model could be used without a scaling factor. Text was added to Section 5.1.2 to clarify those
factors applied to account for the toxicokinetic and dynamic components of interspecies
extrapolation and intraspecies variability.
Inhalation reference concentration (RfC) for dichloromethane
B5. A chronic RfC for dichloromethane has been derived from a 2-year inhalation
bioassay in rats (Nitschke et al., 1988a). Please comment on whether the selection of this
study as the principal study is scientifically supported and clearly described. Please
identify and provide the rationale for any other studies that should be selected as the
principal study.
Comments: Five reviewers supported the selection of Nitschke et al. (1988a) as the principal
study. One reviewer did not consider the inhalation study by Nitschke et al. (1988a) an
appropriate basis for RfC derivation because of limitations in the liver lesion data (hepatic
vacuolization) reported in this study. One reviewer suggested that a graphical display of
endpoint data based on internal dose metrics would strengthen the choice of the principal study.
One reviewer did not comment on this question because it was outside this reviewer's area of
expertise.
Response: A response that addresses the critical effect (hepatic vacuolization) is provided under
RfD Charge Question B6. A response that addresses the recommendation for an exposure
response array based on internal dose metrics is provided under RfD Charge Question Bl.
Comments: Two reviewers noted the value of the findings from epidemiological studies of
neurological effects in workers exposed to dichloromethane as supportive data for the RfC
derived from the animal data. One reviewer suggested that consideration be given to studies that
identified elevated levels of COHb (an effect identified in humans) or the Savolainen et al.
(1981) neurological study in rats as possible principal studies.
Response: EPA agrees that the potential for a significant health risk due to dichloromethane-
generated CO should be evaluated to assure that the final RfC is protective for that mode of
action. Increased CO and COHb have implications for neurotoxicity as well as cardiovascular
disease risk. In examining this question, inaccuracies in the CO submodel representing
endogenous CO production of the David et al. (2006) PBPK model were noted. Since protection
against CO-induced effects should consider all sources of CO, determining an RfC for
A-21 DRAFT - DO NOT CITE OR QUOTE
-------
dichloromethane based on CO would need to account for the range of ambient CO levels likely
encountered, further complicating the analysis. Thus, a simpler evaluation was performed
involving a comparison of CO exposure rates at the CO National Ambient Air Quality Standard
(NAAQS) versus that produced from dichloromethane metabolism at the candidate RfC based on
liver effects in the rat. This evaulation indicated that an RfC based on CO synthesis would likely
be much higher than the one based on rat liver effects, and hence not sufficiently protective
against liver lesions. Therefore, development of an alternative RfC based on CO synthesis was
not warranted.
Savolainen et al. (1981) observed neurochemical changes in the cerebrum and cerebellum
of rats following a 2 week (6 hours/day, 5 days/week) exposure to 500 ppm, 1,000 ppm, and
1,000 TWA ppm dichloromethane. EPA did not consider this study as a principal study because
of the acute nature of the exposure conditions, the type of effect examined, and the fact that the
effects seen were not as sensitive as the liver effects seen in Nitschke et al. (1988a). EPA
considered the comparison RfC developed based on the Lash et al. (1991) study of neurological
effects among retired dichloromethane-exposed workers to be of higher quality than the
Savolainen et al. (1981) study in rats for the evaluation of neurological effects.
B6. An increase in the incidence of hepatic vacuolation was selected as the critical effect
for the RfC. Please comment on whether the selection of this critical effect is scientifically
supported and clearly described. Please identify and provide the rationale for any other
endpoints that should be selected as the critical effect.
Comments: Five reviewers supported the selection of hepatic vacuolation as the critical effect for
the RfC. One reviewer questioned the biological significance of hepatic vacuolation as the
critical effect, noting that hepatic vacuolation appeared to be a high-dose effect in female rats
only, was incompletely reported in the male rat, and had no human correlate. This reviewer
suggested that these limitations should be noted in the Toxicological Review. One reviewer did
not comment on this question because it was outside this reviewer's area of expertise.
Response: To address the reviewer's questions regarding the male rats and humans, a discussion
of biological relevance of hepatic vacuolation was added to the discussion of the selection of the
critical effect in Section 5.2.1 (Choice of Principal Study and Critical Effect—with Rationale and
Justification). The Nitschke et al. (1988a) study includes a relatively low dose range (< 500
ppm), with incidence rates of hepatocyte vacuolation of 59, 60, 58, and 76% at 0, 50, 200, and
500 ppm, respectively. Data from Burek et al. (1984) (also in female Sprague-Dawley rats) are
based on higher exposure levels (i.e., the lowest exposure was equal to the highest exposure in
Nitschke et al. (1988a). An increased incidence of hepatocyte vacuolation was seen at 500,
1,500, and 3,500 ppm (52, 58, and 65%, respectively) relative to the control incidence (34%).
A-22 DRAFT - DO NOT CITE OR QUOTE
-------
The lowest concentration in Burek et al. (1984), 500 ppm, was the LOAEL, and this study does
not inform the shape of the curve below this value. Although a linear dose-response is not seen
across the experimental dose range (0 to 500 ppm), an increase is seen at the highest dose, which
is consistent with the dose-related increase observed in the Burek et al. (1984) study. Thus, it
would be incorrect to view the Nitschke et al. (1988a) as reflecting no dose-response.
B7. BMD modeling was applied to the incidence data for hepatic vacuolation to derive the
POD for the RfC. Has the BMD modeling been appropriately conducted and clearly
described? Is the BMR selected for use in deriving the POD (i.e., a 10% increase in
incidence of hepatic vacuolation) scientifically supported and clearly described?
Comments: Six reviewers generally agreed that the application of BMD analysis to derive the
POD for the RfC was appropriate and clearly described. One of these reviewers specifically
noted that the BMR was scientifically supported, clearly described, and consistent with EPA
guidelines. One reviewer did not comment on this question because it was outside this
reviewer's area of expertise. Three reviewers reiterated comments that had also been offered in
response to other charge questions. One of these reviewers commented that selection of a 10%
BMR is not appropriate if the UF that would be applied to a NOAEL (i.e., UFL = 1) is used. This
reviewer indicated that a 5% BMR more closely corresponds to a NOAEL for this type of data
(animal study and quantal response). One reviewer questioned the use of the 1st percentile
human equivalent in the RfC derivation, and one reviewer questioned the use of a toxicokinetic
scaling factor.
Response: A response that addresses the application of a LOAEL-to-NOAEL UF where BMD
modeling was conducted is provided under RfC Charge Question B3.
The reasoning behind the use of the 1st percentile human equivalent dose is discussed in
response to comments under PBPK Modeling Charge Question A3 a. The use of the
toxicokinetic scaling factor is addressed in the response to comments under PBPK Modeling
Charge Questions Alb and A2b and RfD Charge Question B4.
B8. Please comment on the rationale for the selection of the UFs applied to the POD for the
derivation of the RfC. Are the UFs scientifically supported and clearly described in the
document? Please provide a detailed explanation. If changes to an UF are proposed,
please identify and provide a rationale.
Comments: Reviewers generally agreed with application of the UFs for animal-to-human
extrapolation (i.e., UF of 3 to account for the toxicodynamics portion of the interspecies UF) and
for human variability in sensitivity (i.e., intraspecies UF of 3).
A-23 DRAFT - DO NOT CITE OR QUOTE
-------
One reviewer offered a specific comment on the LOAEL-to-NOAEL UF, noting that use
of an UF that would be applied with a NOAEL (i.e., UF of 1) is not appropriate with a BMR of
10%. One reviewer did not comment on this question because it was outside this reviewer's area
of expertise.
Response: A response that addresses the application of a LOAEL-to-NOAEL UF where BMD
modeling was conducted is provided under RfC Charge Question B3.
Comment: Two reviewers agreed with the selection of a database UF of 10. One reviewer did
not think that a database UF of 10 was necessary because of the relatively large database for
dichloromethane, and suggested that either a value of 3 or 1 would be more appropriate. Another
reviewer stated that the database UF of 10 was not well justified, noting the relatively large
database for dichloromethane and that it is inconsistent to have different database UFs for
different routes of exposure for a systemic toxicant such as dichloromethane. Another reviewer
also disagreed with the database UF of 10 because, although neurodevelopmental effects have
not been the focus of studies of dichloromethane exposure, other structurally related volatile
organic compounds have not been shown to result in this type of effect. This reviewer also noted
that maternal CO levels would not be raised at the proposed value of the RfC.
Response: EPA agreed with the recommendation to change the RfC database UF to 3, the same
value as is used for the RfD database UF. The inhalation database for dichloromethane is large,
but several concerns raised by specific studies support a database UF of 3. Specifically, the
database UF of 3 for the RfC is applied because of concerns with respect to the limitations of the
study protocol of the two-generation reproductive toxicity study, and concerns regarding the
neurodevelopmental and immunotoxicity studies that are not adequately addressed by the
available studies. As one reviewer suggested, EPA determined the levels of CO at the human
equivalent concentration used for deriving the RfC using the human PBPK model (David et al.,
2006). The predicted levels of CO generated by the model were lower than levels at which
neurodevelopmental toxicity has been observed. Therefore, EPA agrees that it is unlikely that
the CO generated from dichloromethane metabolism would result in neurodevelopmental
toxicity. However, this analysis does not account for potential concerns of neurodevelopmental
toxicity associated with the parent compound or possibly other metabolites. Dichloromethane
exposure is known to produce neurotoxicity in humans and adult animals (see Sections 4.1.2.9
and 4.4.3), and these effects cannot be explained solely by CO (Karlsson et al., 1987; Rosengren
et al., 1986; Putz et al., 1979; Stewart et al., 1972a, b). The parent compound can pass through
the placental barrier (Withey and Karpinski, 1985; Anders and Sunram, 1982). Data from
Aranyi et al. (1986) demonstrated evidence of portal-of-entry (i.e., pulmonary)
immunosuppression following a single 100 ppm dichloromethane exposure for 3 hours in CD-I
A-24 DRAFT - DO NOT CITE OR QUOTE
-------
mice, an exposure level that is lower than the POD for the liver effects that serve as the critical
effect for the RfC. Increased risk of pulmonary infectious diseases, particularly bronchitis-
related mortality, is also suggested by some of the cohort studies of exposed workers (Radican et
al., 2008; Gibbs et al., 1996; Gibbs, 1992; Lanes et al., 1993, 1990), providing additional support
for the potential importance of this localized immunosuppression. In consideration of these
deficiencies in the database, a database UF of 3 is supported. The discussion of the database UF
in Section 5.2.4 was revised to more clearly describe this rationale.
Comments: One reviewer repeated a suggestion (also made in response to RfD Charge Question
B4) that an additional UF be included to account for uncertainties in dichloromethane
metabolism.
Response: A response to this comment is provided under RfD Charge Question B4. Briefly, the
interspecies scaling factor and the use of the 1st percentile internal human equivalent
concentration rather than the mean of the population distribution, as predicted by the human
PBPK model, sufficiently account for toxicokinetic uncertainties, which include those from
metabolism.
Comments: One reviewer reiterated points previously presented pertaining to the use of PBPK
modeling to account for interspecies differences in toxicokinetics and a scaling factor to account
for interspecies differences in clearance of metabolites, and questions concerning limitations of
the dose metric that is available for use in the PBPK models (i.e., a metric based on the rate of
metabolite production, rather than tissue concentration).
Response: Responses to these comments are provided under PBPK Charge Questions Alb and
A2b and RfD Charge Question B4.
C. Carcinogenicity of Dichloromethane
Cl. Under the EPA's 2005 Guidelines for Carcinogen Risk Assessment
(www.epa.gov/iris/backgrd.html), dichloromethane is likely to be carcinogenic to humans
by all routes of exposure. Is the cancer weight of evidence characterization scientifically
supported and clearly described?
Comments: Three reviewers indicated that the descriptor of dichloromethane as "likely to be
carcinogenic to humans by all routes of exposure" was scientifically justified and clearly
described; a fourth reviewer noted this was not a primary area of expertise but that the document
provided a clear description of the information.
A-25 DRAFT - DO NOT CITE OR QUOTE
-------
Three reviewers disagreed with the "likely" cancer weight-of-evidence categorization.
One of these reviewers stated that the limited evidence of animal carcinogenicity and largely
negative epidemiology data more appropriately supported a descriptor of "possible human
carcinogen." The second reviewer stated that existing data better supports a descriptor of
"suggested to be carcinogenic to humans" and the third reviewer considered dichloromethane
"unlikely to be carcinogenic to humans via the oral exposure pathway." Reasons provided by
these reviewers for considering that the available data did not support a descriptor of "likely to
be carcinogenic to humans" included the following:
(1) Two reviewers questioned the relevance of liver tumors in male B6C3Fi mice to humans,
noting the high incidence of spontaneous liver tumors in male mice of this strain and the greater
GST metabolic activity in the mouse that should result in much greater susceptibility of the
mouse to dichloromethane-induced hepatocarcinogenicity than humans. One of these reviewers
also commented that the higher alveolar ventilation rate, cardiac output, and dichloromethane
blood:air partition coefficient in the mouse would lead to a greater systemic absorption of
inhaled dichloromethane, and thus higher internal doses, in mice compared with rats and in rats
compared with humans.
(2) One reviewer noted that there is a greater prevalence of Clara cells, which contain relatively
high levels of CYP2E1 and GST-T1, in the bronchioles of the mouse compared with rats and
humans, and that this difference should be considered when evaluating the relevance of lung
tumors seen in the mouse bioassay.
(3) One reviewer noted the negative results from the rat oral exposure study (Serota et al.,
1986a), and one reviewer noted that "no evidence of carcinogenicity" was seen in two animal
species (in oral and inhalation studies in rats, and in an inhalation study in hamsters). This
reviewer stated that the findings that at least 2 species were negative for tumors is suggestive that
the agent may not be a carcinogenic concern to humans, and that a classification of "suggested to
be carcinogenic to humans" would be better supported by the data.
(4) One reviewer characterized EPA's analysis of the Serota et al. (1986b) drinking water study
in the B6C3Fi mouse as a "reanalysis" using a different statistical approach and control groups
than had been used by Serota et al. (1986b) which leads to a small but statistically significant
increase in liver tumors. This reviewer did not agree with EPA's approach, but instead agreed
with the interpretation by the authors that dichloromethane was negative for carcinogenicity by
the oral route of exposure. Another reviewer also noted that the dose-response pattern seen in
mice was considered negative by Serota et al. (1986b), and that presence of statistically
significant findings in "some, but not all of doses" is a weak finding. This reviewer also pointed
to the similarity between the results in the exposed groups and data from historical controls as an
additional factor that weakens the findings.
A-26 DRAFT - DO NOT CITE OR QUOTE
-------
(5) One reviewer indicated that more weight should be given to the epidemiology studies that
"did not provide clear, statistically significant evidence of hepatic and/or lung tumors" in
dichloromethane-exposed workers.
Response: The cancer weight-of-evidence descriptor for dichloromethane of "likely to be
carcinogenic to humans" was retained based on evidence presented in Section 4.7. Responses to
the reviewers' comments follow and clarifications to Toxicological Review were incorporated as
appropriate:
(1) B6C3Fi mice are relatively susceptible to liver tumors and mouse liver tumors can occur with
a relatively high background. For these reasons, use of mouse liver tumor data in risk
assessment has been a subject of controversy (King-Herbert and Thayer, 2006). EPA, in the
absence of mode-of-action data or other information that establishes lack of human relevance,
considers mouse liver tumors (as well as other rodent tumors) to be relevant to humans.
Interspecies differences in the metabolic activity of GST (which is relatively higher in the mouse
compared with humans) do not indicate a lack of human relevance but rather species differences;
however, the PBPK modeling accounts for interspecies differences in the amount of the relevant
metabolite(s) formed (which is relatively higher in the mouse). A discussion of these issues
related to mouse liver tumor relevance was expanded in the Toxicological Review Section
4.7.3.2 (General Conclusions About the Mode of Action for Tumors in Rodents and Relevance
to Humans).
(2) The net difference in total lung CYP activity (versus liver) in humans and rats versus mice,
which is at least partly due to differences in Clara cell prevalence, is accounted for by the PBPK
model. The role of Clara cells per se in the development of mouse lung tumors is not known and
reliably quantifying or predicting the micro-dosimetry in and around individual Clara cells is
beyond the current state of knowledge. (At least one PBPK model exists, for styrene, that
attempts to predict dosimetry in the Clara cell, but the data to validate such a model are not
available.)
(3) The EPA's Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a) indicate that
negative data from two animal species can suggest a lack of carcinogenic concern for humans.
This guidance document also notes, however, that the descriptor is based on the synthesis of all
available data, that is data from positive and negative animal assays, mode of action information,
and data from epidemiological studies. Thus, the results from the rat and hamster studies do not
in themselves serve as a basis for a classification of "suggested to be carcinogenic to humans," as
they must be viewed within the context of the other data, as described in Section 4.7 (Evaluation
of Carcinogenicity). With respect to the interpretation of the rat data, although EPA agrees that
the oral rat study (Serota et al., 1986a) is negative with respect to liver carcinogenicity, it is not
clear that all of the rat data should be categorically described in this way. As noted by another
A-27 DRAFT - DO NOT CITE OR QUOTE
-------
reviewer, the potential relevance of the rat leukemia and mammary tumor data should also be
considered.
(4) The analysis that was described by a reviewer as a EPA reanalysis was contained in the
original full report of the 2-year mouse bioassay prepared by Hazleton Laboratories (1983). The
paper based on these data, published by Serota et al. (1986b), inaccurately reported that the cut-
point for statistical significance testing was 0.05, rather than the value of 0.0125 that was used.
Although it is true that the increase in the liver tumor incidence in the 60 mg/kg-day group was
not statistically significant, the presence of statistically significant increases in the 100-, 125-,
and 250-mg/kg-day exposure groups does not suggest a random pattern of response. A plateau
in the response at the highest dose is seen, as was described by EPA, but this plateau does not
negate the pattern across the dose groups. With respect to comparisons with historical controls,
the incidence of liver tumors in the control groups (19%) was almost identical to the mean seen
in the historical controls from this laboratory (17.8% based on 354 male B6C3Fi mice). This
comparison to historical control data provides no indication that the observed trend is being
driven by an artificially low tumor rate in controls and no indication that the experimental
conditions resulted in a systematic increase in the incidence of hepatocellular adenomas and
carcinomas. For these reasons, EPA considers the characterization of the Serota et al. (1986b)
results in Section 4.7.2 to be accurate and transparent.
(5) There are several limitations in the available cohort studies that EPA considered in evaluating
the data from these studies. These limitations, summarized in Section 4.1.3.7, include the low
statistical power, particularly within the context of relatively rare cancers (including liver cancer
and hematopoietic cancers), the lack of women and thus lack of data pertaining to breast cancer,
the use of mortality rather than incidence data (of particular concern for cancers with a relatively
high survival rate, such as non-Hodgkin lymphoma), missing job history data for a majority of
cohort members in one study (Lanes et al., 1993), and the inability to create an inception cohort
in another study (Gibbs et al., 1996).
Comment: One reviewer who supported the cancer classification of "likely to be carcinogenic to
humans" recommended additional discussion of the range of sites observed in human studies
(specifically, hematopoietic cancers). In particular, this reviewer noted the focus of the
assessment on lung and liver tumors could be too narrow and has implications for comparisons
of potency based on the animal data.
Response: Additional epidemiological studies of leukemia and lymphoma risk in workers
exposed to dichloromethane were identified and added to Section 4.1.3.6.6 (Case-control studies
of lymphoma, leukemia, and multiple myeloma), and the discussion of the NTP (1986)
mononuclear cell leukemia results was expanded in Section 4.2.2.2.1 (Chronic inhalation
exposure in F344/N rats (Mennear et al., 1988; NTP, 1986)). In addition, as noted in response to
PBPK Charge Question A3a, the issue of site concordance and implications with respect to
A-28 DRAFT - DO NOT CITE OR QUOTE
-------
toxicokinetic modeling and uncertainties were also added to Section 4.7.3.1 (Hypothesized Mode
of Action) and Section 5.4.5 (Uncertainties in Cancer Risk Values), respectively.
C2. A mutagenic mode of carcinogenic action is proposed for dichloromethane. Please
comment on whether this determination is scientifically supported and clearly described.
Please comment on data available for dichloromethane that may support an alternative
mode of action.
Comments: Five reviewers agreed that a mutagenic mode of carcinogenic action is supported.
One reviewer observed that the in vitro data are probably sufficient to conclude that
dichloromethane is an in vitro mutagen, but there are no in vivo data in rodents to address
whether dichloromethane can induce mutations either in target or nontarget tissues. This
reviewer concluded that therefore the data are insufficient to prove that dichloromethane acts via
a mutagenic mode of carcinogenic action. The same reviewer indicated that there is no evidence
to strongly conclude that dichloromethane acts via a nonmutagenic mode of action and that one
must therefore conclude that the MOA for tumor induction is unknown. This reviewer
commended the presentation of the series of tables of genotoxicity data and also suggested that a
table(s) be added that would summarize, for a particular rodent species, target tissue, and
exposure route, data outlining key events in the mode of action timeline to support analysis of
temporality and dose-response concordance. Another reviewer noted that the evidence
pertaining to DNA damage comes from studies of very high exposures (inhalation or oral) to
dichloromethane, and would be expected to be minimal or negligible at very low exposures or in
species with low GST-T1 activity. This reviewer suggested that discussion of these points
within the mode of action section would result in a more balanced presentation of the positive
and negative findings, specifically with respect to the relevance of the mode of action to humans.
Response: The conclusion that dichloromethane induces cancer by a mutagenic mode of action
was retained; Section 4.7.3 (Hypothesized Mode of Action) was revised consistent with the
reviewers' suggestions, specifically noting key limitations of the available evidence (e.g., lack of
studies demonstrating induction of specific mutations in vivo, inability to detect highly reactive
DNA adducts in vivo, much of the evidence comes from high-dose studies in mice, the greater
GST activity in mice compared with humans), and making a clearer distinction between
measures of mutagenicity and measures of genotoxicity. Despite the specific data gaps noted,
the overall weight of the evidence supports the conclusion that dichloromethane operates through
a mutagenic mode of action. Therefore, EPA disagrees with one reviewer's determination that in
the absence of sufficient data to definitively prove a mutagenic mode of action, it must be
concluded that the mode of action for dichloromethane-induced tumors is unknown. In addition,
EPA concluded that the available studies provide sufficient data to indicate that the hypothesized
A-29 DRAFT - DO NOT CITE OR QUOTE
-------
mutagenic mode of action is relevant to humans. With regard to the suggestion for an additional
table, EPA notes that Table 4-34 presents in vivo mutagenicity and genotoxicity assays by cell
type and species; a new table (Table 4-44) summarizing the data pertaining to strength,
consistency, and specificity of association, dose-response concordance, and temporality was
added to Section 4.7.3.1.1 (Experimental support for the hypothesized mode of action). The
data from human cell lines were also added to this section (Olvera-Bello et al., 2010).
Quantitative cancer assessment - oral exposure
C3. A 2-year drinking water study in mice (Scrota et al., 1986b) was selected for the
derivation of an oral slope factor (OSF) for dichloromethane. Please comment on whether
the selection of this study for quantitation is scientifically supported and clearly described.
Please identify and provide the rationale for any other studies that should be considered.
Comments: Four reviewers supported the use of the 2-year drinking water study in mice (Serota
et al., 1986b) for quantitation, although one of these reviewers noted the marginal statistical
significance of the tumor incidences. One reviewer considered the discussion pertaining to this
choice to be clear, but that it was not within the reviewer's primary area of expertise. Two
reviewers disagreed with the use of Serota et al. (1986b). One of these reviewers disagreed with
the "reanalysis" of the mouse liver tumor findings by EPA, and supported the interpretation by
Serota et al. (1986b), i.e., that dichloromethane was negative for carcinogenicity by the oral route
of exposure. This reviewer, and one other reviewer, suggested that the chronic inhalation study
by NTP be used as an alternative basis for the OSF (through use of route-to-route extrapolation
using PBPK modeling). One reviewer also suggested that the analysis of combined datasets
from oral and inhalation routes, or analysis based on the arithmetic mean of slope factors for the
two exposure routes, should also be considered.
Response: Comments on EPA's analysis of the Serota et al. (1986b) study are addressed under
Carcinogenicity Charge Question Cl. EPA presented a route-to-route extrapolation in Section
5.4.1.6 (Alternative Derivation Based on Route-to-Route Extrapolation). The rationale for
selection of the oral slope factor based on the oral exposure study over the value derived by
route-to-route extrapolation was clarified in Section 5.4.5 (Uncertainties in Cancer Risk Values).
C4. The OSF was calculated by linear extrapolation from the POD (lower 95% confidence
limit on the dose associated with 10% extra risk for liver tumors in male mice). The OSF is
based on an analysis of the most sensitive of the human subgroups, the GST-T1+/+
genotype, using mean internal dose predictions for that subgroup. Please comment on
A-30 DRAFT - DO NOT CITE OR QUOTE
-------
whether this approach is scientifically supported and clearly described. Has the modeling
been appropriately conducted and clearly described?
Comments: Three reviewers agreed with the extrapolation approach and OSF calculation
developed by EPA. Two reviewers disagreed with the use of linear extrapolation, indicating that
this approach ignores repair systems and other biological processes that would make a threshold
model more appropriate. One reviewer did not comment on this question because it was outside
this reviewer's area of expertise.
Response: Consideration of repair systems and other biological processes that occur at low
doses would require a biologically-based (toxicodynamic) model that accounts for the biological
processes involved in a response. However, sufficient data to develop a biologically-based
toxicodynamic model for dichloromethane are not available. As noted in EPA's Guidelines for
Carcinogen Risk Assessment (U.S. EPA, 2005a), "In the absence of data supporting a
biologically based model for extrapolation outside of the observed range, the choice of approach
is based on the view of mode of action of the agent arrived at in the hazard assessment. If more
than one approach (e.g., both a nonlinear and linear approach) are supported by the data, they
should be used and presented to the decisionmaker." Dichloromethane is hypothesized to have a
mutagenic MOA for carcinogenicity, which is consistent with application of a linear
extrapolation approach. Therefore, consistent with 2005 Cancer Guidelines and in the absence
of data to support a toxicodynamic model, a linear extrapolation was retained.
Comments: Four reviewers agreed with the use of the mean internal dose predictions based on
the GST-T1+ + genotype, the group that is thought to be most sensitive to the carcinogenicity of
dichloromethane (mediated through the GST metabolic pathway). One reviewer did not consider
derivation of the OSF for the most sensitive population to be clearly justified, indicating the need
for clarification regarding the "realism and scientific validity of this approach in light of the use
of a probabilistic PBPK model that already accounts for the population distribution of parameters
of relevance."
Response: The probabilistic PBPK model allows the prediction of a distribution of internal
doses in the U.S. population as a whole. To determine a cancer slope factor, one must select a
point or percentile from within that overall distribution at which the cancer risk is to be assessed.
In assessments without probabilistic models, one is effectively attempting to determine the
overall average population cancer risk, assuming no additional knowledge to inform
differentiation of risk among the population. This average population risk would correspond to
the mean internal dose, or 50* percentile, of the model-predicted distribution for the population
as a whole. Because there is information about sub-population sensitivity for dichloromethane
A-31 DRAFT - DO NOT CITE OR QUOTE
-------
that is not typically available, the choice was made to estimate risk for the average (i.e., mean) of
the GST-T1+ + sub-population. This issue is also discussed under PBPK Modeling Charge
Question A3 a.
Comments: One reviewer noted that although the use of the GST-T1+ + genotype as the most
sensitive group was justified, the use of the tissue-specific dose level in the liver (for GST-T1+/+)
largely ignores the possibility of tumor sites other than liver, which could be proportionally more
significant in the human than the mouse. This reviewer suggested that this issue be addressed in
the discussion of uncertainties.
Response: As noted in response to PBPK Modeling Charge Question A3 a and Carcinogenicity
Charge Question Cl, discussion of the issues of site concordance and GST activity in various
tissues was added to Section 4.7.3.1 (Hypothesized Mode of Action) and Section 5.4.5
(Uncertainties in Cancer Risk Values).
Quantitative cancer assessment - inhalation exposure
C5. A 2-year cancer bioassay in mice (NTP, 1986) was selected for the derivation of an
inhalation unit risk (IUR) for dichloromethane. Please comment on whether the selection
of this study for quantitation is scientifically supported and clearly described. Please
identify and provide the rationale for any other studies that should be considered.
Comments: Six reviewers agreed with the use of the NTP (1986) 2-year inhalation cancer
bioassay for the derivation of the inhalation unit risk. One of these reviewers, however, repeated
concerns (also expressed in previous charge questions) regarding the relevance of mouse liver
tumors to humans, the relatively high number of Clara cells in mouse bronchioles, and the
relatively high GST activity in the mouse compared with humans. One reviewer reiterated the
recommendation that additional discussion of issues regarding interspecies extrapolation at sites
other than liver and lung be addressed in the discussion of uncertainties. One reviewer did not
comment on this question because it was outside this reviewer's area of expertise.
Response: The NTP (1986) bioassay was retained as the principal study for derivation of the
IUR. Responses to the issues raised by two reviewers are addressed under Carcinogenicity
Charge Question 1.
C6. The IUR was calculated by linear extrapolation from the POD (lower 95% confidence
limit on the dose associated with 10% extra risk for lung or liver tumors in male mice)
taking into consideration total cancer risk by determining the upper bound on the
A-32 DRAFT - DO NOT CITE OR QUOTE
-------
combined risk for male lung and liver tumors. The IUR is also based on the analysis of the
most sensitive of the human subgroups, the GST-T1+/+ genotype, using mean internal dose
predictions for that subgroup. Please comment on whether this approach is scientifically
supported and clearly described. Has the modeling been appropriately conducted and
clearly described?
Comments: Four reviewers generally supported the extrapolation and modeling approach used
by EPA, but three of these reviewers posed questions about the procedures used for combining
risk across the two tumor sites (comments discussed in more detail below). One reviewer
disagreed with the use of the linear extrapolation (as also noted in response to Carcinogenicity
Charge Question C4).
Response: A response pertaining to the use of the linear extrapolation is provided under
Carcinogenicity Charge Question C4.
Comments: Four reviewers agreed with the use of the mean internal dose predictions based on
the GST-T1+/+ genotype, the group that is thought to be most sensitive to the carcinogenic
capacity of dichloromethane (mediated through the GST metabolic pathway). One reviewer
repeated the recommendation offered in response to other charge questions that the uncertainties
section should provide additional discussion of issues regarding interspecies extrapolation at
sites other than liver and lung. Two reviewers did not comment on this question.
Response: The site concordance issue was addressed in response to PBPK Charge Question A3a
and Carcinogenicity Charge Question Cl.
Comment: One reviewer did not support the use of the sensitive population (as had also been
expressed in response to Carcinogenicity Charge Question C4). This reviewer recommended
instead that the mean of the entire population, using a modeling approach that specifically
addressed sources of variability in the population, be used rather than the mean in the most
sensitive group (the GST-T1+/+ genotype). This reviewer also asked why the cancer potency
derivations were based on the mean value of the sensitive population, and the noncancer
reference values were based on the 1st percentile of the human equivalent doses or concentrations
in the full population (all GST-T1 genotypes).
Response: By definition, the RfD and RfC should be protective of the entire population,
including sub-populations; this point and the use of the 1st percentile specifically are discussed at
greater length in the response to points under PBPK Charge Question A3a. Cancer risks,
however, have historically been estimated based on an average population risk. As discussed in
A-33 DRAFT - DO NOT CITE OR QUOTE
-------
response to comments under Carcinogenicity Charge Question C4, in the case of
dichloromethane, average cancer risk was estimated for a sensitive group based on genotype, a
group that is estimated to represent approximately one-third of the entire population. Focus on
this portion of the population, which would be considered a higher risk population, is warranted
based on current understanding of genetic variation, metabolism, and mode of action.
Comments: One reviewer did not think the use of a risk estimation based on the combined risk of
lung and liver tumors was justified. Two reviewers raised concerns regarding the assumption of
normality of the underlying risk distributions of the liver and lung tumors. One of these
reviewers noted that the procedure assumes a normal distribution of the BMD and assumes
independence (i.e., no correlation) between the presence of lung and liver tumors. If these
assumptions were not met, the 95% confidence limit calculated for the combined lung and liver
tumors risk is approximate. This reviewer suggested an alternative approach, based on
calculation of the incidence of lung and/or liver tumors in each individual mouse in the NTP
(1986) study (using the individual mouse data in the appendices), and using benchmark dose
modeling of this combined risk. The other reviewer noted the potential for the lack of normality
in the probability distributions to result in an underestimation of the resulting 95% upper
confidence level on the summed risk. This reviewer provided a reference (Salmon and Roth,
2010) that discusses this issue in more detail.
Response: In situations in which tumors at multiple sites are seen in an animal bioassay, a unit
risk estimate based on tumor incidence at only one of these sites will result in an underestimation
of the total carcinogenic potency. Consistent with the recommendations of the NRC (1994) and
with EPA's Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), a statistically
appropriate upper bound on composite risk was estimated to gain understanding of composite
risk across multiple tumor sites. This rationale is noted in the discussion of the combined risk
(Section 5.4.2.5, Cancer IUR). EPA agrees that the procedure used is approximate, and EPA
considered the suggestion of using a calculation based on number of tumor-bearing animals for
an estimate of combined tumor risk. This type of calculation may not be appropriate, however, if
as suggested in the NTP (1986) mouse data, independence between tumor sites is present (NRC,
1994; Bogen, 1990). EPA reviewed the paper by Salmon and Roth (2010), and found that it
supports the procedure used by the EPA in the situation presented by the dichloromethane
assessment. Specifically, the combined risk was generated from the first order models for male
mice. As noted by Salmon and Roth (2010), "In simple cases where the majority of the variation
in the data is explained by a linear model, the resulting distribution shape is a single peak,
although this may have a long tail at the high end of the estimate range depending on how much
unexplained variation is present. Where the separate tumor incidences to be added together all
fit this description, the procedure used by U.S. EPA (2002) of assuming a normal shape of the
A-34 DRAFT - DO NOT CITE OR QUOTE
-------
likelihood density function in order to predict a 95% confidence limit on the combined
distribution will produce reasonable, if not exact, results."
Additional Reviewer Comments
Comment: One reviewer asked why the PBPK model development in humans included ages <18
as a specific group for scaling Vmax, given available data indicating that CYP2E1 expression
increased only up to age 3 months, and given the larger liver size (normalized to body weight) in
young children.
Response: Use of the age group categorized as <18 years was chosen because this represents the
age range during which growth occurs. While Johnsrud et al. (2003) showed that the activity per
mg microsomal protein in infants (<3 months) was not significantly different from adult levels,
assuming the mg protein/g liver is about constant, this means that the total CYP2E1 activity in a
person will increase with age, as the size of the liver increases. Rather than assuming direct
proportionality with liver or body weight, a power-function was fit to the data of Johnsrud and
found that the relationship was more closely represented by direct proportionality (coefficient =
0.88) than the allometric coefficient that was otherwise used in the model by David et al. (2006)
(coefficient = 0.7). The relationship used by David et al., which yields a higher activity per g
liver in the child than the adult, over-predicted the data of Johnsrud et al. The equation was
adjusted to match those data. Figure B-l 1 (upper panel) shows that liver size decreases with age
up to 17 years, based on data. Lacking data on changes in the fraction of blood flow to the liver
with age, the distribution function used to describe that fraction has no dependence on age of the
individual.
Comments: One reviewer also raised the question of the reversibility of the neurological effects
seen with dichloromethane, as well as other solvents, noting the relatively short recovery periods
observed post-exposure in animal studies.
Response: Lash et al. (1991) is the only human study of dichloromethane that supports an
examination of neurotoxicity endpoints following a period of no exposure to the chemical. Lash
et al. (1991) examined the effects of dichloromethane among retired workers (mean length of
retirement among the study participants was 5 years) and found decrements in attention and
reaction time in complex tasks among the retired workers. Although limited by small sample
size and low statistical power of the study, the findings of Lash et al. (1991) suggest that
neurological effects of dichloromethane may not be fully reversible. EPA has not identified
other studies of dichloromethane or other chlorinated solvents (e.g., trichloroethylene and
tetrachloroethylene) that were designed specifically to examine the long-term effects after
A-3 5 DRAFT - DO NOT CITE OR QUOTE
-------
cessation of a solvent exposure. EPA does not consider the available database adequate to
support conclusions about the reversibility of neurologic effects of dichloromethane.
PUBLIC COMMENTS
Comments: A commenter supported the application of the human probabilistic PBPK model
used in the assessment, including incorporation of variability in CYP2E1 activity and variation in
GST-T1 activity that results from the different GST-T1 genotypes. This commenter also
supported use of cancer potency values derived for the population presumed to have the greatest
sensitivity to carcinogenic effects via the GST-mediated metabolic pathway, i.e., the GST-T1+ +
genotype. This commenter raised several issues, some of which were also addressed in
comments by the review panel, as delineated below:
1) In situations where the dose metric is a rate of metabolism (production) rather than a tissue
concentration, the commentator asked if a PBPK model could be used to provide the interspecies
conversion instead of using the scaling factor.
Response: The rationale for the use of the scaling factor was discussed in response to PBPK
Charge Questions Alb and A2b. It is because of the absence of experimentally-derived, species-
specific data pertaining to the relationship between clearance of CYP metabolites in rodents and
humans that the scaling factor based on body weight0'75 scaling is warranted. EPA agrees that if
these types of data were available, it would be preferable to use them to evaluate and potentially
modify or eliminate the scaling factor.
2) A commenter questioned the use of liver effects in rats for the derivation of reference values,
stating that these effects were of low toxicological concern and that evidence of liver effects in
humans was lacking. The commenter suggested that COHb (resulting from CYP2E1 metabolism
of dichloromethane) should be evaluated as a relevant health effect, using the standards for CO.
The commenter also supported the use of the human neurotoxicological data from Cherry et al.
(1983) and Lash et al. (1991) for derivation of the RfC, and suggested that a route-to-route
extrapolation from these data be used for to derive RfD.
Response: As noted in response to RfC Charge Question B5, increased CO and COHb have
implications for neurotoxicity as well as cardiovascular disease risk, and EPA agrees that the
potential for a significant health risk due to dichloromethane-generated CO should be
considered. EPA's evaluation is described in response to Charge question B4. EPA maintains
however, that the liver effects observed in the rat are of toxicological concern. Although the
three epidemiological studies examining serologic measures of liver function (Soden, 1993;
A-36 DRAFT - DO NOT CITE OR QUOTE
-------
Kolodner et al., 1990; Ott et al., 1983c) do not provide clear evidence of hepatic damage in
dichloromethane-exposed workers, these data are limited and the absence, presence, or extent of
hepatic damage has not been established. EPA agrees that the Lash et al. (1991) data are the
most relevant neurological data available, since this is a study examining persistent effects in
retired workers (mean 5 years post-retirement). The potential RfC derived based on these data
(0.55 mg/m3) is essentially the same as the RfC derived from hepatic lesions in the rat (0.6
mg/m3).
As noted in response to Carcinogenicity Charge Question C3, EPA presented a route-to-
route extrapolation in Section 5.4.1.6 (Alternative Derivation Based on Route-to-Route
Extrapolation). The basis for selection of the oral slope factor based on the oral exposure study
over the value derived by route-to-route extrapolation was clarified in Section 5.4.5
(Uncertainties in Cancer Risk Values).
Within that framework, EPA explored the suggested option of estimating an RfD using a
route-to-route extrapolation from the Lash et al. (1991) data. This procedure required several
steps and assumptions. To calculate an average internal dose for the exposed worker population,
the dose metric used was area-under-the-curve (AUC) of the blood concentration of parent
dichloromethane. Since the effects observed in Lash et al. (1991) were long-term, it was
assumed that this average AUC was a better predictor of effect than peak concentration. It was
also assumed that the workers were exposed 8 hours/day, 5 days/week, and the AUC was
averaged over a full week. The distribution of ages over which the workers were exposed was
estimated from the demographic details provided by Lash et al. (1991), and the internal dose for
each sample individual was estimated at the mid-point of his exposure period. The average age
of the simulated worker population at time-of-exposure was 50.3 years. All other model
parameters were then sampled randomly from the distributions otherwise used for the PBPK
model. The estimated mean blood dichloromethane AUC for the exposed workers was 11.1 mg-
h/L. The corresponding distribution of oral exposures for the population as a whole was then
estimated, using the model distributions for the full U.S. population and the assumed oral
ingestion pattern (total intake divided among six boluses over the course of the day), for a
sample of 10,000 individuals. The mean, median, 5*, and 1st percentiles of that distribution were
12.4, 11.1, 5.7, and 4.3 mg/kg-d, respectively. Using the resulting 1st percentile exposure rate
(4.3 mg/kg-d) allows the PBPK model to account for toxicokinetic variability and uncertainty. A
total UF of 100 would then be applied: 3 for intrahuman toxicodynamic uncertainty; 10 for
LOAEL to NOAEL extrapolation; and 3 for database uncertainty (as used for the RfD based on
rat lesions as reported by Serota et al., 1986a). The resulting alternative RfD of 0.04 mg/kg-day
is approximately sevenfold higher than the RfD based on rat liver lesions (0.006 mg/kg-day).
The primary uncertainty in this analysis is that it is based on effects seen an average of 5 years
post-exposure, and so does not capture what is likely to be stronger effects seen at the time of
exposure. Other uncertainties in this analysis are introduced by attempting to simulate the
A-37 DRAFT - DO NOT CITE OR QUOTE
-------
demographics of the exposed workers, the choice of exposure-period mid-point (versus age in
last year of exposure, for example), and the choice of dose metric (dichloromethane AUC rather
than metabolite production). The latter two uncertainties stem from lack of knowledge of the
mechanism for the neurological effect. This analysis demonstrates that the RfD based on rat
liver lesions should also be protective of potential neurological effects. Given the uncertainty in
the level of exposure and degree of response evaluated by Lash et al. (1991), as well as the
approximation of worker demographics, this analysis was not added to the Toxicological
Review.
3) A commenter questioned the approach to the derivation of the RfD and RfC, including the use
of the BMDLio (rather than the BMD) in conjunction with the scaling factor to account for
interspecies differences in metabolite clearance, use of the 1st percentile of the human equivalent
dose (or concentration), and use of UFs to account for interspecies and intraspecies
toxicodynamic variability.
Response: The methodology used to derive the RfD and RfC is consistent with EPA's A Review
of the Reference Dose and Reference Concentration Processes (U.S. EPA, 2002). The scaling
factor is used in place of the toxicokinetic component of the default UF for interspecies
extrapolation because there is no PBPK sub-model for the toxic metabolite(s) to otherwise
extrapolate the metabolite dose-rate from rodents to humans. The 1st percentile of the
distribution of human exposures (rather than an average human equivalent exposure) is used
instead of the toxicokinetic component of the default UF for intraspecies variability. In the
absence of chemical-specific information, it is appropriate to use UFs to account for interspecies
and intraspecies toxicodynamic variability.
4) A commenter provided a set of comments that had also been submitted as a public comment
on the draft IRIS assessment of trichloroethylene relating to the use of the 1st percentile of the
human equivalent dose (and concentration) for the derivation of the RfD and RfC. The
commenter stated that the allowance for inter-human toxicokinetic variability double counts and
misconstrues the nature of the dose-response curve. In particular, the commenter stated if one
assumes a tolerance distribution-based dose-response curve for quantal endpoints, an
experimental animal dose corresponding to a low (e.g., 1%) response may reflect either
toxicokinetic or toxicodynamic sensitivity, with no readily available way to split the components
out. The commenter stated that the method employed in the assessment seems to implicitly
assume that all of the variability in dose-response among rats is attributable to toxicodynamics
and that all the variation in response at a given dose in humans is attributable to toxicokinetics.
The commenter concluded that the net result is to yield an RfC that is overcorrected for human
inter-individual variation to a degree that is not possible to know with the available analyses.
A-3 8 DRAFT - DO NOT CITE OR QUOTE
-------
Response: The method employed neither assumes that all of the variability in dose-response
among rats is due to toxicodynamics, nor that all of the variation among humans is due to
toxicokinetics. EPA's approach uses the BMDLio determined from rat dose-response data (i.e.,
the lower confidence limit on a measure of dose or exposure expected to result in a 10%
response in rats), not a 1% response level. As described in greater detail in EPA's draft
Benchmark Dose Technical Guidance Document (U.S. EPA, 2000b), the BMDLio is used as a
point of departure for reference value derivation in place of a NOAEL or LOAEL; hence the
same suite of UFs and potential replacements apply. In particular, these include a UFA of 10 for
animal-to-human extrapolation and UFn of 10 for variability among humans. The commenter is
in part suggestingthat some of the uncertainty and variability accounted for by these UFs is
already accounted for by use of the BMDL instead of the BMD. This is not the case; however,
use of the lower bound on the BMD accounts for uncertainty inherent in a given study (e.g., a
smaller number of animals used in a given study results in a relatively lower BMDL than a study
with a larger number of animals).
For the noncancer assessment, the probabilistic human PBPK model is assumed to
explicitly account for toxicokinetic variability and uncertainty among the human population;
distributions of HECs and HEDs are thereby calculated for humans. Use of the 1st percentile
rather than the mean of a HEC or FED distribution replaces the toxicokinetic portion of the UFn
so this UF is reduced from 10 to 3, with the remaining UF of 3 accounting for toxicodynamic
uncertainty. An alternative would be to use the mean FIEC or FLED from the distribution
calculated and to then apply the full UFn of 10; this alternative would result in lower RfD and
RfC values than those derived in the current assessment, all other factors being kept the same.
Thus, the total correction for interindividual human variation is less than would be obtained with
a non-probabilistic PBPK analysis.
The comment that the RfC over-corrects for human interindividual variation assumes that
the commenter knows that the variation is less than the correction actually used. While EPA
recognizes that the probabilistic human PBPK model may overpredict the toxicokinetic
variability to a small extent, given the extensive data incorporated into this model any such error
is expected to be small. Since the actual extent of human toxicodynamic variability is unknown,
it could be larger than the remaining UFn of 3. The commenter's conclusion that the RfC is
overcorrected is therefore unsupported by any data.
5) A commenter stated that application of a database UF >1 was not justified because ATSDR
did not identify developmental neurotoxicity as a priority data need for dichloromethane, and
because the Aranyi et al. (1986) data pertaining to lung resistance to infection and bactericidal
activity should not be considered an effect because it was based on an acute exposure (3-hour,
100 ppm exposure) and the concurrent controls had a low mortality rate.
A-39 DRAFT - DO NOT CITE OR QUOTE
-------
Response: In the 2000 ATSDR Toxicological Profile for Methylene Chloride, data needs are
identified by category. The categories that were evaluated were: genotoxicity, reproductive
toxicity, developmental toxicity, immunotoxicity, neurotoxicity, epidemiological and human
dosimetry studies. Specific data needs were identified by ATSDR for
reproductive/developmental, neurotoxicological, and immunotoxicological endpoints, which is in
agreement with this Toxicological Review. Developmental neurotoxicity was not a category that
was specifically evaluated in the ATSDR Toxicological Profile.
The variability in mortality rates and bactericidal activity in the Aranyi et al. (1986) study
was discussed in the Section 4.4.2 (Immunotoxicity Studies in Animals). A recent study by
Selgrade and Gilmour (2010) used a similar method to Aranyi et al. (1986). Susceptibility of
CD-I mice to respiratory infection and mortality due to Streptoccocus zooepidemicus exposure
and the ability of pulmonary macrophages to clear bacterial infection were examined in mice
exposed for 3 hours to trichloroethylene at concentrations of 5, 10, 25, 50, 100, or 200 ppm or
chloroform at concentrations of 100, 500, 1000, or 2000 ppm. This study supports the
methodology used by Aranyi et al. (1986), and suggests that these types of immunosuppressive
endpoints may represent sensitive effects of chemical exposure.
6) A commenter stated that EPA did not adequately value the contribution of cohort studies and
also did not fully acknowledge limitations of case-control studies, in particular possible
dichloromethane exposure misclassification in the Heineman et al. (1984) study. A paper
describing these issues by Dell et al. (1999)10 was referenced by the commenter.
Response: EPA expanded a discussion of exposure assessment issues in the description of the
Heineman et al. (1984) case-control study (Section 4.1.3.6.1).
7) A commenter noted that the genotoxic effects of dichloromethane are due to the GST-T1
metabolic pathway, and that these effects are not likely to be relevant to humans because of the
higher activity of this pathway in mice compared to humans.
Response: As noted in response to Charge Question Cl, interspecies differences in the metabolic
activity of GST-T1 do not establish a lack of human relevance. The PBPK modeling accounts
for interspecies differences in the amount of the relevant metabolite(s) formed.
8) A commenter disagreed with EPA's interpretation of the Serota et al. (1986b) oral exposure
bioassay data. The commenter did not believe this study should be used to derive the oral slope
1 The commenter referred to "Dell et al. (2003)", but the correct reference for the paper is Dell et al. (1999).
A-40 DRAFT - DO NOT CITE OR QUOTE
-------
factor, recommending instead route-to-route extrapolation of data from the NTP (1986)
inhalation study.
Response: Please refer to response Carcinogenicity Charge Question Cl regarding EPA's
evaluation of the Serota et al. (1986b) data, based on the full report of this study prepared by
Hazleton Laboratories (1983).
A-41 DRAFT - DO NOT CITE OR QUOTE
-------
APPENDIX B: HUMAN PBPK DICHLOROMETHANE MODEL
B.I. HUMAN MODEL DESCRIPTION
The basic model structure used by David et al. (2006) was that of Andersen et al. (1987)
with the addition of the CO submodel of Andersen et al. (1991), refinements from the Marino et
al. (2006) mouse model, and an inclusion of CYP metabolism in rapidly perfused tissue (see
Figure B-l).
1 A GST-< 1 1 > CYP —
^ Gas >
Exchange
Fat
Richly
Perfused
1
Slowly
Perfused
Liver
Lung — >
CYP
*
CO Sub
Model
I t
Alveolar
Air
1 t
t
Endogenous
Production
Figure B-l. Schematic of the David et al. (2006) PBPK model for
dichloromethane in the human.
In order to incorporate known variability in human physiology and metabolism of
dichloromethane into internal dosimetry, Monte Carlo analysis of the human model was
performed to derive probability distributions of internal dose, as reported in David et al. (2006),
but with changes in some of the key distributions as described below. The shape of the resulting
dose distribution can be used to quantify the variability and uncertainty in internal dose with
respect to variability in human physiology and variability and uncertainty in dichloromethane
metabolism. The human model was run repeatedly using a random sample of each parameter
from its respective parameter distribution in each iteration. Internal doses predicted for all
iterations collectively defined a distribution for internal dose. The Monte Carlo analysis was run
for 10- or 20,000 iterations. Repeated Monte Carlo analyses (at 10,000 iterations each) yielded
99th percentile values of internal dose in the liver or lung that differed by <2%. Normal or
log-normal distributions of physiological and metabolic parameter and partition coefficient
values were described by a mean, SD, and in most cases upper and lower truncation bounds.
Physiological parameter and partition coefficient values were initially taken from the literature as
described in David et al. (2006) and their distributions were assumed to be true variability
(physiological parameters) or a level of uncertainty and variability (partition coefficients),
neither of which could be meaningfully informed by the dichloromethane pharmacokinetic data.
B-l
DRAFT - DO NOT CITE OR QUOTE
-------
Hence, the distributions for physiological parameters and partition coefficients were not updated
in the Bayesian analysis of David et al. (2006).
The first-order oral absorption rate constant, ka = 5.0/hour, was used in conjunction with
human drinking water exposures simulated as six discrete drinking water episodes for specified
times (25, 10, 25, 10, 25, and 5% of total daily intake at hours 0, 3, 5, 8, 11, and 15 of each day)
(Reitz et al., 1997). Metabolic parameter distributions were derived from multiple human data
sets by using MCMC calibration, also described in David et al. (2006).
The variability of genotypic expression of GST-T1 activity (the mechanism for
GST-mediated metabolism of dichloromethane) was simulated as a uniform discrete distribution
of the three GST-T1 genotypes (+/+, +/-, -/-) with varying activities in the liver and lung. The
genotype frequency was based on data from Haber et al. (2002), with a frequency of genotypes
of 32, 48, and 20% in the +/+, +/-, and -/- groups, respectively. GST activities measured by
Warholm et al. (1994) for the three genotypes in a group of 208 healthy male and female subjects
from Sweden were scaled by David et al. (2006) to obtain distributions of kfc for each genotype
that, when weighted by estimated frequencies of the genotypes in U.S. populations, would result
in a distribution of kfc activities with a mean equal to 0.852/hour-kg0'3, which is the mean
estimate of the population-mean value of kfc obtained from the Bayesian analysis. The resulting
distributions of internal lung and liver dose in human populations would have a theoretical
probability of 20% for zero exposure to GST-mediated metabolites, and hence zero cancer risk
for that 20% of the population. The final parameter distributions used by David et al. (2006) are
summarized in Table B-l.
B.2. REVISIONS TO PARAMETER DISTRIBUTIONS OF DAVID ET AL. (2006)
An evaluation of the David et al. (2006) model and parameterization was undertaken,
focusing on the adequacy of the characterization of parameter distributions in the full human
population. EPA's conclusion is that the reported distributions for physiological parameters in
particular, but also key metabolic parameters, only represented a narrow set of adults (with the
exception of BW) or failed to include the parameter uncertainty from the Bayesian analysis.
Therefore, supplemental data sources were chosen to define a number of the physiological
parameter distributions in a way that should fully characterize the variability in the human
population for individuals between 6 months and 80 years of age. Since many physiological
parameters vary with age and gender, a structured approach will be used where an individual's
age and then gender may be selected from the overall population distribution for these
characteristics (the male:female ratio in the population declines with age, for example).
Dosimetry simulations can then be run for each such individual to obtain an overall population
distribution of internal doses. Thus, each dosimetry distribution represents a "snapshot" of the
dosimetry in a given individual at a given age and age- and gender-appropriate sampled body
composition (fraction of BW in each tissue group). Finally, the sampling for two key metabolic
B-2 DRAFT - DO NOT CITE OR QUOTE
-------
Table B-l. Parameter distributions used in human Monte Carlo analysis for
dichloromethane by David et al. (2006)
Parameter
BW
QCC
VPR
QFC
QLC
QRC
QSC
Body weight (kg)
Cardiac output (L/hr-kg° 74)
Ventilation:perfusion ratio
Fat
Liver
Rapidly perfused tissues
Slow perfused tissues
Distribution
Mean
(arithmetic)
70.0
16.5
1.45
0.05
0.26
0.50
0.19
SD
21.0
1.49
0.203
0.0150
0.0910
0.10
0.0285
Source
Humans3
Humans3
Humans3
Humans3
Humans3
Humans3
Humans3
Tissue volumes (fraction B W)
VFC
VLC
VLuC
VRC
VSC
Fat
Liver
Lung
Rapidly perfused tissues
Slowly perfused tissues (muscle)
0.19
0.026
0.0115
0.064
0.63
0.0570
0.00130
0.00161
0.00640
0.189
Humans3
Humans3
Humans3
Humans3
Humans3
Partition coefficients
PB
PF
PL
PLu
PR
PS
Blood:air
Fatblood
Liverblood
Lung: arterial blood
Rapidly perfused tissue :blood
Slowly perfused tissue (muscle :blood)
9.7
12.4
1.46
1.46
1.46
0.82
0.970
3.72
0.292
0.292
0.292
0.164
Humansb
Ratsb
Ratsb
Ratsb
Ratsb
Ratsb
Metabolism parameters
Vmaxc
Km
Al
A2
FracR
Maximum metabolism rate (mg/hr-kg0 7)
Affinity (mg/L)
Ratio of lung Vmax to liver Vmax
Ratio of lung KF to liver KF
Fractional CYP2E1 capacity in rapidly perfused tissue
9.34
0.433
0.000993
0.0102
0.0193
1.73
0.146
0.000396
0.00739
0.0152
Calibration0
Calibration0
Calibration0
Calibration0
Calibration0
First order metabolism rate (/hr-kg° 3)
kfc
Homozygous (-/-)
Heterozygous (+/-)
Homozygous (+/+)
Oral absorption
k,
First-order oral absorption rate constant (/hr)
0
0.676
1.31
0
0.123
0.167
5.0
Calibration0
Calibration0
Calibration0
Reitz et al.
(1997) (point
estimate)
3U.S. EPA (2000d) human PBPK model used for vinyl chloride.
bAndersen et al. (1987). Blood:air partition measured by using human samples; other partition coefficients based
on estimates from tissue measures in rats.
°Bayesian calibration based on five data sets (see text for description); posterior distributions presented in this table.
Source: David et al. (2006).
B-3
DRAFT - DO NOT CITE OR QUOTE
-------
parameters representing metabolism by CYP2E1 (i.e., Vmaxc) and GST-T1 (i.e., kfc) was adjusted
to explicitly account for both the interindividual variability and the parameter-value uncertainty
among humans.
In estimating the human equivalent exposure levels for noncancer endpoints, some of
which can be presumed to be relatively short-term effects possibly occurring from exposures of
several weeks or months, using the distribution of such dosimetry "snapshots" should provide
precisely the correct distribution to estimate overall population risk. For estimating cancer risk
where risk is due to the cumulative exposure over months or years, however, the ideal approach
would be to simulate the time-course of internal doses in a given individual tracked over a
lifetime or significant portion thereof. But doing so would require estimating time-courses for
each physiological and metabolic parameter in the individual over that time-period, a task which
would be far more complicated than the structured "snapshot" approach used here. For example,
while the CYP2E1 activity in an individual at age 12 is probably predictive of the activity in that
individual at age 70 (e.g., someone who has above-average CYP2E1 activity when younger may
well continue to be above average throughout his or her lifetime), we simply do not have the
information or model structure to predict the time-dependences. Further, we know, for example,
that some individuals who are lean in their youth may become obese by middle-age, while others
(through lifestyle-changes) change in the opposite direction; and these changes may be reversed
by the time the individual reaches 70 or 80 years of age.
Therefore a "life-course" dosimetry for specific individuals has not been calculated. For
calculating HECs for noncancer effects, however, this means that the exposure level is identified
such that 99% of the population at a given time is predicted to have an internal dose at or below a
POD (defined as an internal dose level). It then seems highly likely that if one were to track
individual exposure over time, one would also find that this equivalent exposure keeps 99% of
the population from exceeding the POD. In short, if 99% of individuals' snapshot-internal doses
are below the POD (i.e., 99% of the internal doses in a cross-section of society on a given day), it
can be anticipated that no more than 1% of all people will exceed that POD at any point in their
life-times, even though the model simulations did not specifically track the changes in internal
dose with age. For cancer risk, it can likewise be assumed that the average internal dose per unit
exposure in the population as a whole (or the GST-T1 +/+ portion of the population) at a given
point in time is a good estimate of the average one would estimate if the internal dose was
tracked over the lifetime in the same population—where the distribution of physiological and
metabolic characteristics at a given age is the same as used to estimate the average internal dose
distribution for a cross-section of society.
B-4 DRAFT - DO NOT CITE OR QUOTE
-------
B.3. CY2E1 AND GST-T1
This evaluation incorporated additional data concerning the variability in CYP2E1
activity among humans, based on Lipscomb et al. (2003). The Lipscomb et al. (2003) study was
based on in vitro analysis of liver samples from 75 human tissue donors (activity towards
trichloroethylene and measurements of protein content) to estimate a distribution of activity in
the population. The distribution used by David et al. (2006) for hepatic CYP2E1 activity for
dichloromethane (Vmaxc) was a truncated log-normal distribution with GM = 9.34 mg/hour-kg0'7,
GSD = 1.14, lower bound = 6.33 (68% of mean), and upper bound = 13.8 (218% of mean).
However, Lipscomb et al. (2003; Table IV), analyzing data from a larger set of human tissue
donors, derived an ultimate distribution for CYP2E1 activity with TCE (in units of pmol
oxidized/minute/g liver) with GSD = 1.7274, 5th percentile = 40.7% of the mean, and 95th
percentile = 245.8% of the mean. These data support a wider distribution in CYP2E1 activity
than had been used in the David et al. (2006) model, with approximately a sixfold range between
the upper and lower bounds in Lipscomb et al. (2003) and a twofold range in David et al. (2006).
Since the distribution for Vmaxc (CYP2E1) parameterized by the posterior-distribution
parameters in Table 4 of David et al. (2006) represents the population mean and uncertainty in
that mean, that uncertainty is not reduced or replaced by the knowledge of variability gained
from the data of Lipscomb et al. (2003). Therefore, in EPA's Monte-Carlo simulations, a two-
dimensional sampling process was used for Vmaxc. First the population-mean value of Vmaxc,
Vmaxc,mean, was sampled from the range of uncertainty represented by a log-normal distribution
with GM = 9.34 mg/hour-kg0'7 and GSD = 1.14 mg/hour-kg0'7 (values converted from linear-
space mean [SD] of 9.42 [1.23] mg/hour-kg0'7 reported by David et al. [2006]) and upper/lower
bounds of 7.20 and 12.11 mg/hour-kg0'7, respectively (± 2 SD in log-space). (The linear space or
arithmetic mean is shown in Table B-l). After obtaining the sample population mean
(Vmaxe,mean), an individual Vmaxc value was then obtained by sampling from the log-normal
distribution with that mean but GSD = 1.73 as obtained from the data of Lipscomb et al. (2003).
Further, since even the data available to Lipscomb et al. (2003) were limited, and the log-normal
distribution is naturally bounded to be greater than zero, a nontruncated distribution was used for
this second dimension (step) of parameter sampling.
For GST-T1-mediated metabolism characterized by the rate coefficient, kfc, David et al.
(2006) replaced their estimates of population mean, uncertainty, and variability (the latter is not
reported by David et al. [2006] but would have been estimated along with the uncertainty as part
of the Bayesian analysis) with a measure of population variability alone. The population
variability was obtained by using the known distribution of GST-T1 genotypes in the U.S.
population (from Haber et al., 2002) and the genotype-specific activity distributions from
Warholm et al. (1994), scaled to have the same mean value as the overall mean estimate of the
population mean obtained by David et al. (2006): 0.852 kg°'3/hour. This treatment, however,
fails to incorporate the uncertainty in the population mean characterized by the CV for kfc in
B-5 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4 of David et al. (2006), which is 0.711. Therefore, like Vmaxc, EPA chose to use a two-
dimensional sampling technique for kfc. First, kfc,mean is sampled from a log-normal distribution
with GM = 0.6944 kg°3/hour and GSD = 1.896 kg°3/hour (converted from the linear-space mean
and CV of 0.852 kg°3/hour and 0.711, respectively) with upper/lower bounds of 0.193 kg°3/hour
and 2.50 kg°'3/hour, respectively (± 2 SD in log-space). After obtaining the sample population
mean (kfc,mean), an individual's genotype was sampled from the discrete incidence distribution
(32% chance to be GST-T1 +/+, 48% chance to be +/-, and 20% chance to be -/-; Haber et al.,
2002). Given those genotype frequencies, the interindividual variability was then characterized
by reseating the activity distributions from Warholm et al. (1994), but with the upper and lower
bounds set to zero and mean + 5 SDs, respectively. David et al. (2006) also used zero for the
lower bounds, but set upper bounds to mean + 3 SDs. However, in keeping with the decision to
use an unbounded [log-normal] distribution for CYP2E1, the GST-T1 upper bounds were set at 5
SDs above the mean to assure characterization of the upper end of the human distribution. This
reseating and choice of bounds yields:
0 far GST-Tl-I-,
k/C mean x ^(°-8929, 0.1622 0 < x < 1.704) for GST-Tl +/-,
k/C mean X N(l '786' °-2276 0 < JC < 2.924) far GST-Tl +/+,
where N(u, o | LB < x < UB) is the truncated-normal distribution with mean = \i and SD
= G, bounded between LB and UB. To be clear, when the GST-Tl genotype is sampled as
indicated above, the mean value for the tri-model distribution defined here for kfc is kfc,mean, the
mean for the GST-Tl +/- subpopulation is one-half that for the GST-Tl +/+ subpopulation, and
the CV for each subpopulation is the same as used by David et al. (2006) (from the results of
Warholm et al., 1994).
To assure that the age-dependence of CYP2E1 is properly characterized, particularly for
children, the data of Johnsrud et al. (2003)11 was analyzed. This study measures CYP2E1
activity and other parameters from individuals up to 18 years of age. For a significant subset of
the individuals in that study, values were available for the liver CYP2E1 content (activity/mg
microsomal protein), liver weight, and BW. For individuals 14-18 years old there appeared no
significant trend in the CYP2E1 activity, and the average BW was 69.6 kg, essentially identical
to the value of 70 kg used as a representative adult. Therefore, the data from the 14- to 18-year-
old individuals were assumed to represent adult values, and hence, an evaluation of the change
versus adult could be made by normalizing to these data. In particular, assuming that the
microsomal content per gram of liver is constant across the age-range considered here
"individual data supplied by the corresponding author D. Gail McCarver, to Paul Schlosser, U.S. EPA.
B-6 DRAFT - DO NOT CITE OR QUOTE
-------
(6 months-18 years), the total activity for an individual relative to that in an adult could be
estimated as:
Vmax(individual)/Vmax(14-18) =
CYP2El(individual)-LW(individual)/[CYP2El(14-18)-LW(14-18)]
where Vmax(individual), CYP2E1 (individual), and LW(individual) are the individual's total
activity, activity (per mg microsomal protein), and liver weight, respectively, while Vmax(14-18),
CYP2E1(14-18), and LW(14-18) are the respective average values for individuals 14-18 years
old. (If not normalized, Vmax = CYP2El-msp-LW, where msp is the microsomal content [mg/kg
liver], but if msp is the same in all individuals, it drops from the equation when dividing by the
14-18-year-old average.) These normalized activities are plotted against the relative BW,
BW(individual)/BW(14-18) in Figure B-2.
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
+ Data
BWXDJO scaling
BWUSS scaling
0 0.2 0.4 0.6 0.8 1
BW/BW(14-18)
1.2
1.4
Source: Johnsrud et al. (2003).
Figure B-2. Total CYP2E1 activity (Vmax) normalized to the average total
activity in 14-18 year-old individuals (VmaxIH-lSJ) plotted against
normalized BW for individuals ranging from 6 months to 18 years of age.
The data in Figure B-2 are compared to two model predictions: the allometric-based
prediction used by David et al. (2006) that Vmax will scale as BW0'7, and an alternate scaling
obtained by fitting to these data, BW088. Both alternatives do a fairly good job of representing
the average trend in the data, but the scaling by BW°7 tends to under-predict the data in the range
of 0.4-0.8 for BW/BW(14-18). Therefore, this fitted coefficient will be used when estimating
B-7
DRAFT - DO NOT CITE OR QUOTE
-------
the total activity for individuals <18 years of age. However, since these data indicate a slower
trend (rise in Vmax with BW) for normalized BW -0.7 and above, and there are no data to
indicate that total activity continues to increase so rapidly in adults over 70 kg, the coefficient
will be kept at 0.7 for individuals >18 years of age.
Using the allometric coefficient of 0.88, the normalized Vmaxc values were computed in
the same group of individuals as:
Vmaxc(individual)/Vmaxc(14-18) =
[Vmax(individual)/BW(individual)°-88]/[Vmax(14-18)/BW(14-18)°-88].
To test the revised allometric function versus the data, these values were then plotted
against individual age, and a linear regression was performed, as shown in Figure B-3. While
the slight trend in the regression indicates that not all of the age-dependence may be captured by
this function, the low R2 and small value of the slope indicate that the observation is not
statistically significant and that further attempts to explicitly account for age dependence would
lead to minimal improvement. This representation of the data also clearly shows that the overall
variability in the scaled activity (Vmaxc) is fairly constant across ages: approximately sixfold,
ranging from -0.3 to 1.8. This is the same range observed in adults by Lipscomb et al. (2003), as
noted above. Thus, these data support the use of a constant variability (GSD) in simulating
population variability in CYP2E1 activity for children above 6 months of age, as well as adults.
B-8 DRAFT - DO NOT CITE OR QUOTE
-------
1.8
1.6
oo"1.4
4
1.2
o
0.8
o
OB 0.6
> 0.4
0.2
0
y = 0.01 47x + 0.7907
R2 = 0.0485
9
Age (y)
12
15
18
Source: Johnsrud et al. (2003).
Figure B-3. Body-weight scaled CYP2E1 activity (Vmaxc) normalized to the
average scaled activity in 14-18 year-old individuals (VmaXc[14-18]) plotted
against age individuals ranging from 6 months to 18 years of age.
Note that scaling CYP2E1 activity by BW°'88 for children and by BW° 7 for adults, rather
than per liver weight (which is expected to scale as BWLO with only a 5% CV in liver fraction),
leads to a lower range in CYP2E1 than scaling by BW1 would indicate. In particular, the
distribution predicted by the model for total CYP2E1 activity (mg/hour, upper/lower bound,
given BW between 7 and 130 kg) will be about twofold less than if one assumed the activity
varied as liver weight and simply multiplied the two sets of upper/lower bounds (BW and
activity/g liver). However, given the database for this analysis from adults and children, the
resulting distribution is expected to provide a good prediction of variability in the overall
population.
Unfortunately, there is not a rich data set for the age-dependence of GST-T1 such as is
available for CYP2E1. Strange et al. (1989) examined the developmental patterns for two other
GST classes, mu and pi, and found that the child:adult activity for mu followed a similar pattern
as CYP2E1, increasing from a low level near birth over time, but generally being higher than the
ratio for CYP2E1 in a given age-range. GST-pi, however, was expressed at 21 times adult
values in children up to 1 year of age, then declined with age, so it is difficult to draw specific
conclusions regarding age-dependent variation in GST-T1 from these data (i.e., a quantitative
analysis of the activity data for other GST classes will not be used to estimate the variation in
GST-T1).
B-9
DRAFT - DO NOT CITE OR QUOTE
-------
As is, the model uses a first-order rate constant for GST-T1 which is scaled as BW"°3.
n 9s
While the more recent "standard" for scaling of first-order constants is as BW" , the difference
between these two is small and BW"°3 was used in the Bayesian model calibration. Because
first-order constants are multiplied by tissue volumes in calculating total metabolism rates, and
tissue volume scales approximately as BW1, scaling the GST-T1 constant by BW"°3 is equivalent
to scaling a Vmaxby BW°7. Even though the results cited above for GST-pi have the opposite
trend, data on CYP2E1 activity discussed above and the trend in GST-mu activity are both at
least qualitatively consistent with this scaling. Therefore this scaling is assumed to appropriately
account for age-dependent changes in GST-T1 activity for individuals over 6 months of age via
the explicit dependence on BW.
B.4. ANALYSIS OF HUMAN PHYSIOLOGICAL DISTRIBUTIONS FOR PBPK
MODELING
While the BW distribution in the David et al. (2006) PBPK model used ranges from 7 to
130 kg, thus covering 6-month-old children to obese adults, there are age-dependent changes and
gender-dependent differences in ventilation rates and body fat that are not explicitly included.
To more accurately reflect the distribution of physiological parameters in the entire population,
the unstructured distributions of David et al. (2006) for certain primary or key parameters were
replaced with distributions based on available information that specifically accounts for the
population distributions of age and gender and the age- and gender- specific distributions or
functions for BW, QCC, alveolar ventilation, body fat (fraction), and liver fraction. In the
following, v ~ U[0,l] indicates that v is a random sample from the uniform distribution from 0 to
1.
B.4.1. Age
1 9
U.S. Census Bureau statistics for 6 months to 80 years of age were normalized
(population for 6 months to 1 year assumed to be one-half of 0- to 1-year population), and the
resulting quantiles were plotted against the corresponding ages with a polynomial function fit, as
shown in Figure B-4. A sample individual's age can be determined by using the polynomial,
given v ~ U[0,l]. (Alternately, age can be specified.)
12 Available at http://factfinder.census.gov/servlet/DatasetMainPageServlet: use "enter a table number" or "list all
tables" to select tables QT-P1 and QT-P2 for the (entire) United States.
B-10 DRAFT - DO NOT CITE OR QUOTE
-------
y = 165.86x4-253.19x3+ 113.27x2
0 0.2 0.4 0.6 0.8
Empirical cumulative population fraction
Figure B-4. U.S. age distribution, 6 months to 80 years (values from U.S.
Census Bureau).
B.4.2. Gender
U.S. Census Bureau statistics for fraction of males versus age in 5-year intervals were
plotted and an empirical function was fit, as shown in Figure B-5. Given the individual's age,
the gender is randomly selected as male if v ~ U[0,l] is less than or equal to the polynomial and,
otherwise female. (Alternately, it can be specified.)
0.55
Fraction = 0.513-M25.3-aaer
33.74 + (125.36-age)4
0.25
20
40 60
Age (years)
80
100
Figure B-5. U.S. age-specific gender distribution (values from U.S. Census
Bureau).
B-ll
DRAFT - DO NOT CITE OR QUOTE
-------
B.4.3. BW
Portier et al. (2007) reported statistical mean and SD for BW as a function of age and
gender based on the NHANES IV. Portier et al. (2007) also described empirical functions fit to
these results; however, those functions were found to be exquisitely sensitive to the fitted
parameters to the point that entering the functions by using the four significant figures for
parameters given by Portier et al. (2007) gave results that significantly deviated from the results
as shown by the authors in plots at higher age ranges. Therefore, a somewhat different
functional form was chosen. In particular for BW mean and SD for each gender, a function of
the form was fitted:
fexpt/70/V! ((agec - age) 110)], age agec
9 T
wherepolyfa) = a-x + b[-x + C[-x , i = 1 and 2, age is in years, agec is a "cut" age, dividing the
early-age function from the later age function, andpofy[ (i = 1 or 2). Because there is no explicit
constant term inpofyi and the linear coefficient, a, is common to the two functions, the functions
will automatically satisfy the condition of being equal and of having equal slope (first derivative)
at age = agec, so the overall function will be smooth and continuous.
The functions were fitted to the summarized data based on minimizing a weighted sum of
r\
square errors, error = Xnage x [data(age) - function(age)] , where nage was the number of
observations for the age. The resulting parameter values are listed in Table B-2, and the curve
fits are shown in Figure B-6. Given the age and gender from Sections B.4.1 and B.4.2, the BW
is then randomly selected from a normal distribution, with the resulting mean and SD truncated
at the 1st and 99th percentile.
B-12 DRAFT - DO NOT CITE OR QUOTE
-------
Table B-2. Parameters for BW distributions as functions of age and gender
Parameter
agec
a
bi
c\
b2
b2
Male BW (kg)
Polynomial
parameters for
mean"
21
4.406
-0.0285
-0.729
0.115
0.0048
Polynomial
parameters for SDa
16
2.87
0.06
-2.56
0.96
0.0448
Female BW (kg)
Polynomial
parameters for
mean"
16
4.146
-0.147
-1.36
0.44
-0.0278
Polynomial
parameters for SDa
13
2.574
-0.358
-2.55
1.16
-0.0861
aMean or SD =
where
I exp [polyl ((agec - age) 110)], age < age c
[exp[poly2((agec -age)HO)}, age>agec
pofyi(x) = a-x + b.-x2 + c.-x3, i = 1 and 2
100
100
1.5 3 4.5 6
Age (years/10)
7.5
1.5 3 4.5 6
Age (years/10)
7.5
1.5 3 4.5 6
Age (years/10)
7.5
J
1.5 3 4.5 6
Age (years/10)
7.5
Figure B-6. Function fits to age-dependent data for BW mean and SDs for
males and females in the United States (values from Portier et al. [2007]).
An example output BW distribution, from a Monte Carlo simulation for ages 0.5-
rs, both genders, is shown in Figure B-7. The range, 6.6-131.4 kg, is only slightly larger
80 years, both genders,
B-13
DRAFT - DO NOT CITE OR QUOTE
-------
than that set by David et al. (2006) (i.e., 7-130 kg). But the bimodal form is an unexpected but
reproducible result, presumably occurring because the fraction of a given life span spent at
intermediate BW values is smaller, as evidenced by the most rapid growth rate occurring
between ~7 and 18 years of age (Figure B-6).
0.02
20
40
60 80
BW (kg)
100
120
140
Figure B-7. Example BW histogram from Monte Carlo simulation for 0.5- to
80-year-old males and females in the United States (simulated n = 10,000).
B.4.4. Alveolar Ventilation
Clewell et al. (2004) tabulated values for the alveolar ventilation constant QAlvC
(L/hour-kgOJ5) for males and females at different ages (QAlvC is multiplied by BW0'75 to obtain
the total rate). Smooth functions of age were fitted to those results to use as age- and gender-
specific mean values, shown in Figure B-8. Arcus-Arth and Blaisdell (2007) reported GSD
values for respiration rates for 0-18 years of age; a smooth function was fit to those results and
the value at 18 years was assumed to apply for all adults greater than age 18 (Figure B-9). An
individual's QAlvC was then selected from a log-normal distribution with the resulting mean and
GSD (given age and gender) truncated between the 5* and 95* percentile.
B-14
DRAFT - DO NOT CITE OR QUOTE
-------
GO
-« 30 '
S 20 -
< 10 -
a
5 -
L Respiration constant
v
* * ^ Males = 13.6 + 13.3*exp(-0.05*age)
"^^^ l^ * : : :
Females = 10.7 + 22.1*exp(-0.08*age)
0 20 40 60 80 100
Age (years)
Figure B-8. Mean value respiration rates for males and females as a function
of age (values from Clewell et al. [2004]).
1.65
1.6
1.55
Q 1.5
01.45
1.4
1.35 -\
1.3
Respiration variance
y = -0.1948X3 + 0.6095X2 - 0.3978x + 1.4261
R2 = 0.8823
0.5 1 1.5
Age (years/10)
Figure B-9. GSDs for respiration rates for males and females as a function of
age (values from Arcus-Arth and Blaisdell [2007]).
B.4.5. QCC
Clewell at al. (2004) provide a function that will be taken to represent the mean for the
QCC, QCCmean = 56.906 x (1.0 - e'0-681 x exP[°-0454 x QAlvC]) - 29.747. However, using this function
alone will tightly link cardiac flow and ventilation rates, rather than using a distribution in the
VPR (VPR = QAlv/QC = QAlvC/QCC) as was done by David et al. (2006). Since a distribution
is already defined for QAlvC, above, a VPR subject to variability will be estimated but
renormalized to match the ratio of QAlvC and QCCmean as defined above. In effect, this means
QCCsampie = QCCmean x VPRmean/VPRSamPie will be chosen, where VPRmean and VPRsampie are the
mean and a random sample from the distribution defined by David et al. (2006). Clewell et al.
B-15
DRAFT - DO NOT CITE OR QUOTE
-------
(2004) suggest scaling QCC and alveolar ventilation as BW°'75, while David et al. (2006) used
BW°'74. While virtually identical, it is noted that the implementation here uses BW°'75.
B.4.6. Fat Fraction
Tabulated values from Clewell et al. (2004) show indistinguishable values for the fraction
of BW as fat (VFC) for males and females up to 7 years of age, after which they diverge.
Polynomial functions were fit separately for 0-7 years, 7-20 years, and 20-80 years, with the
latter two ranges being gender-specific, as shown in Figure B-10. The ratio of the resulting
gender- and age-specific value to the VFCmean from David et al. (2006) was then used to scale the
bounded normal distribution as specified by David et al. (2006) for the selected age and gender.
B-16 DRAFT - DO NOT CITE OR QUOTE
-------
0.3
0.25 -
c
° 0.2 -
5 0.15 -
M-
g 0.1
0.05 H
0
0 to 7 years
Male
Female
Poly. (Female)
y = 0.1612x3 + 0.0846X2 - 0.3083x+ 0.2709
R2 = 0.9984
0.2 0.4 0.6
Age (years/10)
0.8
0.3
0.25 H
c
O 0.2
i 0.1 -
0.05 -
0
y = -0.0458x2 + 0.2082x + 0.0274 Female
R2 = 0.9942
y = -0.0057X2 + 0.0293X + 0.1303
R2 = 0.9941
7 to 20 years
0.6 0.85
1.1 1.35 1.6
Age (years/10)
1.85
2.1
0.6
0.5 H
.20.4
0.1 H
o
y = -0.0024x3+ 0.0355x2-0.115x + 0.3678
y = -0.0015X + 0.0384X + 0.0908
R2 = 0.8969
20 to 90 years
1.5
3.5 5.5
Age (years/10)
7.5
9.5
Figure B-10. Fraction body fat (VFC) over various age ranges in males and
females (data from Clewell et al. [2004]).
B.4.7. Liver Fraction
Tabulated values from Clewell et al. (2004) showed an interesting age dependence for the
liver fraction of BW, VLC, as shown in Figure B-l 1. While female VLC values were somewhat
higher than males between 18 and 40 years, and the difference may be statistically significant,
the overall variation is not large; the two were indistinguishable during other age-ranges, and
model predictions are not overly sensitive to VLC. Therefore, only the male data were used.
Smooth functions were fit separately for 0-18 years and 18-80 years and used to scale the
distribution from David et al. (2006) for the given age, as was done for body fat.
B-17
DRAFT - DO NOT CITE OR QUOTE
-------
0.05
0.04 -
0.03 -
0.02 -
0.01 -
0
y = -0.0036X2 - 0.0051X + 0.0395
R2 = 0.8995
0 to 17 years
Male liver
Female liver
Poly. (Male liver)
0 0.3 0.6 0.9 1.2 1.5
Age (years/10)
1.8
0.028
0.026 -
0.024 -
0.022 -
0.02 -
0.018
17 to 83 years
y = -0.0004X2 + 0.0034X + 0.0169
R2 = 0.8984
Male liver
Female liver
Poly. (Male liver)
3456
Age (years/10)
Figure B-ll. Fraction liver (VLC) as a function of age (data from
Clewell et al. [2004]).
B.4.8. Tissue Volume Normalization
While not explicitly stated by David et al. (2006), total tissue volume must remain
roughly the same as a fraction of BW. While this fraction could also change with age, gender,
and other characteristics, it was assumed that any change in it would be modest and not
significantly affect model predictions, given the fairly broad distribution implemented for total
BW. This normalization was applied irrespective of those factors. Therefore, after drawing
sample values of the tissue volume fractions from each of their respective distributions, the
fractions were then normalized to a total fraction of 0.9215, which is the sum of the mean values
for the fractions for the distributions as described by David et al. (2006). The remaining body
mass is taken to be bone, teeth, hair, nails, and any other minimally or nonperfused components.
B.5. SUMMARY OF REVISED HUMAN PBPK MODEL
The resulting set of parameter distribution characteristics, including those used as defined
by David et al. (2006), are described in Table B-3. The metabolic parameter statistics reported in
Table 4 of David et al. (2006) are summary statistics of the converged parameter chains obtained
in that analysis for the population mean of each parameter. As such, those statistics (means and
B-18
DRAFT - DO NOT CITE OR QUOTE
-------
CVs) are assumed to represent the most likely value of the mean and the degree of uncertainty in
that mean. However, for the metabolic parameters other than Vmaxc (representing CYP2E1
activity) and kfc (GST-T1 activity), EPA considers it reasonable to assume negligible variability
among the population compared to the estimated uncertainties. So while EPA's objective is to
account for both variability and parameter uncertainty in the population, the statistics for those
other parameters (Km, Al, A2, and FracR) were used as is to define population distributions. For
Vmaxc and kfc, two-dimensional sampling routines were used, as described in detail above, to
explicitly account for both uncertainty and the known high degree of interindividual variability.
Distributions for a number of the physiological parameters, which are assumed to represent a
well known degree of variability, were also revised from those used by David et al. (2006).
B-19 DRAFT - DO NOT CITE OR QUOTE
-------
Table B-3. Parameter distributions for the human PBPK model for dichloromethane used by EPA
Parameter
BW
Body weight (kg)
Distribution
Shape
Normal
(Geometric)
mean"
SD/GSD3
f(age, gender)
Lower
bound
1st %tile
Upper
bound
99th %tile
Section or source
B-4.3;NHANESIV
Flow rates
QAlvC
vprv
QCC
Alveolar ventilation (L/hr/kg° 75)
Variability in ventilation/perfusion ratio
Cardiac output (L/hr/kg° 75)
Normal
Log-normal
f(age, gender)
1.00
QCCmean=/QAlvC)
f(age)
0.203
5th %tile
0.69
95th %tile
1.42
QCC = QCCmean/vprv
B-4.4; mean: Clewell et al. (2004);
SD: Arcus-Arth and Blaisdell (2007)
VPR/VPRmean of David et al. (2006)
B-4.5; Clewell et al. (2004) (mean)
Fractional flow rates (fraction of cardiac output)
QFC
QLC
QRC
QSC
Fat
Liver
Rapidly perfused tissues
Slow perfused tissues
Normal
Normal
Normal
Normal
0.05
0.26
0.50
0.19
0.0150
0.0910
0.10
0.0285
0.0050
0.010
0.20
0.105
0.0950
0.533
0.80
0.276
David et al. (2006); after sampling from
these distributions, normalize:
_. QC-QiC
Q Z0c
Tissue volumes (fraction BW)
VFC
VLC
VLuC
VRC
VSC
Fat
Liver
Lung
Rapidly perfused tissues
Slowly perfused tissues
Normal
Normal
Normal
Normal
Normal
f(age, gender)
f(age)
0.0115
0.064
0.63
0.3-mean
0.05-mean
0.00161
0.00640
0.189
0.1 -mean
0.85 -mean
0.00667
0.0448
0.431
1.9- mean
1.15-mean
0.0163
0.0832
0.829
Fat mean: B-4.6; (Clewell et al., 2004);
Liver mean: B-4.7; (Clewell et al.,
2004); otherwise David et al. (2006);
after sampling from these distributions,
normalize:
^ Q.92\5-BW-ViC
'• !
-------
Table B-3. Parameter distributions for the human PBPK model for dichloromethane used by EPA
Parameter
Distribution
Shape
(Geometric)
mean"
SD/GSD3
Lower
bound
Upper
bound
Section or source
Metabolism parameters (based on Monte Carte calibration from five human data sets)
* maxC,me
an
/
Vmaxc
Km
Al
A2
FracR
Population mean /
individual maximum metabolism rate
(mg/hr/kgxvmax)
Affinity (mg/L)
Ratio of lung Vmax to liver Vmax
Ratio of lung KF to liver KF
Fractional MFO capacity in rapidly
perfused tissue
Log-normal
Log-normal
Log-normal
Log-normal
Log-normal
Log-normal
9.34
* maxC,mean
0.41
0.00092
0.0083
0.0152
1.14
1.73
1.39
1.47
1.92
2.0
7.20
(none)
0.154
0.000291
0.00116
0.00190
12.11
(none)
1.10
0.00292
0.0580
0.122
B-3; mean: David et al. (2006);
Individual GSD: Lipscomb et al. (2003);
Xvmax = 0.88 for age <18.
Xvmax = 0.70 for age >18.
GM & GSD values listed here, converted
from arithmetic mean and SD values of
David et al. (2006)
First order metabolism rate (/hr/kg° 3)
KfC,mean
kJC lkJC,mean
Population average
Homozygous (-/-)
Heterozygous (+/-)
Homozygous (+/+)
Log-normal
N/A
Normal
Normal
0.6944
kfc = 0
0.8929
1.786
1.896
-
0.1622
0.2276
0.1932
-
0
0
2.496
-
1.704
2.924
Adapted from David et al. (2006);
kfcmean is first sampled, then the relative
individual value, kfc/kfCmean, given the
genotype; kfc is then the product.
""Arithmetic mean and SD listed for normal distributions; GM and GSD listed for log-normal distributions.
B-21
DRAFT - DO NOT CITE OR QUOTE
-------
The Monte-Carlo sampling approach used effectively assumes that all the parameters are
distributed independently, ignoring the covariance that was likely represented in the actual
posterior chains. This approach will tend to overestimate the overall range of parametric
variability and uncertainty and, hence, distribution of dose metrics in the population compared to
what one would obtain if the covariance were explicitly included. Thus, if the covariance (i.e.,
the variance-covariance matrix) for the set of parameters had been reported by David et al.
(2006) it could have been used to narrow the predicted distribution of internal doses or
equivalent applied doses, but lacking such information, the approach used will not underestimate
risk or overestimate lower bounds on human equivalent exposure levels. However, this source of
overestimation in variability and uncertainty is probably offset to some extent by the fact that the
analysis leaves out the degree of interindividual variability for Km, Al, A2, and FracR that
should have also been estimated by the Bayesian analysis.
B-22 DRAFT - DO NOT CITE OR QUOTE
-------
APPENDIX C. RAT DICHLOROMETHANE PBPK MODELS
The critical studies and data chosen for derivation of a recommended RfD and RfC were
based on nonneoplastic liver lesions in rats inhaling dichloromethane for 2 years (Nitschke et al.,
1988a; Serota et al., 1986a). For this reason, a PBPK model for inhaled and orally absorbed
dichloromethane in rats was needed to provide estimates of internal dosimetry for dose-response
modeling and to extrapolate internal liver doses from rats to humans. Several deterministic
PBPK rat models have been reported in the scientific literature (Sweeney et al., 2004; Andersen
et al., 1991; Reitz, 1991; Reitz et al., 1988a, b; EPA 1988b, 1987a, b; Andersen et al., 1987;
Gargas et al., 1986). Unlike in the mouse study (Marino et al., 2006), however, no probabilistic
models are available in which the uncertainty in model parameters was reduced by utilizing
multiple data sets for parameter estimation. Rat data were not available that would allow for
MCMC calibration of individual metabolic parameters for the CYP or GST pathways. For
example, the MCMC calibration of the mouse model (Marino et al., 2006) relied on inhalation
data using trans-l,2-dichloroethylene, a CYP2E1 inhibitor, (not available for the rat) to
specifically estimate GST metabolism in isolation of CYP metabolism, thereby improving the
estimate of metabolic flux through the competing pathways. Thus, the selected model includes
parameter values estimated by deterministic methods only. In order to use the latest data for
dichloromethane toxicokinetics in rats, an assessment was conducted of multiple rat models (or
modified versions of those models) to select the most appropriate model for use in the derivation
of the RfD and RfC.
C.I. METHODS OF ANALYSIS
C.I.I. Selection of Evaluation Data Sets and PBPK Models
Published studies of dichloromethane metabolism and toxicokinetics in rats were
reviewed to identify data sets for use in model evaluation and possible calibration. Toxicokinetic
data were available for:
• blood levels of dichloromethane, the percent saturation of hemoglobin as COHb, and
expired dichloromethane and CO following intravenous injection (Angelo et al.,
1986b);
• dichloromethane air concentrations in closed chamber experiments (Gargas et al.,
1986);
• dichloromethane and %COHb blood levels during and after a 4-hour open chamber
(constant concentration) inhalation exposure (Andersen et. al., 1991, 1987);
C-1 DRAFT - DO NOT CITE OR QUOTE
-------
• %COHb levels from 30 minutes to 12 hours postexposure following a single oral
dose of 526 mg/kg (6.2 mmol/kg in Oleum pedum tauri vehicle) (Pankow et al.,
199la); and
• Cumulative dichloromethane (mg) expired up to 96 hours following gavage doses of
250, 500, 1,000, or 2,000 mg/kg in corn oil or water (Kirschman et al., 1986).
Three variations on the PBPK models of Andersen et al. (1991, 1987) were assessed for
the ability to predict these data. The model structure is depicted in Figure C-l. In each model,
metabolism involves two competing pathways: the GST pathway, described with a linear first-
order kinetic model, and the CYP pathway, described with a saturable Michaelis-Menten kinetic
model.
DCM model
Air
\N. GST-*— i r-^CYP-
Alveolar Lung
air tissue
Oral _
exposurt
Fat
Richly
perfused
Slowly
perfused
—> Gl tract
3
Liver
CO model
Endogenous
production
Figure C-l. Schematic of the PBPK model for dichloromethane in the rat.
Variation A is a hybrid of Andersen et al. (1991) and Andersen et al. (1987) in that it
included both CO production resulting from CYP-mediated metabolism (Andersen et al., 1987)
and lung metabolism of the parent compound (Andersen et al., 1991). Andersen et al. (1987)
based the lung-to-liver metabolism ratios (Al and A2 for CYP- and GST-mediated metabolism,
respectively) on the reaction rates reported by Lorenz et al. (1984). Instead EPA used: 1) the
reaction rates reported by Reitz et al. (1989); 2) the levels of microsomal and non-microsomal
C-2
DRAFT - DO NOT CITE OR QUOTE
-------
(cytosolic) protein reported by Litterst et al. (1975); and 3) the relative lung and liver tissue
weights from Andersen et al. (1987).
Specifically for CYP activity, Reitz et al. (1975) observed metabolic rates of 4.10 and
0.16 nmol/min/mg microsomal protein with 5 mM dichloromethane using rat liver and lung
preparations, respectively. Litterst et al. (1975) measured 0.027 and 0.0094 mg microsomal
protein per kg rat liver and lung, respectively, and the fraction of body for rat liver and lung from
Andersen et al. (1987) for a 0.3 kg rat is 4.0 and 1.16%, respectively; Andersen et al. (1987) used
an allometric relationship for lung mass, Vmng = 0.0115*BW°'99, so the lung fraction varies with
body weight. Based on these values, 99.8% and 0.2% of total CYP activity is estimated to be in
rat liver and lung, respectively. Therefore, the liver CYP allometric constant from Andersen et
al. (1991), Vmaxc, was changed from 4.0 to 3.992 mg/h/kg0'7 and Al was set to 0.002. This value
is significantly less than 0.0558, the value estimated and used by Andersen et al. (1987). The
ratio of activity per mg microsomal protein in liver versus lung from Reitz et al. (1989) is 0.039,
fairly close to the value of Al used by Andersen et al. (1987). This suggests that the discrepancy
between EPA's value of Al and that of Andersen et al. (1987) is that the previous value did not
account for differences in microsomal protein content and tissue mass.
Similarly for GST activity, Reitz et al. (1989) observed 7.05 and 1.0 nmol/min/mg
cytosolic protein for rat liver and lung preparations, respectively, at 40 mM dichloromethane.
While Litterst et al. (1975) did not report the amounts of cytosolic protein per se in their
preparations, they did report protein amounts in the supernatant from separation of microsomal
protein that EPA used as effectively being cytosolic: 0.0531 and 0.0556 mg protein/kg tissue in
liver and lung, respectively. Using the same liver and lung tissue fractions as for the CYP
calculation, one can then estimate that 95.9% of GST activity is in the rat liver, and hence the
liver allometic constant for GST activity, kfc, was changed from 2.0 to 1.917 kg°'3/h. Since the
rate constant, kf = kfc/BW03, is an activity per volume of tissue, the calculation of A2 (lung:liver
activity ratio) is then calculated using the activity per mg cytosolic protein and amount of
cytosolic protein per kg tissue to obtain the relative activity per kg tissue: A2 = 0.149. Using kfc
= 1.917 kg°3/h and A2 = 0.149 yields the same total GST activity for a 0.3 kg rat as using kfc =
2.0 kg°'3/h with no activity in the lung. This value of A2 is close to that used by Andersen et al.
(1987), 0.136.
Simulations of the intravenous and inhalation data described above with Variation A
were visually indistinguishable from simulations with the model set to exactly match the one
used by Andersen et al. (1991); i.e., with Vmaxc and kfc set at the values specified in Table 1 of
Andersen et al. and Al = A2 = 0, there was no metabolism in the lung compartment (results not
shown). Hence, Variation A was thereafter considered to be effectively identical to the original
model of Andersen et al. (1991).
Variation B tests the hypothesis that, with the distribution of GST activity between the
lung and liver, statistically significant improvements in the fit of the model to the data outlined
C-3 DRAFT - DO NOT CITE OR QUOTE
-------
above can be achieved. Andersen et al. (1991) explicitly adjusted the three primary metabolic
constants, Vmaxc, Km, and kfc for CYP- and GST-mediated metabolism. EPA noted that while
the text of Andersen et al. state that the constant for CO yield per unit of CYP metabolism, PI,
was set at 70% (the value used by Gargas et al., 1986), the value listed in Table 1 of Andersen et
al. (1991) for the rat was PI = 0.8; i.e., 80%. The published figures of Andersen et al. (1991)
could be reproduced with PI = 0.8 but not with 0.7. Further, Gargas et al. (1986) indicated that
the value they used, 0.7, was itself fit to data. Thus, it appears that both Gargas et al. (1986) and
Andersen et al. (1991) fit PI to their in vivo PK data, arriving at slightly different values; EPA
also included this parameter among those to be numerically fitted.
Variation C: As will be shown, Variation B fits the considered data set significantly
better than using the original parameter set (Variation A). In order to also described oral
exposures, a GI sub-model was added comprised of an upper and a lower GI compartment, with
associated rate constants for transfer from upper to lower GI (ki2) and absorption from the upper
(ka) and lower GI (ka2). This variation was created by fitting the parameters adjusted in (B)
along with the GI rate constants to the inhalation and intravenous data used previously along
with oral dosimetry data, to obtain a global optimum for the adjusted parameters.
C.1.2. Analysis
While attempting to reproduce the fits to CO-inhalation kinetic data of Andersen et al.
(1991), EPA found that the CO sub-model constants were not entirely self-consistent. In
particular, the background amount of total blood CO given for rats in Table 1 of Andersen et al.
(1991) (ABcoc = 0.117 mg/kg) corresponds to a (background) blood carboxyhemoglobin = 0.7%
saturation (value given in text of Andersen et al [1991]). But the endogenous production rate
constant (RENcoc = 0.035 mg/hr/kg0'7) is insufficient to support that background level, even
with their listed background air concentration (COINH = 2.2 ppm CO). Instead, simulations
with this constant level of CO in the air and endogenous production lead to a predicted steady-
state CO level of 0.5% rather than 0.7%. Also, some data sets proposed for simuation in this
assessment had background COHb levels different from 0.7%. Therefore, in addition to the
changes described above, algebraic equations were included in the code ("INITIAL" section in
acslX) such that the endogenous CO production rate (RENco) and initial total CO mass in the
blood (AeL2o) were then calculated to match specified time-zero percent carboxyhemoglobin
(%COHbo) and background CO air concentration (COINH), rather than using a fixed
background CO production rate. These equations were subject to a conditional (if-then-else)
statement that allows the user to either set the constants exactly as listed by Andersen et al.
(1991) or use the new algebraic equations to set RENcoc and AeL2o to match a specified
%COHbo and make that level the steady-state COHb level in the absence of dichloromethane
exposure. This last change did not introduce any adjustable parameters, but rather made the
choice of endogenous production level easier, with %COHbo being the user-set variable rather
C-4 DRAFT - DO NOT CITE OR QUOTE
-------
than having to calibrate the background production rate to match the observed background
percent COHb.
Analysis Details
Model variations were implemented using the acslX simulation software (version 3.0.1.6,
The Aegis Technologies Group, Huntsville, AL). Parameter-estimation was performed using the
Nelder-Mead algorithm as implemented in the acslX package. Initial values of Vmaxc, kfc, PI,
Al and A2 were those described above for Variation A, while that for Km was the value used by
Andersen et al. (1991) (0.4 mg/L).
In attempting to evaluate any differences between model simulations with Variation A
compared to the original model of Andersen et al. (1991) (i.e, with no lung metabolism) two
small but noticeable internal discrepancies were noted in Andersen et al. (1991) and handled as
follows. First, while the model and parameters of Andersen et al. (1991) were nominally
designed to include an endogenous background level of CO production and COHb = 0.7%
saturation, simulations shown in Figures 3 and 4 of Andersen et al. (1991) appear to begin with
COHb = 0% (and in Figure 4, approach 0% at the end of the observation period). If there is
endogenous production and background CO exposure consistent with 0.7% COHb, this COHb
level should be present at the start of any exogenous dichloromethane exposure and be asymptot-
ically approached after an exposure ends. EPA's simulations were so conducted, including
Variation A. Hence the simulation results (plots) differ slightly from those in Andersen et al.
(1991), the difference being noticeable near the start of exposure. While the difference may not
appear large in the data plots, it could significantly impact the goodness-of-fit (likelihood)
calculations. Since the background CO and COHb levels are set identically for both model
variations, this choice should not unfairly bias the comparison between them.
The other slight discrepancy in Andersen et al. (1991) is that, while Table 1 specifies a rat
BW = 0.22 kg and Vmaxc = 4.0 mg/h/kg0'7, these values lead to Vmax = 1.386 mg/h. However, the
legends of Figures 3 and 4 in Andersen et al. (1991) state that Vmax was 1.46 mg/h for those
simulations. Since this Vmax would occur for a 0.237 kg rat with Vmaxc = 4.0, this discrepancy
was resolved by using that BW instead of 0.22 kg, without adjusting Vmaxc, for those particular
simulations. Like the preceding choice, this input parameter specification was made uniformly
for both model variations and may contribute to differences between EPA's Variation A
simulations and Figures 3 and 4 of Andersen et al. (1991).
Variation B specifically involves comparing the goodness of fit of Variation A, in which
metabolism is distributed between lung and liver but the effective parameters are otherwise
identical to Andersen et al. (1991), to the goodness of fit obtained when Vmaxc, Km, kfc, and PI
are fit to the data, with other parameters as specified in Table 1. For this test goodness of fit is
determined as the log-likelihood function (LLF) value reported by acslX. From a strictly
statistical viewpoint, Variation B involves using the same number of adjustable parameters as set
C-5 DRAFT - DO NOT CITE OR QUOTE
-------
by Andersen et al. (1991), hence any improvement in fit could be considered significant.
However, in deference to the precedent of a previously published scientific paper, EPA used the
test described by Collins et al. (1999): two times the change in the LLF was compared to the
value of the %2 distribution with four degrees (since four parameters are numerically fitted) at a
/>-value of 0.05.
Variation B was obtained by fitting first the closed-chamber gas uptake data of Gargas et
al. (1986), the intravenous data of Angelo et al. (1986b), and the open-chamber (200 and 1014
ppm) inhalation data of Andersen et al. (1987). Initial estimates of the GI sub-model parameters,
ka, ki2, and ka2, were then obtained by keeping the metabolic parameters constant at Variation B
values while calibrating the oral constants against the oral exposure data of Angelo et al.
(1986b): blood and liver dichloromethane levels, expired dichloromethane levels (all time-
points) and expired CO levels (24-hour cumulative). Only the 24-hour cumulative CO
exhalation data were used to inform the estimate ka (earlier time-points not used) for reasons
explained below. Final parameter estimates were obtained by performing a global optimization
in which all parameters were fit simultaneously to the combined iv, inhalation, and oral exposure
data. Results of a 30-minute exposure to 5,159 ppm dichloromethane reported by Andersen et al.
(1991), (data not used for EPA's model calibration), were also used to compare Variation A and
C predictions. Two other data sets from gavage exposures (Pankow et al., 1991a; Kirschman et
al., 1986) were also used in the evaluation of the model's ability to predict oral exposure data.
The statistical error model used by acslX also depends on a coefficient, the hetero-
scedasticity, which allows for variation between an absolute error model (e.g., unrelated to the
magnitude of the response variable) and a relative error model (e.g., experimental noise is
assumed to be proportional to the response variable). The heteroscedasticity for each observed
response was also fitted to the data (adjusted by acslX) along with any fitted metabolic
parameters.
C.2. RESULTS
Parameter values for Variations A-C are listed in Table C-l. Two versions of the PBPK
model for dichloromethane (Variations A and B) were first evaluated for goodness of fit to a
common set of inhalation and intravenous pharmacokinetic data. The oral absorption and GI
transfer constants (ka, ka2, and k^) were only calibrated as described above for the model that
was deemed best with statistical justification: Variation B. The resulting metabolic parameters,
Variation C, were almost identical to those obtained for Variation B and simulation results for
the two were visually indistinguishable. Therefore only results for Variations A and C are shown
when comparing model predictions to data below. The endogenous level of carboxyhemoglobin,
%COHbo, was set to 0.7% and the background CO concentration set to 2.2 ppm, as stated in the
text of Andersen et al. (1991). However, for studies that used radio-labeled dichloromethane and
C-6 DRAFT - DO NOT CITE OR QUOTE
-------
only tracked the radio-labeled CO produced from that dichloromethane, the endogenous levels
were not a factor and hence set to zero.
Table C-l. Parameter values used in rat PBPK models
Parameter
Variation A
Variation B
Flow rates
QCC (L/hr-kg074)
VPR
15.9
0.94
Same as
Variation C
Variation A
Fractional flow rates (percent of QCC)
Fat
Liver
Rapidly perfused tissues
Slowly perfused tissues
9
20
56
15
Same as
Variation A
Tissue volumes (percent BW)
Fat
Liver
Lung (scaled as BW° ")
Rapidly perfused tissues
Slowly perfused tissues
7
4
1.15
5
75
Same as
Variation A
Partition coefficients
Blood:air
Fatblood
Liverfolood
Lung/arterial blood
Rapidly perfused tissue^lood
Slowly perfused tissue (muscle)/blood
19.4
6.19
0.732
0.46
0.732
0.408
Same as
Variation A
Metabolism and absorption parameters
VmaxC, max CYP metabolic rate in liver (mg/h-kg0 7)
Km CYP affinity (mg/L)
kfC, 1st order GST metabolic rate in liver (kg0 3/h)
PI, yield of CO from CYP metabolism (no units)
Fl, correction factor for CO exhalation rate
Al, ratio of lung Vmax to liver Vmax
A2, ratio of lung kf to liver kf
ka, oral absorption constant (1/h)
LLF, log-likelihood for fit to data in Figures C3-C6
3.992
0.4
1.917
0.8
1.21
0.002
0.149
5.0
-657.9
3.97
0.509
2.46
0.68
1.21
0.002
0.149
ND
-591.2"
3.97
0.510
2.47
0.68
1.21
0.002
0.149
4.31
-591.3**
1 Difference from Variation A is statistically significant atp < 0.001, assuming a chi-square
distribution with four degrees of freedom.
b Difference from Variation B is not statistically significant, assuming a chi-square distribution
with four degrees of freedom.
C-7
DRAFT - DO NOT CITE OR QUOTE
-------
C.2.1. Evaluation of Model Structure for Description of Carboxyhemoglobin Levels
The model of Andersen et al. (1991; Variation A) calculated the amount of COHb by
assuming instantaneous equilibration between the free CO and hemoglobin-bound CO in the
blood. However, when Angelo et al. (1986b) observed the kinetics of dichloromethane in blood
and exhaled dichloromethane and CO after oral administration, they found that blood
dichloromethane peaked rapidly in <30 minutes. The rate of exhalation of dichloromethane,
shown by the rate of rise in the exhaled dichloromethane data in Figure C-2A where the plateau
in exhaled dichloromethane levels from 6 to 24 hours, indicates that blood dichloromethane
levels had dropped to nearly 0 by 6 hours. Exhaled CO, however, shows a more gradual rise
than dichloromethane, continuing between 6 and 24 hours. Pankow et al. (1991a) measured
COHb in blood after dichloromethane exposure, albeit at a higher dose (526 mg/kg versus
200 mg/kg by Angelo et al. [1986b]), and found that blood COHb levels did not peak until
6 hours after dosing (Figure C-2B), while Angelo et al. (1986b) found that blood
dichloromethane had declined to -4% of the peak level (10-minute concentration) by 6 hours.
Some delay between peak dichloromethane and peak (cumulative) CO levels is predicted by the
published (1991) and modified (Variations A to C) Andersen et al. (1991) models because CO is
produced from oxidative dichloromethane metabolism. However, EPA found that none of the
PBPK models could simultaneously describe both the very short-time peak in blood
dichloromethane with the concurrent rapid rise in exhaled dichloromethane levels (Figure C-2A)
and the late peak blood CO (Figure C-2B) and slower rate of CO exhalation (Figure C-2A). The
final models were able to adequately predict dichloromethane exhalation data and
dichloromethane blood concentration data but overestimated CO exhalation data (these findings
are illustrated for Variation C later in the results section).
EPA hypothesized that the models' inability to simultaneously fit these data might
specifically be due to the assumption that CO only distributes into the blood, rather than other
body tissues. Cho et al. (2001) measured CO transport in the perfused rat hindlimb and showed
significant distribution into the tissue (compared to albumin or sucrose, for which distribution
into tissue was minimal) and estimated a volume of distribution of 2.3 L/kg and permeability-
area (PA) product of 4.86 L/(h-kg). Cho et al. (2001) also obtained a fractional recovery of only
45% of the infused CO after 10 minute of collection. Since modeling CO kinetics with a high
degree of accuracy was not an objective of the current assessment, EPA conducted only a limited
exploration of an alternate CO model that included a tissue compartment with transport between
that compartment and the blood set using the PA value of Cho et al. (2001). With this
compartment included, the kinetics of CO and COHb were slowed significantly from those
obtained with the Andersen et al. (1991) model structure, better matching the shape of the
concentration-time TK data (results not shown).
C-8 DRAFT - DO NOT CITE OR QUOTE
-------
70
0 60
Q
» 50
re
0) 40
1 30
5 20
10
A f»
/ *
/ — ""
/ ^•'""" ^
-------
determined by the rate of metabolism of dichloromethane and the CO-yield parameter, PI, while
the rate of elimination of CO is determined by physiological parameters and adjustment constant,
Fl, that multiplies the CO exhalation equation. Andersen et al. (1991) used a 2-hour exposure to
500 ppm CO to estimate Fl and another CO-sub-model parameter (results shown in Figure 2 of
Andersen et al., 1991). EPA found that this fit to the CO-exposure data was highly sensitive to
Fl, hence concluded that its value was well-characterized and should not be adjusted. Altering
the rate of dichloromethane metabolism would degrade the fit to the dichloromethane data and
the primary purpose of the model is to describe dichloromethane kinetics. Adjusting
physiological parameters (e.g., cardiac output or respiration rate) in an attempt to fit the CO
exhalation would only hide the error and reduce the model's reliability in predicting other
aspects of dosimetry, particularly the pharmacokinetics of dichloromethane and its rates of
metabolism, which are critical for risk assessment. The only remaining adjustable parameter was
then PI (CO yield from CYP metabolism of dichloromethane); while changing its value could
improve the fit to the short-term CO data, it would simultaneously degrade the fit to the long-
term data, or total measured CO yield. In summary, it appears that given the model structure and
assumptions, one cannot explain the difference between the exhalation time-courses for
dichloromethane and CO/COHb shown in Figure C-2 while retaining realistic physiological
parameters and matching the measured total CO yield. As described above, a limited analysis
indicated that including distribution of CO into body tissues in the model would at least partly
resolve this apparent discrepancy (results not shown). Further improvement might be obtained
by also including the actual kinetics of FIb-CO binding and disassociation (i.e., using kinetic
constants determined in vitro).
While the analysis of the CO and COHb data above led to the conclusion that the purpose
of the model for chronic-exposure risk assessment was best served by not using some of those
data (i.e., the short-term dichloromethane and CO exhalation data), the combination of
dichloromethane concentration data in various compartments as described below (primarily
blood and air) and the more limited long-term CO exhalation data were considered adequate and
necessary to estimate model parameters for dichloromethane. Kinetic COHb data were used for
model calibration, however, since those were predicted well by the model, and comparisons of
model predictions to the CO data are shown to further illustrate the discrepancy in kinetics and
the agreement in long-term mass balance. If the model was to be used for evaluation of the risk
of acute carbon monoxide exposure, then it would be necessary to demonstrate adequate
reproduction of the COHb data at all times.
C.2.2. Evaluation of Prediction of Uptake, Blood and Liver Concentrations, and
Expiration of Dichloromethane
One data set used for model calibration is the closed chamber experiments of Gargas et
al. (1986) (100-3,000 ppm dichloromethane). Because actual chamber concentrations can differ
C-10 DRAFT - DO NOT CITE OR QUOTE
-------
from target concentrations, a simple exponential function was fit to the first four data points at
each concentration and used to extrapolate back to an actual initial concentration. The initial
concentrations so obtained and used in the modeling were 107, 498, 1,028, and 3,206 ppm versus
the target concentrations of 100, 500, 1,000, and 3,000 ppm. Predictions using Variations A and
C both fit the observed chamber air concentrations reasonably well (Figure C-3), but Variation C
fit the 3000 ppm data more exactly from 2 hours on (Figure 3, middle panel). The difference
between Variations A and C is diminished at 1000 ppm (Figure C-3, middle and bottom panels),
but where a difference can be seen (Figure C-3, bottom pane. 2.5 hours and beyond) Variation C
again fits the data more exactly. Predictions with the two models at 500 and 100 ppm were
essentially identical.
C-l 1 DRAFT - DO NOT CITE OR QUOTE
-------
C Data
Model A
Model C
500
400
300 -
200 -
100
0.5
1.5
2 2.5
Time (h)
3.5
4.5
Figure C-3. Observations of Gargas et al. (1986; data points) and predictions
for Variations A and C (curves) for respiratory uptake by three rats of 100-
3,000 ppm dichloromethane in a 9-L closed chamber. Target initial
concentrations were 100, 500,1000, and 3000 ppm. A simple exponential
curve was fit to the first four data points of each data set to identify initial
conditions for each simulation.
C-12
DRAFT - DO NOT CITE OR QUOTE
-------
Model simulations were next compared to measurements of blood dichloromethane
concentration and the fraction (%) of dichloromethane exhaled after intravenous injections of 10
or 50 mg/kg (Angelo et al., 1986b) (Figure C-4). Intravenous data are considered particularly
useful for evaluating metabolic rate constants because the kinetics and complexity of uptake by
other routes of exposure have been bypassed. Blood concentration simulations using the two
models were almost identical (Figure C-4, upper panel) and match the observed blood
dichloromethane levels fairly well: simulations are within a factor of 3 of the data (often closer)
and the slope of the model curves after the initial distribution phase (~5 minutes) are close to that
of the data. However, both models over-predicted the observed data to some extent, particularly
the 50 mg/kg dose, for which all of the observed data points are over-predicted, whereas the 10
mg/kg simulations match the first three data points well. Because the upper panel of Figure C-4
is plotted on a log-y scale, the discrepancy between model simulations and the 50 mg/kg data
appears similar to the discrepancy versus the 10 mg/kg data at later time points, but in fact is
much larger in an absolute sense.
C-13 DRAFT - DO NOT CITE OR QUOTE
-------
1,000
2r ioo
£
O
Q
-a
o
£
00
10
1
70
Model A, 10 mg/kg
- —Model C, 10 mg/kg
Model A, 50 mg/kg
• — Model C, 50 mg/kg
D 10 mg/kg data
• 50 mg/kg data
D
10
20
Time(min)
30
40
D 10 mg/kg data
• 50 mg/kg data
10 mg/kg model A
50 mg/kg model A
^— 10 mg/kg model C
— — —50 mg/kg model C
2
Time(h)
Figure C-4. Observations (data points) of Angelo et al. (1986b) and
predictions for model Variations A and C (curves) of dichloromethane
(DCM) blood concentrations (upper panel) and amount of dichloromethane
exhaled (lower panel) following 10 and 50 mg/kg intravenous DCM injections
in rats.
C-14
DRAFT - DO NOT CITE OR QUOTE
-------
The most likely explanation for the over-prediction of all four models versus the
experimental observations in Figure C-4 is some shortcoming in model structure or parameter
specification (e.g., the model lacks an explicit blood-volume compartment, which could
significantly impact predictions for this particular route of administration). The discrepancy
appears to exist primarily in the initial distribution phase, as the slope of the simulated clearance
curve closely matches that of the data beyond 10 minutes (0.17 hours). When the simulations are
adjusted downwards to match the observed level at 10 minutes after dosing, the fit to the
remaining blood dichloromethane data was excellent (results not shown). Thus, it appears that
the persistent error between model predictions and data, especially for the 50 mg/kg dose,
beyond the initial distribution phase would be eliminated if a change in the model were
introduced that only impacted that initial distribution phase. In contrast, Variation A matches the
3000 ppm closed-chamber data almost exactly at 1.75 hours (Figure C-3), but then deviates from
the data as time progresses.
There were also significant differences between model predictions of % dichloromethane
dose expired after intravenous injections and observed amounts (Figure C-4, lower panel).
Variation C predictions are slightly closer to the data than Variation A at 50 mg/kg, while the
two models were almost identical at 10 mg/kg. Like the blood concentration data, if simulation
results are adjusted downward to match the initial measurements of % exhaled dichloromethane,
the subsequent predictions match the remaining data points much better (results not shown).
Thus, the error again appears to come from the initial distribution phase.
Model fits to the open-chamber inhalation data of Andersen et al. (1987) are shown in
Figure C-5. While both variations fit the dichloromethane blood concentrations reasonably well
during the 4-hour inhalation phase at 1000 ppm, neither fit the 1000 ppm post-exposure
clearance data well. Both variations also over-predict the initial uptake phase at 200 ppm (Figure
C-5, lower panel); while the difference appears larger than occurred at 1000 ppm (upper panel),
this is only due to the difference in y-axis scale of the two plots and in fact is about the same as
occurs at 1000 ppm. In contrast to the 1000 ppm results, the clearance-phase data at 200 ppm
are fairly well predicted. The fit of Variation C to the 1000 ppm data is only slightly better than
Variation A, while the two are almost identical at 200 ppm.
C-15 DRAFT - DO NOT CITE OR QUOTE
-------
Data
Model A
Model C
Time (h)
Figure C-5. Observations of Andersen et al. (1987; data points) and
simulations (curves) for models A and C for dichloromethane (DCM) in rat
blood from inhalation of 200 and 1000 ppm DCM for 4 hours.
The first clear difference between model predictions with Variations A and C are in the
simulations of COHb after 200 and 1014 ppm inhalation exposures, shown in Figure C-6, with
data from Andersen et al. (1991). Variation A fits the peak COHb levels better than Variation B,
and both tend to over-predict the 1-hour data (first time point) and the last two or three time
C-16
DRAFT - DO NOT CITE OR QUOTE
-------
points. Here Variation C fits the data better than Variation A. For example, at 1014 ppm,
Variation A over-predicts 16 of the 24 data points, while Variation C is less biased, over-
predicting only 13 of the data points. Similarly at 200 ppm, Variation A over-predicts 17 of the
21 data points, while variation C only over-predicts 12 of the 21 measurements.
o
o
12
10
8
6
4
2 H
O
O
0
0
10
7 -
6
5
4
3 H
2
1
0
1014 ppm
• Data
Model A
Model C
200 ppm
0
3
Time (h)
Figure C-6. Observations of Andersen et al. (1991; data points) and
simulations (curves) for models A and C for percent saturation of
carboxyhemoglobin (% COHb) in rat blood from inhalation of 200 and 1014
ppm dichloromethane (DCM) for 4 hours.
C-17
DRAFT - DO NOT CITE OR QUOTE
-------
A quantitative comparison of Variations A and C is presented in Table C-2, showing data
(Andersen et al., 1991; NTP, 1986) and model predictions for the fraction (%) of 50 and 200
mg/kg oral gavage doses exhaled as dichloromethane or CO over 24 hours following exposure.
Variation A over-predicts 3 of the 4 measurements by a factor of 1.03-1.04 (ratio of predicted
value to mean measured value), and over-predicts the exhaled CO for a dose of 50 mg/kg by a
factor of 1.20, indicating a slight underestimation of total dichloromethane metabolism and over-
estimation of CO yield from that metabolism. The ratio of Variation C-predicted values to the
mean measured values was 1.00-1.01 for three of the four values and 0.86 for the fourth. Thus
the overall error from Variation C versus the exhalation data in Table C-2 is less than A
although, like Variation A, there is one value that is much farther off. In this case, however it is
an under-prediction of the amount of CO exhaled at 200 mg/kg. The difference between
Variation C and the mean measured values was less than 1% of the dose, while Variation A
differed from the measured values by up to 3% of the dose.
Table C-2. Comparison of PBPK model Variations A and C predictions of
the amount of dichloromethane either exhaled unchanged or as carbon
monoxide (CO)
Dose
metric
% DCM
%DCM
%CO
%CO
Dose
(mg/kg)
50
200
50
200
Measured*
64.2 ±4.3
77.6 ±0.8
16.2 ±1.2
7.2 ±1.0
Variation A*
66.8
79.9
19.4
7.5
Variation C*
64.7
77.3
16.3
6.24
Variation
A/mean
1.04
1.03
1.20
1.04
Variation
C/mean
1.01
1.00
1.01
0.87
* Amounts are the cumulative percent of total dichloromethane dose collected in exhaled breath at 24
hours after oral gavage by Angelo et al. (1986b). Measured values are mean ± SD of reported mean
values for three groups of six rats each; the first group was studied without prior exposure, while the
second and third groups were studied on the seventh and fourteenth daily exposure, respectively. (Data
sets were combined since no trend in dosimetry with repeated dosing over this period was detected at
either 50 or 200 mg/kg.)
A final comparison of Variations A and C is shown in Figure C-7: predictions of blood
COHb during and after a 30-minute inhalation exposure to 5159 ppm dichloromethane,
conducted by Andersen et al. (1991). It must first be noted that Andersen et al. (1991) changed
several parameters from the primary set listed in Table 2 of their publication when depicting
their model fit to these data (as noted in the legend of their Figure 4). In order to provide a
relevant comparison of the two models, the parameters listed in their Table 2 (Variation A in
Table C-l here) were used. However, the Vmax listed in Figure 4 of Andersen et al. (1991) is
consistent with that value of Vmaxc, given a rat BW = 0.24 kg. Therefore this BW value was
used for both simulations shown. With the exception of that adjustment in BW, it can be said
that these are data to which Variation C have not been fit.
C-18
DRAFT - DO NOT CITE OR QUOTE
-------
c
jj
_0
o>
o
a;
X
o
JJ
5
u
13
O
_o
QQ
10 H
8-
6-
4-
2-
O
O
o Data
Model A
Model C
Exposure period
0
3 4
Time (h)
Figure C-7. Observations of Andersen et al. (1991; data points) and
simulations (curves) for models A and C for percent saturation of
carboxyhemoglobin (% COHb) in rat blood from inhalation of 5159 ppm
dichloromethane (DCM) for 30 min. The model Variation A simulation uses
the parameters listed in Table 1 of Andersen et al. (1991) rather than those
listed in Figure 4 of that publication to provide a fair comparison of the two
variations (i.e., both are restricted to a single set of parameters). However, a
value of Vmax consistent with that listed by Andersen et al. could be obtained
by setting the rat body weight to 0.24 kg, so this was done for both
simulations.
Model Variation A clearly fits the data in Figure C-7 better than Variation C. As noted
just above, these data were not used in estimating the parameters for Variation C. Given that
Andersen et al. (1991) used different parameters in showing their model's ability to reproduce
these data, and the methods do not state that these data were used in model fitting, it would
appear that this is also the case for Variation A. Both models capture the overall shape of the
time-course fairly well, specifically showing a continued rise in COHb levels until 2.3-2.5 hours,
almost 2 hours after exposure ends. The difference between Variation C and the data is less than
30% of the measured values, except for the point at 1.5 hours, which appears to be an outlier. So
for data to which a model has not been fit, this level of agreement would generally be considered
good. As will be shown in the next section, Variation C actually predicts higher CYP
metabolism than Variation A. Since CO is assumed to be produced from CYP metabolism, the
lower amount of COHb predicted here can be attributed to the lower CO yield estimated for
Variation C (0.68) versus that used in Variation A (0.8). These results therefore suggest that this
yield is not a constant but increases with concentration. Nevertheless the possibility that
C-19
DRAFT - DO NOT CITE OR QUOTE
-------
Variation C under-predicts high-concentration CYP metabolism (and Variation A even more so),
should be noted.
In summary, both model Variations A and C fit the data considered (Figures C-3 to C-7
and Table C-3) fairly well, although some sets (e.g., exhaled dichloromethane in Figure C-4) less
well than others. In an absolute sense there are no substantial differences between the goodness
of fit to the fitted data for each model, and where differences occur versus the fitted data,
Variation C is consistently better in describing the overall data set. For example, in Table C-2
the total discrepancy between Variation C and the data is less than that versus Variation A,
although there is one measurement for which A is better. The resulting difference in the
likelihood function computed by acslX, shown in Table C-l, is over 66 log-units indicating a
high statistical significance. The likelihood calculated by acslX is only approximate; in
particular the calculation assumes independence of each datum which is clearly not the case for
the gas uptake data shown in Figure C-3. Nevertheless the difference in approximate likelihood
values reflects a systematic reduction in the total relative error with Variation C versus Variation
A.
Variation C does fit the COHb data in Figure C-7 less well than Variation A, although the
difference is no more than that shown for the 50 mg/kg data in Figure C-4 and the 200 ppm data
in Figure C-5. Given the shortcomings that appear in the CO/COHb sub-model discussed
elsewhere in this Appendix and that Variation C otherwise predicts higher CYP metabolism than
Variation A, Variation A's superior fit to this one data set was discounted relative to the
quantitative improvement of Variation C to the fitted data. As shown in Section C.2.3, the two
variations give rise to non-trivial differences in internal dose metric predictions. These
differences are considered sufficient justification to consider Variation C over Variation A.
C.2.3. Evaluation of Relative Flux of CYP and GST Metabolism of Dichloromethane
The estimated hepatic metabolism through the CYP or GST pathways during simulated
medium- and high-level inhalation exposures of rats to dichloromethane for model Variations A
and C is shown in Table C-3. Chronic exposures of 200, 1,000, 2,000, and 4,000 ppm, reflecting
exposures used by Andersen et al. (1991) and NTP (1986) were simulated. Only a slight
difference was seen in predicted CYP dosimetry between Variations A and C, but a larger
difference in GST dosimetry. Variation C predicted 1.5-3% less CYP metabolite production in
the liver than Variation A, depending on the exposure. However GST-mediated metabolism was
25-28% higher for Variation C than Variation A. At 200 ppm where the bulk of metabolism was
via the CYP pathway, the result was almost identical total metabolism, but at 4000 ppm
Variation C predicted 18% higher total metabolism. Thus, while model fits to experimental data
shown in the previous section were mostly similar for Variations A and C, the predicted GST
and total metabolism internal dose metrics under bioassay conditions show a greater difference.
C-20 DRAFT - DO NOT CITE OR QUOTE
-------
Table C-3. Effect of PBPK model variation on predicted dichloromethane
metabolite production in the liver of (male) rats from inhalation exposures3
Model
variation
A
C
A
C
A
C
A
C
Exposure
(ppm)
200
200
1,000
1,000
2,000
2,000
4,000
4,000
CYP-mediated
metabolite
production
(mg/L liver/d)
658
640
893
877
915
901
979
964
GST-mediated
metabolite
production
(mg/L liver/d)
64.3
82.6
598
747
1384
1725
2873
3581
Total
metabolite
production (mg/L
liver/d)
722
723
1491
1625
2298
2626
3852
4544
GST:CYP
metabolite
production
ratio
0.098
0.129
0.67
0.85
1.5
1.9
2.9
3.7
"Inhalation exposures of 200 or 1,000 ppm for 4 hr/d (Andersen et al., 1991) or 2,000 or 4,000 ppm
dichloromethane for 6 hr/d, 5 d/wk for 2 yrs (NTP, 1986).
To elucidate the exposure ranges where internal doses are linear and where CYP
metabolism saturates, weekly average CYP and GST liver metabolic rates were simulated for
Variation A and C over a wide span of inhalation concentrations, given a regimen of 6 hours/day,
5 days/week, with results shown in Figure C-8. Metabolite production is predicted to be
approximately linear up to -20 ppm for both models. From 20-100 ppm CYP metabolism is still
largely linear, but the small degree of saturation in that pathway which begins in that range leads
to a larger relative increase, hence greater-than-proportional increase, in the GST pathway. This
occurs because lower relative CYP metabolism leaves more dichloromethane available for the
GST pathway. CYP metabolism becomes mostly saturated between 100 and 1,000 ppm but is
not fully saturated even at 2,000 ppm. CYP metabolism is not predicted to be fully saturated by
2,000 ppm because exposures are only 6 hours/day, and blood concentrations fall quickly after
each exposure. Since dichloromethane blood levels still increase with exposure level during that
exposure-off period (because they are higher at the end of the exposure-on period), there is also
increased metabolite production with exposure level during these exposure-off periods.
However, the increase in metabolism per exposure-ppm is much less in this high-concentration
range than occurs in the lower-concentration range because blood concentrations are falling
throughout the exposure-off period. Depending on the model, GST metabolism is predicted to
become dominant above 1,100-1,300 ppm.
C-21
DRAFT - DO NOT CITE OR QUOTE
-------
500 n
20-,
f— GST, model A
-----CYP, model A
— GST, model C
CYP, model C
500 1,000 1,500
Exposure concentration (ppm)
Figure C-8. Simulation results using model Variations A and C for weekly
average metabolic rates by the GST and CYP pathways for 6 hours/day, 5
days/week inhalation exposures.
C.2.4. Evaluation of Model Predictions of Oral Absorption of Dichloromethane
For Variation C, the final set of parameters for which results are shown here was
numerically fit by global optimization to: 1) the inhalation and intravenous exposure data shown
previously; 2) data for blood and liver dichloromethane levels, and total expired dichloromethane
and CO levels, measured by Angelo et al. (1986b) for rats exposed to gavage doses of 50 and
200 mg/kg dichloromethane; and 3) blood COHb levels observed after 526 mg/kg bolus doses by
Pankow et al. (1991a). The parameter that specifically determines oral absorption identified
through this global optimization is the GI absorption constant, ka, for which the optimized value
was ka = 4.31/h. Resulting fits to the oral exposure data are shown in Figures C-9 and C-10.
(Note that for the % expired as CO, panel C of Figure C-9, the percentage declines with
increased concentration, balancing the increased percentage exhaled as dichloromethane.)
C-22
DRAFT - DO NOT CITE OR QUOTE
-------
150
D 50 mg/kg data
— — —50 mg/kg sim
50 mg/kg, alt KA
200 mg/kg data
200 mg/kg sim
200 mg/kg, alt KA
1 2
B
8 12 16 20 24
Time (h)
50 mg/kg data
1 — 50 mg/kg sim
50 mg/kg, alt KA
200 mg/kg data
200 mg/kg sim
200 mg/kg, alt KA
234
Time (h)
Figure C-9. Observations of Angelo et al. (1986b; data) and model Variation
C predictions for: (A) percent dose expired as dichloromethane (DCM); (B)
blood DCM; (C) percent expired as CO; and (D) liver DCM in rats given a
single dichloromethane gavage dose of 50 and 200 mg/kg, using a
numerically fitted GI absorption rate constant (ka = 4.31/h, heavy lines) and
an alternate value of the constant (ka = 0.62/h, thin lines).
C-23
DRAFT - DO NOT CITE OR QUOTE
-------
.Q
I
o
o
10
9
8
7-
6-
5-
4-
3-
2-
1 -
0-
Data
Model (global-fit parameters;
Re-fit absorption constant
0
10
12
Time (h)
Figure C-10. Model predictions with of blood carboxyhemoglobin (COHb,
percent of total Hb) from a single gavage dose of 526 mg/kg DCM in rats,
compared to the data of Pankow et al. (1991a). Model simulations performed
with model C (heavy black line, ka = 4.31/h) or with an alternate value of ka =
0.62/h (thin grey line).
The model fits to the oral PK data of Angel o et al. (1986b), shown in Figure C-9, are fair
though not as good as one might like. (In Figure C-9, the heavy lines are for the model with the
global-fit ka.) As discussed in Section 3.1, the model was not expected to fit the shorter-time CO
exhalation data (Figure C-9C) and indeed the model predicts much more rapid exhalation than
observed up to 6 hours, though it matches the total exhaled CO (cumulative amount) at 24 hours.
In fact a similar pattern occurs with exhaled dichloromethane in Figure C-9A, although the
discrepancy is not as large as for CO. The dichloromethane levels measured in liver and blood
are also over-predicted: only slightly so for blood (Figures C-9B) but by approximately threefold
for liver. It should be noted that the timing of the concentration peak and subsequent clearance
phase match the shape of the data reasonably well. Since the over-prediction of liver and blood
concentrations suggests that the absorption rate constant is too high (and as will be seen below,
reducing ka can improve the fit to other data), the change in predicted dosimetry from using a
lower value of ka is shown in Figure C-9 (thin lines). The result illustrates that reducing ka alone
not only reduces the predicted peak concentration but also alters the predicted fraction converted
to CO (and overall metabolic efficiency) and the timing of the peak tissue levels. More
specifically, reducing ka with other parameters held constant: 1) reduces the fraction of
dichloromethane expected to be exhaled (Figure C-9 A); 2) increases the amount predicted to be
converted to and exhaled as CO (Figure C-9B); and 3) shifts the predicted peak blood and liver
C-24
DRAFT - DO NOT CITE OR QUOTE
-------
concentrations from ~ 9 minutes and 1 minutes, respectively to ~ 45-55 minutes, much later than
shown by the data.
Another possible explanation for the over-prediction of the liver concentration data, in
particular, is that with a delay of even a few minutes between exsanguination and tissue removal
from the experimental animals used, metabolism is likely to continue and tissue concentrations
could drop significantly. What the model predicts at 15 minutes after a bolus exposure, for
example, the nominal time of data collection, does not account for this continued metabolism.
The potential for such an effect to significantly alter internal dosimetry was illustrated by
Sweeney et al. (1996) for butadiene.
Pankow et al. (1991a) measured blood COHb following a gavage dose of 525 mg/kg,
data which might also be used to estimate ka. Model fits to these data are shown in Figure C-10.
The heavy, black line is the fit obtained with the globally-fitted oral absorption rate constant.
The thinner grey line is the result of re-fitting ka to just these data while holding all other
parameters constant. While the globally-fitted model matches the initial data, up to ~ 4 hours,
fairly well, it then predicts a rapid decline in COHb levels whereas the data show that blood
levels remain elevated to ~ 8 hours, followed by a more gradual decline. The re-fitted value of ka,
0.62/hour, describes this particular data set fairly well. Unfortunately, as shown above, this
alternate absorption constant gives a much poorer fit to the other oral dosimetry data. Thus,
while it is possible to improve the fit to specific data by choosing different values of ka, doing so
reduces the overall model quality. As discussed previously, the inability to describe the kinetics
of COHb (with the globally-fit parameters) may also be due to the assumption that CO does not
distribute into muscle and other tissues: the more gradual rise and fall shown by the data in
Figure C-10 might be better predicted if diffusion of CO into and out of tissues, especially
slowly-perfused tissues, was included.
Another factor in attempting to simulate the CO data along with other data sets is the
difference in vehicle: Pankow et al. (1991a) used Oleum pedum tauri while Angelo et al.
(1986b) used water with PEG. Withey et al. (1983) compared oral uptake rates for four
halogenated hydrocarbons, including dichloromethane, and found that dosing in corn oil versus
water lead to a significant decrease in the rate and apparent extent of uptake, with lower peak
concentrations and a longer time-to-peak occurring with corn oil vehicle. The kinetics of orally
administered dichloromethane with the water/PEG vehicle (Figure C-9), clearly show the blood
concentration already clearly falling from the peak at 30 minutes (panel B) and most of the
dichloromethane exhalation being complete by 4 hours (panel A). Further, the difference in
percent exhaled as dichloromethane and CO between the 50 mg/kg and 200 mg/kg data (panels
A and C) is explained by the model as resulting from saturation of the CYP pathway, which
occurs because of the peak dichloromethane concentrations that result in turn because there is a
rapid initial absorption rate. If the rate of absorption from the GI tract is reduced to match the
slower rise in COHb levels observed by Pankow et al. (1991a) over the first 4-6 hours (data in
C-25 DRAFT - DO NOT CITE OR QUOTE
-------
Figure C-10), the result is that most of the absorbed dose is predicted to be metabolized on first-
pass, so blood dichloromethane levels remain below 5 mg/L, and the predicted fraction exhaled
as dichloromethane is much lower than observed, while the dose-dependence of the fraction
exhaled as CO is lost (results not shown). Thus, while it is possible that the dichloromethane and
exhaled CO kinetics of Pankow et al. (1991a) did differ that much from those observed by
Angelo et al. (1986b), it is not possible to model both data sets with a single set of parameters,
given the current model structure. Even if the apparent peak at 6 hours in Figure C-10 is due to
experimental noise or variability (i.e., COHb concentrations plateau from ~ 4-8 hours), the slow
rise in COHb from 0-4 hours requires either a significant change in model parameters or
structure to describe.
One should also note that the CO exhalation data in Figure C-9C show continued
elimination between 6 and 24 hours, a time-range when liver and blood dichloromethane levels
have fallen to near zero, indicating that absorption is nearly complete. This suggests that the
extended period of COHb elevation observed by Pankow et al. (1991a) in fact represents what
occurred (but was not measured) in the experiments of Angelo et al. (1986b). If that is the case
then the discrepancy between the model simulation (with global-fit parameters) and data in
Figure C-10 must occur because the model structure does not fully describe CO/COHb kinetics,
rather than a simple difference in absorption kinetics.
Model Variation C simulations of total expired dichloromethane for oral doses with both
globally fit parameters (values in Table C-l) and absorption constants re-fit to the Pankow et al.
(1991a) data (values in legend of Figure C-10) are compared to the data of Kirschman et al.
(1986) in Table C-4. While the Pankow-fit parameters give somewhat lower fraction exhaled,
e.g., 70.9% versus 78.4% predicted with the global-fit parameters at 250 mg/kg, the difference is
much less than that in the data, in the direction expected. In particular, with slower absorption
one expects the liver concentration of dichloromethane to remain lower, in which case there is
less metabolic saturation, allowing a higher fraction of the dose to be metabolized, thereby
leaving a lower fraction to be exhaled. Since the two parameter sets give such similar results,
however, only the global-fit parameter-based simulations will be specifically discussed. While
the model-predicted exhalation is more than double that measured with the corn-oil vehicle at
250 mg/kg, the over-prediction is only 27% versus the corn-oil observation at 2000 mg/kg.
More relevant to bioassay exposures, the model prediction is only 18% higher than the observed
amount exhaled using the water vehicle at 250 mg/kg and only 4% higher than observed for a
500 mg/kg dose in water. The most obvious explanation for the difference in fraction exhaled
with corn oil versus water vehicle is that absorption is slower with corn oil, allowing for more
efficient (less saturated) metabolism, hence a lower fraction exhaled. The value of the first
absorption constant, ka, fit to the Pankow et al. (1991a) data is 14% of the global-fit value.
These 250 mg/kg corn oil vehicle exhalation data can be matched if ka is reduced to 0.087/h: 2%
of the globally-fitted value. It is possible that other factors are also involved; the balance
C-26 DRAFT - DO NOT CITE OR QUOTE
-------
between saturable and first-order metabolism in the model may be different in the animals used
by Kirschman et al. (1986) versus the model, since the model fails to capture the full reduction in
percent exhaled for the water vehicle between 500 and 250 mg/kg. But neither is it implausible
that absorption should be substantially slowed by the corn oil vehicle. Since the model does
reproduce the water vehicle data fairly well, these results are otherwise a reasonable if limited
validation of the model.
Table C-4. Observations and predictions of total expired dichloromethane
resulting from gavage doses in ratsa
Dose
(mg/kg)
250
500
1,000
2,000
Dichloromethane exhaled (% of dose)
Observations
Corn oil vehicle
36.7
Not reported
45.7
65.7
Water vehicle
66.5
78.3
Not reported
Not reported
Predictions
Global-fit
78.4
81.2
82.9
83.8
Pankow-fit
71.4
77.2
80.6
82.6
"The y-axis of Figure 2 of Kirschman et al. (1986) appears to be mislabeled as mg instead of mg/kg; mg would
result in unrealistic values. Assuming this is supposed to be mg/kg, and assuming average value of 250 g for a
F344 rat, an expiration of 1,300 mg/kg given a dose of 2,000 mg/kg corresponds to 65% of the administered dose.
This value is consistent with their observation in mice where 55% was observed expired as dichloromethane from a
dose of 500 mg/kg, with the percentage expired increasing with dose.
Source: Kirschman etal. (1986).
A final comparison of model Variation C predictions to exhaled dichloromethane data is
shown in Figure C-l 1. Model simulations for 1 and 50 mg/kg bolus oral doses are shown along
with corresponding data from McKenna and Zempel (1981), as well as the data of Angelo et al.
(1986b) at 50 mg/kg. The difference between the results of Angelo et al. (1986b; used in model
calibration) and McKenna and Zempel (1981; not used for calibration) at 50 mg/kg dose shows
the level of inter-laboratory variability. While the model matches the 50 mg/kg data of
McKenna and Zempel (1981) quite well up to 1.5 hours, it then approaches a maximum around
65% versus the 72.5% observed by those investigators. The estimated fraction exhaled for 1
mg/kg dose is likewise slightly below the observations for that exposure. However the
discrepancies - less than 10% of the measured values - are well within what might result from
experimental variability in both cases. Thus, these results are taken as confirmation of the
model's ability to predict dosimetry at low exposure levels.
C-27
DRAFT - DO NOT CITE OR QUOTE
-------
70
5 60
50
o 40
TJ
W o
"5 20
0
0
* 1 mg/kg data
1 mg/kg sirn
• 50 mg/kg data
^^— 50 mg/kg sim
Q Angelo et al. (1986b) 50 mg/kg ^
3 4
Time (h)
Figure C-ll. Comparison of model Variation C predictions to
dichloromethane (DCM) exhalation data of McKenna and Zempel (1981)
(data not used for model calibration) from bolus oral exposures to 1 and 50
mg/kg dichloromethane, along with 50 mg/kg bolus oral data of Angelo et al.
(1986) (data used for model calibration).
C.3. MODEL OPTION SUMMARY
The revised model presented here, Variation C, is a rat PBPK model represented by the
basic model structure of Andersen et al. (1991) with the inclusion of lung dichloromethane
metabolism via CYP (0.2% of liver) and GST (14.9% of liver) pathways (estimated from Reitz et
al., 1989) and with liver metabolic parameters and CO yield factor re-calibrated against multiple
data sets. Inclusion of lung metabolism in this model provides increased biological realism
compared to the model of Andersen et al. (1991) and the re-calibration provides better overall
model agreement with the available rat data sets, using a single set of parameters. While the
fraction of CYP metabolism estimated in the rat lung is not quantitatively significant, all three
metabolic parameters for the liver (Vmaxc, Km, and kfc) were likely impacted to some extent
during the re-optimization based on the presence of lung GST metabolism, since splitting GST
metabolism into the lung compartment reduces the extent of flow-limitation on that pathway
Our evaluation also revealed some deficiencies of the available models. Oral exposure
data show a slower rise to peak COHb levels and slower rate of exhalation of CO than predicted
by the model, with the model-simulated rates for CO and COHb being largely determined by the
kinetics of dichloromethane. Thus, it is not possible to fit those CO data using the existing CO
C-28
DRAFT - DO NOT CITE OR QUOTE
-------
sub-model and parameters, without significantly degrading the fit to the parent dichloromethane
data. Cho et al. (2001) used a perfused rat hind-limb preparation to demonstrate that CO
permeates muscle tissue (part of the slowly perfused tissue group in the PBPK model) to a large
extent, slowing its wash-out from the hind-limb significantly versus radio-labeled albumin or
sucrose. In fact distribution of CO into muscle and other body tissues was incorporated into a
human CO PBPK model by Bruce et al. (2008). The CO sub-model, on the other hand, assumes
that CO is not distributed beyond the blood compartment. That the CO sub-model does not
describe its distribution into other tissues could explain the discrepancy between the observed
CO and COHb kinetics and model predictions.
Cho et al. (2001) also found that a significant fraction of CO was retained by the perfused
tissue, estimating that the long-term fractional recovery to only be 45%. The yield factor fitted
in the PBPK model here, PI, must then in part account for CO which is in fact produced but not
exhaled. That PI was estimated to be 0.68, or 68%, for Variation C suggests that either less than
45% of CO is sequestered in the dichloromethane pharmacokinetics experiments or that its
production is underestimated. Since the purpose of this analysis was to identify a model that best
estimates dichloromethane kinetics in the rat with only moderate adjustments to the model
structure, and elaboration of the CO sub-model would result in additional parameters that need to
be given values or identified with the data, the structure of the CO sub-model was left intact and
only the yield factor was adjusted.
Reitz et al. (1997) employed a ka value of 5.0 hours"1 for dichloromethane in deriving
acute- and intermediate-duration oral minimal risk levels. These investigators cited previous
work in which a ka value of 5.0/hour resulted in reasonable kinetic predictions for bolus or
drinking water exposures of 1,1,1-trichloroethane in rats (Reitz et al., 1988c), while a value of
5.4/hour provided good agreement between toxicokinetic observations and predictions for bolus
doses of trichloroethylene in rats (Fisher et al., 1989). Pastino and Conolly (2000) fitted ka =
4.15/h for ethanol absorption in rats. Thus, the value of ka obtained here, 4.31/h, is consistent
with other oral kinetic data. It must also be noted that since the model assumes 100% absorption
of an oral dose, there is no other route of elimination except systemic circulation. However, at
lower doses or dose-rates a larger fraction of the dose will be eliminated on first pass through the
liver, hence never reach systemic circulation, which can give the appearance of lower
bioavailability. Finally, when long-term exposure patterns (comparable to chronic bioassays) are
simulated, the average rate of absorption must equal the average (measured or estimated) rate of
ingestion, independent of the values for these constants.
In summary, we have examined a common PBPK model structure with three parameter
sets, (Variations A-C, to describe dichloromethane dosimetry in rats as candidates for use in risk
assessment where the purpose is to estimate internal doses of dichloromethane that occurred
during various bioassays. In comparing model predictions to a variety of data, one can say that
while all of the model variations do a fairly good job of fitting some of the data, none of them fit
C-29 DRAFT - DO NOT CITE OR QUOTE
-------
all of the data very well, and there are some data for which none of the models provides a
particularly good fit. With respect to the dichloromethane kinetics from inhalation and
intravenous exposures shown in Figures C-3 to C-5, there was very little difference between
predictions with the unadjusted parameters (Variation A) and the final revised parameters
(Variation C). However, predictions of COHb levels after inhalation exposures were 10-20%
lower with Variation C (Figures C-6 and C-7) while Variation C predicted a ~ 25% higher rate of
GST-mediated metabolism (Figure C-8, Table C-3) and hence a lower fraction of
dichlorom ethane exhaled unchanged after oral gave (Table C-2) compared to Variation A. The
rate of CYP metabolism is only slightly lower with Variation C versus A. While some of these
differences are small, the estimated likelihood (LLF in Table C-l) indicates that the re-fitted
parameters provide a significantly better fit to the data used for calibration. This evaluation also
indicates that the existing model structure for CO and COHb does not adequately describe the
corresponding data for any of the parameter sets examined and that attempts to specifically use
those data in setting key parameters would compromise the accuracy of the key dichloromethane
dosimetry estimates. Therefore, Variation C is judged to be sufficiently better than the original
model (Variation A) to support the use of Variation C instead of Variation A. Variation C was
able to simulate exhaled dichloromethane data after oral dosing to which it had not been fit
reasonably well (Figure C-l 1) and likewise provided fair agreement with water-vehicle data
from another source (Table C-4, Global-fit predictions at 250 and 500 mg/kg). Hence, the model
is expected to adequately predict rat internal dosimetry (dichloromethane blood concentrations or
rates of metabolism) under bioassay exposure conditions.
C-30 DRAFT - DO NOT CITE OR QUOTE
-------
APPENDIX D. SUMMARY OF BENCHMARK DOSE (BMD) MODELING OF
NONCANCER ENDPOINTS
D.I. ORAL RfD: BMD MODELING OF LIVER LESION INCIDENCE DATA FOR
RATS EXPOSED TO DICHLOROMETHANE IN DRINKING WATER FOR 2 YEARS
(SEROTA ET AL., 1986a)
BMD and BMDL refer to the model-predicted dose (and its lower 95% confidence limit)
associated with 10% extra risk for the incidence of liver foci/areas of cellular alteration in male
and female F344 rats given dichloromethane in drinking water for 2 years (Serota et al., 1986a)
(Table D-l).
Table D-l. Incidence data for liver lesions and internal liver doses based on
various metrics in male and female F344 rats exposed to dichloromethane in
drinking water for 2 years (Serota et al., 1986a)
Sex
Male
(BW =
380 g)
Female
(BW =
229 g)
Nominal (actual)
daily intake
(mg/kg-d)
0(0)
5(6)
50 (52)
125 (125)
250 (235)
0(0)
5(6)
50 (58)
125 (136)
250 (263)
Rat liver
lesion incidence"
52/76 (68%)
22/34 (65%)
35/38 (92%)c
34/35 (97%)c
40/41 (98%)c
34/67(51%)
12/29 (41%)
30/41 (73%)c
34/38 (89%)c
31/34(91%)c
Rat internal liver doseb
CYP
0
131.6
723.9
1,170.5
1,548.0
0
132.8
801.4
1,261.5
1,672.4
GST
0
2.60
81.5
276.5
612.1
0
2.47
93.3
307.8
705.6
GST and
CYP
0
134.2
805.3
1,446.9
2,160.1
0
135.3
894.7
1,569.3
2.378.0
Parent
AUC
0
0.58
18.1
61.3
135.7
0
0.47
17.8
58.6
134.4
"Liver foci/areas of cellular alteration; number affected divided by total sample size.
blnternal doses were estimated using a rat PBPK model from simulations of actual daily doses reported by the study
authors. CYP dose is in units of mg dichloromethane metabolized via CYP pathway/L tissue/d; GST dose is in
units of mg dichloromethane metabolized via GST pathway/L tissue/d; GST and CYP dose is in units of mg
dichloromethane metabolized via CYP and GST pathways/L tissue/d; and Parent AUC dose is in units of mg
dichloromethane x hrs/L tissue.
'Significantly (p < 0.05) different from control with Fisher's exact test.
Source: Serota et al. (1986a).
All available dichotomous models in the BMDS (version 2.0) were fit to male and female
rat internal tissue doses of dichloromethane metabolized by the CYP pathway and incidences for
animals with these liver lesions observed at the time of death (Table D-2). (The quantal model is
identical to the one-stage multistage model and so is not included in this set of models). The
male rats exhibited a greater sensitivity compared to the female rats (based on lower BMDLio
D-l
DRAFT - DO NOT CITE OR QUOTE
-------
values for all of the models), and thus, the male data are used as the basis for the RfD derivation.
The logistic model was the best fitting model for the male incidence data based on AIC value
among models with adequate fit (U.S. EPA, 2000b). (If more than one model shares the lowest
AIC value, BMDLio values from these models may be averaged to obtain a POD. However, this
average is not a well-defined lower bound, and should be referred to only as averages of
BMDLioS. U.S. EPA does not support averaging BMDLs in situations in which AIC values are
similar, but not identical, because the level of statistical confidence is lost and because there is no
consensus regarding a cut-off between similar and dissimilar AIC values.) Results for this model
are presented below.
Table D-2. BMD modeling results for incidence of liver lesions in male and
female F344 rats exposed to dichloromethane in drinking water for 2 years,
based on liver-specific CYP metabolism dose metric (mg dichloromethane
metabolism via CYP pathway per liter liver tissue per day)
Sex and model"
BMD10
BMDL10
x2
goodness of fit
/7-value
AIC
Males
Gamma3
Logistic1"
Log-logistic3
Multistage (I)3
Probit
Log-probit3
Weibull3
138.124
70.69
188.81
57.08
81.83
176.71
105.85
41.28
51.42
38.74
39.68
62.57
66.89
40.69
0.58
0.69
0.80
0.64
0.64
0.77
0.52
185.46
183.85
184.85
184.05
184.01
184.93
185.67
Females
Gamma3
Logistic
Log-logistic3
Multistage (I)3
Probit
Log-probit3
Weibull3
287.41
137.58
340.15
100.41
145.33
336.41
240.71
81.95
109.45
95.35
74.33
118.47
143.31
80.40
0.49
0.53
0.57
0.40
0.53
0.57
0.43
233.19
231.99
232.88
232.72
231.97
232.87
233.43
3These models in U.S. EPA BMDS version 2.0 were fit to the rat dose-response data shown in Table 5-1 by using
internal dose metrics calculated with the rat PBPK model. Details of the models are as follows: Gamma and
Weibull models restrict power >1; Log-logistic and Log-probit models restrict to slope >1, multistage model restrict
betas >0; lowest degree polynomial with an adequate fit is reported (degree of polynomial noted in parentheses).
bBolded model is the best-fitting model in the most sensitive sex (males), which is used in the RfD derivation.
Source: Serotaetal. (1986a).
D-2
DRAFT - DO NOT CITE OR QUOTE
-------
Logistic Model, Male Rats (Scrota et al., 1986a), CYP Metabolism (Rate of Production)
Metric
Logistic Model with 0.95 Confidence Level
T3
£
o
c
o
•*=
o
(0
0.9
0.8
0.7
0.6
0.5
Logistic
BMDL BMD
200
400
600
800
dose
1000
1200
1400
1600
16:5204/282011
Figure D-l. Predicted (logistic model) and observed incidence of noncancer
liver lesions in male F344 rats exposed to dichloromethane in drinking water
for 2 years (Scrota et al., 1986a).
Logistic Model. (Version: 2.12; Date: 05/16/2008)
Input Data File: C:\Usepa\BMDS21\Data\logSerota_new_CYPSetting.(d)
Gnuplot Plotting File: C:\Usepa\BMDS21\Data\logSerota_new_CYPSetting.plt
Thu Apr 28 16:52:22 2011
BMDS Model Run
The form of the probability function is:
P[response] = I/[1+EXP(-intercept-slope*dose)]
Dependent variable = Effect
Independent variable = Dose
Slope parameter is not restricted
Total number of observations = 5
Total number of records with missing values = 0
Maximum number of iterations = 250
DRAFT - DO NOT CITE OR QUOTE
-------
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
background = 0 Specified
intercept = 0.670583
slope = 0.00188727
the user,
intercept
slope
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by
and do not appear in the correlation matrix )
intercept
1
-0.47
Variable
intercept
slope
Model
Full model
Fitted model
Reduced model
AIC:
slope
-0.47
1
Parameter Estimates
Estimate
0.659731
0.00220361
95.0% Wald Confidence Interval
Std. Err. Lower Conf. Limit Upper Conf. Limit
0.210128
0.00049774
0.247887
0.00122806
1.07157
0.00317916
Analysis of Deviance Table
Log(likelihood) # Param's Deviance Test d.f.
-89.2097 5
-89.925 2 1.43061 3
-106.616 1 34.8133 4
183.85
P-value
0.6984
<.0001
0
131
723
1170
1548
Dose
.0000
.6000
.9000
.5000
.0000
Est
0.
0.
0.
0.
0.
. Prob.
6592
7211
9051
9623
9832
Expe
50,
24,
34,
33,
40,
Gooc
scted
.099
.516
.393
.680
.312
Ines
0
52
22
35
34
40
s of Fit
bserved
.000
.000
.000
.000
.000
Size
76
34
38
35
41
S
Re
0
-0
0
0
-0
caled
sidual
.460
.962
.336
.284
.380
ChiA2 =1.48
d.f. = 3
P-value = 0.6880
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 70.6874
BMDL = 51.4235
D-4
DRAFT - DO NOT CITE OR QUOTE
-------
D.2. INHALATION RfC: BMD MODELING OF LIVER LESION INCIDENCE DATA
FOR RATS EXPOSED TO DICHLOROMETHANE VIA INHALATION FOR 2 YEARS
(NITSCHKE ET AL., 1988a)
BMD and BMDL refer to the model-predicted dose (and its lower 95% confidence limit)
associated with 10% extra risk for the incidence of hepatic vacuolation in female F344 rats
exposed to dichloromethane via inhalation for 2 years (Nitscke et al., 1988a) (Table D-3).
Table D-3. Incidence data for liver lesions (hepatic vacuolation) and internal
liver doses based on various metrics in female Sprague-Dawley rats exposed
to dichloromethane via inhalation for 2 years (Nitschke et al., 1988a)
Sex
Male
Female
(BW =
229 g)
Exposure
(ppm)
0
50
200
500
0
50
200
500
Liver lesion
incidence"
22/70(31)
Not reported
Not reported
28/70 (40)
41/70 (59%)
42/70 (60%)
41/70 (58%)
53/70 (76%)c
Rat internal liver doseb
CYP
GST
GST and
CYP
Parent
AUC
Not modeled because results from male rats were not provided for the
50 and 200 ppm groups
Not modeled because results for middle two doses were not reported
0
285.3
665.3
782.1
0
6.17
93.2
360.0
0
291.4
758.5
1,142.1
0
1.18
17.8
68.6
"Number affected divided by total sample size.
Internal doses were estimated using a rat PBPK model using exposures reported by study authors (50 ppm =
174 mg/m3, 200 ppm = 695 mg/m3, and 500 ppm = 1,737 mg/m3) and are weighted-average daily values for 1 wk of
exposure at 6 hrs/d, 5 d/wk. CYP dose is in units of mg dichloromethane metabolized via CYP pathway/L tissue/d;
GST dose is in units of mg dichloromethane metabolized via GST pathway/L tissue/d; GST and CYP dose is in
units of mg dichloromethane metabolized via CYP and GST pathways/L tissue/d; and Parent AUC dose is in units
of mg dichloromethane x hrs)/L tissue.
Significantly (p < 0.05) different from control with Fisher's exact test.
Source: Nitschke etal. (1988a).
All available dichotomous models in the BMDS (version 2.0) were fit to male and female
rat internal tissue doses of dichloromethane metabolized by the CYP pathway and incidences for
animals with these liver lesions observed at the time of death (Table D-4). (The quantal model is
identical to the one-stage multistage model and so is not included in this set of models). The log-
probit model was the best fitting model for the female incidence data based on AIC value among
models with adequate fit (U.S. EPA, 2000c). (If more than one model shares the lowest AIC
value, BMDLio values from these models may be averaged to obtain a POD. However, this
average is not a well-defined lower bound, and should be referred to only as averages of
BMDLioS. U.S. EPA does not support averaging BMDLs in situations in which AIC values are
D-5
DRAFT - DO NOT CITE OR QUOTE
-------
similar, but not identical, because the level of statistical confidence is lost and because there is no
consensus regarding a cut-off between similar and dissimilar AIC values. Results for this model
are presented below.)
Table D-4. BMD modeling results for incidence of liver lesions in female
Sprague-Dawley rats exposed to dichloromethane by inhalation for 2 years,
based on liver specific CYP metabolism metric (mg dichloromethane
metabolized via CYP pathway per liter liver tissue per day)
Model3
Gamma3
Logistic
Log-logistic3
Multistage (3)a
Probit
Log-probita'b
Weibulf
BMD10
622.10
278.31
706.50
513.50
279.23
737.93
715.15
BMDL10
227.29
152.41
506.84
155.06
154.52
531.82
494.87
x2
goodness of fit
/7-value
0.48
0.14
0.94
0.25
0.14
0.98
0.95
AIC
367.24
369.77
365.90
368.54
369.76
365.82
365.88
aThese models in U.S. EPA BMDS version 2.0 were fit to the rat dose-response data shown in Table 5-5 by using
internal dose metrics calculated with the rat PBPK model. Gamma and Weibull models restrict power >1; Log-
logistic and Log-probit models restrict to slope >1, multistage model restrict betas >0; lowest degree polynomial
with an adequate fit reported (degree of polynomial in parentheses).
bBolded model is the best-fitting model in the most sensitive sex (females), which is used in the RfC derivation.
Source: Nitschke et al. (1988a).
D-6
DRAFT - DO NOT CITE OR QUOTE
-------
Log Probit Model, Female Rats (Nitschke et al., 1988a), CYP Metabolism (Rate of
Production) Metric
•
I
o
13
(0
0.85
0.8
0.75
0.7
0.65
0.6
0.55
0.5
0.45
LogProbit Model with 0.95 Confidence Level
LogProbit
BMDL
BMD
100
200
300
500
400
dose
16:5504/282011
Figure D-2. Predicted (log-probit model) and observed incidence of
noncancer liver lesions in female Sprague-Dawley rats inhaling
dichloromethane for 2 years (Nitschke 1988a).
600
700
800
Probit Model. (Version: 3.1; Date: 05/16/2008)
Input Data File: C:\Usepa\BMDS21\Data\lnpNitschke_new_CYPSetting.(d)
Gnuplot Plotting File: C:\Usepa\BMDS21\Data\lnpNitschke_new_CYPSetting.plt
Thu Apr 28 16:55:00 2011
BMDS Model Run
The form of the probability function is:
P[response] = Background
+ (1-Background) * CumNorm(Intercept+Slope*Log(Dose)) ,
where CumNorm(.) is the cumulative normal distribution function
Dependent variable = Effect
Independent variable = Dose
Slope parameter is restricted as slope >= 1
Total number of observations = 4
D-7
DRAFT - DO NOT CITE OR QUOTE
-------
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
User has chosen the log transformed model
Default Initial (and Specified) Parameter Values
background = 0.585714
intercept = -7.71354
slope = 1
Asymptotic Correlation Matrix of Parameter Estimates
the user,
background
intercept
Limit
Variable
background
intercept
slope
( *** The model parameter(s) -slope
have been estimated at a boundary point, or have been specified by
and do not appear in the correlation matrix )
background intercept
1 -0.37
-0.37 1
Parameter Estimates
95.0% Wald Confidence Interval
Estimate Std. Err. Lower Conf. Limit Upper Conf.
0.590372
-120.151
18
0.0339907
0.346802
NA
0.523751
-120.831
0.656992
-119.471
NA - Indicates that this parameter has hit a bound
implied by some ineguality constraint and thus
has no standard error.
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood) # Param's Deviance Test d.f.
-180.889 4
-180.909 2 0.0403892 2
-184.186 1 6.5937 3
P-value
0.98
0.08604
365.818
Goodness of Fit
Dose
0.0000
285.3000
665.3000
782.1000
Est. Prob.
0.5904
0.5904
0.5907
0.7571
Expected
41.326
41.326
41.350
52.998
Observed
41.000
42.000
41.000
53.000
Size
70
70
70
70
Scaled
Residual
-0.079
0.164
-0.085
0.001
ChiA2 =0.04 d.f. = 2
Benchmark Dose Computation
P-value = 0.9800
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0.1
Extra risk
0.95
737.929
531.817
D-8
DRAFT - DO NOT CITE OR QUOTE
-------
APPENDIX E: SUMMARY OF BENCHMARK DOSE (BMD) MODELING OF
CANCER ENDPOINTS
E.I. ORAL CANCER SLOPE FACTORS: BMD MODELING OF LIVER TUMOR
INCIDENCE DATA FOR MICE EXPOSED TO DICHLOROMETHANE IN DRINKING
WATER FOR 2 YEARS (SEROTA ET AL., 1986b; HAZLETON LABORATORIES,
1983)
BMDio and BMDLio refer to the model-predicted dose (and its lower 95% confidence
limit) associated with 10% extra risk for the incidence of hepatocellular adenoma and carcinoma
in male mice given dichloromethane in drinking water for 2 years (Serota et al., 1986b; Hazleton
Laboratories, 1983) (Table E-l). Multistage models were fit to male mouse internal tissue doses
of dichloromethane metabolized by the GST pathway and incidences for animals with liver
tumors observed at the time of death. Different polynomial models were compared based on
adequacy of model fit as assessed by overall ^ goodness of fit (p-value > 0.10) and examination
of residuals at the 0 dose exposure (controls) and in the region of the BMR. Due to the lack of a
monotonic increase in tumor response at the high dose, the model did not adequately fit the data.
In consideration of the region of interest (i.e., low-dose risk estimation), the highest dose group
was excluded. The modeling of the remaining four dose groups exhibited an adequate fit to the
data. The predicted BMDio and BMDLio for the incidence data are 73.0 and 39.6 mg
dichloromethane metabolized via GST pathways per liter tissue per day, respectively, for the
internal liver metabolism metric, and 3.05 and 1.65 mg dichloromethane metabolized via GST
pathway in lung and liver/kg-day, respectively, for the whole body metabolism metric (Table E-
2).
E-l DRAFT - DO NOT CITE OR QUOTE
-------
Table E-l. Incidence data for liver tumors and internal liver doses, based on
GST metabolism dose metrics, in male B6C3Fi mice exposed to
dichloromethane in drinking water for 2 years
Sex
Male
(BW =
37.3 g)
Nominal (actual) daily
intake (mg/kg-d)
0(0)
60 (61)
125 (124)
185 (177)
250 (234)
Mouse liver
tumor incidence"
24/125 (19%)
51/199(26%)
30/99 (30%)
31/98(32%)
35/123 (28%)
Mouse internal liver
metabolism doseb
0
17.5
63.3
112.0
169.5
Mouse whole body
metabolism dosec
0
0.73
2.65
4.68
7.1
"Hepatocellular carcinoma or adenoma combined. Mice dying prior to 52 wks were excluded from the
denominators. Cochran-Armitage trends-value = 0.058. P-values for comparisons with the control group were
0.071, 0.023, 0.019, and 0.036 in the 60, 125, 185, and 250 mg/kg-d groups, respectively, based on statistical
analyses reported by Hazleton Laboratories (1983).
bmg dichloromethane metabolized via GST pathway/L liver/d. Internal doses were estimated from simulations of
actual daily doses reported by the study authors.
°Based on the sum of dichloromethane metabolized via the GST pathway in the lung plus the liver, normalized to
total BW (i.e., [lung GST metabolism (mg/d) + liver GST metabolism (mg/d)]/kg BW). Units = mg
dichloromethane metabolized via GST pathway in lung and liver/kg-d.
Sources: Serota et al. (1986b); Hazleton Laboratories (1983).
Table E-2. BMD modeling results and tumor risk factors for internal dose
metric associated with 10% extra risk for liver tumors in male B6C3Fi mice
exposed to dichloromethane in drinking water for 2 years, based on liver-
specific GST metabolism and whole body GST metabolism dose metrics
Internal
dose metric"
Liver-
specific
Whole-body
BMDS
modelb
MS (1,1)
MS (1,1)
x2
goodness of
fit/7-value
0.56
0.56
Mouse
BMD10C
73.0
3.05
Mouse
BMDL10C
39.6
1.65
Allometric-
scaled human
BMDL10d
5.66
0.24
Tumor risk factor6
Scaling = 1.0
2.53 x 1Q-3
-
Allometric-
scaled
1.77 x 1Q-2
4.24 x 1Q-1
"Liver specific dose units = mg dichloromethane metabolized via GST pathway per liter tissue per d; Whole-body
dose units = mg dichloromethane metabolized via GST pathway in lung and liver/kg-d).
bThe multistage (MS) model in U.S. EPA BMDS version 2.0 was fit to the mouse dose-response data shown in
Table 5-11 using internal dose metrics calculated with the mouse PBPK model. Numbers in parentheses indicate
(1) the number of dose groups dropped in order to obtain an adequate fit, and (2) the degree polynomial of the
model.
°BMD10 and BMDL10 refer to the BMD-model-predicted mouse internal and its 95% lower confidence limit,
associated with a 10% extra risk for the incidence of tumors.
dMouse BMDL10 divided by (BWhuman/BWmouse)025 = 7.
eDichloromethane tumor risk factor (extra risk per unit internal dose) derived by dividing the BMR (0.1) by the
mouse BMDL10 and by the allometric-scaled human BMDL10, for the scaling =1.0 and allometric-scaled risk
factors, respectively.
Modeling results are presented in the subsequent sections for the tissue-specific liver-
metabolism metric (Section E. 1.1) and the whole-body metabolism metric (Section E. 1.2).
E-2
DRAFT - DO NOT CITE OR QUOTE
-------
E.I.I. Modeling Results for the Internal Liver Metabolism Metric
Scrota et al. (1986b), Hazleton Laboratories, 1983: Internal liver dose-response, highest
dose dropped
1-degree polynomial
Multistage Cancer Model with 0.95 Confidence Level
0.4
0.35
0.3
o
'o 0.25
(0
0.2
0.15
0.1
Multistage Cancer
Linear extrapolation
BMDL
BMD
20
40
60
80
100
dose
13:4602/192009
Figure E-l. Predicted and observed incidence of animals with hepatocellular
carcinoma or adenoma in male B6C3Fi mice exposed to dichloromethane in
drinking water for 2 years, using liver-specific metabolism dose metric
(Scrota et al., 1986b; Hazleton Laboratories, 1983).
Multistage Cancer Model. (Version: 1.7; Date: 05/16/2008)
Input Data File: C:\USEPA\IRIS\DCM\Serota\highdosedropped\lMulSerMS_.(d)
Gnuplot Plotting File:
C:\USEPA\IRIS\DCM\Serota\highdosedropped\lMulSerMS_.plt
Thu Feb 19 13:46:49 2009
BMDS Model Run
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl) ]
The parameter betas are restricted to be positive
Dependent variable = incidence
E-3
DRAFT - DO NOT CITE OR QUOTE
-------
Independent variable = dose
Total number of observations = 4
Total number of records with missing values = 0
Total number of parameters in model = 2
Total number of specified parameters = 0
Degree of polynomial = 1
Maximum number of iterations = 250
Relative Function Convergence has been set to: 2.22045e-016
Parameter Convergence has been set to: 1.49012e-008
**** We are sorry but Relative Function and Parameter Convergence ****
**** are currently unavailable in this model. Please keep checking ****
**** the web sight for model updates which will eventually ****
**** incorporate these convergence criterion. Default values used. ****
Default Initial Parameter Values
Background = 0.218634
Beta(l) = 0.00136788
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.7
Beta(l) -0.7 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0.218642 * * *
Beta(l) 0.00144288 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model Log(likelihood) # Param's Deviance Test d.f. P-value
Full model -296.282 4
Fitted model -296.87 2 1.1754 2 0.5556
Reduced model -299.126 1 5.68747 3 0.1278
AIC: 597.74
Goodness of Fit
Dose
0.0000
17.5000
63.3000
112.0000
Est. Prob.
0.2186
0.2381
0.2868
0.3352
Expected
27.330
47.387
28.398
32.853
Observed
24.000
51.000
30.000
31.000
Size
125
199
99
98
Scaled
Residual
-0.721
0. 601
0.356
-0.397
ChiA2 =1.16 d.f. = 2 P-value = 0.5585
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 73.0211
BMDL = 39.6034
E-4 DRAFT - DO NOT CITE OR QUOTE
-------
BMDU = 335.18
Taken together, (39.6034, 335.18 ) is a 90 % two-sided confidence
interval for the BMD
Multistage Cancer Slope Factor = 0.00252504
E-5 DRAFT - DO NOT CITE OR QUOTE
-------
E.1.2. Modeling Results for the Whole Body Metabolism Metric
Scrota et al. (1986b), Hazleton Laboratories, 1983: Internal whole-body metabolism dose-
response in mice, highest dose dropped
1-degree polynomial
Multistage Cancer Model with 0.95 Confidence Level
•
I
C
o
13
ro
0.4
0.35
0.3
0.25
0.2
0.15
0.1
Multistage Cancer
Linear extrapolation
BMDL
BMD
2 3
dose
17:3302/21 2009
Figure E-2. Predicted and observed incidence of animals with hepatocellular
carcinoma or adenoma in male B6C3Fi mice exposed to dichloromethane in
drinking water for 2 years, using whole-body metabolism dose metric (Scrota
et al., 1986b; Hazleton Laboratories, 1983).
Multistage Cancer Model. (Version: 1.7; Date: 05/16/2008)
Input Data File: C:\USEPA\IRIS\DCM\Serota\highdosedropped\lMulSerMS_.(d)
Gnuplot Plotting File:
C:\USEPA\IRIS\DCM\Serota\highdosedropped\lMulSerMS_.plt
Sat Feb 21 17:33:49 2009
BMDS Model Run
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl) ]
E-f
DRAFT - DO NOT CITE OR QUOTE
-------
The parameter betas are restricted to be positive
Dependent variable = incidence
Independent variable = dose
Total number of observations = 4
Total number of records with missing values = 0
Total number of parameters in model = 2
Total number of specified parameters = 0
Degree of polynomial = 1
Maximum number of iterations = 250
Relative Function Convergence has been set to: 2.22045e-016
Parameter Convergence has been set to: 1.49012e-008
****
****
We are sorry but Relative Function and Parameter Convergence ****
are currently unavailable in this model. Please keep checking ****
**** the web sight for model updates which will eventually ****
**** incorporate these convergence criterion. Default values used. ****
Default Initial Parameter Values
Background = 0.218649
Beta(l) = 0.032703
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.7
Beta (1)
-0.7
Parameter Estimates
Variable Estimate Std. Err.
Background 0.218662 *
Beta(l) 0.0344939 *
Indicates that this value is not calculated.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
AIC:
Log(likelihood)
-296.282
-296.871
-299.126
597.741
# Param's
4
2
1
Deviance Test d.f.
P-value
1.17643
5.68747
0.5553
0.1278
Goodness of Fit
Dose
0.0000
0.7310
2.6470
4.6840
Est. Prob.
0.2187
0.2381
0.2868
0.3352
Expected
27.333
47.385
28.397
32.853
Observed
24.000
51.000
30.000
31.000
Size
125
199
99
98
Scaled
Residual
-0.721
0. 602
0.356
-0.396
ChiA2 =1.17
d.f. = 2
P-value = 0.5582
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD =
BMDL =
BMDU =
3.05447
1.65649
14.0263
E-7
DRAFT - DO NOT CITE OR QUOTE
-------
Taken together, (1.65649, 14.0263) is a 90 % two-sided confidence
interval for the BMD
Multistage Cancer Slope Factor = 0.0603686
E-8 DRAFT - DO NOT CITE OR QUOTE
-------
E.2. CANCER IUR: BMD MODELING OF LIVER AND LUNG TUMOR INCIDENCE
DATA FOR MALE MICE EXPOSED TO DICHLOROMETHANE VIA INHALATION
FOR 2 YEARS (MENNEAR ET AL., 1988; NTP, 1986)
BMDio and BMDLio refer to the model-predicted dose (and its lower 95% confidence
limit) associated with 10% extra risk for the combined incidence of adenoma and carcinoma of
the liver or lung of male B6C3Fi mice inhaling dichloromethane for 2 years (Mennear et al.,
1988; NTP, 1986) (Table E-3).
Table E-3. Incidence data for liver and lung tumors and internal doses
based on GST metabolism dose metrics in male B6C3Fi mice exposed to
dichloromethane via inhalation for 2 years
Sex,
tumor type
Male, liver0
Male, lunge
BW(g)
-
34.0
32.0
-
34.0
32.0
External
dichloromethane
concentration
(ppm)
0
2,000
4,000
0
2,000
4,000
Mouse
tumor incidence
22/50 (44%)d
24/47(51%)
33/47 (70%)
5/50 (10%)d
27/47 (55%)
40/47 (85%)
Mouse internal
tissue dose"
0
2,363.7
4,972.2
0
475.0
992.2
Mouse whole body
metabolism doseb
0
100.2
210.7
0
100.2
210.7
Tor liver tumors: mg dichloromethane metabolized via GST pathway/L liver tissue/d from 6 hrs/d, 5 d/wk
exposure; for lung tumors: mg dichloromethane metabolized via GST pathway/L lung tissue/d from 6 hrs/d,
5 d/wk exposure.
bBased on the sum of dichloromethane metabolized via the GST pathway in the lung plus the liver, normalized to
total BW (i.e., [lung GST metabolism (mg/d) + liver GST metabolism (mg/d)]/kg BW). Units = mg
dichloromethane metabolized via GST pathway in lung and liver/kg-d.
°Hepatocellular carcinoma or adenoma. Mice dying prior to 52 wks were excluded from the denominators.
dStatistically significant increasing trend (by incidental and life-table tests; p < 0.01).
eBronchoalveolar carcinoma or adenoma. Mice dying prior to 52 wks were excluded from the denominators.
Sources: Mennear et al. (1988); NTP (1986).
Multistage models were fit to male mouse internal tissue doses of dichloromethane
metabolized by the GST pathway and incidences for animals with liver tumors observed at the
time of death. The predicted BMDio and BMDLio for the liver tumor incidence data are
913.9 and 544.4 mg dichloromethane metabolized via GST pathways per liter liver per day,
respectively, for the internal liver metabolism metric, and 38.7 and 23.1 mg dichloromethane
metabolized via GST pathway in lung and liver/kg-day, respectively, for the whole body
metabolism metric (Table E-4). For lung tumors, the BMDio and BMDLio are 61.7 and 48.7 mg
dichloromethane metabolized via GST pathway per liter tissue per day, respectively, for the
lung-specific metric, and 13.1 and 10.3 mg dichloromethane metabolized via GST pathway in
lung and liver/kg-day, respectively, for the whole body metabolism metric.
E-9
DRAFT - DO NOT CITE OR QUOTE
-------
Table E-4. BMD modeling results and tumor risk factors associated with 10% extra risk for liver and lung tumors
in male B6C3Fi mice exposed by inhalation to dichloromethane for 2 years, based on liver-specific GST
metabolism and whole body GST metabolism dose metrics
Internal dose
metric"
Tissue-specific
Whole body
Male, liver
Male, lung
Male, liver
Male, lung
BMDS
modelb
MS(1)
MS(1)
MS(1)
MS(1)
x2
goodness of fit
/7-value
0.40
0.64
0.40
0.66
Mouse BMD10C
913.9
61.7
38.7
13.1
Mouse BMDL10C
544.4
48.6
23.1
10.3
Allometric-
scaled human
BMDL10d
77.8
7.0
3.3
1.5
Tumor risk factor"
Scaling = 1.0
1.84 x 1Q-4
2.06 x 1Q-3
-
-
Allometric-scaled
1.29 x 1Q-3
1.44 x 1Q-2
3.03 x 1Q-2
6.80 x 1Q-2
"Tissue specific dose units = mg dichloromethane metabolized via GST pathway per liter (liver or lung) tissue per d; whole-body dose units = mg dichloromethane
metabolized via GST pathway in lung and liver/kg-d).
bThe multistage (MS) model in EPA BMDS version 2.0 was fit to the mouse dose-response data shown in Table 5-17 using internal dose metrics calculated with the
mouse PBPK model. Numbers in parentheses indicate: (1) the degree polynomial of the model.
°BMD10 and BMDL10 refer to the BMD-model-predicted mouse internal dose and its 95% lower confidence limit, associated with a 10% extra risk for the incidence
of tumors.
dMouse BMDL10 divided by (BWhuman/BWmouse)025= 7.
eDichloromethane tumor risk factor (extra risk per unit internal dose) derived by dividing the BMR (0.1) by the mouse BMDL10 and by the allometric-scaled human
BMDL10, for the scaling =1.0 and allometric-scaled risk factors, respectively.
E-10
DRAFT - DO NOT CITE OR QUOTE
-------
Modeling results are presented in the subsequent sections for the tissue-specific liver-
metabolism metric for liver tumors (Section E.2.1), tissue-specific lung metabolism metric for
lung tumors (section E-2.2), and the whole-body metabolism metric for liver tumors
(Section E.2.3) and lung tumors (Section E.2.4).
E.2.1. Modeling Results for the Internal Liver Metabolism Metric, Liver Tumors.
Mennear et al. (1988); NTP (1986): Internal Liver Dose-Response for Liver Tumors in
Male Mice
1-degree polynomial
T3
£
O
c
O
•*=
O
(0
0.8
0.7
0.6
0.5
0.4
0.3
Multistage Cancer Model with 0.95 Confidence Level
BMDL
Multistage Cancer
Linear extrapolation
BMD
1000
2000 3000
dose
4000
5000
16:0802/192009
Figure E-3. Predicted and observed incidence of animals with hepatocellular
carcinoma or adenoma in male B6C3Fi mice exposed by inhalation to
dichloromethane for 2 years, using liver-specific metabolism dose metric
(Mennear et al., 1988; NTP, 1986).
Multistage Cancer Model. (Version: 1.7; Date: 05/16/2008)
Input Data File: C:\USEPA\IRIS\DCM\NTP\lung\male\lMulNTPMS_.(d)
Gnuplot Plotting File: C:\USEPA\IRIS\DCM\NTP\lung\male\lMulNTPMS_.plt
Thu Feb 19 16:08:19 2009
E-ll
DRAFT - DO NOT CITE OR QUOTE
-------
BMDS Model Run
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl) ]
The parameter betas are restricted to be positive
Dependent variable = incidence
Independent variable = dose
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 2
Total number of specified parameters = 0
Degree of polynomial = 1
Maximum number of iterations = 250
Relative Function Convergence has been set to: 2.22045e-016
Parameter Convergence has been set to: 1.49012e-008
**** We are sorry but Relative Function and Parameter Convergence ****
**** are currently unavailable in this model. Please keep checking ****
**** the web sight for model updates which will eventually ****
**** incorporate these convergence criterion. Default values used. ****
Default Initial Parameter Values
Background = 0.406706
Beta(l) = 0.00012805
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.69
Beta(l) -0.69 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0.421771 * * *
Beta(l) 0.000115283 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model Log(likelihood) # Param's Deviance Test d.f. P-value
Full model -95.4892 3
Fitted model -95.8368 2 0.695297 1 0.4044
Reduced model -99.1316 1 7.28482 2 0.02619
AIC: 195.674
Dose
0.0000
2363.7000
4972.2000
Est. Prob.
0.4218
0.5597
0.6740
Goodness of Fit
Expected Observed Size
21.089
26.305
31.680
22.000
24.000
33.000
50
47
47
Scaled
Residual
0.261
-0. 677
0.411
ChiA2 =0.70 d.f. = 1 P-value = 0.4042
E-12 DRAFT - DO NOT CITE OR QUOTE
-------
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 913.932
BMDL = 544.35
BMDU = 2569.01
Taken together, (544.35 , 2569.01) is a 90 % two-sided confidence
interval for the BMD
Multistage Cancer Slope Factor = 0.000183705
E-13 DRAFT - DO NOT CITE OR QUOTE
-------
E.2.2. Modeling Results for the Internal Lung Metabolism Metric, Lung Tumors.
Mennear et al. (1988); NTP (1986): Internal Lung Dose-Response for Lung Tumors in
Male Mice
1-degree polynomial
T3
£
O
c
O
•*=
O
(0
0.8
0.6
0.4
0.2
Multistage Cancer Model with 0.95 Confidence Level
Multistage Cancer
Linear extrapolation
BMDL BMD
200
400 600
dose
800
1000
21:2002/192009
Figure E-4. Predicted and observed incidence of animals with carcinoma or
adenoma in the lung of male B6C3Fi mice exposed by inhalation to
dichloromethane for 2 years, using liver-specific metabolism dose metric
(Mennear et al., 1988; NTP, 1986).
Multistage Cancer Model. (Version: 1.7; Date: 05/16/2008)
Input Data File: C:\USEPA\IRIS\DCM\NTP\lung\male\lMulNTPMS_.(d)
Gnuplot Plotting File: C:\USEPA\IRIS\DCM\NTP\lung\male\lMulNTPMS_.plt
Thu Feb 19 21:20:36 2009
BMDS Model Run
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl) ]
The parameter betas are restricted to be positive
E-14
DRAFT - DO NOT CITE OR QUOTE
-------
Dependent variable = incidence
Independent variable = dose
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 2
Total number of specified parameters = 0
Degree of polynomial = 1
Maximum number of iterations = 250
Relative Function Convergence has been set to: 2.22045e-016
Parameter Convergence has been set to: 1.49012e-008
****
****
We are sorry but Relative Function and Parameter Convergence ****
are currently unavailable in this model. Please keep checking ****
**** the web sight for model updates which will eventually ****
**** incorporate these convergence criterion. Default values used. ****
Default Initial Parameter Values
Background = 0.0642604
Beta(l) = 0.00181622
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.56
Beta(l) -0.56 1
Variable
Background
Beta(l)
Parameter Estimates
95.0% Wald Confidence Interval
Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
0.0980033 * * *
0.00170868 * * *
Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
AIC:
Log(likelihood)
-68.0892
-68.199
-99.8132
140.398
# Param' s
3
2
1
Deviance
0.219579
63.4479
Test d
1
2
P-value
0.6394
<.0001
Dose
0.0000
475.0000
992.2000
Est. Prob.
0.0980
0.5994
0.8345
Goodness of Fit
Expected Observed Size
4.900
28.171
39.219
5.000
27.000
40.000
50
47
47
Scaled
Residual
0.047
-0.349
0.306
ChiA2 =0.22 d.f. = 1
Benchmark Dose Computation
P-value = 0.6408
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0.1
Extra risk
0.95
61.6618
48.628
E-15
DRAFT - DO NOT CITE OR QUOTE
-------
BMDU =
80.2137
Taken together, (48.628 , 80.2137) is a 90
interval for the BMD
two-sided confidence
Multistage Cancer Slope Factor =
0.00205643
E.2.3. Modeling Results for the Whole Body Metabolism Metric, Liver Tumors. Mennear
et al. (1988); NTP (1986): Internal Whole-Body Metabolism Dose-Response for Liver
Tumors in Male Mice
1-degree polynomial
T3
£
o
c
o
'
o
(0
0.8
0.7
0.6
0.5
0.4
0.3
BMDL
Multistage Cancer Model with 0.95 Confidence Level
Multistage Cancer
Linear extrapolation
BMD
50
100
dose
150
200
17:5902/21 2009
Figure E-5. Predicted and observed incidence of animals with hepatocellular
carcinoma or adenoma in male B6C3Fi mice exposed by inhalation to
dichloromethane for 2 years, using whole-body metabolism dose metric
(Mennear et al., 1988; NTP, 1986).
Multistage Cancer Model. (Version: 1.7; Date: 05/16/2008)
Input Data File: C:\USEPA\IRIS\DCM\NTP\liver\male\lMulNTPMS_.(d)
Gnuplot Plotting File: C:\USEPA\IRIS\DCM\NTP\liver\male\lMulNTPMS_.plt
Sat Feb 21 17:59:59 2009
BMDS Model Run
E-16
DRAFT - DO NOT CITE OR QUOTE
-------
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl) ]
The parameter betas are restricted to be positive
Dependent variable = incidence
Independent variable = dose
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 2
Total number of specified parameters = 0
Degree of polynomial = 1
Maximum number of iterations = 250
Relative Function Convergence has been set to: 2.22045e-016
Parameter Convergence has been set to: 1.49012e-008
**** We are sorry but Relative Function and Parameter Convergence ****
**** are currently unavailable in this model. Please keep checking ****
**** the web sight for model updates which will eventually ****
**** incorporate these convergence criterion. Default values used. ****
Default Initial Parameter Values
Background = 0.406695
Beta(l) = 0.00302163
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.69
Beta(l) -0.69 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0.421768 * * *
Beta(l) 0.00272018 * * *
Indicates that this value is not calculated.
Analysis of Deviance Table
Model Log(likelihood) # Param's Deviance Test d.f. P-value
Full model -95.4892 3
Fitted model -95.8372 2 0.696122 1 0.4041
Reduced model -99.1316 1 7.28482 2 0.02619
AIC: 195.674
Dose
0.0000
100.2000
210.7000
Est. Prob.
0.4218
0.5597
0. 6740
Goodness of Fit
Expected Observed Size
21.088
26.307
31.679
22.000
24.000
33.000
50
47
47
Scaled
Residual
0.261
-0. 678
0.411
ChiA2 =0.70 d.f. = 1 P-value = 0.4039
Benchmark Dose Computation
Specified effect = 0.1
E-17 DRAFT - DO NOT CITE OR QUOTE
-------
Risk Type = Extra risk
Confidence level = 0.95
BMD = 38.733
BMDL = 23.0698
BMDU = 108 .885
Taken together, (23.0698, 108.885) is a 90 % two-sided confidence
interval for the BMD
Multistage Cancer Slope Factor = 0.00433467
E-18 DRAFT - DO NOT CITE OR QUOTE
-------
E.2.4. Modeling Results for the Whole Body Metabolism Metric, Lung Tumors. Mennear
et al. (1988); NTP (1986): Internal Whole-Body Metabolism Dose-Response for Lung
Tumors in Male Mice
1-degree polynomial
Multistage Cancer Model with 0.95 Confidence Level
T3
£
o
o
'
O
(0
0.8
0.6
0.4
0.2
Multistage Cancer
Linear extrapolation
BMDL BMD
50
100
dose
150
200
21:1702/21 2009
Figure E-6. Predicted and observed incidence of animals with carcinoma or
adenoma in the lung of male B6C3Fi mice exposed by inhalation to
dichloromethane for 2 years, using whole-body metabolism dose metric
(Mennear et al., 1988; NTP, 1986).
Multistage Cancer Model. (Version: 1.7; Date: 05/16/2008)
Input Data File: C:\USEPA\IRIS\DCM\NTP\lung\male\lMulNTPMS_.(d)
Gnuplot Plotting File: C:\USEPA\IRIS\DCM\NTP\lung\male\lMulNTPMS_.plt
Sat Feb 21 21:17:36 2009
BMDS Model Run
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl) ]
The parameter betas are restricted to be positive
E-19
DRAFT - DO NOT CITE OR QUOTE
-------
Dependent variable = incidence
Independent variable = dose
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 2
Total number of specified parameters = 0
Degree of polynomial = 1
Maximum number of iterations = 250
Relative Function Convergence has been set to: 2.22045e-016
Parameter Convergence has been set to: 1.49012e-008
**** We are sorry but Relative Function and Parameter Convergence ****
**** are currently unavailable in this model. Please keep checking ****
****
****
the web sight for model updates which will eventually
incorporate these convergence criterion. Default values used.
Default Initial Parameter Values
Background = 0.0659119
Beta(l) = 0.00855407
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.56
Beta(l) -0.56 1
****
****
Parameter Estimates
Limit
Variable
Background
Beta(l)
Estimate
0.0980803
0.00807004
Std. Err.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf.
Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
AIC:
Log(likelihood)
-68.0892
-68.1887
-99.8132
140.377
# Param' s
3
2
1
Deviance
0.198975
63.4479
Test d
1
2
P-value
0.6555
<.0001
Dose
0.0000
100.2000
210.7000
Est. Prob.
0.0981
0.5982
0.8353
Goodness of Fit
Expected Observed
4.904
28.116
39.259
5.000
27.000
40.000
Size
50
47
47
Scaled
Residual
0.046
-0.332
0.291
ChiA2 =0.20
d.f. = 1
P-value = 0.6569
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0.1
Extra risk
0.95
13.0558
10.2947
E-20
DRAFT - DO NOT CITE OR QUOTE
-------
BMDU = 16.9865
Taken together, (10.2947, 16.9865) is a 90 % two-sided confidence
interval for the BMD
Multistage Cancer Slope Factor = 0.00971371
E-21 DRAFT - DO NOT CITE OR QUOTE
-------
APPENDIX F. COMPARATIVE CANCER IUR BASED ON FEMALE MICE DATA
Using the male B6C3Fi mouse data from a 2-year inhalation exposure study (Mennear et
al., 1988; NTP, 1986), the recommended cancer lURs are 7 x 10'9 (ug/m3)'1 and 5 x
10"9 (ug/m3)'1 for the development of liver and lung cancer, respectively, based on the mean for
the GST-T1+/+ population. These values were derived using a tissue-specific GST metabolism
dose metric with allometric scaling. The combined human equivalent IUR values for both tumor
O Q 1
types is 1 x 10" (ug/m )" . As described in detail below, the resulting combined human
equivalent IUR values for both tumors did not differ appreciably by gender.
BMDio and BMDLio refer to the model-predicted dose (and its lower 95% confidence
limit) associated with 10% extra risk for the combined incidence of adenoma and carcinoma of
the liver or lung of female B6C3Fi mice inhaling dichloromethane for 2 years (Mennear et al.,
1988; NTP, 1986) (Table F-l).
Table F-l. Incidence data for liver and lung tumors and internal doses
based on GST metabolism dose metrics in female B6C3Fi mice exposed to
dichloromethane via inhalation for 2 years
Sex,
tumor type
Female, liver0
Female, lunge
BW(g)
-
30.0
29.0
-
30.0
29.0
External
dichloromethane
concentration
(ppm)
0
2,000
4,000
0
2,000
4,000
Mouse
tumor incidence
3/47 (6%)d
16/46 (35%)
40/46 (87%)
3/45 (6%)d
30/46 (65%)
41/46 (89%)
Mouse internal
tissue dose"
0
2,453.2
5,120.0
0
493.0
1,021.8
Mouse whole body
metabolism doseb
0
104.0
217.0
0
104.0
217.0
"For liver tumors: mg dichloromethane metabolized via GST pathway/L liver tissue/d from 6 hrs/d, 5 d/wk
exposure; for lung tumors: mg dichloromethane metabolized via GST pathway/L lung tissue/d from 6 hrs/day,
5 d/wk exposure.
bBased on the sum of dichloromethane metabolized via the GST pathway in the lung plus the liver, normalized to
total BW (i.e., [lung GST metabolism (mg/d) + liver GST metabolism (mg/d)]/kg BW). Units = mg
dichloromethane metabolized via GST pathway in lung and liver/kg-d.
°Hepatocellular carcinoma or adenoma. Mice dying prior to 52 wks were excluded from the denominators.
dStatistically significant increasing trend (by incidental and life-table tests; p < 0.01).
eBronchoalveolar carcinoma or adenoma. Mice dying prior to 52 wks were excluded from the denominators.
Sources: Mennear et al. (1988); NTP (1986).
Multistage models were fit to the female mouse internal tissue doses of dichloromethane
metabolized by the GST pathway and incidences for animals with liver tumors observed at the
F-l
DRAFT - DO NOT CITE OR QUOTE
-------
time of death. The predicted BMDio and BMDLio for the liver and lung tumor incidence data
are shown in Table F-2.
F-2 DRAFT - DO NOT CITE OR QUOTE
-------
Table F-2. BMD modeling results and tumor risk factors associated with 10% extra risk for liver and lung
tumors in female B6C3Fi mice exposed by inhalation to dichloromethane for 2 years, based on liver-specific GST
metabolism and whole body GST metabolism dose metrics
Internal dose
metric"
Liver-specific
Whole body
Female, liver
Female, lung
Female, liver
Female, lung
BMDS
modelb
MS (2)
MS(1)
MS (2)
MS(1)
X2
goodness of fit
/7-value
0.53
0.87
0.53
0.88
Mouse BMD10C
1,224.1
51.2
51.9
10.8
Mouse BMDL10C
659.7
40.7
28.0
8.6
Allometric-
scaled human
BMDL10d
94.2
5.8
4.0
1.2
Tumor risk factor*5
Scaling = 1.0
1.52 x 10'4
2.46 x 10'3
-
-
Allometric-scaled
1.06 x 10'3
1.72 x 10'2
2.50 x 10'2
8.14 x 10'2
aLiver specific dose units = mg dichloromethane metabolized via GST pathway per liter tissue per d; whole-body dose units = mg dichloromethane metabolized via
GST pathway in lung and liver/kg-d.
bThe multistage (MS) model in EPA BMDS version 2.0 was fit to the mouse dose-response data shown in Table 5-17 using internal dose metrics calculated with
the mouse PBPK model. Numbers in parentheses indicate: (1) the degree polynomial of the model.
°BMD10 and BMDL10 refer to the BMD-model-predicted mouse internal dose and its 95% lower confidence limit, associated with a 10% extra risk for the
incidence of tumors.
dMouse BMDL10 divided by (BWhuman/BWmouse)° 25= 7.
"Dichloromethane tumor risk factor (extra risk per unit internal dose) derived by dividing the BMR (0.1) by the mouse BMDL10 and by the allometric-scaled
human BMDL10, for the scaling =1.0 and allometric-scaled risk factors, respectively.
DRAFT - DO NOT CITE OR QUOTE
-------
A probabilistic PBPK model for dichloromethane in humans adapted from David et al.
(2006) (see Appendix B) was used with Monte Carlo sampling to calculate distributions of
internal lung, liver, or blood doses associated with chronic unit inhalation (1 ug/m3) exposures.
The model was then executed by using the external unit exposure as input, and the resulting
human equivalent internal dose was recorded. This process was repeated for 10,000 iterations to
generate a distribution of human internal doses. The resulting distribution of lURs shown in
Table F-3 was derived by multiplying the human internal dose tumor risk factor (in units of
reciprocal internal dose) by the respective distributions of human average daily internal dose
resulting from a chronic unit inhalation exposure of 1 ug/m3 dichloromethane. Risk estimates
were slightly higher for liver tumors and essentially equivalent for lung tumors in males
compared to females.
F-4 DRAFT - DO NOT CITE OR QUOTE
-------
Table F-3. lURs for dichloromethane based on PBPK model-derived internal liver and lung doses in B6C3Fi
female mice exposed via inhalation for 2 years, based on liver-specific GST metabolism and whole body
metabolism dose metrics, by population genotype
Internal dose
metric and scaling
factor"
Tissue-specific,
allometric -scaled
Tissue-specific,
scaling = 1.0
Whole-body,
allometric -scaled
Population
genotype1"
GST-T1+/+
GST-T1+/+
Mixed
Mixed
GST-T1+/+
GST-T1+/+
Mixed
Mixed
GST-T1+/+
GST-T1+/+
Mixed
Mixed
Tumor
type
Liver
Lung
Liver
Lung
Liver
Lung
Liver
Lung
Liver
Lung
Liver
Lung
Human tumor
risk factor0
1.06 x 1Q-3
1.72 x 10'2
1.06 x 10'3
1.72 x 10'2
1.52 x 10'4
2.46 x 10'3
1.52 x 10'4
2.46 x 10'3
2.50 x 10'2
8.14 x 10'2
2.50 x 10'2
8.14 x 10'2
Distribution of human internal
dichloromethane doses from 1 jig/m3
exposure"1
Mean
6.61 x 1Q-6
3.89 x 10'7
3.71 x 10'6
2.20 x 10'7
6.61 x 10'6
3.89 x 10'7
3.71 x 10'6
2.20 x 10'7
1.80 x 10'7
1.80 x 10'7
1.01 x 10'7
1.01 x 10'7
95th
percentile
2.21 x 1Q-5
1.24 x 10'6
1.43 x 10'5
8.06 x 10'7
2.21 x 10'5
1.24 x 10'6
1.43 x 10'5
8.06 x 10'7
6.38 x 10'7
6.38 x 10'7
4.00 x 10'7
4.00 x 10'7
99th
percentile
4.47 x 1Q-5
2.42 x 10'6
3.03 x 10'5
1.69 x 10'6
4.47 x 10'5
2.42 x 10'6
3.03 x 10'5
1.69 x 10'6
1.41 x 10'6
1.41 x 10'6
9.43 x 10'7
9.43 x 10'7
Resulting candidate human
lUR'Cug/m3)1
Mean
7.0 x 1Q-9
6.7 x 10'9
3.9 x 10'9
3.8 x KT9
1.0 x 10'9
9.6 x 10'10
5.6 x 10'10
5.4 x 10'10
4.5 x 10'9
1.5 x KT8
2.5 x 10'9
8.2 x 10'9
95th
percentile
2.4 x 1Q-8
2.1 x 10'8
1.5 x KT8
1.4 x 10'8
3.4 x 10'9
3.1 x KT9
2.2 x 10'9
2.0 x 10'9
1.6 x 10'8
5.2 x 10'8
1.0 x 10'8
3.3 x 10'8
99th
percentile
4.7 x 1Q-8
4.2 x 10'8
3.2 x 10'8
2.9 x 10'8
6.8 x 10'9
6.0 x 10'9
4.6 x 10'9
4.2 x 10'9
3.5 x 10'8
1.2 x 10'7
2.4 x 10'8
7.7x 10"8
aTissue specific dose units = mg dichloromethane metabolized via GST pathway per liter tissue (liver or lung, respectively, for liver and lung tumors) per d; whole-
body dose units = mg dichloromethane metabolized via GST pathway in lung and liver/kg-d.
bGST-Tl+/+ = homozygous, full enzyme activity; mixed = population reflecting estimated frequency of genotypes in current U.S. population: 20% GST-T"'", 48%
GST-T1+A, and 32% GST-T1+/+ (Haber et al., 2002).
°Dichloromethane tumor risk factor (extra risk per unit internal dose) derived by dividing the BMR (0.1) by the allometric-scaled human BMDL10 or by the mouse
BMDL10 (from Table 5-18) for the allometric-scaled and scaling =1.0 risk factors, respectively.
dMean, 95th, and 99th percentile of the human PBPK model-derived probability distribution of daily average internal dichloromethane dose resulting from chronic
exposure to 1 ug/m3 (0.00029 ppm).
Derived by multiplying the dichloromethane tumor risk factor by the PBPK model-derived probabilistic internal doses from daily exposure to 1 ug/m3.
F-5
DRAFT - DO NOT CITE OR QUOTE
-------
For the female mouse, the combined human equivalent IUR values for both tumor types
10~8 (ug/m3)'1 in the most sensitive (GST-T14
value that was obtained using the male mouse data.
is 1 x 10~8 (ug/m3)'1 in the most sensitive (GST-T1++) population (Table F-4), which is the same
F-6 DRAFT - DO NOT CITE OR QUOTE
-------
Table F-4. Upper bound estimates of combined human lURs for liver and lung tumors resulting from lifetime
exposure to 1 ug/m3 dichloromethane based on liver-specific GST metabolism and whole body metabolism dose
metrics, by population genotype, using female mouse data for derivation of risk factors
Internal dose
metric and scaling
factor3
Tissue-specific,
allometric-scaled
Tissue-specific,
scaling = 1.0
Whole-body,
allometric-scaled
Population
genotype1"
GST-T1+/+
Mixed
GST-T1+/+
Mixed
GST-T1+/+
Mixed
Tumor site
Liver
Lung
Liver or lung
Liver
Lung
Liver or lung
Liver
Lung
Liver or lung
Liver
Lung
Liver or lung
Liver
Lung
Liver or lung
Liver
Lung
Liver or lung
Upper bound
IURC
7.0 x 1Q-9
6.7 x ID'9
3.9 x ID'9
3.8 x ID'9
1.0 x 10'9
9.6 x lO'10
5.6 x lO'10
5.4 x lO'10
4.5 x ID'9
1.5 x ID'8
2.5 x ID'9
8.2 x ID'9
Central
tendency IURd
3.8 x ID'9
5.3 x ID'9
9.1 x ID'9
2.1 x ID'9
3.0 x ID'9
5.2x lO'9
5.4 x lO'10
7.6 x lO'10
1.3 x 1(T9
3.0 x 10'10
4.3 x 10'10
7.3 x 10'10
2.4 x ID'9
1.2 x ID'8
1.4 x ID'8
1.4 x ID'9
6.7 x ID'9
7.9 x ID'9
Variance of
tissue-specific
tumor risk6
3.87 x ID'18
7.12 x ID'19
1.22 x ID'18
2.28 x ID'19
7.89 x lO'20
1.42 x 10'21
2.48 x 10'20
4.54 x 10'21
1.59 x ID'18
4.11 x ID'18
5.00 x 1Q-19
1.29 x ID'18
Combined
tumor risk SDf
2.1 x ID'9
1.2 x ID'9
3.1 x lO'10
1.7 x 10'10
2.4 x ID'9
1.3 x ID'9
Upper bound on
combined tumor risk8
(jig/m3)-1
1.3 x ID'8
7.1 x ID'9
1.8 x 10'9
1.0 x 10'9
1.8 x ID'8
1.0 x ID'8
aTissue specific dose units = mg dichloromethane metabolized via GST pathway per liter tissue (liver or lung, respectively, for liver and lung tumors) per d; whole-
body dose units = mg dichloromethane metabolized via GST pathway in lung and liver/kg-d.
bGST-Tl+/+ = homozygous, full enzyme activity); mixed = population reflecting estimated frequency of genotypes in current U.S. population: 20% GST-T"'",
48% GST-T1+/; and 32% GST-T1+/+ (Haber et al., 2002).
"Estimated at the human equivalent BMDL10 (0.1/BMDL10) (see Table F-2).
Estimated at the human equivalent BMD10 (0.1/BMD) (see Table F-2).
Calculated as the square of the difference of the upper bound and central tendency lURs divided by the t statistic, 1.645.
Calculated as the square root of the sum of the variances for liver and lung tumors.
Calculated as the product of the cumulative tumor risk SD and the / statistic, 1.645, added to the sum of central tendency lURs.
F-7
DRAFT - DO NOT CITE OR QUOTE
-------
APPENDIX G. COMPARATIVE CANCER IUR BASED ON BENIGN MAMMARY
GLAND TUMORS IN RATS
Data for mammary gland tumors in male and female F344 rats following exposure to
airborne dichloromethane were used to develop a comparative IUR for dichloromethane
(Mennear et al., 1988; NTP, 1986) (Table G-l). Significantly increased incidences of mammary
gland or subcutaneous tissue adenoma, fibroadenomas, or fibromas were observed in male rats at
4,000 ppm, while mammary gland adenomas or fibroadenomas were increased in female rats
exposed 6 hours/day, 5 days/week for 2 years at concentrations > 1,000 ppm. Significant
decreases in survival were observed in the treated groups of both sexes. The at-risk study
populations (represented by the denominators in the incidence data) were determined by
excluding all animals dying prior to 52 weeks.
Table G-l. Incidence data for mammary gland tumors and internal doses
based on different dose metrics in male and female F344 rats exposed to
dichloromethane via inhalation for 2 years
Sex
Male
Female
BW(g)
-
390.5
385.2
384.8
-
245.5
244.3
242.2
External dichloromethane
concentration (ppm)
0
1,000
2,000
4,000
0
1,000
2,000
4,000
Rat tumor incidence3
1/50 (2%)c
1/50 (2%)
4/50 (8%)
9/50 (18%)
6/49 (12%)c
13/50 (26%)
14/50 (28%)
23/50 (46%)
Rat internal dose, AUC in
slowly perfused tissue1"
0
93.3
196.3
403.2
0
93.2
196.1
402.7
aMale tumors include mammary gland or subcutaneous tissue adenoma, fibroadenomas, or fibroma. Female
tumors include mammary gland adenoma or fibroadenomas. Rats dying prior to 52 wks were excluded from the
denominators.
bAverage daily AUC for dichloromethane in slowly perfused tissue (mg x hr/L) (see text for rationale for using this
dose metric).
Statistically significant increasing trend (p < 0.01).
Sources: Mennear et al. (1988); NTP (1986).
The rat PBPK model of Andersen et al. (1991) (modified as described in Appendix D)
was used to simulate inhalation exposures of 6 hours/day, 5 days/week and calculate long-term
daily average internal doses for the 2-year bioassays (Mennear et al., 1988; NTP, 1986). Study-,
group-, and sex-specific mean BWs for rats were used. The modified PBPK model did not
include a compartment for the mammary gland nor did it account for metabolism of
dichloromethane occurring in the mammary gland. The selected internal dose metric for
G-l
DRAFT - DO NOT CITE OR QUOTE
-------
mammary gland tumors was the average daily AUC for dichloromethane in slowly perfused
tissue; the mammary gland and the regions around it consist primarily of fatty tissue, which is
slowly perfused tissue. The role of CYP- or GST-mediated metabolism in the mammary gland is
uncertain. GST-T1 (Lehmann and Wagner, 2008) and CYP2E1 (El-Rayes et al., 2003; Hellmold
et al., 1998) expression has been detected in human mammary tissue, and it is also possible that
some metabolites enter systemic circulation from the liver and lung where they are formed.
Figure G-l shows the comparison between inhalation external and internal doses in the liver and
lung, respectively, using this dose metric for the rat and the human.
10,000
(U
8
T3
"iu
E
;-:-:.:^^:-::::v j :^^^i^^^i---±--i:-
H H—-H--
100 ^^^S
J. L-- .........
10 *Z=Z^ZZ*r_
..L—-L—J.--L-J-
Rat
D Human mixed GST
* Human GST +./-
* Human GST +/ +
•==H*S:::£:
I 1 1 L.-.J
-h—-I—-,<— M-++* H-
100 1,000
Inhalation concentration (ppm)
Average daily doses were calculated from simulated rat exposures of 6 hours/day,
5 days/week, while simulated human exposures were continuous. The GST
metabolism rate in the human population for the mixed GST-T1 group (+/+, +/-,
and -/-) in the current U.S. population was estimated as the mean of a simulated
sample of 3,000 individuals at each exposure concentration, based on GST-T1
polymorphism data from Haber et al. (2002). The results for the GST-T1 +/- and
+/+ subpopulations were then calculated as the means of the subsets of the mixed
population sample with the respective genotypes.
Figure G-l. PBPK model-derived internal doses (daily average AUC for
dichloromethane in slowly perfused tissue) in rats and humans and their
associated external exposures (ppm) used for the derivation of cancer lURs
based on mammary tumors in rats.
G-2
DRAFT - DO NOT CITE OR QUOTE
-------
The multistage model was fit to the rat mammary gland tumor incidence and PBPK
model-derived internal dose data to derive rat internal BMDio and BMDLio values associated
with 10% extra risk (Table G-2).
Table G-2. BMD modeling results associated with 10% extra risk for
mammary gland tumors in F344 rats exposed by inhalation to
dichloromethane for 2 years based on AUC for dichloromethane in slowly
perfused tissue
Sex
Male
Female
Tumor type
Mammary gland or
subcutaneous tissue adenoma,
fibroadenoma, or fibroma
Mammary gland adenoma or
fibroadenoma
BMDS
model"
MS(1)
MS(1)
x2
goodness of fit
/7-value
0.53
0.72
Rat
BMD10C
275.1
90.9
Rat
BMDIV
172.2
61.5
Tumor risk
factord
5.81 x 1Q-4
1.63 x 1Q-3
aThe multistage (MS) model in EPA BMDS version 2.0 was fit to each of the two sets of rat dose-response data
shown in Table G-l using internal dose metrics calculated with the rat PBPK model. Numbers in parentheses
indicate: (1) the degree polynomial of the model.
°BMD10 and BMDL10 refer to the BMD-model-predicted rat internal dose (average daily AUC for dichloromethane
in slowly perfused tissue [mg x hr/L]) and its 95% lower confidence limit associated with a 10% extra risk for the
incidence of tumors.
dDichloromethane tumor risk factor (extra risk per average daily AUC for dichloromethane in slowly perfused
tissue [mg x hr/L]) was derived by dividing the BMR (0.1) by the rat BMDL10. The rat BMDL10 is assumed to be
equivalent to human BMDL10; humans exposed to the same average daily AUC for dichloromethane in slowly
perfused tissue as rats will have the same risks for mammary tumors.
Rat mammary gland tumor risk factors (extra risk per unit internal dose) were calculated
by dividing 0.1 by the rat internal BMDLio. Because this risk factor is based on the internal
concentration of dichloromethane rather than a rate of reaction, it is assumed that the human risk
factor is equal to that of the rat (i.e., that humans exposed for a 70-year lifetime to the same
weekly average AUC of dichloromethane will have the same risk as rats exposed for 2 years).
The human PBPK model (David et al., 2006) was used to calculate a distribution of human
average daily AUCs for dichloromethane in slowly perfused tissue resulting from chronic
inhalation exposure to a unit concentration of 1 ug/m3 dichloromethane (0.00029 ppm). The
distribution of lURs for mammary gland tumors was generated by multiplying the human tumor
risk factor for each sex by the distribution of internal doses from chronic human exposure to
1 ug/m3 dichloromethane. Because this analysis is not based on the assumption that either
metabolic pathway is or is not influencing the cancer risk, this distribution was derived by using
weights reflecting the estimated frequency of GST-T1 genotypes in the current U.S. population
(20% GST-Tl"7", 48% GST-T1+/", and 32% GST-T1+/+). As shown in Table G-3, the mean
O •? Q 1
human IUR based on mammary gland tumors in rats is 4 x 10" and 1 x 10" (ug/m )" based on
DRAFT - DO NOT CITE OR QUOTE
-------
male and female rat-derived risk factors, respectively. Identical values were obtained using
slowly perfused tissue as the internal dose metric.
Table G-3. lURs for dichloromethane based on benign mammary tumors
and PBPK model-derived internal doses in F344N rats exposed via inhalation
for 2 years based on AUC for dichloromethane in slowly perfused tissue dose
metric
Sex,
tumor type
Male
Female
Human
tumor risk
factor"
5.81 x 1(T4
1.63 x 1(T3
Distribution of human internal
dichloromethane doses from 1 jig/m3
exposure1"
Mean
6.83 x 1(T5
6.83 x 1(T5
95th
percentile
1.08 x 1(T4
1.08 x 10'4
99th
percentile
1.33 x 10'4
1.33 x 10'4
Resulting candidate human
IURC
Mean
3.97 x 10'8
1.11 x 10'7
95th
percentile
6.25 x 10'8
1.75 x 10'7
99th
percentile
7.76 x 10'8
2.18 x 10'7
aDichloromethane tumor risk factor (extra risk per average daily AUC for dichloromethane in slowly perfused tissue
[mg x hr/L]) was derived by dividing the BMR (0.1) by the rat BMDL10 (from Table G-2). The rat BMDL10 is
assumed to be equivalent to human BMDL10; humans exposed to the same average daily AUC for dichloromethane
in slowly perfused tissue as rats will have the same risks for mammary tumors.
Vlean, 95th, and 99th percentile of the human PBPK model-derived probability distribution of daily average internal
dichloromethane dose (average daily AUC for dichloromethane in slowly perfused tissue [mg x hr/L]) resulting from
chronic inhalation exposure to a unit concentration of 1 ug/m3 (0.00029 ppm), based on a distribution of
GST-T1 genotypes that reflects the frequency distribution in the current U.S. population (Haber et al., 2002).
TVIean, 95th, and 99th percentile of a distribution of human lURs (extra risk per ug/m3) derived by multiplying the
dichloromethane tumor risk factor by the PBPK model-derived probabilistic distribution of human internal
dichloromethane doses from unit dichloromethane inhalation exposure.
G-4
DRAFT - DO NOT CITE OR QUOTE
-------
APPENDIX H: SOURCE CODE AND COMMAND FILES FOR
DICHLOROMETHANE PBPK MODELS
The following is a copy of the primary acslXtreme code (.csl file for implementation
under acslXtreme v.3.0.1.6) used for the dichloromethane simulations. Portions of the code
which had been commented out (i.e., unused code) were deleted for brevity. Note that most
parameters are set in the subsequent script/m-files, used to specify simulations and call the .csl
code.
While the code indicates modifications by G Diamond and M Lumpkin (current or and
former employees of the EPA contractor, SRC, Inc.), these are changes from the version of the
code as received by them to bring it into alignment with the version as described in the
publications of Marino et al. (2006) and David et al. (2006) (changes noted by comments).
Other changes allow for the calculation of various metrics to improve time-efficiency for
computational convergence; for example, calculating an average AUC over only the last week of
simulated exposure requires a much shorter overall simulated time for the calculation to give the
steady-state or very-long-term average, when the approach to steady-state or repeating periodic
solution occurs over the first days or weeks of exposure. Thus, the code exactly replicates the
published models when the published parameter values are used.
PROGRAM: DCM_2010_EPA.csl
! Code from Reitz et al. 1997, Addressing Data Needs for Methylene Chloride with
! Physiologically Based Pharmcokinetic Modeling (Appendix I)
! Translated by GDiamond (05/2004); DCM.CSL.Reitz.
! Revision date 18-Dec-96 by RHR (Peak moved to dynamic)
! Modified by GDiamond (08/2004) based on Andersen et al. 1987 (TAP 87:185-205)
land 1991 (TAP 108:14-27):
! Deleted brain compartment
! Added lung compartment, with lung metabolism
! Adjusted metabolism parameter values to match Andersen et al. 1987
! Adjusted physiological parameters to match Andersen et al. 1987
! Modifed by M Lumpkin (11/2005) (MHL) to include extrahepatic MFO metabolism and
! CO kinetics in blood
! Additional comments and changes by Paul Schlosser (PS), U.S. EPA (7/2008 - 2/2010)
! Removed unused (legacy) code bits, added comments, PS 9/2009
INITIAL
! Simulation, T=hour
NSTP= 1000 ! Initital integration cycle length at CINT/1000
! MERROR ALU=0.00001 ! Error tolerance for Gear
H-1 DRAFT - DO NOT CITE OR QUOTE
-------
CONSTANT POINTS=96.0 ! Number of points in plot
CINTERVAL CINT=0.1 ! Communication interval
CONSTANT TEND=25.0 ! Termination time (hr)
TSTOP=TEND
! Initial values for 250 g rat
! Body masses and fractional volumes
CONSTANT BW=0.25 ! Body weight (kg) MHL
CONSTANT VBL2C=0.059 ! Fractional volume of blood MHL
CONSTANT VFC=0.07 ! Fractional volume fat
CONSTANT VLC=0.04 ! Fractional volume of liver
CONSTANT VRC=0.05 ! Fractional volume of rapidly-perfused tissues
CONSTANT VSC=0.75 ! Fractional volume of slowly-perfused tissues **
CONSTANT VLUC=0.0115 ! Allometric scaling factor for lung volume * *
! Tissue masses (kg)
VTOT=VFC+VLC+(VLUC/(BW**0.01))+VRC+VSC
! Total volume fractions constrained to sum to 0.9215 (7.85% carcass) MHL
VBL2=VBL2C*BW ! Blood
VF=VFC*BW*0.9215/VTOT ! Fat
VL=VLC*BW*0.9215/VTOT ! Liver
VR=VRC*BW*0.9215/VTOT ! Rapidly-perfused tissue
VS=VSC*BW*0.9215/VTOT ! Slowly-perfused tissue
VLU=VLUC*(BW**0.99)*0.9215/VTOT ! Lung
! Flow constants and fractions
CONSTANT VPR=0.42 ! Ventilation/perfusion ratio (for QP calc) MHL
CONSTANT DSPC=0.15 ! Fractional lung dead space MHL
CONSTANT QCC= 15.0 ! Allometric constant for cardiac output
CONSTANT QCCR=10.0 ! For 'resting' period PMS 8/4/09
CONSTANT QCSW=0 ! Set to 1 to use QCCR (resting cardiac rate) up to TCHNG
CONSTANT QLC=0.20 ! Fractional flow to liver
CONSTANT QFC=0.09 ! Fractional flow to fat
CONSTANT QSC=0.15 ! Fractional flow to slowly-perfused tissue
CONSTANT QRC=0.56 ! Fractional flow to rapidly-perfused tissue * * MHL
! Flow rates (L/hr)
IF (QCSW) THEN
QC=QCCR*BW**0.74 ! Cardiac output if resting, PMS 8/4/09
ELSE
H-2 DRAFT - DO NOT CITE OR QUOTE
-------
QC=QCC*BW**0.74 ! Cardiac output if not resting, PMS 8/4/09
ENDIF
QP=QC*VPR ! Alveolar ventilation rate MHL
QPDP=QP/DL/PAIR ! MHL
QTOT=QFC+QLC+QRC+QSC! MHL
QL=QLC*QC/QTOT ! Liver MHL
QF=QFC*QC/QTOT ! Fat MHL
QR=QRC*QC/QTOT ! Rapidly-perfused tissue MHL
QS=QSC*QC/QTOT ! Slowly-perfused MHL
! Partition coefficients
CONSTANT PL=0.732
CONSTANT PF=6.19
CONSTANT PS=0.408
CONSTANT PR=0.732
CONSTANT PLU=0.732
CONSTANT PB= 19.4
! Liver:blood
! Fatblood
! Muscle:blood
! Liver:blood
! Lung:blood **
! Blood:air
! Metabolism
CONSTANT VMAXC=4.0
CONSTANT KM=0.4
CONSTANT FRACR=0.1
CONSTANT KFC=2.56
CONSTANT MOLWT=85.0
CONSTANT MWCO=28.0
CONSTANT A 1=0.416
CONSTANT A2=0.137
! Allometric scaling constant for VMAX
! Michaelis constant for MFO pathway (mg/L)
! Oxidative metab in rapidly perfused MHL
! Allometric scaling constant for KF
! Molecular weight of DCM
! Molecular weight of CO MHL
! Ratio of specific activities of MFO, lung/liver MHL
! Ratio of specific activities of GST, lung/liver MHL
VMAX=VMAXC*BW**0.7
VMAXR=VMAX*FRACR
KF=KFC/(BW**0.3)
CONSTANT AFFG=0.0
! Maximum rate of MFO pathway (mg/hr)
!MHL
! First-order rate constant for GSH pathway
! 1.57e-4 ! Affinity constant (I/Km) for GST pathway
! Used to test impact of low affinity / slight saturation PMS 12/09
! CO Submodel params MHL
CONSTANT SOLCO=0.03
CONSTANT DLC=0.060
CONSTANT RENCOC=0.035
CONSTANT ABCOC=0.117
CONSTANT HBTOT=10.0
CONSTANT Pl=0.80
L12=P 1 * MWCO/MOLWT
! mg/L/mmHG, MHL
MHL
Rate of endogenous CO production MHL
Cone of background CO (mg/kg) MHL
Cone of hemoglobin (mmoles/L) MHL
CO yield factor MHL
!PMS
DRAFT - DO NOT CITE OR QUOTE
-------
CONSTANT F1=1.21 ! CO elimination factor MHL
CONSTANT MMM= 197 ! Haldane coefficient, PMS 1/11
CONSTANT COINH=0.0 ! Controlled CO inhalation concentration (ppm) MHL
PICO=COINH/1402.5 ! Controlled CO concentration in inhaled air (mg/L) MHL
CONSTANT COBGD=2.2 ! Background CO cone, in air (ppm), PMS 1/11
14/16/32 mmoles/L MHL
! MHL
! Pressure of air (mm Hg) MHL
! Density of CO (mg/L) MHL
! mg/L/mmHG, MHL
! MHL
CONSTANT O2=0.13
BO2=O2/MMM
CONSTANT PAIR=713.0
CONSTANT RHO=1102.0
CONSTANT SOL=0.03
DL=DLC*(BW**0.92)
CONSTANT PCTHBCOO=1.0 ! Endogenous/background HBCO, PMS 2/10
CONSTANT RENCOS=1.0 ! Switch to selectively turn off bgd CO, PMS 2/10
if (RENCOC .EQ. 0.0) THEN
! Calculations of initial amounts of CACO/RENCO, PMS 2/10
PCOBG=COBGD*RENCOS/1402.5 ! Background CO cone, in air (mg/L), PMS 1/11
FO = RENCOS*PCTHBCOO/100.0
HBCOO=FO*HBTOT ! PMS, 2/10
CACOO = HBCOO + BO2*FO/(1.0-FO)
ABL20=CACOO*VBL2C*BW*MWCO ! PMS, 2/10
COFREEO=(O2/MMM)*FO/( 1.0-FO) ! Cone of free CO in blood (mmol/L) MHL
PCCOO=COFREEO*MWCO/SOLCO
RENCO=DL*RHO*F1*QPDP*(PCCOO - PCOBG)/(1.0+QPDP)
ELSE
PCOBG=COBGD*RENCOS/1402.5 ! Background CO cone, in air (mg/L), PMS 1/11
RENCO=RENCOC*(BW**0.7)! MHL
ABL20=ABCOC*BW
ADTCOO=0.0
END IF
PEAKPCTHBCO = RENCOS*PCTHBCOO !Capture peak CO concentration
! Exposure schedule
CONSTANT CONC=0.0
CONSTANT TCHNG=6.0
CONSTANT TDUR=24.0
CONSTANT TCHNG2= 120.0
CONSTANT TDUR2= 168.0
CIZONE=1.0
! Inhaled concentration (ppm)
! Exposure pulse 1 width (hr)
! Exposure duration (hr)
! Exposure pulse 2 width (hr)
! Exposure duration 2 (hr)
! Start with inhalation on
H-4
DRAFT - DO NOT CITE OR QUOTE
-------
PEAK = 0.0 ! Zero peak concentration in brain
CONSTANT IVDOSE=0.0 ! IV infusion (mg/kg) MHL
CONSTANT TIV=0.0028 ! IV dosing time (his, or 10 seconds) MHL
CONSTANT VCHC= 10.0 ! Uncorrected rat chamber volume (L) MHL
CONSTANT CC=0 ! Flag for closed chamber MHL
CONSTANT NCH=5 ! Number of animals in chamber MHL
CONSTANT KL=0.0 ! MHL
VCH=VCHC-(NCH* BW)
! Oral Gavage Dosing added by MHL
CONSTANT BOLUS=0.0
TOTALBOLUS=BOLUS*BW
TOTALDOSE=(IVDOSE+BOLUS)*BW
IVORBOL=(TOTALDOSE.GT.O.O)
CONSTANT DRCONC=0.0 ! User-specified concentration in drinking water (mg/L)
CONSTANT FIXDRDOSE=0.0! Set to constant DW dose in mg/kg-day
! Assume 70 kg human drinks 2 L/day, calculate rodent allometrically
constant RSTDY=0.0 ! Same except for zero-order infusion
ZLSTDY=RSTDY* BW/24.0
DRVOL=0.102*BW**0.7 ! Daily water intake, based on body weight
DRDOSE=DRVOL*DRCONC + FIXDRDOSE*BW ! Total dose from water (mg (in a day))
DDOSE=DRDOSE/BW ! Daily dose (mg/kg/day)
NEWDAY=0.0 ! To reset area-under-curve values each 24 hours
PEAKPCTHBCO=0.0
CONSTANT KA=5.0 ! Rate constant for absorption
CONSTANT KA2= 1.67
CONSTANT K 12=3 5
! Periodic drinking water intake schedule
! Assume T=0 is 7 AM (T+4=l 1 AM)
INTEGER I
1=2
DIMENSION DRTIME(32) ! Store drinking times in array
DIMENSION DRPCT(32) ! Store drinking percentages array
CONSTANT DRTIME=32*0.0, DRPCT=32*0.0
GASD=MOLWT/2445 0.0
IF (DRDOSE.GT.0.0) SCHEDULE drink.AT.DRTIME(2) ! Assume DRTIME(1)=0 ???
! and initial drink applied here.
CUMORALDOSE=DRPCT(1)*DRDOSE !To calculate input to GI (mg)
! Switching 'lastday' and 'lastwk' to be defined by discrete blocks, below, PS 3/2009
H-5 DRAFT - DO NOT CITE OR QUOTE
-------
lastday=0.0
IF (TSTOP.GT.24) SCHEDULE Way .AT. TSTOP-24
lastwk=0.0
if (TSTOP.GT.168) SCHEDULE Iwk .AT. TSTOP-168
CIDAY=1.0
CIWK=1.0
SCHEDULE idayoff .AT.TCHNG
SCHEDULE idayon .AT.TDUR
SCHEDULE iwkoff.AT.TCHNG2
SCHEDULE iwkon .AT.TDUR2
Izzstopflag = .FALSE.
END ! End of INITIAL section of program
DYNAMIC
DISCRETE idayoff
CIDAY=0.0
QC=QCC*BW**0.74
SCHEDULE idayoff .AT. (T+TDUR)
END
DISCRETE idayon
CIDAY=1.0
IF (QCSW) THEN
QC=QCCR*BW**0.74
ENDIF
SCHEDULE idayon .AT. (T+TDUR)
END
DISCRETE iwkoff
CIWK=0.0
SCHEDULE iwkoff .AT. (T+TDUR2)
END
DISCRETE iwkon
CIWK=1.0
SCHEDULE iwkon .AT. (T+TDUR2)
END
CIZONE = CIDAY*CIWK
DISCRETE Way ! Value -> 1 for last 24 h, PS 3/2009
lastday=1.0
end ! ofIday
DISCRETE Iwk ! Value -> 1 for last 168 h, PS 3/2009
H-6 DRAFT - DO NOT CITE OR QUOTE
-------
lastwk=1.0
end ! of Iwk
DISCRETE drink ! Loop for drinking water intake schedule
CUMORALDOSE = CUMORALDOSE + DRPCT(I)*DRDOSE
1=1+1
IF (I .EQ. 32) THEN
1=1
NEWDAY=NEWDAY + 24.0
ENDIF
SCHEDULE drink.AT.(NEWDAY+DRTIME(I))
End ! Drink
DERIVATIVE
ALGORITHM IALG=2! Gear for stiff systems
! Following are daily averages over final week of simulations, PS 8/2008
WAVGLIVCYPDOSE=integ(lastwk*RAMlL,0.0)/(7.0*VL)
WAVGLIVGSTDOSE=integ(lastwk*RAM2L,0.0)/(7.0*VL)
WAVGLUNGGSTDOSE=integ(lastwk*RAM2LU,0.0)/(7.0*VLU)
WAVGLUNGCYPDOSE=integ(lastwk*RAMlLU,0.0)/(7.0*VLU)
WAVGWBDYGSTDOSE=integ(lastwk*(RAM2L+RAM2LU),0.0)/(7.0*BW)
WAVGWBDYCYPDOSE=integ(lastwk*(RAMlL+RAMlLU),0.0)/(7.0*BW)
WAVGAUCV=integ(lastwk* CV,0.0)/7.0
WAVGAUCS=integ(lastwk*CS,0.0)/7.0
WAVGAUCL=integ(lastwk*CL,0.0)/7.0
! Following are daily averages calculated from final day's dose-rate, PS 8/2008
LDAYLIVCYPDOSE=integ(lastday* RAM 1 L,0.0)/VL
LDAYLIVGSTDOSE=integ(lastday*RAM2L,0.0)/VL
LDAYLUNGGSTDOSE=integ(lastday*RAM2LU,0.0)/VLU
LDAYWBDYGSTDOSE=integ(lastday*(RAM2L+RAM2LU),0.0)/BW
LDAYWBDYCYPDOSE=integ(lastday*(RAMlL+RAMlLU),0.0)/BW
LD AYLIV AUC=integ(lastday * CL,0.0)
LDAYAUCV=integ(lastday*CV,0.0)
LDAYAUCS=integ(lastday*CS,0.0)
LDAYPCTHB=integ(lastday*PCTHBCO,0.0)/24.0
AVRFROMDCM=1000.0*integ(lastday*RFROMDCM,0.0)/(BW*24.0)
! GI compartment for drinking water inputs
RSTOM=-(KA+K 12) * STOM
H-7 DRAFT - DO NOT CITE OR QUOTE
-------
R12=K12*STOM
RLGY=KA2*LGY
STOM=INTEG(RSTOM,TOTALBOLUS)+CUMORALDOSE
LGY=INTEG(R12-RLGY,0.0)
R= 0.21*pulse(0.0,1.0,0.015)*pulse(0.0,24.0,12.0)/(12*0.015) + &
0.79*pulse(0.0,0.6,0.015)*pulse(12.0,24.0,12.0)/(20*0.015)
RIV=IVDOSE*BW/TIV ! IV dose rate (mg/hr) MHL
blip=STEP(tiv)
IV=RIV*(1.0-blip)
! Chamber calcs MHL
RACH=CC*NCH*QP*((CA1/PB)-CCH)-(KL*ACH) ! MHL
ACH=INTEG(RACH,CONC*GASD*VCH) ! MHL added initial condition
CCH=CC*ACH/VCH ! MHL
CCHPPM=CCH/GASD! MHL original
! If/else statements removed by use of multipliers; PMS 7-23-08
CI=(1.0-CC)*CONC*CIZONE*GASD ! Convert to mg/L
DIDOSE=24.0*QP*INTEG(CI,0.0)/(BW*(T+1.0E-8)) ! Daily inhaled dose (mg/kg-d)
! CAl=Arterial blood concentration from gas exchange region
! to lung tissue compartment (mg DCM/L)
CA1=(QP* CI+QP* CCH+QC* CV)/((QP/PB)+QC) ! * *
! AX=Amount eliminated by exhalation (mg DCM)
CX=CA1/PB ! Concentration in air leaving gas exchange region
RAX=QP*CX
AX=INTEG(RAX, 0.0)
! Concentration DCM in exhaled air
CEX1=0.7*CX + 0.3*CI
! Assumes mixing with 30% of inhaled that only goes to deadspace
RAEX1=CEX1*QP ! MHL
AEX1=INTEG(RAEX1,0.0) ! MHL
PCTIVEXH=100.0*IVORBOL*AX/(TOTALDOSE + l.OE-8)
! % IV/BOLUS dose exhaled as DCM, MHL
! Amount in lung tissue (mg DCM)
RAM 1 LU=A 1 * VMAX* CA* (VLU/VL)/(KM+CA)
H-8 DRAFT - DO NOT CITE OR QUOTE
-------
AM1LU=INTEG(RAM1LU, 0.0)
RAM2LU=A2*KF*CA*VLU/(1.0+AFFG*CA)
RALU=QC* (CA1 -CA)-RAM 1LU-RAM2LU
ALU=INTEG(RALU, 0.0)
CLU=ALU/VLU
AUCLU=INTEG(CLU,0.0)
CA=CLU/PLU ! Amount in arterial blood to body
! AF=Amount in fat (mg DCM)
RAF=QF*(CA-CVF)
AF=INTEG(RAF, 0.0)
CF=AF/VF
CVF=CF/PF
!AL=Amount in liver (mg DCM)
RAL=QL*(CA-CVL)-RAM1L-RAM2L+KA*STOM+KA2*LGY+ZLSTDY
RAM1L=VMAX*CVL/(KM+CVL) !**
RAM2L=KF* CVL* VL/( 1.0+AFFG* CVL) ! * *
AL=INTEG(RAL, 0.0)
CL=AL/VL
CVL=CL/PL
AUCL=INTEG(CL, 0.0) !**
! AS= Amount in slowly-perfused tissues
RAS=QS*(CA-CVS) I-RAMS ! VMAXS=0, so not needed, PS 7/2008
AS=INTEG(RAS, 0.0)
CS=AS/VS
CVS=CS/PS
! AR=Amount in rapidly-perfused tissues
RAMR=VMAXR*CVR/(KM+CVR) ! MHL
AMR=INTEG(RAMR,0.0) ! Total metabolized (MFC) in RP, MHL
RAR=QR* (CA-CVR)-RAMR
AR=INTEG(RAR, 0.0)
CR=AR/VR
CVR=CR/PR
! AMl=Amount metabolized in MFO pathway (mg DCM)
RAMl=RAMlLU+RAMlL+Ramr !+ams ! AMS = 0, term not needed, PS 7/2008
TAM1 =INTEG(RAM 1,0.0)
! DDAM2=Amount metabolized in GSH pathway (mg DCM)
H-9 DRAFT - DO NOT CITE OR QUOTE
-------
RAM2=RAM2LU+RAM2L
TAM2=INTEG(RAM2,0.0)
! Total rate/amount metabolized (MFO plus GSH pathways)
RAMTOT=RAM1+RAM2 !**
AMTOT=INTEG(RAMTOT, 0.0)
! CV=Average venous blood concentration (mg DCM/L)
CV=(QF*CVF+QL*CVL+QS*CVS+QR*CVR+IV)/QC !MHL added RIV
CVP=0.23*CV ! Venous plasma cone
ABL=(CA+CV)*VBL2
AUCBL=INTEG(ABL, 0.0)
AUCV=INTEG(CV, 0.0) ! OLD
! CO appearence in and elimination from blood, MHL
ACO=ABL2/(VBL2*MWCO) !MHL
BO2=O2/Mmm ! MHL
CHBT=HBTOT ! MHL
HBCO=((ACO+BO2+CHBT)-SQRT((ACO+BO2+CHBT)*(ACO+BO2+CHBT)-4.0*ACO*CHBT))/2.0
! Cone of HB-bound CO (mmol/L) MHL
COFREE=ACO-HBCO ! Cone of free CO in blood (mmol/L) MHL
PICO=COINH/1402.5 ! CO concentration in inhaled air (mg/L) MHL
PCTHBCO=100.0*HBCO/HBTOT ! Percent carboxyhemoglobin in blood
PEAKPCTHBCO = MAX(LDAYPCTHB,PCTHBCO) ! Capture peak CO concentration
WAVGHBCO=integ(lastwk*PCTHBCO,0.0)/168.0
! COE = Amount of expired CO, MHL
PCCO=COFREE*MWCO/SOL
PACO=(PCCO+(CIZONE*PICO+PCOBG)*QPDP)/(1.0+QPDP)
! Background term, PCOBG, added to above, PMS 1/11
RCOE=DL*RHO*(PCCO-PACO)*F1
COE=INTEG(RCOE,0.0) ! Expired amount of CO (mg) MHL
COEC=0.7*COE/QP ! Cone expired CO (mg/L)
! Multiplier changed from 2/3 to 0.7 in above to be consistent with...
! other expired concentration calculations, PS 7/2008.
COECPPM=COEC*24450.0/MWCO
PCTIVCOEXH=100.0*IVORBOL*COE/(TOTALDOSE*MWCO/MOLWT+ l.OE-8) ! % dose exhaled
as CO MHL
! ABL2 = Amount of CO in blood (mg) MHL
RFROMDCM=(RAM1L+RAM1LU+RAMR)*L12 ! RAMS left out since not used, =0
! Inputs calculated for mass balance
H-10 DRAFT - DO NOT CITE OR QUOTE
-------
AENCO = INTEG(RENCO, 0.0) ! Amount produced endogenously (mg)
AFROMDCM = INTEG(RFROMDCM, 0.0) ! Amount produced from metabolism of DCM
RABL2=RFROMDCM+RENCO-RCOE
ABL2=INTEG(RABL2,ABL20) ! MHL
CABL2=ABL2/VBL2 ! MHL
! AI=Total mass input (mg DCM)
RAI=(QP*(CI+CCH))+RIV !MHL
AI=INTEG(RAI, 0.0)
! TMASS=Mass balance (mg DCM)
MASSBAL=MASSIN-STOM-AF-AL-ALU-AS-AR-ABL-MASSOUT
MASSIN=AI+CUMORALDOSE + TOTALBOLUS
MASSOUT=AX+AMTOT
! Mass Balance for CO
MASSBALCO= 100*(-ABL2+AENCO-COE+AFROMDCM)/(AENCO+AFROMDCM+le-10)
! IF (T .GE. TSTOP) zzstopflag = .TRUE. ! Termination condition for model run
END ! End of DERIVATIVE section of program
TERMT(T.GE.TEND)
END ! End of Dynamic section of program
TERMINAL
AVGT=24.0/(T+1.0e-8)
METDOS = AVGT*RAMTOT/VL
LDAYREFDOSE = LDAYLIVCYPDOSE + LDAYLIVGSTDOSE ! MHL
WAVGREFDOSE = WAVGLIVCYPDOSE + WAVGLIVGSTDOSE ! PS, 8/2008
LDAYWBDYCYPGSTDOSE = LDAYWBDYCYPDOSE + LDAYWBDYGSTDOSE
METABRATIO = RAM1/(RAM2+1.0E-12)
DDOSECHECK = CUMORALDOSE*avgt/BW! Should equal DDOSE
END ! End of TERMINAL section of program
END ! End of PROGRAM
Computational Files (.m files)
The following are files (functions) created as .m files to support the Monte Carlo
statistical sampling and subsequent calculations for the DCM analysis.
Utility files used for all human simulations
H-l 1 DRAFT - DO NOT CITE OR QUOTE
-------
function v=normbnd(mu, sigma, lo, up)
% [[[
% FILE: normbnd.m
%
% NORMBND - Returns a random sample from a truncated normal distribution with
% mean = mu, standard deviation = sigma, lower bound = lo, & upper bound = up.
% Modified by Paul Schlosser, U.S. EPA, 7/2008
if(lo>=up)
v = NaN; disp('Lower bound must be less than upper bound.')
return;
end
FL = normcdf(lo, mu, sigma); % normal probability of lower bound
FU = normcdf(up, mu, sigma); % normal probability of upper bound
p = rand*(FU - FL) + FL; % rand = uniform random(0,l)
v = norminv(p, mu, sigma);
end
function v=lnormbnd(mu, sigma, lo, up)
% [[[
% FILE: Inormbnd.m
%
% LNORMBND - Returns a truncated LOGnormal distribution with distribution with
% mean = mu, standard deviation = sigma, lower bound = lo, & upper bound = up.
% Modified by Paul Schlosser, U.S. EPA, 7/2008
v = exp(normbnd(log(mu), log(sigma), log(lo), log(up)));
end
function v=prctile(y,x)
% FILE: prctile.m
% Computes percentiles of vector y at percentile values x
% If x = [50 90 95], that's 50th, 90th, 95th
% Created by Paul Schlosser, U.S. EPA, July, 2008
% [[[
ifany((x<0) (x>100))
disp('Error! All percentiles must be between 0 and 100!');
return
end
dy = sort(y);
ps = 100*((l:length(dy))-0.5)/length(dy);
v = x;
for id= 1 : length(x)
ifx(id)
-------
% File clearT.m
% Creates list of variable names for tracking, rnames, and
% performs other initializations for human Monte Carlo chains.
% Created by Paul Schlosser, U.S. EPA, 7/2008
%
rnames=[''age'';''BW'';''DRVOL'';''DRCONC'';''DDOSE'';''CONC'';''VMAXC'';''KM'';''FRACRM;MKFCM;
"Al";"A2";"BW";"VLC";"VFC";"VRC";"VLUC";"VSC";"QAlvC";"QCC";"QLC";"QFC";"QSC";
"QRC";"PL";"PLU";"PF";"PS";"PR";"PB";"GSTGT";"CV";"LDAYLIVGSTDOSE";
"LDAYLIVCYPDOSE";"LDAYLUNGGSTDOSE";"LDAYWBDYGSTDOSE";"LDAYREFDOSE";"L
DAYAUCV";
"LDAYAUCS";"PEAKPCTHBCO";"LDAYPCTHB";"AVRFROMDCM";"WAVGAUCV";"WAVGAU
CS"];
sr='['; rns=length(rnames);
for ir=l:(rns-l)
sr=[sr,ctot(rnames(ir)),','];
end
sr=[sr,ctot(rnames(rns)),'];'];
% save current model values before perturbations
save @file='dcm_human.sav'
runo=[];
% prepare time history values.
prepare @clear T CV
WESITG=0; WEDITG=0; CONC=0.0; IVDOSE=0.0; DRCONC=0.0; FIXDRDOSE=0.0;
IVORBOL=0.0;
COINH=0; RENCOS=1;
DRPCT = [0.25, 0.1, 0.25, 0.1, 0.25, 0.05,zeros(l,26)];
DRTIME= [0.0, 3.0, 5.0, 8.0, 11.0, 15.0, 16+[1:26]/10];
TEND=600; TCHNG=2000; TCHNG2=2000; TDUR=2000; TDUR2=2000;
start @NoCallback
exist popn; % Test to see if 'popn' defined.
if~ans % If not...
popn="mix"; % Mix of GST types.
end
exist agem; % Check if agem defined ...
if~ans % If not...
agem=0;
end
exist gendm; % Check if gendm defined ...
if~ans % If not...
gendm="both"; % males and females
end
function nn=findnames(nm,rnames)
%
% File fmdnames.m
%
% Function to find indices of specific names in list nm
% within the larger list (of all variables), rnames.
% Created by Paul Schlosser, U.S. EPA, 7/2008
%
nn=[];
for n= 1: length(nm)
H-13 DRAFT - DO NOT CITE OR QUOTE
-------
n 1 =find(rname s==nm(n)) ;
if isempty(nl)
disp(['Error, ',ctot(nm(n)),' not in saved variable list.']); return
end
nn=[nn,nl];
end
end
% File human_parl.m
%
% Code for selecting human model parameters from MC distribution for
% dichlormethane PBPK model from probability density functions as
% described by David et al (2006), with *most of* the revisions as
% described in the U.S. EPA IRIS Toxicological Review for Dichloromethane,
% Appendix B.
% ** This code uses "one -dimensional" sampling distributions for CYP
% (VMAXC) and GSTT-1 (KFC). The sampling of KFC is as described by David
% et al (2006). The sampling for VMAXC uses the mean value of David et
% al (2006) but substitutes a larger geometric standard deviation and
% uses log-normal distribution without bounds. Further details below.
% Genotypic distribution of GSTT-1 activity taken from Tables 8 and 9
% of Environ report for Eastman Kodak.
% Arithmetic means and SDs for lognormal distributions converted
% by ML to geometric mean/SD to match acsl m-file Inormbnd.m values.
%
% Programmed by Michael Lumpkin (ML)
% Syracuse Research Corporation, 1 1/2005
% Modified by Paul Schlosser (PS), U.S. EPA 7/2008 and 1/2010
%
% Gender parameter 'gendm' can be "both" (default), "male", or "female".
% Age parameter 'agem' can be 0 (default) to simulate the full range
% from 0.5-80 years, or any value in that range for a specific age.
% Specific parameters, as noted below, set for those life-stages.
% choose uniform discrete distribution for GSTT-1 genotypes.
if popn=="++"
GSTGT = 0; % for +/+ only
elseif popn=="+-"
GSTGT =1;% for +/- only
elseif popn=="~"
GSTGT = 2; % for -/- only
else
r=rand(l)*gsmult; GSTGT=(r>0.32)+(r>0.8); %For general/mixed population
popn = "mix of +/+, +/-, and -/-";
end
% selection of GSTT1 activity distribution based on ethnic distribution
% upper bounds changed from mean + 3*SD to mean + 5*SD, P.S., 2/2009
if (GSTGT ==0)
KFC = (5.87/0.852)*normbnd(1.31, 0.167, 0.0, 2.145);
o/
/o
elseif (GSTGT ==1)
KFC = (5.87/0.852)*normbnd(0.676, 0.123, 0.0, 1.291); %
else
H-14 DRAFT - DO NOT CITE OR QUOTE
-------
KFC = 0.0;
end
AFFG=0.0; % 1.57e-4 % Affinity constant (I/Km) for GSTT-T1 pathway
% Allows for some saturation of the GSTT-1 pathway if set > 0, P.S., 1/2010
KA=5.0; %From Reitz et al (1997); this parameter has no variability data and is
% described as point estimate
VMAXC=lognorm(9.34, 1.73); %lnormbnd(9.34, 1.73, 3.8, 23.0);
%geometric mean and GSD, converted from arith mean/SD of 9.42 and 1.23
% VMAXC switched to unbounded distribution, PS, 2/2009
KM=lnormbnd(0.410, 1.39,0.154, 1.10);
%geometric mean and SD, converted from arith mean/SD of 0.433 and 1.46
FRACR=lnormbnd(0.0152, 2.0, 0.0019, 0.122);
%geometric mean and SD, converted from arith mean/SD of 0.0193 and 0.0152
Al=lnormbnd(0.00092, 1.47, 0.000291, 0.00292);
%geometric mean & SD, converted from arith mean/SD of 0.000993 & 0.000396
A2=lnormbnd(0.0083, 1.92, 0.00116, 0.0580);
%geometric mean and SD, converted from arith mean/SD of 0.0102 and 0.00739
age=agem;
if age==0 % agem=0 => age = random from population, otherwise leave as set
p=rand;
age = 165.86*pA4 - 253.19*pA3 + 113.27*pA2 + 53.356*p + 0.5;
end
VMP = 0.7 + 0.18*(age<18);
% power for scaling Vmax is 0.88 if age < 18, otherwise 0.7
% Change in VMP introduced 2/2009, PS
gend=gendm;
if ~((gendm=="male")|(gendm=="female")) % if not defined as one
gend="female";
if rand()<(0.513*((125.3 - age)A4)/(33.7A4 + (125.36 - age)A4));
gend="male";
end
end
ifmod(RUNN,50)==0
disp(['GSTGT group: ',ctot(popn),'. GSTGT = ',num2str(GSTGT),'.']);
disp(['Simulation for ',num2str(age),' year old ',ctot(gend),'; gender mix: ',ctot(gendm)]);
disp('');
end
% Human physiologic parameters
if gend=="female"
aged=(16-age)/10;
bwmean = 4.146 - 0.147*aged - 1.36*agedA2 + 0.44*agedA3;
if aged<0
bwmean = 4.146 - 0.147*aged - 0.0278*agedA2 - 0.00095*agedA3;
end
aged=(13-age)/10;
bwSD = 2.574 - 0.358*aged - 2.55*agedA2 + 1.16*agedA3;
if aged<0
bwSD = 2.574 - 0.358*aged - 0.0861*agedA2 - 0.00469*agedA3;
H-l 5 DRAFT - DO NOT CITE OR QUOTE
-------
end
else % males
aged=(21-age)/10;
bwmean = 4.406 - 0.0285*aged - 0.729*agedA2 + 0.115*agedA3;
if aged<0
bwmean = 4.406 - 0.0285*aged + 0.0048*agedA2 + 0.0018*agedA3;
end
aged=(16-age)/10;
bwSD = 2.87 + 0.06*aged - 2.56*agedA2 + 0.96*agedA3;
if aged<0
bwSD = 2.87 + 0.06*aged + 0.0448*agedA2 + 0.0067*agedA3;
end
end
BW=norminv((0.01+0.98*rand),exp(bwmean),exp(bwSD));
QAlvmean = 13.6 + 13.3*exp(-0.05*age);
if gend=="female"
QAlvmean = 10.7 + 22.1*exp(-0.08*age);
end
Qgsd = -0.1948*(age/10)A3 + 0.6095*(age/10)A2 - 0.3978*(age/10) + 1.4261;
ifage>16.81
Qgsd=1.554;
end
QAlvC = QAlvmean*exp(norminv((0.05+0.9*rand),0,log(Qgsd)));
QCCmean = 56.906*(1.0 - exp(-0.681*exp(0.0454*QAlvC))) - 29.747;
QCC = QCCmean/lnormbnd(1.0, 0.203, 0.69, 1.42);
VPR = QAlvC/QCC;
vlm=-0.0036*(age/10)A2 - 0.005l*(age/10) + 0.0395;
if age>17
vlm=-0.0004*(age/10)A2 + 0.0034*(age/10) + 0.0169;
end
VLC=vlm*normbnd(1.0, 0.05, 0.85, 1.15);
vfm=0.1612*(age/10)A3 + 0.0846*(age/10)A2 - 0.3083*(age/10) + 0.2709;
if((age>7)&(age<=20))
vfm=-0.0458*(age/10)A2 + 0.2082*(age/10) + 0.0274;
if gend=="male"
vfm= -0.0057*(age/10)A2 + 0.0293*(age/10) + 0.1303;
end
elseif age>20
vfm=-0.0024*(age/10)A3 + 0.0355*(age/10)A2 - 0.115*(age/10) + 0.3678;
if gend=="male"
vfm= -0.0015*(age/10)A2 + 0.0384*(age/10) + 0.0908;
end
end
VFC=vfm*normbnd(1.0, 0.3, 0.1, 1.9);
VRC=normbnd(0.064, 0.0064, 0.0448, 0.0832);
VLUC=normbnd(0.0115, 0.00161, 0.00667, 0.0163);
VSC=normbnd(0.63, 0.189, 0.431, 0.829);
vtr=0.9215/(VFC+VLC+VLUC+VRC+VSC);
VFC = VFC*vtr; VLC=VLC*vtr; VLUC=VLUC*vtr; VRC=VRC*vtr; VSC=VSC*vtr;
H-16 DRAFT - DO NOT CITE OR QUOTE
-------
VBL2C=0.059;
DSPC=0.15;
QLC=normbnd(0.26, 0.0910, 0.010, 0.533);
QFC=normbnd(0.05, 0.0150, 0.0050, 0.0950);
QSC=normbnd(0.19, 0.0285, 0.105 ,0.276);
QRC=normbnd(0.50 ,0.10, 0.20, 0.80);
%Human partition coefficients for DCM
PL=lnormbnd(1.43, 1.22, 0.790, 2.59);
%geometric mean and SD, converted from arith mean/SD of 1.46 and 0.292
PLU=lnormbnd(1.43, 1.22, 0.790, 2.59);
%geometric mean and SD, converted from arith mean/SD of 1.46 and 0.292
PF=lnormbnd(11.9, 1.34, 4.92, 28.7);
%geometric mean and SD, converted from arith mean/SD of 12.4 and 3.72
PS=lnormbnd(0.80, 1.22, 0.444, 1.46);
%geometric mean and SD, converted from arith mean/SD of 0.82 and 1.64
PR=lnormbnd(1.43, 1.22, 0.790, 2.59);
%geometric mean and SD, converted from arith mean/SD of 1.46 and 0.292
PB=lnormbnd(9.7, 1.10, 7.16, 13.0);
%geometric mean and SD, converted from arith mean/SD of 9.7 and 0.97
% File human_par2.m
%
% Code for selecting human model parameters from MC distribution for
% dichlormethane PBPK model from probability density functions as
% described by David et al (2006), with revisions as described in the
% U.S. EPA IRIS Toxicological Review for Dichloromethane, Appendix B.
%
% ** This code uses "two-dimensional" sampling distributions for CYP
% (VMAXC) and GSTT-1 (KFC), as described in the current assessment.
% Also see comments further details below.
%
% Genotypic distribution of GSTT-1 activity taken from Tables 8 and 9
% of Environ report for Eastman Kodak.
% Arithmetic means and SDs for lognormal distributions converted
% by ML to geometric mean/SD to match acsl m-file Inormbnd.m values.
%
% Programmed by Michael Lumpkin (ML)
% Syracuse Research Corporation, 1 1/2005
% Modified by Paul Schlosser (PS), U.S. EPA 7/2008 and 1/2010
%
% Gender parameter 'gendm' can be "both" (default), "male", or "female".
% Age parameter 'agem' can be 0 (default) to simulate the full range
% from 0.5-80 years, or any value in that range for a specific age.
% Specific parameters, as noted below, set for those life-stages.
%'par2'
% choose uniform discrete distribution for GSTT-1 genotypes.
if popn=="++"
GSTGT = 0; % for +/+ only
elseif popn=="+-"
GSTGT =1;% for +/- only
H-17 DRAFT - DO NOT CITE OR QUOTE
-------
elseifpopn=="~"
GSTGT = 2; % for -/- only
else
r=rand(l)*gsmult; GSTGT=(r>0.32)+(r>0.8); %For general/mixed population
popn = "mix of+/+, +/-, and -/-";
end
% selection of GSTT1 activity distribution based on ethnic distribution
% upper bounds changed from mean + 3*SD to mean + 5*SD, P.S., 2/2009
% From Table 4 of David et al., for kfc, mu = 0.852, CV = 0.711
% Lognorm tranform: m = mu/sqrt(CVA2 + 1) = 0.6944
% and s = exp(sqrt(ln(CVA2 + 1))) = 1.896
%kfcm=lnormbnd(m,s,m/(sA2),m*(sA2))/((0.48/2)+0.32);
% The last term (divisor) accounts for relative weighting and activity
% of the three genotypes.
kfcm=lnormbnd(0.6944,1.896,0.1932,2.496);
if (GSTGT ==0)
KFC = kfcm*normbnd(1.786, 0.2276, 0.0, 2.924); % Relative activity in +/+ pop'n
%=kfcm*normbnd(l, 0.167/1.31, 0.0, 1.0+5*0.167/1.31)/0.56;
% 0.167 and 1.31 are s.d and mean, respectively, for+/+ from Table 2 of David et al.
elseif (GSTGT == 1)
KFC = kfcm*normbnd(0.8929, 0.1622, 0.0, 1.704); % Relative activity in +/- pop'n
%=kfcm*normbnd(l, 0.123/0.676, 0.0, 1.0+5*0.123/0.676);
% 0.123 and 0.676 are s.d and mean, respectively, for +/- from Table 2 of David et al.
else
KFC = 0.0;
end
AFFG=0.0; % 1.57e-4 % Affinity constant (I/Km) for GSTT-T1 pathway
% Allows for some saturation of the GSTT-1 pathway if set > 0, P.S., 1/2010
KA=4.31; 5.0; %from Reitz et al (1997); this parameter has no variability data and is described as point
estimate
K12=0; KA2=0;
vmaxcm=lnormbnd(9.34,1.14,7.20,12.11);
%geometric mean and GSD, converted from arith mean/SD of 9.42 and 1.23
% lower/upper bounds = GM/(GSDA2) and GM*(GSDA2); i.e., +/- 2 SD in log-space
% bound calculations done with GM and GSD *not rounded* to 2 decimal places
VMAXC=lognorm(vmaxcm, 1.73); %lnormbnd(9.34, 1.73, 3.8,23.0);
% VMAXC switched to unbounded distribution, PS, 2/2009
KM=lnormbnd(0.410, 1.39, 0.154, 1.10); %geometric mean and SD, converted from arith
mean/SD of 0.433 and 1.46
FRACR=lnormbnd(0.0152, 2.0, 0.0019, 0.122); %geometric mean and SD, converted from arith
mean/SD of 0.0193 and 0.0152
Al=lnormbnd(0.00092, 1.47, 0.000291, 0.00292); %geometric mean & SD, converted
from arith mean/SD of 0.000993 & 0.000396
A2=lnormbnd(0.0083, 1.92, 0.00116, 0.0580); %geometric mean and SD, converted from arith
mean/SD of 0.0102 and 0.00739
%Ifun-commented, the block rescales the parameters to the means obtaine
% for the DiVincenzo and Kaplan (1981) data set; PS 4/2011
% VMAXC = VMAXC* 10.2/9.42; %
% KM = KM*2.06/0.433; %
% KFC = KFC*5.87/0.852; %
% A1=A1*0.00111/0.000993; %
% A2 = A2*0.0177/0.0102; %
% FRACR = FRACR*0.0379/0.0193; %
% -— END OF RESCALING BLOCK —
H-l 8 DRAFT - DO NOT CITE OR QUOTE
-------
age=agem;
if age==0 % agem=0 => age = random from population, otherwise leave
p=rand;
age = 165.86*pA4 - 253.19*pA3 + 113.27*pA2 + 53.356*p + 0.5;
end
VMP = 0.7 + 0.18*(age<18); % power for scaling Vmax is 0.88 if age < 18, otherwise 0.7
% Change in VMP introduced 2/2009, PS
gend=gendm;
if ~((gendm=="male")|(gendm=="female")) % if not defined as one
gend="female";
if rand()<(0.513*((125.3 - age)A4)/(33.7A4 + (125.36 - age)A4));
gend="male";
end
end
ifmod(RUNN,50)==0
disp(['GSTGT group: ',ctot(popn),'. GSTGT = ',num2str(GSTGT),'.']);
disp(['Simulation for ',num2str(age),' year old ',ctot(gend),'; gender mix: ',ctot(gendm)]);
disp('');
end
% Human physiologic parameters
if gend=="female"
aged=(16-age)/10;
bwmean = 4.146 - 0.147*aged - 1.36*agedA2 + 0.44*agedA3;
if aged<0
bwmean = 4.146 - 0.147*aged - 0.0278*agedA2 - 0.00095*agedA3;
end
aged=(13-age)/10;
bwSD = 2.574 - 0.358*aged - 2.55*agedA2 + 1.16*agedA3;
if aged<0
bwSD = 2.574 - 0.358*aged - 0.0861*agedA2 - 0.00469*agedA3;
end
else % males
aged=(21-age)/10;
bwmean = 4.406 - 0.0285*aged - 0.729*agedA2 + 0.115*agedA3;
if aged<0
bwmean = 4.406 - 0.0285*aged + 0.0048*agedA2 + 0.0018*agedA3;
end
aged=(16-age)/10;
bwSD = 2.87 + 0.06*aged - 2.56*agedA2 + 0.96*agedA3;
if aged<0
bwSD = 2.87 + 0.06*aged + 0.0448*agedA2 + 0.0067*agedA3;
end
end
BW=norminv((0.01+0.98*rand),exp(bwmean),exp(bwSD));
QAlvmean = 13.6 + 13.3*exp(-0.05*age);
if gend=="female"
QAlvmean = 10.7 + 22.1*exp(-0.08*age);
end
Qgsd = -0.1948*(age/10)A3 + 0.6095*(age/10)A2 - 0.3978*(age/10) + 1.4261;
ifage>16.81
Qgsd=1.554;
end
QAlvC = QAlvmean*exp(norminv((0.05+0.9*rand),0,log(Qgsd)));
QCCmean = 56.906*(1.0 - exp(-0.681*exp(0.0454*QAlvC))) - 29.747;
H-19 DRAFT - DO NOT CITE OR QUOTE
-------
QCC = QCCmean/lnormbnd(1.0, 0.203, 0.69, 1.42);
VPR = QAlvC/QCC;
vlm=-0.0036*(age/10)A2 - 0.0051*(age/10) + 0.0395;
if age>17
vlm=-0.0004*(age/10)A2 + 0.0034*(age/10) + 0.0169;
end
VLC=vlm*normbnd(1.0, 0.05, 0.85, 1.15);
vfm=0.1612*(age/10)A3 + 0.0846*(age/10)A2 - 0.3083*(age/10) + 0.2709;
if((age>7)&(age<=20))
vfm=-0.0458*(age/10)A2 + 0.2082*(age/10) + 0.0274;
if gend=="male"
vfm= -0.0057*(age/10)A2 + 0.0293*(age/10) + 0.1303;
end
elseif age>20
vfm=-0.0024*(age/10)A3 + 0.0355*(age/10)A2 - 0.115*(age/10) + 0.3678;
if gend=="male"
vfm= -0.0015*(age/10)A2 + 0.0384*(age/10) + 0.0908;
end
end
VFC=vfm*normbnd(1.0, 0.3, 0.1, 1.9);
VRC=normbnd(0.064, 0.0064, 0.0448, 0.0832);
VLUC=normbnd(0.0115, 0.00161, 0.00667, 0.0163);
VSC=normbnd(0.63, 0.189, 0.431, 0.829);
vtr=0.9215/(VFC+VLC+VLUC+VRC+VSC);
VFC = VFC*vtr; VLC=VLC*vtr; VLUC=VLUC*vtr; VRC=VRC*vtr; VSC=VSC*vtr;
DSPC=0.15;
QLC=normbnd(0.26, 0.0910, 0.010, 0.533);
QFC=normbnd(0.05, 0.0150, 0.0050, 0.0950);
QSC=normbnd(0.19, 0.0285, 0.105 ,0.276);
QRC=normbnd(0.50 ,0.10, 0.20, 0.80);
%Human partition coefficients for DCM
PL=lnormbnd(1.43, 1.22, 0.790, 2.59); %geometric mean and SD, converted from arith mean/SD of
1.46 and 0.292
PLU=lnormbnd(1.43, 1.22, 0.790, 2.59);%geometric mean and SD, converted from arith mean/SD of
1.46 and 0.292
PF=lnormbnd(l 1.9, 1.34, 4.92, 28.7); %geometric mean and SD, converted from arith mean/SD of
12.4 and 3.72
PS=lnormbnd(0.80, 1.22, 0.444, 1.46); %geometric mean and SD, converted from arith mean/SD of
0.82 and 1.64
PR=lnormbnd(1.43, 1.22, 0.790, 2.59); %geometric mean and SD, converted from arith mean/SD of
1.46 and 0.292
PB=lnormbnd(9.7, 1.10, 7.16, 13.0); %geometric mean and SD, converted from arith mean/SD of 9.7
and 0.97
% CO sub-model
RENCOS=0;
VBL2C=normbnd(0.059, 0.0177, 0.0148, 0.103);
DLC=normbnd(0.058, 0.02262, 0.01, 0.1);
ABCOC=0; %normbnd(0.1, 0.75, 0.01, 1);
RENCOC=0; %normbnd(0.05, 0.75, 0.01, 1);
HBTOT=normbnd(10, 0.5, 9, 11);
MMM=normbnd(178.3, 14.264, 152,210);
Pl=normbnd(0.71, 0.1278, 0.5, 1);
Fl=normbnd(0.85, 0.1275, 0.5, 1);
H-20 DRAFT - DO NOT CITE OR QUOTE
-------
COBGD=0; %normbnd(2.2, 2.2, 0, 4.4);
% File: finish.m
% Programmed by Paul Schlosser, U.S. EPA, 8/2008, rev. 1/2010
% Performs final analysis on saved results in runo from simulations.
%
disp([num2str(size(runo,l)),' simulations completed.']);
contsim=0; p=ctot(popn); p=p(l:min([3,length(p)]));
eval(['save runo @file=',astp,num2str(agem),p,'_',ctot(gendm), ...
'_',model(7:10),'.csv @format=ascii @separator=comma']);
eval(['save @file=',astp,num2str(agem),p,'_',ctot(gendm),'_',model(7:10),'.mat']);
disp(['GSTT-l group: ',ctot(popn)]);
aget=['a ',num2str(agem),' year-old;'];
if agem==0
aget='0.5-80 years of age;';
end
gendt=[ctot(gendm),' population'];
ifgendm=="both"
gendt='males and females';
end
disp(['Simulation for ',aget,gendt,'.']);
disp( ['Metric =',dtxt,'.']);
res=[];
pcs=num2str(percs( 1));
for n=2:length(percs)
pcs=[pcs,', ',num2str(percs(n))];
end
for n=l :length(nm)
disp([ctot(nm(n)),' mean, median, percentiles = ',pcs]);
r=[mean(runo(:,nn(n))),median(runo(:,nn(n))),prctile(runo(:,nn(n)),percs/gsmult)]*mult;
disp(r); disp(''); res=[res;r];
end
Files used specifically for cancer analysis
% File: straightsims.m
% Created by Paul Schlosser, U.S. EPA, 8/2008
% Runs Monte Carlo (MC) simulations withOUT search (for human equivalent exposures)
% Requires set exposure, NRUNS and rnames (list of variables to save)
% PBPK parameters set using MC selection by file human_pars.m.
exist contsim; % Test to see if contsim defined.
if~ans % If not...
contsim=0; % Not a continuation.
end
if contsim==0 % When starting a new set of simulation;
% set contsim=l if continuing an interupted chain.
runo=[]; rns=length(rnames); ns=l;
else
H-21 DRAFT - DO NOT CITE OR QUOTE
-------
ns=RUNN;
end
for RUNN = ns:NRUNS
ifmod(RUNN,50)==0
disp(['Remaining runs = ',num2str(NRUNS-RUNN)])
end
human_pars; start @NoCallback
runo=[runo;eval(sr)];
end
% File: Human drinking water MCA OSF.m
%
% Programmed by Michael Lumpkin
% Syracuse Research Corporation, 10/2007
% Modified by Paul Schlosser, U.S. EPA 8/2008, rev. 1/2010
%
% This script file sets up the control parameters to simulate human unit
% drinking water exposure (1 mg/kg/day) and calls straightsim which
% generates a Monte-Carlo chain for the internal doses (identified
% in text array nn) to be used in calculating oral slope factors (OSFs).
%
% Test to see if contsim defined.
exist contsim
if~ans % If not...
contsim=0; % Not a continuation.
end
if contsim==0 % When starting a new set of simulation;
% set contsim=l if continuing an interrupted chain.
clearT
nm=["LDAYPCTHB";"PEAKPCTHBCO";"AVRFROMDCM"];FIXDRDOSE=0.006;dtxt='
avg % HbCO; Peak %'; mult=l;
% Calculate distrib of PCTHBCO given RfD (0.006 mg/kg-day), PMS 2-14-11
nm=["LDAYLIVGSTDOSE";"LDAYWBDYGSTDOSE"];dtxt='n/a';
FIXDRDOSE= 1.0; mult= 1.0/FIXDRDOSE; % Drinking fixed mg/kg-day
nn=findnames(nm,rnames);
if (length(nn) ~= length(nm))
disp("Not all names in list nm.")
return
end
TEND=95.0; NRUNS=10000; CINT=0.1; % Total iterations for Monte Carlo analysis
gendm="both"; % Gender mix; choose "male, "female", or "both"
agem=0; % Age "mix"; if 0 draws from distribution (0.5-80 years)
% otherwise agem value is used exclusively
popn="++"; % GSTGT "++","+-", or "mix" of+/+, +/-, and -/-
model='human_par2'; % Choose model between 'human_parl' (original) and 'human_par2' (+
uncertainty)
astp='OSF_age'; percs=[95 99]; gsmult=1.0;
end
straightsims;
finish
H-22 DRAFT - DO NOT CITE OR QUOTE
-------
% File: Human inhalation MCA IUR.m
%
% Programmed by Michael Lumpkin
% Syracuse Research Corporation, 1 1/2007
% Modified by Paul Schlosser, 7/2008, rev. 1/2010
%
% This script file sets up the control parameters to simulate human unit
% inhalation exposures (1 mcg/m3) and calls straightsim which generates
% a Monte-Carlo chain for the internal doses (identified in text array
% nn) to be used in calculating inhalation unit risks (lURs).
% [[[
exist contsim; % Test to see if contsim defined.
if~ans % If not...
contsim=0; % Not a continuation.
end
if contsim==0 % When starting a new set of simulation;
% set contsim=l if continuing an interupted chain.
clearT;CINT=1.0;
TEND=24* 15; CONC=1 .Oe-6/GASD; CC=0; % Daily dose-rates now calculated from
% final day of simulations, & only need to go to 60 hr to reach SS
% CONC in ppm; 0.00029 ppm = 1 ug/m3 DCM = le-3 ug/L = le-6 mg/L
nm=["LDAYLIVGSTDOSE";"LDAYLUNGGSTDOSE";"LDAYWBDYGSTDOSE";"CV";"LD
AYAUCS"];
astp='IUR_age'; mult= 1.0e-6/(CONC*GASD); percs=[95 99]; dtxt=' n/a';
%CONC=2/(1000*GASD);nm=["LDAYPCTHB";"PEAKPCTHBCO";"AVRFROMDCM"];
dtxt=' avg % HbCO; Peak %'; mult= 1 ;
% Calculate distrib of PCTHBCO given RfC (0.6 mg/mmA3), PMS 2-14-1 1
nn=findnames(nm,rnames);
if (length(nn) ~= length(nm))
disp("Not all names in list nm.")
return
end
NRUNS= 10000; %Total iterations for Monte Carlo analysis
gendm="both"; % Gender mix; choose "male, "female", or "both"
agem=0; % Age "mix"; if 0 draws from distribution (0.5-80 years)
% otherwise agem value is used exclusively
popn="++"; % GSTGT "++", "+-", or "mix" of +/+, +/-, and -/-
gsmult = 1.0; % Multiplies draw for GST individual selection, divides the percentiles at end.
model='human_par2'; % Choose model between 'human_parl' (original) and 'human_par2' (+
uncertainty)
end
straightsims; finish
Files used specifically for non-cancer analysis
searchsim_refdose.m
% Created by Paul Schlosser, U.S. EPA, 8/2008
% Runs Monte Carlo (MC) simulations WITH search for human equivalent
% exposures, using variable LDAYREFDOSE ('reference dose based on
% last day of simulated exposure ~ presumed "periodicity").
% Requires exposure variable named = expnm = "DRCONC" or "CONC",
-------
% target value = heqt, relative tolerance = hetol, initial set
% exposure (CONC or DRCONC value >0) , NRUNS, and rnames (list of
% variables to save)
% PBPK parameters set using MC selection by file human_pars.m.
% [[[
% set contsim=l if continuing an interupted chain.
if contsim==0 % When starting a new set of simulations
if ~(expnm=="CONC" | expnm=="DRCONC")
disp('Variable expnm must be "CONC" or "DRCONC", in double quotes.');
return
end
if expnm=="CONC" % atxt used as variable command below
DRCONC=0.0; atxt='CONC = cl;'; c2=CONC;
else
CONC=0.0; atxt='DRCONC = cl;'; c2=DRCONC;
end
cl=c2; runo=[]; rns=length(rnames); ns=l; cls=['cl = c2*heqt/',dtxt,';'];
vls=['vl = ',dtxt,';']; v2s=[V2 = ',dtxt,';'];
else % Continuing a set of simulations
ns=RUNN;
end
for RUNN = ns:NRUNS
disp(['Remaining runs = ',num2str(NRUNS-RUNN)])
eval(model);
% Following block calculates daily dose resulting in specified internal dose
start @NoCallback
nstep=l; eval(v2s); eval(cls); eval(atxt);
start @NoCallback
eval(vls); hetl=hetol;
while (abs((vl/heqt)-l)>hetol); % Specify corresponding dose metric
en = abs((heqt-v2)*(cl-c2)/(vl-v2) + c2); % Linear interpolation to heqt
c2=cl; v2=vl; cl=cn; eval(atxt); % Assign values
start @NoCallback
eval(vls);
nstep=nstep+l;
if nstep==20
hetol=hetol*10; nstep=l;
disp('WARNING! hetol increased to allow convergence!')
end
end
hetol=hetl; runo=[runo;eval(sr)];
end
% File: human drinking water MCA RfD.m
%
% Programmed by Michael Lumpkin
% Syracuse Research Corporation, 1 1/2005
% Modified by Paul Schlosser, U.S. EPA 8/2008 and 1/2010
%
% This script file sets up the control parameters to simulate human
% drinking water exposure for a given internal dose (heqt) and calls
% searchsim_refdose which generates a Monte-Carlo chain for the human
-------
% equivalent administered daily dose (HEAD; mg/kg/d).
%
exist contsim;
if ~ans
contsim=0;
end
if contsim==0 % When starting a new set of simulation;
% set contsim=l if continuing an interupted chain.
clearT
nm=["DDOSE"]; nn=fmdnames(nm,rnames);
if (length(nn) ~= length(nm))
disp("Not all names in list nm.")
return
end
hetol=1.0e-4; %relative tolerance
expnm="DRCONC"; DRCONC=9.0; % 35 mg/L is 1 mg/kg/day for 70 kg human drinking 2L
TEND=95.0; NRUNS=10000; %Total iterations for Monte Carlo analysis
gendm="both"; % Gender mix; choose "male, "female", or "both"
agem=0; % Age "mix"; if 0 draws from distribution (0.5-80 years)
% otherwise agem value is used exclusively
popn="mix"; % GSTGT "++","+-", "-", or "mix" of+/+, +/-, and -/-
% dtxt = dose metric to use
% heqt = target for HEQ search; set value for specific dose metric
dtxt='LDAYREFDOSE'; dtl='W;
dtxt='LDAYLIVGSTDOSE'; dtl='G';
dtxt='LDAYLIVAUC'; dtl=A'; heqt=ll.l;
dtxt='LDAYLIVCYPDOSE'; dtl=Y'; heqt=13.31;
model='human_par2'; % Choose model between 'human_parl' (original) and 'human_par2' (+
uncertainty)
astp=['RfD_',dtl,'_',num2str(heqt),'_age']; percs=[5 1]; mult=l; gsmult=l;
end
searchsim refdose; finish
H-25 DRAFT - DO NOT CITE OR QUOTE
-------
% File: Human inhalation MCA RfC.m
%
% Programmed by Michael Lumpkin
% Syracuse Research Corporation, 11/2005
% Modified by Paul Schlosser, U.S. EPA, 7/2008 and 1/2010
%
% This script file sets up the control parameters to simulate human
% inhalation exposure for a given internal dose (heqt) and calls
% searchsim_refdose which generates a Monte-Carlo chain for the human
% equivalent cocnetration (HEC; mg/mA3).
%
exist contsim;
if ~ans
contsim=0;
end
if contsim==0 % When starting a new set of simulation;
% set contsim=l if continuing an interupted chain.
use clearT
nm=["CONC"]; nn=fmdnames(nm,rnames);
if (length(nn) ~= length(nm))
disp("Not all names in list nm.")
return
end
TEND=96; NRUNS= 10000; %Total iterations for Monte Carlo analysis
expnm="CONC"; CONC=67; % CONC in ppm; 0.29 ppm = 1 mg/m3 DCM
hetol=1.0e-5; %relative tolerance
gendm="both"; % Gender mix; choose "male, "female", or "both"
agem=0; % Age "mix"; if 0 draws from distribution (0.5-80 years)
% otherwise agem value is used exclusively
popn="mix"; % GSTGT "++","+-", "--", or "mix" of+/+, +/-, and -/-
gsmult = 1.0; % Multiplies draw for GST individual selection, divides the percentiles at end.
% dtxt = dose metric to use
% heqt = target for HEQ search; set value for specific dose metric
% The last statement below is the metric and target value used.
% Move desired one to end and save this file before running.
dtxt='LDAYLIVAUC'; dtl=A'; heqt=0.0562;
dtxt='LDAYLIVGSTDOSE'; dtl='G'; heqt=10.97;
dtxt='LDAYREFDOSE'; dtl='W; heqt=76.71;
dtxt='LDAYLIVCYPDOSE'; dtl=Y'; heqt=130.03;
model='human_par2'; % Choose model between 'human_parl' (original) and 'human_par2' (+
uncertainty)
astp=['RfC_',num2str(heqt),'_age']; mult=GASD*1000; percs=[5 1];
end
searchsim_refdose; finish
H-26 DRAFT - DO NOT CITE OR QUOTE
-------
Files used for mouse dose analyses
% File: mouse_set.m
% Programmed by Paul Schlosser, U.S. EPA, 7/2008
% Clears previous variables and sets parameters for mouse simulations.
% [[[
% prepare time history values.
prepare @clear @all
CINT=0.1; CONC=0.0; IVDOSE=0.0; BOLUS=0.0; DRCONC=0.0; FIXDRDOSE=0.0;
WESITG=0; WEDITG=0; CC=0;
% from Reitz et al. 1997, Table 1
DRPCT = [0.233, 0.1, 0.1, 0.1, 0.233, 0.234];
DRTIME= [0.0, 4.0, 8.0, 12.0, 16.0, 20.0];
TEND=336; % 2 weeks
TCFiNG=6; % daily exposure duration = 6 hours/day
TCFiNG2=120; % weekly dose width = 5 days/week =120 hours
TDUR=24; % daily dose period = 24 hours
TDUR2=168; % weekly exposure period = 7 days =168 hours
% U.S. EPA (1988) reference value for B6C3F1 mice: males=0.0373, females=0.0353
BW=0.0373;
% Mouse uptake & metabolism parameters (defined by MCMC calibration)
KA=5.0; VMAXC=9.27; KM=0.574; FRACR=0.0; KFC=1.41; AFFG=0.0; Al=0.207; A2=0.196;
% Mouse physiologic parameters (from prior distributions or MCMC
% calibration)
VLC=0.04; VFC=0.04; VRC=0.05; VLUC=0.0115; VBL2C=0.059; VSC=0.78;
QCC=24.2; VPR=1.45; DSPC=0.15; QLC=0.24; QFC=0.05; QSC=0.19; QRC=0.52;
%Mouse partition coefficients for DCM
PL=1.6; PLU=0.46; PF=5.1; PS=0.44; PR=0.52; PB=23.0;
start NoCallBack
% File: Mouse drinking water National Coffee 1983.m
%
% Programmed by Michael Lumpkin
% Syracuse Research Corporation, 01/2007
% Modified by Paul Schlosser, U.S. EPA, 8/2008
%
% This run time file sets species-specific constants and exposure
% values for National Coffee Association study of DCM in drinking water
% in the mouse; parameterized to run with DCM.07.rev3.csl.
% [[[
use mouse_set
BW=0.0373; TEND=4*7*24; runo=[];
% Mouse exposure parameters: DDOSE males: 61, 124, 177, or 244 mg/kg-day
forFIXDRDOSE=[61, 124, 177,244]
start @NoCallback;
-------
save runo @file='mouseOSFdrink.csv' @format=ascii @separator=comma
% File: Mouse Inhalation NTP 1986.m
%
% Programmed by Michael Lumpkin
% Syracuse Research Corporation, 1 1/2007
% Modified by Paul Schlosser, U.S. EPA, 7/2008
%
%This run time file sets species-specific constants and exposure
%values for NTP (1986) inhalation exposures of DCM in the mouse.
% [[[
use mouse_set
CINT=0.01;
% Kodak value; study average male: 2k=0.034, 4k=0.032; female 2k=0.030, 4k=0.029
TEND=5*7*24; Bws=[0.034, 0.032; 0.030,0.029]; runo=[]; Cs=[2000, 4000];
fors=[12]
forc=[l,2];
BW=Bws(s,c); CONC=Cs(c);
start @NoCallback
runo=[runo;[BW,CONC,WAVGLIVGSTDOSE,WAVGLUNGGSTDOSE,WAVGWBDYGSTDOSE]];
end
end
runo
%%% Saving last simulation results to file
run=[_t _cv _cvl _cvf _cvs _cvr _cab!2 _wavglivgstdose _wavglunggstdose _ddose];
save run @file='RUNOUT_NTP_inhal.csv' @format=ascii @separator=comma
Files used for rat dose analyses
% File: rat_set_D.m
% Programmed by Paul Schlosser, U.S. EPA, 9/2009
% Clears previous variables and sets parameters for rat simulations.
% Rat model "D" is a baseline for other models.
% [[[
% prepare time history values.
prepare @clear @all
WESITG=0; WEDITG=0;
% Rat exposure controls
CINT=0.01; CONC=0; IVDOSE=0; DRCONC=0; FIXDRDOSE=0; BOLUS=0; CC=0; RSTDY=0;
%DRTIME=[0:12, 12.6:0.6:23.4];
%DRPCT=[(0.21/12)*ones(12,l);(0.79/20)*ones(20,l)];
DRPCT = [0.233, 0.1, 0.1, 0.1, 0.233, 0.234, zeros(l,26)];
DRTIME = [0.0, 4.0, 8.0, 12.0, 16.0, 20.0, 21.1:0.1:23.6];
TEND=25; % hours
TCFING=6; % daily exposure = 6 hours
TCHNG2=120; % weekly dose width = 5 days/week = 120 hours
TDUR=24; % daily dose period = 24 hours
TDUR2=168; % weekly exposure period = 7 days = 168 hours
-------
VMAXC=3.992; KM=0.4; Al=0.002; A2=0.149; KFC=1.917; KZERC=0;
% Rat physiologic parameters (from Andersen et al 1991)
BW=0.220; % males:0.380 kg, females:0.229 kg, from EPA (1988)
VLC=0.04; VFC=0.07; VRC=0.05; VLUC=0.0115; VBL2C=0.059; VSC=0.75;
QCC=15.9; VPR=0.94; DSPC=0.15; QLC=0.20; QFC=0.09; QSC=0.15; QRC=0.56;
%Rat partition coefficients for DCM (Andersen et al 1991)
PL=0.732; PLU=0.46; PF=6.19; PS=0.408; PR=0.732; PB=19.4;
%CO submodel parameters
DLC=0.060; HBTOT=10; Pl=0.80; Fl=1.21; MMM=197.0; COINH=0;
COBGD=2.2; O2=0.13; PAIR=713; RHO=1102; SOL=0.03;
ABCOC=0; RENCOC=0; RENCOS=1; PCTHBCOO=0.7; % Bgd total blood CO ...
% ... and production to be calculated in csl to match PCTHBCOO
start @NoCallBack
VMAXC = 3.97; KM = 0.51; KFC = 2.47; PI = 0.68;
KA = 4.31; K12 = 35.0; KA2 = 1.67;
% File: rat_set_A.m
% Programmed by Paul Schlosser, U.S. EPA, 2/201 1
% Clears previous variables and sets parameters for rat simulations.
% Rat model "A".
% [[[
use rat_set_D
VMAXC=3.992; KM=0.4; Al=0.002;
KFC=1.917; A2=0.149; KA=5.0; K12=0;
P1=0.8;ABCOC=0.117;RENCOS=1;
RENCOC=0.035; COBGD=2.2;% File: rat_set_B.m
% Programmed by Paul Schlosser, U.S. EPA, 2/201 1
% Clears previous variables and sets parameters for rat simulations.
% Rat model "B".
% [[[
use rat_set_D
VMAXC = 3.97; KM = 0.509; KFC = 2.46; PI = 0.68;
%use inhivfit_rat-params % Un-comment this to use acslX-saved params
% File: rat_set_C.m
% Programmed by Paul Schlosser, U.S. EPA, 2/201 1
% Clears previous variables and sets parameters for rat simulations.
% Rat model "C".
% [[[
use rat_set_D
VMAXC = 3.97; KM = 0.51; KFC = 2.47; PI = 0.68;
KA = 4.31; K12 = 35.0; KA2 = 1.67;
K12=0; KA2=0; % For 2nd GI compartment [not used]
%File: Rat inhalation Nitschke 1988a.M
% Programmed by Michael Lumpkin
% Syracuse Research Corporation, 10/2007
-------
% Andersen et al (1987, 1991) simulating Nitschke et al. (1988a) exposures
% of DCM in the male and female rats.
% [[[
use rat_set_C
TEND=5*7*24; runo=[]; BW=0.229; TCHNG=6; TCHNG2=120;
% males:0.380 kg, females:0.229 kg, from U.S. EPA (1988)
for CONC=[50, 200, 500]
start @NoCallback
runo=[runo;[BW,CONC,DIDOSE,WAVGREFDOSE,WAVGLIVGSTDOSE,
WAVGLIVCYPDOSE,WAVGDAILYAUCL]];
end
runo
save runo @file='Rat_inhal_Nitschke.csv' @format=ascii @separator=comma
% File: Rat inhalation Burek et al. (1984).m
%
% Programmed by Michael Lumpkin
% Syracuse Research Corporation, 10/2007
% Modified by Paul Schlosser, U.S. EPA, 7/2008
%
% This run time file sets species-specific constants values for modified Andersen et al
% (1987, 1991) simulating Burek et al. (1984) exposures of DCM in male & female rats
use rat_set_C
TEND=5*7*24; TCHNG=6; TCHNG2=120; runo=[];
for BW=[0.523, 0.338]
for CONC=[500, 1500, 3500]
start @NoCallback
runo=[runo;[BW,DIDOSE,CONC,WAVGREFDOSE,WAVGLIVGSTDOSE,
WAVGLIVCYPDOSE,WAVGDAILYAUCL]];
end
end
runo
save runo @file='Rat_inhal_Burek.csv' @format=ascii @separator=comma
% File: Rat drinking water Serota 1986a.m
%
% Programmed by Michael Lumpkin
% Syracuse Research Corporation, 1/2007
% Modified by Paul Schlosser, U.S. EPA, 7/2008
%
% This run time file sets species-specific constants values for modified Andersen et al
% (1987, 1991) simulating Serota et al. (1986a) exposures of DCM to male & female rats
use rat_set_C
%DRCONC values from Serota et al. (1986a)
%males:0, 44.00, 381.36, 916.74, 1723.47
% females: 0, 37.80, 365.41, 856.82, 1656.94
TEND=7*24; runo=[]; BW=0.380; % Males
for FIXDRDOSE=[6, 52, 125, 235]
start (SNoCallBack
-------
runo=[runo;[BW,DDOSE,LDAYREFDOSE,LDAYLIVGSTDOSE,LDAYLIVCYPDOSE,
LDAYLIVAUC]];
end
BW=0.229; % Females
for FIXDRDOSE=[6, 58, 136, 263]
start @NoCallBack
runo=[runo;[BW,DDOSE,LDAYREFDOSE,LDAYLIVGSTDOSE,LDAYLIVCYPDOSE,
LDAYLIVAUC]];
end
runo
save runo @File="Serota_rat_DW.csv" @Format=ascii @Separator=comma
% File: Rat inhalation NTP 1986.M
%
% Programmed by Michael Lumpkin
% Syracuse Research Corporation, 1 1/2007
% Modified by Paul Schlosser, U.S. EPA, 7/2008
%
% This run time file sets species-specific constants values for modified Andersen et al
% (1987, 1991) simulating NTP (1986) exposures of DCM to male & female rats
use rat_set_A
use rat_set_C
CINT=0.1;
bws=[0.3905, 0.3852, 0.3848, 0.2455, 0.2443, 0.2422];
TEND=6*24*7; concs=[l 2412 4]* 1000; runo=[];
for n=l length(bws)
BW=bws(n); CONC=concs(n); start @NoCallback
runo=[runo;[BW,CONC, WAVGLIVCYPDOSE,WAVGLIVGSTDOSE, WAVGREFDOSE,
(WAVGLIVCYPDOSE/WAVGLIVGSTDOSE),WAVGAUCV,WAVGAUCS]];
end
runo
save runo @File="Rat_inhal_NTP.csv" @Format=ascii @Separator=comma
o/0
% File: Figure 5_3.m
% Programmed by Paul Schlosser, U.S. EPA, 9/2009
% Creates dichloromethane exposure-dose relationship for humans
% versus the rat shown in Figure 5-3 of the IRIS assessment
o/0
use rat_set_C
ods=[10:10:100,120,140,170,200,220,250];
TEND=7*24; ratr=[]; BW=0.380; % Male rats
forFIXDRDOSE=ods
start @NoCallBack
ratr=[ratr;[LDAYLIVGSTDOSE,LDAYLIVCYPDOSE,LDAYREFDOSE]];
disp(['Rat, oral dose = ',num2str(FIXDRDOSE),', livGSTdose =
',num2str(LDAYLIVGSTDOSE),'.']);
end
clearT; TEND=95.0; humr=[]; contsim=0;
H-31 DRAFT - DO NOT CITE OR QUOTE
-------
NRUNS=1000; % NRIMS = Total iterations for Monte Carlo analysis
gendm="both"; % Gender mix; choose "male, "female", or "both"
agem=0; % Age "mix"; if 0 draws from distribution (0.5-80 years)
% otherwise agem value is used exclusively
popn="mix"; gsmult=l; % GSTGT "++", "+-", or "mix" of+/+, +/-, and -/-
model='human_par2'; % Choose between 'human_parl' (original) and 'human_par2' (+
uncertainty)
nm=["LDAYLIVGSTDOSE";"LDAYLIVCYPDOSE";"LDAYREFDOSE"];
nn=fmdnames(nm,rnames);
if (length(nn) ~= length(nm))
disp("Not all names in list nm.")
return
end
for FIXDRDOSE=ods
rres=[FIXDRDOSE]; straightsims;
for n=l :length(nm)
rres=[rres,[mean(runo(:,nn(n))),prctile(runo(:,nn(n)),[5 95])]];
end
disp(['Human, dose = ',num2str(FIXDRDOSE),', internal doses = ...']); rres
humr=[humr;rres];
end
save @File="Fig5_3.mat"
plot(ods,ratr(:,l),ods,humr(:,2),ods,humr(:,4),'Fig5_3a.aps');
plot(ods,ratr(:,2),ods,humr(:,5),ods,humr(:,6),ods,humr(:,7),'Fig5_3b.aps');
plot(ods,ratr(:,3),ods,humr(:,8),ods,humr(:,9),ods,humr(:,10),'Fig5_3c.aps');
o/0
% File: Figure 5_7.m
% Programmed by Paul Schlosser, U.S. EPA, 9/2009
% Creates dichloromethane exposure-dose relationship for humans
% versus the rat shown in Figure 5-7 of the IRIS assessment
o/0
use rat_set_C
ccs=[10:10:60,80,100,150,200:100:600,800,1000];%,1500,2000:1000:5000];
TEND=2*7*24; ratr=[]; BW=0.229; TCHNG=6; TCHNG2=120;
for CONC=ccs
start @NoCallBack
ratr=[ratr;[CONC,WAVGLIVGSTDOSE,WAVGLIVCYPDOSE,WAVGREFDOSE]];
disp(['Rat, cone = ',num2str(CONC),', livGSTdose = ',num2str(WAVGLIVGSTDOSE),'.']);
end
save ratr @file='Fig5_7-rat.csv' @format=ascii @separator=comma
clear!; TEND=95.0; huml=[]; contsim=0;
NRUNS=1000; % NRUNS = Total iterations for Monte Carlo analysis
gendm="both"; % Gender mix; choose "male, "female", or "both"
agem=0; % Age "mix"; if 0 draws from distribution (0.5-80 years)
% otherwise agem value is used exclusively
popn="mix"; gsmult=l; % GSTGT "++", "+-", or "mix" of+/+, +/-, and -/-
model='human_par2';
% Choose between 'human_parl' (original) and 'human_par2' (+ uncertainty)
H-32 DRAFT - DO NOT CITE OR QUOTE
-------
nm=["LDAYLIVGSTDOSE";"LDAYLIVCYPDOSE";"LDAYREFDOSE";
"LDAYLUNGGSTDOSE"];
nn=fmdnames(nm,rnames);
if (length(nn) ~= length(nm))
disp("Not all names in list nm.")
return
end
for CONC=ccs
rres=[CONC]; straightsims;
for n=l :length(nm)
rres=[rres,[mean(runo(:,nn(n))),prctile(runo(:,nn(n)),[5 95])]];
end
disp(['Human, cone = ',num2str(CONC),', internal doses = ...']); rres
huml=[huml; rres];
end
savehuml @file='Fig5_7-human.csv' @format=ascii @separator=comma
save @File="Fig5_7.mat"
plot(ccs,ratr(:,2),ccs,huml(:,2),ccs,huml(:,4),'Fig5_7a.aps')
plot(ccs,ratr(:,3),ccs,huml(:,5),ccs,huml(:,6),ccs,huml(:,7),'Fig5_7b.aps')
plot(ccs,ratr(:,4),ccs,huml(:,8),ccs,huml(:,9),ccs,huml(:,10),'Fig5_7c.aps')
o/0
% File: Figure 5_9nlO_rat_sense.m
% Programmed by Paul Schlosser, U.S. EPA, 9/2009
% Performs sensitivity analysis for rat oral (Figure 5-9) and inhalation
% (Figure 5-10) dichloromethane exposures, and writes results to file
% 'ratsense_Figs59n!0.csv' for plotting in Excel
o/0
use rat_set_C
FIXDRDOSE=10; TEND=7*24; ratr=[]; BW=0.380; % Male rats
start @nocallback
rO=[LDAYLIVGSTDOSE,LDAYLIVCYPDOSE,LDAYREFDOSE];
pars=["QCC","VPR","VLC","VSC","PB","VMAXC","KA","A2","KFC"];
for pt=pars
pO=eval(pt); setbase(pt,1.01*pO); start @nocallback
r=[LDAYLIVGSTDOSE,LDAYLIVCYPDOSE,LDAYREFDOSE];
setbase(pt,0.99*pO); start @nocallback
r=50*(r-[LDAYLIVGSTDOSE,LDAYLIVCYPDOSE,LDAYREFDOSE])./rO;
setbase(pt,pO);
ratr=[ratr;r];
end
FIXDRDOSE=0; CONC=500; TEND=5*7*24; BW=0.229; TCHMG=6; TCHNG2=120;
start @nocallback
rO=[WAVGLIVGSTDOSE,WAVGLIVCYPDOSE,WAVGREFDOSE];
for pt=pars
pO=eval(pt); setbase(pt,1.01*pO); start @nocallback
r=[WAVGLIVGSTDOSE,WAVGLIVCYPDOSE,WAVGREFDOSE];
H-33 DRAFT - DO NOT CITE OR QUOTE
-------
setbase(pt,0.99*pO); start @nocallback
r=50*(r-[WAVGLIVGSTDOSE,WAVGLIVCYPDOSE,WAVGREFDOSE])./rO;
setbase(pt,pO);
ratr=[ratr;r];
end
ratr
save ratr @file="ratsense_Figs59nl0.csv" @format=ascii @separator=comma
o/0
% File: Figure 5_14.m
% Programmed by Paul Schlosser, U.S. EPA, 9/2009
% Creates dichloromethane exposure-dose relationship for humans versus the mouse
% shown in Figure 5-14 of the IRIS assessment; results are written to file
% "Fig5-14_human.csv" for plotting in Excel
o/0
use mouse_set
ods=[10:10:100,120,140,170,200,220,250];
TEND=7*24; mour=[]; BW=0.033; TCHNG=2000; TCHNG2=2000;
model='human_par2';
for FIXDRDOSE=ods
start @NoCallBack
mour=[mour;WAVGLIVGSTDOSE];
disp(['Mouse, dose = ',num2str(FIXDRDOSE),',
livGSTdose = ',num2str(WAVGLIVGSTDOSE),'.']);
end
save mour @file="Fig5-14_mouse.csv" @format=ascii @separator=comma
clear!; TEND=95.0; hum2=[]; contsim=0;
NRUNS=1000; % NRUNS = Total iterations for Monte Carlo analysis
gendm="both"; % Gender mix; choose "male, "female", or "both"
agem=0; % Age "mix"; if 0 draws from distribution (0.5-80 years)
% otherwise agem value is used exclusively
gsmult=l; nm=["LDAYLIVGSTDOSE"]; nn=fmdnames(nm,rnames);
if isempty(nn)
disp(" Variable name not in list nm.")
return
end
forFIXDRDOSE=ods
rres=[FIXDRDOSE]
for popn=["mix","+-","++"] % GSTGT "++", "+-", or "mix" of+/+, +/-, and -/-
straightsims;
for n=l :length(nm)
rres=[rres,[mean(runo(:,nn(n))),prctile(runo(:,nn(n)),[5 95])]]
end
disp(['Human',ctot(popn),', dose = ',num2str(FIXDRDOSE),', internal doses = ...']); rres
end
hum2=[hum2; rres];
end
save hum2 @file="Fig5-14_human.csv" @format=ascii @separator=comma
H-34 DRAFT - DO NOT CITE OR QUOTE
-------
save @file="Fig5_14.mat"
plot(ods,mour,ods,hum2(:,2),ods,hum2(:,4), ods,hum2(:,5),ods,hum2(:,6),ods,hum2(:,7), ...
ods,hum2(:,8),ods,hum2(:,9),ods,hum2(:,10),Tig5_14.aps')
o/0
% File: Figure 5_15.m
% Programmed by Paul Schlosser, U.S. EPA, 9/2009
% Creates dichrlomethane exposure-dose relationship for humans versus the
% mouse shown in Figure 5-15 (panels A and B) of the IRIS assessment;
% results are written to file "Fig5-15_human.csv" for plotting in Excel
use mouse_set
ccs=[10:10:100,120,150,200,230,300,330,400,430,500:100:1000,1300,2000:1000:5000];
TEND=5*7*24; mour=[]; BW=0.031; TCHNG=6; TCHNG2=120;
for CONC=ccs
start @NoCallBack
mour=[mour;[CONC,WAVGLIVGSTDOSE,WAVGLUNGGSTDOSE]];
disp(['Mouse, cone = ',num2str(CONC),', livGSTdose = ',num2str(WAVGLIVGSTDOSE),'.']);
end
clearT; TEND=95.0; hum3=[]; contsim=0;
NRUNS=1000; % NRUNS = Total iterations for Monte Carlo analysis
gendm="both"; % Gender mix; choose "male, "female", or "both"
agem=0; % Age "mix": draws from distribution (0.5-80 years)
model='human_par2'; gsmult=l;
nm=["LDAYLIVGSTDOSE";"LDAYLUNGGSTDOSE"];nn=findnames(nm,rnames);
if (length(nn) ~= length(nm))
disp("Not all names in list nm.")
return
end
for CONC=ccs
rres=[CONC];
for popn=["+-","++"] % GSTGT "++", "+-", or "mix" of+/+, +/-, and -/-
straightsims;
for n=nn
rres=[rres,[mean(runo(:,n)),prctile(runo(:,n),[5 95])]];
end
if popn=="mix"
forij=[0,l]
runo2=runo(runo(:, 31 )==ij,:);
for n=l :length(nm)
rres=[rres,[mean(runo2(:,nn(n))),prctile(runo2(:,nn(n)),[5 95])]];
end
end
end
disp(['Human, cone = ',num2str(CONC),', internal doses = ...']); rres
end
hum3=[hum3;rres];
end
H-3 5 DRAFT - DO NOT CITE OR QUOTE
-------
savehumS @file="Fig5-15_human.csv" @format=ascii @separator=comma
save @File="Fig5_15.mat"
% Liver dose plot...
%plot(ccs,mour(:,2),huml(:,l),huml(:,2),huml(:,l),huml(:,4), ...
% ccs,hum3(:,2),ccs,hum3(:,3),ccs,hum3(:,4), ...
% ccs,hum3(:,8),ccs,hum3(:,9),ccs,hum3(:,10),Tig4_3.aps')
plot(ccs,mour(:,2),huml(:,l),huml(:,2),ccs,hum3(:,2),ccs,hum3(:,8),'Fig5_15a.aps')
% Lung dose plot...
%plot(ccs,mour(:,l),huml(:,l),huml(:,ll),huml(:,l),huml(:,13), ...
% ccs,hum3(:,5),ccs,hum3(:,6),ccs,hum3(:,7), ...
% ccs,hum3(:,ll),ccs,hum3(:,12),ccs,hum3(:,13),'Fig4_4.aps')
plot(ccs,mour(:,3),huml(:,l),huml(:,ll),ccs,hum3(:,5),ccs,hum3(:,ll),'Fig5_15b.aps')
o/0
% File: Fig5_16tol8_mouse_sense.m
% Programmed by Paul Schlosser, U.S. EPA, 9/2009
% Runs sensitivity analyses for dichloromethane exposures in mice,
% producing results plotted in Figures 5-18 to 5-18 of the IRIS assessment.
% Values saved to file "mousesense_Figs5_16to!8.csv" for plotting in Excel.
o/0
use mouse_set
CINT=0.001; TEND=(3*7*24)-CINT;
CONC=2000; TCHNG=6; TCHNG2=120; mour=[]; BW=0.031; % average mouse
start @nocallback
rO=WAVGLIVGSTDOSE;
pars=["QCC","VPR","VLC","VSC","PB","VMAXC","KA","A2","KFC"];
for pt=pars
pO=eval(pt);
setbase(pt,1.01*pO); start @nocallback
r=WAVGLIVGSTDOSE;
setbase(pt,0.99*pO); start @nocallback
r= 50*(r-WAVGLIVGSTDOSE)/rO;
setbase(pt,pO);
mour=[mour,r];
end
DRCONC=500; CONC=0; start @nocallback
rO=LDAYLIVGSTDOSE; m5=[];
for pt=pars
pO=eval(pt); setbase(pt,1.01*pO); start @nocallback
r=LDAYLIVGSTDOSE;
setbase(pt,0.99*pO); start @nocallback
r= 50*(r-LDAYLIVGSTDOSE)/rO;
setbase(pt,pO);
m5=[m5,r];
end
mour=[mour;m5];
CONC=500: DRCONC=0: start (S)nocallback
H-36 DRAFT - DO NOT CITE OR QUOTE
-------
rO=WAVGLUNGGSTDOSE; m5=[];
for pt=pars
pO=eval(pt); setbase(pt,1.01*pO); start @nocallback
r=WAVGLUNGGSTDO SE;
setbase(pt,0.99*pO); start @nocallback
r= 50*(r-WAVGLUNGGSTDOSE)/rO;
setbase(pt,pO);
m5=[m5,r];
end
mour=[mour;m5]
save mour @file="mousesense_Figs5_16tol8.csv" @format=ascii @separator=comma
o/0
% File: FigC3 rat inhal Gargas 86.m
% Figure C-3 (creates 2 sub-plots)
%
% Programmed by Michael Lumpkin
% Syracuse Research Corporation, 10/2007
% Modified by Paul Schlosser, U.S. EPA, 9/2009
%
% This run time file sets species-specific constants values for modified
% Andersen et al (1987, 1991) simulating Burek et al. (1984) exposures
% of dichloromethane in the male and female rats
o/0
dchl=[0.103, 93.73; 0.2178, 81.12; 0.3707, 66.26; 0.5543, 52.58; 0.7227, 41.73;
0.8834, 34.085; 1.044, 27.84; 1.22, 22.095; 1.3807, 18.312; 1.55, 14.532;
1.717, 12.58; 1.893, 10.28; 2.061, 8.3948];
dch5=[0.1015, 448.9; 0.216, 400.0; 0.3843, 326.7; 0.5522, 282.85; 0.7128, 234.4;
0.8806,208.9; 1.056, 183.5; 1.231, 163.5; 1.384, 145.7; 1.5514, 133.7;
1.719, 119.14; 1.887, 106.2; 2.0395, 95.99; 2.192, 85.5385; 2.360, 75.13;
2.5276, 65.99; 2.688, 57.96; 2.8484, 49.45; 3.024, 40.99; 3.1996, 36.00;
3.360, 31.165; 3.5356, 25.46; 3.719, 21.10; 3.864, 18.00; 4.025, 15.585;
4.193, 13.297; 4.353, 11.51];
dchlO=[0.0973, 927.1; 0.2195, 826.0; 0.3957, 627.6; 0.5556, 575.7; 0.7235, 513.1;
0.8836, 457.2; 1.059, 419.5; 1.226, 401.9; 1.3936, 379.6; 1.561, 348.2;
1.721, 333.66; 1.873, 329.1; 2.0556, 301.9; 2.200, 285.1; 2.3754, 265.4;
2.5426, 258.0; 2.695, 236.7; 2.87, 226.8; 3.030, 211.1; 3.1975, 193.66;
3.365, 182.9; 3.532, 175.24; 3.7076, 156.175; 3.875, 145.4; 4.035, 135.3;
4.195, 124.1; 4.340, 112.2; 4.515, 101.5; 4.683, 89.12];
dch30=[0.1058, 2958.6; 0.2288, 2280; 0.3819, 1862.5; 0.5419, 1684; 0.7095, 1545;
0.8848, 1397; 1.067, 1338.6; 1.227, 1264; 1.387, 1229; 1.5615, 1195;
1.7136, 1178.5; 1.881, 1129; 2.056, 1051; 2.208, 1052; 2.383, 1008;
2.5424, 994.2; 2.6945, 966.46; 2.869, 939.6; 3.0366, 913.45; 3.196, 914.2;
3.356, 863.3; 3.531, 851.6; 3.713, 816.1; 3.880, 805.0; 4.032, 794.0;
4.192, 760.75; 4.352, 750.4; 4.5114, 708.6; 4.686, 709.2; 4.853, 699.6;
5.0055, 670.3; 5.173, 642.26; 5.34, 633.53];
H-37 DRAFT - DO NOT CITE OR QUOTE
-------
formodel=['A"C']
eval(['use rat_set_',model]);
BW=0.225; TEND=6; TCHNG=7; CC=1; VCHC=9; NCH=3; CINT=0.05;
prepare @clear T CCHPPM
res=[];
for CONC=[107, 498, 1028, 3206]
start @nocallback
res=[res,_t,_cchppm];
end
eval(['save res @file=gargas-inh-sim-',model,'.csv @format=ascii @separator=comma'])
plot(dchl(:,l),dchl(:,2),dch5(:,l),dch5(:,2),dchlO(:,l),dchlO(:,2), ...
dch30(:,l),dch30(:,2),res(:,l),res(:,2),res(:,3),res(:,4), ...
res(:,5),res(:,6),res(:,7),res(:,8),['gargas86',model,'.aps'])
end
dchl=[dchl;dch5;dchlO;dch30];
save dchl @file='gargas-inh-dat.csv' @format=ascii @separator=comma
o/0
% File: FigC4n5 Angelo_IV_comp.m
% Figures C-4 and C-5 (creates up to 4 sub-plots for each)
% Programmed by Michael Lumpkin
% Syracuse Research Corporation, 10/2007
% Modified by Paul Schlosser, U.S. EPA, 9/2009
% This run time file sets species-specific constants values for the
% modified Andersen et al (1987, 1991) PBPK model for dichloromethane,
% simulating Angelo et al. (1986b) IV exposures of DCM in rats
d!0bl=[2, 17.624; 5, 8.6635; 10, 3.533; 15, 2.858; 20, 1.7125;
30, 0.826; 40, 0.6192];
d50bl=[2, 62.686; 5, 34.105; 10, 16.56; 15, 12.33; 20, 12.80;
30, 10.25; 40, 5.561];
d!0ex=[0.3333, 32.245; 0.6667, 38.56; 1, 41.396; 4, 44.42];
d50ex=[0.3333, 35.38; 0.6667, 45.95; 1, 51.38; 4, 57.59];
for model=['A 'C'] %['A 'B' 'C' 'D'] % Use set of letters for models to test
eval(['use rat_set_',model]);
prepare @clear T CV PCTIVEXH
TEND=10; CINT=0.01;
IVDOSE=10
start @nocallback
res=[_t,_cv,_pctivexh]; r2=[_t*60,_cv,_t,_pctivexh];
IVDOSE=5; start @nocallback
res=[res,_t,_cv,_pctivexh]; r2=[r2,_t*60,_cv,_t,_pctivexh];
IVDOSE=25; start @nocallback
res=[res,_t,_cv,_pctivexh]; r2=[r2,_t*60,_cv,_t,_pctivexh];
H-3 8 DRAFT - DO NOT CITE OR QUOTE
-------
IVDOSE=50; start @nocallback
res=[res,_t,_cv,_pctivexh];r2=[r2,_t*60,_cv,_t,_pctivexh];
plot(dlObl(:,l),dlObl(:,2),d50bl(:,l),d50bl(:,2),...
res(:,l)*60,res(:,2),_t*60,_cv,res(:,4)*60,res(:,5), ...
res(:,7)*60,res(:,8),['angivbl',model,'.aps'])
plot(dl0ex(:,l),d!0ex(:,2),d50ex(:,l),d50ex(:,2), ...
res(:,l),res(:,3),_t,_pctivexh,res(:,4),res(:,6), ...
res(:,7),res(:,9),['angivex',model,'.aps'])
eval(['save r2 @file=Angelo_IV_sims_',model,'.csv @format=ascii @separator=comma'])
end
o/0 --------------------------------------------------------------------
% File: FigC6 rat_inhal_Andersen.m
% Figures C-6 (creates multiple sub-plots)
%
% Programmed by Michael Lumpkin
% Syracuse Research Corporation, 10/2007
% Modified by Paul Schlosser, U.S. EPA, 9/2009
%
% This run time file sets species-specific constants values for the
% modified Andersen et al (1987, 1991) PBPK model for dichloromethane,
% simulating Andersen et al. (1987) inhalation exposures of DCM in rats
o/0
dan2=[l, 2.6634; 2, 4.290; 3, 4.9915; 4, 4.464; 4.25, 1.3334; 4.5, 0.5513;
4.75, 0.4234; 5, 0.2618; 5.25, 0.1206; 5.5, 0.1049];
danlO=[l, 29.34; 2, 49.50; 3, 47.10; 4, 50.72; 4.25, 14.92; 4.5, 9.225;
4.75, 6.260; 5, 3.811; 5.25, 1.3085; 5.5, 0.6515; 5.75, 0.6025];
can2=[0.9942, 3.967; 1.027, 4.121; 1.039, 3.798; 2.045, 7.151;
2.0565, 6.598; 2.057, 5.7836; 3.053, 8.4295; 3.076, 7.493;
3.086, 8.7215; 4.073, 9.094; 4.074, 8.003; 4.084, 8.940; 4.5545, 5.302;
4.597, 6.853; 4.608, 7.1145; 5.078, 3.5525; 5.079, 2.723; 5.0795, 2.124;
5.302, 1.464; 5.313, 1.971; 5.325, 1.142];
canlO=[1.006, 4.004; 1.018, 4.305; 1.021, 5.114; 2.0494, 9.257;
2.061, 9.488; 2.068, 7.9395; 3.047, 10.42; 3.0535, 8.777; 3.055, 9.170;
4.030, 6.449; 4.030, 10.31; 4.036, 12.11; 5.0275, 7.448; 5.061, 7.656;
5.067, 9.598; 5.5285, 4.978; 5.540, 5.001; 5.552, 9.115; 6.028, 1.999;
6.030, 6.460; 6.049, 1.653; 6.510, 0.4305; 6.525, 2.072; 6.539, 2.973];
Tab3=[];
for model=['A 'C']; %['A 'B' 'C' T>']
eval(['use rat_set_',model]);
prepare @clear T CV PCTHBCO
start @nocallback
QCSW=0; % No work/rest change
TEND=7; CONC=200; CINT=0.05; TCHNG=4; BW=0.237; start @nocallback
start @nocallback
re s=[_t,_cv,_pcthb co];
TEND=4*24; start @nocallback
Tab3=[Tab3;[CONC,LDAYLIVCYPDOSE, LDAYLIVGSTDOSE,(LDAYLIVCYPDOSE
LDAYLIVGSTDOSE),(LDAYLIVCYPDOSE/LDAYLIVGSTDOSE)]];
H-39 DRAFT - DO NOT CITE OR QUOTE
-------
ifmodel=='C'
CONC=950; TEND=7; start @nocallback
res=[res,_cv];
end
%CONC=750; start @nocallback
CONC=1000; TEND=7; start @nocallback
res2=[_t,_cv,_pcthbco]; res=[res,res2];
plot(dan2(:,l),dan2(:,2),danlO(:,l),danlO(:,2), ...
res(:,l),res(:,2),res2(:,l),res2(:,2),res(:,l),res(:,4),['andinhbl^model,'.aps'])
plot(can2(:,l),can2(:,2),canlO(:,l),canlO(:,2), ...
res(:,l),res(:,3),_t,_pcthbco,['andinhhb',model,'.aps'])
eval(['save res @file=Andersen87_inh_sims_',model,'OO.csv @format=ascii ...
@separator=comma'])
TEND=4*24; start @nocallback
Tab3=[Tab3;[CONC,LDAYLIVCYPDOSE, LDAYLIVGSTDOSE,(LDAYLIVCYPDOSE
LDAYLIVGSTDOSE),(LDAYLIVCYPDOSE/LDAYLIVGSTDOSE)]];
end
Tab3=Tab3([l,3,2,4],:)
save can2 @file=andersen_200ppm_co_data.csv @Format=Ascii @Separator=comma
o/0
% File: FigC7_Andersen_91_Fig4.m
% Programmed by Paul Schlosser, U.S. EPA, 1/2011
% Creates plot shown in Figure C-7 of the dichloromethane IRIS
% assessment. Figure C-7 shows model A and C simulations versus
% % HbCO data of Andersen et al (1991) for a 30-min inhalation
% exposure to 5159 ppm DCM. The model A simulations use the
% parameters as listed in Table 1 of Andersen et al. (1991) rather
% than those listed in the caption of Figure 4, for a fair "apples
% to apples" comparison to model C; i.e., in both cases this is a
% validation of the parameters/model fit to other data, rather than
% an alternate model w/ the same struchture.
o/0
prepare @clear T PCTHBCO
% Anderse et al. (1991) rat timecourse data for inhalation of
% 5159 ppm DCM for 30 min.
% time (hr), % HbCO
load @file=Andersen_91_Fig4.csv @Format=Ascii; dat=Andersen_91_Fig4;
use rat_set_A
TEND=7; BW=0.24; CINT=0.1; CONC=5159; TCHNG=0.5; start @nocallback
start @nocallback
tl=_t; pl=_pcthbco;
use rat_set_C
TEND=7; BW=0.24; CINT=0.1; CONC=5159; TCHNG=0.5; start @nocallback
start @nocallback
plot(dat(:,l),dat(:,2),tl,pl,_t,_pcthbco,[0,0.5],[0.3,0.3],'and91fig4.aps')
H-40 DRAFT - DO NOT CITE OR QUOTE
-------
0/0
% File: FigCS rat inhal exp_dose.m
% Programmed by Paul Schlosser, U.S. EPA, 1/2011
% Creates exposure-dose relationship for rat inhalation exposures to
% dichloromethane, for models A and C, as shown in Figure C-8 of the
% IRIS assessment
o/0
use rat_set_C
prepare @clear T CV PCTHBCO
bwm=mean([0.3905, 0.3852, 0.3848, 0.2455, 0.2443, 0.2422]); res=[];
formodel=['A"C'];
eval(['use rat_set_',model]);
prepare @clear T CV PCTHBCO
TEND=6*24*7; CINT=0.1; TCHNG=6.0; BW=bwm; r=[0,0,0]; start @nocallback
for CONC=[200,1000,2000,4000] %[10:10:100, 100:20:300, 400:100:1000, ...
1200:200:2000]
start @nocallback
r=[r;[CONC,WAVGLIVGSTDOSE,WAVGLIVCYPDOSE]];
end
res=[res,r]
end
plot(res(:,l),res(:,2),res(:,l),res(:,3), ...
res(:,4),res(:,5),res(:,4),res(:,6),'ratlinear.aps')
% File: FigC9nlO rat_oral_fits.m
% Programmed by Paul Schlosser, U.S. EPA, 2/2011
% Creates plots shown in Figure C-9 (4 panels) and Figure C-10 of the
% dichloromethane IRIS assessment. Figure C-9 shows model C fits to
% data of Angelo et al (1986b) for 50 and 200 mg/kg bolus oral
% exposures. Figure C-10 is fit of model D with KA = 4.31 or = 0.62.
% Results now written to an Excel file where plots were created.
load @file=angelo_86_50mg.csv @Format=Ascii; da50=angelo_86_50mg; % 50 mg/kg
load @file=angelo_86_200mg.csv @Format=Ascii; da200=angelo_86_200mg; % 200 mg/kg
% Angelo et al. (1986b) rat timecourse data from
% oral dosing of 50 or 200 mg/kg
% time (hr), blood DCM, liver DCM, % exhaled DCM, % exh CO
fit=0 % Set to 0 to not use all_fit-params
use rat_set_C
if fit
use allfit_rat-params
end
prepare @clear T CV CL PCTIVEXH PCTIVCOEXH PCTHBCO HBCO
RENCOC=0; RENCOS=0; TEND=25; BW=0.259; BOLUS=50; CINT=0.05; start @nocallback
start @nocallback
ra50=[_t, _cv, _cl, _pctivexh, _pctivcoexh];
H-41 DRAFT - DO NOT CITE OR QUOTE
-------
pth="\\AA.AD.EPA.GOV\ORD\RTP\USERS\K-Q\pschloss\NetMyDocuments\DCM\rat
PBPBK DCM paper figuresVRat DCM PBPK Figure 8.xls"
xlsWrite(pth,"gargas-oral-dat","Gl :K501 ",ra50)
use KAfit-params
start @nocallback
ra50=[ra50, _cv, _cl, _pctivexh, _pctivcoexh];
xlsWrite(pth,''gargas-oral-dat'V'Rl:U50r,ra50(:,6:9))
use rat_set_D
if fit
use allfit_rat-params
end
RENCOC=0; RENCOS=0; TEND=25; BW=0.259; BOLUS=200; CINT=0.05; start
@nocallback
ra200=[_t, _cv, _cl, _pctivexh, _pctivcoexh];
xlsWrite(pth,"gargas-oral-dat","Ml:P501",ra200(:,2:5))
use KAfit-params
start @nocallback
ra200=[ra200, _cv, _cl, _pctivexh, _pctivcoexh];
xlsWrite(pth,"gargas-oral-dat","Wl:Z501",ra200(:,6:9))
plot(da50(:,l),da50(:,2),ra50(:,l),ra50(:,2),da200(:,l),da200(:,2), ...
ra200(:,l),ra200(:,2),ra50(:,l),ra50(:,6),ra200(:,l),ra200(:,6),'angbl.aps')
plot(da50(:,l),da50(:,3),ra50(:,l),ra50(:,3),da200(:,l),da200(:,3), ...
ra200(:,l),ra200(:,3),ra50(:,l),ra50(:,7),ra200(:,l),ra200(:,7),'angli.aps')
plot(da50(:,l),da50(:,4),ra50(:,l),ra50(:,4),da200(:,l),da200(:,4), ...
ra200(:,l),ra200(:,4),ra50(:,l),ra50(:,8),ra200(:,l),ra200(:,8),'angexd.aps')
plot(da50(:,l),da50(:,5),ra50(:,l),ra50(:,5),da200(:,l),da200(:,5), ...
ra200(:,l),ra200(:,5),ra50(:,l),ra50(:,9),ra200(:,l),ra200(:,9),'angexc.aps')
% Pankow et al. '91; COHb following a single gavage dose
% of 6.2 mmol/kg DCM (526 mg/kg) in 259 gram male Wistar rats
load @file=Pankow_91_526mgHb.csv @Format=Ascii; dpank=Pankow_91_526mgFIb;
use rat_set_C
if fit
use allfit_rat-params
end
RENCOS=1; TEND=12.5; BOLUS=526; BW=0.259; PCTHBCOO=1.0; CINT=0.05;
start @nocallback
ps=[_t,_pcthbco];
%use KAfit-params
KA=0.62;
start @nocallback
% See comment on line 15 above re. KA
plot(dpank(:,l),dpank(:,2),ps(:,l),ps(:,2),_t,_pcthbco,'pankhb.aps')
ra50=[ra50,ra200];
save ra50 @file=Angelo-oral-sim.csv @format=ascii @separator=comma
% File: FigCl l_McKennay_oral_rat.m
% Programmed by Paul Schlosser, U.S. EPA, 1/2011
H-42 DRAFT - DO NOT CITE OR QUOTE
-------
% Creates plots shown in Figure C-l 1 of the dichloromethane IRIS
% assessment. Figure C-l 1 shows model C fits to data of McKenna
% and Zempel (1981) for disposition of 14C-DCM for 1 and 50 mg/kg
% bolus oral exposures.
o/0
use rat_set_C
prepare @clear T PCTIVEXH
TEND=4.7; BW=0.25; CINT=0.1;
RENCOS=0.0; COINH=0.0; PSCOO.O; PCTHBCOO=0.0; RENCOC=0.0;
% McKenna and Zempel (1981) rat timecourse data from
% oral dosing of 1 or 50 mg/kg
% time (hr), % exhaled DCM in 30-min interval, ...
% cumulative % exhaled DCM
load @file=McKenna81_lmg_exhDCM.csv @Format=Ascii; dal=McKenna81_lmg_exhDCM;
% 1 mg/kg data
load @file=McKenna81_50mg_exhDCM.csv @Format=Ascii;
da50=McKenna81_50mg_exhDCM; % 50 mg/kg data
% Also showing Angelo et al. (1986b) data, also 50 mg/kg, for comparison
load @file=angelo_86_50mg.csv @Format=Ascii; d50=angelo_86_50mg; % 50 mg/kg
BOLUS=1; start @nocallback
t=_t, cex= _pctivexh;
TEND=48; start @nocallback
pl=PCTIVCOEXH;
BOLUS=50; start @nocallback
pl=[l,50;pl,PCTIVCOEXH]
TEND=6.2; start @nocallback
plot(dal(:,l),dal(:,3),t,cex,da50(:,l),da50(:,3),_t,_pctivexh, d50(:,l), d50(:,4),'mckenna.aps')
o/0
% File: FigureG-l_rat-human-inahl-AUC-dosim.m
% Programmed by Paul Schlosser, U.S. EPA, 4/2011
% Creates exposure-dose relationship for humans versus the rat
% shown in Figure G-l of the IRIS assessment.
o/o
ccs=[10:10:60,80,100,150,200:100:600,800,1000,1500,2000:1000:5000];
clear!; TEND=95.0; huml=[]; contsim=0;
NRUNS=3000; % NRUNS = Total iterations for Monte Carlo analysis
gendm="both"; % Gender mix; choose "male, "female", or "both"
agem=0; % Age "mix"; if 0 draws from distribution (0.5-80 years)
% otherwise agem value is used exclusively
popn=["mix"]; % "mix" of+/+, +/-, and -/-
% +/+ and +/- subsets are sampled in for-CONC loop below;
model='human_par2'; % Choose between 'human_parl' (original) and 'human_par2' (+
uncertainty)
nm=["LDAYAUCS";"GSTGT"]; nn=findnames(nm,rnames); gsmult=l;
if (length(nn) ~= length(nm))
disp("Not all names in list nm.")
return
H-43 DRAFT - DO NOT CITE OR QUOTE
-------
end
for CONC=ccs
straightsims; res=mean(runo(:,nn(l))); rres=[CONC,res]; % Results for full population
disp(['Human [',ctot(popn),'], cone = ',num2str(CONC),', AUCSlowly = ',num2str(res)]);
res=mean(runo(runo(:,nn(2))==l,nn(l))); rres=[rres,res]; % Results for full GST +/-
% nn(2) is the column index for the GST type identifier (GSTGT), which has values:
% 0 for +/+, 1 for +/-, and 2 for -/-. runo(:,nn(2))==x then returns a column of
% 1's and O's that selects the rows when the index is x.
disp(['Human [+/-], cone = ',num2str(CONC),', AUCSlowly = ',num2str(res)]);
res=mean(runo(runo(:,nn(2))==0,nn(l))); rres=[rres,res]; % Results for full GST +/+
disp(['Human [+/+], cone = ',num2str(CONC),', AUCSlowly = ',num2str(res)]);
huml=[huml; rres];
end
savehuml @file='FigG-l_human.csv' @format=ascii @separator=comma
save @File="FigG-l.mat"
% Then rat simulations
use rat_set_C
TEND=2*7*24; ratr=[]; BW=0.229; TCHNG=6; TCHNG2=120;
for CONC=ccs
start @NoCallBack
ratr=[ratr; [CONC,W AVGAUC S]];
disp(['Rat, cone = ',num2str(CONC),', AUCSlowly = ',num2str(WAVGAUCS),'.']);
end
save ratr @file='FigG-l_rat.csv' @format=ascii @separator=comma
plot(ccs,ratr(:,2),ccs,huml(:,2),ccs,huml(:,3),ccs,huml(:,4),'FigG_l.aps')
H-44 DRAFT - DO NOT CITE OR QUOTE
------- |