DRAFT - DO NOT CITE OR QUOTE
SERA
EPA/63 5/R-08/013C
www. epa. gov/ir is
TOXICOLOGICAL REVIEW
OF
ACRYLONITRILE
(CAS No. 107-13-1)
In Support of Summary Information on the
Integrated Risk Information System (IRIS)
January, 2010
NOTICE
This document is an Interagency Science Consultation 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.
11
DRAFT - DO NOT CITE OR QUOTE
-------
CONTENTS —TOXICOLOGICAL REVIEW OF ACRYLONITRILE
(CAS No. 107-13-1)
LIST OF TABLES vii
LIST OF FIGURES xvi
LIST OF ABBREVIATIONS AND ACRONYMS xviii
FOREWORD xxii
AUTHORS, CONTRIBUTORS, AND REVIEWERS xxiii
1. INTRODUCTION 1
2. CHEMICAL AM) PHYSICAL INFORMATION 3
3. TOXICOKINETICS 5
3.1. ABSORPTION 5
3.1.1. Studies in Humans 5
3.1.2. Studies in Animals 5
3.2. DISTRIBUTION 7
3.3. METABOLISM 14
3.3.1. Oxidation of AN to CEO 15
3.3.2. Interaction of AN with GSH 23
3.3.3. Covalent Binding of AN and Its Metabolites to Subcellular Macromolecules 28
3.4. ELIMINATION 34
3.4.1. Studies in Humans 34
3.4.2. Studies in Animals 34
3.4.2.1. Exhalation 34
3.4.2.2. Fecal Excretion 35
3.4.2.3. Urinary Excretion 35
3.5. PHYSIOLOGICALLY BASED TOXICOKINETIC MODELS 40
4. HAZARD IDENTIFICATION 53
4.1. STUDIES IN HUMANS—EPIDEMIOLOGY AND CASE REPORTS 53
4.1.1. Oral Exposure 53
4.1.2. Inhalation Exposure 53
4.1.2.1. Acute Exposure 53
4.1.2.2. Chronic Exposure 55
4.1.3. Dermal Exposure 115
4.1.3.1. Acute Exposure 115
4.1.3.2. Chronic Exposure 117
4.1.4. Ocular Exposure 117
4.2. SUBCHRONIC AND CHRONIC STUDIES AND CANCER BIOASSAYS IN
ANIMALS—ORAL AND INHALATION 120
4.2.1. Oral Exposure 121
4.2.1.1. Subchronic Studies 121
4.2.1.2. Chronic Studies 127
iii DRAFT - DO NOT CITE OR QUOTE
-------
4.2.2. Inhalation Exposure 156
4.2.2.1. Subchronic Studies 156
4.2.2.2. Chronic Studies 157
4.3. REPRODUCTIVE/DEVELOPMENTAL STUDIES—ORAL AND INHALATION . 164
4.3.1. Studies in Humans 164
4.3.2. Studies in Animals 167
4.3.2.1. Oral Studies 167
4.3.2.2. Inhalation Exposure 175
4.3.2.3. Intraperitoneal Administration 177
4.3.3. In Vitro Studies 179
4.4. OTHER DURATION- OR ENDPOINT-SPECIFIC STUDIES 180
4.4.1. Acute Toxicity Data 180
4.4.1.1. Effects of AN on the GI Tract 184
4.4.1.2. Effects of AN on the Kidney 186
4.4.1.3. Effects of AN on the Adrenal Gland 186
4.4.1.4. Neurotoxic Effects of AN 187
4.4.1.5. Effects on Hearing 188
4.4.2. Immunotoxicity of AN 191
4.5. MECHANISTIC DATA AND OTHER STUDIES IN SUPPORT OF THE MODE OF
ACTION 196
4.5.1. Mode-of-Action Studies 196
4.5.1.1. Noncancer Endpoints 196
4.5.1.2. Cancer Effects 210
4.5.2. Genotoxicity Studies 223
4.5.2.1. Studies in Humans 223
4.5.2.2. In Vivo Tests in Mammals 227
4.5.2.3. Short-term Tests: Bacteria, Fungi, Drosophila, Others 230
4.5.2.4. Mammalian Cell Short-term Tests 235
4.6. SYNTHESIS AND EVALUATION OF MAJOR NONCANCER EFFECTS AND
MODE OF ACTION—ORAL AND INHALATION 255
4.6.1. Oral 255
4.6.2. Inhalation 259
4.6.3. Mode-of-Action Information 263
4.6.3.1. GI Effects 263
4.6.3.2. Neurotoxicity 264
4.6.3.3. Reproductive/Developmental Effects 266
4.6.3.4. Hematological Effects 267
4.6.3.5. Immunotoxicity 268
4.6.3.6. Covalent Binding to Sulfhydryl Groups 268
4.7. EVALUATION OF CARCINOGENICITY—SYNTHESIS OF HUMAN, ANIMAL,
AND OTHER SUPPORTING EVIDENCE, CONCLUSIONS ABOUT HUMAN
CARCINOGENICITY, AND LIKELY MODE OF ACTION 269
4.7.1. Summary of Overall Weight of Evidence 269
4.7.2. Synthesis of Human, Animal, and Other Supporting Evidence 269
4.7.3. Mode-of-Action Information 272
4.7.3.1. Hypothesized Mode of Action: Direct Mutagenic Mode-of-Action 273
4.7.3.2. Experimental Support for the Hypothesized Mode of Action 273
4.7.3.3. Other Possible Modes of Action 284
4.7.3.4. Possible Modes of Action for Forestomach Tumors 292
iv DRAFT - DO NOT CITE OR QUOTE
-------
4.7.3.5. Possible Modes of Action for Hepatomas 293
4.7.3.6. Possible Modes of Action for Tumors of Intestines, Tongue, Zymbal Gland,
and Harderian Gland 294
4.7.3.7. Possible Modes of Action for Lung Cancer 294
4.8. SUSCEPTIBLE POPULATIONS AND LIFE STAGES 294
4.8.1. Possible Childhood Susceptibility 294
4.8.2. Possible Geriatric Susceptibility 297
4.8.3. Possible Gender Differences 298
4.8.4. Genetic Polymorphisms 301
4.8.4.1. CYP450 301
4.8.4.2. Glutathione S-transferases 302
4.8.4.3. EH 302
5. DOSE-RESPONSE ASSESSMENTS 304
5.1. ORAL REFERENCE DOSE (RID) 304
5.1.1. Choice of Principal Study and Critical Effect 304
5.1.2. Methods of Analysis—Including Use of PBTK Modeling 308
5.1.2.1. PBTK Modeling 308
5.1.2.2. BMD Modeling 309
5.1.3. RfD Derivation—Including Application of Uncertainty Factors (UFs) 314
5.1.4. Oral Data Array for Noncancer Endpoints 315
5.1.5. Previous RfD Assessment 318
5.2. INHALATION RfC 318
5.2.1. Choice of Principal Study and Critical Effect 318
5.2.2. Methods of Analysis 321
5.2.3. RfC Derivation—Including Application of UFs 322
5.2.4. Inhalation Data Array for Noncancer Endpoints 323
5.2.5. Previous Inhalation Assessment 325
5.3. UNCERTAINTIES IN THE ORAL RfD AND INHALATION RfC 325
5.4. CANCER ASSESSMENT 327
5.4.1. Choice of Study/Data—with Rationale and Justification 327
5.4.2. Dose-Response Data 328
5.4.2.1. Human Occupational Data 328
5.4.2.2. Rat Oral Data 329
5.4.2.3. Rat Inhalation Data 333
5.4.3. Dose-Response Modeling 334
5.4.3.1. Human Occupational Data 334
5.4.3.2. Rat Oral Data 335
5.4.3.3. Rat Inhalation Data 342
5.4.4. Cancer Risk Values 345
5.4.4.1. Oral CSFs 345
5.4.4.2. Inhalation Unit Risk 347
5.4.4.3. Quantitative Analysis of Early Life Exposure Scenarios 349
5.4.5. Uncertainties in Cancer Risk Values 359
5.4.5.1. Oral Cancer Assessment 359
5.4.5.2. Inhalation Cancer Assessment 365
DRAFT - DO NOT CITE OR QUOTE
-------
6. MAJOR CONCLUSIONS IN THE CHARACTERIZATION OF HAZARD AND DOSE
RESPONSE 369
6.1. HUMAN HAZARD POTENTIAL 369
6.2. DOSE-RESPONSE 371
6.2.1. Oral RID 371
6.2.2. Inhalation RfC 372
6.2.3. Oral CSF 374
6.2.4. Cancer Inhalation Unit Risk 375
7. REFERENCES 376
APPENDIX A. SUMMARY OF EXTERNAL PEER REVIEW AND PUBLIC
COMMENTS AND DISPOSITION A-l
APPENDIX B. BENCHMARK DOSE CALCULATIONS B-l
APPENDIX B-l. NONCANCER ORAL DOSE-RESPONSE ASSESSMENT (RfD):
BENCHMARK DOSE MODELING RESULTS EMPLOYING THE INCIDENCE OF
FORESTOMACH LESIONS (HYPERPLASIA AND HYPERKERATOSIS) IN
MALE AND FEMALE SPRAGUE-DAWLEY RATS, F344 RATS, AND B6C3Fi
MICE CHRONICALLY EXPOSED ORALLY TO AN FOR 2 YEARS B-l
APPENDIX B-2. NONCANCER INHALATION DOSE-RESPONSE ASSESSMENT
(RfC): BMD MODELING RESULTS EMPLOYING THE INCIDENCE DATA FOR
NONNEOPLASTIC NASAL LESIONS IN RATS EXPOSED TO AN BY
INHALATION FOR 2 YEARS (TABLES B-6 THROUGH B-9) B-l 16
APPENDIX B-3. CANCER ORAL DOSE-RESPONSE ASSESSMENT: BMD DOSE
MODELING RESULTS FOR TUMOR INCIDENCE DATA FROM RATS
CHRONICALLY EXPOSED TO AN IN DRINKING WATER B-l23
APPENDIX B-4. CANCER INHALATION DOSE-RESPONSE ASSESSMENT: BMD
MODELING RESULTS FOR TUMOR INCIDENCE DATA FROM RATS
CHRONICALLY EXPOSED TO AN VIA INHALATION B-246
APPENDIX B-5. ANALYSIS TO ASSESS COMBINING TUMOR INCIDENCE DATA
FROM TWO CANCER BIO AS SAYS EMPLOYING SPRAGUE-DAWLEY
RATS B-283
APPENDIX B-6. ESTIMATION OF COMPOSITE CANCER RISK FROM EXPOSURE
TO AN BY COMBINING RISK ESTIMATES ACROSS MULTIPLE TUMOR
SITES B-289
APPENDIX B-7. STATISTICAL ANALYSIS OF BLAIR ET AL. (1998) B-293
APPENDIX C. PBPK MODEL DESCRIPTIONS AND SOURCE CODE C-l
APPENDIX D. UNCERTAINTIES ASSOCIATED WITH CHEMICAL-SPECIFIC
PARAMETERS EMPLOYED IN THE PBPK MODEL FOR AN DOSIMETRY
IN HUMANS D-l
vi
DRAFT - DO NOT CITE OR QUOTE
-------
LIST OF TABLES
3-1. Recovery of radioactivity from male Sprague-Dawley rats exposed to 5 or 100 ppm
[1-14C]-AN for 6 hours via inhalation 6
3-2. Recovery of radioactivity after a single gavage dose of 0.1 or 10 mg/kg [1-14C]-AN to
male Sprague-Dawley rats 6
3-3. Percentage recovery of radioactivity in tissues of male Wistar rats following a single
oral dose of radiolabeled AN 9
3-4. Apparent kinetic parameters of CEO formation from AN in B6C3F1 mice, F344 rats,
and humans 20
3-5. Apparent kinetic parameters for glutathione conjugation of AN and CEO at pH 6.5 26
3-6. Glutathione conjugation of AN and CEO with or without microsomal or cytosolic GST
from rat, mouse, and human liver preparations 26
3-7. Urinary excretion of thioethers derived from AN 38
4-1. Clinical signs in 144 subjects accidentally exposed to AN 54
4-2. Distribution of select incidence and mortalities among wage workers and all workers
at an AN plant 56
4-3. Distribution of select incidence and mortalities among wage workers and all workers
at an AN plant 59
4-4. Distribution of select incidence and mortalities among wage and salary workers at
an AN plant 61
4-5. Distribution of select mortalities among exposed workers in two AN plants 63
4-6. Distribution of select mortalities among exposed workers 64
4-7. Crude and adjusted hazard ratio estimates for 100 ppm-year increase in cumulative
exposure 65
4-8. Hazard ratio estimated for select cancer mortality by lagged cumulative exposure for
100 ppm-year increase in cumulative exposure 66
4-9. Distribution of select mortalities among AN-exposed and unexposed workers 73
4-10. Lung cancer mortality of AN-exposed workers stratified by cumulative dose
and latency 74
4-11. Distribution of select mortalities among AN-exposed and unexposed workers 80
vii DRAFT - DO NOT CITE OR QUOTE
-------
4-12. Distribution of select mortalities among AN-exposed and unexposed workers 83
4-13. Distribution of mortality among AN-exposed workers 87
4-14. Summary of relative regression analyses for cancer of the bronchus, trachea,
and lung 88
4-15. Distribution of observed lung cancer deaths among AN-exposed workers, using
regional rates for comparison 89
4-16. ORs of lung cancer for AN exposure 93
4-17. Derived all-cancer SMRs for major cohort studies on AN exposure 96
4-18. Derived lung cancer SMRs for major cohort studies on AN exposure 98
4-19. Industrial AN exposure, levels of AN and thiocyanate in urine, and prevalence of
physical signs of adverse effects in workers exposed to AN at six acrylic fiber
factories in Japan 102
4-20. Comparison of reproductive outcomes in wives of exposed and control males at a
chemical fiber pi ant 108
4-21. Comparison of reproductive outcomes between exposed and control females at a
chemical fiber pi ant 108
4-22. Epidemiology studies of noncancer outcomes among cohorts of workers exposed
to AN 117
4-23. Effect on SCV in male Sprague-Dawley rats exposed to AN via gavage for
12 weeks 124
4-24. Incidence of nonneoplastic lesions in Sprague-Dawley rats exposed to AN in
drinking water for 2 years 129
4-25. Selected tumor incidences in response to AN administered to Sprague-Dawley rats
in drinking water for up to 2 years 130
4-26. Histopathologic location of astrocytomas in the CNS of male Sprague-Dawley rats
administered AN in drinking water for 2 years 132
4-27. Incidence of nonneoplastic lesions in Sprague-Dawley rats exposed to AN in
drinking water for 2 years 134
4-28. Selected tumor incidences in Sprague-Dawley rats exposed to AN in drinking water
for up to 2 years 136
viii
DRAFT - DO NOT CITE OR QUOTE
-------
4-29. Incidence of nonneoplastic lesions in Sprague-Dawley rats exposed to AN by gavage
for 20 months 138
4-30. Cumulative incidence of tumors in response to AN administered to Sprague-Dawley
rats by gavage for up to 2 years 139
4-31. Incidences of nontumorous lesions in F344 rats exposed to AN in drinking water for
2 years 143
4-32. Selected tumor incidences in F344 rats exposed to AN in drinking water for 2 years .... 144
4-33. Incidence and severity of nonneoplastic lesions in B6C3Fi mice exposed by gavage
to AN for 2 years 146
4-34. Incidences of selected neoplastic lesions in B6C3Fi mice exposed by gavage to AN
for 2 years 147
4-35. Incidence of tumors in female Sprague-Dawley rats exposed to AN in drinking
water for up to 46 weeks 151
4-36. Summary of chronic oral toxicity studies of AN: noncancer effects in rats
and mice 151
4-37. Summary of chronic oral toxicity studies of AN: cancer effects in rats and mice 153
4-38. Effect on SCV in male Sprague-Dawley rats exposed to AN via inhalation for
24 weeks 156
4-39. Incidence of histopathological lesions of the nasal turbinates in Sprague-Dawley
rats exposed to AN via inhalation for 2 years 158
4-40. Incidence of dose-related noncancerous histopathological lesions in Sprague-Dawley
rats exposed to AN via inhalation for 2 years 159
4-41. Cumulative incidence of tumors in Sprague-Dawley rats exposed to AN via
inhalation for up to 2 years 160
4-42. Comparison of carcinogenic effects of chronic exposure to AN at 60 ppm starting
either in utero or in adulthood, in Sprague-Dawley rats 162
4-43. Incidence of fetal abnormalities among litters of Sprague-Dawley rats following
maternal exposure to AN on GDs 6-15 168
4-44. Morphological alterations in GD 12 fetuses of Sprague-Dawley rats exposed to
100 mg/kg AN on GD 10 169
4-45. Group-specific reproductive indices in three generations of Sprague-Dawley rats
receiving AN in drinking water 172
IX
DRAFT - DO NOT CITE OR QUOTE
-------
4-46. Group-specific pup weights in three generations of Sprague-Dawley rats receiving
AN in drinking water 173
4-47. Incidence of fetal malformations among litters of Sprague-Dawley rats exposed to
AN by inhalation 175
4-48. Effects of AN on organ weight, clinical chemistry, and biochemical parameters
when administered to male Wistar rats via inhalation 182
"3
4-49. Time course of the effect of AN administration on [ H]-thymidine uptake into
mouse splenocytes under the influence of different mitogens in vitro 193
4-50. Summary of immunotoxicity studies of AN 194
4-51. Effect of AN on RBC metabolic intermediates following a single oral dose 196
4-52. Detection of N -(2-oxoethyl)guanine after i.p. administration of 50 mg/kg AN or
CEO to male F344 rats 212
4-53. Summary of studies on the mutagenicity/genotoxicity of AN 239
4-54. Formation of 8-oxodG in DNA from tissues of male Sprague-Dawley and F344 rats
exposed to AN in drinking water for 21 days 251
4-55. Summary of studies on the indirect mutagenicity or genotoxicity of AN 253
4-56. Noncancer effects in animals repeatedly exposed to AN by the oral route 235
4-57. Noncancer effects in human workers and animals repeatedly exposed to AN
by inhalation 259
4-58. 8-OxodG in brain DNA and brain tumor incidence in male F344 rats exposed to
AN in drinking water 287
4-59. 8-OxodG in brain DNA and brain tumor incidence in male Sprague-Dawley rats
exposed to AN in drinking water 288
5-1. Incidences of forestomach lesions (hyperplasia or hyperkeratosis) in Sprague-Dawley and
F344 rats exposed to AN in drinking water for 2 years 305
5-2. Incidences of forestomach lesions (hyperplasia or hyperkeratosis) in
male and female B6C3Fi mice administered AN via gavage for 2 years 306
5-3. Candidate RfDs based on BMD modeling of the incidence of nonneoplastic
forestomach lesions (hyperplasia or hyperkeratosis) in male and female
Sprague-Dawley rats exposed to AN in drinking water for 2 years 310
x
DRAFT - DO NOT CITE OR QUOTE
-------
5-4. Candidate RfDs based on BMD modeling of the incidence of nonneoplastic
forestomach lesions (hyperplasia or hyperkeratosis) in male and female F344 rats
exposed to AN in drinking water for 2 years 311
5-5. Candidate RfDs based on BMD modeling of the incidence of nonneoplastic
forestomach lesions (hyperplasia or hyperkeratosis) in male and female
B6C3Fi mice exposed to AN via gavage for 2 years 312
5-6. Results of dose-response analyses of incidence data for selected nasal lesions in
male and female Sprague-Dawley rats exposed by inhalation to AN for 2 years 321
5-7. Incidence of CNS tumors in Sprague-Dawley and F344 rats exposed to AN in
drinking water for 2 years 329
5-8. Incidence of mammary gland tumors in F344 and Sprague-Dawley rats exposed to
AN in drinking water for 2 years 330
5-9. Tumor incidences Sprague-Dawley and F344 in rats exposed to AN in drinking water
for 2 years 331
5-10. Incidences of intestinal, CNS, Zymbal gland, tongue, and mammary gland tumors
in Sprague-Dawley rats exposed to AN via inhalation for 2 years 333
5-11. Four different dose metrics, two external and two internal, based on doses
employed in studies of Sprague-Dawley and F344 rats exposed to AN in
drinking water for 2 years 335
5-12. BMD modeling results using tumor incidence data from male and female
Sprague-Dawley and F344 rat studies in which animals were exposed to AN in
drinking water for 2 years 336
5-13. Site-specific oral CSFs for AN based on BMD modeling of tumor incidence data in
rats and predicted CEO levels in blood (AUC/24 hours) of rats and humans
assuming episodic exposure to AN 338
5-14. Site-specific oral CSFs for AN based on BMD modeling of tumor incidence data
in rats and predicted AN levels in blood (AUC/24 hours) of rats and humans
assuming episodic exposure to AN 339
5-15. Site-specific oral CSFs for AN based on BMD modeling of tumor incidence
data in rats and BW scaling to the 3/4 power to convert from rat to human
administered doses 340
5-16. Two different dose metrics, one external and one internal, based on administered
air concentrations of AN employed in a 2-year bioassay in Sprague-Dawley rats 341
5-17. BMD modeling results using tumor incidence data from male and female
Sprague-Dawley rats exposed to AN via inhalation for 2 years and CEO
concentration in blood predicted from an EPA-modified PBTK model 342
XI
DRAFT - DO NOT CITE OR QUOTE
-------
5-18. Site-specific IURs for AN based on BMD modeling of tumor incidence data in
Sprague-Dawley rats and PBTK modeling of CEO levels in blood (AUC/24 hours)
of rats and humans 343
5-19. Site-specific IURs for AN based on BMD modeling of tumor incidence data in
Sprague-Dawley rats exposed to AN via inhalation 343
5-20. Estimated human equivalent oral CSFs for composite risk based on tumor
incidence in AN-exposed rats and predicted CEO-AUC levels in blood 345
5-21. Estimated human equivalent composite IURs based on tumor incidence in
AN-exposed rats and predicted CEO-AUC levels in blood 347
5-22. Tumor incidences in Sprague-Dawley rats following inhalation exposure to
60 ppm AN for 104 weeks, starting either in utero (GD 12) or as adults at
13 wks of age 350
5-23. Tumor incidences in female Sprague-Dawley rats exposed to AN in drinking water
for approximately 46 weeks, starting either in utero or at 5 weeks of age 353
5-24. Overall extra risks of multiple tumor incidence, for chronic exposure beginning
either in early-life or in adulthood; based on Sprague-Dawley rats exposed to AN 355
5-25. Application of ADAFs for estimating human cancer risk from 70-year oral or
inhalation exposures to AN from ages 0 to 70 357
5-26. Summary of uncertainty in the AN oral cancer risk assessment 359
5-27. Summary of uncertainty in the AN inhalation cancer risk assessment 364
B-l. Incidences of forestomach lesions (hyperplasia or hyperkeratosis) in Sprague-Dawley and
F344 rats exposed to AN in drinking water for 2 years B-l
B-2. Incidences of forestomach lesions (hyperplasia or hyperkeratosis) in male and female
B6C3Fi mice administered AN via gavage for 2 years B-3
B-3. Summary of the BMD modeling results based on the incidence of forestomach lesions
(hyperplasia or hyperkeratosis) in male and female
Sprague-Dawley rats exposed to AN in drinking water for 2 years B-4
B-4. Summary of the BMD modeling results based on the incidence of forestomach lesions
(hyperplasia or hyperkeratosis) in male and female F344 rats
exposed to AN in drinking water for 2 years B-52
B-5. Summary of the BMD modeling results based on the incidence of forestomach lesions
(hyperplasia or hyperkeratosis) in male and female B6C3Fi mice exposed to AN via
gavage for 2 years B-83
Xll
DRAFT - DO NOT CITE OR QUOTE
-------
B-6. Incidence data for selected nasal lesions in Sprague-Dawley rats exposed by
inhalation to AN for 2 years B-l 16
B-7. A summary of BMDS (version 1.3.2) modeling results based on incidence of
hyperplasia of mucus-secreting cells in male Sprague-Dawley rats exposed to
AN via inhalation for 2 years B-l 17
B-8. A summary of BMDS (version 1.3.2) modeling results based on incidence of
flattening of respiratory epithelium in female Sprague-Dawley rats exposed to
AN via inhalation for 2 years B-l20
B-9. Incidence of forestomach (nonglandular) tumors in Sprague-Dawley rats exposed to
AN in drinking water for 2 years B-l24
B-10. Summary of BMD modeling results based on incidence of forestomach
(nonglandular) tumors in Sprague-Dawley rats exposed to AN in drinking water
for 2 years B-l24
B-l 1. Incidence of CNS tumors in Sprague-Dawley rats exposed to AN in drinking water
for 2 years B-l39
B-12. Summary of BMD modeling results based on incidence of CNS tumors in
Sprague-Dawley rats exposed to AN in drinking water for 2 years B-l39
B-l3. Incidence of Zymbal gland tumors in Sprague-Dawley rats exposed to AN
in drinking water for 2 years B-l54
B-14. Summary of BMD modeling results based on incidence of Zymbal gland
tumors in Sprague-Dawley rats exposed to AN in drinking water for 2 years B-l54
B-l 5. Incidence of tongue tumors in Sprague-Dawley rats exposed to AN in drinking
water for 2 years B-l69
B-16. Summary of BMD modeling results based on incidence of tongue tumors in
Sprague-Dawley rats exposed to AN in drinking water for 2 years B-l69
B-17. Incidence of mammary gland tumors in Sprague-Dawley rats exposed to AN in
drinking water for 2 years B-l84
B-l 8. Summary of BMD modeling results based on incidence of mammary gland
tumors in Sprague-Dawley rats exposed to AN in drinking water for 2 years B-l84
B-19. Incidence of forestomach (nonglandular) tumors in F344 rats exposed to AN in
drinking water for 2 years B-l92
B-20. Summary of BMD modeling results based on incidence of forestomach
(nonglandular) tumors in F344 rats exposed to AN in drinking water for 2 years B-l93
Xlll
DRAFT - DO NOT CITE OR QUOTE
-------
B-21. Incidence of CNS tumors in F344 rats exposed to AN in drinking water for
2 years B-208
B-22. Summary of BMD modeling results based on incidence of CNS tumors in
F344 rats exposed to AN in drinking water for 2 years B-208
B-23. Incidence of Zymbal gland tumors in F344 rats exposed to AN in drinking water
for 2 years B-223
B-24. Summary of BMD modeling results based on incidence of Zymbal gland tumors in
F344 rats exposed to AN in drinking water for 2 years B-223
B-25. Incidence of mammary gland tumors in F344 rats exposed to AN in drinking water
for 2 years B-23 8
B-26. Summary of BMD modeling results based on incidence of mammary gland tumors
in F344 rats exposed to AN in drinking water for 2 years B-238
B-27. Incidence of intestinal tumors in Sprague-Dawley rats exposed to AN in air for
2 years B-247
B-28. Summary of BMD modeling results based on incidence of intestinal tumors in
Sprague-Dawley rats exposed to AN in air for 2 years B-247
B-29. Incidence of CNS tumors in Sprague-Dawley rats exposed to AN in air for
2 years B-253
B-30. Summary of BMD modeling results based on incidence of CNS tumors in
Sprague-Dawley rats exposed to AN in air for 2 years B-253
B-31. Incidence of Zymbal gland tumors in Sprague-Dawley rats exposed to AN in air
for 2 years B-262
B-32. Summary of BMD modeling results based on incidence of Zymbal gland tumors
in Sprague-Dawley rats exposed to AN in air for 2 years B-262
B-33. Incidence of tongue tumors in Sprague-Dawley rats exposed to AN in air for
2 years B-272
B-34. Summary of BMD modeling results based on incidence of tongue tumors in Sprague-
Dawley rats exposed to AN in air for 2 years B-272
B-3 5. Incidence of mammary gland tumors in Sprague-Dawley rats exposed to AN in air
for 2 years B-278
B-36. Summary of BMD modeling results based on incidence of mammary gland tumors
in Sprague-Dawley rats exposed to AN in air for 2 years B-278
xiv
DRAFT - DO NOT CITE OR QUOTE
-------
B-37. Summary of PODs for composite cancer risk associated with episodic oral
exposure to AN, using CEO-AUC levels in blood as dose metric and multiple
tumor incidence data in rats B-291
B-38. Summary of PODs for composite cancer risk associated with inhalation exposure
to AN, based on multiple tumor incidence data in rats and CEO-AUC levels
in blood B-291
B-39. Estimated human oral CSFs for AN based on multiple tumor incidence data in
rats and CEO-AUC levels in blood B-292
B-40. Estimated human IURs for AN based on multiple tumor incidence data in rats and
CEO-AUC levels in blood B-292
C-l. Rat and human PBPK model parameter values C-2
D-l. Tissue:blood PCs D-2
D-2. Impact of method of estimation of PCs on the peak AN and CEO model predictions D-3
D-3. PBPK mass balance predictions for an 8-hour human exposure to 2 ppm AN D-6
D-4. Sensitivity of AN and CEO metrics for continuous inhalation exposure D-9
D-5. Sensitivity of AN and CEO metrics for continuous oral exposure D-10
D-6. Estimated CVs for AN and CEO metrics D-12
D-7. Parameter contributions to overall CVs for inhalation exposure D-13
D-8. Parameter contributions to overall CVs for oral exposure D-l4
xv
DRAFT - DO NOT CITE OR QUOTE
-------
LIST OF FIGURES
2-1. Chemical structure of AN 3
3-1. Scheme for the metabolic transformation of AN 14
3-2. Structure of the PBTK model for AN and CEO 41
3-3. Intravenous exposure, dosimetry, and model fits 46
3-4a. Inhalation exposure, CEO concentrations 47
3-4b. Inhalation exposure, AN concentrations 48
3-5a. Oral exposure, CEO concentrations 49
3-5b. Oral exposure, AN concentrations 50
3-5c. Oral exposure, urinary excretion, and Hb binding 51
5-1. Comparison of LOAELs for noncancer effects across target organs following oral
exposure to AN in animals 316
5-2. Comparison of LOAELs for noncancer effects in human workers following inhalation
exposure to AN 323
5-3. Comparison of tumor responses to inhalation exposure of acrylonitrile, by age at start
of exposure (offspring at GD 12, adults at 13 weeks), sex, and length of exposure, for
Sprague-Dawley rats (Maltoni et al., 1988) 351
5-4. Comparison of composite oral CSFs derived from tumor incidence data in four
different sex/strain/species of rats exposed chronically to AN. For each
sex/strain/species combination, two different dose metrics were employed:
(1) CEO concentration in blood, and (2) human equivalent administered dose 360
5-5. Comparison of Composite IURs derived from: (1) tumor incidence data in male and
female Sprague-Dawley rats exposed chronically to AN, and (2) neurological effects
in humans exposed to AN occupationally. In deriving the animal-based IURs, two
different dose metrics were employed: (1) predicted CEO concentration in blood, and
(2) human equivalent administered AN concentration in air 365
C-l. Human inhalation exposure level vs. internal AN concentration C-5
C-2. Human inhalation exposure level vs. internal CEO concentrations C-6
C-3. Human oral exposure level vs. internal AN concentration for continuous exposure C-8
C-4. Human oral exposure level vs. internal CEO concentration for continuous exposure C-8
xvi DRAFT - DO NOT CITE OR QUOTE
-------
C-5. Human oral exposure level vs. internal AN concentration for episodic exposure C-9
C-6. Human oral exposure level vs. internal CEO concentration for episodic exposure C-9
xvii
DRAFT - DO NOT CITE OR QUOTE
-------
LIST OF ABBREVIATIONS AND ACRONYMS
ABS
AN butadiene styrene
ABT
1 -aminobenzotriazole
ADAF
age-dependent adjustment factor
ADUR
AN-derived undialyzable radioactivity
AIC
Akaike's Information Criterion
ALT
alanine aminotransferase
AMAP
amplitude of the motor action potential
AN
acrylonitrile
ASAP
amplitude of the sensory action potential
AST
aspartate aminotransferase
ATP
adenosine triphosphate
ATPase
adenosine triphosphatase
ATSDR
Agency for Toxic Substances and Disease Registry
AUC
area under the curve
BAL
broncho alveolar lavage
BMC
benchmark concentration
BMCL
95% lower bound of the BMC
BMD
benchmark dose
BMDL
95% lower confidence limit of benchmark dose
BMDS
benchmark dose software
BMR
benchmark response
BrdU
bromodeoxyuridine
BSO
buthionine sulfoximine
BUN
blood urea nitrogen
BW
body weight
CA
chromosomal aberration
CAIII
carbonic anhydrase III
CAP
compound action potential
CASRN
Chemical Abstracts Service Registry Number
CEMA
2-cyanoethyl mercapturic acid
CEO
2-cyanoethylene oxide
CEVal
N-(2-cyanoethyl)v aline
CF
correction factor
CHL
Chinese hamster lung
CHO
Chinese hamster ovary
CI
confidence interval
CNS
central nervous system
con-A
concanavalin-A
CSF
cancer slope factor
CV
coefficient of variation
CYP450
cytochrome P450
DEM
diethylmaleate
DEX
dexamethasone
dGMP
deoxyguanosine-5'-monophosphate
DMSO
dimethyl sulfoxide
DNA
deoxyribonucleic acid
xviii
DRAFT - DO NOT CITE OR QUOTE
-------
DPOAE
distortion product otoacoustic emission
DTH
delayed-type hypersensitivity
EC
exposure concentration
ECG
el ectrocardi ogram
EH
epoxide hydrolase
ERK
extracellular signal-regulated kinase
EROD
ethoxyresorufin-O-deethylase
FISH
fluorescence in situ hybridization
FSH
follicle stimulating hormone
GAPDH
glyceraldehyde-3-phosphate dehydrogenase
GD
gestation day
GEI
gastric erosion severity index
GI
gastrointestinal
GJIC
gap junction intercellular communications
GSH
reduced glutathione
GSSG
glutathione disulfide
GST
glutathione-S-transferase
GSTM
GST of the |i subclass
y-gtp
y-glutamyl transpeptidase
Hb
hemoglobin
HbCO
carboxyhemoglobin
HbO
oxyhemoglobin
HEC
human equivalent concentration
HEP
human equivalent dose
HMPA
hexamethylphosphoramide
HPLC
high performance liquid chromatography
hprt
hypoxanthine guanine phosphoribosyl transferase
IC50
median inhibitory concentration
Ig
immunoglobulin
i.p.
intraperitoneal
IPCS
International Programme on Chemical Safety
IRIS
Integrated Risk Information System
IUR
inhalation unit risk
i.v.
intravenous
KCN
potassium cyanide
LC50
median lethal concentration
LD50
median lethal dose
LDH
lactate dehydrogenase
LEC
95% lower bound of exposure concentration
LH
luteinizing hormone
LOAEL
lowest-observed-adverse-effect level
LPS
lipopolysaccharide
MCMC
Markov chain Monte Carlo
MCV
motor conduction velocity
MDA
malondialdehyde
mill
microsomal epoxide hydrolase
MEK
mitogen-activated/ERK-activating kinase
MEL
melatonin
MetHb
methemoglobin
XIX
DRAFT - DO NOT CITE OR QUOTE
-------
MN
micronucleus, micronuclei
MNU
methylnitrosourea
MP
microsomal protein
4MP
4-methylpyrazole
NAC
N-acetylcysteine
NADP(H)
nicotinamide adenine dinucleotide phosphate (reduced)
NAS
National Academy of Sciences
NCI
National Cancer Institute
NCTB
Neurobehavioral Core Test Battery
NHA
normal human astrocyte
NIHL
noise-induced hearing loss
NIOSH
National Institute for Occupational Safety and Health
NMR
nuclear magnetic resonance
NOAEL
no-ob served-adverse-effect level
NRC
National Research Council
NTP
National Toxicology Program
OBN
octave band of noise
OHC
outer hair cell
OR
odds ratio
OTC
L-2-oxothiazolidine-4-carboxylic acid
8-oxodG
8-oxodeoxyguanosine, also referred to as 8-oxo-7,8-dihydro-
2'deoxyguanosine and 8-hydroxy-2'-deoxyguanosine
PB
phenobarbital
PBN
phenyl -N-terti ary-butylnitrone
PBPK
physiologically based pharmacokinetic
PBTK
physiologically based toxicokinetic
PC
partition coefficient
PCR
polymerase chain reaction
PH
phorone
PHA
phytohemagglutinin
PK
protein kinase
PMA
phorbol 12-myristate 13-acetate
PND
postnatal day
POD
point of departure
PRL
prolactine
RBC
red blood cell
RfC
reference concentration
RfD
reference dose
RLC
rat liver cell
RNA
ribonucleic acid
ROS
reactive oxygen species
RR
relative risk
SAED
serially additive expected dose
s.c.
subcutaneous
SCE
sister chromatid exchange
SCV
sensory conduction velocity
SD
standard deviation
SDH
sorbitol dehydrogenase
SEER
Surveillance, Epidemiology and End Results
XX
DRAFT - DO NOT CITE OR QUOTE
-------
SG
USEPA Supplemental Guidance for Assessing Susceptibility from Early-
Life Exposure to Carcinogens
SGPT
serum glutamate pyruvate transaminase
SHE
Syrian hamster embryo
SIR
standardized incidence ratio
SMR
standardized mortality ratio
SOD
superoxide dismutase
SRBC
sheep red blood cell
STS
sodium thiosulfate
TAU
taurine
TEARS
thiobarbituric acid-reactive substances
TCPO
1,1,1 -trichloropropane-2,3 -oxide
6-TG
6-thioguanine
TNF-a
tumor necrosis factor
TPO
3,3,3-trichloropropylene oxide
TRX
trolox
TWA
time-weighted average
UCL
upper confidence limit
UDS
unscheduled DNA synthesis
UF
uncertainty factor
U.S. EPA
U.S. Environmental Protection Agency
v/v
volume/volume
WBC
white blood cell
WHO
World Health Organization
WT
wild-type
XXI
DRAFT - DO NOT CITE OR QUOTE
-------
FOREWORD
The purpose of this Toxicological Review is to provide scientific support and rationale
for the hazard and dose-response assessment in IRIS pertaining to chronic exposure to
acrylonitrile. It is not intended to be a comprehensive treatise on the chemical or toxicological
nature of acrylonitrile.
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).
XXll
DRAFT - DO NOT CITE OR QUOTE
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS
CHEMICAL MANAGER
Diana Wong, Ph.D., DABT
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
AUTHORS
Diana Wong, Ph.D., DABT
Ted Berner, MS
John Fox
Rosemarie Hakim, Ph.D.
Karen Hogan, MS
Amanda Persad, Ph.D.
Leonid Kopylev, Ph.D.
Paul Schlosser, Ph.D.
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
Roglio Tornero-Valez, Ph.D.
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
CONTRACTOR SUPPORT
Lutz W.D. Weber, Ph.D., DABT (First Draft)
George Holdsworth, Ph.D. (First Draft)
Janusz Z. Byczkowski, Ph.D., D.Sc., DABT (First Draft)
Donna Cragle, Ph.D. (First Draft)
Virginia H. Sublet, Ph.D. (First Draft)
Oak Ridge Institute for Science and Education
Oak Ridge Associated Universities
Oak Ridge, TN
Margaret Fransen, Ph.D. (Second Draft)
Michael Lumpkin, Ph.D. (Second Draft)
Jennifer Rhoades, B.S. (Second Draft)
Peter McClure, Ph.D., DABT (Second Draft)
Environmental Science Center
SRC, Inc.
North Syracuse, NY
XXlll
DRAFT - DO NOT CITE OR QUOTE
-------
INTERNAL EPA REVIEWERS
James Allen, Ph.D.
NHEERL
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC
Dave Bayliss, Ph.D.
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
Jerry Blancato, Ph.D.
NERL
U.S. Environmental Protection Agency
Research Triangle Park, NC
Anthony DeAngelo, Ph.D.
NHEERL
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC
Rob Dewoskin, Ph.D.
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC
Lynn Flowers, Ph.D., DABT
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
Kate Guyton, Ph.D., DABT
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
Karl Jensen, Ph.D.
NHEERL
Office of Research and Development
xxiv
DRAFT - DO NOT CITE OR QUOTE
-------
U.S. Environmental Protection Agency
Research Triangle Park, NC
Jennifer Jinot
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
Channa Keshava, Ph.D.
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
John Lipscomb, Ph.D.
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Cincinnati, OH
Robert MacPhail, Ph.D.
NHEERL
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC
Jeff Ross, Ph.D.
NHEERL
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC
Cheryl Scott
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
Lawrence Valcovic, Ph.D.
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
xxv
DRAFT - DO NOT CITE OR QUOTE
-------
John Whalan
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
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 acrylonitrile
(AN). 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 (MOA). 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
"3
deleterious effects during a lifetime. The inhalation RfC (expressed in units of mg/m ) 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 (j,g/m air breathed.
Development of the hazard identification and dose-response assessments for AN has
followed the general guidelines for risk assessment as set forth by the National Research Council
(NRC, 1983). U.S. Environmental Protection Agency (U.S. 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, 1988), Guidelines for Developmental Toxicity Risk Assessment (U.S. EPA, 1991), Interim
1
DRAFT - DO NOT CITE OR QUOTE
-------
Policy for Particle Size and Limit Concentration Issues in Inhalation Toxicity (U.S. EPA,
1994a), Methods for Derivation of Inhalation Reference Concentrations and Application of
Inhalation Dosimetry (U.S. EPA, 1994b), Use of the Benchmark Dose Approach in Health Risk
Assessment (U.S. EPA, 1995), Guidelines for Reproductive Toxicity Risk Assessment (U.S. EPA,
1996), Guidelines for Neurotoxicity Risk Assessment (U.S. EPA, 1998a), 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 Risks 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 June 2009.
2
DRAFT - DO NOT CITE OR QUOTE
-------
2. CHEMICAL AND PHYSICAL INFORMATION
AN (CASRN 107-13-1) is a three-carbon alkene carrying a nitrile substituent group as
part of the terminal carbon atom (carbon 1) (Figure 2-1). Synonyms for the compound include
vinyl cyanide, propenenitrile, and cyanoethylene, and there are a variety of trade names. AN is a
colorless, flammable, and volatile liquid with a weakly pungent onion- or garlic-like odor. Some
physical and chemical properties are shown below (NLM, 2003; IPCS, 2002; ATSDR, 1990).
H
I
hk ^C-
9 %
N
H
Figure 2-1. Chemical structure of AN.
Formula:
Molecular weight:
Melting point:
Boiling point:
Density:
Log K0w-
Log Koc:
Vapor pressure:
Henry's law constant:
Conversion factors:
C3H3N
53.06
-83°C
77.4°C
0.806 g/mL (at 20°C)
-0.92
-0.07
100 mm Hg at 22.8°C
8.8 x 10"5 atm-m3/mol
"3
1 ppm = 2.17 mg/m
"3
1 mg/m = 0.46 ppm
AN is a commercially important chemical with a wide range of uses in the chemical
industry. It is used in the production of acrylic and modacrylic fibers, plastics (AN butadiene
styrene [ABS] and AN-styrene resins), and nitrile rubbers and as an intermediate in the
production of other important chemicals, such as adiponitrile and acrylamide. AN is used in the
plastics industry in the formation of surface coatings and adhesives. It is a chemical intermediate
in the synthesis of antioxidants, pharmaceuticals, and dyes and, in general, for processes
requiring the introduction of a cyanoethyl group into a molecule (NLM, 2003). AN is also used
in clinical practice in the form of dialysis tubing (Mulvihill et al., 1992). AN was used
occasionally as a fumigant insecticide for stored grain. A measure of the commercial importance
of AN may be judged by the amount produced in a given year. The Agency for Toxic
Substances and Disease Registry (ATSDR, [1990]) reports that 1,112,754 metric tons of AN
were produced in the United States in 1987. Production had increased to 1,455,735 metric tons
DRAFT - DO NOT CITE OR QUOTE
-------
by 1995 (ACS, 2003). Exposure of the general public to AN can potentially occur through
migration of residual monomer in polymeric products via contact with food or water. Exposure
to airborne AN is possible among members of the general population living in the vicinity of
emission sources such as acrylic fiber or chemical manufacturing plants or waste sites (ATSDR,
1990). In addition, smokers are expected to be exposed to AN, which has been detected in
cigarette smoke at levels of 3.2-15 mg per cigarette (IARC, 1999).
4
DRAFT - DO NOT CITE OR QUOTE
-------
3. TOXICOKINETICS
Although only limited information on the toxicokinetics of AN in exposed humans is
available, a substantial body of evidence has accumulated on the absorption, distribution,
metabolism, and excretion of AN in experimental animals. The overall conclusion that can be
drawn is that AN is rapidly and nearly completely absorbed, widely distributed among the
tissues, and biochemically transformed into several discrete metabolic products that are excreted
in the urine and, to a much lesser extent, in feces and expired air. Characterization of the
excretory products of AN, with support from in vitro and in vivo kinetic data on its
biotransformation, has contributed to a picture of how AN is rearranged into a series of
metabolic products. Two primary processes appear to be involved: (1) interaction with reduced
glutathione (GSH) and (2) cytochrome P450 (CYP450) 2E1-mediated formation of
2-cyanoethylene oxide (CEO). Each product of these processes can undergo further metabolic
transformations. For example, CEO can be hydrolyzed to cyanide with further transformation to
thiocyanate. Alternatively, CEO can itself interact with GSH, resulting in the formation of a
number of metabolites, several of which have been identified in the urine of experimental
animals.
Important data elements that have contributed to the EPA's current understanding of the
toxicokinetics of AN are summarized in the following sections.
3.1. ABSORPTION
3.1.1. Studies in Humans
There are few data on the absorption of AN in humans. However, Jakubowski et al.
(1987) provided some information on the topic by administering AN via inhalation to six male
"3
volunteers for 8 hours at concentrations of either 5 or 10 mg/m from a chamber. Lung
ventilation and retention of AN in the lungs of these individuals were measured by determining
the concentrations of AN in the inhaled and expired air. A respiratory retention of 52% was
estimated, based on 90 minutes to 8 hours of observation.
3.1.2. Studies in Animals
Most of the available information on the absorption of AN has come from studies in
experimental animals. Young et al. (1977) carried out a series of experiments in which
[1-14C]-AN was administered to male Sprague-Dawley rats via the oral, inhalation, or
intravenous (i.v.) route. Semiquantitative evidence of extensive absorption of AN has come
from an inhalation experiment of Young et al. (1977) in which four rats/group were exposed
nose only to 5 or 100 ppm AN vapor for 6 hours. Following the exposure, the animals were kept
in metabolism cages and excreta were collected for 220 hours. Total recovered doses estimated
5
DRAFT - DO NOT CITE OR QUOTE
-------
from total recovered radioactivity for the low and high exposure levels were 0.7 and 10.2 mg/kg
[1-14C]-AN, respectively. Substantial amounts of the recovered dose were found in urine, feces,
and tissues, which indicated that AN has the capacity to be absorbed via the inhalation route
(Table 3-1).
Table 3-1. Recovery of radioactivity from male Sprague-Dawley rats
exposed to 5 or 100 ppm [1-14C]-AN for 6 hours via inhalation
Site of recovery
5 ppm
100 ppm
Percentage of recovered dose (mean ± standard deviation [SD])
Urine
68.50 ±9.38
82.17 ±4.21
Feces
3.94 ±0.97
3.15 ± 0.82
14co2
6.07 ± 1.58
2.60 ±0.83
Body
18.53 ±4.68
11.24 ±2.85
Cage wash
2.95 ±3.95
0.85 ±0.58
Total dose (mg/kg)
0.7
10.2
Source: Young et al. (1977).
In the gavage experiments of Young et al. (1977), animals were kept in metabolic cages
after a single dose of either 0.1 or 10 mg/kg [1-14C]-AN (vehicle not stated), and the fate of the
radiolabel was monitored for 72 hours. As shown in Table 3-2, only about 5% of the
administered radiolabel was recovered in the feces after 72 hours. By contrast, most of the
administered radiolabel was recovered in the urine, carcass, and skin, with smaller fractions of
the administered radiolabel expired as 14CC>2. These data suggest that at least 95% of
administered AN was absorbed.
Table 3-2. Recovery of radioactivity after a single gavage dose of 0.1 or
10 mg/kg [1-14C]-AN to male Sprague-Dawley rats
Recovery site
0.1 mg/kg
10 mg/kg
Percent of dose
Urine
34.22 ±6.26
66.68 ± 10.6
Feces
5.36 ± 1.43
5.22 ±1.17
Expired C02
4.56 ± 1.82
3.93 ± 1.79
Carcass
24.24 ±5.02
16.04 ± 1.87
Skin
12.78 ±1.17
10.57 ±4.55
Total recovery
81.2
102.4
Source: Young et al. (1977).
Ahmed et al. (1983) administered a single oral dose of 46.5 mg/kg AN to male Sprague-
Dawley rats in distilled water, using 50 jj,Ci/kg of either [2,3-14C]- or [1-14C]-AN as a tracer.
6
DRAFT - DO NOT CITE OR QUOTE
-------
Only 8-10% of administered radioactivity from the [2,3-l4C]-AN dose had been excreted in
feces 72 hours after dosing, compared with 2% from orally administered [1 -l4C]-AN during the
same period. Thus, AN was readily absorbed at the gastrointestinal (GI) tract. The time course
of radioactivity released from feces showed a major spike of [2,3-14C]-AN-derived radiolabel
after 72 hours. The delayed release suggests that at least a portion of this released radioactivity
may have resulted from biliary elimination of absorbed AN. The major peak of released
radiolabel from [1-14C]-AN was at 24 hours postexposure, possibly indicative of a different
metabolic fate for the nitrile portion of the molecule.
Kedderis et al. (1993a) administered [2,3-14C]-AN orally to male F344 rats (0.09-
28.8 mg/kg) and male B6C3Fi mice (0.09-10 mg/kg). Three to five percent of the administered
dose was recovered in the feces of rats after 72 hours. Between 2 and 8% of the dose was
recovered in the feces of male B6C3Fi mice. These data indicated the near-complete absorption
of the compound when administered via the oral route.
3.2. DISTRIBUTION
The toxicokinetic experiments of Young et al. (1977) provided data on the deposition of
radiolabel when male Sprague-Dawley rats were exposed to [1-14C]-AN via inhalation, gavage,
or i.v. injection. As shown in Table 3-1, an average of 11.24% of total recovered dose
(10.2 mg/kg [1-14C]-AN) was obtained in the tissues when a group of four rats were exposed to
100 ppm of [1-14C]-AN for 6 hours via inhalation, and excreta were collected for 220 hours.
Another group exposed to 5 ppm [1-14C]-AN had an average of 18.53% of the recovered
radiolabel (0.7 mg/kg [1-14C]-AN) deposited in the tissues. When the fate of radiolabel
administered by gavage was monitored, combining the recoveries of the administered
radioactivity in carcass and skin gave a value of 37% for tissue deposition at the lower AN
concentration (0.1 mg/kg) vs. 27% at the higher concentration (10 mg/kg) (Table 3-2), indicating
possible metabolic saturation at the higher dose. The percentage of expired CO2 was lower for
the high-dose group than the low-dose group (3.93 vs. 4.56%), providing support for possible
metabolic saturation.
Young et al. (1977) examined the distribution of the radiolabel among the major organs
and tissues after oral and i.v. administration of [1-14C]-AN. Radioactivity was detected at a
range of tissues, including lungs, liver, kidneys, stomach, intestines, skeletal muscle, heart,
spleen, brain, thymus, testes, skin, carcass, and blood cells. Of these, the stomach, red blood
cells (RBCs), and skin appeared to be the most important deposition sites for radiolabeled AN or
its metabolites, regardless of the route of administration, dose level, or time. Radioactivity found
in the stomach was highest after either oral or i.v. route and was not due to unabsorbed AN since
similar results were obtained with either i.v. or oral dosing.
In a follow-up time course experiment, Young et al. (1977) administered 10 mg/kg
[1-14C]-AN intravenously via the tail vein to three male Sprague-Dawley rats and examined the
7
DRAFT - DO NOT CITE OR QUOTE
-------
distribution of radioactivity to the stomach and adrenal glands. The level of radioactivity in the
stomach and its contents increased from 5 minutes to 24 hours postinjection. On the other hand,
radioactivity was high in the adrenal gland at 5 minutes but decreased 13-fold in 24 hours.
Radioactivity was observed in the bile of a single bile-duct-cannulated rat, indicating biliary
excretion. Maximal radioactivity in bile was observed after 15 minutes but declined at later time
points.
When both sexes of adult Sprague-Dawley rats (including some pregnant animals) and
two cynomolgus monkeys were exposed orally (26 mg/kg) or intravenously (13 mg/kg) to single
doses of [1-14C]-AN, whole-body autoradiography from 20 minutes after injection primarily
showed radioactivity in the bile, intestinal contents, and urine (Sandberg and Slanina, 1980).
Other organs showing accumulation of isotope in the autoradiogram were blood, liver, kidney,
lung, and adrenal cortex, in which the activity declined slowly during 24 hours. There also was
uptake of the label in the stomach mucosa and hair follicles. Distribution of radioactivity in fetal
tissue following i.v. and oral administration to pregnant rats was uniform and showed a low
uptake compared to maternal tissue. An exception was found in the eye lens, in which the
radioactivity exceeded that of the maternal blood (Sandberg and Slanina, 1980).
Jacob and Ahmed (2003a) used whole-body autoradiography to examine the distribution
of 11.5 mg/kg [2-14C]-AN administered orally or intravenously to male F344 rats 5 minutes and
8, 24, and 48 hours postexposure. Levels of radioactivity per gram of tissue were highest
5 minutes after oral dosing for stomach lumen and mucosa, small intestine lumen and mucosa,
liver, nasal mucosa, spleen, and kidney; other tissues (including the lung, brain, spinal cord,
thyroid, and testis) had peak levels at 8 hours. Covalently bound radioactivity was detected
48 hours later in stomach mucosa, blood, and hair follicles. Five minutes following i.v.
administration, the highest levels of radioactivity per gram of tissue were detected in the lung,
liver, spleen, small intestine lumen, kidney, epididymis, and adrenal gland. At 24 hours, tissues
that showed peak levels included bone marrow, brain, lacrimal gland, and testis. Tissues with
the highest level at 24 hours included the lung, liver, and bone marrow. The levels of covalent
bound radioactivity (nCi/g) were higher 48 hours after i.v. exposure compared with oral
exposure, with bound radioactivity retained in liver, spleen, bone marrow, lung, kidney, and
adipose tissue ranging from 2.5 to 23 times higher following i.v. exposure. The only organs that
retained higher levels of covalently bound radioactivity 48 hours following oral exposure were
the stomach mucosa (3x) and heart blood (5.7x). The total radioactive dose retained in animals
after i.v. and oral exposures were 70 and 38%, respectively. Jacob and Ahmed (2003a)
concluded that the metabolism and distribution of AN is greatly influenced by the portals of
entry, with a higher amount of AN metabolized and excreted following oral exposure compared
with exposure by i.v. injection. Rapid delivery of AN after i.v. treatment resulted in fast
conjugation and/or covalent interaction of the parent compound with biological molecules,
resulting in minimal metabolism and excretion in urine or feces.
8
DRAFT - DO NOT CITE OR QUOTE
-------
Sapota (1982) administered 40 mg/kg AN in saline, containing either 40 j_iCi/kg
[1,2-14C]-AN or [1-14C]-AN to male Wistar rats, either via gavage or intraperitoneally. Tissue
distribution of radioactivity as measured by liquid scintillation counting after intraperitoneal
(i.p.) and oral administration showed that the highest specific radioactivity in tissue (nCi/g) was
found in RBCs, liver, and kidneys. Tissue-wide recovery of the radiolabel from either
[1,2-14C]-AN or [1-14C]-AN at 2, 8, and 24 hours after a single oral dose is shown in Table 3-3.
Statistically significant differences were observed in the distribution of radioactivity from the
two forms of labeled AN in RBCs, plasma, liver, and kidney at 8 and 24 hours after
administration. More rapid loss of tissue radioactivity in the liver, kidneys, and brain was also
observed after oral administration of [1-14C]-AN than [1,2-14C]-AN. These results suggested
different pathways for disposition and biotransformation of the cyano and vinyl moieties of the
AN molecule.
Table 3-3. Percentage recovery of radioactivity in tissues of male Wistar
rats following a single oral dose of radiolabeled AN
Target organ/
tissue
Recovered radioactivity (percent of dose in tissue)
Distribution from [1-14C]-AN
Distribution from [1,2-14C]-AN
2 h
8 h
24 h
2 h
8 h
24 h
RBCs
5.36
4.82a
5.45
5.31
1.2T
6.72
Plasma
2.63
4.00a
0.50a
1.93
1.92a
1.70a
Liver
6.13
1.21a
1.00a
7.00
5.98a
2.6T
Kidney
1.17
0.2T
0.15a
0.82
0.77a
0.30a
Spleen
0.22
0.10
0.08
0.14
0.17
0.10
Lung
0.36
0.25
0.25
0.30
0.27
0.11
Brain
0.25
0.12a
0.09
0.24
0.25a
0.12
Total
16.1
10.8
7.5
15.7
16.6
11.7
*p < 0.05; significant difference between [14CN]-AN and [1,2-14C]-AN at the same time point.
Source: Sapota (1982).
In an in vivo study on the interaction of orally administered 46.5 mg/kg [1-14C]-AN or
5 mg/kg K14CN with rat blood, Farooqui and Ahmed (1982) reported that up to 94% of 14C from
AN in RBCs was covalently bound to cytoplasmic and membrane proteins. On the other hand,
90% of the radioactivity from K14CN in erythrocytes was bound to the heme fraction of
hemoglobin (Hb), indicating that CN liberated from potassium cyanide (KCN) interacted with
heme. In addition, distribution of 14C from erythrocytes of rats treated with [1-14C]-AN showed
that more than 40% of total radioactivity was localized in membrane residue, 20—35% in the
globin fraction, and 11—25% in the heme fraction. In contrast, 70% of 14C from K14CN in red
cells was localized in the heme fraction, 14—25% in globin, and 5—10% in cell membrane. The
study authors concluded that KCN interacted with rat blood mainly through liberation of CIST,
9
DRAFT - DO NOT CITE OR QUOTE
-------
which was bound to heme. Since AN was found to be mainly covalently bound to cell
membranes, AN might cause damage to RBCs by mechanisms other than the release of CN ,
In male Wistar rats that received 40 mg/kg AN (about half the median lethal dose [LD50])
by gavage in water, peak blood levels of AN (2 (j,g/mL) were detected 1.5 hours after dosing
(Shibata et al., 2004). By 2.5 hours after dosing, blood levels of AN had dropped to less than
0.2 [j,g/mL. Ten hours after dosing, AN was still detectable in blood at 50 ng/mL.
Ahmed et al. (1983, 1982) studied the distribution of AN administered to male Sprague-
Dawley rats by gavage with a dose equivalent to one-half the LD50 (46.5 mg/kg). In the first
experiment (Ahmed et al., 1982), the dose additionally contained 50 jaCi/kg [1-14C]-AN. In the
second experiment (Ahmed et al., 1983), both [2,3-14C]-AN (both carbons in the vinyl moiety
were radiolabeled) and [1-14C]-AN were studied in rats administered an oral dose of 46.5 mg/kg.
Radioactivity from both forms of labeled AN or its metabolites initially was sequestered
mainly in the stomach and stomach content, followed by the rest of the GI tract, including small
and large intestines. The GI tract contained the highest levels of radioactivity up to 72 hours
before beginning to decline, suggesting that AN or its metabolites were re-secreted in the
stomach (Ahmed et al., 1983). In addition, radioactivity became widely distributed in all tissues
within 1-6 hours after dosing, with liver, kidney, and blood showing higher radioactivity than
muscle, fat, and bone (Ahmed et al., 1983, 1982). Heart, spleen, brain, and thymus showed
maximum concentrations between 3 and 6 hours. By 24 hours after administration, the levels of
radioactivity found in liver, kidney, and lung began to decline, resulting in 10- to several
100-fold reductions from peak concentrations over the course of the 10-day experiment.
However, in several tissues, the decline of radioactivity was much lower: at 10 days postdosing,
radiolabel concentrations had declined only 2.4-fold in skin, 2.9-fold in blood, 3.8-fold in spleen,
and 4.9-fold in eyes, as compared with peak levels.
Two differently labeled AN preparations were administered to rats to elucidate potential
differences in distribution and metabolism between the cyano and vinyl groups. One important
finding of these studies was that radioactivity in blood was predominantly in the RBCs,
especially for radioactivity from [2,3-14C]-AN. Radioactivity from [1-14C]-AN in plasma was
higher than that from [2,3-14C]-AN. For radioactivity from [1-14C]-AN in RBCs, 40% was
localized in membrane residue, 20-35% in the globin fraction, and 11—25% in the heme fraction.
In contrast, 50% of radioactivity from [2,3-14C]-AN was in the membrane fraction, 45% in the
globin fraction, and only a trace amount in the heme fraction.
In addition, compared to [1-14C]-AN administered to animals, the percentage of covalent
binding of [2,3-14C]-AN to proteins was significantly higher even 72 hours after dosing.
Subcellular distribution of radioactivity from [2,3-14C]-AN was also different from that derived
from [1-14C]-AN. For [2,3-14C]-AN, the cytosol fraction attained the lowest covalent protein
binding in tissues. The percentage of covalently bound radioactivity in tissues relative to the
total increased four- to fivefold over that of the 1-hour level. At 72 hours after administration,
10
DRAFT - DO NOT CITE OR QUOTE
-------
the highest bound radioactivity was in the mitochondrial fractions of kidney, spleen, lung, and
heart. However, in liver, the microsomal fraction contained the highest radioactivity (Ahmed et
al., 1983).
For [1-14C]-AN, covalent binding to macromolecules in tissues remained unchanged over
time. Cytosol contained the highest levels of total radioactivity in the six tissues (liver, kidney,
spleen, brain, lung, and heart) selected for the study of subcellular distribution. Twenty to 40%
of total radioactivity was bound to nuclear, mitochondrial, or microsomal fractions. Only 6-14%
of total radioactivity was bound to cytosol over 6 hours (Ahmed et al., 1982).
Farooqui and Ahmed (1983a) demonstrated the irreversible binding of radiolabel from
[2,3-14C]-AN to proteins and nucleic acids in vivo. AN was administered by gavage as a bolus
dose of 46.5 mg/kg (one-half LD50) in distilled water to male Sprague-Dawley rats (3-4/group).
Proteins were extracted by chloroform-isoamyl alcohol-phenol, and ribonucleic acid (RNA) and
deoxyribonucleic acid (DNA) were isolated by hydroxyapatite chromatography. Protein binding
at 1 hour after dosing was highest in spleen and stomach, followed by liver, brain, and kidney.
Protein binding in spleen and stomach declined after 1 hour, whereas binding in liver, kidney,
and brain increased. At 6 hours after dosing, protein binding fell to lower values in spleen and
stomach but increased in other tissues. Binding plateaued in all tissues between 6 and 48 hours,
with levels in spleen > liver > stomach > kidney > brain. Binding to RNA was highest in liver,
stomach, and brain, with liver attaining a maximum by 6 hours and stomach and brain by
24 hours. The subsequent decline until 48 hours was also slow. DNA binding did not reach a
maximum until 24 hours after dosing, with levels in brain > stomach > liver. Again, the decline
of DNA-bound radioactivity during the following 24 hours was slow. Binding of AN to DNA in
brain, stomach, and liver was 56, 45, and 5 [j,mol AN per mol DNA, respectively, at 24 hours.
The study authors also calculated a covalent binding index for DNA, defined as the ratio of
([j,mol AN bound per mol DNA) to (mmol AN applied per kg body weight [BW]). The values
for brain, stomach, and liver were 65, 52, and 6, respectively.
Silver et al. (1987) examined the distribution of 100 mg [1-14C]-AN in female Sprague-
Dawley rats after i.v. injection to investigate why this procedure induced acute hemorrhagic
necrosis of the adrenal gland 2 hours after administration and why this damage was more
prominent with i.v. injection than with oral administration. Total radioactivity was found to be
highest in the blood, liver, kidney, duodenum, and adrenals 15-90 minutes following i.v.
injection of the radiolabeled compound. (This result was largely in agreement with the whole-
body autoradiographic findings of Sandberg and Slanina [1980] in rats and monkeys.) Total
radiolabel in blood increased over this time period, whereas the total radiolabel in other organs
remained constant or decreased with time. The level of covalently bound radiolabel in the
adrenals was lower than that observed in blood, liver, kidney, forestomach, and glandular
stomach.
11
DRAFT - DO NOT CITE OR QUOTE
-------
Silver et al. (1987) also administered the same dose of [1-14C]-AN in water by gavage to
female Sprague-Dawley rats (four/group) and investigated the distribution of radiolabel up to
24 hours after dosing. Total radioactivity was highest during the first 8 hours after dosing in the
blood, GI tract, liver, and kidney. At 24 hours after dosing, the highest total radioactivity was
found in the blood, forestomach, and glandular stomach. The highest level of covalently bound
radioactivity was found in these same tissues during the first 8 hours after dosing and remained
highest in the blood and forestomach at 24 hours after dosing. The study authors concluded that
their observations would not support a role of covalent binding in the hemorrhagic effect of AN
on the adrenals. Rather, the initial high concentrations of radiolabel from AN might play a role
in the action of AN on the adrenal gland. However, it should be noted that in both studies, the
cyano carbon, not the vinyl carbons, was labeled.
The rapidity with which AN was biotransformed and distributed to the brain as its
epoxide metabolite, CEO, was reported in two studies (Kedderis et al., 1993b; Roberts et al.,
1991). Roberts et al. (1991) administered 4 mg/kg AN to male F344 rats and B6C3Fi mice
(three/group) and measured CEO levels in blood at 0.5, 1, 4, or 24 hours after dosing. Higher
levels of CEO were found in rat blood than in mouse blood. In addition, CEO was cleared from
mouse blood in 4 hours, but was cleared in rat blood in 24 hours. The dose dependence of CEO
concentrations in blood was also evaluated. Blood CEO was measured 0.5 hours after oral
dosing in F344 rats given either 0, 1, 4, 10, or 30 mg/kg AN and in B6C3Fi mice given either 0,
1, 4, 8, or 10 mg/kg AN. Blood CEO concentrations increased with dose in rats and mice but at
higher concentrations in rats at the same doses.
In the first experiment by Kedderis et al. (1993b), three male F344 rats and three male
B6C3Fi mice were administered 10 mg/kg AN in water by gavage. The rats were sacrificed
10 minutes after dosing, while the mice were sacrificed 5 minutes after dosing. CEO
concentrations from blood and brains of rats and mice were measured. Higher CEO
concentrations were found in the blood and brains of rats than in mice (13% higher in blood and
23% higher in brain). In addition, CEO concentration in rat blood 10 minutes after oral
administration was about twice the concentration previously reported by Roberts et al. (1991) at
30 minutes after oral dosing of 4 mg/kg AN to three male F344 rats. On the other hand, CEO
concentration in mice 5 minutes after oral administration was about 10 times higher than that
reported at 30 minutes after oral dosing of 4 mg/kg AN to three male B6C3Fi mice (Roberts et
al., 1991). These results suggested that CEO was rapidly cleared in both rats and mice and that
the clearance of CEO in mice was more rapid than in rats.
Kedderis et al. (1993b) also administered 3 mg/kg [2,3-14C]-CEO orally to F344 rats and
B6C3Fi mice to determine the tissue distribution of radioactivity from labeled CEO after 2 and
24 hours. Radioactivity from labeled CEO was widely distributed in major organs of rats and
mice 2 hours after administration, with the highest level of radioactivity found in the stomach
and intestines of rats and mice. However, radioactivity detected in the stomach and intestines of
12
DRAFT - DO NOT CITE OR QUOTE
-------
mice was only about 15 and 40%, respectively, of that detected in rats, suggesting that mice
absorbed CEO more rapidly than rats (Kedderis et al., 1993b). By 24 hours, radioactivity
decreased by 71-90% in all tissues, including the brain, liver, and lung. Stomach and intestines
continued to retain the highest level of radioactivity, probably due to covalent binding of CEO to
macromolecules in these organs.
Burka et al. (1994) monitored the tissue distribution of radiolabel derived from
[2-14C]-AN after oral dosing at 0.87 mmol/kg (46 mg/kg) to untreated, phenobarbital
(PB)-pretreated, or SKF 525A pretreated male F344 rats (three/group). PB induces a number of
CYP450 isozymes, including CYP2B1 and CYP2B2, whereas SKF 525A is a general inhibitor
of CYP450. After 24 hours, about 10% of the administered dose in untreated rats was present in
the blood with a further 4% sequestered in the tissues. In PB-pretreated rats, a 40% increase in
AN-derived radioactivity was found in both the liver and glandular stomach when compared
with rats treated with AN alone, with no changes in other tissues. However, AN-derived
radioactivity in most tissues of SKF 525A pretreated rats were up to 278%) higher, suggesting the
involvement of CYP450 metabolism in the disposition of AN and/or its metabolites. (AN reacts
more rapidly with tissue nucleophiles than CEO; hence, decreasing its oxidative metabolism to
CEO would increase tissue binding of radiolabel.) Because PB pretreatment had little effect on
tissue distribution, the isoforms of CYP450 induced by PB are probably not the ones involved in
the metabolism of AN. It is known that AN is metabolized by CYP2E1, not CYP2B1 or
CYP2B2, to CEO (see Section 3.3). PB might increase AN-derived radioactivity in the liver and
stomach by inducing other enzymes, such as nicotinamide adenine dinucleotide phosphate
(NADPH)-cytochrome CYP450 reductase.
Ahmed et al. (1996a) monitored the tissue distribution of AN-derived radioactivity in
F344 rats (four/group) up to 48 hours following i.v. injection of 11.5 mg/kg of [2-14C]-AN
(50 (j,Ci/kg). The study authors used whole-body autoradiography to chart a time course of tissue
deposition and obtained the highest levels of activity in lung (998 nCi/mg), intestinal contents
(752 nCi/mg), liver (713 nCi/mg), and spleen (539 nCi/mg) 5 minutes after dosing. Other tissues
with high radiolabel at this time point were the kidney (283 nCi/mg), epididymis (266 nCi/mg),
adrenal gland (241 nCi/mg), intestinal mucosa (245 nCi/mg), heart-blood (166 nCi/mg), bone-
marrow (178 nCi/mg), thyroid (121 nCi/mg), adipose tissue (169 nCi/mg), and lacrimal gland
(122 nCi/mg), while the brain, spinal cord, and testis had the lowest levels of radioactivity. At
8 hours after dosing, the contents of the large intestine, especially the cecum, had the highest
level of radioactivity (852 nCi/mg). Radioactivity in brain (92 nCi/mg), lacrimal gland
(294 nCi/mg), and thyroid (211 nCi/mg) peaked at 24 hours after dosing, while radioactivity
level in bone marrow (698 nCi/mg) peaked at 48 hours. Covalent bound radioactivity, as
determined after acid-extraction techniques on freeze-dried sections, was observed in the spleen,
liver, bone marrow, and lung.
13
DRAFT - DO NOT CITE OR QUOTE
-------
3.3. METABOLISM
A proposed scheme for the metabolic pathways of AN in mammals is shown in
Figure 3-1. The scheme has been developed as a result of studies on the identification of urinary
metabolites following acute exposure (Fennell and Sumner, 1994; Kedderis et al., 1993a; Fennell
et al., 1991; Turner et al., 1989; Miiller et al., 1987; Tardiff et al., 1987; Gut et al., 1985;
Kopecky et al., 1980; Langvardt et al., 1980), measurements of the in vivo modulation of
AN-induced toxicological and biochemical changes by enzyme inhibitors or inducers (Wang et
al., 2002; Sumner et al., 1999; Burka et al., 1994; Kedderis et al., 1993c; Pilon et al., 1988a, b;
Ghanayem and Ahmed, 1986; Ahmed and Abreu, 1981; Abreu and Ahmed, 1980), determination
1996a), and analysis of metabolites formed in vitro by subcellular fractions of liver, lung, and
kidney in response to AN administration (Mostafa et al., 1999; Kedderis et al., 1995; Kedderis
and Batra, 1993; of the subcellular distribution of metabolic products of AN (Nerland et al.,
2001; Ahmed et al., Roberts et al., 1991, 1989; Hogy, 1986; Geiger et al., 1983; Ahmed and
Abreu, 1981; Guengerich et al., 1981; Abreu and Ahmed, 1980).
rsch2ch2cn«
HOOC—ch2—s—ch2ch2cn
S-(2-cyanoethyl)thioacetic acid
H,0
cn— c=
:N
Acrylonitrile
GSH
glutathione
GS CH2CH2CN
S-(2-cyanoethyl)glutathione
Ac-N-Cys S CH2CH2CN
N-acetyl-S-(2-cyanoethyl)cysteine
NGs
H
ch./N,
spontaneous
""rearrangement
O
Cyanoacetaldehyde
ch2
Cyanoethanol
OH
OH
ch2
Cyanoacetic acid
o
CH2 OH
I
CHO G1yco1 Cyanide
epoxide/ aldehyde
^ / hydrolase
/\
ch2—CH—CN CH2OH
rhodanese
CN -SCN
Thiocyanate
2-Cyanoethylene
oxide
GS-
CHCN
I
ch2oh
Ac-N-Cys—S—CHCN
N-acetyl-S-(l-cyano-2-hydroxyethyl)cysteine
OH
I
GS—CH2—CHCN
Cyanohydrin
i
GS—CH2CHO '
CN"
Cyanide
rhodanese
- SCN
Thiocyanate
Ac-N-Cys S— CH2CH2OH
N-acetyl-S-(2-hydroxyethyl)cysteine
S-2-oxoethylglutathione
\ o
Cys S—CH2COOH —||
S-(carboxymethyl)cysteine HOOC CH2 S CH2COOH
I Thionyldiacetic acid
Ac-N-Cys —S CH2COOH
N-acetyl-S-(carboxymethyl)cysteine
Source: National Toxicology Program (NTP) (2001).
Figure 3-1. Scheme for the metabolic transformation of AN.
14
DRAFT - DO NOT CITE OR QUOTE
-------
Primary components of the scheme are formation of CEO by the action of mixed function
oxidases (predominantly CYP4502E1), detoxification of AN by interaction with GSH, and
covalent binding of AN reactive metabolite to other biological macromolecules.
In rats, CYP2E1 is present in the liver and widespread among other tissues. This isoform
can be induced in the tongue, esophagus, forestomach squamous epithelia (Shimizu et al., 1990),
and intestinal mucosa (Subramanian and Ahmed, 1995). It is present, albeit at levels 10-
20 times lower than in liver, in such tissues as kidney and lung. It is also present in the brain
(Geng and Strobel, 1993; Sohda et al., 1993). The formation of CEO has important implications
for the toxicity of AN, because the intermediate has been proposed as the principal carcinogenic
metabolite of AN. However, CEO can undergo a number of further transformations. These
include the interaction with GSH to form a series of cysteine or N-acetyl cysteine derivatives and
the production of cyanide via the action of epoxide hydrolase (EH). Subsequent detoxification
of cyanide to thiocyanate is thought to occur under the action of rhodanese (Kopecky et al.,
1980).
3.3.1. Oxidation of AN to CEO
Evidence for the oxidation of AN to CEO and its subsequent transformations came from
a number of studies. Abreu and Ahmed (1980) studied the in vitro conversion of AN to cyanide
in subcellular fractions of liver from Sprague-Dawley rats (also reported in Ahmed and Abreu
[1981]). The metabolic activity was localized in the microsomal fraction and required NADPH,
MgCh, and oxygen for maximal activity. Determination of the kinetic parameters (Km and Vmax)
of the transformation of AN to cyanide pointed to a higher affinity and faster rate of product
formation in microsomes from rats pretreated with CYP450-inducing agents, such as
Aroclor 1254 and PB. (Six AN concentrations ranging from 10 to 300 mM were used for each
preparation.) The Km values calculated for the PB and Aroclor 1254 preparations were 54.8 and
40.9 mM, respectively, and were lower than the control (190 mM). Pretreatment of rats or
addition to incubation mixtures with agents that inhibit CYP450 activity, such as SKF 525A or
cobalt chloride, reduced the amount of cyanide formed by rat liver microsomes.
Abreu and Ahmed (1980) studied the effect on cyanide formation when 1,1,1-trichloro-
propane-2,3-oxide (TCPO), a specific inhibitor of EH, was added to the microsomal incubation
mixtures. Production of cyanide was dose-dependently reduced to 19% of control levels at
TCPO concentration of 1 x 10"2 M. Abreu and Ahmed (1980) also tested the effect of sulfhydryl
compounds on microsomal metabolism of AN, as measured by the rate of cyanide formation.
Only cysteamine decreased cyanide formation; other sulfhydryl compounds, including GSH and
cysteine, enhanced the rate of cyanide formation from AN.
Abreu and Ahmed (1980) suggested that probably more than one step was involved in the
enzymatic conversion of AN to cyanide. The study authors proposed the initial product of AN
oxidation to be CEO, which could then undergo a number of alternative transformations, one of
15
DRAFT - DO NOT CITE OR QUOTE
-------
which would be non-enzymatic conversion to cyanide. That another transformation might
involve EH was indicated by the decrease in cyanide formation following administration of the
inhibitor TCPO. Because cyanide production was enhanced in the presence of sulfhydryl
compounds, such as GSH, chemical interaction of CEO with GSH could lead to formation of
cyanohydrin. Rearrangement of this cyanohydrin to an aldehyde could result in the release of
cyanide. Cysteamine might diminish cyanide formation due to its inhibition of CYP450-
dependent metabolism (Buckpitt et al., 1979).
Geiger et al. (1983) studied the conversion of AN to its metabolic products in isolated rat
hepatocytes and demonstrated the formation of CEO and its hydrolysis to cyanide, which itself
was transformed and detected as thiocyanate.
In vitro incubation of AN with liver microsomes isolated from male F344 rats pretreated
with inhibitors or inducers of specific members of the CYP450 family of mixed function
oxidases indicated that CYP2E1 is the major catalyst in the oxidation of AN to CEO (Kedderis et
al., 1993c). The rate (Vmax) at which rat liver microsomes oxidized AN to CEO was increased
more than fivefold from Vmax of 366 pmol CEO/minute-mg for untreated rats following acetone
pretreatment, although Km was increased from 11 to 19 [xM. Because acetone is a potent inducer
of CYP2E1, the data suggest that this isoform is a primary catalyst of AN epoxidation in rats.
Treatment with P-naphthoflavone to induce CYP1 Al and CYP1A2 or with dexamethasone
(DEX) to induce the CYP3 A enzymes increased Vmax only less than twofold, but Km was
increased by 3.5 and 5.2-fold, respectively, for AN epoxidation in these two pretreatment
systems. Treatment with PB to induce CYP2B1 and CYP2B2 slightly decreased the Vmax but
increased the Km in microsomes from rats. These studies demonstrated that other forms of
CYP450 (CYP2B1, CYP2B2, and the 3 A enzymes) can oxidize AN but with specific activities
much lower than CYP2E1.
The effect of a number of CYP450 inhibitors on epoxidation of 1.2 mM AN by rat
hepatic microsomes was investigated (Kedderis et al., 1993c). Neither SKF 525A nor
metyrapone were effective inhibitors, retaining 87% of control activity. After treatment of rats
with DEX or PB, SKF 525A became a more effective inhibitor, retaining 45 and 47% of control
activity. Metyrapone also became a more effective inhibitor of epoxidation of AN after DEX
treatment. The CYP450 ligand, 1-phenylimidazole, was a potent inhibitor of AN epoxidation
(4%> of control activity). Chlorzoxazone (27%), ethanol (42%), and diethyldithiocarbamate
(17%>) also inhibited this pathway. The changes in the degree of inhibition of epoxidation of AN
following DEX and PB treatments could be interpreted as multiple CYP450 enzymes from rat
hepatic microsomes were capable of oxidizing AN.
Antibodies to CYP2E1 (sheep or goat anti-rabbit CYP2E1) inhibited more than 85% of
AN epoxidation in liver microsomes from untreated or acetone-treated rats but only 40 and 60%
inhibition following DEX and PB treatment, respectively (Kedderis et al., 1993c), suggesting
that CYP450 enzymes other than CYP2E1 might participate in the epoxidation. However, it
16
DRAFT - DO NOT CITE OR QUOTE
-------
should be noted that the AN concentration in these in vitro studies was high (1 mM). Forms of
CYP450 enzymes other than CYP2E1 might have been recruited to AN metabolism.
Kedderis et al. (1993c) also investigated the kinetics of the epoxidation of 1.2 mM AN by
using human hepatic microsomes. Km and Vmax values for oxidation of AN by liver microsomes
from six uninduced individuals ranged from 12 to 18 |iM and from 129 to 315 pmol/minute-mg,
respectively. Antibodies to CYP2E1 produced 58-70% inhibition of AN epoxidation catalyzed
by the six human liver microsomal preparations. This suggests that, while CYP2E1 was the
major catalyst of AN epoxidation in humans, other isoforms of CYP450 may also be involved.
Sumner et al. (1999) investigated the role of CYP450 in the metabolism of AN in mice.
Three male wild-type (WT) mice and three to four male CYP450 2El-null mice were treated
13
orally with either 2.5 mg or 10 mg/kg of [1,2,3- C] AN. Urinary metabolites in samples
13
collected over 24-h were characterized using C nuclear magnetic resonance (NMR). In WT
mice, urinary metabolites of CEO predominated, with metabolites derived from GSH
conjugation at the 3-carbon of CEO accounting for 67-71%. Metabolites from GSH conjugation
at the 2-carbon accounted for about 13%. Metabolites from direct GSH conjugation with the
parent compound, AN, accounted for 15-21%). In the urine of CYP2El-null mice, however,
only metabolites from direct GSH conjugation were detected. Sumner et al. (1999) interpreted
their data as indicating that CYP2E1 may be the only CYP450 involved in the metabolism of AN
in mice.
Subramanian and Ahmed (1995), attempting to characterize the specific intestinal
toxicity of AN, incubated microsomes isolated from male Sprague-Dawley rat intestinal mucosa
in vitro with AN in the presence of NADPH. AN metabolism to cyanide was enhanced by the
addition of sulfhydryl compounds such as GSH, cysteine, and D-penicillamine. AN metabolism
to cyanide was also enhanced following the induction of microsomal proteins (MPs) by treating
rats with PB (inducer of CYP2B1), P-naphthoflavone (inducer of CYP1A1), and
4-methylpyrazole (inducer of CYP2E1). AN metabolism to cyanide was inhibited to 8 and 20%>
of control, respectively, when dimethyl sulfoxide (DMSO) or ethanol (competitive inhibitors of
CYP2E1) was added to the incubation mixtures.
Subramanian and Ahmed (1995) showed that the intestinal CYP450 isoform had a high
affinity for AN, with a Km of 1.1 [xM and a Vmax of 1,250 pmol/mg protein/minute. Addition of
DMSO in varying concentrations (final 30 mM) increased the Km of the reaction to 10 [xM, but
Vmax remained unchanged. Since DMSO is a specific substrate and competitive inhibitor for
CYP2E1, these studies indicated that CYP2E1 was the main CYP450 isoform that bio-activates
AN in the intestine. In addition, anti-P450 3a immunoglobulin (Ig)G (which cross-reacts with rat
CYP2E1) caused a concentration-dependent inhibition of the metabolism of AN to cyanide in the
ethanol-induced intestinal microsomes (Subramanian and Ahmed, 1995). These results showed
that CYP2E1 was the main intestinal mucosa enzyme metabolizing AN to cyanide.
17
DRAFT - DO NOT CITE OR QUOTE
-------
Similarly, Abdel-Aziz et al. (1997) demonstrated the metabolism of AN to cyanide when
AN was incubated in vitro with NADPH and a microsomal fraction prepared from Sprague-
Dawley rat testis. The Vmax of this reaction was 65 pmol CNVmg protein/minute, and the Km
was 88.6 [jM AN. Addition of SKF 525A or benzimidazole (competitive inhibitors of CYP450)
to the incubation mixture inhibited the formation of cyanide, whereas microsomes obtained from
PB-treated rats increased activation of AN to cyanide. Thus, AN was metabolized in rat testis
via CYP450 mixed function oxidase. Addition of GSH, L-cysteine, D-penicillamine, or
2-mercaptoethanol also enhanced the release of cyanide from AN.
The capability of rat kidney to metabolize AN to cyanide was demonstrated by Mostafa
et al. (1999) in an in vitro study that investigated the mechanism by which AN caused renal
toxicity. In renal subcellular fractions from Sprague-Dawley rats, the metabolism of AN to
cyanide was highest in the microsomal fraction. An NADPH-generating system in the presence
of magnesium ions was required for maximal activity. The Vmax of this reaction was 118 pmol
CNVmg protein/minute, and the Km was 160 |iM AN. Metabolism of AN to cyanide was
increased when microsomes were obtained from PB-, ethanol-, 4-methylpyrazole-, and
3-methylcholanthrene-treated rats. On the other hand, addition of SKF 525 A or benzimidazole
to the incubation mixture inhibited AN metabolism. These data suggested that AN was
metabolized in the kidney via a CYP450-dependent mixed function oxidase system. Addition of
GSH, L-cysteine, cysteamine, D-penicillamine, or 2-mercaptoethanol to the incubation mixture
enhanced AN metabolism.
Ahmed and Patel (1981) carried out a series of single-dose gavage experiments on male
Sprague-Dawley rats and male Swiss mice at fractions of the LD50 values of AN and KCN. (The
LD50 of AN is 93 mg/kg in rats and 27 mg/kg in mice; the LD50 of KCN is 10 mg/kg in rats and
8.5 mg/kg in mice.) Cyanide was measured and detected in blood, liver, kidney, and brain of
both rats and mice 1 hour after administration in a dose-dependent manner. However, cyanide
concentrations from metabolism of one LD50 AN in blood and tissues of rats were significantly
lower than those produced from one LD50 of KCN. On the other hand, comparable
concentrations of cyanide in blood and tissues were observed after one LD50 AN or KCN was
administered to mice. Blood and liver contained higher amounts of cyanide per unit volume than
kidney and brain (the other two organs evaluated).
Observed signs of toxicity were also different in rats and mice administered an LD50 of
AN. Rats developed severe cholinomimetic signs including salivation, lacrimation, diarrhea,
wheezing on expiration, and peripheral vasodilatation within 10 minutes after administration of
93 mg/kg AN. These signs were not observed in rats treated with KCN. Severe central nervous
system (CNS) effects such as depression, convulsions, and asphyxia were observed in rats 10-
20 minutes after treatment with 10 mg/kg KCN (LD50). These CNS signs of cyanide toxicity
were observed in AN-treated rats 2-3 hours after dosing. No physiological adverse effects were
observed in rats receiving 0.25 LD50 AN. Mild salivation, diarrhea, and vasodilation were
18
DRAFT - DO NOT CITE OR QUOTE
-------
observed after one-half LD50 AN in rats. However, in mice treated with equitoxic dose (LD50
27 mg/kg) AN, CNS signs identical to those observed after KCN was administered were
observed. These results demonstrated species differences in the toxicity and metabolism of AN.
Moreover, Ahmed and Patel (1981) showed that pretreatment of rats with Aroclor 1254
and PB increased AN metabolism to cyanide in rats, and pretreatment of rats with C0CI2 or SKF
525A decreased blood cyanide concentrations. These results showed that the AN
transformations demonstrated by Abreu and Ahmed (1980) in vitro also could take place in vivo.
In addition, increased metabolism of AN to cyanide would increase CNS effects. However,
acute AN toxicity was also manifested as cholinomimetic signs, which were not from cyanide.
Shibata et al. (2004) employed headspace gas chromatography to simultaneously measure
blood levels of AN and its metabolite, hydrogen cyanide, following oral administration to rats.
Plasma and urinary thiocyanate concentrations were also measured by the colorimetric method.
Male Wistar rats that received 40 mg/kg AN (about half the LD50) by gavage in water showed
toxic signs such as tachycardia 1 hour later. Peak blood levels of AN (2 (j,g/mL) and cyanide
(0.7 ng/mL) were detected 1.5 hours after dosing. By 2.5 hours after dosing, blood levels of AN
had dropped to less than 0.2 ng/mL and blood levels of cyanide decreased to 0.1 (j,g/mL; at that
time, thiocyanate was detected in plasma (20 ng/mL). Plasma thiocyanate concentrations rose
over time, peaking at 5 hours (31.3 (j,g/mL). At the same time, excretion of thiocyanate in urine
began to increase significantly. Ten hours after dosing, AN was still detectable in plasma at
50 ng/mL, but cyanide had decreased to a background level of about 5 ng/mL. The cumulative
urinary elimination of thiocyanate gradually increased, and at 10 hours, about 1.2 mg thiocyanate
was excreted into the urine. This amount was calculated to be 7% of the total administered AN.
Urinary AN level was not measured.
The capacity for formation of CEO from AN has been demonstrated in F344 rat liver
microsomes, lung microsomes, and isolated lung cells. The rate of CEO formation in rat lung
was cell specific, with the Clara cell-enriched fraction having a rate of CEO formation 7 times
greater than other cell fractions (Roberts et al., 1989). The overall rate of CEO formation was
about 15 times greater in the livers than the lungs (Roberts et al., 1989).
Roberts et al. (1991) provided data on the kinetics of CEO formation in liver and lung
microsomes isolated from male F344 rats, B6C3Fi mice, and humans (Table 3-4). While CEO
was produced in vitro by lung and liver microsomes in both rats and mice, the metabolite was
produced at a greater rate in liver compared with lung and in mice vs. rats. These data
potentially implicated the liver as the primary site of CEO formation after oral challenge with
AN but suggested differences in the kinetics of CEO formation between species. The rate of
CEO formation in microsomes isolated from human livers was comparable to that of F344 rats,
but about 4 times lower than that of B6C3Fi mice. The average rate of CEO formation in liver
microsome samples from six human donors was 501 ±112 pmol/minute-mg protein. (The
almost eightfold variation in enzymatic activity among these human samples appeared to
19
DRAFT - DO NOT CITE OR QUOTE
-------
correlate with the amount of CYP450 in each preparation.) However, after oral administration of
AN, the concentration of CEO in mouse blood was about one-third that in rat blood at all doses
and time points tested. Thus, blood CEO concentration did not correlate with rate of microsomal
CEO formation, suggesting that species differences in detoxification of CEO might play a role in
determining CEO concentrations in blood. The single human lung microsome sample tested in
the study formed 0.55 pmol CEO/minute-mg protein, which was much lower than that for rat
liver or lung microsomes.
Table 3-4. Apparent kinetic parameters of CEO formation from AN in
B6C3F1 mice, F344 rats, and humans
Tissue
Species
V max (pmol/min-mg protein)
Km (|iM)a
Liver
Mouse
2,801a
67
Rat
66T
52
Human
501b
Not available
Lung
Mouse
570a
1,229
Rat
45a
1,854
Human
0.55°
Not available
aValues are the mean of eight replicates.
bValue is the mean of six human donors.
°Value is from one human donor.
Source: Roberts et al. (1991).
Guengerich et al. (1981) used a reconstituted enzyme system containing purified rat liver
CYP450 and NADPH-P450 reductase and a NADPH-generating system to oxidize AN in vitro
to a metabolite that they identified colorimetrically as CEO. The extent of CEO accumulation
was decreased by the addition of purified rat liver EH to the incubation medium. When 0.5 mM
CEO was incubated with 30 ng/mL purified EH, the rate that CEO was hydrolyzed was
5.5 nmol/minute. The rate of disappearance of CEO (due to nonenzymatic hydrolysis) was
1.7 nmol/minute in the absence of EH or in the presence of inactivated EH. HCN was released at
a rate of 1.5 nmol/minute during the hydrolysis of CEO by EH. (The rate of HCN release was
0.2 nmol/minute in the absence of EH.) The study authors suggested that the reason the HCN
release was not stoichiometric with epoxide disappearance might be due to a finite level of
cyanohydrin existing in solution. A Km of 0.8 mM and a Vmax of 300 nmol/minute-mg based on
disappearance of CEO was estimated for the hydrolysis of CEO by purified rat liver microsomal
EH. The half-life of CEO in 0.1 M potassium phosphate was estimated to be about 2 hours at
37°C (Guengerich et al., 1981).
Kopecky et al. (1980) also investigated the role of EH and the generation of hydrogen
cyanide from AN metabolism. AN was incubated in vitro with liver microsomes isolated from
2+
female Wistar rats, with and without cofactors (NADP, Mg , etc.) for 60 minutes. After the
20
DRAFT - DO NOT CITE OR QUOTE
-------
incubation, the mixtures were adjusted to either pH 1.8 (acidic processing) or 6.3 (alkaline
processing). Cyanide released, in the presence of cofactors, was found to be fourfold higher
under alkaline processing condition than acidic processing condition. The role of EH in AN
metabolism in rats was supported by the increase in cyanide released after alkaline processing,
indicating the existence of a cyanohydrin intermediate (glycolaldehyde cyanohydrin) in the
biotransformation of AN. (Cyanohydrins generally decompose spontaneously to hydrogen
cyanide and a carbonyl compound at pH higher than 7.) Moreover, when 3,3,3-trichloro-
propylene oxide (TPO), a potent inhibitor of EH, was added to the incubation mixture, the
conversion of AN to cyanide was decreased by 70%. This result also provided evidence for the
participation of the cyanohydrin in AN metabolism because the hydration of CEO to
glycolaldehyde cyanohydrin was significantly inhibited by TPO.
Kedderis and Batra (1993) compared the rates of CEO hydrolysis, enhanced by liver
cytosol and microsomes from rats, mice, and humans, to the background rate of non-enzymatic
hydrolysis displayed by the chemical. [2,3-14C]-CEO was incubated at pH 7.3 and 37°C for
5 minutes with liver cytosol or microsomes. [2,3-14C]-CEO and its hydrolysis products were
separated by high performance liquid chromatography (HPLC). The identity of the hydrolysis
products could not be determined and did not correspond to aldehydes. Human hepatic
microsomes enhanced the formation of hydrolysis products of CEO, whereas both human hepatic
cytosol and liver cytosol and microsomes from F344 rats and B6C3Fi mice had no effect on
hydrolysis product formation from CEO. The study authors concluded that rodent hepatic
microsomal and cytosolic EHs were not active toward CEO, contrary to conclusions developed
by Guengerich et al. (1981) on rat purified microsomal EH and the conclusions by Kopecky et al.
(1980).
One possible explanation that EH activity was not observed by Kedderis and Batra
(1993) in microsomes from rats and mice was that their enzymatic reactions had not been
optimized. No cofactors were used in the incubation mixture, and the incubation duration was
only for 5 minutes. Both Kopecky et al. (1980) and Guengerich et al. (1981) used an NADPH-
generating system in their incubation mixture.
Kedderis and Batra (1993) also showed that the heat-labile human EH activity was
inhibited by the specific inhibitor, 1,1,1-trichloropropene oxide (median inhibitory concentration
[IC50] of 23 |iM), indicating that EH was the catalyst in the hydrolysis of CEO. The half-life of
CEO in sodium phosphate buffer (pH 7.3), as estimated from hydrolysis by human liver
microsomes, was 99 minutes. Estimated Km using liver microsomes from six individuals ranged
from 0.6 to 3.2 mM. Vmax ranged from 8.3 to 18.8 nmol hydrolysis products/minute-mg protein.
The affinity of the human liver microsomal EH for CEO was relatively low, suggesting that the
contribution of the hydrolysis pathway to the clearance of CEO would be small at low substrate
concentrations. Increase in microsomal hydrolysis was observed after treatment of mice and rats
21
DRAFT - DO NOT CITE OR QUOTE
-------
with PB or acetone, suggesting that CEO hydrolysis is inducible. However, treatment of rats and
mice with AN did not induce hepatic EH activity towards CEO (Kedderis and Batra, 1993).
Studies that demonstrated that EH is present in rats and humans are also available.
Immunoblot analysis of MPs was used by de Waziers et al. (1990) to measure EH in different
organs and tissues of rats and humans. They reported that EH occurred in rat liver microsomes
at 165 (J,g/mg protein and in human liver microsomes at 170 (J,g/mg protein. Therefore, the
concentrations of EH in MP are similar in rats and humans. Guengerich et al. (1979) also
reported that multiple forms of EH exist in rats and humans.
EH activity was demonstrated in mice in a recent study, contrary to the conclusion by
Kedderis and Batra (1993). El Hadri et al. (2005) demonstrated that microsomal EH was present
in WT mice, which metabolized AN administered by gavage to cyanide in a dose- and time-
dependent manner. Blood cyanide levels in microsomal EH-null mice treated with a gavage
dose of 0.047-0.38 mmol/kg AN were lower than levels in similarly treated WT mice. Blood
cyanide level was also largely abolished in CYP2El-null mice and in WT mice pretreated with a
nonselective CYP inhibitor, 1-aminobenzotriazole (ABT), confirming that CYP2E1 was the key
enzyme for the epoxidation of AN and the subsequent formation of cyanide. AN-treated
CYP2E1- and mEH-null mice showed less severe symptoms of cyanide poisoning (labored
breathing, lethargy, and trembling) than similarly treated WT mice (El Hadri et al., 2005).
Significantly higher levels of AN-derived blood cyanide levels were observed in male mice than
in female mice, suggesting gender-related differences in toxicity. Western blot analysis also
demonstrated that expression of soluble EH was greater in male than female mice.
Detection and identification of urinary metabolites of AN from male F344 rats or B6C3Fi
13
mice exposed orally to [1,2,3- C]-AN (10 or 30 mg/kg for rats, 10 mg/kg for mice), using
13
C NMR spectroscopy, also offered more information of its possible metabolic interactions
(Fennel and Sumner, 1994; Fennel et al., 1991). As detailed in Section 3.4, some of the
hypothetical metabolites of AN shown in Figure 3-1 were detected in the urine of animals
exposed to AN. A major urinary metabolite in rats was N-acetyl-S-(2-cyanoethyl)cysteine, from
conjugation of AN with GSH. Other metabolites, formed following oxidation of AN to CEO and
subsequent conjugation to GSH, were identified as N-acetyl-S-(2-hydroxyethyl)cysteine,
thiodiglycolic acid, S-carboxylmethylcysteine, and thionyldiacetic acid, all derived from addition
of GSH to the 3-position of CEO. Thiocyanate was detected in urine as a metabolite of released
cyanide. Moreover, N-acetyl-S-(l-cyano-2-hydroxyethyl)cysteine was formed after addition of
GSH to the 2-position of CEO (Fennel and Sumner, 1994). These metabolites were also found in
mouse urine.
Species differences in the extent of AN metabolism via oxidation to CEO, and
subsequent conjugation of CEO with GSH, may exist. Fennell et al. (1991) and Fennell and
Sumner (1994) noted differences in the relative abundance of these urinary metabolites in mice
compared with rats. After oral administration of 10 mg/kg AN to mice, 80% of the urinary
22
DRAFT - DO NOT CITE OR QUOTE
-------
metabolites were derived from CEO, most notably thiodiglycolic acid and S-(carboxymethyl)-
cysteine. By contrast, these metabolites made up only 60% of metabolites in the urine from rats
administered orally with 10 or 30 mg/kg AN. This difference indicated that more CEO was
produced in the mouse than in rats. In addition, the ratio of metabolites derived from glutathione
conjugation of CEO at the 2- and 3-positions determined the amount of cyanide released, since
cyanide is released from CEO metabolites conjugated at the 3-position. This ratio was 0.43 in
rats and 0.21 in mice, indicating that a greater percentage of the CEO produced in mice was
metabolized to release cyanide. Thus, mice were likely to be exposed to a higher cyanide level
produced from CEO, possibly accounting for the greater acute toxicity of AN in the mouse
(Fennel and Sumner, 1994).
Wang et al. (2002) confirmed the central role of CYP2E1 in the metabolism of AN to
cyanide via CEO. Male WT and CYP2El-null mice were dosed by gavage with 0, 2.5, 10, 20, or
40 mg/kg AN, and cyanide was measured in blood and tissues. Expression of CYP2E1 and EH
was monitored concurrently using Western blot techniques. Cyanide concentrations in blood and
tissues of AN-treated WT mice increased dose dependently but remained at background levels in
CYP2El-null mice or control WT mice. Results from Western blots showed CYP2E1 to be well
expressed in the liver, kidney, and lung of WT mice and not detected in tissues of CYP2El-null
mice. EH was equally expressed in both WT and CYP2E1 mice, supporting the hypothesis that
CYP2E1-mediated oxidation of AN is an early step in the metabolism of AN to cyanide. The
role of cyanide in the acute toxicity of AN was confirmed by the lack of acute symptoms of AN
toxicity in CYP2El-null compared with WT mice. Pretreatment of WT mice with a universal
CYP450 inhibitor, ABT, likewise blocked cyanide formation and abolished the symptoms of
acute toxicity. Wang et al. (2002) concluded that the metabolism of AN to CEO was exclusively
catalyzed by CYP2E1.
3.3.2. Interaction of AN with GSH
A succession of research findings showed that AN conjugates with GSH, both non-
enzymatically or by being catalyzed by glutathione transferase. Young et al. (1977) suggested
that since the toxicity of AN was due to the parent compound (AN) or its oxidative metabolites,
the cyanoethylation of sulfhydryl-containing compounds, such as GSH or cysteine, by AN
represented a detoxification mechanism. This metabolic pathway was shown indeed to play a
role in the detoxification of AN (Ghanayem and Ahmed, 1986; Ghanayem et al., 1985; Appel et
al., 1981). The toxicity of AN would be expected to increase in severity as the GSH level
becomes depleted (Benz et al., 1997a).
Kopecky et al. (1980) administered 0.75 mmol/kg AN or 0.5 mmol/kg [1-14C]-AN to
female Wistar rats by different routes (oral, i.p., subcutaneous [s.c.], and i.v.) and measured
radioactivity and thiocyanate excreted in the urine. "Non-thiocyanate" metabolites excreted in
urine constituted about two-thirds of the administered dose. Paper chromatography of the
23
DRAFT - DO NOT CITE OR QUOTE
-------
metabolites in urine identified AN mercapturic acid as the key metabolite. Kopecky et al. (1980)
proposed that there were at least two pathways for AN metabolism. The minor route was
oxidative metabolism to cyanide, which was further metabolized to thiocyanate and other "non-
thiocyanate" metabolites. The major route was conjugation with glutathione, catalyzed by
glutathione S-alkenetransferases, to N-acetyl-S-(2-cyanoethyl)cysteine.
Further support for this proposition came from Geiger et al. (1983), who studied AN
metabolism in isolated F344 rat hepatocytes. GSH levels and AN-protein binding were
measured after incubating rat hepatocytes with [1-14C]- or [2,3-14C]-AN. GSH-adduct levels
were determined by chromatographic procedures of aliquots of the trichloroacetic acid
supernatant. Exposure to AN at 5 or 10 mM resulted in decrease of GSH levels to 15-20% of
controls within 10 minutes. The primary radiolabeled product was S-(2-cyanoethyl)glutathione,
and not S-(2-oxoethyl)GSH (the compound formed by reaction of CEO with GSH in the
presence of purified GSH transferase).
Indirect evidence for the involvement of GSH in the metabolism of AN was provided by
Langvardt et al. (1980, 1979), who used gas chromatography-mass spectrometry and gas
chromatography-infrared spectroscopy to identify urinary components in male Sprague-Dawley
rats 16 hours after exposure to [1-14C]- or [2,3-14C]-labeled AN by gavage. They identified two
major components: the first was N-acetyl-S-(2-cyanoethyl)cysteine, which they assumed to be a
product of AN conjugation with GSH, and the other was thiocyanate. A third metabolite,
N-acetyl-S-(2-cyanoethyl)cysteine, was tentatively identified and was proposed by Langvardt et
al. (1980) to have resulted from the action of GSH on the epoxide intermediate. The authors
speculated that the detoxification of AN likely involved conjugation with GSH, and the toxicity
of AN was likely affected by the status of GSH pools in target tissues, since a rapid and dose-
dependent decrease in GSH stores in the liver, lungs, kidney, and adrenals was observed by
Szabo et al. (1977) after i.v. injection of 1-15 mg/100 g AN to Sprague-Dawley rats. A sharp
decrease in cerebral GSH concentrations occurred between 5 and 15 mg/100 g AN and correlated
with the occurrence of mortality. On the other hand, oral dosing of 0.002-0.05% AN to rats for
21 days resulted in up to 25% increase in hepatic GSH and might represent a rebound
phenomenon (Szabo et al., 1977).
Benz et al. (1997a) studied the time and dose dependence of the depletion of tissue GSH
and tissue cyanide and the covalent binding to tissue after s.c. injection of 0, 20, 50 (LDio),
80 (LD50), or 115 mg/kg (LD90) AN to male Sprague-Dawley rats. GSH levels in liver were the
most sensitive marker of AN exposure and were depleted by 50% at 20 mg/kg, a dose without
overt toxicity. At 50 mg/kg, the threshold dose for overt toxicity, GSH was depleted by >85%
and followed by a rapid recovery of 60% at 4 hours. Liver GSH was depleted almost completely
within 30 minutes when rats were injected with 80 mg/kg AN. The depletion was sustained
through 120 minutes and followed by 40% recovery through the end of study period of 4 hours.
Blood and brain GSH were more resistant to the GSH depleting effects of AN and were depleted
24
DRAFT - DO NOT CITE OR QUOTE
-------
less extensively in a dose-dependent manner as the doses were in the toxic range. (The highest
dose of 115 mg/kg depleted only 40% brain GSH at 2 hours.) In addition, brain GSH levels
showed little capacity for recovery during the study period, unlike liver and kidney. Glandular
and forestomach GSH were also dose-dependently depleted by AN treatment and were unable to
recover within the study period.
GSH depletion was accompanied by a dose-dependent increase of cyanide in the blood
and brain during the first 60 minutes. At the lowest dose of 20 mg/kg, blood and brain cyanide
declined after 60 minutes. At the higher doses, blood and brain cyanide continued to increase to
120 minutes and then declined. Covalent binding of AN to tissue protein increased in all tissues
rapidly during the first 30 minutes at all doses. At 20 mg/kg, covalent binding reached a plateau
at 30 minutes. At the three higher doses, covalent binding continued to increase after 30 minutes
and reached a plateau level by 2-4 hours. Benz et al. (1997a) concluded that when liver GSH
was depleted, detoxification of AN was terminated. Acute AN toxicity became apparent, and a
sustained increase in covalent binding to tissue protein was observed.
Further experiments by Ahmed and coworkers (Ahmed et al., 1983, 1982; Ghanayem and
Ahmed, 1982) confirmed the involvement of hepatic GSH in AN metabolism. For example,
when bile was collected from male Sprague-Dawley rats given a single oral dose of 46.5 mg/kg
AN containing 12 j_iCi/kg [1-14C]-AN, four metabolites were isolated and characterized in biliary
extracts at 6 hours after treatment. The two main metabolites were S-cyanoethyl glutathione and
N-acetyl-S-(2-cyanoethyl)cysteine (Ghanayem and Ahmed, 1982). Both were glutathione
conjugates of AN. Pretreatment of rats with diethyl malate (a glutathione-depleting agent)
significantly decreased or abolished all of the metabolites in the bile. The study authors
proposed GSH conjugation as a major pathway of AN metabolism, probably catalyzed by
glutathione transferases. The product, S-cyanoethyl glutathione, is further metabolized to
N-acetyl-S-(2-cyanoethyl)cysteine. Nearly 27% of the administered dose was excreted in the
bile after 6 hours (Ahmed et al., 1982).
Kedderis et al. (1995) compared the kinetics of AN and CEO interaction with GSH in
vitro by measuring the formation of conjugates when [2,3-14C]-labeled AN or CEO were
incubated for a very short time (20 seconds for mice, 30 seconds for rats) with [glycine-
2- HJ-GSH in the presence of microsomal and cytosolic subcellular fractions of human, rat, or
mouse liver. Because of the rapid non-enzymatic reaction of AN and CEO with GSH at pH 7.3,
the steady-state kinetics of GSH conjugation were determined at pH 6.5. HPLC-mass
spectrometry was used to separate and identify the conjugates, which included S-(2-cyanoethyl)-
glutathione from AN; and S-(l-cyano-2-hydroxyethyl)glutathione and S-(2-cyano-2-hydroxy-
ethyl)-glutathione from CEO. The apparent kinetic parameters for the conjugation reactions
(Vmaxapp and Kmapp) were estimated by fitting the Michaelis-Menten equation to the data, giving
estimates of VmaXapp for the conjugation reactions catalyzed by mouse cytosolic enzymes that
were four to six times greater than those for rat cytosolic enzymes at pH 6.5 (Table 3-5). These
25
DRAFT - DO NOT CITE OR QUOTE
-------
data suggest that mouse liver cytosolic glutathione-S-transferase (GST) has a greater capacity for
conjugating AN and CEO than do rat liver enzymes. Initial velocity studies were also carried out
with microsomes and cytosol from human liver, providing data to suggest that GSH conjugation
of AN with human liver cytosol was broadly similar to that of rodent liver cytosol (Table 3-6).
Table 3-5. Apparent kinetic parameters for glutathione conjugation of AN
and CEO at pH 6.5
Fixed substrate
Species
Vmaxapp (nmol/min-mg)a
Kmapp (GSH) (mM)a
AN
Rat
3.32 ±0.29
17.47 ±3.60
Mouse
21.68± 1.75
23.39 ±3.99
CEO
Rat
2.00 ±0.20
9.36 ±2.27
Mouse
8.34 ±0.75
15.12 ± 2.84
"Values are the means ± SDs of three determinations.
Source: Kedderis et al. (1995).
Table 3-6. Glutathione conjugation of AN and CEO with or without
microsomal or cytosolic GST from rat, mouse, and human liver
preparations
Species"
Substrate
Nanomoles of product
Non-enzymatic
Microsomal
Cytosolic
Rat
AN
25.5 ±2.0
32.6 ± 1.9b
36.1 ±3.0b
Mouse
31.3 ± 1.2b
36.6 ± 1.4b
Human
19.5-24.5°
25.2-34.4°
Rat
CEO
15.5 ± 1.3
16.6 ±1.1
25.2 ± l.lb
Mouse
18.9 ± 1.0b
25.9 ± 3.7b
Human
11.2-14.3°
12.6-14.5°
aRodent values are means ± SDs of three determinations.
Statistically significantly greater than the non-enzymatic reaction (p < 0.05) as calculated by the authors.
°Range of values for subcellular fractions prepared from six human subjects.
Source: Kedderis et al. (1995).
In the same report, Kedderis et al. (1995) described the determination of the initial
velocities of non-enzymatic vs. enzymatic GSH conjugation of AN and CEO at pH 7.3 in the
absence or presence of GST (Table 3-6). Both substrates reacted rapidly with GSH in a non-
enzymatic reaction, and addition of hepatic microsomes or cytosol from rats and mice
statistically significantly enhanced the rate of product formation from AN (p < 0.05). Initial
velocities indicated that AN conjugated more effectively with GSH than did CEO. Hepatic
cytosol enhanced the rate of product formation from CEO to a greater extent (up to 52% higher)
than did microsomes, suggesting that rodent cytosolic GST is more active toward CEO than
26
DRAFT - DO NOT CITE OR QUOTE
-------
microsomal GST. Under physiological conditions (pH 7.3), addition of liver cytosol from four
out of six individuals similarly enhanced GSH conjugation with AN but not CEO. Human liver
microsomes did not enhance the velocity of CEO or AN conjugation with GSH, suggesting that
human microsomal GST forms do not catalyze this reaction (Table 3-6).
Similarly, Guengerich et al. (1981) studied the reaction of AN and CEO with glutathione
in vitro. The pseudo first-order rate constant for disappearance of GSH at pH 7.7 (37°C) was
0.28/minute, when 0.5 mM was mixed with 0.1 M AN. The rate constant for the disappearance
of GSH in the presence of 0.1 mM CEO was 0.11/minute.
Guengerich et al. (1981) also incubated [1-14C]-AN or -CEO with GSH at pH 6.5 in the
presence of cytosolic fraction from rat liver, human liver, or rat brain to determine GST activity.
Rat liver cytosol showed GSH S-transferase activity towards AN, with a Km of 33 mM and a
Vmax of 57 nmol/minute-mg protein. Human liver and rat brain cytosolic fraction had no activity
towards AN. Rat liver also showed activity towards CEO, with activity in rat brain more than an
order of magnitude lower.
The importance of sulfhydryl compounds in the detoxification of AN was shown in a
number of studies, particularly by the demonstration that treatment with exogenous sulfhydryl
compounds could protect the organism from the harmful effects of AN. Appel et al. (1981)
reported that sulfhydryl compounds such as cysteine were effective antidotes for both orally and
intraperitoneally administered lethal doses of AN in rats. N-acetyl-cysteine was ineffective
when given intraperitoneally and less effective than cysteine when administered orally. The
GSH-depleting effect of a single s.c. dose of AN administered to male Sprague-Dawley rats and
the subsequent GI bleeding and gastric mucosal necrosis could be blocked by pretreatment with
sulfhydryl-containing agents such as L-cysteine or cysteamine (Ghanayem and Ahmed, 1986;
Ghanayem et al., 1985).
Benz et al. (1990) studied the effectiveness of D- or L-cysteine and N-acetyl-D-cysteine
or N-acetyl-L-cysteine in the detoxification of acutely administered AN by determining the s.c.
LD50 of AN in male Sprague-Dawley rats either administered AN alone or in combination with
individual antidotes. The LD50 of AN alone was determined to be 74.7 mg/kg. The LD50 of AN
when combined with other antidotes ranged from 93.3 mg/kg (with N-acetyl-D-cysteine) to
151.4 mg/kg (with L-cysteine). The antidote protective index ranged from 1.25 for N-acetyl-D-
cysteine to 2.03 for L-cysteine. Thus, N-acetyl-D-cysteine was less effective than other antidotes
in reducing acute lethality. Measurement of urinary N-acetyl-S-cyanoethyl cysteine, which was
derived from conjugation with GSH pathway, following s.c. injection of 50 mg/kg AN alone or
AN plus an antidote, indicated that none of the antidotes significantly increased the excretion of
this metabolite.
Blood cyanide levels were also measured in rats at 0.5, 1, 2, 4, and 6 hours following s.c.
injection of 50 mg/kg AN or AN plus an antidote. Benz et al. (1990) showed that all of the
antidotes, except N-acetyl-D-cysteine, lowered blood cyanide levels. Since N-acetyl-D-cysteine
27
DRAFT - DO NOT CITE OR QUOTE
-------
was the least effective antidote, the antidotal effectiveness of these cysteine enantiomers was
related to their cyanide detoxification mechanism. Discussing these findings, Borak (1992)
pointed out that, because both D- and L-cysteine provided equivalent protection against AN
poisoning but only L-cysteine could be incorporated into GSH, the antidotal effects of these
compounds may be unrelated to GSH repletion. In fact, since the potency of each antidote was
proportional to their ability to lower cyanide levels, Borak (1992) suggested that their effects
may be due to the ability of cysteine derivatives to serve as sulfur donors for the detoxification of
cyanide via rhodanese-mediated transformation of cyanide to thiocyanate. The protection
provided by these antidotes for cyanide poisoning from AN exposure, however, does not
necessarily extend to other forms of AN-induced toxicity.
3.3.3. Covalent Binding of AN and Its Metabolites to Subcellular Macromolecules
When isolated hepatocytes from male F344 rats were incubated with 1 mM [2,3-14C]-AN
for 2 hours (Hogy, 1986; Geiger et al., 1983), a radiolabeled protein adduct was formed that
could be characterized after the removal of residual AN and other low molecular weight products
by dialysis. Analysis of the hydrolyzed non-dialyzable fraction indicated that about 75% of the
radioactivity was contained in a single major peak that was chromatographically identified as
S-(2-carboxyethyl)cysteine, the hydrolysis product of S-(2-cyanoethyl)cysteine, indicating direct
cyanoethylation of cysteinyl residues by AN. When isolated hepatocytes were incubated with
2 mM [2,3-14C]-AN for 60 minutes, Geiger et al. (1983) estimated the level of irreversible
binding to protein to be 8.1 nmol/mg protein, a level of alkylation corresponding to modification
of 1 in every 900 amino acid residues or roughly 1 of every 20 cysteine groups. In contrast to
AN-protein binding, detection of binding of [2,3-14C]-AN-derived radiolabel to DNA and RNA
in hepatocytes was limited by the low level of radioactivity that could be incorporated into
[14C]-AN because of polymerization and microsynthesis problems and by the high level of
protein binding. Therefore, hepatocytes were incubated with 2 mM [2,3-14C]-AN for 60 minutes
in the presence of extracellular calf thymus DNA (0.5 mg/mL) (Hogy, 1986; Geiger et al., 1983).
RNA and both intracellular and extracellular DNA were isolated. The isolated RNA contained
53 pmol AN metabolites per mg RNA of which 9 pmol/mg RNA could be attributed to the
contaminating protein. The isolated extracellular DNA contained 47 pmol AN adducts per mg
DNA, of which 30 pmol/mg DNA could be attributed to protein. A portion of incubated
extracellular DNA was further purified; alkylation of extracellular DNA was not observed at a
detection limit at 3.5 x 105 bases (Geiger et al., 1983).
Hogy and Guengerich (1986) treated an F344 rat intraperitoneally with 0.6 mg/kg
[2,3-14C]-CEO to assess macromolecular binding in liver and brain 1 hour after treatment. DNA
and RNA in liver and brain were isolated, and the amounts were estimated. Bound radioactivity
was estimated by liquid scintillation counting. Covalent binding of CEO-derived radioactivity
was detected in both liver and brain protein at rates of 1.1 and 1.0 alkylations per 106 amino
28
DRAFT - DO NOT CITE OR QUOTE
-------
acids, respectively. However, no covalent binding to DNA or RNA could be detected at the
level of 0.3 alkylations per 106 bases in liver and brain.
In another experiment by Hogy and Guengerich (1986), liver and brain DNA were
isolated from three rats 2 hours after treatment with 50 mg/kg AN i.p. or 6 mg/kg CEO i.p.
Using thin layer chromatography with fluorescent plates, the study authors were able to detect
N7-(2-oxoethyl)guanine at 0.014 and 0.032 alkylations per 106 DNA bases in liver for rats treated
with CEO and AN, respectively. This DNA adduct was apparently derived from CEO,
consistent with its recovery after AN and CEO treatment. The apparent lineage of this DNA
adduct was different from the protein adduct, S-(2-cyanoethyl)cysteine, described previously,
which was derived from AN.
N -(2-oxoethyl)guanine was only at the level of detection in rat brain DNA. In addition,
l,N6-ethenoadenosine or l,N6-ethenodeoxyadenosine were not detected in liver DNA, using
HPLC with fluorescence detector in the analyses of these two adducts. Binding of AN and/or
CEO to DNA is also discussed in Section 4.5.1.2.1.
As discussed in Section 3.3.2, when 0, 20, 50, 80, or 115 mg/kg [2,3-14C]-AN was
injected subcutaneously into male Sprague-Dawley rats, GSH became almost completely
depleted (>95%) in liver at 80 mg/kg within 30 minutes, while blood and brain GSH were more
resistant to the depleting effect of AN. Brain and blood GSH were not affected at 20 mg/kg.
The amount of cyanide in blood and brain increased dose dependently in the first hour after
dosing (Benz et al., 1997a). Covalent binding to tissue proteins increased in a dose-dependent
fashion during the first 30 minutes at all doses, with binding to blood proteins being 3-4 times
greater than in any other tissue. Benz et al. (1997a) suggested that GSH depletion in liver was
related to AN toxicity and covalent binding.
The effect of GSH depletion on the irreversible binding of AN to tissue macromolecules
has been studied in male F344 rats exposed to 4 mg/kg [2,3-14C]-AN either by inhalation (Pilon
et al., 1988a) or gavage (Pilon et al., 1988b). Binding of radiolabel to tissue macromolecules
was evaluated in control rats or rats depleted of GSH by an i.p. injection of phorone (PH)/
buthionine sulfoximine (BSO) about 30 minutes prior to AN exposure. GSH contents in control
rats were as follows: liver (17.3 (j,mol/g), kidney (4.5 (j,mol/g), lung (3.1 (j,mol/g), stomach
(5.3 (j,mol/g), brain (3.9 (j,mol/g), and blood (4.2 (j,mol/g). A significant depletion of GSH was
produced in liver (43%), kidney (42%), and lung (22%) after PH/BSO treatment. No significant
depletion of GSH was observed in blood, brain, or stomach 30 minutes after a combined
PH/BSO treatment.
In the inhalation studies (Pilon et al., 1988a), three rats were exposed to initial AN
concentrations of 0, 25, 50, 100, 500, or 750 ppm in a closed-circuit inhalation chamber for
240 minutes. AN was not replenished during the exposure, and the decrease in chamber AN
concentration was monitored by taking samples every 10 minutes during the exposure. Uptake
of AN vapor by control rats showed two distinct phases: an initial, rapid phase that lasted about
29
DRAFT - DO NOT CITE OR QUOTE
-------
60 minutes, followed by a slower phase. An uptake rate of 4.82 mg/kg-hour was estimated for a
concentration of 100 ppm using the uptake curve for the rapid phase. In GSH-depleted rats, the
mortality rate was higher. The rate of AN uptake was increased in the rapid phase but decreased
at 500 and 750 ppm. In the slow phase, uptake was similar to that in control rats for
concentrations below 200 ppm, but was elevated at 500 and 750 ppm.
Radioactivity irreversibly associated with tissue macromolecules was measured in control
rats 1, 2, 4, 6, 12, or 24 hours after the AN dose. In GSH-depleted rats, radioactivity was
measured 1, 6, or 24 hours postexposure. In most tissues, the concentration of AN-derived
undialyzable radioactivity (ADUR) reached a maximum in <1 hour in both control and
GSH-depleted rats. (In the brain, ADUR levels reached a maximum in 2 hours.) The time
course of ADUR levels showed a plateau from 1 to 6 hours in studied organs of control rats,
followed by a decrease thereafter during the next 6 hours. This was then followed by an increase
between 12 and 24 hours in the lung and kidney (but not in the brain, stomach, and liver). The
time course of ADUR levels in blood showed high level 1 hour after AN administration,
decreased from 6 to 12 hours, and increased from 12 to 24 hours, resulting in a maximum level at
24 hours.
Total radioactivity recovered from brain, stomach, liver, kidney, and blood decreased by
54% in GSH-depleted rats compared with controls at both 1 and 24 hours. In GSH-depleted rats,
ADUR levels remained constant in all organs evaluated throughout the 24-hour postexposure
period and were lower than those in controls. The kidney was the most affected organ, with an
average 52% decrease in ADUR levels, followed by the liver (44%). ADUR levels in the brain,
stomach, and lung of GSH-depleted rats were 31% lower when compared with controls. Blood
ADUR levels were decreased by 50% in GSH-depleted rats.
Irreversible binding of ADUR with total nucleic acids (RNA + DNA) of brain, stomach,
and liver was also evaluated in control and GSH-depleted rats 1 hour after an inhalation dose of
4 mg/kg (40 (j,Ci) AN vapor. In control rats, ADUR in total nucleic acids was found to be
highest in the brain (63 pmol AN equivalents/mg), followed by that in the stomach (20 pmol AN
equivalents/mg) and the liver (12 pmol AN equivalents/mg). In GSH-depleted rats, ADUR in
nucleic acids in the brain was about 50% lower than that in controls, but no change was detected
in stomach or liver. ADUR in DNA could not be detected in any tissue of control or
GSH-depleted rats. The study authors suggested that ADUR in DNA was not detected because
the analytical method used was not sensitive enough to detect radiolabel bound to DNA.
Nevertheless, a preferential binding of ADUR with brain RNA was observed in both control and
GSH-depleted rats, with ADUR in brain RNA about 50% lower in GSH-depleted rats.
It is unlikely that the observed decrease in radiolabel binding in GSH-depleted rats
treated with PH/BSO was due to treatment effect on CYP2E1. BSO is a selective inhibitor of
GST; PH depletes GSH in liver, kidney, and lung but not in the blood or brain by conjugation
30
DRAFT - DO NOT CITE OR QUOTE
-------
with GSH. PH was reported not to affect the hepatic microsomal mixed-function oxidase system
(Younes et al., 1986).
In the oral studies on male F344 rats, total radioactivity recovered in the tissues examined
was similar in control and GSH-depleted rats at both 1 and 24 hours. However, significantly
higher recovered radioactivity was found in stomach, lung, and blood of GSH-depleted rats than
of control rats 1 hour after dosing, while the radioactivity in kidney was lower in GSH-depleted
rats (Pilon et al., 1988b).
The concentration of ADUR in brain reached a maximum in <1 hour for both control and
GSH-depleted rats and remained constant for 24 hours. In GSH-depleted rats, 60-80% increase
in ADUR was measured in brain at 1, 6, and 24 hours after dosing. The highest ADUR level was
found in stomach and increased with time in control rats. ADUR levels in liver and kidney of
control rats was characterized by an increase over the first 6 hours and a decrease between 5 and
24 hours. GSH-depletion resulted in an increase in ADUR levels in liver, lung, kidney, stomach,
blood, and brain between 6 and 24 hours after the dose.
GSH depletion also caused a significant increase in ADUR levels in total nucleic acid
(DNA + RNA) in both brain and stomach (one and a half- and threefold, respectively) 6 hours
after the dose. No change was found in liver. ADUR associated with DNA was detected in
stomach tissue of control rats only. Pilon et al. (1988b) suggested that ADUR levels reflected
the relative concentration of covalently bound radioactivity in control and GSH-depleted rats and
that the reaction of AN with protein and other macromolecules was responsible for the rapid
increase in ADUR at <1 hour. The slower increase in ADUR in metabolically competent organs
(liver, kidney, and lung) of control rats and all organs of GSH-depleted rats might represent the
binding of CEO to macromolecules. Urinary excretion of thiocyanate, a final metabolite from
the epoxide pathway of AN metabolism, was increased twofold in GSH-depleted rats. Since
urinary thiocyanate is indicative of CEO formation, Pilon et al. (1988a) interpreted their results
as indicating that more CEO was formed after GSH-depletion.
Farooqui and Ahmed (1983a) also reported covalent binding of [2,3-14C]-AN to protein,
DNA, and RNA of tissues of male Sprague-Dawley rats treated with a single oral dose of
46.5 mg/kg. DNA from tissue homogenate was isolated by extraction with chloroform/isoamyl
alcohol/phenol and application of the aqueous extract to hydroxy apatite chromatography. DNA
alkylation was higher in brain and stomach than that in the liver, with highest levels of covalent
binding in the brain. The covalent binding indices for the liver, stomach, and brain at 24 hours
after dosing were 5.9, 51.9, and 65.3, respectively.
Ahmed et al. (1992a) demonstrated the covalent binding of radiolabel from [2,3-14C]-AN
to testicular DNA after a single gavage of radiolabeled AN (46.5 mg/kg) to male Sprague-
Dawley rats. In a time course study, bound activity was shown to be greatest after 30 minutes
(8.93 ± 0.80 [j,mol AN bound per mol nucleotide). Using an identical experimental protocol,
Ahmed et al. (1992b) demonstrated the capacity of AN to bind covalently to DNA in the lung.
31
DRAFT - DO NOT CITE OR QUOTE
-------
Binding was associated with a 28-41% decrease in replicative DNA synthesis at time points up
to 24 hours after dosing.
Jacob and Ahmed (2003a) used whole-body autoradiography to examine the distribution
of [2-14C]-AN administered orally or intravenously to male F344 rats. Two days after oral
dosing, covalently bound radioactivity was retained at higher levels in the gastric mucosa, blood,
and hair follicles. After i.v. injection, covalently bound radioactivity was retained in liver,
spleen, bone marrow, adipose tissue, and lung. The amount of radioactivity associated with
covalent binding was lower for oral dosing than for i.v. injection, which was consistent with the
higher recovery of radioactivity excreted following oral dosing compared with injection (see
Section 3.4.2).
Covalent binding of [2,3-14C]-AN to tissue protein and globin was also studied in male
Sprague-Dawley rats after a single s.c. injection of 1.2-115 mg/kg (Benz et al., 1997b).
Covalent binding to tissue protein reached completion in 1-4 hours and was linear in the low
dose range (1.2-50 mg/kg), with the relative order (in descending order) as follows: blood >
kidney = liver > forestomach = brain > glandular stomach > muscle. Covalent binding to globin
followed a similar dose-response curve. Benz et al. (1997b) also measured an N-(2-cyanoethyl)-
valine (CEVal) adduct of globin at this dose range. This adduct was formed by reaction of AN
with the NEb-terminal residue of globin (Osterman-Golkar et al., 1994) and represented only
0.2% of total AN binding to globin. However, regression of tissue protein binding vs. globin
total covalent binding or globin CEVal adduct indicated that both globin biomarkers could be
used as surrogates for the amount of AN bound to tissue protein.
Using a similar dosing regimen, Nerland et al. (2001) employed sodium dodecyl sulfate-
polyacrylamide gel electrophoresis to separate labeled proteins isolated from subcellular
fractions of liver from treated rats. Binding of AN was found to be associated preferentially with
GST of the |i subclass (GSTM). Within this subclass, GSTM1 was labeled about seven times
more strongly than GSTM2, while, from the a-subclass, only GSTA3 was labeled (at about
1/35 the strength of GSTM1). No label was associated with GSTA1 or GSTA2. The site of
binding was identified as exclusively cysteine 86. Since this particular cysteine residue in
rGSTMl appeared to have been targeted specifically, the study authors hypothesized that high
reactivity at cysteine 86 was due to its potential interaction with histidine residue at position 84,
which would lower the pka of cysteine 86, increasing reactivity towards sulfhydryl reagents.
These data would suggest that tissue proteins containing cysteine residues with an abnormally
low pka value would be likely targets for AN. In an in vivo experimental approach, Nerland et
al. (2003) demonstrated that subcutaneously administered AN preferentially bound to the
cysteine 186 residue of carbonic anhydrase III (CAIII) in rat liver.
A considerable body of evidence demonstrated the ability of AN to bind to intercellular
proteins, in particular to Hb. Osterman-Golkar et al. (1994) reported a method for quantifying an
N-terminal cyanoethyl-valine adduct, CEVal, the product of reaction between AN and the
32
DRAFT - DO NOT CITE OR QUOTE
-------
N-terminal valine of Hb. The method was based on the N-alkyl Edman procedure involving the
derivatization of globin with pentafluorophenyl isothiocyanate and gas chromatography-mass
spectrometry analysis. Osterman-Golkar et al. (1994) showed that the method was applicable to
experimental animals exposed to AN in drinking water and to humans exposed to AN in tobacco
smoke. Hb from smokers (10-20 cigarettes/day) with a daily intake of 0.5-5 jag AN per kg BW
contained about 90 pmol CEVal/g, whereas adduct levels in the Hb of nonsmokers were below
the detection limit of about 20 pmol/g Hb. This utility was confirmed by Tavares et al. (1996)
for occupationally exposed workers and for smokers.
A subsequent study by Bergmark (1997) showed that the Hb of smokers contained
adducts of ethylene oxide and acrylamide as well as AN. In nonsmokers, the CEVal Hb adduct
of AN was below the detection limit of <2 pmol/g of globin. In the 10 smokers studied, the
levels of this adduct ranged from 25 to 178 pmol/g (mean 106 pmol/g) and correlated with the
number of cigarettes smoked per day (correlation coefficient = 0.94).
Fennell et al. (2000) determined CEVal from blood samples of 16 nonsmokers and
32 smokers and reported that CEVal Hb adducts increased with increased cigarette smoking.
The estimated CEVal level from smoking was 170 fmol per mg Hb per pack-day. Two
participants in a smoking cessation program showed a gradual reduction of CEVal levels (Perez
et al., 1999). Thus, the use of the Hb adduct CEVal may have utility as a biomarker to assess
low-level exposure to AN (in the region of 50 ppb), even in a complex mixture of toxicants,
although smoking would be a confounding variable. Fennell et al. (2000) also reported that the
null genotypes for GSTM1 or GSTT1 had little effect on CEVal levels when compared to active
genotypes.
Borba et al. (1996) measured CEVal as a marker of AN exposure in occupationally
exposed workers in an acrylic fiber factory. The values for CEVal among the subjects were 8.5-
70.5 pmol/g globin in controls, 635.2-4,603.5 pmol/g globin for continuous polymerization
workers, and 93.9-4,746 pmol/g globin for maintenance workers. These findings pointed to the
ready formation of AN adducts with Hb in an occupational setting.
CEVal Hb adduct was also used as a follow-up dose monitor after accidental exposure of
four cleaning workers to AN in an AN-containing tank wagon (Bader and Wrbitzky, 2006). On
day 25 after exposure, CEVal adduct levels in Hb ranged from 640 (blood sample partly
hemolyzed) to 2,020 pmol/g globin for the three smokers and was 566 pmol/g globin for the
nonsmoker, indicating residual AN adducts from the accidental exposure. On day 175, CEVal
adduct levels were 81-276 pmol/g globin for the smokers and 2 pmol/g globin for the nonsmoker
and represented background CEVal of the study participants according to their smoking status.
For both the smokers and nonsmoker, the adduct concentrations in blood declined linearly with
time. Linear regression analysis of the data estimated a total elimination interval of 148 days,
longer than the standard lifespan of 126 days of erythrocytes. Linear regression analysis also
allowed estimation of the initial adduct levels of the workers on the day of the accident and
33
DRAFT - DO NOT CITE OR QUOTE
-------
estimation of the exposed AN concentrations based on the correlation from the former German
exposure equivalent that 3 ppm AN yields 17,200 pmol/g globin.
3.4. ELIMINATION
3.4.1. Studies in Humans
"3
In an inhalation study of six male volunteers exposed to 7.8-10 mg/m AN for 8 hours
(Jakubowski et al., 1987), 44-58% (mean = 52%) of the inhaled AN was absorbed, of which
about 22% of absorbed AN was metabolized and excreted in the urine over 31 hours from the
start of exposure as N-acetyl-S-(2-cyanoethyl)-L-cysteine (2-cyanoethyl mercapturic acid
[CEMA]). Elimination followed first-order kinetics, with a half-life of about 8 hours. The
authors concluded that individual urinary CEMA levels were not a useful measure for exposure
to AN.
3.4.2. Studies in Animals
Three pathways are involved in the elimination of AN from an organism: exhalation,
urinary excretion, or fecal excretion.
3.4.2.1. Exhalation
In a study by Young et al. (1977), male Sprague-Dawley rats were exposed to [1-14C]-AN
via inhalation (5 or 100 ppm) or a single oral dose (0.1 or 10 mg/kg). Exhalation of AN as CO2
within 6 hours after dosing decreased with increasing dose, from 6.1 to 2.6% of the dose
following inhalation exposure and from 4.6 to 3.9% after gavage (see Tables 3-1 and 3-2).
When Ahmed et al. (1983) administered a single oral dose of 46.5 mg/kg AN to male
Sprague-Dawley rats in distilled water, using 50 jj,Ci/kg of either [2,3-14C]- or [1-14C]-AN as
tracer, the recovery of total dose in expired 14CC>2 varied from 2% for [2,3-14C]-AN to 12% for
[1-14C]-AN 24 hours after dosing. Burka et al. (1994) administered 0.87 mmol/kg (46.2 mg/kg)
[2-14C]-AN by gavage to male F344 rats. About 2% of the dose was expired as volatile organic
components, predominantly unchanged AN, while 11% was liberated as 14CC>2 24 hours after
dosing.
Jacob and Ahmed (2003a) compared excretion of 11.5 mg/kg [2-14C]-AN administered
orally or intravenously to male F344 rats. In the 48 hours after oral dosing, 61% of the
radioactive dose was excreted, with 4% in exhaled CO2, 4% in urine, and 53% in feces.
Following i.v. administration, 30% of the dose was eliminated over 48 hours, with 2% in expired
air, 8%> in urine, and 21% in feces. In the 8 hours after oral dosing, 3% of the radioactive dose
was in exhaled CO2, 2% in urine, and 48% in feces. Following i.v. administration, 2% of the
dose was in exhaled CO2, 3% in urine, and 0.2% in feces. Jacob and Ahmed (2003a) concluded
that these results indicated a significant difference in biological fate of AN following i.v. or oral
treatment.
34
DRAFT - DO NOT CITE OR QUOTE
-------
3.4.2.2. Fecal Excretion
Young et al. (1977) also investigated the fecal excretion of AN in male Sprague-Dawley
rats. This route of excretion showed little dependence on the route of administration, amounting
to 3-4% of the dose following inhalation exposure and about 5% following gavage within
6 hours of dosing. Young et al. (1977) also investigated the possibility of biliary excretion with
subsequent reabsorption from the intestines in one male rat. A cannula was inserted between the
common bile duct and the duodenum for sampling of bile, and [14C]-AN was given
intravenously. Based on the concentrations of radioactivity in bile, urine, RBCs, and plasma as a
function of time, the rate of elimination from the bile was found to be faster than from other
fluids during the first 6 hours after dosage. The initial half-life for excretion of radioactivity via
bile was estimated to be about 15 minutes. One unidentified metabolite was excreted in the bile
and underwent enterohepatic circulation to the GI tract, contributing partially to 5% of the dose
excreted in feces.
Kedderis et al. (1993a) estimated that 3-5% of AN doses were excreted in feces of F344
rats (0.09-28.8 mg/kg orally), while 2-8% of the dose were recovered from the feces of B6C3Fi
mice (0.09-10 mg/kg orally). In either species, the differences in fecal excretion were not
related to the administered dose. Farooqui and Ahmed (1982) administered a single oral dose of
46.5 mg/kg [1-14C]-AN to male Sprague-Dawley rats, the highest percentage of 14C excreted in
feces (2%) occurred between 12 and 24 hours after dosing. At the end of 10 days, 2.5% of the
dose was excreted in feces. In another study, male Sprague-Dawley rats were given a single
gavage dose of 46.5 mg/kg [1-14C]-AN; four radioactive peaks were identified in biliary extracts
at 6 hours after treatment. The two major metabolites in bile were GSH conjugates of AN:
S-cyanoethyl glutathione and N-acetyl-S-(2-cyanoethyl)cysteine (Ghanayem and Ahmed, 1982).
Nearly 27% of the dose appeared in the bile after 6 hours (Ghanayem and Ahmed, 1982; Ahmed
et al., 1982). The GI tract contained the highest level of radioactivity up to 72 hours, suggesting
resecretion of AN metabolites to the stomach or binding of metabolites to the stomach mucosa
(Ahmed et al., 1982).
3.4.2.3. Urinary Excretion
A substantial number of studies demonstrated the rapid urinary elimination of AN or its
metabolites when AN was administered to experimental animals via the oral or inhalation routes
(Burka et al., 1994; Fennell and Sumner, 1994; Kedderis et al., 1993a; Ahmed et al., 1983, 1982;
Young et al., 1977). For example, when Young et al. (1977) exposed male Sprague-Dawley rats
to radiolabeled AN via inhalation (5 or 100 ppm) or a single oral dose (0.1 or 10 mg/kg), most of
the radiolabel was recovered in the urine, with much lower proportions of the initial dose in feces
or expired air (see Table 3-1 and 3-2). Urinary excretion increased with dose, from 69 to 82% of
35
DRAFT - DO NOT CITE OR QUOTE
-------
the dose following inhalation exposure and from 34 to 67% after gavage (Tables 3-1 and 3-2).
Excreted dose in urine was mostly unidentified metabolites.
In another study, Sapota (1982) administered 40 mg/kg AN, containing either 40 j_iCi/kg
[1,2-14C]- or [1-14C]-AN, in saline to male Wistar rats, either via gavage or intraperitoneally. In
parallel to the depletion of radiolabel in tissues at 24 hours postexposure, 82-93% of the dose
was eliminated from the body in the urine, with 3—7% exhaled unchanged in the breath in
24 hours, independent of the route of administration and the position of the radiolabel. However,
when Farooqui and Ahmed (1982) administered an oral dose of 46.5 mg/kg [1-14C]-AN to male
Sprague-Dawley rats, 40% of the radioactivity was excreted in urine over 24 hours after dosing.
At the end of 10 days, 61% of the total dose was excreted in urine. Similarly, when Ahmed et al.
(1983) gave a single oral dose of 46.5 mg/kg AN to male Sprague-Dawley rats in distilled water,
using 50 [j,Ci/kg of either [2,3-14C]- or [1-14C]-AN as tracer, 40% of the radioactivity from
[1-14C]-AN but 60% of [2,3-14C]-AN-derived radiolabel were detected in the urine in the initial
24 hours. Neither of the two studies provided details that might explain the discrepancy.
However, the observed discrepancy might indicate the impact of different strains used in the
studies.
The single gavage dose experiment of Burka et al. (1994), in which male F344 rats were
placed in metabolic cages after receiving 0.87 mmol/kg (46.2 mg/kg) [2-14C]-AN, resulted in
67%) of the load being voided to the urine after 24 hours. This proportion of the recovered load
was similar to the 55-56%) value obtained when male F344 rats or B6C3Fi mice were orally
exposed to [1,2,3-13C]-AN (Fennell and Sumner, 1994). However, Ahmed et al. (1996a)
recovered only about 4% of the counts in 48-hour urine samples when F344 rats were injected
intravenously with 11.5 mg/kg [2-14C]-AN. About 27% of the load was eliminated via all routes
combined. In this study, much higher levels of tissue binding were observed than in other
studies.
Identification of the urinary metabolites derived from AN metabolism was attempted in
several studies. Langvardt et al. (1980) administered [2,3-14C]- and [1-14C]-AN orally to male
Sprague-Dawley rats, thereby specifically tracing the fate of the vinyl and cyano groups of AN.
Two main urinary metabolites, thiocyanate and N-acetyl-S-(2-cyanoethyl)cysteine, were
identified. Thiocyanate, formed from the epoxide metabolite CEO, was the predominant urinary
metabolite following oral dosing with 30 mg/kg of [1-14C]-AN and accounting for 54% of the
injected radioactivity within 16 hours after dosing. On the other hand, thiocyanate only
accounted for 1% of the urinary radioactivity after dosing with [2,3-14C]-AN. The second
metabolite, N-acetyl-S-(2-cyanoethyl)cysteine, a mercapturic acid derived from direct
conjugation between AN and GSH, constituted 18% of the radiolabel derived from [1-14C]-AN
and 28%) from [2,3-14C]-AN. Another metabolite, tentatively identified as N-acetyl-3-carboxy-
5-cyanotetrahydro-l,4-2H-thiazine, constituted 19% of the label from [1-14C]-AN and 35% from
[2,3-14C]-AN. Evidently, this metabolite was formed from conjugation of CEO with GSH.
36
DRAFT - DO NOT CITE OR QUOTE
-------
Langvardt et al. (1980) found another four minor metabolites that were not identified
structurally.
Tardiff et al. (1987) monitored the urinary metabolites of AN-treated male Sprague-
Dawley rats 24 hours after i.v. or i.p. single doses (0.6, 3, or 15 mg/kg) or inhalation exposure to
4, 20, and 100 ppm for 6 hours. Three major metabolites were measured: thiocyanate, N-acetyl-
S-(2-hydroxyethyl)cysteine, and N-acetyl-S-(2-cyanoethyl)cysteine, that were also detected by
Langvardt et al. (1980). These three excreted metabolites represented 50-54% of the
administered dose in urine within 24 hours, independent of the dose. Tardiff et al. (1987) found
that following i.p. or i.v. administration of AN, N-acetyl-S-(2-hydroxyethyl)cysteine and
thiocyanate each represented between 5 and 10% of the urinary metabolites; the major
metabolite was N-acetyl-S-(2-cyanoethyl)cysteine, representing 74-78%) of the urinary
metabolite when AN was administered intraperitoneally or intravenously. However, following
inhalation exposure of AN, only 8%> of the dose was excreted as N-acetyl-S-(2-cyanoethyl)-
cysteine, and the major urinary metabolite was thiocyanate at 20 and 100 ppm. Moreover,
N-acetyl-S-(2-hydroxyethyl)cysteine was excreted in larger amounts than N-acetyl-S-(2-cyano-
ethyl)cysteine. These results also showed that the percentage of dose excreted as urinary
thiocyanate increased with dose when AN was administered by inhalation. The study authors
concluded that the route of administration of AN had an important influence on the pattern of
metabolic excretion.
Shibata et al. (2004) investigated the urinary excretion of thiocyanate in male Wistar rats
that received 40 mg/kg AN (about half the LD50) by gavage in water. Urinary excretion of
thiocyanate became measurable at the time of peak plasma thiocyanate levels, 5 hours after
dosing. Excretion of urinary thiocyanate gradually increased so that at 10 hours after dosing,
about 1.2 mg thiocyanate (7%> of administered dose) had been excreted into urine.
Gut et al. (1981) administered [1-14C]-AN to male Wistar rats (dose not given) by the
oral, i.v., i.p., and s.c. routes. Total excretion of radioactivity after 48 hours was reported to be
close to 100%) following oral administration but 75-84%) following the other routes of
administration. The patterns of urinary radioactivity elimination were also different: after
parenteral administrations, elimination of radioactivity was highest within the first 4 hours after
dosing, with much smaller amounts for the remaining 44 hours. After oral dosing, between
6 and 8%> of the dose was eliminated in urine for the time periods 4, 8, and 12 hours after dosing
with much lower amounts thereafter. However, the major difference was found to be urinary
thiocyanate elimination: 23%> during the 48 hours after oral dosing and 4%> following i.p., 4.6%>
following s.c., and 1.2% following i.v. administration.
In another study (Gut et al., 1985), male Wistar rats were exposed via inhalation to 57,
125, or 271 mg/m AN for 12 hours. A constant ratio between thiocyanate and the sum of
thioether compounds (AN mercapturic acids) in urine was found throughout the three doses.
Average total amounts of thioethers excreted in urine during 12 hours of exposure were 24, 63,
37
DRAFT - DO NOT CITE OR QUOTE
-------
or 83 [j,mol/kg for 57, 125, or 271 mg/m AN, respectively. Average total amounts of
-3
thiocyanate excreted were 14, 24, and 49 [j,mol/kg for 57, 125, and 271 mg/m AN, respectively.
The ratio of thioethers to thiocyanate was about 2:4 during the 12 hours of exposure and was
similar to that from oral exposure to AN. Thus, thioethers, not thiocyanate, were the major
urinary metabolites following inhalation and oral exposure. These results were different from
those reported by Langvardt et al. (1980) and Tardiff et al. (1987).
Miiller et al. (1987) quantified four urinary metabolites and unchanged AN following
inhalation exposure of male Wistar rats to 1-100 ppm of AN for 8 hours (thiocyanate was not
measured). At 24 hours postexposure, N-acetyl-S-(2-cyanoethyl)cysteine was the primary
urinary metabolite, followed by N-acetyl-S-(2-hydroxyethyl)cysteine, thiodiglycolic acid (also
known as thiodiacetic acid), S-carboxyethyl-L-cysteine, and unchanged AN. The excretion
pattern of AN and its metabolites was dependent on the inhalation exposure concentrations. The
study authors proposed cyanoethyl mercapturic acid was the most sensitive indicator metabolite
of AN exposure at levels of 5 ppm.
Fennell et al. (1991) also measured GSH-derived metabolites, but not thiocyanate, in the
urine of male F344 rats (10 or 30 mg/kg) and male B6C3Fi mice (10 mg/kg), following oral
13
administration of [1,2,3- C]-AN. The results of this study are shown in Table 3-7. N-acetyl-
S-(2-cyanoethyl)cysteine was formed by conjugation of AN with GSH, whereas the other
metabolites were from reaction of CEO with GSH. The results support the finding that rats and
mice differ in the way they metabolize AN, and mice evidently metabolize more AN through the
CYP2E1-mediated formation of CEO.
Table 3-7. Urinary excretion of thioethers derived from AN
Metabolite (see Figure 3-2)
Rat (30 mg/kg)
Mouse (10 mg/kg)
Percent of total metabolites
N-acetyl-S-(2-cyanoethyl)cysteine
42.8 ±4.8
20.5 ±2.0
N-acetyl-S-(2-hydroxyethyl)cysteine
26.7 ± 1.8
22.3 ± 1.2
N-acetyl-S-(l-cyano-2-hydroxyethyl)cysteine
17.4 ±2.2
13.9 ± 3.2
N-acetyl-S-(carboxymethyl)cysteine and thiodiacetic acid
7.4 ±2.9
43.2 ±3.5
Thionyldiacetic acid
5.7 ±0.21
Not detected
Source: Fennell et al. (1991).
Kedderis et al. (1993a) studied the dose dependence of the urinary excretion of AN
metabolites in male F344 rats and male B6C3Fi mice. In rats during the 72 hours following oral
doses of 0.09-28.8 mg/kg [2,3-14C]-AN, 73-99% of the dose was excreted in urine, while 3-5%
was found in feces. In mice receiving 0.09-10 mg/kg [2,3-14C]-AN, 83-94% of the dose was
excreted in urine and 2-8% in feces. Excretion of radioactivity by both routes was not dose
dependent in either species.
38
DRAFT - DO NOT CITE OR QUOTE
-------
The position of the radiolabel did not allow detection of thiocyanate. Radiochromato-
grams of urine from rats or mice identified five major peaks, two of which contained more than
one compound. Following administration of [2,3-14C]-CEO, two of the five peaks were not
found, indicating that those peaks were derived from direct conjugation of AN without metabolic
activation to CEO. The sum of the percent of total radioactivity from CEO conjugate-derived
peaks was higher than that from the AN conjugate-derived peaks in both rats and mice. None of
the metabolites appeared to be glucuronides. Kedderis et al. (1993a) detected four of the five
metabolites shown in Table 3-7, but could not identify thionyldiacetic acid. In addition,
S-(2-cyanoethyl)-thioacetic acid was detected in mouse urine only, likely a degradation product
of N-acetyl-S-(2-cyanoethyl)cysteine.
The excretion of metabolites derived from CEO was approximately linear with the AN
dose in both rats and mice. However, the urinary excretion of N-acetyl-S-(2-cyanoethyl)cysteine
increased nonlinearly with increasing dose of AN, an effect that was much more pronounced in
rats than in mice. This probably indicated the presence of a competing pathway, namely,
epoxidation of AN, with the conjugation of AN with GSH. The fraction of the total dose
recovered as metabolite from CEO was 0.5 in rats and 0.67 in mice. The study authors estimated
that the ratio of AN epoxidation to GSH conjugation ranged from 4.8 to 1.3 as the AN dose
increased in rats and from 5.7 to 2.7 as the dose increased in mice. Kedderis et al. (1993a) also
pointed specifically to the species differences detected in this study—a roughly 10-fold higher
excretion of thiodiglycolic acid (thiodiacetic acid) in mice as compared with rats and measurable
excretion of S-(2-cyanoethyl)thioacetic acid in mice. This urinary metabolite could not be
detected in rats at all.
Sumner et al. (1999) treated three male WT and four male CYP2El-null mice
(C57BL/GN x Svl29) orally to 0, 2.5, or 10 mg/kg [1,2,3-13C]-AN and usedNMR spectroscopy
to characterize AN metabolites in urine samples collected over 24 hours. WT mice excreted
metabolites derived from CEO (80-85% of total excreted) and from direct GSH conjugation with
AN (15-21% of total excreted), with the largest percentage of metabolites from conjugation of
GSH with the 3-carbon of CEO. CYP2El-null mice displayed only metabolites derived from
direct GSH conjugation with AN in their urine following administration of 2.5 or 10 mg/kg
[1,2,3-13C]-AN. This confirmed the role of CYP2Elin the oxidation of AN to CEO and its
subsequent transformation to a range of other products. Since CYP2El-null mice did not excrete
metabolites that would be produced by oxidation by other CYP450s, CYP2E1 may be the only
CYP450 enzyme involved in the metabolism of AN. In addition, CYP2El-null mice excreted
about the same percentage of administered dose as the WT mice, indicating CYP2El-null mice
compensated for the CYP2E1 deficiency by producing more metabolites from direct conjugation
of AN with GSH.
Taken together, the animal data indicate that AN can be exhaled as parent compound at a
low percentage of the administered dose, probably increasing at high doses (e.g., 10 mg/kg).
39
DRAFT - DO NOT CITE OR QUOTE
-------
Fecal excretion amounts to about 5% of a given dose. The biliary pathway leading to fecal
excretion has not been well characterized. To a small extent, similar to fecal excretion, AN is
metabolized completely and exhaled as CO2. Both pathways appear to be quite independent of
the administered dose. Urinary excretion of AN metabolites has been well characterized but is
not without contradictory findings. There appears to be little doubt that mice metabolize more
AN via the CYP2E1-mediated oxidative pathway than do rats. To what extent this pathway is
likely to be overcome by large doses, or what contribution the GSH conjugation makes with
varying doses of AN and different species exposed, is not known.
3.5. PHYSIOLOGICALLY BASED TOXICOKINETIC MODELS
A physiologically based toxicokinetic (PBTK) model for AN was previously developed
in rats (Kedderis et al., 1996; Gargas et al., 1995) and extended to describe the dosimetry of both
AN and CEO in humans (Sweeney et al., 2003). The PBTK model structure consists of two
parallel modules, one for AN and one for CEO, interlinked by the rate of oxidative metabolism
of AN to CEO in the liver, essentially as described by Gargas et al. (1995). Each module
consists of seven dynamic tissue compartments representing the lung, slowly perfused tissues,
fat, well-perfused tissues, brain, stomach, and liver (Figure 3-2). All perfusion-limited tissue
compartments are linked through blood flow, following an anatomically accurate, typical,
physiologically based description (Andersen, 1991).
40
DRAFT - DO NOT CITE OR QUOTE
-------
AN CEO
Inhalation
Drinking
water
Metabolism
> Metabolism
Arterial
blood
Venous
blood
Venous
blood
Arterial
blood
Brain
Brain
Stomach
Stomach
Other well
perfused
tissue
Liver
Lung
Other well
perfused
tissue
Slowly
perfused
tissue
Lung
Slowly
perfused
tissue
Fat
Fat
Liver
Source: Sweeney et al. (2003).
Figure 3-2. Structure of the PBTK model for AN and CEO.
Because AN and CEO are retained by the tissue in each compartment according to their
tissue/blood partition coefficients (PCs) (which were measured in vitro), the concentrations of
both chemicals in venous blood (leaving the tissue) are lower than those in arterial blood during
the equilibration phase (except CEO in the liver). Therefore, the rate of change in the amounts
of both chemicals in each tissue compartment is given by the difference between concentration in
blood entering and exiting the tissue multiplied by the blood flow. Simple differential equations
for each compartment, corrected for the non-enzymatic conjugation with GSH, are integrated
over time, giving the amounts of AN and CEO present in the tissue (Kedderis et al., 1996) (see
equation below). Therefore, knowing (from the literature) the actual volume of each tissue,
concentrations of AN and CEO in each tissue can be calculated over time.
41
DRAFT - DO NOT CITE OR QUOTE
-------
dA/dT = 0,fCA CV,J (Kso x CV, x GSH) x V,
where
Qi = blood flow rate to the target tissue (i)
CA = concentration of AN in arterial blood
CVi = concentration of AN in venous blood
Kso = rate constant for the conjugation of AN with glutathione (GSH)
J7i = volume of the target tissue (i)
For the lung compartment, with two mass inputs (mixed venous blood and inhaled air)
and two outputs (arterial blood and exhaled air), the amount of either chemical in alveolar air is
in equilibrium with the amount in lung blood at the steady state. Thus, concentrations of AN and
CEO in arterial blood can be calculated from simple mass balance equations. Such calculations
take into account the alveolar ventilation rate and the rate of blood flow through the lung, a
parameter made equal to cardiac output (both known from the literature) and corrected for
binding to Hb and blood sulfhydryls.
For the liver compartment, with mass input from blood and two outputs (venous blood
and metabolism, excluding biliary excretion that was not considered), the chemical mass transfer
is given by the difference between concentrations in portal and venous blood multiplied by
hepatic blood flow and corrected for: (1) metabolism of AN to yield CEO (calculated from the
Michaelis-Menten equation, subtracted from the mass of AN, and added to the mass of CEO),
(2) enzymatic hydrolysis of CEO (also described by the Michaelis-Menten equation), and (3) the
first-order conjugation with GSH. A simplified scheme of the mass flow in the PBTK model for
AN and CEO is shown in Figure 3-2 (Sweeney et al., 2003).
The model was initially calibrated in rats for three routes of AN administration—oral, i.v.
(bolus), and inhalation (Kedderis et al., 1996)—by manually adjusting the metabolic parameters
Vmax and Km for AN oxidation, first-order constants for AN- and CEO-GSH conjugation, and
first-order constant for absorption of AN from the GI tract, guided by the approximate statistical
likelihood calculation of SimuSolv. Absorption through the skin, which has been estimated at
about 1% of that through the lung (Rogaczewska, 1975), was not considered in this model.
Chemical-specific modeling parameters were either: (1) measured in vitro in rats (PC [Teo et al.,
1994]; macromolecular interaction constants [Gargas et al., 1995]), (2) fitted to the experimental
data by the model (metabolism parameters: Vmax and Km), or (3) estimated from the literature
(CEO elimination constants).
The model assumed that rats do not have EH activity, based on data from Kedderis and
Batra (1993). As discussed previously in Section 3.3.1, this assumption is probably incorrect,
based on data from de Waziers et al. (1990), Guengerich et al. (1981), and Kopecky et al. (1980).
This PBTK model did not scale allometrically EH activity to humans, whose liver microsomes
display EH activity toward CEO (Kedderis and Batra, 1993). Instead, Kedderis (1997) estimated
42
DRAFT - DO NOT CITE OR QUOTE
-------
this activity in humans in vivo based on the ratio of EH to CYP450 activity in subcellular hepatic
fractions multiplied by the CYP450 activity in vivo. In addition, the human blood-to-air PC for
AN was determined experimentally (Kedderis and Held, 1998). Because no in vivo
pharmacokinetic data were available to validate the human model, human in vitro data were
scaled to experimental data obtained from rats by using a "parallelogram approach."
The model by Kedderis et al. (1996) overestimates the oral exposure of venous blood
CEO concentration by 3- to 10-fold. This is a systematic bias, which suggests that the model
parameters need to be reevaluated or that the model is not capturing an important kinetic process
for CEO clearance. In particular, for all tissues except the liver, the measured rise in CEO
concentration for the first 10-15 minutes after oral and i.v. exposures is much slower than
predicted by the model, and the model continues to overpredict the oral data for all measured
time points, with the overprediction in the brain being the worst. Kedderis et al. (1996)
suggested that the overestimation of CEO concentration in blood at early time points is "most
likely due to a large intrahepatic first-pass effect." This hypothesis has not been further
evaluated, but, even if correct, the model structure has not been revised to simulate this
phenomenon. The model continues to overestimate CEO levels in blood (albeit to a lesser
degree) at later time points when the hypothesized first-pass effect would be less of a
contributing factor.
It should also be noted that Kedderis et al. (1996) did not compare their model
predictions to data on the urinary elimination of AN-GSH and CEO-GSH conjugates and on total
activity bound to Hb, as was done in a previous version of the model (Gargas et al., 1995).
While fits to those data may have informed the parameters reported by Kedderis et al. (1996),
this is clearly not the case in this instance. Also, the blood and tissue time-course data shown by
Kedderis et al. (1996) do not include all of the points shown in the previous publication.
Since Kedderis and colleagues (1996) did not include EH activity in their rat model (but
other evidence strongly indicates significant CEO hydrolysis be EH), did not perform a more
global, numerical optimization of their parameters, may not have included the urinary and Hb
data, and the subsequent model fits consistently overpredicted CEO pharmacokinetics in blood
and brain to a large extent, the model parameterization was revised to hopefully improve the
model characterization of the rat pharmacokinetic data and provide a more sound footing for
human-rat dosimetry comparisons and hence risk extrapolation, at least making the results
numerically reproducible. Guengerich et al. (1981) measured CEO hydrolysis with purified rat
EH and obtained a Vmax of 0.3 [j.mol/minute-mg protein and a Km of 800 [jM. Since the highest
blood levels observed for CEO were ~2 |iM (0.1 (j,g/mL), to a good approximation hydrolysis
can then be described using a first-order rate constant of 0.3/800 = 3.75 x 10"4 L/minute-mg EH
protein. The EH content of rat liver was measured by de Waziers et al. (1990), who found
0.165 mg EH/mg MP and a value of 40 mg MP/g liver from Ploemen et al. (1997) can be
applied.
43
DRAFT - DO NOT CITE OR QUOTE
-------
Finally, the liver fraction for the rat as used in the model 40 g/kg BW was applied,
assuming a standard 0.25 kg rat. The rate constant for a standard rat would then be l
-------
the range of those used in the PK studies. Since GSH depletion will be least significant at the
lowest doses used, and the most interest is in the low-dose range for extrapolation, it was decided
to reestimate the model parameters by using only the lowest concentration pharmacokinetic data:
the oral doses at or below 3 mg/kg, i.v. dose of 3.4 mg/kg, and inhalation concentration of
186 ppm. Also, of the data used to estimate parameters, it was determined that the last AN blood
measurement from both the 3 mg/kg oral data and the 3.4 mg/kg i.v. data appeared to be outliers
since they were well above the otherwise log-linear clearance curve defined by the preceding
data, so they too were not used in the estimation. (Model simulations of all exposure
concentrations and doses will be shown as compared with the data, but only the data for the
exposure values stated two sentences above were used for fitting.)
The following figures show the model fits obtained with the parameters of Kedderis et al.
(1996) as compared with the EPA's revised parameters. (All parameter values are listed in Table
C-l of Appendix C.) The term "fits" is used here to describe the closeness of model simulations
to the data, recognizing that only a subset of the data, as described in the preceding paragraph
and indicated in the figure legends, was actually used in parameter estimation.) The most
remarkable aspect of the revised fits is how close they are to identical to the original model fits.
Results are nearly identical for most of the i.v., inhalation, and oral (Figures 3-3 to 3-5a) blood-
and tissue-time-course data. The revised fits to the CEO blood- and tissue-time-course data after
oral dosing are slightly worse than in the original model (Figure 3-6b), but then the fits to the
urine and especially the Hb-binding data (Figure 3-6c) are considerably better than those
obtained with the parameters of Kedderis et al. (1996). Since only a few of the fits were
degraded slightly, while others were improved, the revised model parameters are considered to
represent the overall data set at least as well as the original. The fact that the revised parameters
are specifically defined by the low-dose data, which are closest to the range where the model will
be applied and are obtained using an objective criteria and numerically reproducible methods
(the approximate likelihood function with numerical optimization), gives further support for their
use.
45
DRAFT - DO NOT CITE OR QUOTE
-------
0.3
0.1
0.03
0.01
o
E 0.003
c
o
IS
0.001
Kedderis et al. 1996
- 4
CEO in blood
K M
1+
XX
O -x,X
+4
++
-------
Kedderis et al. 1996
Revised parameters
CEO in b ood
\.o +
1
CEO in liver
+
" V o #
+
186 pprn data
186 ppm fit
o
254 ppm data
-254 ppm fit
+
291 ppm data
•291 ppm fit
i—[
'
-------
Kedderis et al. 1996
Revised parameters
AN n blood
* BOn
+ 186 ppm data
186 pprn fit
o 254 pprn data
254 ppm fit
+ 291 pprn data
291 DDtTl fit
AN in liver
!\ v +
3.4 2.8 3
Time (h)
Note: Only 186 ppm data were used to estimate revised parameters.
Figure 3-4b. Inhalation exposure, AN concentrations.
48
DRAFT - DO NOT CITE OR QUOTE
-------
0.1
0.01
0.001
=s, 0.1
O)
E
E
o
ta
Kedderis et al. 1996
1 ^ 1
+ +
+
Revised parameters
nr.
CEO in blooc
CEO in blood
o
+
c
a>
o
C
O
o
0.01
0.001
0.1
~ uz
CEO in liver
/-¦
Q __,
O O
OO
0.01
0.001
+ 3 mg/kg data
3 mg/kg fit
o 10 mg/kg data
10 mg/kg fit
+ 30 mg/kg data
30 mg/kg fit
r ocroo'
i
CEO in brain
o
o
o
o
o
o
o
o
o
o
0
0.2
0.4
0.6 0
Time (h)
0.2
Note: Only 3 mg/kg data were used to estimate revised parameters.
Figure 3-5a. Oral exposure, CEO concentrations.
0.4
0.6
49 DRAFT - DO NOT CITE OR QUOTE
-------
30
10
3
1
0.3
10
Kedderis et al. 1996
Revised parameters
O)
c
o
"ro
1
c
Q>
c 0.3
o
a
10
3
1
0.3
0.1
+ +
AN in liver
AN in brain
AN in blood
+ 3 mg/kg data
3 mg/kg fit
o 10 mg/kg data
--10 mg/kg fit
+ 30 mg/kg data
3o mg/kg fit
0
0.2
0.4
0.6 0
Time (h)
0.2
Note: Only 3 mg/kg data were used to estimate revised parameters.
Figure 3-5b. Oral exposure, AN concentrations.
0.4
0.6
50 DRAFT - DO NOT CITE OR QUOTE
-------
Kedderis et al. 1996
Revised parameters
CM
C
"D
C
3
O
.Q
"D
a>
a>
u
X
d>
+¦>
c
3
o
E
<
12
9
6
3
1.6
1.2
0.8
0.4
0*
0
2b
I
20
-
15
AN-GSH in urine
¦
10
+
+
5
O)
+
E
^ oJ
H^<+
" ¦
o
o
CEO-GSH
in urine
o
.
r
x ~~~
J* .
Hb binding
10
20 30 0
Dose (mg/kg)
10
20
30
Note: points are data; lines are model fits or simulations. Only data for doses
<2 mg/kg were used to estimate revised parameters.
Figure 3-5c. Oral exposure, urinary excretion, and Hb binding.
Since the human metabolic parameters were extrapolated from in vitro measurements by
using the relationship between in vitro measured and in vivo estimated values from the rat
(Sweeney et al., 2003), it is appropriate to update the human metabolic parameters and model in
parallel with the EPA's update of those in the rat. For EH, Sweeney et al. (2003) used the in
vivo:in vitro relationship for the CYP450-mediated metabolism, since there had been no parallel
relationship for hydrolysis. As we are now choosing to extrapolate the rat EH from in vitro to in
51
DRAFT - DO NOT CITE OR QUOTE
-------
vivo based only on enzyme content, microsomal content, liver fraction, and BW, a parallel
approach for the human would be to do the same, rather than using a correction factor based on a
different class of enzymes. In the case of humans, Kedderis and Batra (1993) measured EH
activity in vitro by using hepatic microsomes, so we do not need to use the amount of enzyme
per mg MP in the calculation. Kedderis and Batra (1993) determined a Vmax and Km EH-
mediated hydrolysis of CEO using liver samples from six individual humans. The lowest
estimated Km in the group was 600 [xM, so again, we choose to describe the metabolism as first
order, using the ratio of Vmax/Km. The ratio of Vmax/Km was first calculated for each individual
since Vmax and Km tend to be statistically correlated due to the way they are estimated, and an
average value for the ratio was then determined to be 7.02 x 10"6 L/minute-mg MP. We can then
apply the value of 56.9 mg MP per g liver from Lipscomb et al. (2003), the liver fraction of
25.7 g/kg BW, and the standard value of 70 kg for a human. The rate constant for a standard
human is then k£H = (7.02 x 10"6 L/minute-mg MP) x (56.9 mg MP/g liver) x (25.7 g liver per kg
BW) x (70 kg BW) x (60 minutes/hour) = 43.1 L/hour, assuming it also scales as BW0'7, kEHC =
0 7 0 7
kEH/(0.70 kg) ' = 2.20 L/hour-kg ' . Recall that the value for rats was estimated to be
3.92 L/hour-kg0'7.
The original PBTK model for humans was assessed for its sensitivity to changes in key
input parameters, and the expected variability in CEO concentrations in humans under different
AN exposure scenarios was estimated (Sweeney et al., 2003). In addition to updating the CEO
hydrolysis rate constant for the human model (and using a first-order equation for that reaction)
as previously discussed, the ratio of Vmax for the oxidation step as estimated in vivo vs. measured
in vitro for the rat was used to estimate the human in vivo Vmaxc, and the enzymatic GSH
conjugation rate constants for AN (kFc) was likewise estimated from the rat estimated in vivo vs.
measured in vitro ratio (kFC2 was unchanged during the reestimation). Since the rat values for
Vmaxc and kFc were revised, and these values were each one leg of the metabolic extrapolation
"parallelogram," the human in vivo values should be accordingly varied. The result is that the
ratio of revised/original human values for each of these constants is simply equal to the
respective revised/original in vivo values for the rat constants; the revised human values are
0 7 0 3
Vmaxc = 22.1 mg/hour-kg ' and kFc = 77 kg ' /hour. Finally, since the original model
extrapolation assumed that the oral absorption constant, kA, and the Hb-binding constants, kH and
kjj2, were the same in humans as in rats, these assumptions were retained, updating the human
parameters appropriately.
Appendix C provides model source codes written in acsXtreme (AEgis Technologies,
Huntsville, AL) and Matlab (The Mathworks, Inc., Natick, MA) that were used to model
AN/CEO pharmacokinetics with the PBTK models of Kedderis et al. (1996, as revised) and
Sweeney et al. (2003) in this evaluation.
52
DRAFT - DO NOT CITE OR QUOTE
-------
4. HAZARD IDENTIFICATION
4.1. STUDIES IN HUMANS—EPIDEMIOLOGY AND CASE REPORTS
4.1.1. Oral Exposure
No studies were identified that addressed the exposure of human beings to AN via the
oral route.
4.1.2. Inhalation Exposure
4.1.2.1. Acute Exposure
"3
In a study on six male volunteers, a single 8-hour inhalation exposure of 5-10 mg/m AN
produced no subjective symptoms, such as headache, nausea, or general weakness (Jakubowski
etal., 1987).
A report by Chen et al. (1999) collated 144 acute AN poisoning cases that had occurred
between 1977 and 1994 in China. Each case involved a brief workplace accidental exposure to a
high concentration of AN. There were few reliable data on the levels of exposure in these cases,
"3
although transient concentrations of AN were thought to range from 40 to 560 mg/m (18-
"3
258 ppm) for 60 cases and may have been over 1,000 mg/m for the remaining 84 cases. All but
9 of the 144 subjects were males, ranging in age from 18 to 53 years old. Forty-two of the cases
were considered to have resulted in severe acute AN poisoning, while the rest fell into the mild
acute category.
Table 4-1 summarizes the incidence of symptoms and signs of toxicity that were evident
in the 144 cases. Other changes in monitored biochemical or physiological parameters included
transient increases in peripheral white blood cell (WBC) count to greater than 10 x 109/L in
66 cases and a number of apparent fluctuations in clinical chemistry parameters. Out of
120 subjects whose liver function was monitored, seven subjects showed abnormal increases in
alanine aminotransferase (ALT) (slight), aspartate aminotransferase (AST), and cholylglycine
(i.e., glycocholic acid). Up to 23 times the normal serum concentration of the latter compound
was detected. The seven subjects with abnormal liver functions were poisoned at low
concentrations in air (40-79 mg/m [18-36 ppm]) but for a comparatively long duration of acute
exposure (36 hours). Blood levels of GSH were depressed up to 50% in the severe cases, and
urine thiocyanate was increased up to fourfold.
53
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-1. Clinical signs in 144 subjects accidentally exposed to AN
Symptoms
Cases (%)
Symptoms
Cases (%)
Dizziness
144 (100)
Congestion of pharynx
105 (73)
Headache
144 (100)
Hoarseness
13(9)
Feebleness
144 (100)
Pallor
108 (75)
Sore throat
87 (60)
Profuse diaphoresis
95 (66)
Chest tightness
144 (100)
Rough breathing sound
18 (13)
Cough
16(11)
Rapid heart rate
36 (25)
Dyspnea
118(82)
ECG abnormalities3
15 (10)
Nausea
133 (92)
High blood pressure
20 (13)
Vomiting
95 (66)
Low blood pressure
5(4)
Abdominal pain
97 (68)
Liver tenderness
9(7)
Numbness of the limbs
50 (40)
Hepatomegaly and splenomegaly
7(5)
Fainting
104 (72)
Coma
7(5)
Convulsion
46 (32)
Hyperactive knee jerk
137 (95)
aAll 15 instances of electrocardiogram (ECG) abnormalities occurred in cases of severe poisoning.
Source: Chen et al. (1999).
These changes disappeared after poisoned subjects were removed from the accident site
and underwent treatment that included amyl nitrite via inhalation and an i.v. injection of 3%
sodium nitrite followed by 50% sodium thiosulfate (STS). Other treatments included infusion
with glucose, adenosine triphosphate (ATP), coenzyme A, and DEX along with oxygen
inhalation to prevent the development of cerebral edema and to protect brain and liver cells.
In a study that was possibly reflective of combined inhalation and dermal exposure to
AN, Wilson et al. (1948) reported that workers exposed to AN vapors at 16-100 ppm (35-
"3
217 mg/m ) for periods of 20-45 minutes while involved in cleaning operations in polymerizer
facilities complained of dull headache, fullness in the chest, irritation of all mucous membranes
(eyes, nose, and throat), and a feeling of apprehension and nervous irritability. Some workers
also complained of intolerable itching of the skin but had no clinically demonstrable dermatitis.
Workers who had direct skin contact with AN displayed a sequence of symptoms, including skin
irritation and erythema, followed by bleb formation and then desquamation with slow healing
(Wilson et al., 1948). An earlier report by Wilson (1944) reported the onset of nausea, vomiting,
weakness, headache, fatigue, diarrhea, and an "oppressive feeling" in the upper respiratory
passages following exposure to "mild concentrations" of the compound. Several cases of
reversible mild jaundice and one case of slow-to-reverse severe jaundice were identified and
associated with occasional liver tenderness and low-grade anemia, but it was unclear whether
they were related to inhalation or dermal contact exposures to only AN or to a combination of
AN with other industrial rubber manufacturing chemicals such as butadiene and styrene.
54
DRAFT - DO NOT CITE OR QUOTE
-------
4.1.2.2. Chronic Exposure
4.1.2.2.1. Cancer epidemiology
Cohort studies
A retrospective cohort study involving workers at a DuPont textile fibers plant in
Camden, South Carolina, was the first epidemiologic study addressing the potential
carcinogenicity of AN (O'Berg, 1980). Shortly thereafter, reports from other cohorts of
AN-exposed workers, based in Germany, the Netherlands, the United Kingdom, and the United
States, were published. Other smaller-scaled occupational cohorts have also been utilized to
better assess the association between AN exposure and adverse human health effects.
Early studies did not utilize quantitative measures of exposure nor were they able to
control for smoking (an important consideration when lung cancer is an outcome of interest).
These factors were being taken into account by the late 1980s. The following section provides a
summary of some of the major cohort studies conducted both within and outside the United
States.
The DuPont studies
Researchers from DuPont conducted a series of studies among male workers potentially
exposed to AN (Symons et al., 2008; Wood et al., 1998; Chen et al., 1987; O'Berg et al., 1985;
O'Berg, 1980). The original study was conducted in South Carolina, with a follow-up study a
few years later. A second DuPont plant, located in Virginia, was the site for another cohort
study. Finally, Wood et al. (1998) combined the cohorts from South Carolina and Virginia to
further investigate the relationship between AN exposure and cancer. Most recently, Symons et
al. (2008) updated this combined cohort, adding 11 years of follow-up for a total follow-up
period of over 50 years.
The first epidemiologic study addressing the potential carcinogenicity of AN (O'Berg,
1980) examined cancer incidence and mortality in a cohort of 1,345 male workers from a DuPont
textile plant who were potentially first exposed to AN between 1950 and 1966. The cohort was
chosen from a larger group of more than 10,000 workers, based on examination of work history
(wage roll employees and some salaried employees), recollection of plant supervisors (wage roll
maintenance employees and salaried employees), and survey of salaried employees (salary roll
employees). No individual or area exposure monitoring data were available during the period of
study. Several surrogate measures of potential exposure were utilized in the analyses, including
job title, initial period of first exposure, length of exposure, and payroll (assuming that wage roll
employees have higher potential exposure than supervision or salary roll employees).
Workers were followed through the end of 1976, thus allowing a latency period of at least
10 years for all surviving members of the cohort. Cancer incidence was ascertained for all
cohort members who were diagnosed during their employment, and cause of death was
determined for all cohort members who died either while employed or after termination.
55
DRAFT - DO NOT CITE OR QUOTE
-------
Comparison rates for cancer incidence and mortality were derived from similarly collected data
in a company-wide cancer incidence and mortality registry system. Additional mortality
information was supplied by the Social Security Administration, and death certificates were
obtained for all known deaths. Other standard comparison rates were also utilized and presented;
however, the company-wide data represented the most comparable control group.
After observing only one incident case of cancer in the 217 employees with <6 months of
exposure, examination of the incidence data was concentrated on the 1,128 employees who had
more than 6 months of exposure. Among this group, cancer incidence was stratified by interval
of diagnosis: 1956-1964, 1965-1969, and 1970-1976. The cancer incidence was statistically
significantly elevated only in the last interval with 19 cases observed compared to 10.1 expected
(standardized incidence ratio [SIR] = observed ^ expected = 1.88, 95% confidence interval [CI]
= 1.17-2.88, p < 0.01), based on company-wide rates. Seventeen of the 19 cases were among
wage roll employees, the remaining two were salary roll employees. The difference between the
observed and expected number of cases among the wage employees was statistically significant
(SIR = 2.05, 95% CI = 1.23-3.21,/? < 0.01) (Table 4-2).
Table 4-2. Distribution of select incidence and mortalities among wage
workers and all workers at an AN plant
Wage workers
All workers
Cause of
disease/death
Observed
cases
SIR3
(CI)
Observed
deaths
SMRa
(CI)
Observed
cases
SIR3
(CI)
Observed
deaths
SMRa
(CI)
Less than 6 months of exposure
All cancers
1
0.77
(0.04-3.19)
3
1.30
(0.33-3.55)
1
0.67
(0.03-3.29)
3
1.25
(0.32-3.40)
Respiratory cancer
0
0.00
1
1.11
(0.06-5.48)
0
0.00
1
1.11
(0.06-5.48)
Greater than 6 months of exposure
All cancers
(1956-1976)
20
1.30
(0.82-1.97)
15
1.16
(0.68-1.87)
24
1.26
(0.82-1.85)
17
1.13
(0.68-1.78)
All cancers
(1970-1976)
17
2.05b
(1.23-3.21)
11
1.41
(0.74-2.45)
19
1.88b
(1.17-2.88)
13
1.44
(0.80-2.41)
Respiratory cancer
(1956-1976)
8
2.35b
(1.09-4.47)
6
1.30
(0.53-2.71)
8
1.95
(0.91-3.71)
7
1.35
(0.59-2.66)
Respiratory cancer
(1970-1976)
6
2.86b
(1.16-5.94)
5
1.61
(0.59-3.57)
6
2.40b
(0.97-4.99)
6
1.71
(0.69-3.57)
aSIR and SMR standardized against company-wide rates.
Statistically significant (p < 0.05).
SMR = standardized mortality ratio
Source: Amended from O'Berg et al. (1980).
56
DRAFT - DO NOT CITE OR QUOTE
-------
Eight incident respiratory cancer cases were observed in the cohort of 1,345 workers. All
eight of the respiratory cancers occurred in the group of employees who had >6 months of
exposure, and all were among wage roll employees. Six of the eight respiratory cancers occurred
in the 1970-1976 time interval compared to 2.1 expected based on company-wide rates (SIR =
2.86, 95% CI = 1.16-5.94,p < 0.05) (Table 4-2). The numbers were too small to perform any
other cancer-specific analyses.
In order to consider the effects of latency and level of exposure, further analyses were
performed looking at the group of workers who had first been exposed between 1950 and 1952,
when AN was first used at the plant and when exposures were known to have been highest.
Statistically significant differences between the number of observed and expected cases were
found among the wage roll employees in the diagnosis interval from 1970 to 1976, with
17 cancers observed and 5.6 expected (SIR = 3.04, 95% CI = 1.83-4.78), with six of these cancer
cases being respiratory cancers with an expected value of 1.5 (SIR = 4.0, 95% CI = 1.62-8.32).
This group is presumed to have the longest latency and highest exposures to AN.
Other factors related to significant differences in cancer incidence between observed and
expected cases among the wage roll employees during the time period 1970-1976 were:
(1) being a mechanic (10 observed vs. 3.7 expected), (2) duration of exposure of more than
10 years (9 observed vs. 2.8 expected), and (3) exposure above "low level" (13 cancers observed
vs. 5.5 expected), low level not defined. Similar relationships were found among respiratory
cancers except for duration of exposure of more than 10 years.
Incidence data among wage roll employees were examined by duration of exposure,
excluding exposures of <6 months. The duration categories were <5, 5-9, and >10 years. The
SIRs for all cancers in these three duration of exposure categories were 0.7 (95% CI = 0.23-
1.79), 1.2 (95% CI = 0.49-2.50), and 2.3 (95% CI = 1.18-4.14), respectively.
A total of 89 deaths was observed in the entire cohort of 1,345 workers by the end of
1976. The mortality analyses using company-wide rates for comparison did not show any large
deviations between observed and expected numbers for all cancers or for respiratory cancers
alone. Observed and expected numbers of cancer deaths between 1970 and 1976 for the wage
roll employees with >6 months of exposure were not markedly different, with 11 cancers
observed vs. 7.8 expected (standardized mortality ratio [SMR] = 1.41, 95% CI = 0.74-2.45) and
5 respiratory cancers observed vs. 3.1 expected (SMR = 1.61, 95% CI = 0.59-3.57) (Table 4-2).
The excess mortality observed is of interest, since statistical significance may have been
hindered by the small sample size.
The use of company-wide rates as a comparison may be problematic because the
company-wide rates are based on other plants where exposure to carcinogenic agents may be
possible. Additionally, company-wide rates include both exposed and unexposed workers, thus
weakening the ability to observe increased risk among the exposed workers. The use of this
population-based referent may hide true associations between AN and cancer. Additionally, the
57
DRAFT - DO NOT CITE OR QUOTE
-------
number of incident cancer cases and cancer-specific mortalities was too small for further
analysis. However, the fact that the cancer/respiratory cancer comparisons appeared to be
stronger for workers with longer latency, longer exposure, jobs with higher exposure, and work
during the early years of observation when exposures were highest added to the weight of the
evidence.
Another potential limitation of this study was under-ascertainment of incident lung cancer
cases. Two confirmed lung cancer cases that should have been included in the cohort were
omitted because of clerical error. A validity check of a subsample of 465 employees revealed
five more omissions, suggesting that a problem existed with ascertainment of lung cancer cases.
The omission of these cases may have resulted in the underestimation of the risk estimate for the
incidence of lung cancer. Also, subjective exposure assessment could have resulted in
misclassification of exposure. If the exposure misclassification was random, it would bias the
risk ratio toward the null value. If it was nonrandom because those assessing AN exposure
believed it could cause illness, the effect on the risk estimate is unpredictable. However, other
metrics used in this study, including duration of exposure, time of diagnosis, and job category,
had internal validity and supported the finding of an association between AN exposure and
cancer. Another limitation is the indication that workers may have been exposed to other
chemicals in addition to AN. Lastly, there was no adjustment for smoking. The EPA AN health
assessment (U.S. EPA, 1983) adjusted the results of this analysis based on smoking rates in the
plant and among U.S. blue-collar workers, estimating that the expected number of cancer cases
would be 15% higher and would reduce, but not eliminate, the association between AN and
cancer in this study. Finally, the small sample size and limited follow-up restricted the ability to
detect most cancers.
A second DuPont study extended the follow-up of the above cohort through 1983 for
cancer incidence and 1981 for cancer mortality, adding 7 and 5 years of follow-up, respectively
(O'Berg et al., 1985). The exposure assessment was not updated, and case ascertainment and
calculation of expected incidence and mortality were unchanged from the above study. To
evaluate dose-response relationships and latency, cumulative exposure was categorized into three
groups (<2, 2-12, and >13 years), and latency was dichotomized into "less than 20 years" or
"20 or more years." Men with <6 months of exposure were included in the analyses. As with
the earlier study, DuPont cancer registry data were used as a comparison for the cancer incidence
rates observed in the exposed cohort. DuPont and U.S. death rates were used as a comparison
for the mortality data in the exposed cohort.
There were 43 incident cancer cases, including 10 lung cancer cases and 6 prostate cancer
cases (Table 4-3). The authors did not comment on the problems with ascertainment of incident
lung cancer cases described in the O'Berg (1980) study. Two additional cases of lung cancer
were identified. No significant differences between observed all-cancer incidence or even lung
cancer incidence and expected values were found among workers (Table 4-3). There was a
58
DRAFT - DO NOT CITE OR QUOTE
-------
statistically significant association between AN exposure and prostate cancer incidence (SIR =
6 observed -M.8 expected = 3.3, 95% CI = 1.35-6.93). All prostate cancer cases were observed
among wage workers who had at least 20 years of latency. No other significant relationship
between latency and cumulative exposure was detected.
Table 4-3. Distribution of select incidence and mortalities among wage
workers and all workers at an AN plant
Wage workers
All workers
Cause of
disease/death
Observed
cases
SIRab
(CI)
Observed
deaths
SMRa
(CI)
Observed
cases
SIRab
(CI)
Observed
deaths
SMRa
(CI)
All causes
-
-
139
1.18
(1.00-1.39)
-
-
155
1.15
(0.98-1.34)
All cancers
37
1.25
(0.89-1.70)
31
1.15
(0.79-1.61)
43
1.17
(0.86-1.56)
36
1.14
0.81-1.56)
Lung cancer
10
1.67
(0.85-2.97)
12
1.17
(0.63-2.00)
10
1.39
(1.41-4.95)
14
1.21
(0.69-1.98)
Prostate
6
4.00°
(1.62-8.32)
1
1.11
(0.05-5.48)
6
3.33°
(1.35-6.93)
1
1.00
(0.05-4.93)
aSIR and SMR standardized against company-wide rates.
Calculated based on observed and expected values provided.
"Statistically significant (p < 0.05).
Source: Amended from O'Berg et al. (1985).
A total of 155 deaths was reported, 36 of which were attributed to cancer. No significant
differences between observed and expected deaths were reported among either wage workers or
salary workers (Table 4-3). Twelve lung cancer deaths were observed vs. 10.2 expected (SMR =
1.17, 95% CI = 0.63-2.00), and 1 prostate cancer death was observed vs. 0.9 expected (SMR =
1.11, 95% CI = 0.05-5.48). The mortality comparisons were less elevated than the morbidity
comparisons, indicating that a longer follow-up period may be needed. The observation that
prostate cancer was elevated in the incidence data but not in the mortality data could reflect the
fact that prostate cancer has a relatively high 5-year survival rate.
In summary, O'Berg et al. (1985) added approximately 7 years of follow-up to the
previous DuPont cohort but did not aid in refining knowledge about the potential for a
relationship between AN exposure and lung-cancer incidence. The population was exposed to
chemicals other than AN, and these additional exposures were not accounted for in the analyses.
The small number of lung cancer and prostate cancer cases made the interpretation of the
stratified analyses difficult, and the observed numbers of cases were not different from expected
numbers of cases in either the highest cumulative exposure group or the longest latency group.
This study shares the same limitations as the O'Berg (1980) study. Smoking status, which was
not available in O'Berg et al. (1985), may have acted as a confounder. As the numbers of
59
DRAFT - DO NOT CITE OR QUOTE
-------
incident cases and cancer-specific deaths are small, and concomitant exposure to other chemicals
is a potential limitation, this study does not provide a substantial basis for the evaluation of the
relationship between AN exposure and cancer incidence.
Chen et al. (1987) assembled a cohort of 1,083 mostly white male workers from a
DuPont plant in Waynesboro, Virginia, that produced the acrylic fiber Orion®. The
manufacturing process was similar to that at the Camden plant except there was greater distance
between the process areas at the Waynesboro facility (Wood et al., 1998). Workers employed at
the plant between 1944 and 1970 who had a potential for AN exposure were included. The
cohort consisted of 805 wage roll employees and 278 salary roll employees. Potential exposure
was based on the review of work histories. No quantitative information was available for
exposure levels to AN; however, each job was classified as having either low, moderate, or high
levels of AN exposure. Because data prior to 1957 were not available, the investigators analyzed
only deaths between 1957 and 1981.
As with previous DuPont cohort studies, morbidity in the exposed cohort was compared
with company-wide cancer incidence rates, while mortality was compared with company-wide
rates and national rates. A total of 92 deaths was observed, with 21 attributed to cancer. For
both wage workers and salary workers, the number of observed deaths was significantly lower
than expected (Table 4-4). No excess of observed cancer deaths was observed in either worker
category. Among the wage workers, the five observed lung cancer deaths were not statistically
different than expected based on U.S. rates (SMR = observed ^ expected = 0.59, 95% CI = 0.22-
1.32). The significant deficit in observed deaths likely indicates, at least in part, the presence of
a healthy worker effect, incomplete cohort identification, and possibly incomplete ascertainment
of mortality data. It should be noted that death information was gathered for active and
pensioned employees, and names of terminated employees were submitted to the Social Security
Administration for identification of vital status. Vital status for 20 people (1.8%) was unknown.
No lung cancer deaths were observed among the salary workers. As for other cause-specific
deaths, no significant trends were detected among either wage or salary workers by time period
or duration of exposure. Analyses by the level of AN exposure and cumulative exposure did not
show significant differences between observed deaths and expected values.
60
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-4. Distribution of select incidence and mortalities among wage and
salary workers at an AN plant
Cause of
disease/death
Wage workers
Salary workers
Total
Observed
deaths
SMRusa
(CI)
SMRDuPonta
(CI)
Observed
deaths
SMRusa
(CI)
SMRDuPonta'b
(CI)
Observed
cases
SIRDuPonta'b
(CI)
All causes
68
0.57b
(0.44-0.78)
0.77b
(0.61-0.98)
24
0.41b
(0.27-0.60)
0.66b
(0.43-0.97)
-
-
All cancers
18
0.75
(0.44-1.16)
0.88
(0.54-1.37)
3
0.24b
(0.06-0.66)
0.3 lb
(0.08-0.85)
37
1.01
Lung cancer
5
0.59
(0.22-1.32)
0.66
(0.24-1.46)
0
Not
determined
Not
determined
5
0.72
(0.26-1.61)
Prostate
cancer
1
1.11
(0.06-5.48)
1.11
(0.06-5.48)
1
2.0
(0.03-2.47)
2.0
(0.03-2.47)
5
2.63
(0.96-5.83)
"Calculated based on observed and expected values provided.
bStatistically significant (p < 0.05).
SMRDuPont = SMR based on DuPont Registry; SMRu s = SMR based on U.S. statistics
Source: Amended from Chen et al. (1987).
Analyses of cancer incidence or morbidity reported 37 incident cancers in the cohort
(Table 4-4). Twenty-seven were among the wage roll employees (vs. 26.0 expected) and
10 were among the salary roll employees (vs. 10.5 expected). No increase was noted in lung
cancer incidence, with 5 cases observed in the total group vs. 6.9 expected (SIR = observed ^
expected = 0.72, 95% CI = 0.26-1.61). The incidence of prostate cancer was increased, with
5 observed cases compared to 1.9 expected based on the company-wide database (SIR = 2.63,
95% CI = 0.96-5.83).
In summary, no statistically significant observed excess in overall cancer incidence or
mortality was found among AN workers, but the presence of a healthy worker effect is likely.
An increase in prostate cancer incidence was noted. As there is a general high rate of survival
for prostate cancer cases, cancer incidence rather than mortality should be considered in the
evaluation of AN exposure and disease. Some of the study's disadvantages that limit the
interpretation of the results include the small sample size and number of cancer cases, subjective
exposure assessment, lack of smoking information, and lack of stratification by latency and dose
in the data analysis. There was the possibility of under-ascertainment of cases, particularly
prostate cancer cases, by relying on the DuPont Cancer Registry as the source of information on
cancer incidence. The results of this study support the findings of an excess in prostate cancer by
O'Berg et al. (1985), but the aforementioned study limitations may have hindered the
observation of an association between lung cancer and AN.
Wood et al. (1998) combined the Virginia and South Carolina cohorts described above
(Chen et al., 1987; O'Berg et al., 1985) and analyzed 2,428 male workers with an added decade
of follow-up. The exposure assessment was updated, taking into account plant histories,
61
DRAFT - DO NOT CITE OR QUOTE
-------
descriptions of manufacturing processes, and changes that would affect exposures, a matrix of
job titles, and work area names, documentation of personal protective equipment use, personal
and area air sampling data from 1975, and descriptions of working conditions by long-term
employees. Environmental air samples confirmed changes in plant processes, engineering, and
ventilation. Exposure levels for the period before 1975 were estimated from information
provided by a panel of "knowledgeable employees." Wood et al. (1998) estimated peak AN
exposure in parts per million for a 40-hour workweek for each job title/work area combination
and averaged this over a year. Peak exposure level categories, reported as the interval averages,
were low (0.11 ppm), moderate (1.1 ppm), high (11 ppm), and very high (30 ppm). Other
exposure variables analyzed included latency (<20 and >20 years of observation), duration of
exposure (<5, 5-9, and >10 years), and cumulative exposure (<10, >10-<50, >50-<100, and
>100 ppm-years). There was no information on exposure to other chemicals or on smoking
status of the subjects.
The cohort included all employees who worked in exposed areas in either plant until the
plants were closed. The DuPont Cancer Registry was used to identify incident cases of cancer
among employees during employment. Thus, incident cancer cases occurring after tenure at
DuPont would not have been identified. Vital status of past employees was determined through
a review of the National Death Index (1979-1991) and the Social Security Administration (all
years). Vital status of living employees was confirmed through pension records, motor vehicle
records, and credit bureau reports. Expected numbers of deaths were derived from the DuPont
mortality files and the U.S. population. The period of follow-up was extended through 1991 in
the South Carolina plant and 1990 in the Virginia plant. At the end of the follow-up period,
approximately 18% of the cohort was deceased (n = 454). The entire cohort had a total of
72,083 person-years of follow-up for the mortality analyses and 29,461 person-years for the
morbidity analyses. More than half of the study population was born before 1930, and 82% were
first exposed before 1971, thus allowing adequate follow-up to examine latency and adequate
attained age to examine most cancer outcomes. A total of 37% of the population had a
cumulative exposure level estimated at 50 or more ppm-years.
The SMR analysis revealed that the 454 deaths observed in the cohort were significantly
below the expected number of deaths based on U.S. rates (SMR = 0.69, 95% CI = 0.62-0.75).
There were 126 deaths from cancer, which include 46 lung cancer deaths, 27 digestive cancer
deaths, and 11 prostate cancer deaths (Table 4-5). The SMR for all cancers was also
significantly lower than expected (SMR = 0.78, 95% CI = 0.64-0.93), indicating a possible
healthy worker effect. Incomplete ascertainment of the endpoint of death may have also played a
role in the low SMR value, as only 18% of the cohort was deceased at the time of the follow-up.
Among the site-specific cancer mortalities, an excess number of prostate cancer deaths was
reported (SMR = 1.29, 95% CI = 0.64-2.30). A comparison of the cohort death rates with those
derived from the DuPont mortality registry showed SMRs that were slightly different than those
62
DRAFT - DO NOT CITE OR QUOTE
-------
derived from the U.S. data but did not provide any definitive differences. For all comparisons,
there was no evidence of a relationship with any measure of exposure. For respiratory cancer, a
nonsignificant excess of observed deaths was noted for the highest exposure level within the
stratified analysis (SMR = 1.23 based on 23 observed deaths, 95% CI = 0.80-1.85).
Cancer incidence in the cohort was compared with company-wide incidence rates derived
from the DuPont Cancer Registry. There were no significant elevations reported in any cancer-
specific category (Table 4-5), although an excess in prostate cancer cases was reported. There
were sufficient deaths in several categories to allow examination of the patterns of incidence by
the four measures of exposure utilized in the mortality analyses. Specifically, all cancers,
digestive organ cancers, respiratory system cancer, and prostate cancer were examined in the
exposure analyses. There was no evidence of a relationship with any measure of exposure for all
cancers, respiratory cancer, or digestive organ cancers.
Table 4-5. Distribution of select mortalities among exposed workers in two
AN plants
Cause of
disease/death
South Carolina cohort
Virginia cohort
Combined cohort
Observed
deaths
SMRa
(CI)
Observed
deaths
SMRa
(CI)
Observed
deaths
SMRa
(CI)
Observed
cases
SIR3
(CI)
All causes
271
0.76b
(0.67-0.85)
185
0.60
(0.51-0.69)
454
0.69b
(0.62-0.75)
-
-
All cancers
77
0.88
(0.69-1.10)
50
0.66
(0.49-0.87)
126
0.78b
(0.64-0.93)
101
0.97
(0.79-1.18)
Lung cancer
35
1.06
(0.74-1.47)
11
0.40b
(0.20-0.71)
46
0.76
(0.56-1.02)
17
0.81
(0.48-1.28)
Digestive
cancer
12
0.57b
(0.29-0.99)
15
0.81
(0.45-1.33)
27
0.69
(0.45-1.00)
22
0.89
(0.56-1.34)
Prostate
cancer
5
1.23
(0.40-2.86)
6
1.32
(0.48-2.88)
11
1.29
(0.64-2.30)
12
1.58
(0.82-2.76)
aSIR and SMR standardized against company-wide rates.
Statistically significant (p < 0.05).
Source: Amended from Wood et al. (1998).
In summary, this study provided better exposure assessment than previous studies of this
group of workers. Additional follow-up and the combination of two small cohorts enhanced the
information available for inclusion; however, the apparent healthy worker effect may have
masked a relationship between AN exposure and death from or incidence of lung cancer or other
cancers.
Symons et al. (2008) provided an update based on the combined Virginia and South
Carolina cohorts from Wood et al. (1998), adding 11 years of follow-up. The exposure
assessment was the same as described in Wood et al. (1998) with exposure being based on a job-
63
DRAFT - DO NOT CITE OR QUOTE
-------
exposure matrix, documentation of personal protective equipment use, personal and area air
sampling data, and descriptions of working conditions by long-term employees. Vital statistics
were obtained through the DuPont Epidemiology Registry and verified by the National Death
Index. New in this update was the assignment of mean intensity values based on estimated
intensity categories ranging from <0.2 to >20 ppm. These intensity categories were coupled with
duration of employment to determine cumulative exposure. For SMR calculations, the expected
number of deaths were derived both the U.S. population and regional DuPont employees. To
reduce potential bias associated with the healthy worker effect, the Occupational Mortality
Analysis Program (OC-MAP) was utilized for SMR calculations. For the estimation of relative
mortality risk via a hazard ratio, a log linear model for AN exposure was assumed for all-cause
and cause-specific outcomes.
Of the 2,559 workers from Wood et al. (1998), 11 workers were found to be exposed to
AN for <6 months and thus were excluded from the cohort in Symons et al. (2008). A total of
839 deaths (32%) were observed in this updated cohort of 2,548 male workers. Of these,
240 deaths were due to cancer, with 88 deaths being specific to lung cancer (Table 4-6). One
quarter of the updated cohort were found to be in the mean intensity exposure group of
<2.0 ppm, while, two-thirds were in the 2.0-19.9 ppm mean intensity exposure category. The
number of deaths, especially cause-specific deaths, in these exposure groups was not reported.
Table 4-6. Distribution of select mortalities among exposed workers
Cause of death
South Carolina cohort
Virginia cohort
Combined cohort
Observed
deaths
SMRa
(95% CI)
Observed
deaths
SMRa
(95% CI)
Observed
deaths
SMRa
(95% CI)
SMRb
(95% CI)
All causes
481
0.98
(0.89-1.07)
358
0.85°
(0.76-0.95)
839
0.92°
(0.86-0.98)
0.69°
(0.64-0.74)
All cancers
144
1.00
(0.84-1.18)
96
0.81°
(0.67-0.99)
240
0.92
(0.81-1.04)
0.73°
(0.64-0.82)
Lung cancer
61
1.14
(0.88-1.49)
27
0.64°
(0.42-0.93)
88
0.92
(0.75-1.14)
0.74°
(0.60-0.91)
Prostate cancer
12
0.93
(0.48-1.63)
13
1.12
(0.60-1.92)
25
1.02
(0.66-1.51)
0.91
(0.59-1.35)
aBased on company regional rates.
''Based on U.S. population rates.
Statistically significant (p < 0.05).
Source: Amended from Symons et al. (2008).
Symons et al. (2008) also reported hazard ratio estimates for 100-ppm years increase in
cumulative exposure (Table 4-7). In addition to the variable 'exposure term' used in the crude
model, the adjusted model utilized in generating hazard ratio estimates included 'birth period'
(6-decade ordinal variable) and 'employment in South Carolina start-up group' (binary
64
DRAFT - DO NOT CITE OR QUOTE
-------
indicator). The authors noted a slight significant increase in all-cause mortality associated with
increasing cumulative exposure to AN in the crude model but did not find any other statistically
significant findings. Colorectal and brain cancers had elevated hazard ratios but were not
statistically significant in either the crude or adjusted models.
Table 4-7. Crude and adjusted hazard ratio estimates for 100 ppm-year
increase in cumulative exposure
Cause of death
Crude
Adjusted"
Hazard ratio
95% CI
Hazard ratio
95% CI
All causes
1.12b
1.04-1.21
1.05
0.97-1.14
All cancers
1.07
0.92-1.24
1.00
0.86-1.17
Lung cancer
1.04
0.81-1.33
0.95
0.73-1.23
Prostate cancer
0.81
0.48-1.35
0.78
0.46-1.32
Colorectal cancer
1.13
0.74-1.71
1.16
0.75-1.81
Brain and CNS
1.37
0.59-3.16
1.03
0.38-2.78
aModel includes exposure term, birth period, and employment in South Carolina start-up group.
Statistically significant (p < 0.05).
Source: Amended from Symons et al. (2008).
Hazard ratios estimates were also derived based on lagged cumulative exposure.
Increasing exposure lags resulted in decreasing hazard ratios for all cancers, lung cancer, and
colorectal cancer. However, an increase in hazard ratio estimates over lagged cumulative
exposure was observed for brain and CNS cancer and lymphatic and hematopoietic cancer. It
should be noted that the number of events for these cancers were small (Table 4-8).
65
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-8. Hazard ratio estimated for select cancer mortality by lagged
cumulative exposure for 100 ppm-year increase in cumulative exposure
Cause of death
5-Yr lag
(n = 2,224)
10-Yrlag
(n = 2,066)
15-Yrlag
(n = 1,907)
Events
Hazard ratio
(95% CI)
Events
Hazard ratio
(95% CI)
Events
Hazard ratio
(95% CI)
All cancers
210
1.03
(0.87-1.23)
199
0.98
(0.80-1.20)
179
0.94
(0.74-1.19)
Lung
75
0.95
(0.70-1.28)
72
0.84
(0.59-1.19)
63
0.80
(0.53-1.22)
Prostate
22
0.86
(0.48-1.53)
19
0.98
(0.50-1.92)
15
0.95
(0.41-2.22)
Colorectal
26
1.06
(0.64-1.76)
24
0.90
(0.49-1.66)
23
0.81
(0.40-1.66)
Brain and CNS
4
1.38
(0.43-4.47)
4
1.55
(0.44-5.47)
4
1.96
(0.49-7.84)
Lymphatic and
hematopoietic
16
1.09
(0.59-2.01)
14
1.29
(0.66-2.54)
13
1.27
(0.57-2.86)
Source: Amended from Symons et al. (2008).
In summary, Symons et al. (2008) followed workers for over 50 years and, among the
33% mortality in the cohort, reported that no statistically significant observed excess in overall
mortality was found among AN workers. Unlike the previous studies based on the DuPont
cohort, Symons et al. (2008) did not provide cancer incidence data or address previous issues
with this cohort such as smoking status and concomittent exposure to other chemicals. This
update did attempt to reduce potential bias from the healthy worker effect with specialized
statistical program. Differences between U.S. population-based SMRs and regional worker-
based SMR indicated that there may still be a strong healthy worker effect. Although Symons et
al. (2008) categorized workers in different mean intensity exposure and cumulative exposure
groups, the authors did not report the observed number of cause-specific deaths in these
categories or provide cause-specific SMR estimates. In deriving hazard ratio estimates, Symons
et al. (2008) adjusted for workers 'employment in South Carolina start-up group.' Based on
O'Berg (1980), it appears that workers in the South Carolina start-up group may have had higher
levels of exposure to AN. Additionally, the metric derived by quantitation of AN exposure and
SMR derivation of this subgroup would have been helpful for better assessing the relationship
between AN exposure and lung cancer.
Overall summary
The DuPont cohort studies were the first to hypothesize an association between AN
exposure and cancer, specifically lung and prostate cancer. It should be noted that the DuPont
cohort studies are limited to male workers and lack appropriate unexposed worker comparison
66
DRAFT - DO NOT CITE OR QUOTE
-------
groups to better assess the risk of developing cause-specific diseases, namely cancer. Even in
the most recent cohort study by Symons et al. (2008), smoking and concomitant exposures were
not adequately addressed in any of the DuPont cohort studies. Interestingly, the prostate cancer
findings in the Wood et al. (1998) study revealed a closer association to duration of employment
rather than exposure. This observation is echoed with brain and CNS cancer and lymphatic and
hematopoietic cancer in Symons et al. (2008).
Excluding Symons et al. (2008), the DuPont cohort studies are the only studies that
evaluate the association between AN exposure and cancer incidence rather than relying on a
surrogate marker such as cause-specific death. For cancer incidence, the cohort studies appeared
to only focus on data from the DuPont cancer registry rather than including data from state or
national tumor registries that have known quality assurance and quality control procedures for
diagnoses.
Symons et al. (2008) assessed risk based on mortality. As with the other DuPont studies,
the reliance on mortality statistics from the U.S. population as a comparison group is less
desirable given the potential for bias from the healthy worker effect. Symons et al. (2008)
attempted to address this issue by employing DuPont regional workers as a comparison group
instead of using company-wide statistics as in other DuPont studies. Information regarding other
chemical exposures among regional workers would aid in reducing the uncertainty of exposure
to other potential carcinogens in this comparison group.
As with its predecessor, Wood et al. (1998), the most recent DuPont cohort study Symons
et al. (2008) may also suffer from exposure misclassification, as no exposure monitoring existed
prior to 1975. This lack of information on exposure potential and level may also subject the
studies to recall bias. However, given the shortcomings of each of the individual DuPont studies,
there is still suggestive evidence that there may be an association between AN exposure and the
incidence of cancer.
American Cyanamid Company
Collins et al. (1989) conducted a retrospective cohort study of 2,671 men who worked at
two American Cyanamid Company plants in Louisiana and Florida from start-up (1951 for the
Louisiana plant, 1957 for the Florida plant) through December 1973. One facility manufactured
AN and other materials, and the other facility utilized large quantities of AN in the
manufacturing of acrylic fiber. All 2,671 study subjects were followed through the end of
1983 for mortality, thus allowing at least 10 years of follow-up for the cohort. According to
Collins et al. (1989) for exposure estimation, industrial hygiene monitoring, which began in
1977, was considered representative of previous exposure levels, with adjustments made for
changes in practices and engineering controls. Any actual measurements that were available
were also used to tailor job-specific exposure. The investigators created four exposure
categories: 0-<0.01, 0.01-0.7, 0.7-7.0, and >7 ppm/year. Exposed workers were defined as
67
DRAFT - DO NOT CITE OR QUOTE
-------
having a cumulative exposure of >0.01 ppm/year. Smoking information was only available for
58% of the workers. Each person in the cohort was coded as smoker (smoked for >3 months),
nonsmoker (smoked for <3 months), or unknown (information not available). The authors used
an internally standardized method of adjustment for age (<45, 45-54, 55-64, 75+ years), race
(white, nonwhite), smoking status (smoker, nonsmoker, unknown), latency (<10, 10-19,
20+ years), and time period (<1965, >1965). In the calculation of the SMR, the expected number
of cause-specific deaths was derived from U.S. general population data on men.
More than 50% of the exposed workers in this study were followed for at least 20 years.
By the end of the study, a total of 237 deaths (92 from the unexposed group and 145 from the
exposed group) were observed, and death certificates were located for 224 of these workers.
From both groups combined, there were 65 cancers, including 23 respiratory cancers, of which
22 were lung cancers. For lung cancer, there was no significant elevation in the SMR for either
the exposed or unexposed group (SMR in unexposed group = 1.08, 95% CI = 0.69-1.61; SMR in
exposed group = 1.01, 95% CI = 0.74-1.35). A similar pattern was observed among the lung
cancer deaths, with no significant elevations reported (SMR in unexposed group = 1.01, 95% CI
= 0.44-2.01 [7 cancer deaths]; SMR in exposed group = 1.00, 95% CI = 0.58-1.61 [15 cancer
deaths]). The SMRs for the AN exposure categories for lung cancer were 1.09 (95% CI = 0.51—
2.08), 0.63 (95% CI = 0.10-2.06), 0.64 (95% CI = 0.20-1.53), and 1.41 (95% CI = 0.68-2.58)
for categories 0-<0.01, 0.01-0.7, 0.7-7.0, and >7 ppm/year, respectively. Using an internally
standardized method that adjusted for smoking, race, latency, age, and time period, the SMRs
were 1.11 (95% CI = 0.52-2.11), 0.72 (95% CI = 0.12-2.36), 0.71 (95% CI = 0.23-1.72), and
1.22 (95%) CI = 0.59-2.23), respectively, for the above exposure categories. None of these risk
estimates or subsequent trend tests was statistically significant. No additional information was
provided with regard to the nature of the trend test.
The inclusion of an analysis for the unexposed group compared to the U.S. population
was helpful and allowed an evaluation of whether elevations in cancer rates in the exposed group
were also observed in the unexposed group and thus were not likely to be related to exposure. It
should be noted that indirect standardization (the use of age-specific mortality rates from the
standard U.S. population to derive regional expected deaths) was used, thus hindering the ability
to compare SMRs across groups. This study attempted to quantify exposure levels and control
for smoking history. The study did allow for approximately 50% of the exposed cohort to have
at least 20 years of follow-up, thus strengthening the possibility that a health effect might be
detected. However, since the endpoint marker used was death, as recorded on the death
certificates, and with less than 10% of the study cohort deceased by the end of the study, the
study may have been premature in attempting to link AN exposure to cancer deaths. This study
may have had insufficient power as portrayed by the observation of only 15 lung cancer deaths in
the exposed group. It should also be noted that this study, as with many other cohorts assembled
to assess the relationship between AN and cancer, consists of only men.
68
DRAFT - DO NOT CITE OR QUOTE
-------
Because the results presented were adjusted by latency, duration of exposure, and time
since exposure, relative risks (RRs) in these strata could not be readily compared with the results
of the studies of Symons et al. (2008), Benn and Osborne (1998), Blair et al. (1998), Swaen et al.
(1998), Wood et al. (1998), O'Berg et al. (1985), and O'Berg (1980). Unlike other studies that
assumed early exposure levels were higher, Collins et al. (1989) assumed that AN exposure
levels in 1977 were representative of the time frame prior to that date. This assumption may
have led to exposure misclassification, resulting in a flattening of the dose-response gradient.
Other limitations included low statistical power to evaluate lung or rarer cancers, particularly in
subgroup analyses, incomplete smoking information, and use of SMRs rather than comparable
unexposed controls. In summary, this study may have been premature in assessing the
relationship between AN exposure to cancer deaths.
Synthetic chemical plant in the U.S.
Waxweiler et al. (1981) examined the mortality rates in a cohort of 4,806 chemical plant
workers who were exposed to many potential carcinogens, including AN. The cohort was
identified as all workers at the synthetic chemicals plant who were first employed between
1942 and 1973. These workers were followed through the end of 1973. Since the majority of
the cohort (63%) were actually hired before 1954, this allowed for at least 20 years of latency
and follow-up for the majority of the workers. However, it should be noted that some workers
had less than 1 year of follow-up, latency, or exposure. The cohort was described as young, with
only 30% of live workers being over 55 years of age.
Mortality rates in the plant cohort were compared to those derived from the U.S. white
male population. Under the assumption that all workers were considered at risk from the first
day of employment, no difference was noted between the observed deaths and the expected
deaths (556 observed deaths vs. 550.2 expected deaths, SMR = 1.01, 95% CI = 0.93-1.10). The
number of observed cancer deaths was higher than expected (109 observed vs. 92.5 expected,
SMR = 1.18, 95%) CI = 0.97-1.42). Upon stratification of cancer deaths, higher numbers of
respiratory system and CNS cancers were observed among the workers (respiratory cancer SMR
= 1.49, 95% CI= 1.09-1.99; CNS cancer SMR = 2.09, 95% CI= 1.02-3.84).
A secondary analysis was performed among workers who had at least 10 years of
exposure. Waxweiler et al. (1981) noted similar findings, though actual values were not
reported. The authors did note that there were 39 observed deaths from cancers of the
respiratory system compared to 25 expected (SMR = 1.56, 95% CI = 1.12-2.11). A similar
elevation had been observed by the authors in a previous publication (Waxweiler et al., 1976)
that focused on the workers with high vinyl chloride monomer exposure; thus, the authors
concluded that the excess lung cancer risk may not solely be due to exposure to a single
chemical.
69
DRAFT - DO NOT CITE OR QUOTE
-------
Serially additive expected dose (SAED) modeling was performed to determine whether
exposure to chemicals, including AN, was associated with excess lung cancer risk. Job histories
were used to assess the potential for exposure to 19 chemicals routinely used at the plant. Each
job was assigned a qualitative exposure rating (from 0 for no exposure to 5 for intimate contact
on the skin or high inhalation potential) for each year of the study. Thirty-five of the 80 (56%)
job categories ranked had no exposure to AN, with another 30% reporting minimal to low levels
of exposure. An exposure score was derived for each chemical for all workers by summing each
exposure rating for each year. The exposure score for each cohort member dying from
respiratory system cancers was compared to the score for a group of workers with a similar birth
year and year of first hire (the "subcohort" from which the case arose). Scores in the
nonrespiratory death subcohorts (expected scores) were subtracted from the observed scores in
the respiratory cancer deaths to obtain an observed-minus-expected cumulative dose difference
per lung-cancer case. For AN, this calculation resulted in a negative unit, reflecting that
expected cumulative dose greatly exceeded the observed cumulative dose. No significant
differences with regard to AN exposure were observed in this analysis.
While this study showed significant increased risks of death from lung and CNS cancer in
a cohort of workers exposed to multiple carcinogens, limitations with the study design hindered
the ability to assess the relationship between the observed deaths and exposure to AN. The
population of workers was potentially exposed to multiple carcinogens, so any effect of exposure
to AN may have been impossible to measure.
Rubber manufacturing plant
Delzell and Monson (1982) analyzed the mortality among workers from a rubber
manufacturing plant in order to determine if potential exposure to AN might be associated with
excess deaths. The study included 327 white males who had been employed at the plant for at
least 2 years between January 1940 and July 1971. These employees were selected from over
15,000 workers because they worked in departments where AN exposure was most likely.
Mortality information was gathered on both active and pensioned workers through July 1978,
allowing at least 7 years of follow-up for most workers in the cohort. Workers without a record
of death were assumed to be alive. Mortality rates were compared with the white male subset of
the U.S. general population rates, stratifying by cause of death, age, and calendar time.
By mid-1978, a total of 74 deaths (~ 22%) were observed, 22 of which were from various
types of cancers. The differences between the observed and expected deaths for all-cause and for
cancer mortality were not statistically significant, with SMRs of 0.8 (95% CI = 0.7-1.0) and
1.2 (95%) CI = 0.8-1.9), respectively. Nine lung cancer deaths were observed as compared to the
5.9 expected (SMR = 1.5, 95% CI = 0.7-2.9). The ability to detect statistical significance may
have been hindered by the small number of deaths. The numbers of deaths from bladder,
lymphatic, and hematopoietic cancers exceeded the expected values, but the number of deaths
70
DRAFT - DO NOT CITE OR QUOTE
-------
were small. Lung cancer deaths were examined by duration of employment and latency. No
trends were noted, and there were no deaths in the group employed for >15 years. However,
7 lung cancer deaths were observed as compared to the 4.1 expected in the group with 15 or
more years of latency since first exposure.
This study was based on a small number of study participants, with less than a quarter
reaching the study endpoint of death. The study provides no quantitative assessment as to the
level of AN exposure. Mixed exposure is also an issue, since it is noted that butadiene, styrene,
and vinyl pyridine were utilized at the plant during the exposure time frame. However, there is
no information in the study as to whether the study participants were potentially exposed to these
other agents. Finally, though lung cancer was specifically analyzed, no mention of data
collection or adjustment for smoking history was reported. These shortcomings limit the weight
that this study carries with regard to assessing the relationship between AN exposure and cancer
mortality.
The Netherlands cohort
Three successive papers (Swaen et al., 2004, 1998, 1992) reported mortality analyses of a
Dutch cohort consisting of 2,842 male workers employed >6 months from 1956 to 1979 in eight
companies that produced AN, latex polymer, acrylic fiber, AN polymers, resins, or acrylamide.
This cohort was compared to 3,961 workers at a neighboring plant during the same time frame
who were not exposed to any known carcinogens in the normal work setting. Use of this
reference population aided in minimizing the impact of a potential healthy worker effect.
In the original study (Swaen et al., 1992), exposure assessment for most companies was
conducted using a job matrix model and air AN samples collected in 1978-1979. The exposure
assessment took into account changes in the production process, industrial hygiene control
measures, and work procedures over time. Potential for exposure misclassification was
enhanced by the extrapolation of exposure monitoring data from one plant to the other seven
plants, for which there were no exposure monitoring data. For all of the companies involved,
most jobs were classified in the following ranges: 0-0.5, 0.5-1, 1-2, and 2-5 ppm. However,
for some jobs, only two categories could be distinguished: 0-2 and 2-5 ppm. For all exposed
workers in the study, a cumulative measure of exposure was derived by multiplying the average
concentration for a particular exposure class by the number of years in that class. Although
workers were characterized by duration of exposure and duration of follow-up, these variables
were not used in the evaluation of the relationship between AN and cancer deaths.
Both the exposed and the unexposed groups were compared with national Dutch death
rates to generate SMRs. No direct comparison of rates in the exposed and unexposed groups was
followed for >20 years after entry into the cohort. Almost 24% of the exposed cohort was
categorized in the highest category of cumulative exposure of >10 ppm-years, a cutoff higher
than the >8 ppm used in the Blair et al. (1998) study.
71
DRAFT - DO NOT CITE OR QUOTE
-------
Mortality information was collected for both groups through the end of 1987, allowing at
least 8 years of follow-up for all surviving study members. A total of 134 deaths was observed
in the exposed group (SMR = 0.78, 95% CI = 0.65-0.92) and 572 deaths in the unexposed group
(SMR = 0.77, 95% CI = 0.71-0.84), undertaken, with the authors citing differences in age and
calendar time between the groups as the rationale. Nearly half of the exposed group worked with
AN for at least 5 years, and 26% of the unexposed group was both observations significantly
lower than expected. As both exposed and unexposed workers were compared to a national rate,
the lower observed deaths may be attributable to the healthy worker effect. Among the workers
exposed to AN, no significant differences were noted between the number of observed and
expected cancer deaths (i.e., 42 observed deaths vs. 50.8 expected, SMR = 0.83, 95% CI = 0.60-
1.10). Analyses of site-specific cancer deaths in this group did not reveal any significant
differences between the observed cancer deaths and the expected values. However, among the
unexposed workers, the number of observed deaths from cancer of the trachea and lung was
significantly lower than expected (i.e., 67 observed deaths vs. 93.3 expected deaths, SMR = 0.72,
95% CI = 0.56-0.90), likely indicating, at least in part, the presence of a healthy worker effect.
Among exposed workers in the highest exposure category, the number of observed lung cancer
deaths exceeded the number expected (SMR = 1.11, 95% CI = 0.48-2.19). Compelling evidence
of the healthy worker effect is provided by the fact that the observed number of deaths among
the unexposed group, regardless of disease category, is significantly lower than the expected
estimates derived from the national population statistics.
In an updated analysis by Swaen et al. (1998), an additional 8 years of follow-up yielded
156 more deaths in the exposed cohort and 411 more deaths in the unexposed group, bringing the
total number of observed deaths to 290 and 983, respectively. The exposure assessment was not
updated because more recent exposures were considered to be negligible. For both groups, the
total number of cancers observed was lower than expected. The exposed group had 97 deaths
from neoplasms observed vs. 110.8 expected (SMR = 0.88, 95% CI = 0.71-1.06) and the
unexposed group had 332 deaths observed vs. 400.4 expected (SMR = 0.83, 95% CI = 0.74-
0.92). With the increased follow-up time, slight, but mainly statistically nonsignificant,
increases in cancer of the brain, large intestine, prostate, and all leukemias for the exposed group
were observed that were not previously evident. While these increases were based on small
numbers, it is noteworthy that intestinal tract and brain are cancer target organs in rat studies.
The association between AN exposure and mortality, all-cancer mortality, and lung
cancer mortality was examined by the following exposure variables: latency, peak exposure,
cumulative exposure, respirator use, and exposure to other carcinogens. Overall, among workers
exposed to AN, the number of observed deaths did not differ greatly from expected.
Swaen et al. (2004) revisited this cohort of AN workers and added 5 years of follow-up to
the analysis, for a minimum of 22 years of follow-up. This updated study added 142 new deaths
for the exposed cohort and 360 deaths to the unexposed cohort, bringing the total number of
72
DRAFT - DO NOT CITE OR QUOTE
-------
observed deaths for 432 and 1,343, respectively. The number of deaths was 2.5-fold higher than
in the original cohort study and accounted for over a quarter of the original study population.
The exposure assessment was not updated from the first study.
SMR analyses were conducted for both the exposed and unexposed cohorts, using the
Dutch general population rates as a comparison. The observed number of cancer deaths in the
exposed group was not significantly different from the expected value (Table 4-9). The study
found that number of respiratory cancer deaths in this group was higher than expected,
67 observed vs. 62.5 expected (SMR = 1.07, 95% CI = 0.83-1.36). Analyses by peak exposure,
respirator use, and possible exposure to cocarcinogens were also performed, yielding no
indication for elevated site-specific cancer risks in any of the subgroups. Additional analyses
were performed by examining the SMRs for lung cancer by various measures of dose and
latency (Table 4-10). An increasing SMR was noted with increasing levels of exposure (i.e.,
0.92, 1.06, and 1.15 for low, medium, and high exposure, respectively).
Table 4-9. Distribution of select mortalities among AN-exposed and
unexposed workers
Cause of death
Exposed workers
Unexposed workers
Observed
Expected
SMR
95% CI
Observed
Expected
SMR
95% CI
All causes
432
467.8
0.92
0.84-1.01
1,343
1,545.2
0.87
0.82-0.923
All cancers
146
164.5
0.89
0.75-1.04
447
519.8
0.86
0.78-0.94a
Lung and
trachea cancer
67
62.5
1.07
0.83-1.36
160
203.8
0.78
0.67-0.923
aStatistically significant (p < 0.05).
Source: Amended from Swaen et al. (2004).
73
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-10. Lung cancer mortality of AN-exposed workers stratified by
cumulative dose and latency
All cancer mortality
Lung cancer mortality
Dose
Observed
Expected
SMR
95% CI
Observed
Expected
SMR
95% CI
Low (<1 ppm/yr)
<10 Yrs latency
0
1.9
-
-
-
0.7
-
-
10-20 Yrs latency
7
7.5
0.93
0.37-1.91
3
2.9
1.03
0.21-2.97
>20 Yrs latency
10
11.3
0.88
0.42-1.62
4
4.0
1.00
0.27-2.53
Total
17
20.7
0.82
0.48-1.31
7
7.6
0.92
0.37-1.89
Moderate (1-10 ppm/yr)
<10 Yrs latency
8
8.4
0.95
0.41-1.87
1
3.2
0.31
0.40- 1.58
10-20 Yrs latency
31
32.1
0.96
0.66-1.37
16
12.4
1.29
0.74-2.09
>20 Yrs latency
39
49.4
0.79
0.56-1.08
19
18.2
1.04
0.63-1.63
Total
78
89.9
0.87
0.69-1.08
36
33.8
1.06
0.75-1.47
High (>10 ppm/yr)
<10 Yrs latency
8
6.0
1.33
0.57-2.62
3
2.5
1.20
0.24-3.44
10-20 Yrs latency
25
21.1
1.18
0.77-1.75
12
8.4
1.43
0.74-2.49
>20 Yrs latency
18
26.3
0.68
0.40-1.08
9
10.0
0.90
0.41-1.70
Total
51
53.4
0.95
0.71-1.26
24
20.9
1.15a
0.75-1.68a
Calculated based on reported data.
Source: Amended from Swaen et al. (2004).
In summary, as with previous analyses of this cohort, the interpretation of the results
from this study is limited by the following: potential misclassification of AN exposure because
of the use of current measures to derive past exposures and the use of subjective information
about exposure, use of a population-based control group, pooling of data from factories with
different kinds of AN production and exposures without adjusting for this, and lack of
information on smoking.
Epidemiology studies based on this Dutch cohort provided better exposure assessment
than studies using the SAED method or similar extrapolation tools. Here, exposure categories
represented estimates based, in part, on actual measurements rather than ordinal ranking. The
number of lung cancer deaths observed among the AN-exposed workers was higher than
expected. However, the following biases regarding the Dutch cohort studies are possible:
potential misclassification of AN exposure because of the use of current measures to estimate
past exposures and the use of subjective information about exposure, use of a population-based
control group, pooling of data from factories with different kinds of AN production and
exposures without appropriate adjustment, and bias based on the lack of smoking information
available. The low SMRs observed indicate the presence of a potential healthy worker effect,
making it more difficult to detect an association with AN.
74
DRAFT - DO NOT CITE OR QUOTE
-------
The Dutch studies cohort utilized an unexposed worker cohort for a "comparison" group.
The use of this unexposed group was not fully explored by developing rates by age and calendar
time in the unexposed group to be used as a basis for developing expected rates in the exposed
group. Instead, the unexposed group was compared to national rates and then used as a
"standard" to see if the patterns of SMRs generated in the exposed cohort looked similar to those
generated in the unexposed cohort. Therefore, the true comparison group used in these studies
was the national population. It should be noted that the lung cancer SMR among workers
exposed to AN was found to be higher than that among unexposed workers. Direct comparison
between the exposed and unexposed workers in terms of an RR was not derived, as the
demographics between the exposed and unexposed workers may have differed. Though no
explicit details were provided, it is noted that the number of person-years and crude mortality
rate are higher in the unexposed workers (recruited from a different plant) than in the exposed
workers. Thus, the expected number of deaths for each group, on which the SMRs are based,
would be dependent on different distributions or standards, a situation in which comparing the
SMRs between the exposed and unexposed groups would not be recommended. Additional
support is provided with the increasing SMR with level of exposure; a progressive increase of
0.92 in the low exposure group to 1.15 in the high exposure group was reported.
Finally, the statistical measure used in all the Dutch cohort studies was the SMR. This
statistic allows for the comparison of cause-specific deaths and not the incidence of disease.
Using deaths as a surrogate for disease may underestimate the true relationship between AN
exposure and cancer, especially when the majority of the cohort population is not deceased. As
with other studies that evaluate AN exposure using SMRs, these Dutch studies were focused on
determining if AN exposure may lead to an excess of cause-specific deaths; however, these
studies lack the appropriate design to address the issue of AN exposure and the incidence of
disease, namely site-specific cancers.
BASF plants in Germany
A mortality study was conducted among workers from 12 BASF plants in Germany
(Thiess et al., 1980). Though none of these plants manufactured AN, this substance, along with
styrene and butadiene, was used in many of their processes. The number of employees in each
plant varied, ranging from 30 to 334 employees. The first uses of AN at these plants did not
occur simultaneously but rather over a course of 14 years, from 1954 to 1968. A total of
1,469 active and former employees who had worked for >6 months processing AN were
identified for mortality follow-up. The cohort was followed through May 15, 1978, and death
certificates were obtained and coded for cause of death. There were 1,081 German workers in
the cohort, and the vital status was traced for 98% of these workers. Tracing was less successful
for the 388 foreign workers in the cohort, with only 56% follow-up.
75
DRAFT - DO NOT CITE OR QUOTE
-------
No measurements of levels of AN exposure were available for use in this study. It was
noted that prior to 1976, AN was handled manually, and as a result, there may have been
increased exposure during that time (Thiess et al., 1980). In later years, closed systems were
utilized, leading to reduced exposures. Although measurements were not available, this study
assumed that workers were occasionally exposed to levels of >20 ppm for short periods, based
on the manual handling of AN at specific work sites prior to 1976. Several other known
carcinogens were used at some of the facilities and could have presented a confounding effect in
terms of cancer outcomes. One plant (Plant 5) was subsequently excluded from the analysis
because of the potential for concurrent exposure to P-naphthylamine.
The comparison rates used for the majority of the analyses were derived from mortality
rates for the Federal Republic of Germany. The 1,469 workers accounted for a total of
15,350 person-years of follow-up. A total of 89 deaths were observed, a value lower than the
99 predicted deaths based on rates for the Federal Republic of Germany (SMR = 0.90, 95% CI =
0.72-1.10). A total of 27 cancer deaths was observed for the entire cohort, compared to the
20.5 expected deaths (SMR = 1.32, 95% CI = 0.89-1.89).
In the analysis that excludes Plant 5, the number of observed deaths dropped to 74, with
the expected deaths at 78.8 (SMR = 0.94, 95% CI = 0.74-1.17). Among these deaths, 20 were
attributed to cancer, with a calculated expected value of 16.1 deaths (SMR = 1.24, 95% CI =
0.78-1.88). In both analyses, the number of cancer deaths exceeded expected estimates, but the
small sample size might have hindered the observation of statistically significant findings. When
cause-specific cancer deaths were examined, a significant difference was noted between the
observed and expected bronchial carcinoma (lung cancer) deaths, regardless of the inclusion of
Plant 5 in the analysis (i.e., 11 observed deaths vs. 5.7 expected in the full cohort and 9 observed
vs. 4.4 expected in the cohort without Plant 5).
Neoplasms of the lymphatic and hematopoietic organs were also observed to be elevated,
but not significantly (i.e., 4 observed vs. 1.7 expected in the full cohort and 4 observed vs.
1.4 expected in the cohort without Plant 5). Because of the small number of deaths in this
category, further stratification or examination was precluded. It should be noted that two of the
four deaths in this category were from Hodgkin's disease, which is significantly different from
the 0.3 deaths expected.
Analyses were conducted, excluding Plant 5, for the group of workers who were followed
for at least 5 years before death or loss to follow-up. This comprised 944 workers, with only
seven bronchial carcinomas reported. These carcinoma deaths were stratified by duration of
exposure into three categories: 0-4, 5-9, and >10 years. There were no bronchial carcinomas in
the 0-4 year exposure category, four in the 5-9-year exposure category and three in the highest
exposure category. For the latter two categories, the observed number of deaths was higher than
expected (SMR4-9 years = 3.86, 95% CI = 1.23-9.31; SMR>i0 years = 2.23, 95% CI = 0.57-6.07).
76
DRAFT - DO NOT CITE OR QUOTE
-------
The study of BASF workers showed a significant excess in lung-cancer deaths in workers
potentially exposed to AN. The follow-up time was not specifically quantified; however,
examination of the exposure (1954-1968) and follow-up (5/15/1978) dates indicates that most
workers in the cohort were followed for at least 9 years. The study is limited by the fact that
only 6% of the study cohort was recorded as deceased; this observation is important since death
certificates were used to determine if an association between AN exposure and certain cancers
existed. An additional limitation of this study is the fact that the cohort was assembled from
12 different plants and some of the plants were acknowledged to have concurrent exposures to
known carcinogens such as P-naphthylamine, vinyl chloride, or solvents. The study did not
indicate when these overlaps in exposures could have occurred and how many workers at the
plant could have been possibly exposed, and no information on the level of exposure was
provided. This limitation coupled with a lack of ability to characterize the workers by job or
exposure level makes this study less informative in trying to assess the relationship between
cancer mortality and AN exposure.
Six factories in the United Kingdom
There have been two successive studies evaluating the mortality rates in a cohort of male
workers potentially exposed to AN at six United Kingdom polymerization and spinning factories
(Benn and Osborne, 1998; Werner and Carter, 1981). For both studies, workers were included if
they were employed for at least 1 year between 1950 and 1968. Werner and Carter (1981)
examined the mortality rates in a cohort of 1,111 men drawn from six factories in England,
Wales, Scotland, and Northern Ireland. This cohort was followed through the end of 1978,
allowing for a minimum 10-year follow-up for all surviving workers. The workers included in
the study were deemed to have the potential for the highest level of AN exposure, since they
were involved in either the AN polymerization process or spinning of acrylic fiber. However, no
exposure monitoring data were available for the period of the study. Additionally, there was
potential for concomitant exposure to styrene and butadiene in the work environment.
Each worker in the cohort was classified according to the length of time spent in a high-
exposure job. This categorization resulted in 934 workers with >1 years and 177 workers with
<1 year of potential for high exposure to AN. The remainder of the analyses focused on the
934 workers with >1 year in a job with potential for high exposure. Examination of the person-
years distribution by age group in these 934 workers revealed that <2% of the person-years was
observed in persons over age 65, while 65% of the person-years was observed in the 15-44-year
age range. Thus, the follow-up time for this cohort may be inadequate to detect excess cancer
deaths.
Among the 934 workers, 68 deaths were observed compared to the 72.4 expected based
on mortality rates for the total male population of England and Wales (SMR = 0.93, 95% CI =
0.73-1.18). Of these observed deaths, 21 were attributed to cancer, with 9 specifically attributed
77
DRAFT - DO NOT CITE OR QUOTE
-------
to cancer of the trachea, bronchus, and lung. The SMRs for all malignant neoplasms and cancer
of the trachea, bronchus, and lung are 1.10 (95% CI = 0.72-1.70) and 1.20 (95% CI = 0.58-
2.17), respectively. Incidentally, the number of observed stomach cancer deaths was
significantly higher than expected (5 observed vs. 1.9 expected, SMR = 2.63, 95% CI = 0.96-
5.83).
The distribution of cancer deaths was also examined by age at death (age groups: 15-44,
45-54, 55-64, and >65 years). No significant differences between the number of observed and
expected all-cause deaths were found in any of the age groups. Of the 14 deaths observed in the
15-44-year age group, 3 deaths were attributed to cancer of the trachea, bronchus, and lung, a
significant increase over the 0.7 expected value (SMR = 4.28, 95% CI = 1.09-11.66). No
significant differences between observed and expected cancer-specific deaths were noticed in
any of the other age groups.
All-cause deaths were also stratified based on the year of first exposure to determine if
cancer risks were higher for those who were exposed in the early years of the factories'
operations (usually a surrogate for longer and often higher exposure and longer latency)
compared with more recent hires. Three time periods were examined: 1950-1958, 1959-1963,
and 1964-1968, with the observed number of deaths being 35, 21, and 12, respectively. No
significant differences between the number of observed and expected deaths were found.
Observed and expected deaths from all cancer types, including cancer of the trachea, bronchus,
and lung, were similar in all three time periods, with only a slight increase in the most recent
time period (i.e., 1964-1968) (all-cancer SMR = 1.36, 95% CI = 0.55-2.83; cancer of the
trachea, bronchus, and lung SMR = 1.87, 95% CI = 0.48-5.10).
In order to further examine latency, an analysis of cancer deaths by length of time since
first exposure was performed. No increases in cancers of the stomach or cancers of the trachea,
bronchus, and lung were noted with increasing time since first exposure. Similarly, the observed
increases in cancers of the stomach and of the trachea, bronchus, and lung could not be related to
duration of exposure, year of first exposure, or latency. It should be noted that the study power
to examine these parameters was quite limited. Furthermore, 25% of the cohort (foreign
workers, who may have had the highest exposure) were lost to follow-up, potentially introducing
a bias resulting from an under-ascertainment of deaths. Another shortcoming of this study was
that the comparison group selected to derive the expected values may not have been
representative of the other regions and the comparisons may have been biased downwards by the
healthy worker effect. This study was also limited by the short length of follow-up and small
number of deaths observed. Aside from the shortcomings of this study, if an effect of AN were
to be observed, one would expect cancer rates to be elevated in the group with the earliest years
of first exposure, which represents the workers with the highest potential for greater exposure,
longest follow-up, and longest latency. The fact that increases were observed in the most recent
group (i.e., those first exposed between 1964 and 1968) weakens the argument that there was an
78
DRAFT - DO NOT CITE OR QUOTE
-------
exposure effect demonstrated in this study, because the latter group probably had lower exposure
levels and fewer than 15 years latency and follow-up.
In the update of the above study, Benn and Osborne (1998) expanded the sample to
include workers employed in the same six AN polymerization and acrylic fiber spinning
factories from 1969 to 1978 and extended follow-up through 1991, allowing for at least 13 years
of follow-up. Added to the study were craftsmen with possible AN exposure, control laboratory
workers, other possibly exposed workers, and unexposed workers. The sample size was
2,763 men employed for >1 year. The exposure assessment in this study was based on a
threshold limit enacted by the British government in 1981, a few measurements from the late
1970s, and an exposure estimated for the period 1958-1977 by a chemist at one of the factories.
The available exposure measurements were found to be lower than the calculated 8-hour time-
weighted average (TWA) (0.4-2.7 vs. 20 ppm, respectively). Job titles were collapsed into three
AN exposure categories: polymer workers and spinners—"high"; craftsmen, control laboratory
workers, and others with possible AN exposure—"other or medium"; and all others into the
"little or no" AN exposure group. No details were provided on how these categories were based
on the previously described exposure estimates. Unlike the earlier study by Werner and Carter
(1981), death rates for England and Wales were used to calculate expected deaths for the
factories that were located in England and Wales. Scottish rates were used for the factories
located in Scotland and Northern Ireland (rates for Northern Ireland were not available).
The 13 years of follow-up resulted in a total 409 observed deaths, a value significantly
lower than the expected (SMR = 0.84, 95% CI = 0.76-0.93). This result demonstrates a likely
healthy worker effect. Stratifying by cause of death, no significant differences were noted
between observed and expected deaths (Table 4-11), except in the case of all-circulatory
diseases, in which the number of observed deaths was significantly lower than expected (SMR =
0.86, 95% CI = 0.75-0.99). Upon stratification by the level of AN exposure, no significant
excess in cause-specific mortality was observed (Table 4-11); however, an excess of lung,
trachea, and bronchial cancers in the highest exposure group was found. In addition, linear
regression analysis found a significant increasing trend between level of AN exposure and
stomach cancer mortality. The distribution of cause-specific deaths was also examined by age of
death, with age categorized in the same manner as by Werner and Carter (1981). No significant
difference was found between the observed number of deaths in each age category and the
expected value. This observation was also noted when all-cancer mortality was stratified by age
group. Interestingly, among the youngest cohort of AN-exposed workers, the number of
observed deaths from respiratory cancers was significantly higher than expected (5 observed vs.
0.8 expected, SMR = 6.10, 95% CI = 2.23-13.51).
79
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-11. Distribution of select mortalities among AN-exposed and
unexposed workers
Exposure level
Cause of death
High
Possible
Little or none
Total
Oa
Ea
SMR
Oa
Ea
SMR
Oa
Ea
SMR
Oa
Ea
SMR
All causes
170
181.2
0.94
(0.80-1.09)
97
124.7
0.78b
(0.63-
0.94)
142
179.6
0.79b
(0.67-
0.93)
409
485.5
0.84b
(0.76-0.93)
All cancers
58
50.1
1.16
(0.89-1.49)
22
35.9
0.61b
(0.39-
0.91)
41
51.1
0.80
(0.58-
1.08)
121
137.1
0.88
(0.74-1.05)
Lung, tracheal,
bronchial
cancer
27
19.1
1.41
(0.95-2.03)
7
13.3
0.52
(0.23-
1.04)
19
19.1
0.99
(0.62-
1.52)
53
51.5
1.03
(0.78-1.34)
Stomach
cancer0
7
4.2
1.66
(0.73-3.30)
3
2.9
1.03
(0.26-
2.81)
1
4.3
0.23
(0.01-
1.15)
11
11.4
0.96
(0.51-1.68)
Circulatory
disease
81
86.9
0.93
(0.74-1.15)
49
59.1
0.83
(0.62-
1.09)
70
86.2
0.81
(0.64-
1.02)
200
232.2
0.86b
(0.75-0.99)
aO = observed deaths; E = expected deaths.
Statistically significant (p < 0.05).
"Statistically significant trend (p < 0.05).
Source: Amended from Benn and Osborne (1998).
Data were also analyzed by stratifying by year of first exposure, time since first exposure,
and length of exposure. The latter variables were divided into the following groups: <5, 5-10,
10-15, and >15 years. When cause-specific deaths, including all cancers and respiratory
cancers, were examined within these categories, no significant differences were noted between
observed and expected values. However, a significant increasing trend was noted for all deaths
and circulatory diseases based on time since first exposure but not with length of exposure. The
analysis that focused on cause-specific mortality and the year of first exposure differed from the
previous study by Werner and Carter (1981) in that workers were divided into the following
three groups: pre-1960, 1960-1968, and post-1968. No differences in all-cause or cancer deaths
were noted between the observed and expected values, but a significant increase in respiratory
cancer deaths was noted in the subgroup of workers that were exposed post-1968 (7 observed
deaths vs. 2.6 expected deaths, SMR = 2.70, 95% CI = 1.18-5.32).
This study increased the number of observed deaths sixfold as compared with the number
of deaths reported in the previous study by Werner and Carter (1981). The number of cancer
deaths, including deaths from lung cancer, also rose proportionally. However, the number of
observed deaths only represented less than 15% of the studied cohort and thus may have
contributed to not finding an overall statistically significant association between lung cancer and
80
DRAFT - DO NOT CITE OR QUOTE
-------
exposure to AN. As seen in the initial study, cancers of the trachea, bronchus, and lung
(respiratory cancers) were elevated in the youngest age group. While significant, this statistic
was based on five subjects in the younger age group. Also noted was the fact that there was a
significant elevation of deaths from cancers of the trachea, bronchus, and lung in the workers
who were first exposed in 1969 and later, rather than those first exposed prior to that year.
Smoking history, an important confounder when assessing cause-specific lung cancer, was not
addressed. The analyses on the high-exposure group included fewer than 800 workers from the
total cohort; therefore, power to detect trends and excess deaths was limited in this part of the
publication. Although not statistically significant, an increased SMR for lung, tracheal, and
bronchial cancers was observed in the high exposure group (SMR = 1.41, 95% CI = 0.95-2.03,
n = 27). The finding of an increase in stomach cancer in the high exposure group is interesting,
but the small total number of stomach cancer deaths (n = 11) makes rigorous analysis difficult.
This study had no quantitative AN levels, use of a population reference group could have
masked associations, and statistical power needed to detect rare cancers such as stomach cancer
was low. There was no evidence of a dose-related trend between increasing AN exposure and
respiratory cancer, which is not surprising given study deficiencies and the lack of quantitative
exposure data.
Acrylic fiber factory in Italy
A retrospective cohort study was conducted with 671 male workers who had at least
12 months of exposure to AN (or mixed exposure to AN and dimethylacetamide) at an acrylic
fiber factory in Venezia, Italy (Mastrangelo et al., 1993). Mortality patterns were examined to
determine whether excess cancer cases were related to these exposures. Occupational exposure
to AN occurred between 1959 and the end of 1988, with mortality tracked until the end of 1990,
thus allowing at least 2 years of follow-up for every person in the cohort. Workers with past
exposure to vinyl chloride or benzidine were excluded from the study cohort. Study participants
were categorized based on their level of AN exposure as follows: (1) high exposure to AN only,
(2) low exposure to AN plus exposure to dimethylacetamide, and (3) episodic exposure to AN
plus exposure to dimethylacetamide. SMRs were calculated based on the observed deaths and
the expected number of deaths in the general population. The expected death rate took into
account age, gender, year, cause, and person-time.
No significant difference in all-cause mortality was observed during 1959-1990, with
only 32 deaths (4.7% of total cohort) being reported. A total of 12 deaths from cancer was
observed in the total cohort. This observation was slightly, but not significantly, higher than the
8.73 expected cancer deaths. None of these cancer deaths was observed among the 100 workers
who were only exposed to high levels of AN. Of the 12 cancer cases, there were 2 lung cancer
cases and 4 intestinal cancer cases. The lung cancer cases were observed among the 272 workers
who had discontinuous but episodic exposure to AN and dimethylacetamide, and the number was
81
DRAFT - DO NOT CITE OR QUOTE
-------
not significantly different than the expected number of lung cancer cases. The intestinal and
colon cancer cases were equally distributed between the groups that had concomitant exposure to
dimethylacetamide and were significantly higher than expected in both groups.
When analyses were stratified by duration of exposure and time since first exposure, no
relationship was found with regard to all-cause mortality or all cancers. Among individuals
exposed for 1-4 years, significant differences were noted between the number of observed
testicular, rectal, intestinal, and colon cancer deaths and the expected values. However, the
number of cancer-specific deaths was small, with only one or two deaths being attributed to the
above cancer types.
In addition to questions about the comparison group, the ability of this study to greatly
inform an assessment of the relationship between AN and cancer is quite limited due to its small
sample size. The follow-up time is also short, as reflected by the fact that less than 5% of the
cohort was deceased by the end of the follow-up period. These factors combined may contribute
to the study's lack of sufficient power to determine if AN exposure is associated with cancer.
National Cancer Institute (NCI) cohort study
Many studies have investigated the relationship between AN exposure and cause-specific
death using a cohort assembled by the NCI (Starr et al., 2004; Marsh et al., 2001, 1999; Blair et
al., 1998). The studies were found to differ by the types of analyses used, comparison groups,
cohort subsets, or years of follow-up.
Blair et al. (1998) assembled a cohort of 25,460 workers (18,079 white males,
4,293 white females, 2,191 nonwhite males, and 897 nonwhite females) who were employed in
AN production or use beginning in the 1950s through 1983. The cohort included workers who
were employed prior to 1984 and after the start-up of AN operations (between 1952 and 1965) at
one of eight plants located in Alabama, Florida, Louisiana, Ohio, Texas, and Virginia. This
method allowed for the examination of both AN-exposed workers and unexposed workers.
Workers were followed through the end of 1989, allowing >6 years of follow-up.
Exposure was assessed for each plant by developing a quantitative estimate for each job,
department, and time period. Sources used to develop the estimates were walk-through surveys,
personal and area monitoring data, and interviews with longtime workers. The exposure
assessment for this study was more detailed (Stewart et al., 1998) than for any previously
published study. More than 10,000 estimates were developed for 3,662 job, department, and
plant combinations for a 30-year period of time. Individual worker exposure estimates were
developed, including estimates for workers whose exposures were difficult to estimate because
of their movement through all areas of a plant (i.e., maintenance workers). The estimation
methods were compared with actual data for validation.
SMR analyses were performed to compare observed mortality in both the exposed and
unexposed groups to expected numbers of deaths based on U.S. race- and gender-specific death
82
DRAFT - DO NOT CITE OR QUOTE
-------
rates. Subsequent analyses used the unexposed worker rates to develop internal comparison rate
ratios adjusted for birth year, plant, calendar time, race, gender, wage, and salary status. These
comparisons alleviate the healthy worker effect that is expected when using external general
population rates. Smoking history was obtained on a sample of workers by using a case-cohort
design to allow statistical adjustment for risk estimates calculated for known smoking-related
cancers, such as lung cancer. A 10% sample of living workers was chosen at random to be
interviewed regarding their smoking history. A 10% sample of persons deceased prior to
1983 was chosen, and next-of-kin interviews were attempted. Also, all brain and lung cancer
deaths that were not chosen in the 10% sample of deceased persons were also selected for next-
of-kin interview.
At the end of the study, the total person-years for the exposed workers was
348,642, while the person-years for the unexposed workers tallied 196,727. More than 66% of
the members of the cohort had at least 20 years of follow-up. A total of 1,217 exposed workers
and 702 unexposed workers were known to be deceased. For all-cause mortality, significant
lower numbers of observed deaths than expected were reported for both AN-exposed and
unexposed groups (Table 4-12). This pattern is also noted for all-cancer mortality among
exposed workers; however, no statistically significant difference between observed and expected
deaths was found among the unexposed workers. A similar observation was made for lung and
tracheal cancer death, with SMR values for AN-exposed and unexposed workers being 0.9 (95%
CI = 0.8-1.1) and 0.8 (95% CI = 0.6-1.1), respectively (Table 4-12). Interestingly, among the
exposed workers, the numbers of observed deaths from pancreatic cancer, lymphosarcoma, and
reticulosarcoma, as well as noncancer deaths like diabetes, cerebrovascular disease, and liver
cirrhosis, were significantly lower than expected values. Upon comparison of the exposed
workers to the unexposed workers, a small excess in lung and tracheal cancer deaths was
observed (see RR in Table 4-12).
Table 4-12. Distribution of select mortalities among AN-exposed and
unexposed workers
Cause of death
Exposed workers
Unexposed workers
RR
(95% CI)
Observed
SMR
95% CI
Observed
SMR
95% CI
All causes
1,217
0.7
0.6-O.T
702
0.7
0.7-0.8a
0.9 (0.8-1.0)
All cancers
326
0.8
0.7-0.93
216
0.9
0.8-1.0
0.8 (0.7-1.0)
Lung and tracheal cancer
134
0.9
0.8-1.1
59
0.8
0.6-1.1
1.2 (0.9-1.6)
Pancreatic cancer
10
0.5
0.3-0.93
13
1.2
0.7-2.1
0.4 (0.2-1.0)
Cerebrovascular disease
37
0.5
0.4-0.73
23
0.5
0.4-0.83
0.9 (0.5-1.6)
aStatistically significant (p < 0.05).
Source: Amended from Blair et al. (1998).
83
DRAFT - DO NOT CITE OR QUOTE
-------
Data were stratified by time since first exposure (<10, >10-20, and >20 years) to examine
any evidence of latency effects. Using the unexposed group for comparison, no significant
increases in cancer deaths were observed among AN exposed workers in any of the latency
intervals examined. The RRs for lung cancer mortality were increased for the longer latency
periods; however, these RRs and the overall trend were not statistically significant. The RRs
were 0.4, 1.6, and 1.3 in the <10-years exposure group, the >10-20-years group, and the
>20 years group, respectively.
Cumulative exposure was examined by stratifying the data into five exposure groups:
<0.13, >0.13-0.57, >0.57-1.5, >1.5-8.0, and >8.0 ppm-years. There was no evidence of a dose-
response effect across the quintiles of exposure when death rates for all cancers combined were
examined. Similarly, lung cancer death rates did not increase across the exposure categories
with rate ratios of 1.1, 1.3, 1.2, 1.0, and 1.5, respectively. A sufficient number of lung cancer
deaths were available to analyze the relationship between latency and cumulative exposure.
Among the lung cancer deaths that occurred within 10 years since first exposure to AN, there
was no significant increase in deaths with increasing exposure. This observation is also noted for
lung cancer deaths occurring 11-19 years after first exposure to AN. For the workers with
>20 years since first exposure, a statistically significant rate ratio of 2.1 (95% CI = 1.2-3.8) was
reported in the highest cumulative exposure quintile; however, there was not a significant
exposure-response trend (p = 0.11).
The risk of lung cancer mortality was also analyzed by other exposure variables,
including duration, intensity, frequency of peak exposures, and cumulative exposures,
considering different lag periods and adjustment factors. Most exposure variables showed a
nonsignificant increase in lung cancer deaths among workers with the highest level of AN
exposure, but none yielded a strong exposure-response gradient or statistically significant
exposure-response trend. No exposure-response patterns were observed for duration of exposure
at any intensity level.
Additional analyses focused on the workers with 20 or more years since first exposure in
order to examine the factors that might contribute to increased risk of lung cancer at the highest
cumulative exposure level in this group. The increased risk was not observed when analyses
were restricted to workers first employed between 1960 and 1969. This could be because
workers who were first employed between 1960 and 1969 would have at least 10 fewer years of
follow-up than those hired before 1960 and would tend to be younger. Also, having a later hire
date would allow workers in this group less time to work and accumulate exposure, so they
would be less likely to be included in the highest quintiles of exposure. This is confirmed by the
fact that 42 lung cancer cases were observed in the two highest quintiles for those first hired
before 1960, and 9 were observed in the two highest quintiles for those first hired between
1960 and 1969. The increase in risk in the highest exposure quintile was evident in both wage
and salary employees and for both fiber and nonfiber plants for this group that was followed for
84
DRAFT - DO NOT CITE OR QUOTE
-------
20 or more years after first exposure. None of the trend tests for cumulative exposure in workers
with 20 or more years since first exposure was significant.
Smoking history was sought on a 10% sample of study subjects. A total of 2,655 workers
was identified for interview, and 1,890 (71%) of this group were interviewed. For lung cancer
deaths, 64 next-of-kin interviews were conducted. Additional analyses were performed by
controlling for smoking status. These analyses did not change previous results, although they did
result in a slight reduction in the risk ratios in the highest quintile of exposure. Blair et al. (1998)
stated that they assumed that if smoking data were available for the full cohort, then the smoking
adjustment would yield the same proportion of effects in the full cohort as in the subcohort.
In summary, the sample size and follow-up time in this study were likely large enough to
detect any substantial elevation of cause-specific cancer deaths. For lung and tracheal cancer
deaths, an elevated RR of 1.2 (95% CI = 0.9-1.6) was observed in AN-exposed workers
compared to the unexposed worker population. With lung cancer deaths as the surrogate for the
development of lung cancer and less than 10% of the worker population being deceased at the
time of the study evaluation, lung cancer risk may have been underestimated. The study authors
state that increased risk with latency or cumulative exposure in separate analyses was not
demonstrated. This observation could be attributable to the reduced level of power in the
subgroup analyses. However, a statistically significant RR of 2.1 (95% CI = 1.2-3.81) was
observed in the highest cumulative exposure quintile among workers with >20 years since first
exposure. Overall, this study provides suggestive evidence with regard to AN exposure and lung
cancer.
Marsh et al. (1999) added 7 years of follow-up and focused on analyzing a subset of the
original cohort of Blair et al. (1998). The subcohort was comprised of 992 white male workers
who had worked for >3 months between 1960 and 1996 at a chemical plant in Ohio. Analyses
included the calculation of SMRs and RRs for categories of exposure and latency for lung cancer
and the calculation of SMRs (based on regional mortality rates) and RRs for stomach, prostate,
large intestine, and lymphohematopoietic cancers by cumulative exposure level.
Exposure assessment in this cohort mirrored that of the original cohort study by Blair et
al. (1998) with the following modifications. A panel of industrial hygienists used all of the
exposure data collected to assign calendar time-specific categories to each job title. The
following categories were designated: <0.2, >0.2-2.0, 2.1-20.0, and >20.0 ppm. Job titles were
also assessed for the potential for exposure to nitrogen products in a qualitative manner
(potential, no potential). Marsh et al. (1999) mentioned that other known potential occupational
hazards at the plant included asbestos, 1,3-butadiene, and depleted uranium; exposure to these
chemicals was not assessed. However, in the exposure assessment by Stewart et al. (1998) on
this cohort, these chemicals were not singled out as impactful potential occupational hazards.
For AN exposure, the quantitative evaluations for each job title were used to compute three time-
dependent measures of exposure for each worker who held any job where AN exposure was
85
DRAFT - DO NOT CITE OR QUOTE
-------
possible. These measures included duration of exposure, cumulative exposure, and average
intensity of exposure. Cumulative exposure was recorded in ppm-years, with workers being
assigned to one of the following categories: >0-0.139, 0.14-0.579, 0.58-1.509, 1.51—
7.999 ppm-years, and >8.000 ppm-years. Unlike the study by Blair et al. (1998), which
measured person-year accumulation from the first day of employment at the plant to the end of
1989 or date of death or last date known to be alive, Marsh et al. (1999) extended their 1960-
1988 job exposure matrix to include new job categories and monitoring data through 1996.
Thus, the exposure assessment by Marsh et al. (1999) covers all jobs held from the beginning of
the plant operation (1960) through the end of 1996.
The total cohort of 992 workers included 474 workers who never worked in a job where
AN exposure was possible and 518 workers who were potentially exposed to AN in at least one
of their jobs. A total of 110 deaths was observed in the cohort through the end of 1996.
Smoking history was collected for 90.3% of the total cohort and 93.2% of the AN-exposed
group. The prevalence of smokers was similar between the two groups with 58% of the
unexposed workers being smokers and 62.5% of the exposed workers being smokers. However,
the prevalence of smoking in the exposed group was associated with the level of cumulative AN
exposure and increased with increasing levels of exposure.
Of the 110 deaths observed, 43 deaths were attributed to cancer. The observed number of
cancer deaths did not differ from expected values based on the U.S. mortality rates or the
regional death rates, with SMRs of 0.98 (95% CI = 0.71-1.32) and 0.97 (95% CI = 0.70-1.31),
respectively. Fifteen of the observed cancer deaths were attributed to respiratory cancers, which
showed a similar pattern of nonsignificant SMRs <1 (Table 4-13). With the exception of bladder
tumors, none of the other site-specific cancers evaluated showed a significant difference between
observed and expected values using either the U.S. mortality rates or the regional mortality rates.
The number of bladder cancer deaths among workers was significantly higher than expected
(SMR = 4.5, 95% CI = 1.23-11.53); however, all the deaths occurred among workers not
exposed to AN.
86
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-13. Distribution of mortality among AN-exposed workers
Exposed workers
Unexposed workers
Overall cohort
Cause of death
Ob
SMR
(95% CI)
Ob
SMR
(95% CI)
Ob
SMRus
(95% CI)
SMRregionai
(95% CI)
RR
(95% CI)
All causes
41
0.58a
0.42-0.79
69
0.78b
0.60-0.98
110
0.64a
0.53-0.77
0.69a
0.57-0.83
0.74
0.5-1.1
All cancers
17
0.88
0.52-1.42
26
1.04
0.68-1.53
43
0.98
0.71-1.32
0.97
0.70-1.31
0.87
0.4-1.7
Respiratory
cancer
9
1.28
0.58-2.42
6
0.64
0.24-1.40
15
0.88
0.49-1.43
0.92
0.51-1.51
1.98
0.6-6.9
Bladder and
other urinary
0
0.00-8.23
4
7.01a
1.91-17.96
4
4.50a
1.23-11.53
3.93a
1.07-10.06
Not
determined
cancer
aStatistically significant (p < 0.05).
Ob = observed deaths
Source: Amended from Marsh et al. (1999).
For respiratory cancers, there were nine observed deaths in the exposed group and six in
the unexposed group, leading to a risk ratio of 1.98 (95% CI = 0.6-6.9) (Table 4-13). This
indicates that there was nearly 2 times the rate of lung cancer deaths in the exposed group as in
the unexposed group. Respiratory cancer deaths were further examined using RR regression
models that were able to adjust for age and calendar time as well as one other potential
confounding factor, such as year of hire or smoking. Two variables were marginally associated
with lung cancer risk in the total cohort (exposed and unexposed workers): duration of
employment and time since first exposure. Smoking was not related to lung cancer risk;
therefore, it is probable that there was misclassification of this risk variable, and use of smoking
information in models of time-related exposure variables to control for a smoking effect was not
possible. Regression analyses performed using categories of duration of exposure, cumulative
exposure, and average exposure compared to no exposure were not more informative than
previous analyses that compared exposed to unexposed rates. For each analysis, the exposed
worker RR was approximately 2 times that for unexposed workers across all categories of
exposure (Table 4-14). The risk ratios, derived in the regression analysis, increased with
exposure level and time since first exposure, though the small numbers of observed deaths may
have hindered the observation of a statistically significant increasing trend.
87
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-14. Summary of relative regression analyses for cancer of the
bronchus, trachea, and lung
Exposure measure
Category
Observed deaths
RR
95% CI
Duration (yrs)
Unexposed
6
1.00
>0-4.9
3
1.71
0.25-8.94
5.0-13.9
3
2.28
0.35-11.38
14.0+
3
2.15
0.34-10.70
Cumulative exposure
Unexposed
6
1.00
(ppm-yrs)
>0-7.9
2
1.96
0.81-12.04
8.0+
7
2.07
0.58-7.58
Cumulative exposure
Unexposed
6
1.00
(ppm-yrs)
>0-7.9
2
1.97
0.18-12.10
8.0-109.9
4
2.15
0.43-9.33
110.0+
3
1.97
0.31-9.42
Average exposure (ppm)
Unexposed
6
1.00
>0-4.9
3
1.97
0.31-9.54
5.0-11.9
3
1.70
0.26-8.26
12.0+
3
2.64
0.42-12.67
Source: Amended from Marsh et al. (1999).
The risk ratio for all-cancer mortality and AN exposure derived by Marsh et al. (1999)
was similar to the risk ratio derived in the larger-scale study by Blair et al. (1998). Despite the
limitations of the small number of observed deaths and the focus on only white males, this study
does indicate that there may be an association between AN exposure and increased risk in
respiratory cancer deaths.
Marsh et al. (2001) performed a sensitivity analysis on data from the original cohort of
Blair et al. (1998), examining dependency of lung cancer RR estimates on selection of referent
populations. Exposure categorization from the Blair et al. (1998) study was retained for this
analysis, but the comparison groups differed. Mortality analyses were performed using U.S.
mortality rates and local county rates. SMRs were calculated for both AN-exposed and
unexposed workers.
Upon comparison with U.S. and regional mortality rates, both exposed and unexposed
workers were found to have significantly lower rates of all-cause mortality. Using the U.S.
morality rates, the SMRs for the unexposed group and the exposed group were 0.75 (based on
702 deaths, 95% CI = 0.7-0.8) and 0.66 (based on 1,217 deaths, 95% CI = 0.6-0.7), respectively.
The SMRs for all-cause mortality utilizing regional rates were nearly identical to those using
U.S. rates. Additionally, the number of lung cancer deaths observed among workers, both
exposed and unexposed, was significantly lower than regional estimates, with SMRs for each
group of workers being 0.74 (95% CI = 0.6-0.9) and 0.68 (95% CI = 0.5-0.9), respectively,
compared with U.S. population rates. These findings indicate a potential healthy worker effect.
88
DRAFT - DO NOT CITE OR QUOTE
-------
Observed lung-cancer deaths were stratified by the cumulative exposure to AN and time since
first exposure to derive RRs and regional-mortality-rate-based SMRs (Table 4-12). Workers
with the highest cumulative exposures and at least 20 years of exposure were twice as likely to
have lung cancer as unexposed workers (RR = 2.1, 95% C.I = 1.2-3.8). The numbers of
observed and expected deaths in this highest exposure, longest duration category were about the
same using external comparisons (SMR = 1.07, 95% CI = 0.7-1.6). This illustrates how the use
of an external comparison population can mask an apparent association; the discrepancy is most
likely attributable, at least in part, to the healthy worker effect. There was no evidence of a
linear exposure-response relationship for lung cancer, based on the categories used to assess time
since first exposure. SMRs in each category did appear to increase with increasing cumulative
exposure category (Table 4-15).
Table 4-15. Distribution of observed lung cancer deaths among AN-exposed
workers, using regional rates for comparison
Cumulative
exposure
(ppm-yrs)
Time since first exposure
Less than 10 yrs
10-19 Yrs
At least 20 yrs
Ob
RR (CI)
SMR (CI)
Ob
RR (CI)
SMR (CI)
Ob
RR (CI)
SMR (CI)
>0-0.13
7
0.4
(0.2-1.2)
0.72
(0.3-1.5)
9
0.5
(0.5-3.2)
0.71
(0.3-1.4)
11
1.1
(0.6-2.2)
0.56
(0.3-1.0)
0.13-0.57
3
0.4
(0.1-1.4)
0.63
(0.1-1.8)
12
2.6a
(1.2-5.7)
1.18
(0.6-2.1)
11
1.0
(0.6-2.2)
0.57
(0.3-1.0)
0.57-1.50
2
0.4
(0.1-1.6)
0.70
(0.1-2.5)
10
2.0
(0.SM.8)
1.01
(0.5-1.8)
16
1.2
(0.5-2.1)
0.71
(0.4-1.2)
1.50-8.00
2
0.4
(0.1-2.0)
0.87
(0.1-3.1)
7
1.2
(0.5-3.1)
0.66
(0.3-1.4)
18
1.2
(0.6-2.2)
0.61
(0.4-1.0)
>8.00
1
0.4
(0.1-3.1)
0.81
(0.02-4.5)
4
0.9
(0.3-1.2)
0.54
(0.2-1.4)
21
2.1a
(1.2-3.8)
1.07
(0.7-1.6)
Statistically significant (p < 0.05).
Ob = observed deaths
Source: Amended from Marsh et al. (2001).
To look for plant-specific risks, the lung-cancer deaths in each of the eight study plants
were analyzed separately by cumulative level of exposure. Only one plant showed an increased
risk of lung-cancer deaths among the exposed workers, with the highest level of exposure (i.e.,
>8 ppm-years) having an SMR of 2.68 (based on 10 deaths, 95% CI = 1.3-4.9). There was no
comparison of process differences in the plants, thus raising uncertainty as to whether workers at
this plant were exposed to other potential carcinogens or if these findings were due to chance. In
addition, the small number of observed lung-cancer deaths in the plant-specific exposure
categories lowers the level of power this study had to discern trends in SMRs that might provide
more information on the association between AN exposure and cancer.
89
DRAFT - DO NOT CITE OR QUOTE
-------
In summary, Marsh et al. (2001) focused on external comparison groups rather than the
available unexposed worker cohort. The use of an external comparison group, rather than a
cohort more comparable to the exposed population (i.e., internal comparison group), may subject
the relationship between exposure and lung cancer to biases such as the healthy worker effect.
Internal comparison groups are generally preferred since they tend to be less influenced by the
healthy worker effect, other sources of selection bias, and confounding (Checkoway et al., 1989).
Marsh et al. (2001) mentioned potential occupational hazards such as asbestos, 1,3-butadiene,
and depleted uranium not being assessed. Stewart et al. (1998) conducted an exposure
assessment for the cohort of workers exposed to AN. Exposure to 340 substances other than AN
was assessed qualitatively. 1,3-Butadiene and depleted uranium were not among the chemicals
with exposures of >20,000 person-years. Additionally, Blair et al. (1998) did not observe any
deaths from asbestosis or mesotheliomas, which would have been indicative of asbestos
exposure. The reanalysis by Marsh et al. (2001), particularly the RR results, does support the
association, found by Blair et al. (1998), between AN exposure and increased lung cancer
mortality risk among workers with the highest exposure level, but the healthy worker effect,
small sample sizes within subcategories, use of external controls, and short follow-up may limit
the detection of stronger support for such an association.
Data from this cohort were reanalyzed employing semiparametric Cox regression models
with time-dependent covariates to estimate additional risk of death from lung cancer for several
different AN occupational exposure scenarios (Starr et al., 2004). The "cumulative exposure
estimate" was a time-dependent covariate, and "plant worked" was a time-independent covariate.
The analysis focused on the largest race-sex group (18,079 white males) from the original Blair
et al. (1998) study. The Cox models allowed for the calculation of the cumulative risk of dying
from a disease by a certain age. The outcome measurement used was the risk of dying from lung
cancer by age 70 years. Baseline rates were developed for the unexposed worker population and
for three different AN exposure scenarios: (1) early intense exposure, (2) long moderate
exposure, and (3) late intense exposure. All scenarios provided 50 ppm-years of cumulative
exposure by age 55 years. The increased number of lung cancers per 1,000 workers that would
develop by age 70 was calculated as 0.77-1.56, 0.74-1.50, and 0.68-1.56, respectively (the
ranges reflect the use of different plant-specific baseline rates). The upper bound of additional
risk was in the range of 7.5-15.1 per 1,000 workers with the upper bound on the exposure
parameter being 0.0048 per ppm-working year. It is important to note that this study analysis did
not control for smoking. Also, based on the extent of exposure misclassification, any exposure-
response association may have been underestimated.
In summary, the study by Blair et al. (1998) reported an elevated risk of lung cancer
deaths among AN-exposed workers as compared to the unexposed workers, particularly among
workers in the highest cumulative exposure quintile with >20 years since first exposure (RR =
2.1, 95% C.I. = 1.2-3.8). However, the short follow-up may have contributed to the study's
90
DRAFT - DO NOT CITE OR QUOTE
-------
inability, overall, to demonstrate an increased risk associated with latency or cumulative
exposure. Similar to Blair et al. (1998), Marsh et al. (1999) found an increased risk of lung
cancer deaths among AN-exposed workers as compared to the unexposed workers, though the
point estimate of 1.98 was not statistically significant. However, Marsh et al. (1999) only
focused on a small subset of the NCI cohort. Though the small sample size and the study
inclusion of only white males reduces the statistical power and generalizability of the study, the
observation of the excess lung cancer risk in Marsh et al. (1999) does provide some weight with
regard to hazard inferences.
Marsh et al. (2001) utilized primarily an external comparison group and, while they
provide better stability of comparison rates than internal controls as used by Blair et al. (1998), a
limitation of external controls as the referent population is the risk of having potential
associations masked by a healthy worker effect or other factors that may be more similar within
an occupational cohort than between the cohort and the general population. The results from
Marsh et al. (2001) indicate the presence of such an effect dampening the observation of an
association between AN and cancer.
Though the studies by Marsh et al. (2001) and Blair et al. (1998) have a relatively large
cohort size, the number of observed deaths is small. These studies, though they rely on cancer-
specific mortality ratios rather than cancer incidence as in the Wood et al. (1998) study, have
several advantages to the Wood et al. (1998) study. The low SMR for overall mortality reported
by Wood et al. (1998) suggests the potential presence of a number of biases, including the
healthy worker effect, incomplete cohort identification, and incomplete ascertainment of the
outcome measure. Furthermore, Blair et al. (1998) provide a better job exposure matrix than
Wood et al. (1998). Although cancer incidence is typically preferred, mortality rates do serve as
a good surrogate for incidence for some cancers like lung cancer. On the other hand, these
mortality studies may miss associations with treatable cancers such as prostate cancer.
The small percentage of deaths in the NCI cohort suggest a young mean age, which can
impact statistical power. Thus, the observation of a statistically significant elevation in SMR
among those with the longest latency and high cumulative exposure is noteworthy. Additional
follow-up of this cohort may be useful to further assess the association between AN exposure
and cancer suggested by the excess in lung cancer deaths that has been observed among workers
exposed to high levels of AN.
Case-control studies
A large case-control study conducted in seven European countries investigated the
association between occupational exposure to vinyl chloride, AN, and styrene and the risk of
lung cancer (Scelo et al., 2004). The study included new cases of lung cancer occurring between
1998 and 2002 in 15 centers in six Central and Eastern European countries and in Liverpool in
the United Kingdom. Controls, consisting of subjects hospitalized in general public hospitals in
91
DRAFT - DO NOT CITE OR QUOTE
-------
the same areas as cases, were frequency matched to cases based on age and gender. Controls had
to have been hospitalized within 3 months of diagnosis of the case and could not have cancer or
any tobacco-related diseases. A total of 3,403 cases and 3,670 controls met the study inclusion
criteria; after exclusions and refusals to participate, the final study group included 2,861 cases
and 3,118 controls.
In order to ascertain exposure and lifestyle information, such as tobacco consumption,
each study participant was interviewed using a standard questionnaire. Exposure assessment was
obtained from the work history portion of the questionnaire, where information was collected for
each job held at >1 year. Experts evaluated the frequency and intensity of exposure to AN,
among other agents, for each job held by each study subject. The following exposure models
were constructed for analysis: (1) duration of exposure, (2) weighted duration of exposure
(which considered both duration and frequency), and (3) cumulative exposure in ppm-years.
Models included age, gender, center, tobacco consumption, and other occupational factors.
A total of 10,555 jobs were held by the 2,861 study participants with lung cancer, and a
total of 11,174 jobs were held by the 3,118 controls. AN exposure was associated with 48 jobs
held by the cases and 26 jobs held by the controls. Thirty-nine of the 2,861 cases and 20 of the
3,118 controls were characterized as being exposed to AN, resulting in a significant odds ratio
(OR) of 2.20 (95% CI = 1.11-4.36) (i.e., cases were 2 times more likely to be exposed to AN
than controls). However, it was found that more than half of the study participants that were
exposed to AN were also exposed to styrene. Further analysis was conducted on individuals not
exposed to styrene (17 cases and 10 controls). This resulted in a similar OR estimate but with a
wider CI due to the smaller sample size (OR = 2.08, 95% CI = 0.82-5.27). Increasing linear
trends for lung cancer were noted for both weighted duration of exposure (p = 0.05) and
cumulative exposure (p = 0.06) (Table 4-16). Additional analyses employed a 20-year lag for
exposures, which did not change the results appreciably. The authors reported an increased risk
of exposure among lung cancer cases diagnosed before the age of 60 (43% of the cases), where
the OR for ever being exposed was 2.79 (95% CI = 1.01-7.70), while the OR for ever being
exposed among those over 60 years was 1.02 (95% CI = 0.35-2.92). An age-exposure
interaction test yielded nonsignificant results.
92
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-16. ORs of lung cancer for AN exposure
Exposure measure
Cases
Controls
OR
95% CI
Weighted duration of exposure (yrs)
Not exposed
2,822
3,098
1.00
0.01-1.00
13
9
2.03
0.72-5.73
1.01-2.25
9
5
2.73
0.73-10.20
>2.25
17
6
2.91
0.87-9.79
Linear trend: p = 0.05
Cumulative exposure (ppm-yrs)
Not exposed
2,822
3,098
1.00
0.01-0.46
13
9
2.03
0.72-5.73
0.47-1.61
10
4
2.76
0.68-11.22
>1.61
16
7
2.87
0.85-9.66
Linear trend: p = 0.06
Source: Amended from Scelo et al. (2004).
Because of the personal interviews conducted with each case and control, this study was
able to provide an in-depth assessment of potential confounding factors, such as smoking history
and lifestyle factors. The exposure assessment, though different than for the recent generation of
cohort studies where actual plant measurements were possible, utilized reasonable and
standardized methods to assign various levels of AN exposure to different jobs. As a multi-
industry study, the possibility of exposure misclassification in this study was probably greater
than in single-industry studies, though exposure misclassification is probably nondifferential
with respect to disease and therefore would serve to lessen the outcome measures (ORs)
calculated. The key observation in this study is the fact that, even after adjustment for
confounding factors such as smoking history, an association between AN exposure and lung
cancer incidence was observed. This observation adds to the weight of evidence in support of an
association between AN and cancer.
Cross-sectional studies
A detailed cross-sectional study was conducted among workers at a Hungarian AN
factory in June 2000 (Czeizel et al., 2004). Of the 888 employees, 72 employees did not work
during the study time frame and 33 refused to participate. The remaining 783 employees were
interviewed, with information gathered on demographics, lifestyle and habits, occupational
exposures, and history of general and occupational diseases, among other factors. Medical
records aided in the validation of the workers' responses. Workers were categorized into three
groups based on level of contact with AN (i.e., direct exposure, indirect or sporadic exposure,
and no exposure). Since the interviews were done with living current workers, cancer
93
DRAFT - DO NOT CITE OR QUOTE
-------
information for any worker who had left the plant or died before the study was conducted was
anecdotal.
It is known that 12 former workers from the factory had died from cancer between
1990 and 1999, and none of these deaths was due to lung cancer. Of the 783 workers
interviewed, 12 workers were found in the interview sample that had cancer. Of these, only one
lung-cancer case was identified. This worker, categorized as having direct exposure to AN,
started working at the plant in 1973 and was diagnosed 15 years later. Five of the 12 cancer
cases were among the group (n = 452) thought to have the highest level of exposure to AN, while
4 cancer cases were noted among workers with no exposure to AN. No significant association
between AN exposure and the development of cancer was observed. The study mentioned three
of its shortcomings: persons with serious disorders who died or had premature pensions were
not included, updated information on occupational exposures was unavailable, and there was
difficulty in identifying appropriate controls. It should be noted that workers were not exposed
exclusively to AN but to a mixture of other chemicals as well. The small number of cancer
incidences and the identified shortcomings hinder the reliability of this study in evaluating the
association between AN exposure and the development of cancer.
Other supporting studies
In 2006, the Ohio Department of Health assessed the burden of cancer among residents of
Addyston, Hamilton County, Ohio, who lived near a thermoplastics manufacturing plant that
emitted AN and 1,3-butadiene into the environment (Ohio Department of Health, 2006). The
study population consisted of invasive cancer cases identified through the Ohio Cancer Incidence
Surveillance System between 1996 and 2003. The incidence of site-specific cancer was
compared to the expected number of cases, the latter being derived from national background
cancer incidence rates from NCI's Surveillance, Epidemiology and End Results (SEER) program
in 1998-2002 and region-specific cancer incidence rates from 1993 to 2003. A total of
55 invasive cancer cases were identified among the 1,010 residents in the area, with an SIR =1.8
(95% C.I. = 1.3-2.3) based on the 1998-2002 SEER age-specific incidence rates. Cancer of the
lung and bronchus was the most common cancer identified (13 cases, 23.6%), followed by
colorectal cancer (10 cases, 18.2%). The number of observed cancer cases in both instances was
higher than expected (lung and bronchus SIR = 3.2, 95% C.I. = 1.7-5.4; colon and rectum SIR =
3.0, 95%) C.I. = 1.5-5.6). SIRs based on the region-specific cancer incidence yielded similar
results. Although cancer incidence in this cohort was higher than expected, the association
between AN exposure and the incidence of lung cancer may be confounded by other risk factors,
such as smoking, that were acknowledged but not controlled in the analyses.
94
DRAFT - DO NOT CITE OR QUOTE
-------
Overall summary of epidemiology data
The early DuPont studies found potential increases in incidence and/or death from lung
cancer and prostate cancer among persons possibly exposed to AN. These studies were the
impetus for generating several additional studies spanning more than 20 years after the first
publication. Exposure assessment was not fully explored until recently, though older studies did
examine qualitative distributions of exposure. The cohorts ranged from a few hundred workers
to over 25,000 workers. The follow-up period for many of the studies was short, not allowing
half of the population to reach a terminal endpoint so that a health outcome such as cancer could
be observed. However, over time, the cohort studies increased in power by increasing the
number of workers, length of follow-up, and sophistication of the exposure assessment.
A composite of the major cohort studies reviewed, along with all-cancer SMRs, is
provided in Table 4-17. Table 4-18 summarizes SMRs for lung cancer, a cancer type that has
been assessed in most of the epidemiology studies reviewed. In both tables, the SMRs are based
on the cohort population most likely to be exposed to AN, as the actual cohort sample size in
many cases included an unexposed worker group. As in most studies, the number of deaths on
which the SMRs are based is a fraction (<33%) of the actual cohort studied; thus, the actual
cohort size may not reflect the relative number of deaths in the SMR calculation. These SMRs
were evaluated by the size of the exposed cohort, number of observed deaths, percentage of
observed deaths within each study, and year of publication to determine if any of these factors
was associated with increased SMR values. For both all-cancer mortality and lung-cancer
mortality, no discernable association was observed between increased SMRs and the size of the
exposed cohort, number of observed deaths, percentage of observed deaths within each study, or
year of publication (data not shown).
95
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-17. Derived all-cancer SMRs for major cohort studies on AN exposure
Reference
Study population
Comparison group
Potentially exposed
cohort
Observed
deaths
All-cancer
SMRa
DuPont
O'Berg (1980)
Male workers exposed to AN between 1950 and 1966 at a
DuPont Plant in South Carolina and followed through 1976.
DuPont Registry
l,128b
17
1.13
(0.68-1.78)
O'Berg et al. (1985)
As above but updated extended to 1983
DuPont Registry
1,345
36
1.14
(0.81-1.56)
Chenetal. (1987)
Male workers exposed to AN between 1944 and 1970 at
DuPont Plant in Virginia and followed through 1983.
White male subset of
U.S. population
1,083
18°
0.75
(0.44-1.16)
Chen et al. (1987)
As above.
DuPont Registry
1,083
18°
0.88
(0.54-1.37)
Wood et al. (1998)
Combined O'Berg et al. (1985) and Chen et al. (1987)
U.S. general population
2,559
126
0.78d
(0.64-0.93)
Wood et al. (1998)
As above.
DuPont Registry
2,559
126
0.86
(0.72-1.02)
Symons et al. (2008)
Update from Wood et al. (1998) with 11 yrs of follow-up
Regional Dupont workers
2,548
839
0.92
(0.81-1.04)
Symons et al. (2008)
As above.
U.S. general population
2,548
839
0.73d
(0.64-0.82)
NCI
Blair etal. (1998)
Workers employed in eight AN-producing facilities from
1950s to 1983, followed through 1989.
U.S. general population
25,460
326
0.80d
(0.70 -0.90)
Blair etal. (1998)
As above.
Unexposed workers
25,460
326
RR = 0.80
(0.7-1.0)
Marsh etal. (1999)
Subset of Blair et al. (1998).
County mortality rates
518
17
0.88
(0.52-1.42)
Marsh etal. (1999)
As above.
Unexposed workers
518
17
RR = 0.87
(0.4-1.7)
Marsh etal. (2001)
Same as Blair et al. (1998)
Regional mortality rates
25,460
-
American Cyanamid Company
Collins et al. (1989)
Male workers at two plants employed between 1951 to 1973,
followed throughl983.
White male subset of
U.S. population
1,774
43
1.01
(0.74-1.35)
Synthetic chemical plant
Waxweiler et al. (1981)
Chemical plant workers employed between 1942 and 1973,
followed through 1973.
White male subset of
U.S. population
4,806
101
1.18
(0.97-1.43)
96
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-17. Derived all-cancer SMRs for major cohort studies on AN exposure
Reference
Study population
Comparison group
Potentially exposed
cohort
Observed
deaths
All-cancer
SMRa
Table 4-17. Derived all-cancer SMRs for major cohort studies on AN exposure
Reference
Study population
Comparison group
Potentially
exposed cohort
Observed
deaths
All-cancer
SMR
Rubber industry
Delzell and Monson (1982)
White male rubber chemical plant workers employed for at
least 2 yrs between 1940 and mid-1971, followed through
mid-1978.
White male subset of
U.S. population
327
22
1.20
(0.77-1.79)
Netherlands cohort
Swaenetal. (1992)
Male workers exposed to AN in eight factories for >6 mos
before mid-1979, followed through 1987.
Dutch general population
2,842
42
0.83
(0.61-1.11)
Swaenetal. (1998)
As above, but followed through 1995.
Dutch general population
2,842
97
0.88
(0.72-1.07)
Swaen et al. (2004)
As above, but followed through 2000.
Dutch general population
2,842
146
0.89
(0.75-1.04)
BASF (Germany)
Thiess et al. (1980)
Male workers from 12 plants followed through mid-1978.
German mortality rates
1,469
27
1.32
(0.89-1.89)
Six factories (U.K.)
Werner and Carter (1981)
Male workers employed for at least 1 yr in one of six
factories from 1950 to 1968, followed through 1978.
Male mortality rates in
England and Wales
934
21
1.10
(0.70-1.65)
Benn and Osborne (1998)
As above but employed from 1969 to 1978, followed through
1991.
Mortality rates from
different European
countries
2,963
121
0.88
(0.73-1.05)
Acrylic fiber factory (Italy)
Mastrangelo et al. (1993)
General population
671
12
1.37
(0.74-2.33)
aSMR may be calculated from article based on available data.
bOnly workers employed for >6 mos.
°Based only on wage workers.
Statistically significant (p < 0.05).
97
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-18. Derived lung cancer SMRs for major cohort studies on AN exposure
Reference
Study population
Comparison group
Potentially exposed
cohort
Observed lung
cancer deaths
Lung cancer
SMRa
DuPont
O'Berg (1980)
Male workers exposed to AN between 1950 and
1966 at a DuPont Plant in South Carolina and
followed through 1976.
DuPont Registry
l,128b
7
1.35
(0.59-2.66)
O'Bergetal. (1985)
As above, but updated and extended to 1983.
DuPont Registry
1,345
14
1.21
(0.69-1.98)
Chenetal. (1987)
Male workers exposed to AN between 1944 and
1970 at DuPont Plant in Virginia and followed
through 1983.
White male subset of
U.S. population
1,083
5°
0.59
(0.22-1.32)
Chenetal. (1987)
As above.
DuPont Registry
1,083
5°
0.66
(0.24-1.46)
Wood etal. (1998)
Combined O'Berg et al. (1985) and Chen et al.
(1987).
U.S. general population
2,559
47
0.74d
(0.55-0.98)
Wood etal. (1998)
As above.
DuPont Registry
2,559
47
0.89
(0.66-1.17)
Symons et al. (2008)
Update from Wood et al. (1998) with 11 yrs of
follow-up
Regional Dupont workers
2,548
88
0.92
(0.75-1.14)
Symons et al. (2008)
As above.
U.S. general population
2,548
88
0.74d
(0.60-0.91)
NCI
Blair et al. (1998)
Workers employed in eight AN-producing facilities
from 1950s tol983, followed through 1989.
U.S. general population
25,460
134
0.90
(0.8-1.1)
Blair et al. (1998)
As above.
Unexposed workers
25,460
134
RR = 1.2
(0.9-1.6)
Marsh etal. (1999)
Subset of Blair et al. (1998).
County mortality rates
518
9
1.32
(0.64-2.42)
Marsh etal. (1999)
As above.
Unexposed workers
518
9
RR = 1.98
(0.96-3.63)
Marsh etal. (2001)
Same as Blair et al. (1998).
Regional mortality rates
25,460
134
0.74d
(0.62-0.87)
American Cyanamid Company
Collins et al. (1989)
Male workers at two plants employed between
1951 and 1973, followed through 1983.
White male subset of
U.S. population
1,774
15
1.00
(0.58-1.61)
98
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-18. Derived lung cancer SMRs for major cohort studies on AN exposure
Reference
Study population
Comparison group
Potentially exposed
cohort
Observed lung
cancer deaths
Lung cancer
SMR
Table 4-18. Derived lung cancer SMRs for major cohort studies on AN exposure
Reference
Study population
Comparison group
Potentially
exposed cohort
Observed lung
cancer deaths
Lung cancer
SMRa
Synthetic chemical plant
Waxweiler et al. (1981)
Chemical plant workers employed between 1942 and
1973, followed through 1973.
White male subset of
U.S. population
4,806
42
1.49d
(1.09-1.99)
Rubber industry
Delzell and Monson
(1982)
White male rubber chemical plant workers employed
for at least 2 yrs between 1940 and mid-1971,
followed through mid-1978.
White male subset of
U.S. population
327
9
1.52
(0.74-2.79)
Netherlands cohort
Swaenetal. (1992)
Male workers exposed to AN in eight factories for
>6 mos before mid-1979, followed through 1987.
Dutch general population
2,842
16
0.82
(0.48-1.30)
Swaenetal. (1998)
As above, but followed through 1995.
Dutch general population
2,842
47
1.10
(0.82-1.45)
Swaen et al. (2004)
As above, but followed through 2000.
Dutch general population
2,842
67
1.07
(0.84-1.35)
BASF (Germany)
Thiess et al. (1980)
Male workers from 12 plants followed through mid-
1978.
German mortality rates
1,469
11
1.85
(0.97-3.22)
Six factories (U.K.)
Werner and Carter (1981)
Male workers employed for >1 yr in one of six
factories from 1950 to 1968, followed through 1978.
Male mortality rates in
England and Wales
934
9
1.20
(0.58-2.20)
Benn and Osborne (1998)
As above but employed from 1969 to 1978, followed
through 1991.
Mortality rates from
different European
countries
2,963
53
1.03
(0.78-1.34)
Acrylic fiber factory (Italy)
Mastrangelo et al. (1993)
General population
671
2
0.77
(0.13-2.54)
aSMR may be calculated from article based on available data.
bOnly workers employed for >6 mos.
99
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-18. Derived lung cancer SMRs for major cohort studies on AN exposure
Reference
Study population
Comparison group
Potentially exposed
cohort
Observed lung
cancer deaths
Lung cancer
SMR
°Based only on wage workers.
Statistically significant (p < 0.05).
100
DRAFT - DO NOT CITE OR QUOTE
-------
To date, the largest cohort assessed in determining the relationship between AN and
cancer is the NCI cohort. The NCI/National Institute for Occupational Safety and Health
(NIOSH) cohort study of Blair et al. (1998) provides the strongest evidence for carcinogenicity,
although this evidence is not conclusive. This study of 25,460 subjects in eight plants was
designed to evaluate a relationship between AN exposure and site-specific cancer, including lung
cancer, an a priori hypothesis. Both AN-exposed and unexposed subjects showed a favorable
all-cause mortality rate compared to the all-cause mortality rate of the U.S. population.
Although only 5% of the cohort had died, this study has a large number of observed deaths that
could be analyzed. The study addressed known problems occurring with earlier studies by
quantifying exposures, estimating the effect of smoking, and using an internal control group of
unexposed workers.
The study by Blair et al. (1998) is of particular interest because a possible association
between AN exposure and death from lung cancer was observed among workers in the highest
level of AN exposure. Specifically, workers with at least 20 years since first exposure and
exposed to a high level of AN were 2 times more likely to die of lung cancer than unexposed
workers. However, though significance was found among workers exposed to high levels of
AN, as with other studies, the exposure-response analysis in the Blair et al. (1998) study did not
provide strong or consistent evidence for a causal association.
The observations of Swaen et al. (2004) of AN-exposed workers in the Netherlands and
of Benn and Osborne (1998) of AN-exposed workers in the United Kingdom support the
findings of Blair et al. (1998) and provide some evidence for carcinogenicity, although lung
cancer SMRs in these studies were not statistically significant. Other supporting evidence comes
from the observation that SMRs for lung cancer among subjects with the highest exposure (as
compared to national mortality rates) were not only >1, but also larger than the SMRs for low- or
no-exposed groups. Furthermore, the statistically significant increased OR for AN exposure in
the lung-cancer case control study of Scelo et al. (2004) provides weight for this association.
This study adjusted for effects related to individual smoking history and to a number of potential
coexposures found in a subject's occupational setting. The finding of lung-cancer risk increasing
with increasing exposure duration or with increasing cumulative exposure provides further
evidence of an association with AN.
Within the body of epidemiologic literature examining a potential relationship between
exposure to AN and cancer in occupational cohorts, the following shortcomings are noted: low
power from small numbers of exposed subjects, lack of quantitative exposure information on
individual study subjects leading to a greater potential for exposure misclassification bias,
assessment of an insensitive outcome, and insufficient follow-up period for cancer latency. In
addition, many of the studies used external comparison groups, and their results were subject to a
downward bias from the healthy worker effect.
101
DRAFT - DO NOT CITE OR QUOTE
-------
An alleged shortcoming of the Blair et al. (1998) study has been the observation that the
AN-exposed workers experienced a lower rate of lung-cancer deaths than in the general
population; however, this is likely a manifestation of the healthy worker effect. The use of
internal controls by Blair et al. (1998) was to draw inferences about AN risks. With cancer
mortality being used as a surrogate for developing cancer associated with AN exposure and with
a low percentage of the cohort having died, there may be an underestimate of the risk associated
with AN exposure. In addition, Selikoff and Seidman (1992) showed based on histopathological
analysis of a comparably large industrial cohort (17,800 asbestos workers) that as much as 13.7%
of lung cancer cases are misdiagnosed on death certificates, both as neoplasms and nonmalignant
diseases, with about 3% as asbestos only diseases (mesothelioma and asbestosis). Therefore,
Blair cohort may have as much as 10% of internal controls that are actually lung cancers. That
would cause a strong bias against finding a statistically significant relationship. Finally,
mortality data may be inadequate for assessing treatable cancers that could result from AN
exposure. The earlier studies based on the DuPont cohort utilized cancer incidence as well as
cancer mortality data. In these studies, although a better outcome measurement (i.e., cancer
incidence) is used, observed rates are compared through an SIR with company-wide or national
population statistics rates, rather than rates from an internal control group. Thus, a strong
healthy worker effect may hinder the observation of a cancer effect of AN exposure. Even in the
most recent DuPont cohort study (Symons et al. 2008) where an attempt was made to reduce the
bias due to healthy worker effect, there may be bias from the health worker effect. It should be
noted that unlike the other DuPont cohort studies, Symons et al. (2008), with over 50 years of
follow-up, did not report cancer incidence. Though the outcome measure of mortality in the
Blair et al. (1998) study is not as sensitive, this study has a relatively large number of lung-
cancer deaths and does compare cause-specific deaths between exposed and unexposed workers
as well as the general population. The resulting risk ratio may not achieve statistical significance
(as mentioned above, several factors may have caused bias to the null), but the point estimate
does indicate a potential excess in lung-cancer deaths among workers exposed to AN.
Furthermore, the observed RR in the group of workers with the highest cumulative exposure and
the longest time since first exposure was statistically significantly increased.
4.1.2.2.2. Epidemiological studies of AN in humans (noncancer effects). Sakurai et al. (1978)
performed a cross-sectional health examination of 102 male workers exposed to AN in six
acrylic fiber manufacturing factories in Japan in 1976. Also examined were 62 nonexposed
matched control workers from polyester fiber manufacturing plants, power supply plants, or
finishing branches of the acrylic fiber plants. The workers were selected by a random sampling
process. By experimental design, all exposed subjects were shift workers who had been exposed
to AN in the workplace for at least 5 years but had no history of exposure to other chemicals.
All subjects underwent medical examinations, and blood and urine samples were collected for
102
DRAFT - DO NOT CITE OR QUOTE
-------
chemical analysis. Clinical chemistry analyses included urinary protein and other parameters,
including Hb, total cholesterol, AST, ALT, alkaline phosphatase, cholinesterase, y-glutamyl
transpeptidase (y-GTP), and lactate dehydrogenase (LDH). Parameters that measured liver
injury were a focus of the clinical examinations because a previous epidemiological study of
576 Japanese acrylic fiber manufacturing workers exposed to <5 or <20 ppm AN between 1960
and 1970 had reported an increase in subjective symptoms and mild injury to the liver in exposed
workers (Sakurai and Kusumoto, 1972). AN and thiocyanate concentrations in urine were also
measured to evaluate individual exposure levels.
Levels of AN in the air (Sakurai et al., 1978) were measured as 8-hour TWA from "spot"
samples (from stationary air samplers) and from exposed subjects wearing personal samplers.
As shown in Table 4-19, the factories and exposed workers were classified into three groups (A,
B, and C), according to their level of AN exposure. SDs for the means and ranges of the
exposure concentrations were not reported. Mean exposure durations for workers in the three
groups of factories were 10.3 years (SD = 4.5) for group A, 10.8 years (SD = 4.4) for group B,
and 12.6 years (SD = 2.1) for group C.
Table 4-19. Industrial AN exposure, levels of AN and thiocyanate in urine,
and prevalence of physical signs of adverse effects in workers exposed to AN
at six acrylic fiber factories in Japan
Factory
group
(n = number
of factories)
Mean 8-hr TWA AN
concentration (ppm)
AN in urine
Og/L)
Thiocyanate
in urine
Og/L)
Percentage of exposed vs. control workers
with physical signs of adverse effects
Spot
samples
Personal
samples
Reddening of
pharynx or
conjunctiva
Palpable
liver
Rashes or
pigmentation
of skin
A (n = 2)
2.1
(n = 116)
0.1
(n = 11)
3.9
(n = 35)
4.50
(n = 19)
19.4 (n = 31)
vs.
18.2 (n = 22)
16.1 (n = 31)
vs.
9.1 (n = 22)
9.7 (n = 31)
vs.
9.1 (n = 22)
cn
II
7.4
(n = 394)
0.5
(n = 37)
19.7
(n = 51)
5.78
(n = 58)
11.3 (n = 53)
vs.
10.0 (n = 30)
15.1 (n = 53)
vs.
10.0 (n = 30)
3.8 (n = 53)
vs.
0 (n = 30)
C (n = 1)
14.1
(n = 98)
4.2
(n = 14)
359.6
(n = 22)
11.41
(n = 14)
50.0 (n= 18)
vs.
30.0 (n= 10)
38.9 (n= 18)
vs.
30.0 (n= 10)
11.0 (n= 18)
vs.
0 (n = 10)
Control
-
-
0
(n = 22)
4.00
(n = 52)
-
-
-
Source: Sakurai et al. (1978).
Although there were some differences in mean age between the groups (38.1 years for
group C, 33.9 years for group B, and 30.5 years for group A), the age distributions of exposed
and control subjects were broadly similar (Sakurai et al., 1978). No statistically significant
differences in mean clinical chemistry parameters were found between exposed workers and
controls. Medical histories of exposed workers and controls suggested a transient AN-related
103
DRAFT - DO NOT CITE OR QUOTE
-------
increase in such symptoms as irritation of the conjunctiva and upper respiratory tract, runny
nose, and skin irritation (for example, at the scrotum). Physical examination of subjects
suggested a slight increase in the incidence of palpable liver, reddening of the conjunctiva and
pharynx, and occurrence of skin rashes (see Table 4-19). However, the incidences of these
findings among the groups did not rise to the level of statistical significance. Likewise, there
were no AN-related changes in blood pressure or neurological findings.
EPA determined 4.2 ppm (average 8-hour TWA from the high exposure group) as an
equivocal lowest-observed-adverse-effect level (LOAEL) for physical signs of eye or throat
irritation, liver enlargement, or skin irritation in male workers with average durations of 10-
12 years of occupational exposure to airborne AN. Limitations to this LOAEL are that
workplace air concentrations across the 10-12 years of exposure were not available, workers
who underwent medical examinations were not necessarily the same as those whose air and urine
were sampled, and no survey of self-reported symptoms was conducted.
A later study of Japanese acrylic fiber workers by some of the same investigators (Muto
et al., 1992) included a survey of self-reported symptoms. A strength of the LOAEL is the
concurrence between the urinary biomarkers of exposure and the measured air concentrations
(see Table 4-19).
Muto et al. (1992) performed another cross-sectional health examination of male workers
in seven Japanese acrylic fiber manufacturing plants in 1988. The seven factories included the
six factories studied in the 1976 cross-sectional health examination of Japanese acrylic fiber
workers (Sakurai et al., 1978). Exposed workers selected for the study were 157 male shift
workers with at least 5 years of experience on production lines. The mean years of exposure for
these workers was 17 ± 6.6 years. A nonexposed control group consisted of 537 male shift
workers in polyester fiber plants, power supply plants, or finishing branches in the acrylic fiber
plants. Controls were similar to exposed workers in average age (42.2 years, controls;
41.9 years, exposed), percentage who drank alcohol (76.5%, controls; 77.1%, exposed), and
percentage who smoked (58.1%, controls; 58.6%, exposed). Subjects underwent a medical
examination that documented past illnesses, work history, exposure to AN, smoking and
drinking habits, subjective symptoms, physical condition, urinalysis, hematology, liver function
blood variables (total bilirubin, AST, ALT, and y-GTP), and chest X-rays. Prevalences of
workers with abnormal values for the medical examination parameters were analyzed
statistically (protein, sugar, or urobilinogen in urine; blood specific gravity >1.055, Hb <14 g/dL,
hematocrit >40%, and RBC count <103/mm3; and bilirubin >1.0 mg/dL, AST >40 enzyme
activity units, ALT >35 units, and y-GTP >80 units). Exposure of workers (reported as 8-hour
TWA concentrations) was assessed by data from workplace area air samples and from personal
air samples collected over a 2-day period for 142 of the 157 exposed workers. Overall, the mean
AN exposure concentrations for the exposed workers were 0.53 ± 0.52 ppm (range, 0.01-
2.80 ppm) based on stationary air samples and 0.62 ± 0.90 ppm (range, 0.01-5.70 ppm) based on
104
DRAFT - DO NOT CITE OR QUOTE
-------
personal air samples. The mean AN exposure measurements were lower than the exposure
measurements made in 1976 (see Table 4-19), suggesting that workplace air concentrations in
the plants had declined over the 12-year period between the studies.
Prevalences of subjective symptoms that were statistical significantly different between
the exposed and control groups included "heaviness in stomach" and decreased libido, with the
symptom described as heaviness in stomach as most significant (Muto et al., 1992). Reported
prevalences for heaviness of stomach were 29.3% in the exposed group vs. 18.7% in the control
group. No significant differences were found in prevalences for other GI effects such as
anorexia, nausea, vomiting, heartburn, or stomachache.
When workers were grouped into those from factories with mean TWA AN
concentrations below 0.3 ppm (four factories; mean stationary and personal air concentrations of
0.27 and 0.19 ppm, respectively; Group A) or above 0.3 ppm (three factories; mean stationary
and personal air concentrations of 0.84 and 1.13 ppm, respectively; Group B), statistically
significant increased prevalences for the following symptoms (compared with controls) were
found for the workers in the factories with higher exposure levels (Group B): decreased libido
(54.9 vs. 40.2% for controls), poor memory (76.1 vs. 64.0% for controls), irritability (35.2 vs.
24. P/o for controls), reddening of conjunctiva (21.1 vs. 11.7% for controls), and eye pain or
lacrimation (32.4 vs. 19.2% for controls). There were no statistically significant differences
between the exposed and control groups in the prevalences of clinically observed physical signs
(skin rashes or reddening of conjunctiva) or abnormal findings in urinalytic, hematological, liver
function, or blood pressure variables. Prevalences of chest X-ray abnormalities were likewise
not different between exposed and control groups.
The mean 8-hour TWA personal air concentration of workers in the Group B factories,
1.13 ppm, was judged by EPA to be a LOAEL for small but statistically significant increased
prevalences of several subjectively reported symptoms (e.g., poor memory and irritability) in the
absence of statistically significant increases in the prevalences of physical signs or abnormal
values for a number of urinalytic, hematological, liver function, or blood pressure variables
(Muto et al., 1992). The average 8-hour TWA personal air concentration in the Group A
factories, 0.19 ppm, was judged to be a no-observed-adverse-effect level (NOAEL). Like the
NOAEL identified in the earlier cross-sectional health examination of Japanese acrylic fiber
workers (Sakurai et al., 1978), a limitation to the Muto et al. (1992) NOAEL is that the exposure
assessments were cross-sectional in nature. Historical measurements of air concentrations were
not available.
Kaneko and Omae (1992) performed a cross-sectional health questionnaire study of
exposed and nonexposed workers from seven acrylic fiber manufacturing plants in Japan. The
questionnaire for surveying subjective symptoms was administered to 1,220 exposed male
workers and 757 nonexposed male workers who were either from the same factory or a close-by
factory of the same company. The selected study population included 504 exposed individuals
105
DRAFT - DO NOT CITE OR QUOTE
-------
and 249 unexposed controls. Subjects who were excluded from the study were workers with
administrative or nonshift jobs, a history of exposure to other chemicals, ages not able to be
matched, or incomplete information on the questionnaire. Workplace air concentrations of AN
were measured on 2 consecutive days in each factory by using a portable gas chromatograph.
Factories were grouped into three exposure groups with the following mean workplace air
concentrations: Group L = 1.8 ppm, Group M = 7.4 ppm, and Group H = 14.1 ppm. Further
information on the exposure measurements was not reported (e.g., SD of means, ranges of
values, or whether or not the reported concentrations represented 8-hour TWA concentrations).
Mean durations of exposure in the three groups were 5.6, 7.0, and 8.6 years for the L, M, and
H groups, respectively. All subjects filled in a detailed questionnaire (232 questions in all) that
was intended to detect neurological condition and subjective symptoms.
The neurological statuses, as assessed with two different analytical methods
(Fukamachi's criteria and Cornell Medical Index profiles), of the AN-exposed workers were
slightly higher than those of control workers, although the differences were not statistically
significant. Subjective symptoms with significantly higher prevalences in the L, M, and
H groups, compared with the nonexposed groups, included headaches, tongue trouble, choking,
lump in throat, fatigue, general malaise, heavy arms, and heavy sweating. The numbers of
subjective symptoms that were significantly more prevalent in exposed workers were as follows:
8 in group L, 19 in group M, and 14 in group H. Only the prevalence of one subjective
symptom, "often feel a choking lump in the throat," had a tendency to increase with increasing
length of exposure to AN in all factories and in group L. EPA identified 1.8 ppm as an equivocal
LOAEL (Group L mean air concentration) for statistically significantly increased prevalences of
subjective symptoms.
In a translated study from China, Chen et al. (2000) examined the health effects of
occupational exposure to AN in 224 workers at an acrylic fiber plant. The exposed group
consisted of 180 males and 44 females, with an average age of 38.6 years (range 19-57 years)
and an average of 13 years of service. The average AN concentration in the work areas at the
"3
plant was reported to be 1.04 mg/m (0.48 ppm). All subjects were given a physical
examination. The results of these investigations and of subjects' hematological and clinical
chemistry parameters were compared with those of unexposed controls. Reported symptoms that
had significantly higher incidence in exposed subjects than in unexposed controls included
headache and dizziness (41 vs. 21%), poor memory (30 vs. 13%), feelings of choking in the
chest (13 vs. 8%>), and loss of appetite (13 vs. 7%). However, of the other parameters evaluated
in this study, all hematological data and all but one of the clinical chemistry parameters gave
closely similar values to those of controls. The exception was the serum activity of y-GTP that
was significantly higher (p < 0.05) in exposed subjects than in controls (44.32 ± 32.21 vs.
40.22 ± 31.06 IU/L). The study authors identified 0.48 ppm as a LOAEL for statistically
significantly increased prevalences of subjective symptoms (including headache, dizziness, poor
106
DRAFT - DO NOT CITE OR QUOTE
-------
memory, and loss of appetite) in workers employed in an acrylic fiber manufacturing plant for an
average of 13 years, without significant changes in most hematological and clinical chemistry
variables except for y-GTP activity.
Lu et al. (2005a) employed the World Health Organization (WHO)-recommended
Neurobehavioral Core Test Battery (NCTB), which includes seven components, to evaluate
neurobehavioral effects of workers exposed to AN in a Chinese plant. The subjects included
81 workers (68 males and 13 females) in the AN-monomer department, 94 workers (67 males
and 27 females) in the acrylic fibers department, and 174 workers (130 males and 44 females) in
the administrative or embroidery departments with no AN exposure. The monomer and fiber
workers represented 96% of the eligible exposed workers in the two departments. Periodic
short-term area sampling between 1997 and 1999 indicated that the geometric means of AN
exposure were 0.11 ppm (range 0.00-1.70 ppm for 390 samples) in the monomer department and
0.91 ppm (range 0.00-8.34 ppm for 570 samples) in the fiber department; no personal sampling
data were collected. As categorized by duration of employment, 23% of monomer workers were
exposed for 1-10 years, 42% for 11-20 years, and 35% for more than 20 years; and 47% of fiber
workers were exposed for 1-10 years, 23% for 11-20 years, and 30% for more than 20 years.
Mean durations of employment were not reported for the exposed groups. Monomer workers
were also potentially exposed to cyanide and fiber workers to methyl methacrylate and heat, but
levels of exposure to these possible confounders were not monitored. The exposed workers were
frequency matched on age (within 5 years) and years of education (within 1 year) with the
unexposed workers. Exposed workers (mean age 40.8 years; range 25-53) were slightly older
than unexposed workers (mean age 36.4 years; range 21-53); the percentage of females and
years of education were roughly similar across groups. All subjects were interviewed for
demographic data, general health status, and lifestyle. All tests were conducted by three
specially trained physicians using a Chinese operational guide of the NCTB.
Results of the analysis revealed that exposure to AN had adverse effects for some
components of the NCTB, indicating neuropsychological impairment; scores from the following
tests were statistically significantly different (p < 0.05) from controls in analyses of covariance
that took into account age, sex, and education level. In the Profile of Mood States test, all scores
for negative moods (anger, confusion, depression, fatigue, and tension) were significantly higher
in the exposed groups than in the unexposed group and higher for monomer workers (41-68%)
higher than controls) than for fiber workers (20-44% higher than controls). Simple Reaction
Time, a test of attention and visual response speed, was longer in the two exposed groups than in
the unexposed group: 16% longer for monomer workers and 10% longer for fiber workers.
Exposed workers performed more poorly (by 21% for monomer workers and 24% for fiber
workers) in the backward sequence of the Digit Span test, a measure of auditory memory, but
fiber workers had better performance in the forward sequence than unexposed workers. Both
groups of exposed workers also had a 4% poorer performance in the Benton Visual Retention
107
DRAFT - DO NOT CITE OR QUOTE
-------
test, a measure of visual perception and memory; scores in the Pursuit Aiming II test, which
assesses fine motor skills and perceptual speed, were 14% lower for monomer workers and 10%
lower for fiber workers compared with controls. Exposure to AN had no significant effect on
scores for manual dexterity in the Santa Ana test or for perceptual speed in the Digital Symbol
test. In examining effects by duration of AN exposure, there was no statistical relationship for
mood scores and duration of exposure. However, there was an insignificant decrease in Simple
Reaction performance with duration of exposure for both monomer and fiber workers. Inverse
relationships were found between performance and duration in both the Digital Symbol test and
total scores of the Digit Span test in the two exposed groups. Decreased performance with
duration of exposure was also found in the Pursuit Aiming II test for exposed monomer workers.
Lu et al. (2005a) mentioned several limitations of this study. One was that the cited
exposure data represented estimates of previous exposure levels and no contemporaneous
personal monitoring data were available. Furthermore, as the NCTB was developed for
populations in Europe and North America, it is not known to what extent cultural differences
may have affected results of the Profile of Mood States test, shown to be sensitive to cultural
differences. In addition, the study authors could not rule out the possibility that examiner drift
may have affected the results. Selection bias among volunteers in different groups was
mentioned as a possible confounding effect, but was discounted on the basis that the
participation among exposed workers was so high. Moreover, largest measures of
neurobehavioral effect occurred on the acrylic fiber workers, who had lower average exposure
level.
EPA determined the results from this study were consistent with the designation of the
average exposure levels for the monomer workers, 0.11 ppm, and the fiber workers, 0.91 ppm, as
LOAELs for small deficits in neurobehavioral tests of mood, attention and response speed,
auditory memory, and motor steadiness but not in tests of manual dexterity or perceptual motor
speed.
The toxicity of AN in an occupational setting has been the subject of a number of other
reports from China, which were collectively submitted to the U.S. EPA as a Toxic Substances
Control Act Test Submissions report (Acrylonitrile Group, 2000). While these studies are
discussed below for hazard identification purposes, some reports lack sufficient detail to support
reported findings of AN-related toxicity. For example, Wang et al. (2000) reported on the
disease incidence of 1,121 AN-exposed workers compared with 489 unexposed controls in a
university in the same area. Subjects in the exposed group were stated to have a higher incidence
of disease of the digestive system (16.95 vs. 8.28%) and the liver (8.12 vs. 1.84%) than in the
control group. The highest prevalence rate of 12.5% for liver disease was found in workers
exposed to AN over 20 years. The prevalence rate of chronic disease also increased with
duration of exposure. However, the report contained no diagnostic information on these
conditions and no information on the levels of AN to which the subjects had been exposed.
108
DRAFT - DO NOT CITE OR QUOTE
-------
Dong and Pan (1995) surveyed reproductive outcomes in 150 male and 105 female
workers at a chemical fiber plant in China. The concentration of AN in the air in work areas of
"3
the plant ranged from 0.53 to 15.5 mg/m (0.24-7.1 ppm; midpoint = 3.7 ppm). Controls in the
study were 110 male and 121 female workers who had not been exposed to AN. The report
compared the reproductive outcomes for the wives of male workers to the wives of male controls
(Table 4-20) and for all female workers at the plant vs. female controls (Table 4-21). As shown
in Table 4-20, the rates of premature delivery, spontaneous abortion, threatened abortion,
stillborn fetuses, and sterility in the wives of male exposed workers all were higher than those in
wives of male controls, with some differences achieving statistical significance.
Table 4-20. Comparison of reproductive outcomes in wives of exposed and
control males at a chemical fiber plant
Parameter
Exposed group" (n = 150)
Controls" (n = 110)
Average age
20.97 ±6.85b
30.04 ±3.37
Average working yrs
3.19 ± 5.35
8.53 ±3.89
Pregnancies
168
113
Sterility (%)
9 (5.00)°
2(1.82)
Life births
159b
113
Normal births
141
104
Premature deliveries (%)
18 (10.71 d)°
4 (3.54)
Late deliveries (%)
6 (3.57)
4 (3.54)
Spontaneous abortion (%)
8 (4.76)°
1 (0.88)
Threatened abortion (%)
5 (2.98)°
1 (0.88)
Stillborn fetuses (%)
4 (2.38)
0(0)
Low birth weight
5 (3.14)
4 (3.84)
aNumbers in parentheses are in %.
bGiven as 7.74 in the report.
"Significantly different from controls (p < 0.05 by Fisher's exact test), as calculated by U.S. EPA.
dSignificantly different from controls (p < 0.001 by Student's t-test), as calculated by U.S. EPA.
Source: Dong and Pan (1995).
Table 4-21. Comparison of reproductive outcomes between exposed and
control females at a chemical fiber plant
Outcome
Exposed group" (n = 105)
Control group" (n = 151)
Average age
29.36 ± 5.12b
20.68 ±4.05
Average number of working yrs
10.19 ± 8.22
10.12 ± 4.34
Number of pregnancies
112
124
Number of alive newborns
105
125
Normal births
101
113
Immature deliveries (%)
8(7.14)
5 (4.03)
Late deliveries (%)
8(7.14)
6 (4.84)
109
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-21. Comparison of reproductive outcomes between exposed and
control females at a chemical fiber plant
Outcome
Exposed group" (n = 105)
Control group" (n = 151)
Spontaneous abortions (%)
1 (0.89)
0(0)
Threatened abortions (%)
3 (2.68)
5 (4.03)
Stillborn fetuses (%)
5 (4.46)°
0(0)
Mortality of newborns (%)
3 (2.86)
0(0)
Sterility (%)
3 (2.86)
1 (0.83)
aNumbers in parentheses are in %.
bSignificantly different from controls (p < 0.001 by Student's t-test).
"Significantly different from controls (p < 0.05), as calculated by the study authors.
Source: Dong and Pan (1995).
Effects on the reproductive outcome in exposed female workers at the plant were not as
striking as in males, when compared with controls (Table 4-21) (Dong and Pan, 1995).
However, the rate of stillborn fetuses in the exposed group was far higher than that of the control
group (5/112 vs. 0/124, p < 0.05). It is unclear how much these differences may be related to the
fact that in the first comparison (Table 4-20), the exposed males were significantly younger than
the controls (p < 0.001 by Student's t-test, as calculated by the reviewers), whereas in the second
case (Table 4-21), the exposed childbearing females were significantly older than the controls
(p < 0.001). The results from this study identify the midpoint of the range of average workplace
air concentrations of AN, 3.7 ppm, as a LOAEL for increased prevalences of adverse
reproductive outcomes in male and female workers employed for averages of about 3 and
10 years, respectively. The adverse outcomes showing statistically significantly increased
prevalences compared with controls were premature deliveries (18/150 vs. 4/110 in controls),
threatened abortion (5/150 vs. 1/110 in controls), and sterility (9/150 vs. 2/110 in controls) for
the male exposed workers and their wives and increased prevalence of stillborn fetuses in female
exposed workers (4/105 vs. 0/151 in controls).
In a follow-up to the study by Dong and Pan (1995), Dong et al. (2000b) expanded the
cohort to include the years 1994 and 1995, resulting in 548 exposed male workers (496 controls)
and 391 exposed female workers (427 controls). Average employment durations for the exposed
males and females were 11.0 ± 4.5 and 10.4 ±3.8 years, respectively. The average ages of the
male and female workers were 33.8 and 32.4 years, respectively. The age, length of service, and
lifestyle of the controls were similar to the exposed workers.
"3
The reported range of AN concentrations in the workplace air, 0.11-15.5 mg/m (0.05-
7.14 ppm, midpoint = 3.6 ppm), was not markedly different from that reported in the initial
study. In the follow-up study, however, Dong et al. (2000b) also evaluated birth defects. The
general trend of results was a slightly higher percentage of abnormal findings in the exposed
cohort, as compared with the initial study, and an incidence of 15.7% birth defects, as compared
110
DRAFT - DO NOT CITE OR QUOTE
-------
with 6.1% in controls. This difference was not statistically significant. There were, however,
some statistically significant findings in the reproductive outcome of exposed females. In the
initial study, Dong and Pan (1995) had reported a significant increase in stillbirths (4.5% in
exposed vs. 0% in controls). In the expanded cohort, this value was lower, 2.7 vs. 0.68%, but
still statistically significant at thep < 0.05 level. There was also a significantly (p < 0.05)
increased incidence of birth defects (21.3 vs. 4.8%), decreased incidence of low birth weights
(3.5 vs. 4.8%>), and highly significant (p < 0.01) increase in premature deliveries (8.2 vs. 3.9%).
The results from the follow-up study identify the midpoint of the range of workplace air
concentrations, 3.6 ppm, as a LOAEL for statistically significantly increased prevalences of
adverse reproductive outcomes (increased stillbirths, birth defects, and premature deliveries) in
female exposed workers employed for an average of 11 years.
Dong et al. (2000a) also carried out an industrial hygiene survey of 68 male and
25 female workers at a Chinese chemical plant, as compared with 58 male and 38 female
workers who had not been exposed to AN. The average age of exposed subjects was 27.9 years,
and their years of employment at the plant ranged from 1 to 15 years. The average age of
controls was 32.3 years. All subjects in the study answered lifestyle questions, including
medical and occupational history, smoking and drinking history, and subjective symptoms. They
also were given a physical examination, including clinical examinations, chest X-rays, liver and
gallbladder ultrasound, electrocardiogram (ECG), clinical chemistry, and blood and urine tests.
Air samples were collected from two workplace areas over a 3-year period and analyzed for AN,
HCN, and acetonitrile concentrations. Annual average AN concentration was not provided in the
"3
report, but measured AN concentrations were over 2 mg/m at all times, with the highest
concentration reaching 22.79 mg/m .
Exposed workers reported the occurrence of headache, dizziness, sleeping disorders, and
a feeling of choking in the chest at a significantly higher prevalence than nonexposed workers.
Without presenting details, Dong et al. (2000a) referred to ultrasound tests of six individuals in
the exposed group, which appeared to show diffuse changes in the liver. In this group, there was
also a higher incidence of trembling of eyes, face, and fingers than in controls. However, all of
the blood pressure, hematological, and clinical chemistry parameters under evaluation (including
liver function tests) for exposed subjects were within the normal range and closely similar to
those of controls. Although the results indicated that increased prevalences of symptoms were
reported by exposed workers compared with nonexposed workers, the reported exposure
information was inadequate to designate a LOAEL for this study.
Li (2000) surveyed reproductive outcomes in 379 female workers in a Chinese AN
manufacturing company and in 511 unexposed control workers from a bed sheet factory and a
biological research institute in 1991. The average age and duration of employment of the (20.8
vs. 7.14%)), premature deliveries (11.62 vs. 4.72%), and congenital defects (25.4 vs. 4.2 exposed
workers were 33.95 years (22.75-54.83 years) and 14.10 years (3.25-34.45 years), respectively.
Ill
DRAFT - DO NOT CITE OR QUOTE
-------
Average age and employment duration of the control workers were reported to be similar to the
exposed group. Monthly workplace air concentration data for 1989 and 1990 were provided by
"3
the factory. The average AN air concentration was reported as 16.35 mg/m (7.5 ppm, range =
"3
0-152.88 mg/m [0-70 ppm]). Statistically significant increased prevalences of the following
reproductive outcomes were found, compared with controls: sterility (2.64 vs. 0.78%),
pregnancy complications %). Li (2000) also divided the exposed group into female workers with
and without exposed male partners. Prevalence rates for several adverse reproductive outcomes
(pregnancy complications, premature delivery, late delivery, stillbirths, and congenital deficits)
were statistically significantly elevated in females with exposed partners, compared with those
with nonexposed partners. This study identifies the average workplace AN air concentration, 7.5
ppm, as a LOAEL for increased prevalences of adverse reproductive outcomes in female AN
manufacturing workers employed for an average of 14 years.
Xiao (2000a) reported on the fasting serum activities of serum glutamate pyruvate
transaminase (SGPT, also known as ALT) in 372 workers exposed to AN for 1-31 years in a
chemical factory in China and compared the levels with those in 186 unexposed workers.
Individuals with SGPT level >19 [j,mol/L-minute were considered positive. The percentage of
individuals with SGPT activity exceeding the benchmark level was compared between the
groups. Significantly higher percentage of positive individuals were found in the exposed group
compared with controls (41.13 vs. 4.8%), with exposed males being more severely affected than
exposed females (50.23 vs. 27.82%). No data were provided on the level of AN exposure.
Xiao (2000b) also reported on the levels of whole blood cholinesterase in 237 workers
exposed to AN in a chemical factory in comparison with those in 184 unexposed workers. AN
measurements were provided for three separate workshops, although it is not evident from the
report whether the measurements were taken as spot samples or from personal samplers. A
colorimetric method was used to measure cholinesterase activity. The average AN
concentrations in air for the three workshops were 7, 3.3, and 3 ppm, respectively. The authors
reported that the levels of whole blood cholinesterase in AN-exposed workers from the three
workshops were more than 50% lower than in controls. Health examination results showed there
was also an apparent increase in the incidence of symptoms related to lowered cholinesterase
activity in exposed subjects compared with controls. These symptoms included neurological
disorder, excessive sweating, trembling, and discomfort in the chest.
In an article published in a Chinese journal, Ding et al. (2003) evaluated mitochondrial
DNA damage in a group of 47 Chinese workers randomly selected from 1,020 active workers in
the chemistry department of a petrochemical company. These workers were exposed to AN at a
3 3
geometric mean concentration of 0.25 mg/m (0.11 ppm) (median 0.36 mg/m [0.17 ppm], range
"3
0-3.70 mg/m ) for an average of 17.3 ±3.8 years. An unexposed control group of 47 persons
was selected from the teachers and staff of a college, with an average length of employment of
18.7 ± 4.1 years. DNA was extracted from peripheral blood samples from each subject and
112
DRAFT - DO NOT CITE OR QUOTE
-------
evaluated using the polymerase chain reaction (PCR) with specific primer pairs to detect
deletions in mitochondrial DNA. Deletions in mitochondrial DNA were detected in
8/47 exposed workers compared with 0/47 nonexposed workers. A deletion rate of 0.00225 ±
0.00171 was calculated based on optical densities from gel scans of the deletion fragment
compared with a fragment synthesized by using primer pairs to a conserved region of
mitochondrial DNA; this deletion rate was statistically significantly different (p < 0.05) from the
rate of 0 for the controls. For studying the effects of aging, a group of 12 healthy nonexposed
retirees from governmental organizations (average age 79.15 ± 3.80 years) and a group of
12 healthy nonexposed high school students (average age 14.23 ± 1.52 years) were also
examined. Deletion fragments were detected in 3/12 elderly subjects and 0/12 young subjects.
The deletion rate for the elderly was calculated as 0.00193 ± 0.00086, which was not
significantly different from that calculated for AN-exposed workers. The study authors
suggested that exposure to AN might have an effect on the molecular process(es) of aging.
Limitations of this study included the lack of reporting for the criteria of selection for the
controls and the gender composition of each group. The results identified the mean workplace
AN air concentration, 0.11 ppm, as a LOAEL for increased prevalence of workers with deletions
in peripheral blood mitochondrial DNA compared with controls.
Ivanescu et al. (1990) used a radioimmunoassay to measure serum levels of testosterone
in three groups of male workers exposed to AN in a chemical factory. Blood samples were taken
from 39 subjects in May 1975, from 109 subjects in March 1976, and from 149 subjects in May
1977. Subjects were between 19 and 40 years old and had been employed at the facility from
6 months to 10 years. Controls in the study consisted of 145 unexposed men. The three groups
of exposed subjects had average serum testosterone concentrations ranging from 3.5 to
4.1 ng/mL. This compared with average values ranging from 5.4 to 7.3 ng/mL in different
subsets of the 145 control subjects. Although the time of blood sampling during the day was
variable in this study and the circadian rhythm of testosterone was unknown, testosterone
concentrations in sera of exposed groups were much lower than in control groups of the same
month. However, no data were presented in the report on the level of exposure to AN or other
chemicals.
An epidemiological report by Czeizel et al. (2000, 1999) examined the incidence of
congenital abnormalities and indicators of germinal mutations in the vicinity of an AN-using
factory in Hungary. The study used the incidence of congenital abnormalities from the
Hungarian Congenital Abnormality Registry in 46,326 infants born between 1980 and 1996 in
30 settlements within a 25-km radius of an AN-using factory. A number of time-space-specific
clusters of abnormalities were identified among the subjects, including pectus excavatum in the
Tata community between 1990 and 1992 (OR = 78.5, 95% CI = 8.4-729.6), undescended testis
in Nyergesujfalu between 1980 and 1983 (OR = 8.6, CI = 1.4-54.3) and at Esztergom between
113
DRAFT - DO NOT CITE OR QUOTE
-------
1981 and 1982 (OR = 4.2, CI = 1.3-13.5), and clubfoot in Tata between 1980 and 1981 (OR =
5.5, CI= 1.5-20.3).
The relationship between the findings of clusters of abnormalities and AN exposure was
uncertain because there were no exposure data in the report. Consequently, it was not known by
how much the mothers of affected infants were exposed to the compound, if at all. Furthermore,
when the incidence of congenital abnormalities was considered for infants born throughout the
entire duration of the study, there were no increased incidences of congenital abnormalities
compared with infants born throughout Hungary.
The study authors stated that there was a technological change at the AN factory in 1984
that resulted in greater environmental protection, implying that releases of AN to the
environment were lower than before the change. This suggests that the cluster of pectus
excavatum obtained at Tata between 1990 and 1992 was unlikely to have been due to AN
exposure. However, the high incidence of undescended testis in Nyergesujfalu between 1980
and 1983 may have been an environmental phenomenon, because the region-wide incidence of
this congenital abnormality appeared to decrease with increasing distance from the factory. In
general, however, it is difficult to draw conclusions about a link between maternal exposure to
AN and the incidence of congenital abnormalities from the data in this study because of a lack of
exposure data.
Wu et al. (1995) (in the English translation of the abstract of an article in Chinese)
reported findings on 477 female workers exposed to AN and 527 controls studied using the
retrospective cohort method. Statistically significant increases in the incidence of pernicious
vomiting, anemia, preterm delivery, and birth defects were found in exposed women. However,
the authors also emphasized several confounding factors in their study, such as illness, medicine
taking, and X-ray exposure during pregnancy.
Other reports of chronic occupational exposure to AN described additional health effects.
Babanov et al. (1959) (as cited in IPCS, 1983) reported nonspecific changes in skin and pale
mucous membranes in addition to vocal cord inflammation in workers exposed to 0.6-6.0 mg/m
(0.3-3 ppm) AN in air for approximately 3 years.
Depression, lowered arterial pressure, diffuse dermographia, and increased sweating in
AN production workers was described by Ageeva (1970) (as cited in IPCS, 1983). Stamova et
al. (1976) (as cited in IPCS, 1983) reported that workers in a polyacrylic fiber plant showed
-3
increased incidence of skin diseases when exposed to air concentrations of 10-25 mg/m (5-
12 ppm) AN. An increase in other "neurasthenic" complaints and diseases was also reported.
However, the presence of other industrial chemicals and uncertainty in the length of exposure
were potential confounding factors in these responses.
Borba et al. (1996) measured CEVal-Hb adducts as a marker of AN exposure in three
groups of occupationally exposed workers in an acrylic fiber factory in Portugal. The groups
comprised 20 administrative workers who were not exposed to AN in the same plant,
114
DRAFT - DO NOT CITE OR QUOTE
-------
14 individuals employed in the continuous polymerization department, and 10 equipment
maintenance workers. Considered a measure of the biologically effective dose, CEVal values
were 8.5-70.5 pmol/g Hb in controls, 635.2-4,603.5 pmol/g Hb for continuous polymerization
workers, and 93.9-4,746 pmol/g Hb for maintenance workers. These findings pointed to the
ready formation of AN adducts with Hb in exposed workers. Unfortunately, the report lacked
critical information about the levels and duration of exposure to AN undergone by the workers of
the different groups and whether such exposure and the adduct formation that resulted were
associated with symptoms of ill health.
Borba et al. (1996) also monitored several internal exposure markers for the three groups
of workers. The induction of CYP450 species was determined by the excretion of D-glucaric
acid (an end product of the glucuronic pathway) in the urine, and formation of malondialdehyde
(MDA) (a final product of lipid peroxidation) in RBCs was determined as a surrogate measure of
oxidative stress. The study authors found no indication of AN-induced CYP450 induction, but a
significant increase was seen for the oxidative stress marker in the group of maintenance
workers. Smoking had no influence on these metrics. Borba et al. (1996) also investigated
markers for genotoxicity (viz., gene reversion activity of urine extracts, chromosomal aberrations
(CAs), and sister chromatid exchanges [SCEs]). These are discussed in the genotoxicity section
(Section 4.5.2).
4.1.3. Dermal Exposure
4.1.3.1. Acute Exposure
A case report by Vogel and Kirkendall (1984) described a 24-year-old ship's officer who
was accidentally sprayed with AN when a valve burst while he was unloading the chemical.
Because the man's face, eyes, and body were covered with AN, it is likely that he was exposed
via the oral and inhalation routes as well as to the skin and eyes. Immediate responses to
exposure included dizziness, flushing, and nausea with vomiting. During hospitalization, acute
toxicological impacts included a rapid pulse rate (100 beats/minute) and a respiratory rate of
16/minute. The subject displayed erythema and mild conjunctivitis, tachycardia, and striking
"3
hematological changes (WBC count of 26,400 cells/cm , of which 76% was polymorphonuclear
leukocytes, 10% lymphocytes, and 7% each basophiles and monocytes). Methemoglobin
(MetHb) concentration was 10.3% on admission. The patient received nitrite/thiosulfate
treatment, underwent dialysis, and, overall, showed steady recovery over his 5-day
hospitalization.
There are a number of additional case studies of the toxic effects of AN resulting from
acute exposure after accidental spillage in the workplace. These support the designation of AN
as a skin irritant. In several cases, erythemas were shown to result from direct dermal contact
with solutions of AN (Davis et al., 1973 [as cited in IPCS, 1983]; Zeller et al., 1969; Wilson et
al., 1948; Dudley and Neal, 1942). The lesions were followed by delayed blistering and burns,
115
DRAFT - DO NOT CITE OR QUOTE
-------
typically 1 or 2 days following exposure (Davis et al., 1973 [as cited in IPCS, 1983]; Zeller et al.,
1969; Babanov et al., 1959; Dudley andNeal, 1942). In some cases, clinically diagnosed
dermatitis was associated with irritation (Bakker et al., 1991; Davis et al., 1973 [as cited in IPCS,
1983]). For example, Davis et al. (1973) (as cited in IPCS, 1983) reported a wide range of
dermal effects from AN contact, including skin dermatitis, local irritation, erythema, swelling,
blistering, and burns. However, dermatitis has not always resulted from AN-induced skin
irritation (Zeller et al., 1969; Babanov et al., 1959; Wilson et al., 1948; Dudley andNeal, 1942).
When Dudley and Neal (1942) investigated the effects of AN exposures to laboratory
animals, an accident in their laboratory resulted in a case of occupational exposure. Symptoms
similar to those described later by Wilson et al. (1948) were reported for a male laboratory
worker who spilled small quantities of liquid AN on his hands. Diffuse erythema on hands and
wrists was evident after 24 hours, with subsequent blistering on the fingertips on day 3. Both
hands became slightly swollen, erythematous, itching, and painful. By day 10 after exposure, the
skin of the fingers had cracked and peeled and the skin was dry and scaly with large areas of
tender new skin.
Zeller et al. (1969) identified 137 cases of accidental exposure to nitriles when reviewing
industrial hygiene data from Germany on accidental exposures to chemical agents in an
occupational setting over a 15-year period. Of the 137 cases, 66 cases related to AN exposure, of
which 50 cases resulted from direct skin contact and the remaining 16 to inhalation or exposure
to AN vapors. All 66 cases were judged to be minor and did not require hospitalization.
However, up to 3 weeks of recuperation was required as a result of direct skin contact with the
compound. Typically, the first symptoms appeared between 5 minutes to 24 hours after initial
dermal contact with AN. For the most part, the workers complained of burning sensations of the
skin, followed by reddening of the exposed area and the formation of blisters at any point during
the first 24 hours. In one case, AN was thought to have diffused through the leather of a shoe on
which it was spilled, resulting in a delay before blisters appeared and delayed healing. There
was no indication of resorptive damages from the dermal exposure in any of the 50 cases.
In another study (Babanov et al., 1959), blistering was also observed on workers' legs
within 6-8 hours of contact with spilled AN, while a diluted (5%), heated (50°C) solution of AN
caused serious skin burns.
A fatal case of dermal contact exposure with AN was described by Lorz (1950), when
application of a delousing agent containing AN resulted in the death of a 10-year-old girl.
Following dermal application of the delousing agent to the scalp, the girl's head was wrapped in
a cloth and she went to bed. Symptoms of nausea, headaches, and dizziness were followed by
repeated vomiting and coma. Cramps and increasing cyanosis were followed by death 4 hours
after application. A similar case of fatal poisoning was reported by Grunske (1949) in which a
3-year-old girl died after reentry into a home that had been treated with an AN-containing
116
DRAFT - DO NOT CITE OR QUOTE
-------
fumigant. In both of these cases, exposure was likely to have occurred via inhalation as well as
the dermal route.
4.1.3.2. Chronic Exposure
In addition to the known skin irritation effects of acute exposures to high-concentration
liquids and vapors, various chronic dermal exposure effects were reported. There was limited
evidence that skin sensitization resulted in dermal allergies to AN. Therefore, an intrinsic
capacity of AN to act as a skin sensitizing agent was suggested.
The abstract of a Japanese language report by Hashimoto and Kobayasi (1961) discussed
the case of a chemical laboratory worker who developed skin lesions through contact with AN.
Although much of the detail remained uncertain, including period and duration of exposure and
the influence of other chemicals on the subject's condition, the lesions apparently spread across
the subject's body from the contact site, consistent with a direct contact allergic reaction.
Bakker et al. (1991) reported that 10 employees at an acrylic fiber factory complained of
skin irritation. While five of the subjects developed irritant dermatitis, the other five subjects
gave positive patch tests with AN, suggesting an allergic reaction. This finding also supported
the designation of AN as a sensitizing agent.
Prolonged dermal contact for 6 weeks to methyl methacrylate, a copolymer of AN,
resulted in an allergic skin sensitization reaction in the case of a 27-year-old man who had his
finger splinted for a torn ligament (Balda, 1975). The unpolymerized AN of the Plexidur
copolymer resulted in strong skin irritation response, with erythema and formation of scaly skin
and scattered blisters. Patch testing confirmed that the allergic sensitization was to AN and not
to a copolymer or the polymerization catalyst, benzoyl peroxide. A contributing factor may have
been the man's prior hyperhydrosis episode of the hands that led to blister formation.
Very few studies hinted at a possible desensitization or adaptive effect for prolonged
chronic exposures. Based on investigations carried out over a period from 1965 to 1971, Zotova
(1975) reported complaints of poor health, which included skin irritation, in workers at an AN
manufacturing facility. Gincheva et al. (1977) (as cited in IPCS, 1983) did not find changes in
the health status of a group of 23 men exposed to 4.2-7.2 mg/m (2-3.3 ppm) of AN for
exposure durations of 3-5 years. Details of the study were not provided.
4.1.4. Ocular Exposure
Secondary routes of exposure to AN include ocular exposure to either AN liquid or
vapor. For example, in the Wilson et al. (1948) study, subjects exposed to AN at concentrations
"3
varying from 16 to 100 ppm (35-217 mg/m ) for 20-45 minutes demonstrated irritation of all
mucous membranes, including the eyes, nose, and throat. In other studies, blepharoconjunctivitis
was reported in workers exposed to the compound (Delivanova et al., 1978). Of 302 workers
117
DRAFT - DO NOT CITE OR QUOTE
-------
examined over 2 years (138 in 1976 and 164 in 1978), 42 had severe cases of
blepharoconjunctivitis related to AN exposure.
In the case report by Vogel and Kirkendall (1984), the mucous membranes of the eyes of
the 24-year-old man had been sprayed with AN. Among other symptoms, mild conjunctivitis but
no apparent corneal clouding was observed. Eye irritations and nasal discharge also were
reported in workers exposed to relatively high levels of AN at an acrylic fiber plant (Sakurai,
2000; Sakurai et al., 1978).
A synopsis of noncancer outcomes of exposure to AN is provided in Table 4-22.
Table 4-22. Epidemiology studies of noncancer outcomes among cohorts of
workers exposed to AN
Reference
Study population
Exposure assessment
Toxic effects/outcome
Muto et al. (1992)
157 Male workers
employed in seven
Japanese acrylic fiber
plants with average of
17 yrs employment;
537 unexposed controls
Mean 8-hr TWAs:
Group A factories =
0.19 ppm; Group B
factories =1.13 ppm
Increased prevalences of subjective
symptoms in Group B factories, such as
heaviness of the stomach, decreased libido,
poor memory, irritability, conjunctival
reddening, and eye irritation; not in Group
A; no significantly increased prevalences
of physical signs or abnormal values in
urinalytic, hematological, liver function, or
blood pressure variables in Group A or B
compared with control
Sakurai et al.
(1978)
102 Male workers in six
Japanese acrylic fiber
plants with averages of
10-12 yrs of employment;
62 matched controls
Mean 8-hr TWAs:
Group A factories =
0.1 ppm; Group B
factories = 0.5 ppm;
Group C factories =
4.2 ppm
Statistically insignificant increase in the
incidence of palpable liver, reddening of
the conjunctiva and pharynx, and skin
rashes compared with controls; survey of
subjective symptoms not performed
Kaneko and
Omae (1992)
502 Male workers in
seven Japanese acrylic
fiber plants with averages
of 5.6, 7.0, and 8.6 yrs of
employment at low-,
medium-, and high-level
workplace exposure;
249 unexposed matched
controls
Group L factories =
1.8 ppm; Group M
factories = 7.4 ppm;
Group H factories =
14.1 ppm: air
concentrations reported
as "means" without
other information
Statistically significant increased
prevalences of subjective symptoms, such
as headaches, tongue trouble, choking lump
in chest, fatigue, general malaise, heavy
arms, and heavy sweating in workers in
Groups L, M, and H factories, compared
with controls
Chen et al. (2000)
224 Workers (180 males
and 44 females) in an
acrylic fiber plant with
average of 13 yrs of
employment;
224 unexposed controls
0.48 ppm
Statistically significant increased
prevalences of subjective symptoms, such
as headache, dizziness, poor memory,
choking feeling in the chest, and loss of
appetite, compared with controls; increase
in serum y-GTP but not in other clinical
chemistry or hematological variables
118
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-22. Epidemiology studies of noncancer outcomes among cohorts of
workers exposed to AN
Reference
Study population
Exposure assessment
Toxic effects/outcome
Lu et al. (2005a)
Chinese acrylic fiber
workers: 81 monomer
workers (68 male and
13 female); 94 fiber
workers (67 male and
27 female);
174 unexposed workers
Mean 8-hr TWAs:
Monomer work areas =
0.11 ppm (0-1.7 ppm);
fiber manufacture work
areas = 0.91 ppm (0-
8.34 ppm)
Small but statistically significant deficits in
tests of neurobehavior in monomer and
fiber workers compared with controls;
significant deficits in tests of mood
(increased scores for anger, confusion,
depression, fatigue, and tension), attention
and response speed, auditory memory, and
motor steadiness; not in tests of manual
dexterity or perceptual motor speed
Dong and Pan
(1995)
255 Workers in a Chinese
chemical fiber plant
(150 males and
105 females) with
averages of 3.2 and
10.2 yrs of employment;
231 unexposed controls
Workplace air
concentrations = 0.24-
7.1 ppm;
midpoint = 3.7 ppm
Statistically significant increased
prevalences of adverse reproductive
outcomes (premature delivery [10.7 vs.
3.5%, wives of exposed males], stillbirths
[4.5 vs. 0%, exposed females], and sterility
[5.0 vs. 1.8%, exposed males]) compared
with controls
Dong et al.
(2000b)
548 Male and 391 female
workers in a Chinese
chemical fiber plant with
averages of 11.0 and
10.4 yrs of employment;
496 male and 427 female
unexposed controls
Workplace air
concentrations = 0.05-
7.14 ppm; midpoint =
3.6 ppm
Statistically significantly increased
prevalences of adverse reproductive
outcomes (increased stillbirths [2.7 vs.
0%], birth defects [21.3 vs. 4.8%], and
premature deliveries [8.2 vs. 3.9%]) in
female workers compared with controls
Dong et al.
(2000a)
93 Workers at a Chinese
chemical fiber plant;
96 unexposed controls
Unclear presentation of
exposure data
Increased prevalences of subjective
symptoms (headache, dizziness, sleeping
disorders, and a feeling of choking in the
chest) compared with controls
Li (2000)
379 Female AN
manufacturing workers
employed an average of
14 yrs; 511 unexposed
controls
Average workplace air
concentration =
7.5 ppm; range = 0-
70 ppm
Statistically significant increased
prevalences of adverse reproductive
outcomes (sterility [2.6 vs. 0.8%],
pregnancy complications [20.8 vs. 7.1%],
premature deliveries [11.6 vs. 4.7%], and
congenital defects [25.4 vs. 4.2%]) in
female exposed workers compared with
controls
Xiao (2000a)
372 Workers exposed to
AN in a chemical factory;
186 unexposed controls
No data
Increase in the serum ALT activity
Xiao (2000b)
237 Workers exposed to
AN in a chemical factory;
184 unexposed controls
Workshop A = 7 ppm;
workshop B = 3.3 ppm;
workshop C = 3 ppm
Reduction in whole blood cholinesterase
activity; subjective symptoms such as
neurological disorder, sweating, trembling,
and discomfort in the chest
Ivanescu et al.
(1990)
39 Subjects (May 1975),
109 subjects (March
1976), 149 subjects (May
1977)
No data
Reduced serum testosterone
Czeizel et al.
(2000, 1999)
Case control study of
babies born with
congenital abnormalities
in the vicinity of an
AN-using factory
No data
Incidence of undescended testis possibly
related to proximity to the AN-using
facility
119 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-22. Epidemiology studies of noncancer outcomes among cohorts of
workers exposed to AN
Reference
Study population
Exposure assessment
Toxic effects/outcome
Wu et al. (1995)
477 Female AN workers;
527 workers
No data
Increased vomiting, anemia, preterm
delivery, and birth defects in pregnant
females
Babanov et al.
(1959)
Unstated number of
exposed subjects
0.3-3.0 ppm
Changes to skin and mucous membranes,
vocal cord inflammation
Ageeva (1970)
(as cited in IPCS,
1983)
Unstated number of
exposed workers
No data
Diffuse dermographia and increased
sweating
Stamova et al.
(1976) (as cited in
IPCS, 1983)
Unstated number of
exposed AN workers
5-12 ppm
Skin reactions
Delivanova et al.
(1978)
302 Exposed workers
No data
Blepharoconjunctivitis
4.2. SUBCHRONIC AND CHRONIC STUDIES AND CANCER BIOASSAYS IN
ANIMALS—ORAL AND INHALATION
AN has been demonstrated to be a multiple-site carcinogen in rat and mouse bioassays.
Chronic oral exposure to AN-induced tumors in the brain or spinal cord, Zymbal gland in the ear
canal, the forestomach, and, to a lesser degree and less consistently, in the female mammary
gland and the tongue in several oral bioassays with F344 rats (Johannsen and Levinskas, 2002b;
Bigner et al., 1986; Biodynamics, 1980b) and Sprague-Dawley rats (Johannsen and Levinskas,
2002a; Quast, 2002; Biodynamics, 1980a, b; Quast et al., 1980a). In addition, oral exposure for
up to 46 weeks led to increases in brain astrocytomas and Zymbal gland in female Sprague-
Dawley rats (Friedman and Beliles, 2002; Litton Bionetics, 1992). Lifetime inhalation
cancer bioassays with Sprague-Dawley rats found exposure-related increased incidences of brain
tumors, Zymbal gland tumors, intestinal tumors, malignant mammary gland tumors, and tongue
tumors (Dow Chemical Co., 1992a; Maltoni et al., 1988, 1977; Quast et al., 1980b). In mice, a
single gavage lifetime bioassay identified the forestomach and the Harderian gland as sites of
tumor development, but elevated incidences of brain, Zymbal gland, or mammary gland tumors
were not found (NTP, 2001). Thus, consistent target organs for the carcinogenicity of AN
include the forestomach, CNS, and Zymbal gland in rats and the forestomach and ocular
Harderian gland in mice. Exposure-related increased incidences of tongue and mammary gland
tumors have been observed, but the responses have not been observed as consistently across
studies as the responses in the forestomach, CNS, and Zymbal gland in rats.
Evidence from two studies of exposure to AN starting in early-life suggests increased
susceptibility to the carcinogenicity of AN compared with chronic exposure during adulthood
only. The evidence for carcinogenicity in these studies, a three-generation reproductive/
developmental study of AN in drinking water (Friedman and Beliles, 2002) and an inhalation
120
DRAFT - DO NOT CITE OR QUOTE
-------
bioassay (Maltoni et al., 1988), is presented in Sections 4.2.1.2.8. and 4.2.2.2, respectively. The
evidence for early-life susceptibility is discussed in Section 4.8.1.
Oral toxicity studies in rats and mice identified the forestomach as the most sensitive
target organ for the noncancer toxicity of oral exposure to AN. Hyperplasia of the forestomach
epithelium occurred at lower oral doses than doses producing other noncancer effects. Increases
in other nonneoplastic lesions were observed in the kidney, Harderian gland, and ovary.
Available data include results from several oral toxicity and cancer bioassays of chronically
exposed rats (see the first paragraph for references) and one oral toxicity and cancer bioassay in
mice (NTP, 2001).
Results from a single chronic inhalation toxicity study in rats (Quast et al., 1980b)
identified the nasal epithelium as a sensitive noncancer toxicity target of AN. Hyperplasia of
mucus-secreting cells and flattening of the respiratory epithelium developed with chronic
exposure to concentrations as low as 20 ppm. A limited number of other subchronic studies in
rats identified impaired sensory nerve conduction (Gagnaire et al., 1998) and mild developmental
effects in the presence of maternal toxicity (Saillenfait et al., 1993; Murray et al., 1978) (see
Section 4.3) as other noncancer effects associated with repeated inhalation exposure to AN.
4.2.1. Oral Exposure
4.2.1.1. Subchronic Studies
Single studies in dogs and mice are the only well-documented standard bioassays for the
subchronic toxicity of AN in experimental animals by the oral route (NTP, 2001; Quast et al.,
1975). A full report of a study in rats is not available (Humiston et al., 1975 [as cited in Quast,
2002]). Additional subchronic studies have evaluated only functional effects of AN exposure in
specific organs, such as the adrenals (Szabo et al., 1984) and the brain (Gagnaire et al., 1998).
These endpoint-specific studies are discussed in Section 4.4.1.
Humiston et al. (1975) (as cited in Quast, 2002) exposed groups of Sprague-Dawley rats
(10/sex/group) to AN in drinking water at concentrations of 0, 35, 85, 210, or 500 ppm for
90 days. Reported intakes of AN for males/females, respectively, were 0/0, 4/5, 8/10, 17/22, or
38/42 mg/kg-day. Water consumption was decreased in a dose-related manner in both sexes,
with females more affected than males. However, the lowest dose affecting males was not
identified in the summary; in females exposed at 35 ppm, 9 of 26 measured intervals showed
statistically significant decreases compared with controls. In the 210 and 500 ppm groups, there
were significant decreases in water consumption, food consumption, and BW, but no information
was provided as to the magnitude of these changes. Exposure to AN had no effect on
hematology, clinical chemistry, or histopathologic findings; urinalysis results were also
unaffected by treatment, except for an increase in specific gravity that was correlated with
increasing dose. Apparently, a number of treatment-related effects was observed in the 85 ppm
group, but neither their identity nor the magnitude of change were specified. The summary
121
DRAFT - DO NOT CITE OR QUOTE
-------
provided by Quast (2002) did not provide sufficient detail to accurately identify NOAEL or
LOAEL values from the study by Humiston et al. (1975) (as cited in Quast, 2002).
Szabo et al. (1984) evaluated the effect of subchronic oral exposure to AN on the
structure and function of the adrenal gland and intestinal tract in rats. Female Sprague-Dawley
rats (three to four per group) were exposed for 7, 21, or 60 days to AN in drinking water at
concentrations of 0, 1, 20, 100, and 500 ppm, representing approximate daily doses of 0, 0.2, 4,
20, and 100 mg/kg. Other groups of rats received equivalent doses of AN by gavage for the
same duration. Water and food intakes were monitored continuously and BWs were recorded
every fourth day. At the end of the studies, blood samples were taken to measure plasma
corticosterone and aldosterone, and a complete necropsy was carried out on all survivors. The
adrenals, thyroid, liver, and one kidney were weighed. Samples of liver, kidney, lung, brain, and
the entire adrenal and thyroid glands were measured for levels of nonprotein sulfhydryls in tissue
homogenates. Adrenals, "other" endocrine organs, stomach, duodenum, liver, kidney, lung, and
heart were evaluated for histopathology.
BWs were reduced by about 25% in rats exposed to 500 ppm in drinking water up to
60 days, but no reduction in BW was observed in rats receiving the equivalent dose of
100 mg/kg-day by gavage (Szabo et al.,1984). Water intake was reduced in rats exposed to
500 ppm in drinking water but increased at the equivalent gavage dose. Adrenal weights were
slightly lower than in controls in the 7- and 21-day exposure groups, but significant increases in
relative adrenal weights were observed after 60 days of exposure to 500 ppm in drinking water or
to >0.2 mg/kg-day by gavage. The kidneys were enlarged in rats exposed to 100 or 500 ppm AN
in drinking water for 60 days or 500 ppm for 21 days.
Histopathological lesions included adrenocortical hyperplasia in all rats exposed by
gavage at >0.2 mg/kg-day for 60 days. Enlarged kidneys and hyperplasia of the gastric mucosa
at the junction of the glandular stomach and forestomach was observed in rats exposed to 100 or
500 ppm in drinking water for 21 or 60 days. No incidence data were provided for these lesions.
Plasma corticosterone levels were significantly decreased by both 500 ppm AN in drinking water
or by gavage in the 21-day study. Even the lowest concentration of 1 ppm AN by gavage
resulted in a 50% reduction in corticosterone. Plasma aldosterone concentration was reduced,
starting from 20 ppm AN. In the 60-day study, significant suppression of plasma corticosterone
was observed after exposure to 100 or 500 ppm AN in drinking water and in all exposure levels
given by gavage.
Levels of nonprotein sulfhydryls (glutathione) were increased by 20-50% in the 7- and
21-day studies but were significantly reduced by 20-30% in the adrenals after 60 days of gavage
exposure to 4-60 mg/kg-day. Increases in nonprotein sulfhydryl concentrations in duodenal
mucosa were observed in the 100 or 500 ppm AN exposure groups in the 7- and 21-day studies
and in all doses after 60 days, but not in the control group of rats that received water by gavage.
For the drinking water study, a NOAEL of 20 ppm (4 mg/kg-day) and a LOAEL of 100 ppm
122
DRAFT - DO NOT CITE OR QUOTE
-------
(20 mg/kg-day) were identified for enlarged kidneys and increases in regional hyperplasia of the
gastric mucosa in female rats exposed for 60 days. For the gavage study, no NOAEL was
identified, but a LOAEL of 0.2 mg/kg-day was identified for adrenocortical hyperplasia in rats
exposed for 60 days.
In a separate experiment, Szabo et al. (1984) evaluated age-dependency in the sensitivity
of Sprague-Dawley rats to the action of AN on the adrenals. Groups of weanling rats (40 g) and
adult rats (190 g) were treated with 0, 0.002% (20 ppm), 0.01% (100 ppm), or 0.05% (500 ppm)
of AN in drinking water for 21 days, or were given by the corresponding amount of AN by daily
gavage for 21 days. The animals were sacrificed, and plasma and adrenals were collected as in
the previous experiments. Young, immature rats treated with AN were found to have lower (up
to 66%) levels of plasma corticosterone and aldosterone than adult rats. These differences were
statistically significant with 0.01% AN given by gavage or 0.05% AN in drinking water.
Quast et al. (1975) exposed beagles (four/sex/group) to 0, 100, 200, and 300 ppm AN
(purity >99%) in drinking water for 6 months; the reported calculated doses were approximately
0, 10, 16, and 17 mg/kg-day in males, respectively, and 0, 8, 17, and 18 mg/kg-day in females,
respectively. Dogs were evaluated daily for clinical signs and weighed weekly; food and water
consumption were calculated from the consumption by groups of dogs penned together. Routine
hematology examinations were conducted on all samples taken from all dogs 20 days before
exposure and on days 83, 130, and 179. Blood and urine samples for standard biochemical
analyses were collected from all dogs 8 days before exposure, on days 84, 135, and 176, and at
termination (day 182 for males and day 183 for females). Serum samples were collected from all
dogs on day 155 to determine total protein concentrations and evaluate the percentages of
specific Igs by electrophoresis. Ophthalmic examinations were conducted pretest and at 85, 112,
and 175 days. All dogs were subjected to gross necropsy at which time organ weights were
recorded for brain, heart, liver, kidneys, and testes; a full set of tissues from all dogs was
examined for histopathology. Samples of liver and kidney were analyzed for nonprotein free
sulfhydryl content.
The following treatment-related effects were observed in dogs exposed to AN in drinking
water for 6 months (Quast et al., 1975). No mortality was observed in the control or 100 ppm
groups, but dogs exposed at 200 ppm (2/4 males and 3/4 females) and 300 ppm (3/4 males and
2/4 females) either died prematurely or were euthanized in a moribund condition. Signs of
toxicity manifested in these dogs included reduced consumption of food and water, decreased
BW, roughened hair coat, and a nonproductive cough. These dogs subsequently exhibited
lethargy, weakness, emaciation, respiratory distress, and terminal depression. Decreased water
consumption was observed throughout the study in males at 300 ppm and sporadically in females
at >200 ppm. Food consumption was reduced in males at 300 ppm and females at >100 ppm. A
supplemental study was conducted on eight female dogs treated with 100 ppm AN for 5 weeks,
and no reduction in food or water consumption was observed.
123
DRAFT - DO NOT CITE OR QUOTE
-------
Substantial reductions in BW were observed among dogs that died (200 and 300 ppm);
group mean weights among surviving male and female dogs were not statistically significantly
different from controls. RBC and Hb counts were reduced by 19-21% in males exposed to
200 ppm for 83 days but not at later time points; females exposed at 300 ppm exhibited 22-26%
decreases in RBC counts after 83 and 130 days of exposure. No treatment-related changes in
hematology parameters were evident after 179 days of exposure. Exposure to AN had no effect
on urinalysis, clinical chemistry, ophthalmoscopic examinations, nonprotein free sulfhydryl
content in liver or kidney, or the electrophoretic behavior of serum proteins. Statistically
significant alterations in organ weight, such as increases in relative kidney weights in males
treated at 100 and 200 ppm, were not biologically significant (less than 10% change).
Histopathologic changes were observed in the esophagus (focal erosions and/or ulcerations in the
middle one-third, dilations, and thinning of the walls) and tongue (increased thickness of
epithelium lining of the dorsal surface) of male and female dogs treated at 200 and 300 ppm.
Lesions of the lung were attributed to parasitic nematode infection that was present in all dogs.
In this study, a NOAEL of 100 ppm (8-10 mg/kg-day) and a LOAEL of 200 ppm (16-17 mg/kg-
day) were identified for early mortality and histopathological lesions of the esophagus and
tongue in male and female dogs.
In a comparative study of the neurotoxicity of nitriles, Gagnaire et al. (1998)
administered AN (>99% purity) at doses of 0, 12.5, 25, or 50 mg/kg-day to male Sprague-
Dawley rats (12/group plus 10 controls) by gavage in olive oil, 5 days/week for 12 weeks. BWs
were measured weekly. After 3, 6, 9, and 12 weeks of treatment and after an 8-week recovery
period (week 20), rats were evaluated for electrophysiological parameters, with testing occurring
16 hours after dosing (or 48 hours for testing after weekends). Electrical stimulation of the tail
nerve was used to assess four neurological properties, the motor conduction velocity (MCV),
sensory conduction velocity (SCV), and amplitudes of the sensory and motor action potentials
(ASAPs and AMAPs).
One high-dose rat died during the first week of treatment, but no other mortality was
observed. High-dose rats showed reduced BW gain that became significant after the fourth week
of treatment, leading to a terminal BW 17% lower than in controls. BWs of mid-dose rats
became significantly lower than controls after the fifth week, resulting in a terminal weight
approximately 7% lower than controls (as estimated by visual inspection of the data graph); this
weight change was not considered biologically significant. Five of 11 high-dose rats exhibited
significant weakness of the hind limbs that somewhat improved during the recovery period. Rats
exposed to AN showed acute signs of behavioral abnormalities, including salivation, locomotor
hyperactivity, and fur wetting within 1 hour of dosing. Gagnaire et al. (1998) attributed these
findings to cholinomimetic effects possibly caused by AN-induced changes in muscarinic
acetylcholine receptors or by alterations in hepatic metabolism. Exposure to AN had no effect
on MCV or AMAP except for a 40% increase in AMAP observed in high-dose rats at the 9-week
124
DRAFT - DO NOT CITE OR QUOTE
-------
time point. The most consistent effect of AN was a significant reduction in SCV compared with
that in controls (Table 4-23), ranging from 7.5 to 15% between weeks 6 and 12 (p < 0.05 to
p < 0.001); this effect abated following the cessation of exposure, but a 10.6% reduction
compared with controls (p < 0.001) was observed after 8 weeks of recovery (week 20). A 25%
reduction in ASAP was also observed in high-dose rats at week 20, but no effect on this
parameter was observed during the treatment period. A NOAEL of 25 mg/kg-day and a LOAEL
of 50 mg/kg-day were identified in the study by Gagnaire et al. (1998), based on reduced BW
and neurotoxic effects (weakness of the hind limbs and reduced SCV) in male rats exposed to
AN by gavage.
Table 4-23. Effect on SCV in male Sprague-Dawley rats exposed to AN via
gavage for 12 weeks
SCV (m/s)a
Group
Exposure (weeks)
Recovery
(mg/kg)
0
3
6
9
12
20
0
34.2 ±0.7
43.2 ±0.6
45.2 ± 1.3
48.4 ± 1.1
46.6 ±0.8
53.8± 1.5
12.5
34.9 ±0.5
42.6 ±0.9
45.7 ±0.7
47.8 ±0.7
46.7 ±0.9
51.8 ± 0.9
25
34.6 ±0.8
40.0 ± 1.0
45.5 ±0.6
46.0 ±0.9
44.4 ±0.9
51.3 ± 0.7
50
35.8± 1.1
42.2 ± 1.1
41.8 ± 1.3b
44.0 ± 1.2°
39.8 ± 0.6°
48.1 ±0.7C
aValues are means ± SDs, n = 12 for treated rats and n = 10 for controls.
Statistically significant compared with controls (p < 0.05), as calculated by the authors.
"Statistically significant compared with controls (p < 0.001), as calculated by the authors.
Source: Gagnaire et al. (1998).
In a comparative study on the effects of nine compounds related to acrylamide (Barnes,
1970), AN was administered by gavage over a period of 7 weeks to six young adult albino rats of
the Porton strain as 15 daily doses of 30 mg/kg, followed by 7 doses of 50 mg/kg, and finally
13 doses of 75 mg/kg. The time-adjusted average dose administered over the course of the study
was 36 mg/kg-day. Rats were weighed weekly, and their gait and stance when walking on a
sloping nonslippery surface including an ascent up a sloping wooden board were evaluated.
Treated rats were also held by the tail in front of a sloping bar, and tested for the ability to grasp
the bar with the front paws, and then grasp it with the hind feet, a reflex typically lost early in
rats with peripheral neuropathies. No data were provided for this experiment, but the study
author reported that there was no evidence of adverse effects.
In a recent study that explored the neurobehavioral effects of AN in rats (Rongzhu et al.,
2007), male Sprague-Dawley rats (10/group) were exposed to 0, 50, or 200 ppm AN in drinking
water. The study authors estimated AN doses to be 0, 4.03, and 13.46 mg/kg-day. Three
neurobehavioral tests, including the open field test, rotarod test, and spatial water maze, were
conducted to evaluate locomotor activities, motor coordination, and learning and memory,
125
DRAFT - DO NOT CITE OR QUOTE
-------
respectively, prior to initiation of exposure and at 4, 8, and 12 weeks of exposure. Thiocyanate
levels in urine were measured in a minimum of five rats from each group at week 12 and were
reported to be 2.79, 6.10, and 25.03 mg/g creatinine, respectively.
Beginning from the sixth week of AN administration, three rats in the 50 ppm AN group
and five rats in the 200 ppm AN group showed behavioral changes. The coat appearance in all
treated rats was soiled, and the main changes were head twitching, trembling, circling,
backwards pedaling, and decreased home-cage activities. The two treatment groups also showed
less BW gain than the control group. In the open field test, there were no significant differences
in start-up latency among the exposed and unexposed groups. The 200 ppm group consistently
had higher locomotor activity than the control group, from pretreatment to 12 weeks of exposure.
There were also no changes in the number of rearing and grooming episodes. Therefore, there
were no uniform changes in exploration and locomotion. In the rotarod test, the maximal and the
total and falling latency in the 50 and 200 ppm groups were significantly decreased in a dose-
and time-dependent manner. In the spatial water maze test, rats in the 200 ppm group had
significantly increased training time and training duration, compared with the control and
50 ppm groups. However, these two parameters in the exposed groups returned to close to
control level at the end of the experiment. Rongzhu et al. (2007) suggested that this reversible
phenomena may be caused by tolerance. The study authors concluded that oral exposure to AN
induced neurobehavioral alterations. The neurochemical mechanisms need to be further
investigated. The LOAEL for neurobehavioral alterations is 4 mg/kg-day.
In a range-finding study for a 2-year bioassay (see Section 4.2.1.2), the National
Toxicology Program (NTP) (2001) treated B6C3Fi mice (10/sex/group) with 0, 5, 10, 20, 40, or
60 mg/kg-day AN (purity >99%) by gavage in water, 5 days/week for 14 weeks; adjusted for
intermittent exposure (5 days/7 days), the intakes were 0, 3.6, 7.1, 14.3, 28.6, and 42.9 mg/kg-
day. Clinical findings were recorded on day 8 and once weekly thereafter. BWs were recorded
before treatment, weekly, and at study termination. Necropsies were performed on all animals,
at which time organ weights were recorded for heart, right kidney, liver, lung, spleen, right testis,
and thymus. Hematology analysis was conducted on all mice surviving at the end of the study.
Complete histopathologic analyses were conducted on all control mice and those treated with
40 and 60 mg/kg-day; males receiving 20 mg/kg-day were also examined. At the end of the
study, 10 males/group in the groups receiving 0, 5, 10, and 20 mg/kg-day were selected for
reproductive evaluations; the left cauda, left epididymis, and left testis were weighed and sperm
samples were evaluated for sperm counts and motility. Also, 10 females/group in the groups
receiving 0, 10, 20, and 40 mg/kg-day were evaluated for vaginal cytology (estrous cycle and
stage length) in the last 12 days of the study before termination.
The following effects were observed in mice exposed by gavage to AN. Aside from one
control male at week 9, there were no deaths at exposures up to and including 20 mg/kg-day.
Mortality at the two highest doses comprised 9/10 males and 3/10 females at 40 mg/kg-day and
126
DRAFT - DO NOT CITE OR QUOTE
-------
all mice treated at 60 mg/kg-day. All deaths occurred on the first day, except for one male in the
40 mg/kg-day group. Slight (2-8%) decreases in BW in treated mice compared with controls
were not biologically significant. Survivors (seven females and one male) receiving 40 mg/kg-
day exhibited lethargy and abnormal breathing immediately after dosing "for several days" but
then appeared to develop tolerance to AN.
Sporadic, statistically significant alterations in hematological parameters included
reductions in platelet counts by 20% in males at 20 mg/kg-day, in leukocyte and lymphocyte
counts (~30%> in males at 20 mg/kg-day and ~37%> in females at 40 mg/kg-day), and in Hb and
RBC counts by 10%> in females at 40 mg/kg-day; other statistically significant changes of
doubtful biological significance in females included reductions in RBC counts by 4% in groups
treated with 5-20 mg/kg-day and in hematocrit by 6% in the 40 mg/kg-day group. Some of these
hematological effects may have been secondary to stomach ulceration observed in some mice.
Absolute and relative heart weights were increased by 20 and 30%, respectively, in males
at 20 mg/kg-day. The absolute weights of the left cauda epididymides were increased by 15% in
groups exposed at 10 and 20 mg/kg-day. However, no histopathological findings were reported
in these organs. The only treatment-related lesions were in the forestomach of females at
40 mg/kg-day: 4/7 with chronic active inflammation (associated with hyperplasia) and 5/7 with
focal epithelial hyperplasia. Two females in this group exhibited focal ulceration of the
forestomach associated with the hyperplasia. No other treatment-related histopathology was
observed. Exposure to AN produced no effects on sperm motility in males at <20 mg/kg-day.
There were no differences in vaginal cytology parameters in females at <40 mg/kg-day. A
NOAEL of 20 mg/kg-day and a LOAEL of 40 mg/kg-day were identified in the NTP (2001)
bioassay, based on hyperplastic lesions in the forestomach of female mice.
4.2.1.2. Chronic Studies
The chronic toxicity and/or carcinogenicity of AN have been evaluated in drinking water
or gavage studies in rats and a gavage study in mice (Johannsen and Levinskas, 2002a, b; Quast,
2002; Biodynamics, 1980a, b; Quast et al., 1980a). Studies by Gallagher et al. (1988), Maltoni et
al. (1988, 1977), and Bigner et al. (1986) were cancer bioassays that did not investigate
nonneoplastic lesions caused by oral exposure to AN. In addition, a three-generation
reproductive toxicity assay involving up to 46 weeks of exposure to AN (Friedman and Beliles,
2002; Litton Bionetics, 1992) provided evidence of AN carcinogenicity. These studies are
discussed below.
4.2.1.2.1. Quast (2002) and Quast et al. (1980a). Quast (2002) and Quast et al. (1980a)
conducted a 2-year toxicity and carcinogenicity study in Sprague-Dawley rats (48/sex/group)
exposed to AN (purity >99%) in drinking water at concentrations of 35, 100, or 300 ppm; groups
of 80/sex receiving untreated drinking water served as controls. Additional interim-sacrifice
127
DRAFT - DO NOT CITE OR QUOTE
-------
groups of 10/sex/dose were exposed for 1 year and analyzed under the same protocol as the main
study. The reported intakes of AN were 0, 3.4, 8.5, and 21.3 mg/kg-day for male rats and 0, 4.4,
10.8, and 25.0 mg/kg-day for female rats. Rats were observed daily for clinical signs of toxicity
and were weighed and examined monthly for palpable masses. Food and water consumptions
were determined for 30 rats/sex/group weekly for the first 3 months of the study and for 1 week
during each of the following months: 4, 5, 6, 7, 9, 11, 12, 15, 18, 21, and 24. Hematology
examinations and urinalysis were conducted on 10 rats/sex in the control and the highest dose
group on days 45, 87, 180, and 355. Additional hematology examinations were conducted on
10 rats/sex from all groups on days 544 (males) and 545 (females) and at study termination on
day 724 (males and females); an additional urinalysis was conducted on 10 rats/sex from all
groups on day 181 to evaluate a dose response for increased urine specific gravity observed
previously at the highest dose group. Clinical chemistry examinations were conducted on
10 rats/sex from the control and highest dose groups on days 46 and 356, on 10 rats/sex from all
groups on days 88, 180, and 550, and on all survivors on day 746. All rats, whether dying
prematurely, sacrificed in a moribund condition, or sacrificed on schedule, were subjected to a
gross necropsy, which included an ophthalmologic examination. Necropsies of rats on scheduled
sacrifice included organ weight determinations for brain, heart, liver, kidneys, and testes.
Complete histopathologic examinations were conducted for rats in the control and 300 ppm
groups in the 1-year interim and 2-year studies, and, based on those results, a set of 22 tissues
was examined microscopically in nearly all rats exposed at 35 and 100 ppm; tumors and other
lesions identified at gross necropsy were also examined microscopically.
Noncancer results
Exposure to AN in drinking water reduced survival, BW, and consumption of food and
water by Sprague-Dawley rats in a dose-related manner (Quast, 2002; Quast et al., 1980a). In
male rats exposed at 300 ppm, survival was significantly reduced compared with controls
(p < 0.05), beginning at 16 months, and none survived after 22 months; survival in other treated
male groups was not significantly different from controls. In female rats, exposure at 300, 100,
and 35 ppm resulted in significantly reduced survival beginning at 10, 12, and 18 months,
respectively, with no high-dose females surviving past month 22; at 24 months, survival was 25,
8.3, 2.1, and 0% for the control and low- to high-dose groups, respectively. The magnitude, time
of onset, and duration of significantly lower BWs in exposed rats compared with controls were
dose related. Only the 17-22 and 17-18% decreases observed in males and females,
respectively, at 300 ppm were biologically significant. Reductions only reached ~8-9% in the
100 ppm groups and ~6-9% in the 35 ppm groups.
The study authors mentioned that BW comparisons near the end of the study tended to be
confounded by geriatric changes, few surviving rats, and excessive tumor growth in exposed rats.
Food intake (g/rat/day) was significantly lower in 100 and 300 ppm groups compared with
128
DRAFT - DO NOT CITE OR QUOTE
-------
controls beginning during the first week of the study and only in females at 35 ppm.
Significantly lower feed intake values for males and females, respectively, compared with
controls were measured on 12/24 and 16/24 occasions at 300 ppm, 8/24 and 11/24 times at
100 ppm, and 9/24 times in females only at 35 ppm. The largest difference in high-dose rats was
a 25% reduction of feed intake in males after 6 months and a 27% reduction in females after
9 months. Feed intakes during the last weeks of the study were not significantly different among
the different groups.
Water intake was reduced during the first 10 days by 36% in males and 38% in females
exposed to 300 ppm AN and remained significantly lower for all 26 measurements throughout
the study. Significantly lower water consumption was also measured 24/26 times for males and
23/26 times for females at 100 ppm and 13/26 times for males and 21/26 times for females at
35 ppm. Male and female rats in the 100 and 300 ppm dose groups showed a lack of normal
grooming later in the study. In addition, signs of nervous system dysfunction (erratic
movements, trembling, circling, and limb weakness) were observed in the higher dose groups
rats in the absence of end-stage kidney disease or pituitary tumors. These signs of nervous
system dysfunction when they occurred in controls were correlated with end-stage kidney
disease or pituitary tumors and occurred more frequently in males than in females. The study
authors were unable to detect microscopic tumors or lesions that would correlate with the
observed clinical signs.
AN had no primary effect on hematological parameters, except for lowered RBC counts
associated with blood loss from ulcerated tumors or nutritional anemia during the later stages of
the study. The only effect of AN on urinalysis was a slight, but statistically significant increase
(1.3-3.0%)) in urine specific gravity noted for 300 ppm group males at five time points (45-
355 days) and females at all six time points (45-545 days). Significant increase in urine specific
gravity was also observed in 100 ppm male and female rats on day 181 and day 544 (females
only). The lack of significant effect in treated males on days 544 and 724 was attributed by
Quast (2002) to higher incidence of advanced chronic renal disease in controls that resulted in an
inability to concentrate urine normally. All 300 ppm male and female rats were dead by
day 724. No toxicological significant alterations were observed in clinical chemistry parameters
or in absolute or relative organ weights in the few treated rats surviving at termination.
Lesions of the forestomach were the most prominent nonneoplastic histopathologic
effects observed in Sprague-Dawley rats (Quast, 2002; Quast et al., 1980a). In the 1-year interim
sacrifice, the only treatment-related noncancer lesion was squamous cell hyperplasia of the
forestomach, which was observed in 10/10 males and 9/10 females in the 300 ppm group and
4/10 males and 7/10 females in the 100 ppm group. At 2 years, the incidence of forestomach
lesions (hyperplasia and/or hyperkeratosis of the squamous epithelium) was dose related and
significantly elevated in males at >100 ppm and in females at >35 ppm; incidences were 15/80,
15/47, 44/48, and 45/48 for males and 20/80, 23/48, 41/48, and 47/48 for females in the 0, 35,
129
DRAFT - DO NOT CITE OR QUOTE
-------
100, and 300 ppm dose groups, respectively. Minimal progressive chronic nephropathy was also
elevated in females treated at >100 ppm; incidences were 37/80, 24/48, 37/48, and 38/48 for
control to high-dose groups. In males, minimal progressive nephropathy was elevated only in
the 300 ppm group. However, significant increase in severe progressive nephropathy was
observed in males in the 35 and 100 ppm groups; incidences were 11/80, 13/47, 13/48, and 10/48
for control to high-dose groups. Gliosis of the brain, with or without perivascular cuffing, was
significantly increased in the low- and middle-dose females (8/48 and 13/48, respectively, vs.
2/80 in controls), although the incidence (5/48) did not reach statistical significance in the high-
dose females. In males, it was present in one rat exposed at 100 ppm and three rats exposed at
200 ppm. A NOAEL was not identified in this drinking water study. A LOAEL of 4.4 mg/kg-
day was identified for increases in forestomach lesions, decreased survival, and gliosis in brain in
female rats exposed to 35 ppm AN in drinking water. Table 4-24 summarizes incidence of
nonneoplastic lesions in Sprague-Dawley rats exposed to AN in drinking water for 2 years.
Table 4-24. Incidence of nonneoplastic lesions in Sprague-Dawley rats
exposed to AN in drinking water for 2 years
AN in drinking water (ppm)
0
35
100
300
Male rats
Dose (mg/kg-d)
0
3.4
8.5
21.3
Forestomach hyperplasia and/or hyperkeratosis
15/80(19%)
15/47 (32%)
44/48 (92%)a
45/48 (94%)a
Kidneys—chronic progressive nephropathy
Severity:
minimal
10/80 (12%)
4/47 (9%)
7/48(15%)
16/48 (33%)a
moderate
10/80 (12%)
5/47 (11%)
10/48 (21%)
16/48 (33%)a
severe
11/80(14%)
13/47 (28%)a
13/48 (27%)a
10/48 (21%)
Female rats
Dose (mg/kg-d)
0
4.4
10.8
25.0
Forestomach hyperplasia and/or hyperkeratosis
20/80 (25%)
23/48 (48%)a
41/48 (85%)a
47/48 (98%)a
Kidneys—chronic progressive nephropathy
Severity:
minimal
37/80 (46%)
24/48 (50%)
37/48 (77%)a
38/48 (79%)a
moderate
17/80 (21%)
13/48 (27%)
8/48 (17)
5/48 (10%)
severe
13/80(16%)
9/48 (19%)
1/48 (2%)
4/48 (8%)
Brain—gliosis and perivascular cuffing
2/80 (3%)
8/48 (17%)a
13/48 (27%)a
5/48 (10%)
Statistically significant at p < 0.05.
Source: Quast (2002).
130
DRAFT - DO NOT CITE OR QUOTE
-------
Cancer results
The drinking water study by Quast (2002) and Quast et al. (1980a) provided evidence of
the carcinogenicity of AN in male and female Sprague-Dawley rats. Tumor findings in the
1-year sacrifices included forestomach papillomas in males: 1/10 at 100 ppm and 7/10 at
300 ppm. In females, the occurrence was 5/10 at 300 ppm. Also observed after 1 year of
exposure were microscopic tumors of the CNS (brain) in 2/10 males and 4/10 females at
100 ppm and 1/10 males and 2/10 females at 300 ppm. Carcinoma of Zymbal gland and benign
and malignant tumors of the mammary gland, also observed after 1 year of exposure, were
considered by the study authors to be related to treatment.
Histopathologic examinations after 2 years of exposure revealed statistically significant
and dose-dependent increase in incidences of several types of tumors (Table 4-25). Squamous
cell papillomas or carcinomas of the forestomach were significantly elevated in males and
females exposed at >100 ppm. Carcinomas of Zymbal gland were significantly increased in
males at 300 ppm and in females at >35 ppm.
Table 4-25. Selected tumor incidences in response to AN administered to
Sprague-Dawley rats in drinking water for up to 2 years
Tissue type
Incidence of tumor formation
Male rats
Female rats
Exposure concentration (ppm)
0
35
100
300
0
35
100
300
Dose (mg/kg-d)
0
3.4
8.5
21.3
0
4.4
10.8
25.0
Astrocytomas only
1/80
8/47a
19/48
23/48a
1/80
17/48a
22/48a
24/48a
Combined astrocytomas/glial
cell proliferation
1/80
12/47a
22/48a
30/48a
1/80
20/48a
25/48a
31/48a
Tongue
(papilloma or carcinoma)
1/80
2/47
4/48
5/48a
0/80
1/48
2/48
12/48a
Forestomach
(papilloma or carcinoma)
0/80
2/47
23/48a
39/48a
1/80
1/48
12/48a
30/48a
Zymbal gland
(carcinoma or adenoma)
3/80
4/47
3/48
16/48a
1/80
5/48a
9/48a
18/48a
Small intestine (mucous
cystadenocarcinoma)
NDb
ND
ND
ND
0/80
1/48
4/4 8a
4/48a
Mammary gland (malignant)
ND
ND
ND
ND
1/80
1/48
3/48
10/48a
Mammary gland
(malignant or benign)
ND
ND
ND
ND
58/80
42/48a
42/48a
35/48
aSignificantly different from controls (p < 0.05), as calculated by the study authors.
bND = not determined.
Sources: Quast (2002); Quast et al. (1980a).
131
DRAFT - DO NOT CITE OR QUOTE
-------
An increase in papillomas or carcinomas of the tongue in males at 300 ppm was also
considered to be related to exposure. Increases were also observed in the incidences of
malignant tumors of the mammary gland. The incidence of malignant or benign mammary
tumors was significantly increased in the low- and high-dose females but decreased in the high-
dose group. Quast (2002) noted that the lower mammary tumor incidence in the 300 ppm group
was likely due to the marked early mortality in this group, despite the earlier occurrence of these
tumors. The incidence of mammary tumors increased considerably in controls during the latter
portion of the study, when few high-dose females survived.
Two diagnostic categories—focal or multifocal glial cell proliferation (suggestive of
early tumors), and focal or multifocal glial cell tumor (astrocytomas)—were used by Quast
(2002) for tumors present in the brain or spinal cord. These diagnoses were mutually exclusive
and primarily based on the size of the lesion, with glial cell proliferation a smaller-sized lesion
than astrocytoma. For enumerating the astrocytomas of the CNS, the incidence of glial
proliferation was combined with the incidence of astrocytomas and was elevated in both males
and females in all exposed groups (Table 4-25).
Distribution of astrocytomas in the various regions of nervous tissue in male and female
Sprague-Dawley rats is shown in Table 4-26. No tumors were found from 0 to 12 months in any
location in either male or female rats. As noted by Quast (2002), the sections of cerebral cortex
contained most of the tumors because of the cortex's larger size. However, astrocytomas were
also found in the cerebellum, brain stem, and spinal cord in male and female rats. Smaller-sized
lesions (glial cell proliferation) had a distribution similar to astrocytomas.
132
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-26. Histopathologic location of astrocytomas in the CNS of male
Sprague-Dawley rats administered AN in drinking water for 2 years
Age at sacrifice
Location of astrocytomas"
Male rats
Female rats
Cerebral
cortex
Cerebellum
Brain
stem
Spinal
cord
Cerebral
cortex
Cerebellum
Brain
stem
Spinal
cord
0 ppm
13-18 mos
0/23
0/23
0/23
0/22
0/11
0/11
0/10
0/11
19-24 mos
1/43
0/43
0/43
0/43
0/48
0/48
0/47
0/46
Terminal kill
0/7
0/7
0/7
0/7
1/20
0/20
0/20
0/20
35 ppm
13-18 mos
2/14
0/14
1/14
0/14
1/13
0/13
1/13
0/13
19-24 mos
5/26
1/26
1/25
0/26
5/30
2/29
6/30
0/30
Terminal kill
1/5
0/5
1/5
0/5
3/4
0/4
2/4
0/4
100 ppm
13-18 mos
3/16
1/16
2/16
1/16
5/19
2/20
1/19
1/19
19-24 mos
13/26
0/26
6/26
1/26
12/24
1/24
9/24
1/24
Terminal kill
2/5
0/5
0/5
0/5
0/1
0/1
1/1
0/1
300 ppm
13-18 mos
10/26
0/26
2/26
1/26
11/23
2/23
4/23
3/22
19-24 mos
11/18
2/18
4/18
2/18
8/11
2/11
3/11
2/11
Terminal kill
0/0
0/0
0/0
0/0
0/0
0/0
0/0
0/0
aNo astrocytomas were found in any locations from 0 to 12 mos.
Sources: Quast (2002); Quast et al. (1980a).
4.2.1.2.2. Johannsett and Levinskas (2002a) and Biodynamics (1980a): drinking water study.
Johannsen and Levinskas (2002a) and Biodynamics (1980a) carried out another 2-year drinking
water study of AN in Sprague-Dawley rats. Groups of 100 rats/sex/group were exposed to 0, 1,
or 100 ppm AN (100% purity) in drinking water; as calculated by the study authors, average
daily doses were 0, 0.09, and 8.0 mg/kg-day , respectively, for males and 0, 0.15, and
10.7 mg/kg-day, respectively, for females. Ten rats/sex/group were selected from these groups
for interim evaluations at 6, 12, and 18 months and at study term, leaving a maximum of
70/sex/group for lifetime exposure. Although the study was designed to last 24 months, because
of high mortality among high-dose rats, surviving males were necropsied after 22 months and
females at 19 months to ensure that a sufficient number of animals would be available for
clinical and pathological analyses. Rats were observed daily for morbidity and mortality and
examined and palpated weekly to detect growths. All rats received ophthalmoscopic
examinations in the period before treatment started and again at the time of necropsy.
BWs were recorded weekly for the first 14 weeks, biweekly between weeks 16 and 26,
and monthly thereafter. Food intake and water consumption were recorded over 3-day periods
133
DRAFT - DO NOT CITE OR QUOTE
-------
for about 25 rats/sex/group at the same intervals at which BWs were recorded. Ten rats/sex from
each group were selected for hematological, clinical chemical, and urinalysis examinations at 6,
12, and 18 months and at study termination (22 months for males and 19 months for females).
All rats, whether dying prematurely, sacrificed in a moribund condition, or sacrificed on
schedule, were subjected to a gross necropsy that included preservation of 40 tissues and organs,
gross lesions, and tissue masses for possible histopathological examination. At interim and
terminal necropsies, weights were recorded for selected organs (brain, pituitary, adrenal, gonads,
heart, kidney, and liver) in 10 rats/sex/group. Interim gross necropsies were performed after 6,
12, and 18 months on 10 rats/sex/group and on the remaining rats at termination. For the
scheduled interim and terminal sacrifices, complete histopathologic examinations were
conducted on 10 rats/sex from the control and 100 ppm groups; on all other rats, a limited set of
tissues was examined microscopically that included potential target organs (brain, ear canal,
spinal cord, and stomach) and any tissue masses observed at necropsy.
Noncancer results
Treatment-related noncancer effects were observed in Sprague-Dawley rats exposed to
AN for 19-22 months (Johannsen and Levinskas, 2002a; Biodynamics, 1980a). Statistically
significant increases in early deaths were observed in high-dose male and female rats after
10 months and became particularly severe during the last 5 months of the study. Reduction in
survival compared with controls was not significant in high-dose males at termination
(22 months) but was significant in high-dose female rats terminated at 19 months. Food and
water consumption in high-dose rats was lower than in controls throughout the study. Reduction
in mean BW compared with controls was observed in high-dose males and females throughout
the study (a significant 10% reduction in males and an insignificant 8% reduction in females at
term). Slight reductions in hematocrit, RBC count, and Hb values compared with controls were
observed during the study in high-dose groups but were generally not statistically significant; 3-
6% reductions in Hb values in high-dose males were statistically significant at the 6- and
18-month interim time points. Exposure to AN had no effect on clinical chemistry or urinalysis
parameters or ophthalmoscopic examinations; there were no treatment-related clinical signs
except for some neurological symptoms at the time of terminal necropsy in a few rats with
astrocytomas of the brain or spinal cord. The only significant treatment-related effects on organ
weights were lower absolute pituitary weights in high-dose males at 12 months and in females at
termination. Significant increase in mean relative kidney and liver weights observed in high-
dose males and females were considered by the study authors to reflect the reduced BWs rather
than target organ effects, since this effect did not occur at other study intervals.
Statistically significant increases in nonneoplastic lesions related to exposure to AN
included renal transitional cell hyperplasia in high-dose females at termination (Table 4-27).
Significant increase in squamous metaplasia of the uterus was observed in high-dose females at
134
DRAFT - DO NOT CITE OR QUOTE
-------
12 months compared with controls (5/10 vs. 0/10), and in low-dose females at 18 months (3/3 vs.
2/10). This effect was not significant in high-dose females at 18 months (4/10), and was not
observed at terminal sacrifice.
Table 4-27. Incidence of nonneoplastic lesions in Sprague-Dawley rats
exposed to AN in drinking water for 2 years
AN in drinking water (ppm)
0a
1
100
Male rats Dose (mg/kg-d)
0
0.09
8.0
Kidney transitional cell hyperplasia
1/15
2/18
0/12
Forestomach squamous cell hyperplasia
Incidence at termination13
14/19
24/26
14/15
Severity: mild
5
4
1
moderate
7
14°
8°
severe
2
6°
5°
Forestomach squamous cell hyperplasia
Incidence in early deaths or unscheduled sacrifices
36/46
31/37
37/42
Severity: mild
9
10
4
moderate
17
15
18°
severe
10
6
15°
Female rats Dose (mg/kg-d)
0
0.15
10.7
Kidney transitional cell hyperplasia
2/27
1/18
8/14d
Forestomach squamous cell hyperplasia
Incidence at termination13
39/48
40/57
17/17
Severity: mild
12
8
4
moderate
18
25
9
severe
9
7
4
Incidence in early deaths or unscheduled sacrifices
14/20
6/12
33/43
Severity: mild
4
1
1
moderate
7
2
21d
severe
3
3
lld
aDrinking water control.
bMales exposed for 22 mos, females for 19 mos.
"Significantly different from controls as calculated by authors (p < 0.05).
dSignificantly different from controls as calculated by authors (p < 0.01).
Sources: Johannsen and Levinskas (2002a); Biodynamics (1980a).
In addition, the incidence of squamous cell hyperplasia was increased in high-dose males
and females at terminal sacrifice and in spontaneous deaths in high-dose groups (Table 4-27).
The severity of squamous cell hyperplasia of the forestomach was significantly greater
(characterized as moderate or severe) than in controls in male rats at termination after exposure
to 1 or 100 ppm AN; a significant increase in severity in high-dose males was also observed at
12 months (not shown) but not at 18 months. Increased incidences in forestomach hyperplasia
135
DRAFT - DO NOT CITE OR QUOTE
-------
characterized as moderate or severe were also observed in high-dose rats of both sexes that died
prematurely or were sacrificed in a moribund condition. A NOAEL for nonneoplastic effects
was not identified in this study. A LOAEL of 1 ppm (0.09 mg/kg-day) was identified for
increased severity of squamous cell hyperplasia of the forestomach in male Sprague-Dawley rats
exposed to AN in drinking water for 2 years (22 months).
Cancer results
Significant increases in the cumulative incidences of neoplasms of the CNS, ear canal
(Zymbal gland), and forestomach were observed in Sprague-Dawley rats exposed to 100 ppm
AN in drinking water for 2 years (Johannsen and Levinskas, 2002a; Biodynamics, 1980a).
Statistically significant, dose-dependent increases were observed for astrocytomas of the brain in
both sexes and astrocytomas of the spinal cord in females. The astrocytomas varied from
discrete solid masses to localized areas of diffuse infiltrations of neoplastic cells. In addition,
adenomas of Zymbal gland (ear canal) in both sexes, carcinomas of Zymbal gland in both sexes,
and squamous cell papillomas of the forestomach in females were increased. Intestinal
adenocarcinomas were also found in high-dose male and female rats (see Table 4-28).
136
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-28. Selected tumor incidences in Sprague-Dawley rats exposed to AN
in drinking water for up to 2 years
Tumor incidence"
Tissue type
Male rats
Female rats
Exposure concentration (ppm)
0
1
100
0
1
100
Dose (mg/kg-d)
0
0.09
7.98
0
0.15
10.7
CNS
Brain (astrocytomas)
2/78
3/75
23/77b
0/79
1/80
32/77b
Spinal cord (astrocytomas)
ND°
ND
ND
0/76
0/79
7/7 8b
Total
2/78
3/75
23/77b
0/79
1/80
3 9/78b
Zymbal gland
Carcinomas
1/80
0/71
14/83b
0/79
0/75
7/7 8b
Adenomas
0/80
0/71
5/83
1/79
0/75
5/78
Total
1/80
0/71
19/83b
1/79
0/75
12/7 8b
Forestomach
Carcinomas
0/78
1/78
3/77
0/80
0/79
0/79
Papillomas
3/78
2/78
8/77
1/80
4/79
7/7 9b
Total
3/78
3/78
1 l/77b
1/80
4/79
7/7 9b
Intestine
Adenocarcinomas
0/40
0/34
2/41
0/78
0/79
2/70
aDenominators are calculated from the total number of animals examined (as reported in Table 4 of Johannsen and
Levinskas, 2002a) minus the animals scheduled for sacrifice at 6 and 12 mos. Thus, incidences are for animals
scheduled for the 18-mo, spontaneous deaths and terminal sacrifices (22 mos for males and 19 mos for females). For
Zymbal gland tumors in the 100 ppm male, one carcinoma and adenoma occurred in the 12-mo sacrifice; therefore,
adjustment to the denominator was only made for the 6-mo sacrifice.
bSignificantly different from controls (p < 0.05), Fisher's exact test.
°ND = no data collected.
Sources: Johannsen and Levinskas (2002a); Biodynamics (1980a).
Johannsen and Levinskas (2002a) reported that most of the tumors occurred in rats after
12 months of exposure. CNS astrocytomas were not detected at the 6- or 12-month interim
sacrifices (at which about 10 rats of each gender were sacrificed from each group). However,
brain astrocytomas and Zymbal gland carcinomas were detected in high-dose females as early as
after 6 months of exposure. One each of Zymbal gland carcinoma, adenoma, and squamous cell
carcinoma of the forestomach was detected at the 12-month interim sacrifice in high-dose males.
This finding was consistent with findings in the other drinking water bioassay with Sprague-
Dawley rats, indicating that some male and female rats exposed to 300 ppm AN developed
tumors after only 7-12 months of exposure (Quast, 2002). Mammary gland carcinomas were
detected in female rats scheduled for the terminal sacrifice but not in male or female rats
scheduled for sacrifice at 6, 12, or 18 months. Incidences of female rats (scheduled for the
137
DRAFT - DO NOT CITE OR QUOTE
-------
terminal sacrifice) with mammary gland carcinomas, however, did not show exposure-related
increases: 13/80, 4/79, and 14/79 for the control, 1 ppm, and 100 ppm female groups,
respectively.
4.2.1.2.3, Johannsett and Levinskas (2002a) and Biodynamics (1980b): gavage study. In
parallel to the chronic drinking water study in Sprague-Dawley rats, Johannsen and Levinskas
(2002a) and Biodynamics (1980b) conducted a lifetime study in which groups of Sprague-
Dawley rats (100 rats/sex/group) received 0, 0.1, or 10 mg/kg-day AN by gavage in water for
7 days/week. The doses were selected to match the daily intake levels in the drinking water
study. The schedule and protocols used in the gavage study for observations of clinical signs,
measurements of food and water consumption and BW, clinical biochemistry analyses, and gross
necropsy and histopathologic analyses were identical to those in the drinking water study.
Ingestion of 10 mg/kg-day AN resulted in increased mortality after 12 months and was
significantly higher in males at termination and in females beginning at month 14 compared with
controls. The study was terminated after 20 months of exposure because of high treatment-
related deaths, with a decreased survival of about 30 and 50% in males and females compared
with their respective controls. BWs showed a significant reduction of about 6% in high-dose
males compared with controls, beginning on week 12 and reaching 13% by termination; there
were no BW effects in exposed females.
Exposure to AN had no effect on food consumption in either sex or on water
consumption in males; high-dose females had slightly increased water consumption during the
first 12 months only. Small reductions in hematocrit, Hb, and RBC counts were observed in
high-dose males at 12 and 18 months and reached statistical significance at terminal sacrifice
(hematocrit, 15%; Hb, 19%; and RBC counts, 18%). Significant increases in absolute mean liver
weights were observed in the high-dose males and females and low-dose males at the 18-month
interval. Relative liver weights for high-dose males and females were significantly increased in
most intervals and might be reflective of lower BWs. Absolute and relative kidney weights of
high-dose males were increased significantly at 18 months and increased insignificantly in both
males and females at study term. The adrenal gland of high-dose males was the organ that
showed the most significant weight increase of 43% at termination.
Noncancer results
Nonneoplastic histopathological effects were observed in male and female Sprague-
Dawley rats exposed to 10 mg/kg-day AN by gavage for 20 months (Table 4-29). A significant
increase in epidermal inclusion cysts was observed in male and female rats, most notably in
animals dying spontaneously. Significant increases in renal transitional cell hyperplasia were
observed in high-dose females at 12-month sacrifice and in high-dose males after 18 months. In
addition, there was a significant increase in the severity of squamous cell hyperplasia of the
138
DRAFT - DO NOT CITE OR QUOTE
-------
forestomach in high-dose males and females. This effect was most noticeable among rats dying
early or at scheduled sacrifices after at least 12 months. High-dose rats showed a significant
elevation in the incidence of moderate or severe hyperplasia. In this gavage study, a NOAEL of
0.1 mg/kg-day and a LOAEL of 10 mg/kg-day were identified for increased severity of
forestomach lesions (squamous cell hyperplasia) in male and female rats.
Table 4-29. Incidence of nonneoplastic lesions in Sprague-Dawley rats
exposed to AN by gavage for 20 months
Dose (mg/kg-d)
0a
0.1
10
Male rats
Skin: Epidermal inclusion cysts (in spontaneous deaths)
2/59
4/68
8/5 6b
Kidney: Transitional cell hyperplasia (18 mos)
3/11
0/8
10/10°
Heart: Cardiomyopathy (18 mos)
3/11
No data
8/10b
Forestomach: Squamous cell hyperplasia
Incidence at terminal sacrifice
10/10
No data
8/10
Severity: mild
2
0
moderate
5
2
severe
3
6
Incidence in early deaths or unscheduled sacrifices
46/58
46/67
49/56
Severity: mild
18
11
2
moderate
19
26
13°
severe
9
10
34°
Female rats
Skin: Epidermal inclusion cysts (in spontaneous deaths)
0/10
No data
4/10b
Kidney: Transitional cell hyperplasia (12 mos)
0/10
0/1
4/10b
Heart: Cardiomyopathy (18 mos)
4/10
No data
3/10
Forestomach squamous cell hyperplasia
Incidence at terminal sacrifice
8/10
No data
9/10
Severity: mild
1
1
moderate
5
3
severe
2
5
Incidence in early deaths or unscheduled sacrifices
47/57
63/69
53/59
Severity: mild
12
12
0
moderate
23
38
14°
severe
12
13
39°
aWater vehicle control.
bSignificantly different from controls as calculated by the study authors (p < 0.05).
"Significantly different from controls as calculated by the study authors (p < 0.01).
Sources: Johannsen and Levinskas (2002a); Biodynamics (1980a).
139
DRAFT - DO NOT CITE OR QUOTE
-------
Cancer results
The carcinogenic effects of AN on Sprague-Dawley rats exposed via gavage were similar
to those in the drinking water studies (Johannsen and Levinskas, 2002a; Biodynamics, 1980b).
Significant increase in tumor incidences were observed in male and female rats dosed with
10 mg/kg-day and included carcinomas of Zymbal gland, astrocytomas of the brain, and tumors
of the squamous epithelium of the forestomach (papillomas and carcinomas in males and
papillomas in females) (Table 4-30). Adenocarcinomas of the intestine were also observed in
high-dose male rats that died or were killed after at least 12 months of exposure. Increases in
carcinomas of the mammary gland were observed in high-dose females that died prematurely.
Table 4-30. Cumulative incidence of tumors in response to AN administered
to Sprague-Dawley rats by gavage for up to 2 years
Tissue type
Tumor incidence"
Male rats (mg/kg-d)
Female rats (mg/kg-d)
0
0.1
10
0
0.1
10
Brain astrocytomas
2/80
0/79
16/77b
1/80
2/78
17/80b
Spinal cord astrocytomas
0/74
0/73
1/77
0/80
0/75
1/79
Zymbal gland papillomas
0/76
0/73
3/76
0/65
0/74
0/74
Zymbal gland carcinomas
1/76
0/73
10/76b
0/65
0/74
9/74b
Zymbal gland adenomas
0/76
1/73
5/76b
1/65
0/74
5/74
Forestomach carcinomas
0/79
0/77
18/79b
0/79
0/79
1/79
Forestomach papillomas
2/79
6/77
19/79b
2/79
4/79
14/7 9b
Intestine adenocarcinomas
0/80
0/80
6/80b
0/40
0/40
1/41
Mammary gland carcinomas
0/80
0/78
0/80
5/80
6/80
21/80b
""Denominators are calculated from the total number of animals examined (as reported in Table 4 of Johannsen and
Levinskas, 2002a) minus the animals scheduled for sacrifice at 6 and 12 mos; thus, incidences are for animals
scheduled for the 18-mo, spontaneous death, and terminal sacrifices (20 mos).
bSignificantly different from controls (p < 0.05), as calculated by the study authors.
Sources: Johannsen and Levinskas (2002a); Biodynamics (1980a).
4.2.1.2.4. Johannsen and Levinskas (2002b); Biodynamics (1980c). A 2-year drinking water
assay was conducted in F344 rats (Johannsen and Levinskas, 2002b; Biodynamics, 1980c).
Groups of rats (100/sex/group) were given AN (100% purity) in drinking water at concentrations
of 0, 1, 3, 10, 30, or 100 ppm; two groups of 100/sex served as untreated controls. The study
authors reported the equivalent average daily doses of AN as 0, 0.1, 0.3, 0.8, 2.5, and 8.4 mg/kg-
day for males and 0, 0.1, 0.4, 1.3, 3.7, and 10.9 mg/kg-day for females. Rats were observed
twice daily for overt signs of toxicity and examined and palpated weekly to detect growths.
Ophthalmoscopic examinations were carried out before testing and again at the time of necropsy.
BWs were recorded weekly for the first 14 weeks, biweekly between weeks 16 and 26, and
monthly thereafter. Food intake and water consumption were recorded over a 3-day period at the
140
DRAFT - DO NOT CITE OR QUOTE
-------
same intervals that BWs were recorded for about 25 rats/sex/group for all groups. At intervals of
6, 12, and 18 months and at study termination (26 months for males and 23 months for females),
10 rats/sex/group from each exposed and control group were sacrificed for clinical pathology and
microscopic evaluation.
At 6, 12, and 18 months and at termination, 10 rats/sex from the 100 ppm group and
5 rats/sex from each of the control groups were selected for hematology, clinical chemistry, and
urinalysis examinations. Lower dose group rats were evaluated as needed to determine possible
dose-response relationships. All rats, whether dying prematurely, sacrificed in a moribund
condition, or sacrificed on schedule, were subjected to a gross necropsy that included
preservation of 40 tissues and organs, all gross lesions, and tissue masses for possible
histopathologic examination. At interim and terminal sacrifice, weights were recorded for
selected organs (brain, pituitary, adrenal, gonads, heart, kidney, and liver) for 10 rats/sex from
treated groups and 5 rats/sex from each control group. For the interim sacrifices, complete
histopathologic examinations were conducted on 10 rats/sex from the 100 ppm group and
5 rats/sex from the two control groups. Potential target organs (e.g., brain, ear canal, spinal cord,
and stomach), gross lesions, and tissue masses were examined microscopically in all study
animals. All surviving females were terminated after 23 months and all males after 26 months.
Terminal necropsies included complete microscopic examinations for 10 rats/sex exposed to
100 ppm AN and 5/sex/group for the two control groups. Potential target organs and suspicious
lesions were examined in all animals in other dose groups.
Noncancer results
Treatment-related noncancer effects were observed in F344 rats exposed to AN in
drinking water for 2 years (Johannsen and Levinskas, 2002b; Biodynamics, 1980c). Statistically
significantly early deaths (p < 0.05), compared with controls, were observed in male and female
rats exposed to 100 ppm, beginning after 14 months of exposure. A statistically significant
decrease in survival was also observed in female rats exposed to 30 ppm but not to 10 ppm,
beginning after 18 months of exposure. Exposure to AN did not result in any overt neurological
impairments or ophthalmoscopic findings.
Statistically significant decrease in BWs (-12% lower than controls, p < 0.01) were
observed in male and female rats exposed to 100 ppm AN. A statistically significant decrease in
BWs of less than 5% was found in male rats exposed to 30 ppm AN and was not biologically
significant. Total food consumption (g/week) was reduced compared with controls in rats
exposed at 100 ppm (more prominently in female rats after week 13), but consumption on a BW
basis (g/kg) for both 100 ppm male and female rats was not significantly different from that for
controls. Total water consumption (mL/week) was significantly lower in the 100 ppm male and
female rats compared with controls. No differences in food and water consumption from
controls were found for groups exposed to <30 ppm AN. Hematological analyses revealed slight
141
DRAFT - DO NOT CITE OR QUOTE
-------
reductions in Hb, hematocrit, and RBC counts in the 100 ppm group, beginning at 12 months,
with Hb significantly lower in female rats at 12 and 18 months (13% at 18 months). Hematocrit
in 100 ppm female rats was 4% lower at 12 months. However, no statistically significant
differences from controls were noted at termination.
Serum alkaline phosphatase levels were significantly elevated by 80 and 65% (p < 0.05),
respectively, in 100 ppm females at 18 and 23 months. Results of clinical chemistry for the
30 and 10 ppm groups were not reported. However, the study authors noted that female rats in
these dose groups also showed significant elevation of this enzyme. A significant 33% increase
in SGPT was also observed in 100 ppm female rats at 18 months but not at other intervals. The
only treatment-related urinalysis finding was a slight increase in urine specific gravity in
100 ppm males at 18 and 26 months. No significant dose-related changes were observed for
absolute organ weights; elevations in relative organ weights (kidney, brain, liver, adrenal, testis,
heart, and pituitary in females) sporadically observed at 100 ppm at intervals throughout the
study were attributed by the study authors to lower BWs rather than target-organ toxicity.
Unlike Sprague-Dawley rats exposed to AN in drinking water (see above, Quast [2002]),
F344 rats did not show forestomach lesions at the 6- or 12-month sacrifices. In rats exposed for
periods >1 year, treatment-related nonneoplastic effects were observed in the forestomach
(Table 4-30) and skin. The incidence of squamous cell hyperplasia or hyperkeratosis of the
forestomach was elevated in male and female rats exposed to 3, 10, and 30 ppm but not to
100 ppm. Johannsen and Levinskas (2002b) suggested that, because the lesions were observed
more frequently in late surviving animals, the reduced survival in 100 ppm group may have been
the reason that no significant increase was observed for that dose group. Because these
forestomach lesions were not observed in rats examined during the 6- or 12-month sacrifices, no
such incidence data are listed in Table 4-31. An increase in epidermal inclusion cysts of the
skin, observed in 4/50 male rats treated at 100 ppm but in no other group, was also considered by
the study authors to be treatment related. A NOAEL of 1 ppm (0.1 mg/kg-day) and a LOAEL of
3 ppm (0.3 and 0.4 mg/kg-day for males and females, respectively) were identified for increases
in squamous cell lesions (hyperplasia and/or hyperkeratosis) of the forestomach in male and
female F344 rats exposed to AN in drinking water for 2 years.
142
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-31. Incidences of nontumorous lesions in F344 rats exposed to AN
in drinking water for 2 years
AN in drinking water (ppm)
0a
1
3
10
30
100
Male rats Dose (mg/kg-d)
0
0.1
0.3
0.8
2.5
8.4
Forestomach squamous cell hyperplasia or
hyperkeratosis'3
11/159
3/80
18/75°
13/80d
17/80°
9/77
Foci of cellular alteration in liver
7/94
5/31
11/51
8/45
14/50°
6/68
Epidermal inclusion cysts
0/61
0/12
0/12
0/11
0/17
4/50d
Chronic nephropathy, bilateral
118/149
57/58d
62/64°
52/54d
57/60d
42/62
Atrophy of seminiferous tubules, bilateral
150/172
72/75d
68/84
64/74
78/78°
64/87
Female rats Dose (mg/kg-d)
0
0.1
0.4
1.3
3.7
10.9
Forestomach squamous cell hyperplasia or
hyperkeratosis'3
4/156
2/80
16/80°
23/74°
13/80°
5/74
Foci of cellular alteration in liver
1/84
2/20
4/3 9d
3/49
0/40
1/70
Chronic nephropathy, bilateral
52/68
10/19
21/26
20/29
23/26
18/50
aDrinking water control.
bMales exposed for 18-26 mos, females for 18-23+ mos; excludes rats sacrificed at 6 or 12 mos. These incidences
were further adjusted to exclude (from the denominators) rats that died between 0 and 12 mos in the study. Rats
dying during this time period were determined from page 6 of Appendix H and Table 1 in Biodynamics (1980c)
and Table 8 in Johannsen and Levinskas (2002b). Unscheduled deaths between 0 and 12 mos in the study occurred
in two female controls, two males at 3 ppm, three females at 10 ppm, and three males and three females at 100 ppm.
"Significantly different from vehicle control as calculated by study authors (p < 0.01).
dSignificantly different from vehicle control as calculated by study authors (p < 0.05).
Sources: Johannsen and Levinskas (2002a); Biodynamics (1980a).
Other microscopic findings included increase in chronic nephropathy and atrophy of
seminiferous tubules, which were significant at 1 ppm. However, the study authors did not
consider these effects to be related to treatment, but rather to be due to the sample selection
procedure that only tissues showing questionable or suspicious lesions were selected for
microscopic examination for the intermediate dose groups (1-30 ppm), resulting in the selection
of tissues from more aged animals.
Cancer results
Increases in tumor incidences (Table 4-32) were observed in multiple organs, following
chronic exposure of F344 rats to AN in drinking water (Johannsen and Levinskas, 2002b;
Biodynamics, 1980c). No tumors were detected in rats sacrificed after only 6 months of
exposure. One female rat exposed to 100 ppm for 12 months had an adenocarcinoma of the
mammary gland that the study authors considered to be possibly related to treatment. Treatment-
related tumors, appearing at the 18-month interim sacrifice, included brain astrocytomas
(one/sex/group) in male and female rats exposed to 30 and 100 ppm, squamous cell papillomas
of the ear canal (Zymbal gland) in 2/9 males at 100 ppm and 1/9 females at 30 ppm, and
squamous cell carcinomas of the latter organ in one female at 100 ppm. AN-related neoplastic
143
DRAFT - DO NOT CITE OR QUOTE
-------
lesions were observed at terminal sacrifice in the brain, spinal cord, forestomach, ear canal
(Zymbal gland), and mammary gland (females only). Significant dose-related elevations in
astrocytomas were observed in the brains of male and female rats exposed to 30 and 100 ppm
and in the spinal cord of male rats exposed to 100 ppm (see Table 4-32). Dose-related increases
in squamous cell adenomas/carcinomas of the ear canal (Zymbal gland) were observed in males
exposed to 30 and 100 ppm and in females at 10-100 ppm. Increased incidences of squamous
cell papillomas or carcinomas of the forestomach, observed in males exposed to 3 and 10 ppm
and in both sexes exposed to 30 ppm, were considered by the study authors to be likely treatment
related despite the lack of a dose response. Fibroadenomas of the mammary gland were slightly
increased in females treated with 10 and 30 ppm. Other lesions considered by the study authors
to be possibly treatment related included a single case of squamous cell carcinoma of the salivary
gland in one male at 100 ppm and squamous cell papillomas of the tongue in female rats
(one/group) exposed to 10 and 30 ppm.
144
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-32. Selected tumor incidences in F344 rats exposed to AN in
drinking water for 2 years
AN in drinking water (ppm)
0a
1
3
10
30
100
Male ratsb Dose (mg/kg-d)
0
0.1
0.3
0.8
2.5
8.4
Brain astrocytoma Total0
2/200°
2/100
1/100
2/100
10/99e
21/99e
Adjustedd
2/160d
2/80
1/78
2/80
10/7 9e
21/76e
Spinal cord astrocytoma
1/196
0/99
0/92
0/98
0/99
4/93f
1/156
0/79
0/70
0/78
0/79
4/70f
Ear canal (Zymbal gland) squamous cell
2/189
1/97
0/93
2/88
7/94e
16/93e
papilloma/adenoma and carcinoma
1/147
1/76
0/73
0/67
2/7 le
14/68e
Forestomach squamous cell papilloma/carcinoma
0/199
1/100
4/97f
4/100f
4/100f
1/101
0/159
1/80
4/7 8f
3/80f
4/80f
1/77
Mammary gland fibroadenoma
2/48
1/4
2/7
2/12
0/7
2/45
Mammary gland carcinoma
1/48
0/4
0/7
0/12
0/7
0/45
Female ratsb Dose (mg/kg-d)
0
0.1
0.4
1.3
3.7
10.9
Brain astrocytoma
1/199
1/100
2/101
4/95
6/100e
23/98e
1/157
1/80
2/80
4/75
6/80e
23/76e
Spinal cord astrocytoma
0/197
0/97
0/99
1/92
0/96
1/91
0/155
0/78
0/79
1/72
0/77
1/69
Ear canal (Zymbal gland) squamous cell
0/193
0/94
2/92
4/90e
5/94e
10/86e
papilloma/adenoma and carcinoma
0/157
0/73
0/73
0/70e
2/7 3e
8/62e
Forestomach squamous cell papilloma/carcinoma
1/199
1/100
2/100
2/97
4/100f
2/97
1/157
1/80
2/79
2/77
4/80f
2/75
Mammary gland fibroadenoma
12/65
5/14
6/14
9/16e
10/22f
9/49
12/156
5/80
6/80
8/79
9/80
9/73
Mammary gland carcinoma
1/65
2/14
0/14
0/16
3/22
2/49
3/156
4/80
0/80
1/78
3/80
6/73
aDrinking water control.
bMales exposed for 26 mos, females for 23+ mos.
°Total cumulative number of rats with lesion.
dThe denominators for incidences excluded rats from the 6- and 12-mo sacrifices and rats that died before the
appearance of the first tumor for each of three tumor sites: CNS, Zymbal gland, and forestomach. The termination
history reports in Appendix C and the individual animal histopathology reports in Appendix H of the Biodynamics
(1980b) report were examined to determine time of death and tumor occurrence for each of the F344 rats. Times of
first detection of tumors were 419 d for forestomach tumors, 495 d for CNS tumors, and 475 d for Zymbal gland
tumors. Due to the limited number of mammary glands examined in most groups, the adjusted denominators
represented the number of animals that were exposed for more than 12 mos for each group; mammary gland tumor
incidences are for animals scheduled for the 18-mo and terminal sacrifices.
"Significantly different from vehicle control as calculated by study authors (cumulative only) (p < 0.01).
Significantly different from vehicle control as calculated by study authors (cumulative only) (p < 0.05).
Sources: Johannsen and Levinskas (2002a); Biodynamics (1980a).
4.2.1.2.5. NTP (2001). NTP (2001) evaluated the toxicity and carcinogenicity of AN (>99%
purity) given by gavage in water to B6C3Fi mice (50/sex/dose) at doses of 0, 2.5, 10, and
20 mg/kg-day, 5 days/week for 2 years. Adjusted for discontinuous exposure (5 days/7 days),
the intakes were 0, 1.8, 7.1, and 14.3 mg/kg-day. The doses were chosen based on results of the
145
DRAFT - DO NOT CITE OR QUOTE
-------
14-week assay described in Section 4.2.1.1. Mice were observed twice daily for clinical signs
that were recorded on day 29, every 4 weeks, and at study termination. BWs were recorded
before the start of the exposure period, then every 4 weeks, and at study termination. Five male
and five female mice were evaluated at 2 weeks and 3, 12, and 18 months for urinalysis
parameters. All mice were subjected to gross necropsies and complete histopathologic
examinations.
Noncancer results
Survival was significantly reduced in male and female mice treated with 20 mg/kg-day;
38/50, 42/50, 39/50, and 14/50 males and 39/50, 32/50, 39/50, and 23/50 females survived to
104-105 weeks with increasing dose. Despite a tendency for BWs in high-dose mice to be
slightly lower compared with controls after 30 weeks of treatment, AN had no significant effect
on terminal BWs; no treatment-related clinical signs were observed.
Exposure to AN significantly increased the incidences of nonneoplastic lesions in male
and female mice compared with control mice (Table 4-33). Statistically significant increases in
the incidences of mild focal or multifocal epithelial hyperplasia of the forestomach were
observed in males treated with 10 or 20 mg/kg-day and in focal or multifocal epithelial
hyperplasia of the forestomach in females treated with 20 mg/kg-day. The hyperplastic lesions
were often accompanied by focal hyperkeratosis and occasionally associated with chronic
inflammation. The incidence of hyperkeratosis (diffuse or focal) of the forestomach was
statistically significantly elevated in males treated at 20 mg/kg-day. AN-treated males showed a
higher incidence of hyperplasia of the Harderian gland, but only the increase observed at
10 mg/kg-day was statistically significantly different from the control. No significant increase
was found in treated females (Table 4-33). Statistically significant elevations were observed in
the incidences for ovarian cysts in females treated with 2.5-20 mg/kg-day and for ovarian
atrophy in females treated with 10-20 mg/kg-day; for both ovarian lesions, the highest
incidences were observed at 10 mg/kg-day. A NOAEL was not identified for noncancer effects
in this gavage study. A LOAEL of 1.8 mg/kg-day (adjusted for continuous exposure) was
identified for increased incidence of ovarian cysts in female mice. A NOAEL of 1.8 mg/kg-day
and a LOAEL of 7.1 mg/kg-day were identified for increased incidences of forestomach
hyperplasia in male mice.
146
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-33. Incidence and severity of nonneoplastic lesions in B6C3Fi mice
exposed by gavage to AN for 2 years
Adjusted dose (mg/kg-d)a
0b
1.8
7.1
14.3
Males
Number of male mice examined
50
50
50
50
Forestomach hyperkeratosis, diffuse/focal
2° (2.5)d
3 (2.0)
7(1.7)
12f( 1.8)
Forestomach epithelial hyperplasia, focal
2 (3.0)
4 (2.3)
8e (2.0)
9f(1.9)
Harderian gland hyperplasia
1 (2.0)
4 (2.3)
7e(3.4)
4 (2.3)
Females
Number of female mice examined
50
50
50
50
Forestomach hyperkeratosis, diffuse/focal
2(1.5)
1 (2.0)
2 (2.0)
4 (2.0)
Forestomach epithelial hyperplasia, focal, or multifocal
2(1.5)
2 (3.0)
5(1.8)
7e(1.6)
Harderian gland hyperplasia
5 (3.0)
4 (3.3)
6 (2.2)
8 (3.5)
Ovarian atrophy
6 (3.0)
8 (3.9)
45f (4.0)
40f (4.0)
Ovarian cyst
12 (2.3)
20e (2.3)
27f (2.1)
19e (2.1)
aDoses administered 5 d/wk.
bVehicle control.
°Number of mice with lesion.
dAverage severity grade of lesions in affected mice: 1 = minimal, 2 = mild, 3 = moderate, 4 = marked.
"Significantly different from vehicle control as calculated by authors (p < 0.05).
Significantly different from vehicle control as calculated by authors (p < 0.01).
Source: NTP(2001).
Cancer results
In mice exposed to AN, treatment-related carcinogenic effects (Table 4-34) were
observed in the forestomach, Harderian gland, and, less consistently, lung and ovary of treated
female mice (NTP, 2001). Histopathologic examination revealed significantly increased
incidences of forestomach squamous cell papillomas and in overall incidence of papillomas or
carcinomas in male and female mice treated with 10 or 20 mg/kg-day. Forestomach squamous
cell carcinomas alone were significantly increased in males at >10 mg/kg-day and in females at
20 mg/kg-day. Incidences of Harderian gland adenoma and adenomas or carcinomaswere
significantly elevated in males at >2.5 mg/kg-day and in females at >10 mg/kg-day. The overall
incidence of alveolar/bronchiolar adenomas or carcinomas was significantly elevated in females
treated at 10 mg/kg-day but not at the highest dose. Increase in benign or malignant granulosa
cell tumor of the ovary was observed in females treated with 10 mg/kg-day AN. However, the
increase was not statistically significant. Significant positive trends (p < 0.001) were observed in
both male and female mice for the incidences of forestomach squamous cell papilloma or
carcinoma and Harderian gland adenoma or carcinoma; there was also a significant positive
trend (p = 0.029) for alveolar/bronchiolar adenoma or carcinoma in treated females. An inverse
relationship between tumor latency and dose was observed for forestomach squamous cell
papillomas or carcinomas in females and for Harderian gland adenomas or carcinomas in males.
147
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-34. Incidences of selected neoplastic lesions in B6C3Fi mice exposed
by gavage to AN for 2 years
Adjusted Dose (mg/kg-d)a
Controlb
1.8
7.1
14.3
Malesc
Number of mice examined
50
50
50
50
Forestomach
Squamous cell papilloma (includes multiple)
3
4
19d
25d
Squamous cell carcinoma
0
0
8d
9d
Squamous cell papilloma or carcinoma
3
4
26e
32e
Harderian gland
Adenoma (includes bilateral)
5
16d
24d
27d
Carcinoma
1
1
4
3
Adenoma or carcinoma
6
16f
IT
30e
Femalesc
Number of mice examined
50
50
50
50
Forestomach
Squamous cell papilloma (includes multiple)
3
6
24d
19d
Squamous cell carcinoma
0
1
1
lld
Squamous cell papilloma or carcinoma
3
7
25e
29e
Harderian gland
Adenoma (includes bilateral)
10
10
25d
23d
Carcinoma
1
0
3
2
Adenoma or carcinoma
11
10
26e
25e
Lung
Alveolar/bronchiolar adenoma or carcinoma
6
6
14s
9
Ovary
Benign or malignant granulosa cell tumor
0
0
4
1
aDoses administered 5 d/wk.
bVehicle control.
°Number of mice with tumor.
dSignificantly different from vehicle control as calculated by authors (p < 0.01).
"Significantly different from vehicle control as calculated by authors (p < 0.001).
Significantly different from vehicle control as calculated by authors (p = 0.014).
8Significantly different from vehicle control as calculated by authors (p = 0.039).
Source: NTP(2001).
4.2.1.2.6. Gallagher et al. (1988). In a cancer bioassay, groups of 20 male Sprague-Dawley-
derived CD rats were exposed to 0, 20, 100, or 500 ppm AN in drinking water for 2 years
(Gallagher et al.,1988). As calculated from the reported AN concentration and average daily
drinking water consumption data, the intakes of AN were 0, 1.5, 7.1, and 28 mg/kg-day. Rats
were weighed weekly, and feed intake and water consumption were measured for 1 week each
month. All rats, whether dying prematurely or sacrificed at termination, were subjected to a
148
DRAFT - DO NOT CITE OR QUOTE
-------
complete gross necropsy. Tissues exhibiting gross lesions and a selection of organs (liver,
stomach, adrenal, kidney, heart, brain, pituitary, and lung) were examined for histopathology.
Significantly reduced survival compared with controls was observed in the high-dose
group after 1 year of exposure but not in the other treatment groups. The last high-dose rat died
shortly before the scheduled termination. BWs were significantly reduced in high-dose rats after
8 months and in mid-dose rats after 16 months; compared with terminal BWs of the control
group, reductions of >50% at 500 ppm and >20% at 100 ppm were estimated from the data
graph. Exposure to AN had no effect on food consumption, but water consumption was reduced
in a dose-related fashion (by 7.5, 11.3, and 30% in the low- to high-dose groups). A NOAEL of
1.5 mg/kg-day and a LOAEL of 7.1 mg/kg-day were identified for reduced BW in rats exposed
to AN for 2 years. This study was limited as to its usefulness for noncancer risk assessment
because few tissues were evaluated for histopathology and noncancer histopathology was not
reported.
Gallagher et al. (1988) reported dose-related increases for the incidences of tumors of the
forestomach and ear canal (Zymbal gland). Four of 20 high-dose rats (20%) had papillomas of
the squamous epithelium of the forestomach, a tumor type not observed in the other groups.
Squamous carcinomas of Zymbal gland were observed in 1/20 (5%) mid-dose rats and
9/20 (45%) high-dose rats but not in low-dose rats or controls. This study provided evidence of
the carcinogenicity of AN to the forestomach and Zymbal gland in male rats. Limitations of the
study in evaluation of AN carcinogenicity included small group sizes, lack of testing in females,
and only limited tissues were evaluated in gross and histopathological examination.
4.2.1.2.7. Bigner et al (1986). In an interim (18-month) report of a 2-year drinking water study
(Bigner et al., 1986), F344 rats were exposed to AN at targeted concentrations of 0 ppm
(51 males and 49 females), 100 ppm (50 rats/sex), and 500 ppm (two subgroups, with a total of
197 males and 203 females). One group of the 500 ppm rats, 147 males and 153 females, was
allocated for studies of morphology, tumor biology, and karyotyping, leaving the remaining rats
for studies of comparative survival and tumor incidences. By using reference average BWs for
F344 rats in a chronic-duration study and an allometric equation deriving drinking water
consumption from BW (U.S. EPA, 1988), the average daily doses of AN were calculated as 0,
12.9, and 64.5 mg/kg-day for males and 0, 14.8, and 74.2 mg/kg-day for females. Rats were
observed twice daily for neurological signs and death and were weighed weekly. Complete
gross necropsies were conducted on all rats. Brains were evaluated for tumors using light and
electron microscopy. The only endpoints related to noncancer effects reported for this study
were mortality, BW, and clinical signs. Incomplete quantitative data were provided for
mortality, which was reported to show dose-response effect in both sexes, occurring earlier in
100 and 500 ppm females than in males. A total of 215 high-dose rats died between month 6 and
18, whereas only "a few" male and female controls were reported to have died by month 18. No
149
DRAFT - DO NOT CITE OR QUOTE
-------
quantitative results were provided for the magnitude or statistical significance of BW reductions
that affected high-dose male rats by the third week and high-dose females "slightly" later. BW
reductions in the mid-dose group occurred in males after 2 months and females sometime during
the second year. The incidences of neurological clinical signs (paralysis, head tilt, circling, and
seizures) were dose related, affecting a total of 45/400 rats treated at 500 ppm, 4/100 rats at
100 ppm, and 0/100 controls exposed for 12-18 months. The reported data were insufficient to
accurately identify a NOAEL or LOAEL for noncancer effects in this study. A final report of
this study was not located.
In 215 rats in the 500 ppm group that died between months 6 and 18, the types of tumors
frequently found included s.c. papillomas, papillomas of the forestomach, and tumors in Zymbal
gland, but no incidence data were provided for these lesions (Bigner et al., 1986). A statistically
significant increase in tumors of the brain bearing similarity to astrocytomas was observed in rats
exposed to 500 ppm AN, with 49 primary brain tumors observed among the 215 rats dying
between months 6 and 18; incidences in other treatment groups were not reported. The tumors
were observed mostly in the cerebral cortex (about 75%) and also in the brain stem and
cerebellum. When the brain tumors were classified according to size, 10/49 of these tumors
were larger than 5 mm, 28/49 were between 1 and 5 mm in diameter, and 11/49 were detected
only microscopically. Although this study provided support for the carcinogenic effect of AN at
multiple sites in rats, the lack of numerical results rendered it inadequate for the purpose of
quantifying cancer risk.
4.2.1.2.8. Maltoni et al. (1988,1977). Maltoni et al. (1988, 1977) reported a cancer bioassay
(designated BT203) in which groups of Sprague-Dawley rats (40/sex) received 5 mg/kg AN by
gavage in olive oil, 3 days/week for 52 weeks; a control group of 75/sex received olive oil alone
on the same schedule. After the exposure period, rats were maintained without further treatment
for the rest of their natural lives (the study ending on week 131). Rats were examined 3 times
daily for their general health status and were subjected to a clinical examination for gross
changes every 2 weeks. Rats were weighed every 2 weeks during the first year and monthly
thereafter. All rats were examined by gross necropsy. All tissues with gross lesions and a
limited set of 12 tissues/organs from each rat were examined microscopically. No statistically
significant increases in tumors were observed in treated rats in this study. However, slight
decrease in tumor latency or increase incidence were observed for some types of tumors
identified in other studies on AN. For example, a carcinoma of Zymbal gland was noted in
1/40 exposed males (latency 84 weeks) and 1/75 control females (latency 90 weeks). Papillomas
and acanthomas of the forestomach were observed in 1/40 exposed males (latency 92 weeks),
4/40 exposed females (average latency 97.5 weeks), and 1/75 control females (latency
54 weeks). Maltoni et al. (1977) recognized the increase in forestomach tumors as treatment
related. Gliomas (a category that includes astrocytomas) appeared in 1/40 treated females
150
DRAFT - DO NOT CITE OR QUOTE
-------
(latency of 33 weeks), 2/75 control females (average latency 104 weeks), and 1/74 control males
(latency 98 weeks). The major limitation of this study as a cancer bioassay was the relatively
short exposure period and the small number of animals in each group. The dose was also low for
a single exposure group study.
4.2.1.2.9. Friedman andBeliles (2002); Litton Bionetics (1992). In a three-generation
reproductive toxicity study in Sprague-Dawley rats, Friedman and Beliles (2002) and Litton
Bionetics (1992) provided further evidence of the possible carcinogenicity of AN (methods and
reproductive/developmental findings of this study are presented in Section 4.3.2). In this study,
15 males and 30 females per dose level (F0 parents) were exposed to 0, 100, and 500 ppm AN in
drinking water for 100 days prior to mating for 6 days. As calculated by the study authors, the
concentrations were equivalent to doses of 0, 11, and 37 mg/kg-day for males and 0, 20, and
40 mg/kg-day for females. A subset of F0, Fl, and F2 females underwent two cycles each of
breeding, then were held for a further 20 weeks after weaning of the second litter prior to
termination.
Considering each generation separately, Fib female breeders in the 500 ppm group
showed statistically significant increased incidences of brain and Zymbal gland tumors, while
increased incidences of tumors in F0 or F2b female breeders in the 100 or 500 ppm groups
compared with controls were consistent with the Fib responses, but not statistically significant
(Table 4-35). These were relatively small groups, however, with low power to detect responses
as high as 10% statistically significant. Across all of the generations, there was a statistically
significant increasing trend in both tumor types, supporting the conclusion that exposure to AN
for up to 51 weeks at 100 or 500 ppm in drinking water was associated with increased tumor
incidence in female Sprague-Dawley rats.
151
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-35. Incidence of tumors in female Sprague-Dawley rats exposed to
AN in drinking water for up to 46 weeks
Generation
Tumor incidence in rats
Brain astrocytomas
Zymbal gland
Exposure concentration (ppm)
0
100
500
0
100
500
Dose (mg/kg-d)
0
20
40
0
20
40
F0
0/19
1/20
2/24
0/19
0/20
2/24
Fib
0/20
1/19
4/1T
0/20
2/19
3/17a
F2b
0/20
1/20
1/20
0/20
0/20
3/20
Total
0/5 9b
3/59
7/6 la
0/5 9b
2/59
8/6 la
aSignificantly different from controls (p < 0.05), as calculated by the authors.
Statistically significant by Cochran-Armitage trend test,/? <0.01.
Sources: Friedman and Beliles (2002); Litton Bionetics (1992).
Table 4-36 summarizes the noncancerous effects observed in chronic oral studies of AN
in rats and mice. Table 4-37 summarizes the cancerous effects observed in chronic oral studies
of AN in rats and mice.
Table 4-36. Summary of chronic oral toxicity studies of AN: noncancer
effects in rats and mice
Strain
number/sex
Exposure
route/
duration"
Doses
Effects
NOAEL /
LOAELa
References
Comments
Rats
F344
100/sex/group
Unexposed
controls =
200/sex
Drinking
water/
2 yrs
M: 0,0.1,0.3,
0.8, 2.5,
8.4 mg/kg-d;
F: 0,0.1,0.4,
1.3, 3.7,
10.9 mg/kg-d
Squamous cell
hyperplasia/
hyperkeratosis of
the forestomach;
decreased
survival in high-
dose groups,
increase in
epidermal
inclusion cysts of
the skin
NOAEL=
0.1 mg/kg-d;
LOAEL =
0.3 mg/kg-d
(M)
0.4 mg/kg-d (F)
Johannsen and
Levinskas
(2002b);
Biodynamics
(1980c)
10 rats/sex/group
were taken from
exposed and
control groups for
interim sacrifice at
6, 12, and 18 mos
152
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-36. Summary of chronic oral toxicity studies of AN: noncancer
effects in rats and mice
Strain
number/sex
Exposure
route/
duration"
Doses
Effects
NOAEL /
LOAEL"
References
Comments
Sprague-
Dawley
100/sex/group
Drinking
water:
22 months
(M);
19 months
(F)
M: 0, 0.09,
8.0 mg/kg-d;
F: 0,0.15,
10.7 mg/kg-d
Squamous cell
hyperplasia of
the forestomach,
decreased
survival in high-
dose groups,
reduction in
absolute/relative
pituitary weight
(F)
LOAEL =
0.09 mg/kg-d
(M)
Johannsen and
Levinskas
(2002a);
Biodynamics
(1980a)
10 rats/sex/group
were taken from
exposed and
control groups for
interim sacrifice at
6, 12, and 18 mos
Sprague-
Dawley
100/sex/group
Gavage/
20 months
0,0.1,
10 mg/kg-d
Squamous cell
hyperplasia of
the forestomach,
small reductions
in hematocrit,
Hb, and RBC
count in high-
dose males,
increase in
absolute/relative
liver weight in
high-dose groups
NOAEL=
0.1 mg/kg-d;
LOAEL =
10 mg/kg-d
Johannsen and
Levinskas
(2002a);
Biodynamics
(1980b)
10 rats/sex/group
were taken from
exposed and
control groups for
interim sacrifice at
6, 12, and 18 mos
Sprague-
Dawley
48/sex/group;
Controls:
80/sex
Drinking
water/
2 yrs
M: 0, 3.4, 8.5,
21.3 mg/kg-d;
F: 0,4.4,
10.8,
25.0 mg/kg-d
Hyperplasia/
hyperkeratosis of
the forestomach,
reduced survival
and BW,
minimal
progressive
nephropathy,
gliosis of the
brain
LOAEL =
4.4 mg/kg-d (F)
Quast (2002);
Quast et al.
(1980a)
F344
50/sex low-
dose;
197 males and
203 females
high-dose.
Controls:
51 -males and
49-females:
Drinking
water/
lifetime
M: 0, 12.9,
64.5 mg/kg-d;
F: 0, 14.8,
74.2 mg/kg-d
Neurological
signs
No data
Bigner et al.
(1986)
Doses calculated
using default
assumptions (U.S.
EPA, 1988);
noncancer effects
not evaluated in
the study
153 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-36. Summary of chronic oral toxicity studies of AN: noncancer
effects in rats and mice
Strain
number/sex
Exposure
route/
duration"
Doses
Effects
NOAEL /
LOAEL"
References
Comments
Mice
B6C3FJ
50/sex/group
Gavage/
2 yrs
0,2.5, 10,
20 mg/kg-d,
5 d/wk;
continuous
exposure-
adjusted
doses: 0, 1.8,
7.14,
14.3 mg/kg-d
Reduced survival
in high dose
group;
hyperkeratosis/
hyperplasia of
the forestomach;
ovarian cysts and
atrophy (F)
LOAEL:
1.8 mg/kg-d
(F); NOAEL =
1.8 mg/kg-d;
(M) LOAEL =
7.1 mg/kg-d
(M)
NTP (2001)
aM = male; F = female.
Table 4-37. Summary of chronic oral toxicity studies of AN: cancer effects
in rats and mice
Strain
number/sex
Exposure
route/
Duration
Doses3
Effects"
References
Comments
Rats
F344
100/sex/group
Unexposed
controls =
200/sex
Drinking
water/
2 yrs
Males: 0,0.1,
0.3,0.8,2.5,
and 8.4 mg/kg-
d;
females: 0, 0.1,
0.4, 1.3, 3.7,
and
10.9 mg/kg-d
Males: increase in brain
astrocytomas and Zymbal
gland tumors, increase in
forestomach tumors; females:
increase in brain
astrocytomas, Zymbal gland
tumors, and forestomach
tumors
Johannsen
and
Levinskas
(2002b);
Biodynamics
(1980c)
10 rats/sex/group
were taken from
exposed and control
groups for interim
sacrifice at 6, 12,
and 18 months
Sprague-
Dawley
100/sex/group
Drinking
water/
22 months
(M);
19 months (F)
Males: 0, 0.09,
and 8.0 mg/kg-
d; females: 0,
0.15, and
10.7 mg/kg-d
Significant increases in
incidences of tumors of the
CNS, Zymbal gland, and
forestomach
Johannsen
and
Levinskas
(2002a);
Biodynamics
(1980a)
10 rats/sex/group
were taken from
exposed and control
groups for interim
sacrifice at 6, 12,
and 18 months
Sprague-
Dawley
100/sex/group
Gavage/
20 months
0, 0.1, and
10 mg/kg-d
Increase in brain
astrocytomas, and tumors of
Zymbal gland, forestomach,
and intestine. Carcinomas of
the mammary gland in
females
Johannsen
and
Levinskas
(2002a);
Biodynamics
(1980b)
10 rats/sex/group
were taken from
exposed and control
groups for interim
sacrifice at 6, 12,
and 18 months
Sprague-
Dawley
48/sex/group;
Controls:
80/sex
Drinking
water/
2 yrs
Males: 0, 3.4,
8.5, and
21.3 mg/kg-d;
females: 0, 4.4,
10.8, and
25.0 mg/kg-d
Increases in CNS tumors,
squamous cell papillomas or
carcinomas of the
forestomach, carcinomas of
Zymbal gland, as well as
benign and malignant tumors
of the mammary gland
Quast (2002);
Quast et al.
(1980a)
154
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-37. Summary of chronic oral toxicity studies of AN: cancer effects
in rats and mice
Strain
number/sex
Exposure
route/
Duration
Doses3
Effects"
References
Comments
F344
50/sex low-
dose; 197 males
and 203 females
high-dose.
Controls:
51 males and
49 females
Drinking
water/
lifetime
M: 0, 12.9, and
64.5 mg/kg-d;
females: 0,
14.8, and
74.2 mg/kg-d
Tumors found included brain
astrocytomas, Zymbal gland
tumors, and papillomas of the
forestomach
Bigner et al.
(1986)
Doses calculated
using default
assumptions (U.S.
EPA, 1988); no
tumor incidence data
were provided
Sprague-
Dawley
40/sex
Control: 75/sex
Gavage/
3 ds/week for
52 weeks; rats
maintained
without
treatment
until natural
death (study
ended week
131)
0, 5 mg/kg in
olive oil
Treatment-related increase in
forestomach tumors
Maltoni et al.
(1977)
Study limitations:
single dose, short
exposure period, and
small number of
animals in each
group
Sprague-
Dawley
20 M/group
Drinking
water/2 yrs
0, 1.5, 7.1, and
28 mg/kg-d
Increase in tumors of the
forestomach and Zymbal
gland
Gallagher et
al. (1988)
Study limitations:
small group size,
lack of testing in
females, and only
limited tissues were
examined for
histopathology.
Sprague-
Dawley
15 M/group,
30 F/group
Drinking
water/
46 weeks
(three-
generation
reproduction/
developmental
study)
0, 11,
37 mg/kg-d
(M); 0, 20, and
40 mg/kg-d (F)
Increase in incidences of
brain and Zymbal gland
tumors
Friedman and
Beliles
(2002);
Litton
Bionetics
(1992)
Study suggested
increased
susceptibility to the
carcinogenicity of
AN from early-life
exposure (see
Section 4.8.1).
Mice
B6C3FJ
50/sex/group
Gavage/
2 yrs
0, 2.5, 10, and
20 mg/kg-d,
5 d/wk,
continuous
exposure-
adjusted doses:
0, 1.8,7.14, and
14.3 mg/kg-d
Increase in tumors of the
forestomach and of Harderian
gland in males and females
NTP (2001)
Overall incidence of
alveolar/bronchiolar
adenomas or
carcinomas was
significantly
elevated in females
at 10 mg/kg-d, but
not at 20 mg/kg-d.
aM = male; F = female.
155 DRAFT - DO NOT CITE OR QUOTE
-------
4.2.2. Inhalation Exposure
4.2.2.1. SubchronicSubchronic Studies
The only subsubchronic study of inhalation effects of AN in animals was a comparative
study of the neurotoxicity of nitriles (Gagnaire et al., 1998). Groups of 12 male Sprague-Dawley
rats were exposed (whole body) to AN vapor at concentrations of 25, 50, or 100 ppm,
6 hours/day, 5 days/week for 24 weeks. A control group of 10 male rats was exposed to filtered
air. BWs were measured weekly. Following 4, 8, 12, 16, 20, and 24 weeks of exposure and an
8-week recovery period (week 32), rats were evaluated for the same electrophysiological
parameters that were tested in parallel experiments on orally exposed rats (see Section 4.2.1.1).
As in the companion study, electrophysiological testing was performed at least 16 hours after
daily exposure (waiting period was 48 hours for weekends). Electrical stimulation of the tail
nerve was used to assess MCVs, SCVs, ASAPs, and AMAPs. No mortality was observed during
the treatment period, but, during the first and second week of the recovery period, 2/12 rats in the
100 ppm group died and one rat each in the 100 and 25 ppm groups was euthanized in week 31
because of tumors in the neck. BW gain in the 100 ppm group was significantly lower than in
controls in weeks 4, 8, 16, and 21-24, such that the BW was 11% lower at the end of week 24.
Rats exposed to AN did not develop weakness of the hind limbs or disturbances in gait.
After 1 or 2 weeks of exposure, rats exposed at >50 ppm exhibited clinical signs of gross toxicity
(wet fur and excessive salivation but not hyperactivity). Excessive salivation was attributed by
the study authors to a cholinomimetic effect of AN. Exposure to AN had no effect on
neurophysiological parameters during the first 8 weeks and no effect on the AMAP at any time
during the study. Statistically significant concentration-dependent SCV reductions of-9%
compared with controls were observed in the 100 ppm group from weeks 12-24 (Table 4-38). In
week 12, an -7% reduction was observed in the 50 ppm group, and in week 24 the SCV was
reduced by 5% in the 25 ppm group (not biologically significant) and by >8% in the 50 and
100 ppm groups. The ASAP was significantly reduced by 14.5-20% in the 50 ppm group and by
29-30% in the 100 ppm group from weeks 16-24. After recovery in week 32, a 21% reduction
in ASAP persisted in the 100 ppm group. Sporadic reductions in MCV were observed in
100 ppm rats beginning in week 16(11%) reduction) but were not concentration dependent in
week 24 or after recovery in week 32. A LOAEL of 25 ppm was identified for reductions in
SCV in rats exposed to AN by inhalation for 24 weeks. A NOAEL was not identified.
156
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-38. Effect on SCV in male Sprague-Dawley rats exposed to AN via
inhalation for 24 weeks
Exposure
(ppm)
SCV (m/s)a
Exposure (weeks)
Recovery
0
12
16
20
24
32
0
35.0 ±0.5
49.7 ±0.8
49.9 ± 1.0
50.3 ±0.5
53.3 ±1.0
53.4 ±0.6
25
35.2 ±0.4
48.2 ±0.7
47.8 ± 1.0
50.2 ±0.7
50.5 ± 0.8b
51.8 ± 0.8
50
35.3 ±0.6
46.3 ± 0.8°
48.0 ± 1.1
50.5 ±0.6
49.1 ±0.5d
51.3 ± 1.0
100
35.8 ±0.5
45.3 ± 1.0C
46.2 ± 0.7b
48.1 ±0.7b
48.4 ± 1.0d
50.4 ±0.8
aValues are means ± SDs (n = 12 for treated, n = 10 for controls).
Statistically significant compared with controls (p < 0.05) as calculated by the study authors.
Statistically significant compared with controls (p < 0.01) as calculated by the study authors.
Statistically significant compared with controls (p < 0.001) as calculated by the study authors.
Source: Gagnaire et al. (1998).
4.2.2.2. Chronic Studies
Two chronic inhalation bioassays in rodents were available: Dow Chemical (1992a)/
Quast et al. (1980b) and Maltoni et al. (1988, 1977).
4.2.2.2.1. Dow Chemical (1992a) and Quast et al. (1980b). Dow Chemical (1992a) and Quast
et al. (1980b) evaluated the effects of AN in Sprague-Dawley rats (100/sex/group) exposed by
inhalation at concentrations of 0, 20, or 80 ppm (0, 43.4, or 173.6 mg/m ) 6 hours/day,
5 days/week for 2 years. Additional groups of 7 and 13 rats/sex/group were exposed and
sacrificed at 6 and 12 months, respectively. BWs were determined 10 times during the first
3 months and monthly thereafter. Rats were observed daily for clinical signs and mortality, and
beginning after 6 months, were examined for palpable masses and dental condition. Moribund
animals or those with ulcerating tumors were sacrificed and subjected to gross necropsy.
Hematology and urinalysis examinations were conducted on 10 rats/sex/group on days 174/175,
365/366, 616/617, and 727/727 for males/females, respectively. The hematology determinations
of male rats was also conducted on day 183 to verify observations made on day 174. Clinical
chemistry analyses were conducted on 10 rats/sex/group on days 176, 372, and 735. Water
consumption was determined for representative male and female rats in each group for the first
8 months. At terminal sacrifice, all rats received an ophthalmologic examination and necropsy
during which organ weights were recorded for brain, heart, liver, kidneys, and testes. Complete
histologic examinations were carried out on all rats in the control and 80 ppm groups at terminal
sacrifice. More than 80% of rats in the 20 ppm group were examined for gross lesions, and
23 selected organs and all tissues with grossly recognized tumors were collected for
histopathology examination. Because there were signs of upper respiratory tract irritation in the
nasal turbinate, about 10 rats/sex/group from the terminal sacrifice were evaluated by light
157
DRAFT - DO NOT CITE OR QUOTE
-------
microscopy. In addition, because of brain lesions observed during drinking water studies, nine
sections from various regions of the CNS of all rats were examined microscopically.
Noncancer results
Inhalation exposure to AN resulted in significant concentration-related noncancer effects
compared with controls (Quast et al., 1980b). Statistically significant (p < 0.05) decreases in
survival with respect to controls were observed in males after 6 months of exposure at 80 ppm
and in females after 10 months of exposure at 80 ppm or 22 months at 20 ppm. The numbers of
rats surviving at termination (out of 100/sex) were 18, 14, and 4 males and 22, 9, and 1 females
in the control, 20 ppm, and 80 ppm groups, respectively. BWs were decreased by about 10-15%
in male and female rats after 9 months of exposure to 80 ppm AN. A significant decrease of less
than 10% was also observed in 20 ppm female rats. By the end of the study, BWs in 20 and 80
ppm females were not significantly different from those in controls.
Hb and RBC counts were significantly lower (by -9%) in rats exposed to 80 ppm for 4-
8 months but not later. However, the study authors considered these changes to be a secondary
effect of reduced growth, tumor formation, and hemorrhage, resulting from exposure and not due
to bone marrow toxicity. Statistically significant increases in water consumption were observed
in both exposed groups of male and female rats during the first 6 months of the study and were
consistent with slightly decreased urine specific gravity measured in 80 ppm groups during that
period. No significant effects on urinalysis parameters were observed after 6 months. Exposure
to AN had no consistent significant effect on clinical chemistry parameters or results of
ophthalmoscopic examinations. A significant elevation (about 26%) in blood urea nitrogen
(BUN) was observed in the 20 and 80 ppm female rats on day 176. SGPT was also elevated by
57%) in the 80 ppm females at that time. However, no significant findings of these parameters
were found upon subsequent evaluation at a later time interval. Hence, Quast et al. (1980b)
considered these changes as secondary responses and not indications of direct renal or
hepatotoxicity from AN exposure. Significant increases in relative weights of brain, heart, and
testes of male rats exposed at 80 ppm were considered by the study authors to be a consequence
of the reduction in BW.
Gross pathological examinations found statistically significantly findings in the nasal
turbinates, lungs, teeth (malocclusion), and liver of the 80 ppm rats. Gross observation of male
rats indicated significant increase in minimal chronic nephropathy in the 80 ppm group
(40/100 vs. 24/100). Significant increase in pneumonia, atelectasis, or edema was found in the
20 and 80 ppm males (14/100, 27/100, and 30/100 for 0, 20, and 80 ppm groups, respectively).
Gross observations of nontumorous changes included enlarged liver in female rats, with
incidence of 2/100, 9/100, and 7/100 in 0, 20, and 80 ppm groups, respectively. (The increased
incidence in the 20 ppm group was statistically significant.)
158
DRAFT - DO NOT CITE OR QUOTE
-------
Statistically significant increases in the incidence of histopathologic lesions of the nasal
turbinates were observed in all rats exposed at 80 ppm and most rats in the 20 ppm group. These
effects were considered by the study authors to be the result of irritant effects of AN
(Table 4-39). These lesions appeared in rats sacrificed after at least 13 months (usually
19 months) of exposure. These inflammatory and degenerative changes included hyperplasia,
flattening, focal erosion, and squamous metaplasia of the respiratory epithelium and hyperplasia
of mucus secreting cells. Flattening of the respiratory epithelium in females and hyperplasia of
mucus-secreting cells in males were both significantly increased at the 20 ppm exposure level.
Table 4-39. Incidence of histopathological lesions of the nasal turbinates in
Sprague-Dawley rats exposed to AN via inhalation for 2 years
Response
Concentration (ppm)
0
20
80
Incidence
Males
Suppurative rhinitis
0/11
1/12
5/10a
Hyperplasia of respiratory epithelium
0/11
4/12
10/103
Focal erosion of mucous lining
0/11
0/12
4/10a
Squamous metaplasia of the respiratory epithelium
0/11
1/12
7/10a
Hyperplasia of mucus-secreting cells
0/11
7/12a
8/10a
Focal inflammation
0/11
1/12
1/10
Flattening of the respiratory epithelium
0/11
2/12
3/10
Females
Suppurative rhinitis
1/11
0/10
2/10
Hyperplasia of respiratory epithelium
0/11
2/10
5/10a
Focal erosion of mucous lining
0/11
1/10
1/10
Squamous metaplasia of the respiratory epithelium
0/11
2/10
5/10a
Hyperplasia of mucus-secreting cells
0/11
2/10
8/10a
Focal inflammation
2/11
6/10
7/10a
Flattening of the respiratory epithelium
1/11
7/10a
8/10a
"Statistically significant (p < 0.05), as calculated by the study authors.
Source: Quast et al. (1980b).
Increase in acute suppurative pneumonia was observed in the lungs of the 80 ppm male
rats during the 7-12-month time interval. A nonsignificant increase was also observed in the
20 ppm group.
Other histopathological observations included an increase in the incidence of gliosis and
perivascular cuffing in the brain of high-dose rats (either sex) and, in males only, minimal
chronic focal progressive nephrosis and formation of keratinized cysts in the thyroid gland
(Table 4-40). Incidences of focal necrosis of the liver were increased in 20 and 80 ppm female
159
DRAFT - DO NOT CITE OR QUOTE
-------
rats. The incidence of AN-related hyperplasia and hyperkeratosis of the nonglandular portion of
the stomach did not achieve statistical significance in either sex but was statistically significant
(p < 0.05) by Fisher's exact test when the data were combined. There were concentration-related
increases in the incidences of several lesions that were secondary to other effects of AN exposure
(numerical data not provided here). These included hepatocellular atrophy without fatty changes
and atrophy of mediastinal fat in 80 ppm male rats, attributed by the study authors to the
decreased feed intake of the rats at the end of the study, and extramedullary hematopoiesis of the
spleen in females, a consequence of the reductions in RBC and Hb counts. An increase in
lymphoid hyperplasia in males at 80 ppm was interpreted by the study authors to be secondary to
tumors of the ear canal and inflammatory changes in the nasal turbinates. A NOAEL was not
identified in this study. A LOAEL of 20 ppm was identified for increased lesions of the nasal
turbinates (hyperplasia of mucus-secreting cells in males and flattening of the respiratory
epithelium in females) and focal necrosis in liver of female rats exposed to AN vapor for 2 years.
Table 4-40. Incidence of dose-related noncancerous histopathological lesions
in Sprague-Dawley rats exposed to AN via inhalation for 2 years
Response
Concentration (ppm)
0
20
80
Incidence
Males
Focal nephrosis (progressive)
22/100
24/100
48/100a
Minimal chronic nephropathy
24/100
21/100
40/100a
Thyroid cyst
4/95
9/97
13/96a
Gliosis and perivascular cuffing (brain)
1/100
2/99
7/99a
Acute suppurative pneumonia
0/100
4/100
10/100a
Pulmonary changes
15/100
26/100a
20/100
Hyperplasia of the nonglandular epithelium (stomach)
8/98
7/100
16/99
Females
Gliosis and perivascular cuffing (brain)
0/100
2/100
8/100a
Focal necrosis in liver
3/100
16/100a
10/100a
Vasculization of spinal myelin (minimal)
42/100
67/100a
47/100
Hyperplasia of the nonglandular epithelium (stomach)
2/99
3/99
7/97
Statistical significance (p < 0.05), as calculated by the study authors.
Source: Quast et al. (1980b).
Cancer results
In rats chronically exposed to AN vapor, there were significant increases in tumors at
multiple sites, several of which also had been affected in oral exposure bioassays (Quast et al.,
1980b). In males and females exposed to 80 ppm AN, there were increased incidences of
160
DRAFT - DO NOT CITE OR QUOTE
-------
astrocytomas of the brain as well as glial cell proliferation that was considered an earlier stage in
the progression to astrocytomas. The incidence data in Table 4-41 combine the incidences of
astrocytomas and glial cell proliferation. The incidence of CNS tumors was also significantly
increased in females exposed to 20 ppm AN and insiginificantly increased in males exposed to
20 ppm AN. Carcinomas of Zymbal gland were significantly elevated in the 80 ppm rats of both
sexes. Other tumor increases observed in the 80 ppm groups were squamous cell papillomas or
carcinomas in the tongue and carcinomas of the intestinal tract of males and adenocarcinomas of
the mammary gland in females. An increase was observed in forestomach tumors in 80 ppm
male rats and nasal turbinate tumors in 80 ppm female rats. Both types of tumors were
considered by the study authors to be treatment related, although the increase was not
statistically significant. Table 4-41 presents incidences of AN-induced target organ-specific
tumor formation. All tumors were found in exposed rats that died or were sacrificed after
12 months of exposure, with the exception of one male rat with a CNS tumor and three female
rats with mammary gland adenocarcinomas that died between 7 and 12 months.
Table 4-41. Cumulative incidence of tumors in Sprague-Dawley rats
exposed to AN via inhalation for up to 2 years
Tissue
Males (ppm)
Females (ppm)
0
20
80
0
20
80
All brain/CNSa
0/96
4/93
22/82b
0/93
8/9 9b
20/89b
Zymbal gland
2/96
4/93
1 l/82b
0/93
1/98
1 l/89b
Intestinal tract
4/96
3/93
17/82b
Not increased
Mammary gland (adenocarcinomas)
0/100
0/100
1/100
9/93
8/98
20/99b
Mammary gland (total, benign and malignant)
4/100
5/100
7/100
88/100
95/100
85/100
Forestomach (squamous cell papilloma)
1/98
1/100
4/99
0/99
0/99
1/97
Tongue
1/95
0/14
7/82b
0/96
0/9
1/91
Nasal turbinate (carcinoma in respiratory
epithelial region)
Not increased
0/11
0/98
2/10
"Incidence data include all brain/CNS tumors (astrocytomas and glial cell proliferation). Male tumor incidence data
are from Tables 22, 25, and 26 of the Quast et al. (1980b) report; female tumor incidence data are from Tables 31,
34, and 35 in the same report. For all incidence data, the denominators excluded rats dying earlier than 12 mos in
the study. These data were ascertained from Tables 22, 25, 31, 34, and 35 in the original study report by Quast et
al. (1980b).
bSignificantly different from controls (p < 0.05) as calculated by the study authors.
Sources: Dow Chemical (1992a); Quast et al. (1980b).
An apparent decrease in the incidence of tumors of the pituitary, adrenals, thyroid, and
pancreas of male and female treated rats, and testes of males, was observed when compared with
controls.
161
DRAFT - DO NOT CITE OR QUOTE
-------
4.2.2.2.2. Maltoni et al (1988,1977). The following studies by Maltoni et al. (1988, 1977)
were cancer bioassays that did not investigate nonneoplastic effects in rats exposed to AN vapor.
These study authors reported a number of experiments in which Sprague-Dawley rats were
exposed to AN by inhalation. In the first (designated BT201 by the authors), 30 rats/sex/group
were exposed to 0, 5, 10, 20, and 40 ppm 4 hours/day, 5 days/week for 52 weeks; the animals
then were allowed to complete their natural life spans, with the final deaths occurring in week
136 (Maltoni et al., 1977). Rats were examined 3 times weekly for general health status and
subjected to a clinical examination for gross changes every 2 weeks. Rats were weighed every
2 weeks during the exposure period and monthly thereafter. All rats were subjected to gross
necropsy. Histopathologic examinations were conducted on all gross lesions and a selection of
about 12 organs and tissues, including the Zymbal glands, interscapular brown fat, salivary
glands, tongue, lungs, liver, kidneys, spleen, stomach, intestine, bladder, and brain.
Exposure to AN had no significant effect on survival or BWs in male or female rats. A
moderate increase in gliomas, forestomach tumors, Zymbal gland carcinomas, and mammary
tumors was found in the treated group. However, increases in the incidences of these tumors in
exposed rats were not statistically significant and could be due to the small number of animals in
each dose group. For example, gliomas were found in 20 and 40 ppm males at 1/30 and 2/30,
respectively, and not in controls and other exposure groups. The average latency of the gliomas
was shorter at the higher concentration (63.5 vs. 84 weeks). Forestomach papillomas and
acanthomas were found in males at 0/30, 1/30, 2/30, and 3/30 for 0, 5, 10, and 40 ppm,
respectively. In females, forestomach tumor incidences were 0/30, 1/30, 2/30, and 1/30 for the
0, 5, 10, and 20 ppm groups. The average latency of forestomach tumors ranged from 103 to
124 weeks. Zymbal gland carcinomas were also found at 1/30 in the 10 ppm group for male and
female and 1/30 for the female 20 ppm group. No Zymbal gland carcinomas were found in the
control and other dose groups.
Moreover, a possible treatment-related effect of AN in the mammary gland of females
was observed, as indicated by the following incidences of benign or malignant tumors in relation
to exposure level: 5/30 (controls), 10/30 (5 ppm), 7/30 (10 ppm), 10/30 (20 ppm), and
7/30 (40 ppm) (Maltoni et al., 1977). Slight discrepancies in the incidence of tumors of the
mammary gland were reported for controls and highest-exposed females in a subsequent report
of this study (Maltoni et al., 1988). In the later study, the incidence of tumors in control females
was listed as 20%, an incidence of 6/30, whereas that of highest-concentration females was listed
as 26.7%, an incidence of 8/30. In the later report, taking both sexes together, the incidences of
encephalic gliomas were 3/60 and 2/60 in the 40 and 20 ppm groups, respectively, and none in
the control and lower-dose groups. Although this cancer study was consistent with other studies
in the identification of tumor sites associated with exposure to AN, it was limited in that the
exposure concentrations were too low, the number of animals per group was too small, and the
exposure duration was too short to result in statistically significant increases in specific tumor
162
DRAFT - DO NOT CITE OR QUOTE
-------
types. However, this study demonstrated that rats with less-than-lifetime exposure and allowed
to recover to their natural life span developed the same tumor types as rats with lifetime exposure
to AN.
The later report by Maltoni et al. (1988) described two additional lifetime exposure
experiments in which Sprague-Dawley rats were exposed to AN by inhalation beginning during
the gestation period. In the first experiment (designated BT4003), 54 adult pregnant females,
beginning on gestation day 12, were exposed to 60 ppm AN for 4 hours/day, 5 days/week for
7 weeks and then 7 hours/day, 5 days/week for 97 weeks. A group of 60 unexposed adult
females served as controls. Gestation was permitted to proceed normally and the offspring were
exposed on the same schedule as the dams. The exposed offspring included 67 males and
54 females, whereas the controls were 158 males and 149 females.
Overall, there was a statistically significant treatment-related increase in the percentage
of dams with malignant tumors at all sites (37 vs. 15%). Increased incidences in exposed dams
compared with controls were observed for several sites, but none of these was statistically
significant: Zymbal gland carcinomas (5.5 vs. 1.7%), mammary gland carcinomas (5.5 vs.
3.3%>), benign or malignant mammary gland tumors (69 vs. 40%>), extrahepatic angiosarcomas
(1.8 vs. 0%>), and encephalic gliomas (5.5 vs. 0%>). No hepatomas were observed in exposed or
unexposed dams. These site-specific results are summarized in Table 4-42.
Table 4-42. Comparison of carcinogenic effects of chronic exposure to AN at
60 ppm starting either in utero or in adulthood, in Sprague-Dawley rats
Percent with tumor
Stage
during
exposure
Exposure
protocol
Sexa
Number
of rats at
startb
Brain
tumors
(encephalic
gliomas)
Zymbal
gland
carcinomas
Hepatomas
Malignant
mammary
tumors
Extra-
hepatic
angio-
sarcomas
Chronic0
F
54
5.5
5.5
0.0
5.5
1.8
Adult only
Unexposed
controls
F
60
0.0
1.7
0.0
3.3
0.0
Chronic
M
F
67
54
16.4
18.5
14.9°
1.8
7.5
1.8
0.0
16.7d
4.4
5.5e
Starting at
GD 12
Subchronicd
M
F
60
60
5.0
3.3
6.7
1.7
1.7
0.0
0.0
6.7
5.0
1.7
Unexposed
controls
M
F
158
149
1.3
1.3
1.3
0.0
0.6
0.0
1.9
5.4
0.6
0.0
aF = female; M = male.
bAnimals were allowed to live until spontaneous death.
°Chronic: 4 hrs/d, 5 d/wk for 7 wks (starting during gestation), followed by 7 hrs/d for 97 wks.
Statistically significantly higher than correspondng control incidence,/? < 0.05.
eSubchronic: 4 hrs/d, 5 d/wk for 7 wks (starting during gestation), followed by 7 hrs/d for 8 wks.
Source: Maltoni et al. (1988).
163
DRAFT - DO NOT CITE OR QUOTE
-------
In contrast, chronically exposed male and female offspring showed statistically
significant increases in the incidences of specific types of tumors: malignant tumors of the
mammary gland in females (16.7 vs. 5.4%; p < 0.05), extrahepatic angiosarcomas in females
(5.5 vs. 0%;p < 0.05), hepatomas in males (7.5 vs. 0.6%; p < 0.05), Zymbal gland carcinomas in
males (14.9 vs. 1.3%;/? < 0 .01), and encephalic gliomas (16.4% in males and 18.5% in females
vs. 1.3% for controls of both sexes; p < 0.01) (see Table 4-42). Higher responses in the exposed
offspring at these sites relative to the dams suggests some early-life susceptibility to AN
carcinogenity. These results are discussed further in Section 4.8.1.
The second exposure experiment (BT4006) involving gestational exposure of Sprague-
Dawley rats started with the same exposure conditions as for experiment BT4003, except that
exposure of the offspring (127 males and 114 females) ended after 15 weeks (Maltoni et al.,
1988). Exposure in this group of offspring was for 4 hours/day, 5 days/week for 7 weeks starting
on GD 12, followed by 7 hours/day, 5 days/week for 8 weeks. All animals were kept under
observation until spontaneous death, at which time they were examined for the presence of
tumors. The control group of offspring was the one used in experiment BT4003 (158 males and
149 females). There was a statistically significant increase in the total incidence of malignant
tumors in exposed offspring compared with controls for both males (31.7 vs. 17.1%,/? < 0.05)
and females (35.0 vs. 17.4%,/? < 0.01). Increases in the following tumors were observed when
compared with controls: Zymbal gland tumors in males and females (4.2 vs. 0.7%); extrahepatic
angiosarcomas in males (5.0 vs. 0.6%); encephalic gliomas in males and females (4.2 vs. 1.3%);
and hepatomas in males (1.7 vs. 0.6%) (see Table 4-42). However, these increases were not
statistically significant.
4.3. REPRODUCTIVE/DEVELOPMENTAL STUDIES—ORAL AND INHALATION
In contrast to the extensive information on cancer and noncancer effects of AN in
experimental animals and exposed human populations, there are comparatively few data that
address the reproductive and developmental toxicity and teratogenic action of AN.
4.3.1. Studies in Humans
There have been some preliminary translated reports of surveys of the reproductive
history of women occupationally exposed to AN in chemical factories in China. While these
studies pointed to a possible link between occupational exposure to AN and adverse effects on
pregnancy, in some cases, there was insufficient information in the reports to assess the validity
of these claims. However, there are studies with usable data. As described in Section 4.1.2.2,
Dong and Pan (1995) provided data on exposure and reproductive outcomes in 150 male and
106 female workers employed at a chemical fiber plant. These subjects, routinely exposed to
-3
AN concentrations ranging from 0.53 to 15.5 mg/m (0.24-7.1 ppm), were compared with
110 male and 121 female workers who had not been exposed to AN. The results of the study
164
DRAFT - DO NOT CITE OR QUOTE
-------
were divided into two sections, one dealing with a comparison of the reproductive outcome for
female employees, exposed vs. unexposed, and the second comparing the reproductive outcome
of the wives of exposed male employees to that of the wives of unexposed male controls. In the
former case, there were few differences in reproductive performance between exposed females
and controls (see Table 4-21 for details), although the incidence of stillborn fetuses in the
exposed group was significantly greater than in controls (5/112 vs. 0/124). However, it should
be noted that there was a statistically significant age difference between the groups
(29.36 ± 5.12 in exposed females vs. 20.68 ± 4.05 in female controls). The wives of exposed
male workers at the plant had poorer reproductive outcomes than the wives of male controls,
even though they were significantly younger than their counterparts. For example, there were
differences in the incidence of live births as a proportion of the total number of pregnancies
(159/168 vs. 113/113), spontaneous abortions (8/168 vs. 1/113), threatened abortions (5/168 vs.
1/113), stillborn fetuses (4/168 vs. 0/113), and sterility (9/150 vs. 2/110). These data suggested a
potential AN-related impact on the sperms and reproductive performance of exposed males.
An elevated incidence of complications in pregnancy was reported also by Li (2000) in
379 female workers exposed to an average AN concentration of 7.5 ppm for >1 year. When
compared with 511 unexposed controls, exposed subjects had a higher rate of overall
complications (20.8 vs. 7.14%) and premature deliveries (11.62 vs. 4.72%). Exposed subjects
were subdivided to create a category where both partners had been exposed to AN. Overall rates
of complications, premature delivery, overdue delivery, and deficiency were greater than in cases
where only the female partner was exposed.
In another Chinese study, Wu et al. (1995) reported statistically significant increases in
the incidence of pernicious vomiting, anemia, preterm delivery, and birth defects in 477 females
exposed to AN in the workplace compared with 527 controls.
The epidemiological report and follow-up by Czeizel et al. (2000, 1999) examined the
environmental distribution of congenital abnormalities in 30 settlements within a 25 km radius of
an AN factory, drawing on data from the Hungarian Congenital Abnormality Registry covering
46,326 infants born between 1980 and 1996. A number of time-space-specific clusters of
abnormalities were identified among the subjects, the most striking of which was the incidence
of pectus excavatum in the community of Tata between 1990 and 1992. This effect was
associated with an OR of 78.5 and a 95% CI of 8.4-729.6. Other clusters of congenital
abnormalities in the vicinity of the factory were undescended testis in the community of
Nyergesujfalu between 1980 and 1983 (OR = 8.6, CI = 1.4-54.3) and at Esztergom between
1981 and 1982 (OR = 4.2, CI = 1.3-13.5) and clubfoot in the Tata community between 1980 and
1981 (OR = 5.5, CI= 1.5-20.3).
As set forth in Section 4.1.2.2, the relationship between the findings of clusters of
abnormalities and AN is uncertain because there are no exposure data in the report (Czeizel et
al., 2000, 1999). Consequently, it is not known by how much the mothers of affected infants
165
DRAFT - DO NOT CITE OR QUOTE
-------
were exposed to the compound if at all. Furthermore, when the incidence of these and other
abnormalities was considered for infants born throughout the duration of the study, there were no
overall increased incidences of congenital abnormalities within the study area compared with
infants born throughout Hungary.
The study authors stated that there was a technological change at the AN factory in 1984
that resulted in greater environmental protection, implying that releases of the compound to the
environment were less after 1984 than before the change. This suggested that the cluster of
pectus excavatum obtained at Tata between 1990 and 1992 was unlikely to have been due to AN
exposure. However, the high incidence of undescended testis in Nyergesujfalu between 1980
and 1983 may have been an environmental phenomenon, because the region-wide incidence of
this congenital abnormality appeared to decrease with increasing distance from the factory. In
general, however, it was difficult to draw conclusions about a link between maternal exposure to
AN and the incidence of congenital abnormalities from the data in this study because of a lack of
exposure data.
Ivanescu et al. (1990) reported lower serum testosterone level in male workers exposed to
AN in a chemical factory (see Section 4.1.2.2.2).
Xu et al. (2003) performed conventional sperm analysis according to WHO guidelines
and investigated DNA strand breakage and sex chromosome aneuploidy in spermatozoa of
30 AN-exposed workers compared with 30 unexposed controls. The mean concentration of AN
"3
at exposure sites was reported to be 0.8 ± 0.25 mg/m (0.37 ppm). Sperm density was
significantly lower in the exposed group (75 x 106/mL) than in the control group (140 x 106/mL).
Sperm number per ejaculum was 205 x 106 in the exposed group, significantly lower than the
280 x 106 spermatozoa in the control. However, there were no significant differences between
the groups in semen volume, sperm motility, viability, or morphology. Xu et al. (2003) used
single cell gel electrophoresis (comet assay) to monitor the incidence of DNA strand breakage of
sperm cells. The rate of comet sperm nuclei was 28.7% in the exposed group, significantly
higher than in the control group (15.0%). Mean comet tail length was 9.8 [j,m in exposed
workers but 4.3 [j,m in control workers. The frequency of sex chromosome aneuploidy in sperm
cells was analyzed using fluorescence in situ hybridization (FISH). Sex chromosome disomy
was found to be 0.69% in the exposed group, significantly higher than 0.35% in controls.
XY-bearing sperm was the most common sex chromosome disomy, with an average rate of
0.37%) in exposed vs. 0.20% in controls. XX- and YY-bearing sperm accounted for an additional
0.09 and 0.23% of sperm in exposed vs. 0.05 and 0.10%> in controls, respectively. Xu et al.
(2003) concluded that AN exposure affected semen quality among occupationally exposed
persons by the induction of DNA strand breakage and sex chromosome nondisjunction.
166
DRAFT - DO NOT CITE OR QUOTE
-------
4.3.2. Studies in Animals
4.3.2.1. Oral Studies
Assessments of reproductive/developmental effects of AN in orally exposed animals are
derived from standard toxicity assays in rats and mice, standard developmental toxicity assays in
female rats, and a three-generation reproductive toxicity assay in male and female rats.
4.3.2.1.1. Standard toxicity assays: reproductive organ pathology. As described in
Section 4.2.1, standard 2-year oral toxicity assays in rats revealed no evidence for increased
reproductive histopathology in males or females exposed to AN at doses as high as 8-25 mg/kg-
day (Johannsen and Levinskas, 2002a, b; Quast, 2002; Biodynamics 1980a, b, c; Quast et al.,
1980a). In addition, there was no evidence for adverse effects of AN on functional reproductive
parameters (sperm morphology, estrous cycle) or the histology of reproductive organs in male
B6C3Fi mice treated with AN by gavage 5 days/week at 20 mg/kg-day or in females at
40 mg/kg-day for 14 weeks (NTP, 2001). In this subchronic study, the weights of the left cauda
epididymides were significantly elevated compared with controls in male mice exposed to
10 and 20 mg/kg-day, but this effect was not considered biologically significant in the absence of
histopathology. In the companion 2-year gavage assay (NTP, 2001), no reproductive
histopathology was observed in male mice treated 5 days/week with AN at doses as high as
20 mg/kg-day, but effects were observed in females (Table 4-33). The incidence of ovarian cysts
was significantly elevated at 2.5, 10, and 20 mg/kg-day, and the incidence of atrophy of the
ovary increased at 10 and 20 mg/kg-day. Atrophy was severe and was characterized by lack of
histologically evident follicle and corpus luteum development with a predominance in interstitial
tissue. The lowest dose of 2.5 mg/kg-day in this study was identified as the LOAEL for ovarian
cysts and atrophy.
4.3.2.1.2. Developmental toxicity assays. Dow Chemical (1992b) and Murray et al. (1978) evaluated
developmental effects in pregnant female Sprague-Dawley rats (29-39/group) that received 10,
25, or 65 mg/kg AN by gavage in water on GDs 6-15; an additional group of 43 controls
received water alone. Dams were observed daily for clinical signs and weighed on GDs 6, 10,
16, and 21. Food and water consumption were monitored at 3-day intervals on GDs 6-21, at
which point all animals were sacrificed and maternal liver weights were recorded. Among the
reproductive parameters evaluated were the numbers and positions of live, dead, and resorbed
fetuses and the number of implantation sites. Developmental toxicity was evaluated by the
weight, sex ratio, and crown-rump length of the fetuses. One-third of the fetuses in each litter
were evaluated for visceral malformations and soft tissue abnormalities; the rest were examined
for skeletal alterations.
Maternal toxicity of AN was most evident in the high-dose group. Systemic effects
observed only at this dose included a single maternal death on GD 6, an increase in clinical signs
167
DRAFT - DO NOT CITE OR QUOTE
-------
(hyperexcitability and excessive salivation) during the dosing period, BW gain reduced 46%
compared with controls by GD 15 (22% by GD 21), water consumption significantly increased
by an unspecified amount on GDs 6-20, and a statistically significant 11% increase in absolute
(but not relative) liver weight compared with controls. Food consumption was significantly
reduced by an unspecified amount compared with controls in high- and mid-dose dams on
GDs 6-8. At necropsy, most (number not specified) dams dosed with 65 mg/kg-day and
3/33 dams dosed with 25 mg/kg-day displayed a thickening of the nonglandular portion of the
stomach. Pregnancy rate was significantly decreased among dams given 65 mg/kg-day AN, with
only 20 of 29 dams producing litters. Uterine staining revealed implantation sites in four
additional dams. A NOAEL for maternal toxicity was identified as 10 mg/kg-day AN, and
25 mg/kg-day was the LOAEL for hyperplasia of forestomach.
Exposure to AN had no statistically significant effect on the average numbers of
implantations/dam, live fetuses/litter, or resorptions/litter; the average sex ratio of litters was not
reported. At 65 mg/kg-day, there were statistically significant reductions in fetal BW (by 7.4%)
and crown-rump length (by 1.8%) compared with controls. The high-dose group also showed a
significant increase in external malformations (short tail in 6/17 litters and short trunk in
3/17 litters); there was also a significant increase in skeletal malformations (missing vertebrae)
(Table 4-43). The defect ranged in severity from the absence of a single lumbar vertebra to the
absence of all sacral and lumbar vertebrae and most thoracic vertebrae. Although the incidence
of malformations at 25 mg/kg-day was not statistically significant (Table 4-43), a minimal
LOAEL of 25 mg/kg-day was identified for fetal malformation in Sprague-Dawley rats. The
NOAEL for fetal toxicity was 10 mg/kg-day.
168
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-43. Incidence of fetal abnormalities among litters of Sprague-
Dawley rats following maternal exposure to AN on GDs 6-15
Type of malformation
AN (mg/kg-d)
0
10
25
65
Number offetuses affected/number of litters examined
External and skeletal malformations
443/38
388/35
312/29
212/17
Visceral malformations
154/38
135/35
111/29
71/17
Number of fetuses (litters) affected
External malformations
Short tail
1(1)
0(0)
2(2)
8 (6)a
Short trunk
0(0)
0(0)
0(0)
3 (3)a
Imperforate anus
0(0)
0(0)
0(0)
2(2)
Visceral abnormalities
Right side aortic arch
0(0)
0(0)
1(1)
1(1)
Missing kidney, unilateral
1 (1)
0(0)
0(0)
1 (1)
Anteriorly displaced ovaries
0(0)
0(0)
1 (1)
1 (1)
Skeletal malformations
Missing vertebrae
1(1)
0(0)
2(2)
8 (6)a
Missing two vertebrae and two ribs
7(1)
0(0)
7(2)
0(0)
Hemivertebrae
0(0)
0(0)
0(0)
0(0)
Total malformed
8(2)
0(0)
10(4)
8 (6)a
aSignificantly different from controls (p < 0.05), as determined by the study authors.
Source: Murray et al. (1978).
Behavioral teratogenicity of AN was examined in the progeny of pregnant Wistar rats
(15/group) that received 0 or 5 mg/kg-day AN by gavage on GDs 5-21 (Mehrotra et al., 1988).
Dams were weighed at intervals, and food and water intakes and the length of gestation were
recorded. At parturition (postnatal day [PND] 0), litters were culled to four males and four
females. On PND 1, pups were sexed and examined for gross external anomalies; litters with
fewer than two/sex were rejected. Eight pups from four litters were examined for behavioral
abnormalities, including tests for spontaneous locomotion and passive avoidance. Excised brains
of 21-day-old pups were assayed for biogenic amines (noradrenaline, dopamine, and
5-hydroxytryptamine) and for the activities of Na+,K+-adenosine triphosphatase (ATPase),
monoamine oxidase, and acetylcholinesterase.
Exposure to AN at 5 mg/kg-day had no significant effect on BW or food and water
intakes in dams or on reproductive parameters such as gestational length, number of viable
offspring, or pup sex ratio. No AN-induced effects were observed on postnatal morphological
development, pup BWs, developmental indices (eye opening or incisor eruption), or tests of
neurological impairment (righting reflex, cliff avoidance, and grip strength). The levels of
biogenic amines were significantly altered in specific brain regions as a result of exposure.
169
DRAFT - DO NOT CITE OR QUOTE
-------
Levels of 5-hydroxytryptamine were increased by 32% in the pons medulla and decreased by
44% in the corpus striatum and 30% in the hippocampus. Levels of noradrenaline were
decreased by 40% in the pons medulla and increased by 81% in the hippocampus. Gestational
exposure to AN resulted in a statistically significant 49% reduction in brain levels of monoamine
oxidase in 21-day-old pups; levels of acetylcholinesterase and Na+,K+-ATPase were not
significantly affected. The biological significance of the enzyme and neurotransmitter changes
was unclear, given that no treatment-related behavioral effects were observed. However,
Mehrotra et al. (1988) noted that alterations in the levels of biogenic amines may become more
prominant after prolonged exposure or exposure to higher doses of AN.
In a study comparing developmental toxicities of aliphatic nitriles in vitro and in vivo,
Saillenfait and Sabate (2000) administered a single dose of 0 or 100 mg/kg AN by gavage in
olive oil to groups of four pregnant Sprague-Dawley rats on GD 10. Dams were sacrificed on
GD 12, and the numbers of uterine implantation sites and fetuses with heartbeats were recorded.
Viable fetuses were examined for defects of the allantois, trunk, and misdirected caudal
extremity (left-sided), which the investigators had found to be typical in embryos exposed to
sodium cyanide. Maternal effects of all of the nitriles (including AN) included increases in
clinical signs (piloerection, prostration, and/or tremors) and unspecified maternal BW loss
between GDs 10 and 12. AN exposure had no effect on the numbers of implants/litter or live
embryos/litter, but significantly increased the incidence of overall poor and abnormal
development and the incidence of misdirected allantois or allantois, trunk, and caudal extremity
misdirected (Table 4-44). The study authors suggested that maternal production of the
metabolite cyanide may have contributed to the developmental toxicity of AN since the
characteristic defects that occurred in embryos exposed to cyanide were found in AN-exposed
embryos. The single applied dose of 100 mg/kg is a LOAEL for maternal and fetal effects.
Table 4-44. Morphological alterations in GD 12 fetuses of Sprague-Dawley
rats exposed to 100 mg/kg AN on GD 10
Response
Group (number of litters examined)
Control (n = 4)
100 mg/kg (n = 4)
Implants/litter
13.25 ± 1.71
13.75 ±0.96
Live embryos/litter
13.0± 1.83
12.5 ± 1.0
Total embryos examined
52
50
Allantois, trunk, and/or caudal extremity misdirected
Embryos affected/embryos examined
0/52
13/463
Litters affected/litters examined
0/4
3/4
aSignificantly different from controls (p < 0.01) as calculated by the reviewers (Fisher's exact test).
Source: Saillenfait and Sabate (2000).
170
DRAFT - DO NOT CITE OR QUOTE
-------
4.3.2.1.3. Reproductive toxicity assay. Friedman and Beliles (2002) and Litton Bionetics (1992)
conducted a three-generation reproductive study in Sprague-Dawley rats (groups of 15 males and
30 females) exposed to 100 or 500 ppm AN in drinking water for 100 days before mating.
Groups of 10 male and 20 female controls received untreated drinking water. As calculated by
the study authors, the concentrations were equivalent to average doses of 0, 11, and 37 mg/kg-
day for males and 0, 20, and 40 mg/kg-day for females. The calculated average doses, averaged
across sexes, were 0, 16, and 39 mg/kg-day. Rats were observed daily, especially for signs of
neurotoxicity (e.g., abnormal gait). Water consumption in the F0 generation was measured twice
a week, food intake was measured weekly, and BWs were recorded every 2 weeks. After
100 days of exposure, rats were paired for mating for 6 days; females not bred after 6 days were
mated to another proven breeding male from the same exposure group. The offspring (Fla) of
the first mating were examined on PNDs 0, 4, and 21. Litters were culled to 10 pups on PND 4
to achieve an equal sex ratio. Litter BWs were recorded on PND 4 and individual pup weights
were recorded on PND 21. Two weeks after removal of the Fla offspring, F0 females were
remated to produce the Fib litter, and those not used in breeding Fla were also mated to produce
Fib pups to ensure a sufficient number of offspring for the F2 generation, although these animals
should have been discarded according to the original study design. The original study design
stipulated discarding Fla pups (those not scheduled for breeding) at weaning, but, because of
high mortality in the 500 ppm group, they were retained to ensure sufficient time-mated females
to use as foster mothers (one-half of each high-dose litter was fostered onto untreated females).
All F0 males used for breeding were discarded after the second mating, and those not used in
breeding were discarded when the Fla litter were weaned.
At weaning (21 days), one male and one female from the unfostered Fib litter (or from
the Fla litter, if needed) were selected as potential breeders for the F2 generation. All
F1 offspring exposed to AN were checked daily for mortality, and necropsies were performed on
all animals either found dead or killed in a moribund condition. The F0 dams and females not
producing a litter were exposed for an additional 20 weeks after weaning the Fib pups. After
that period, they were sacrificed and the sciatic nerve, gastrocnemius muscle, brain, and gross
lesions were examined for histopathology. Fla and Fib rats were sacrificed in week 95 of the
study and necropsied for the examination of papillomas in the stomach and intestines. The
protocol for the subsequent generations (Fib parents and F2a and F2b offspring; F2b parents and
F3a and F3b offspring) was as described for the first generation. F2 females, after an additional
20 weeks of exposure following the weaning of the F3b litters, were necropsied, and sciatic
nerve, gastrocnemius muscle, brain, stomach, and gross lesions were evaluated for
histopathology. F3b rats (10/sex) in the control and 500 ppm groups were randomly selected for
histologic examination.
The following discussion of results does not include cancer data from this study, which
were presented separately in Section 4.2.1.2. In F0 parents, BWs were lower than controls after
171
DRAFT - DO NOT CITE OR QUOTE
-------
4 weeks of exposure to 500 ppm, the reduction reaching 22% in males and 17% in females by
week 10. This BW reduction was accompanied reduction in food intake by —18% in 500 ppm
males and 8% in 500 ppm females for the first 10 weeks. Water consumption was reduced in
males and females by about 50% at the high dose and 20-25% at the low dose for the first
10 weeks. Reduced water consumption and possibly BWs may have been a result of reduced
palatability of treated water. Exposure to AN had no effect on the incidence of neurological or
other clinical signs, male or female fertility indices, or the duration of mating or gestation of
F0 parents.
Compared with controls, no significant changes in fertility index or gestational index
were observed in any of the exposed generations (Table 4-45) (Friedman and Beliles, 2002;
Litton Bionetics, 1992). However, poor fertility among controls for the Fib (50-60%) and
F2b (60-70%)) parents might have limited the capability of this study to detect differences in
fertility between control and treated rats (Table 4-45). The durations of mating and gestation
were also unaffected, except the mating duration of F2b rats for F3a generation. Significant
decreased viability and lactation indices were observed in the 500 ppm F0 parents for the
Fla generation, due to the deaths of pups between 1-4 and 5-21 days, respectively. Significant
decreases in viability index were also observed in 100 and 500 ppm F0 parents for the
Fib generation and in 500 ppm F2b parents for the F3a generation (Table 4-45). In addition,
decreases of about 10—40% in pup weights were observed in the Fla, Fib, F2a, F2b, F3a, and
F3b generations in the 500 ppm groups (Table 4-46). Overall, the results identified a drinking
water concentration of 100 ppm (16 mg/kg-day) as a minimal reproductive toxicity LOAEL for
decrease in lactation index in F2a generations. Although no effects on fertility were observed in
this three-generation study, poor fertility among controls in Fib and F2b parents might have
limited the sensitivity of this assessment. The study also identified a LOAEL of 100 ppm for
reduced viability in F0 parents.
172
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-45. Group-specific reproductive indices in three generations of
Sprague-Dawley rats receiving AN in drinking water
Male3'"
Femalecd
Generation/group
Fertility index
Gestation index®
Viability indexf
Lactation index8
F0 parents for Fla
Control
10/10
18/20
18/18
185/186
138/150
100 ppm
8/10
16/20
16/16
197/201
139/150
500 ppm
10/10
16/20
16/16
166/177h
95/143h
F0 parents for Fib
Control
10/10
16/20
16/16
186/186
137/150
100 ppm
10/10
17/20
17/17
182/202h
132/139
500 ppm
13/15
22/28
22/22
99/109h
87/99
Fib parents for F2a
Control
5/10
10/20
10/10
107/109
91/91
100 ppm
7/10
11/20
11/11
116/124
95/104h
500 ppm
8/10
14/20
14/14
133/140
107/114h
Fib parents for F2b
Control
6/10
10/20
10/10
101/101
82/82
100 ppm
5/10
8/20
8/8
93/97
70/73
500 ppm
8/10
14/20
14/14
138/138
123/123
F2b parents for F3a
Control
6/10
14/20
14/14
161/161
128/131
100 ppm
9/10
13/20
13/13
157/158
124/124
500 ppm
10/10
15/20
15/15
157/166h
134/135
F2b parents for F3b
Control
9/10
14/20
14/14
170/176
106/108
100 ppm
10/10
15/20
15/15
198/198
117/119
500 ppm
10/10
17/20
17/17
170/178
115/125
aDoses in males: 0, 11, or 37 mg/kg-d for 0, 100, or 500 ppm as calculated by the study authors.
fertility index in males = number of males producing a litter/number mated.
°Doses in females: 0, 20, or 40 mg/kg-d for 0, 100, or 500 ppm, as calculated by the study authors.
fertility index in females = number of pregnant females/number mated.
"Gestation index = number of litters born/number of pregnant females.
Viability index = number of pups that survived to PND 4/number of pups born alive.
8Lactation index = number of pups surviving to weaning/number of pups alive on PND 4.
hSignificantly lower than controls (p < 0.05), as calculated by the study authors.
Sources: Friedman and Beliles (2002); Litton Bionetics (1992).
173
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-46. Group-specific pup weights in three generations of Sprague-
Dawley rats receiving AN in drinking water
Generation-concentration
Weight (g)
D 4
D 21 (males)
F1 a-control
11
42
Fla-100 ppm
10
40
Fla-500 ppm
9a
28a
Fib-control
10
39
Flb-100 ppm
9
36
Flb-500 ppm
10
34a
F2a-control
11
39
F2a-100 ppm
10
39
F2a-500 ppm
9a
30
F2b-control
11
53
F2b-100 ppm
10
46
F2b-500 ppm
9
30a
F 3 a-control
10
43
F3a-100 ppm
9
43
F3a-500 ppm
8a
30a
F3b-control
10
50
F3b-100 ppm
10
47
F3b-500 ppm
8a
32a
aSignificantly different from controls (p < 0.05), as calculated by the study authors.
Sources: Friedman and Beliles (2002); Litton Bionetics (1992).
4.3.2.1.4. Male exposure reproductive toxicity studies. Tandon et al. (1988) evaluated reproductive
toxicity in male CD-I mice that received 0, 1, or 10 mg/kg-day AN by gavage in saline for 60
days. The testes of six mice/group were examined histopathologically, and homogenates of
pooled testes (four testes) in each group were assayed for the activities of sorbitol dehydrogenase
(SDH), acid phosphatase, LDH, glucose-6-phosphatase dehydrogenase, and P-glucuronidase.
Exposure to AN decreased the epididymal sperm count by 21 and 45% at the low- and
high-dose group, respectively. However, the decrease was statistically significant (p < 0.05)
only with the 10 mg/kg-day group. Histopathological examination of the testes did not reveal
changes in the low-dose group. In high-dose mice, degenerative changes were seen in 40% of
seminiferous tubules. In addition, the testes of mice dosed with 10 mg/kg-day showed a 12%
increase in the activity of LDH, a 22% decrease in the activity of SDH, a 37% increase in the
activity of P-glucuronidase, and a 16% decrease in the activity of acid phosphatase compared
with controls. AN exposure had no effect on the activity of testicular glucose-6-phosphatase
dehydrogenase. The study authors suggested that the changes in the activities of LDH and SDH
were related to the AN-induced degeneration of germinal epithelium. In this study, a NOAEL of
174
DRAFT - DO NOT CITE OR QUOTE
-------
1 mg/kg-day and a LOAEL of 10 mg/kg-day were identified for the toxicological effects of AN
on the testes of male CD-I mice.
In a range-finding acute toxicity study for a dominant lethal assay, groups of 10 male
F344 rats received 45, 60, 68, 75, or 90 mg/kg-day AN by gavage in 0.9% saline daily for 5 days
(Working et al., 1987). Rats were observed for a total of 42 days from the first administration.
No deaths were observed in groups receiving 45 or 60 mg/kg-day. In the higher dose groups, the
study was terminated early on account of mortality: 40% at 68 mg/kg-day (terminated on day 6),
30%) at 75 mg/kg-day (terminated on day 4), and 30% at 90 mg/kg-day (terminated on day 2).
As a result of this range-finding study, 60 mg/kg-day was selected as the maximum tolerated
dose for the dominant lethal assay. In this assay, groups of 50 male F344 rats received AN by
gavage at 0 or 60 mg/kg-day in 0.9% saline for 5 days. BWs of males were recorded three times
during the week of treatment and then weekly during the mating period of 10 weeks; males were
caged with a different female each week for 6 days. A transient reduction in mean BW (-4%)
compared with controls was observed in rats on treatment days 3 and 5 and on the fourth post-
treatment day; BW gain in treated rats was equivalent to controls beginning the second week of
observation. AN exposure in males had no effect on the incidence of pre- or postimplantation
losses, indicating a negative result in the dominant lethal assay. AN also had no effect on the
fertility of exposed males in any postexposure week.
4.3.2.2. Inhalation Exposure
Information about reproductive/developmental toxicity in animals exposed to AN by
inhalation comes from a standard chronic toxicity assay in rats, developmental toxicity assays in
rats, and a dominant lethal assay in male mice.
4.3.2.2.1. Standard toxicity assays. As described in Section 4.2.2.2, no reproductive
histopathology was observed in male or female Sprague-Dawley rats that were exposed to AN at
concentrations as high as 80 ppm 6 hours/day, 5 days/week for up to 2 years (Dow Chemical
Co., 1992a; Quastetal., 1980b).
4.3.2.2.2. Developmental toxicity assays. Haskell Laboratory (1992a) and Murray et al. (1978)
evaluated developmental toxicity in groups of 30 pregnant Sprague-Dawley rats exposed (whole
body) to 0, 40, or 80 ppm AN vapor for 6 hours/day on GDs 6-15. The parameters examined
were the same as those for the oral exposure study by these authors described in Section 4.3.2.1.
Inhalation exposure to 40 or 80 ppm AN did not result in deaths, changes in appearance, gastric
thickening, or increase in terminal liver weight in dams. Maternal BW gain was significantly
reduced during GDs 6-15 by 47% in the 40 ppm group and by 58% in the 80 ppm group
(a reduction of about 20% in both groups for GDs 6-21). Both exposure groups exhibited
significant (unspecified) decreases in food consumption on GDs 6-9 (but not later intervals) and
175
DRAFT - DO NOT CITE OR QUOTE
-------
increases in water consumption on GDs 9-20. In this study, a NOAEL for maternal effects was
not identified, but a LOAEL of 40 ppm was identified for reduced BW gain in dams.
Gestational exposure to AN had no significant effect on any of the reproductive
parameters (pregnancy rates, numbers of implantations, live fetuses, or resorptions) and no effect
on fetal BW or crown-rump length measurements. Furthermore, no single major malformation
occurred at significantly higher incidence in AN-exposed rats vs. controls. However, as shown
in Table 4-47, there was an increase in the incidence of total major malformations when they
were considered collectively (p < 0.06) for the high-dose group (present in 6/35 litters).
Malformations observed in litters of 80 ppm group included short tail, missing vertebrae, short
trunk, omphalocele, and hemivertebra. In this study, a NOAEL of 40 ppm and a LOAEL of
80 ppm were identified for increases in total malformations in rats.
Table 4-47. Incidence of fetal malformations among litters of Sprague-
Dawley rats exposed to AN by inhalation
Type of malformation
AN concentration (ppm)
0
40
80
Number of fetuses/number of litters examined
External and skeletal malformations
421/33
441/36
406/35
Visceral malformations
140/33
148/36
136/35
Number of fetuses (litters) affected
External malformations
Short tail
0(0)
0(0)
2(2)
Short trunk
0(0)
0(0)
1 (1)
Imperforate anus
0(0)
0(0)
0(0)
Omphalocele
0(0)
1 (1)
1(1)
Visceral abnormalities
Right-sided aortic arch
0(0)
0(0)
0(0)
Missing kidney, unilateral
0(0)
0(0)
0(0)
Anteriorly displaced ovaries
0(0)
0(0)
1(1)
Skeletal malformations
Missing vertebrae
0(0)
0(0)
2(2)
Missing two vertebrae and two ribs
8(1)
2(1)
7(2)
Hemivertebrae
0(0)
0(0)
1(1)
Total malformed
8(1)
3(2)
11 (6)a
sub chronic
"Significantly different from control, as calculated by the study authors (p = 0.06).
Source: Murray et al. (1978).
Saillenfait et al. (1993) included AN in a survey of the relative developmental toxicities
of inhaled aliphatic mononitriles in rats. Pregnant Sprague-Dawley rats (20-21/group) were
exposed (whole body) for 6 hours/day to 0, 12, 25, 50, and 100 ppm AN vapor on GDs 6-20.
176
DRAFT - DO NOT CITE OR QUOTE
-------
Dams were observed daily throughout pregnancy and BWs were recorded on GDs 0, 6, and 21.
All subjects were sacrificed on GD 21, and the uteri were weighed and opened to assess the
numbers of implantations, resorption sites, and live and dead fetuses. The fetuses were
examined for external abnormalities and then split into two equal groups for examination of
skeletal or visceral anomalies.
AN exposure did not cause premature deaths in the dams, but exposure to >25 ppm
caused concentration-related, statistically significant reductions (by 16-45%) in overall BW gain
and concentration-related losses in absolute BW (exclusive of gravid uterus weight) in dams. In
this study, 12 ppm is a NOAEL and 25 ppm is a LOAEL for reduced BW in dams.
AN exposure had no effect on reproductive parameters (pregnancy rate, average number
of implantations, numbers of live fetuses, incidences of nonsurviving implants, or resorptions per
litter). Concentration-dependent, statistically significant reductions in average fetal weight (by
5-15% compared with controls) were observed at >25 ppm. There were no significant increases
in the incidences of external, visceral, or skeletal anomalies in the exposed groups and one
control fetus. In this study, a NOAEL of 12 ppm and a LOAEL of 25 ppm were identified for
significantly reduced fetal BW.
In a dominant lethal assay for AN, Zhurkov et al. (1983) continuously exposed male ICR
"3
mice (20/group, whole body) to AN at concentrations of 0, 20, or 100 mg/m (0, 9.1, or 46 ppm)
for 5 days. After exposure, each male was mated to two unexposed females for 8 weeks.
Females were sacrificed between GDs 13 and 15, and a number of reproductive and
developmental parameters were monitored, including the percentage of pregnant females,
number of corpora lutea, implantations, and live and dead fetuses per female as well as total and
pre- and postimplantation mortality. Exposure to AN did not result in any dominant lethal effect
on male germ cells nor did it cause adverse pre- or postimplantation outcomes. The highest
exposure level in this study, 46 ppm, was a NOAEL for reproductive toxicity. This study was
limited by sparse descriptions of methods and a lack of quantitative reporting of results.
4.3.2.3. Intraperitoneal Administration
In a translated study in Chinese, the effect of AN on spermatogenesis was examined in
groups of male Kunming mice (10 3-4-week-old pubertal and five 6-8-week-old adult mice per
group) treated by i.p. injection (Liu et al., 2004). The animals received AN (>99% purity) at
doses of 1.25, 2.5, or 5.0 mg/kg-day (1/24, 1/12, or 1/6 of the LD50) for 5 days. Negative
controls received physiological saline (10 mL/kg) daily for 5 days, whereas positive controls
received a single injection of 40 mg/kg CP (not identified but presumably cyclophosphamide).
On day 35, mice were sacrificed, the left testicle was selected from five mice per group, and
testicular cell suspensions were evaluated using flow cytometry.
The following dose-related effects observed in the 2.5 and 5.0 mg/kg-day groups were
statistically significantly different from the negative control group (p < 0.05); the same effects
177
DRAFT - DO NOT CITE OR QUOTE
-------
were observed in positive controls (Liu et al., 2004). AN decreased the percentage of haploid
testicular cells (indicative of completed spermatogenic meiosis) by 14.3 and 15% in mid- and
high-dose adult mice, but a slight reduction was not statistically significant in pubertal mice.
The percentage of apoptotic testicular cells was significantly increased by 58.7 and 74.8% in
mid- and high-dose pubertal mice and by 81.5 and 108% in mid- and high-dose adult mice. The
percentages of spermatogenic epithelial cells in Go/Gi phase were significantly reduced by
18 and 20% in mid- and high-dose pubertal mice and by 35.5 and 40.5% in mid- and high-dose
adult mice. The percentage of spermatogenic epithelial cells in G2/M phase was significantly
elevated in adult mice by 31.4 and 32.8% at the mid- and high doses, respectively. AN exposure
had no effect on the percentage of spermatogenic epithelial cells in S phase. The NOAEL and
LOAEL for suppression of spermatogenesis were 1.25 and 2.5 mg/kg-day, respectively.
The teratogenic effects of AN were investigated in groups of pregnant golden hamsters
that were exposed via an i.p. injection on GD 8 (Willhite et al., 1981). The actual numbers of
dams per group were not reported, but data were presented for 3-6 litters in the exposed groups
and 12 litters for the controls injected with sodium chloride. Dose levels were 0, 0.09, 0.19,
0.47, 1.23, and 1.51 mmol/kg (equivalent to 0, 5, 10, 25, 65, and 80 mg/kg). No adverse clinical
symptoms were observed in the dams exposed to up to 1.23 mmol/kg (65 mg/kg) AN by GD 14,
when the dams were sacrificed. No malformations were observed in the offspring. In contrast,
dams exposed to 1.51 mmol/kg (80 mg/kg) showed intense dyspnea, gasping, incoordination,
hypothermia, salivation, and convulsions for 1-5 hours after injection. Additionally, several
fetal abnormalities and malformations were observed at this exposure level, including
encephaloceles and fused or bifurcated ribs. Coadministration of 8.06 mmol/kg STS and
1.51 mmol/kg AN induced neither maternal toxicity nor fetal malformations. Coadministration
of 8.06 mmol/kg STS and a higher dose of AN (1.88 mmol/kg, 100 mg/kg) prevented toxic
symptoms in dams but not the teratological effects in fetuses. The study authors concluded that
the teratogenic action of AN was related to the metabolic release of cyanide.
The effect of AN on rat liver cytochrome P-450 and serum hormone levels were studied
in male Sprague-Dawley rats (4/group) injected with 33 mg/kg AN (i.p.) for 3 consecutive days
(Nilsen et al., 1980). Control animals were treated with 0.9% sodium chloride. Blood were
collected immediately after the animals were sacrificed and serum luteinizing hormone (LH),
follicle stimulating hormone (FSH), and prolactine (PRL) were measured by radioimmunoassay.
Serum corticosterone levels and liver microsomal cytochrome P-450 were also determined.
A significant increase (3%) in BW was reported in the AN treated group when compared
with controls (Nilsen et al., 1980), but relative liver weight was not increased. A significant
decrease in liver microsomal content of CYP450 was observed. Serum corticosterone measured
24 hours after the last AN injection was decreased by 70%, while FSH levels were increased by
118%). Serum PRL levels were decreased by 63% when compared with the controls, while
178
DRAFT - DO NOT CITE OR QUOTE
-------
serum LH levels were unchanged. Nilsen et al. (1980) suggested that the increase in FSH might
be secondary to impaired spermatogenesis in the testes of treated rats.
4.3.3. In Vitro Studies
The embryotoxicity of AN was evaluated in cultures of whole rat embryos by Saillenfait
and coworkers in a series of studies (Saillenfait et al., 2004, 1992; Saillenfait and Sabate, 2000).
As described initially in Saillenfait et al. (1992), the experimental system involved incubating
day 10 embryos from pregnant Sprague-Dawley rats for 26 hours in whole organ culture in a
medium containing AN at concentrations of 76-760 [xmol/L. The growth-related parameters
evaluated were functional yolk-sac circulation, yolk-sac diameter, crown-rump length, head
length, number of somites, number of malformed embryos, incidences of abnormal brain,
malformed caudal extremities, delayed yolk-sac circulation, and defective flexion. The effects of
metabolic activation on the embryotoxicity of AN were evaluated by the inclusion of S9 and
cofactors (NADPH, glucose-6-phosphate) for CYP450-dependent biotransformation in the
incubation system.
Exposure to AN induced concentration-related effects on growth and development.
Functional yolk-sac circulation (circulating erythrocytes) was reduced at >304 |iM and was
completely absent at 760 [xM. Crown-rump length was reduced at 304 |iM and could not be
measured at higher concentrations. A concentration-dependent increase in the incidence of
malformations was observed following in vitro exposure of day 10 fetuses to AN at 152 |iM and
above. AN at 152 and 304 [xM induced malformations in 53 and 100% of the exposed embryos,
respectively. Malformations primarily consisted of a shortened caudal extremity (significant at
152 |iM) and a reduction of the brain (achieving significance at 304 [xM). Other general
malformations were delayed development of the yolk-sac circulation and defective flexion.
Furthermore, growth retardation and severity of malformations induced by 304 |iM AN were
enhanced by the presence of S9 and cofactors, suggesting a role for oxidative biotransformation
in AN embryotoxicity. Addition of 0.1-2.2 mM GSH in the incubation medium reduced the
embryotoxic effects of AN in this system.
Similarly, a later experiment by this research group (Saillenfait and Sabate, 2000)
demonstrated that day 10 rat embryos exposed to AN for 46 hours in vitro showed concentration-
related effects on growth, development, and morphology. Reduced growth (reduced yolk sac
diameter and crown-rump length) occurred at 100 [jM. The head length and somite number were
reduced at 125 [xM. Abnormal development (reduced prosencephalon and mesencephalon) and
maxillary process defects occurred at >125 [xM AN. Defects of the rhombencephalon and the
auditory system were observed less frequently. These effects were characteristic of those
developed by embryos exposed to sodium cyanide in culture. Addition of microsomes and
NADPH to the culture medium containing 150 or 175 [xM AN enhanced the observed growth
179
DRAFT - DO NOT CITE OR QUOTE
-------
retarding and dysmorphogenic effects, especially in increasing the incidence of
rhombencephalon and auditory system defects.
Saillenfait et al. (2004) evaluated the effects of eight aliphatic nitriles on the viability and
differentiation of cultured limb bud cells from Sprague-Dawley rat embryos on GD 13. Limb
bud micromass cultures were exposed for 5 days to AN at concentrations between 0.01 and
0.45 mM (0.5 and 23.9 (j.g/mL), with or without microsomal activation, after which they were
evaluated for cytotoxicity (neutral red uptake assay) and differentiation of chondrocytes (number
and total surface of foci with Alcian blue staining were used as indicator of cell differentiation).
The concentrations that inhibited cell viability and cell differentiation by 50% of concurrent
untreated controls were determined.
The IC50 values were 0.24 mM (13 (j,g/mL) for cytotoxicity (viability) and 0.33-0.38 mM
(18-20 (j,g/mL) for differentiation (total surface of foci and number of foci), respectively. The
ratios of IC50 for cytotoxicity and differentiation were 0.6 for number of foci and 0.7 for total
surface of foci. Microsomal activation had no effect on the results with AN. In parallel
experiments, the IC50 for cytotoxicity in cultured 3T3 cells (differentiated mouse fibroblast cell
line) exposed to AN was 0.065 mM. The relative potency of the tested nitriles in this limb bud
cell culture system matched previously published results for cultured whole embryos, but not
necessarily for in vivo teratogenicity (false negative results were obtained for two of the eight
nitriles). Saillenfait et al. (2004) characterized the response of AN in the micromass culture
assay as equivocal, since it depended on the criteria used to define a positive result. According
to one criterion, AN might be considered to have teratogenic potency because its IC50 was less
than 50 [j,g/mL. However, under the "twofold rule" that defines a positive result by a value
>2 for the ratio (IC50 for cytotoxicity)/(ICso for differentiation), AN would be classified as
having poor potential developmental hazard. As suggested by comparison of the IC50 for
cytotoxicity, embryonic rat limb bud cells were not more vulnerable to AN than differentiated
3T3 mouse fibroblasts. These suggestive results are consistent with the observations in vivo (see
Section 4.3.2) that fetal toxicity from AN occurs only at exposure levels that cause maternal
toxicity.
4.4. OTHER DURATION- OR ENDPOINT-SPECIFIC STUDIES
4.4.1. Acute Toxicity Data
Closely similar values have been reported for the oral LD50 in rats: 78 mg/kg (Benesh
and Cherna, 1959), 93 mg/kg (Smyth and Carpenter, 1948), 90 mg/kg (Sprague-Dawley rats)
(Younger Labs, 1992), and 81 mg/kg (male CF Nelson rats) (Vernon et al., 1990). Lower value
has been published for the oral LD50 in mice: 25 mg/kg (H-strain, sex not stated) (Benesh and
Cherna, 1959) and 36 and 48 mg/kg in male and female mice, respectively (strain not stated)
(Tullar, 1947, as cited in IPCS, 1983). These findings indicate that mice are more susceptible to
the acute toxicity of AN than are rats. Published LD50 values for guinea pigs (56 mg/kg)
180
DRAFT - DO NOT CITE OR QUOTE
-------
(Jedlicka et al., 1958) and rabbits (93 mg/kg) (Paulet and Desnos, 1961) are in the same narrow
range as those for rats and mice. For the inhalation route, median lethal concentration (LC50)
values in rats of 217 ppm (470 mg/m3) (Knobloch et al., 1971) and 333 ppm (777 mg/m3)
(Haskell Laboratory, 1992b) have been reported. However, exposing Sprague-Dawley rats to
1,008 ppm AN for 1 hour failed to produce mortality (Younger Labs, 1992). A 4-hour
LC50 value of 946 ppm (2,053 mg/m ) was reported for Sprague-Dawley rats following nose-
only exposure (WIL Research Laboratories, 2005); no mortality was observed in male or female
"3
rats at exposures as high as 775 ppm (1,682 mg/m ) in this study. However, ataxia, labored
respiration, and hypoactivity were observed at this exposure concentration and higher.
3 3
LC50 values (duration not reported) of 138 ppm (300 mg/m ) in mice and 456 ppm (990 mg/m )
in guinea pigs have been reported (Knobloch et al., 1971).
As set forth in the English abstract of an article in Polish, Knobloch et al. (1971)
investigated the acute and subacute toxicity of AN in BN mice, Wistar rats, and guinea pigs
(strain not given). For rats, s.c. and i.p. LD50 values of 80 and 100 mg/kg, respectively, were
reported. For mice, the LD50 was 34 mg/kg.
An earlier subacute study examined the effect of oral AN administration on the liver of
sodium PB-pretreated (400 (j.mol/kg) or Aroclor 1254-pretreated (300 (j,mol/kg) male and female
Sprague-Dawley rats that received 100 or 500 ppm AN in drinking water for 21 days (Silver et
al., 1982). Other pretreated animals were given 0, 50, 75, 100, or 150 mg/kg AN by gavage for
up to 3 days. Some liver-related biochemical changes were noted as a result of AN exposure,
including a dose-dependent reduction in hepatic nonprotein sulfhydryl concentrations (maximal
reduction of 81% at the highest dose of 150 mg/kg). Serum SDH activity was increased by
fourfold 24 hours after administration of 150 mg/kg AN. On the other hand, SGPT activity was
not significantly altered. Pretreatment with PB or Aroclor 1254 resulted in only a slight
enhancement of AN-induced elevation of serum SDH or SGPT activities. In the experiment in
which female rats were pretreated with either vehicle, PB or Aroclor 1254, and treated with
100 ppm or 500 ppm AN for 21 days, there was a 60% increase in serum SDH activity in
animals receiving 500 ppm AN in drinking water.
In the 3-day studies, there was some evidence of focal superficial necrosis of the liver in
rats receiving high gavage doses (100 or 150 mg/kg). This effect was associated with the
presence of hemorrhagic gastritis and distention of the forestomach. However, while the
biochemical and pathology changes implied a limited perturbation of the liver by AN, light
microscopy showed only minor changes in the histopathology of the organ and no ultrastructural
changes to the liver were evident when evaluated by electron microscopy. Silver et al. (1982)
concluded that liver did not appear to be a target organ in the acute or subacute toxicity of AN.
Dudley andNeal (1942) exposed several species of laboratory animals—rat, guinea pig,
rabbit, cat, dog, rhesus monkey—to AN vapor for 0.5-8 hours at concentrations ranging from
0.063 to 5.3 mg/L (30-2,445 ppm). In rats (Osborne Mendel, sex not stated), 1,260 ppm was
181
DRAFT - DO NOT CITE OR QUOTE
-------
found to be an effective lethal concentration when administered for 4 hours. During or after an
8-hour exposure, 320 ppm AN was fatal. A 4-hour exposure to 260 ppm AN was fatal in rabbits
(strain and sex not stated), while 1,160 ppm appeared to be a fatal concentration in guinea pigs.
Other lethal concentrations were 600 ppm for 1.5 hours in cats and 110-165 ppm for 3 hours in
dogs (breed not given, male and female), while Rhesus monkeys (males and females) tolerated
90 ppm AN with only minor, transitory adverse effects (skin redness, sleepiness). In all species
except guinea pig, AN caused an initial respiratory stimulation, followed by rapid, shallow
breathing, nasal exudate, and watering eyes. In rats, the most striking symptom was reddening
of the skin in the less prominently haired regions (nose, ear, feet). Reddening of the skin was
also observed in the other species, again with the exception of guinea pigs. The study authors
concluded that, in five of the six species investigated, AN-induced symptoms resembled cyanide
poisoning. By contrast, AN acted as a severe pulmonary irritant in guinea pigs. Consistent with
these findings, sodium nitrite (a cyanide antidote) protected five of the species studied, but not
guinea pigs, from AN toxicity. The study authors postulated that guinea pigs may metabolize
AN differently than rats, rabbits, cats, dogs, or monkeys, possibly by transformation to acrolein
rather than cleavage of the cyano group. However, Dudley and Neal (1942) were unable to
extract any cyanide from the tissues of animals that had succumbed to AN exposure.
Knobloch et al. (1971) evaluated the acute toxicity of AN via inhalation exposure
(duration not given) in three species. The LC50 in mice was 0.30 mg/L (138 ppm), in rats
0.47 mg/L (217 ppm), and in guinea pigs 0.99 mg/L (456 ppm). These values confirmed the
roughly twofold species difference between rats and guinea pigs, but were about threefold lower
than values reported by Dudley and Neal (1942).
Two studies by Gut et al. (1985, 1984) evaluated the subchronic toxicity of AN in rats
-3
when administered via inhalation. In both studies, male Wistar rats were exposed to 280 mg/m
AN by inhalation 8 hours/day for 5 days (Gut et al., 1985, 1984). As shown in Table 4-48, BW
was decreased in rats exposed to AN for 5 days. The absolute weight of liver decreased, while
brain weight remained unchanged. Hence, the relative liver weight was significantly decreased,
while relative brain weight increased due to the BW decrease. However, there were no
significant histopathological changes in the lungs, livers, kidneys, or adrenals. The
concentrations of glucose, pyruvate, and lactate were elevated in the brain and blood of exposed
rats (Table 4-48). Treatment-related reductions in the concentrations of serum cholesterol and
triglycerides were found in test animals. The study authors considered the increase in blood
glucose concentration to be the most sensitive indicator of AN exposure under these
experimental conditions (Gut et al., 1984).
182
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-48. Effects of AN on organ weight, clinical chemistry, and
biochemical parameters when administered to male Wistar rats via
inhalation
Parameter (units)
Controls (n)
AN (n)
BW (g)
341 ± 18(8)
287 ± 24a (8)
Relative liver weight (g/100 g BW)
3.39 ±0.29 (8)
2.79 ± 0.08a (8)
Relative brain weight (g/100 g BW)
0.564 ±0.06 (8)
0.647 ± 0.076a (8)
Serum triglyceride (mmol/L)
2.36 ±0.51 (10)
1.47 ± 0.47b (10)
Serum cholesterol (mg/dL)
72.3 ±8.5 (10)
54.6 ± 12.9b (10)
Pyruvate (mmol/L)
Blood
0.080 ±0.012
0.128 ± 0.028a (5)
Brain
0.039 ±0.007
0.068 ±0.014a (5)
Lactate (mmol/L)
Blood
2.259 ±0.577 (5)
4.032 ± 1.061a (5)
Brain
6.002 ±0.258 (5)
8.034 ± 0.4713 (5)
Glucose (mmol/L)
4.10 ± 0.21
10.22 ± 1.26° (5)
Significantly different from controls (p < 0.05).
bSignificantly different from controls (p < 0.001).
Significantly different from controls (p < 0.01).
Source: Gut et al. (1984).
Gut et al. (1984) further examined the dose-response effects of AN on blood glucose
levels by giving food-deprived male Wistar rats a single 12-hour exposure of 0, 57, 125, or
271 mg/m AN. Blood glucose levels were dose-dependently increased at the end of the
exposure period (4.33 ± 0.64 at 57 mg/m3; 10.88 ± 3.74 at 271 mg/m3 vs. 3.46 ± 0.29 mg/m3 in
controls) but declined to normal or below normal levels 24 hours after exposure
3 3
(1.79 ± 0.84 at 271 mg/m vs. 4.73 ± 0.34 mg/m in controls). The simultaneous increase in the
pyruvate and lactate concentrations suggests that AN affected carbohydrate metabolism on the
level of glycolysis as well as on the citric acid cycle level.
In a subsequent study using exposure regimens identical to those in Gut et al. (1984), Gut
et al. (1985) monitored AN-induced changes in sulfhydryl concentrations in the liver and brain
and the appearance of thioethers (AN-mercapturic acid) and thiocyanate in the urine. GSH
concentrations in the liver were significantly reduced compared with controls (3.86 ± 0.41 vs.
7.66 ± 0.58 (j,mol/g) in rats exposed to AN for five consecutive daily 8-hour exposures at
"3
280 mg/m . No GSH depletion was evident in the brain. The protein sulfhydryl level remained
unchanged in the brain but was increased by 17% in the liver, although not statistically
significant. One out of 15 treated animals died, and the study authors suggested that GSH
depletion could have been responsible.
183
DRAFT - DO NOT CITE OR QUOTE
-------
Urinary excretion of thioethers and thiocyanate was proportional to the inhaled
concentration. The ratio between urinary thioethers and thiocyanate in exposed rats was about
2:4 and was not influenced by the exposed AN concentrations (Gut et al., 1985).
Bhooma et al. (1992) evaluated the effect of AN on the procoagulant activity in
pulmonary alveolar macrophages of rats. Six male Wistar rats/group were exposed to 100 ppm
AN 5 hours/day for 5 days. Animals were sacrificed at 1-28 days after the last exposure. The
lungs were lavaged, and alveolar macrophages were collected from the broncho alveolar lavage
(BAL) fluid. Procoagulant activity (the ability of a cell or cell products to accelerate the
conversion of fibrinogen to fibrin) in macrophage and BAL fluid was determined. The
procoagulant activity in the isolated macrophages from exposed animals was about 10-fold
higher than in controls 1 day after AN exposure and then slowly declined to control levels by
day 28 after exposure. Procoagulant activity of BAL fluid remained unaltered in rats sacrificed
up to 7 days after exposure but was elevated in those sacrificed 14 and 28 days after exposure.
The study authors noted that acute lung injury often resulted in deposition of fibrin in alveolar
spaces (Bachofen and Weibel, 1977) and that fibrin and its degradation products have been
implicated in contributing to pulmonary inflammation (Malik et al., 1979; Cuterman et al.,
1977). Since this study showed macrophage-associated procoagulant activity in the lung
following inhalation to AN, the study authors suggested the alveolar macrophages participated in
the pulmonary deposition of fibrin.
Other experimental studies have used acute dosing protocols to study toxicological
effects of AN on various target organs. These studies are summarized in the following sections.
4.4.1.1. Effects of AN on the GI Tract
Two studies have reported GI hemorrhage and gastric erosion in rats administered with
single doses of AN. In the first report (Ghanayem and Ahmed, 1983), a single dose of 50 mg/kg
AN was administered to Sprague-Dawley rats (orally or s.c.). GI bleeding was observed 3 hours
after treatment, with no significant difference in the amount of GI blood loss resulting from s.c.
or oral administration. Thus, AN-induced GI bleeding was not a result of direct irritation of AN
on the GI tract. A time-course study of a single 50 mg/kg dose of AN administered by the s.c.
route indicated that the amount of blood recovered from the stomach was significantly higher
than that of controls at 1, 2, and 3 hours after treatment, while the amount recovered from the
intestinal contents was significantly higher than in controls at 2 and 3 hours. Dose-response
study of s.c. administration of AN and GI bleeding indicated that significantly higher GI
bleeding than in controls occurred at 40, 50, and 70 mg/kg AN, with the maximum at 50 mg/kg
AN.
Pretreatment of rats with the CYP450 enzyme inducer, PB, decreased the AN-induced GI
blood loss by 55%, whereas pretreatment with Aroclor 1254 increased blood loss by 240%
(Ghanayem and Ahmed, 1983). In contrast, pretreatment of rats with CYP450 inhibitors, cobalt
184
DRAFT - DO NOT CITE OR QUOTE
-------
chloride or SKF 525A, prior to AN administration produced significant decreases in blood loss
of 10 and 40%, respectively. Pretreatment of rats with diethylmaleate (DEM), a known depletor
of GSH, prior to AN administration produced no significant change in GI bleeding. In addition,
s.c. administration of a sublethal dose (6 mg/kg) of KCN did not induce GI bleeding when
compared with controls, whereas 50 mg/kg AN produced significant GI bleeding. The study
authors concluded that metabolic activation of AN to a reactive metabolite other than cyanide
(probably CEO) by CYP450 was a prerequisite for AN to induce gastric hemorrhage.
In the second paper, Ghanayem et al. (1985) studied the mechanism of AN-induced
gastric mucosal necrosis in the glandular stomach. Male Sprague-Dawley rats were treated with
a single s.c. dose of AN (50 or 30 mg/kg), and the glandular stomach was removed and evaluated
for histopathology and GSH concentrations. The liver was also removed for GSH determination.
Gastric erosion severity index (GEI) was obtained for each exposure group by multiplying the
mean severity score by the incidence of gastric necrosis in the group. Calculated GEI was found
to be dose and time dependent: higher at 50 mg/kg than 30 mg/kg AN 3 hours after
administration and in rats killed 3 hours as compared with rats killed 1 hour after the same dose
of AN.
Subcutaneous administration of 30, 40, or 50 mg/kg AN also caused a significant
decrease in hepatic GSH concentration 3 hours after treatment, with greater decrease at lower
dose (30 mg/kg) than in high dose (50 mg/kg). A significant decrease in gastric GSH
concentrations was observed 3 hours after treatment at 40 and 50 mg/kg AN. Pretreatment of
rats with various metabolic modulators (CYP450 monooxygenase and GSH) before
administration showed that there was a significant inverse relationship between gastric GSH
concentration and AN-induced gastric erosions. P450 inducers (Na PB and Aroclor 1254) alone
increased GSH levels in the liver. Pretreatment of rats with these P450 inducers inhibited
AN-induced gastric necrosis, and partially blocked AN-induced gastric GSH depletion. SKF
525A, a CYP450 inhibitor, caused a slight depletion of gastric and hepatic GSH and potentiated
the AN-induced gastric necrosis and GSH depletion. In contrast, cobaltous chloride, another
inhibitor of CYP450 enzyme, inhibited AN-induced gastric necrosis and increased both the
hepatic and gastric GSH concentrations. In rats treated with DEM, a depletor of GSH and AN,
or DEM + cysteamine + AN, GEIs were increased up to fivefold when compared with rats
treated with AN alone. Mucosal erosion severity in these two groups was also greatly increased,
with more than 50% of rats showing severe or extensive lesions. Pretreatment of rats with
sulfhydryl-containing compounds (cysteine or cysteamine) protected against AN-induced gastric
necrosis and blocked the depletion of gastric GSH.
In addition, AN-induced gastric erosions could be prevented by pretreatment with
atropine, a muscarinic receptor blocker, suggesting the involvement of muscarinic receptors in
the AN-induced gastric mucosal necrosis (Ghanayem et al., 1985). Activation of acetylcholine
muscarinic receptors is known to increase gastric acid secretion and cause gastric erosions.
185
DRAFT - DO NOT CITE OR QUOTE
-------
Because muscarinic receptors are known to contain sulfhydryl groups in their active site (Ikeda
et al., 1980; Aronstam et al., 1978), Ghanayem et al. (1985) hypothesized that AN inactivated
critical sulfhydryl groups and caused gastric erosions by locally modulating muscarinic
acetylcholine receptors in the stomach.
More recently, Ahmed et al. (1996a) showed accumulation of AN-derived radioactivity
in intestinal contents and intestinal mucosa following i.v. injection of 2-[l4C]-AN to rats. A
recent study by Jacob and Ahmed (2003a) also demonstrated that AN and or its metabolites
accumulated and covalently interacted in GI mucosa of male F344 rats treated either i.v. or orally
with 2-[14C]-AN. These studies supported the hypothesis that AN-induced injury of the GI
mucosa is not due to direct irritation by AN but by metabolic incorporation and macromolecular
interaction of AN in these tissues.
4.4.1.2. Effects of AN on the Kidney
The acute nephrotoxic effect of AN was investigated in single exposure studies in rats
and hamsters by inhalation or i.p. administration. Intraperitoneal injection of Chinese hamsters
with 30 mg/kg AN increased kidney weight and renal GSH concentration 24 hours after injection
(Zitting et al., 1981). Kidney deethylation activity was also decreased. Rouisse et al. (1986)
administered i.p. doses (0, 10, 20, 40, 60, or 80 mg/kg) of AN to male F344 rats (six/dose group).
Urinary volume was increased two- to threefold during the 24-hour period following
administration for all dose groups. Urinary glucose was about 6 times higher in the 20 mg/kg
group than in controls, and 40-60 times higher in the higher dose groups. Urinary excretion of
N-acetyl-P-D-glucosaminidase was increased at the highest doses, up to 80% over controls in the
80 mg/kg group. An increased number of lysosomes or dense bodies in renal proximal tubules
was seen under the light or electron microscope. In the inhalation study, a similar array of
toxicological and clinical chemistry effects relating to kidney structure and function was
observed in male rats (seven/group) exposed to 200 ppm AN for 4 hours (Rouisse et al., 1986).
4.4.1.3. Effects of AN on the Adrenal Gland
An experimental model for the toxicity of AN investigated the capacity of single i.v.
doses of AN to induce acute hemorrhagic necrosis of the adrenal gland (Szabo et al., 1976). This
condition has a parallel in man, the Waterhouse-Friderichsen syndrome, which is characterized
by thrombocytopenia, disseminated intravascular coagulation, and the appearance of fibrin
degradation products in the circulation (Szabo et al., 1976). The condition has been induced in
female Sprague-Dawley rats by a single injection of 150 mg/kg AN into the jugular vein (Szabo
et al., 1980) and typically has been marked by massive "bilateral apoplexy," a usually fatal
hemorrhagic necrosis of the adrenal glands that occurs in 90-100% of the rats within 90-
150 minutes. Thrombocytopenia and a range of associated clinical signs, including tremors,
cyanosis, and ultimately respiratory failure, were observed. Light and electron microscopy
186
DRAFT - DO NOT CITE OR QUOTE
-------
showed that an early event in the onset of this condition was damage to the vascular endothelium
of the adrenal cortex, with parenchymal injury as a late event. However, these AN-induced
lesions could be prevented by pretreatment with the a-adrenergic antagonist phenoxybenzamine,
the a,P-blocker labetalo, or the 11-P-hydoxylase inhibitor metyrapone. Elevation of tissue
sulfhydryl levels by cysteine or GSH reduced the adrenal apoplexy. Dopamine concentrations in
the adrenals increased over time. Because depletion of catecholamines by reserpine, or
medullectomy, could prevent the chemically induced adrenocortical necrosis, the study authors
proposed that cortical damage resulting from AN was associated with vasoactive amines released
from the medulla and/or with metabolites of AN.
4.4.1.4. Effects of AN on Neurological Endpoints
Neurotoxic effects were induced in male Sprague-Dawley rats receiving single gavage or
s.c. doses of 20, 40, or 80 mg/kg AN (Ghanayem et al., 1991). Two distinct phases of
neurological response were evident. The early phase was cholinomimetic in nature, with signs
such as salivation, lacrimation, polyuria, miosis, vasodilatation, gastric secretion, and diarrhea.
The second phase developed 4-5 hours later with toxic signs, including depression, convulsions,
and respiratory failure, followed by death at the higher doses. These CNS effects were observed
in rats treated with 40 and 80 mg/kg and were similar to those caused by cyanide. Pretreatment
of animals with 1 mg/kg atropine, an acetylcholine muscarinic antagonist, abolished the
cholinomimetic toxicity, implicating an involvement of the cholinergic system in some aspects
of acute AN neurotoxicity. Since effects were observed even at the lowest dose, a NOAEL was
not identified, and 20 mg/kg was the LOAEL.
In another study, male Sprague-Dawley rats that were administered s.c. doses of
112 mg/kg AN (LD90) showed a biphasic response consisting of an early phase with tremors and
seizures about 100 minutes after dose administration, followed by severe clonic convulsions that
preceded death at about 3-4 hours (Benz and Nerland, 2005). The effects of preadministered
inhibitors of oxidative metabolism by CYP450 (80 mg/kg SKF 525A, 75 mg/kg
1-benzylimidazole, or 100 or 200 mg/kg metyrapone) and an alternative CYP450 substrate,
ethanol (5,000 mg/kg), on the acute convulsions were examined. Although blood levels of
cyanide, and the development of the first phase of tremors and seizures, were inhibited by
1-benzylimidazole or ethanol, treatment with these two agents did not prevent the terminal
convulsions or the death of rats injected with 112 mg/kg AN. Ethanol, being a CNS depressant,
decreased the incidence of terminal convulsion (5/17 vs. 15/17). These results suggested that the
initial phase of the acute neurotoxically lethal effects may have been due to cyanide, which is
released via the CYP450 metabolic pathway for AN, and that the second phase was mediated by
the parent compound.
Because ethanol showed some effect on lessening the second phase response to AN
(although it did not prevent lethality), several anticonvulsants were examined for their ability to
187
DRAFT - DO NOT CITE OR QUOTE
-------
counteract the acute neurotoxicity and lethality of 112 mg/kg AN. Administration of PB
(25 mg/kg) or phenytoin (150 mg/kg) (but not 144 mg/kg valproic acid) markedly inhibited the
lethal response to 112 mg/kg AN: 9/10 rats died following administration of AN alone or AN
plus valproic acid, whereas 1/10 and 2/10 rats died following administration of AN plus PB or
AN plus phenytoin, respectively. The protection by phenytoin and PB against convulsion and
lethality was not due to inhibition of metabolism of AN to cyanide, since only phenytoin was
able to lower blood cyanide levels (by about 32%), and was much less effective than
1-benzylimidazole or ethanol (97 and 94%, respectively).
4.4.1.5. Effects of AN on Hearing
AN is one of a number of organic compounds that have been shown to promote noise-
induced hearing loss (NIHL) in rats (Fechter, 2004; Fischel-Ghodsian et al., 2004). The
ototoxicity of AN was examined in several experiments in male Long-Evans rats exposed by s.c.
injection (Fechter et al., 2003). The AN used in these experiments was stabilized to minimize
the accumulation of peroxides. Ten rats were anesthetized and surgically prepared for the
assessment of the compound action potential (CAP), which represents the synchronous neural
activity elicited by primary auditory neurons (spiral ganglion cells) and directly measures
auditory threshold sensitivity. Auditory thresholds at each test frequency were measured for
each rat under anesthesia on 20 different occasions before treatment with AN to determine
baseline auditory thresholds. Auditory thresholds for 11 test frequencies between 2 and 40 kHz
were recorded at 5-minute intervals up to 100 minutes postinjection with AN. The acute effect
of 50 mg/kg s.c. AN on auditory sensitivity was studied in five rats; five control rats were
injected only with water. In a second study, the effects of AN on permanent NIHL were
evaluated in six experimental groups (six rats each) that received the following: no treatment;
50 mg/kg AN alone, two injections of 50 mg/kg-day AN on 2 consecutive days, noise alone
(108 dB octave-band noise for 8 hours), single injection of 50 mg/kg AN immediately followed
by noise, and exposure to noise after a second injection of AN.
The acute study showed that exposure to 50 mg/kg AN s.c. alone elevated auditory
threshold temporily and produced a 10-20 dB loss in auditory threshold sensitivity (temporary
threshold shift) in the stimulus range of 8-40 kHz (Fechter et al., 2003). This transient loss
reached a maximum within 10-20 minutes after injection but returned to control levels within
75-100 minutes. In the study on permanent auditory threshold shifts, when rats were tested
3 weeks following AN and AN + noise treatment, exposure to AN alone did not produce a
persistent loss in auditory threshold sensitivity. AN-treated rats had slightly reduced auditory
threshold compared with controls, indicating slightly increased sensitivity (within 5 dB of
controls), but the difference was not statistically significant. Noise treatment alone elevated
auditory thresholds by less than 20 dB at all tested frequencies. However, rats given two
injections of AN followed by noise in the high frequency range of 12-40 kHz exhibited auditory
188
DRAFT - DO NOT CITE OR QUOTE
-------
impairments averaging 27 dB (maximally 40 dB) compared with controls; this shift was
statistically significant compared with other groups (controls, noise alone, AN alone, or no
treatment). Rats given a single injection before exposure to noise showed an average 11 dB
threshold shift in the high-frequency range (12 and 40 kHz). This study demonstrates that
exposure to AN exacerbated NIHL.
Fechter et al. (2003) also assessed blood cyanide and glutathione levels in brain, liver,
and paired cochleae in additional groups of rats exposed to AN. Groups of five rats given 20, 50,
or 80 mg/kg s.c. AN produced peak levels of cyanide, a metabolite of AN, in the blood at 1 hour
(for 20 and 50 mg/kg) and 2 hours (for 80 mg/kg) following injection. Cyanide levels returned
to baseline values within 2, 3, and 4 hours, respectively. Since AN produced maximal auditory
threshold impairment within 20 minutes of administration, before blood cyanide level peaked at
1 hour, the acute ototoxic effect of AN was not likely associated with elevated cyanide levels.
When rats were given a single injection of 50 mg/kg, maximal reductions in glutathione
levels (measured between 15 minutes and 8 hours postinjection) were detected in the brain by
15 minutes (-50%), in the liver by 1 hour (-80%), and in the cochlea by 2 hours (—45%).
Cochlear glutathione levels remained depressed for about 4 hours. Recovery of glutathione
levels to near control levels was achieved in all tissues by 8 hours. Since AN induced transient
cochlear function loss that peaked within 10-20 minutes after injection and recovered within 75-
100 minutes, the acute ototoxic effect of AN could not be associated with GSH level in the
cochlea. Fechter et al. (2003) concluded that, while AN-induced oxidative stress in the cochlea
may play a role by which AN promotes NIHL, the acute ototoxic effect of AN might reflect
other unidentified toxic action of AN in the cochlea.
Fechter et al. (2004) extended their evaluation of the potentiation effect of AN exposure
on NIHL. Two experiments were conducted: the first experiment studied the effects of a single
AN exposure on permanent NIHL, while the second one evaluated the effects with five daily AN
and noise exposures. In both experiments, six male Long-Evans rats were assigned to each
treatment group: AN alone, noise alone, AN + noise, and untreated controls. AN exposed
groups were given s.c. injections of 50 mg/kg-day AN, with or without 4 hours of exposure to
noise (105 dB), and then assessed for auditory threshold sensitivity 4 weeks later.
In the single-day treatment study, AN alone did not alter the auditory threshold, but a
single AN exposure in combination with noise significantly elevated auditory thresholds by an
average of 10 dB above the effect of noise alone. Noise exposure alone increased auditory
thresholds compared with controls by an average of 10 dB in the range of 12-40 kHz (Fechter et
al., 2004). In the 5 consecutive-day treatment study, AN plus noise treatment exacerbated the
impairment induced by noise alone by an average of 30-45 dB at frequencies between 20 and
40 kHz and by no more than 10 dB at frequencies <16 kHz. Similar to the single-exposure
study, repeated AN exposure had no effect on auditory thresholds assessed 4 weeks later.
Repeated noise exposure elevated auditory thresholds about 17 dB above control rats.
189
DRAFT - DO NOT CITE OR QUOTE
-------
When the reactive-oxygen scavenger phenyl-N-tertiary-butylnitrone (PBN) (100 mg/kg
i.p.) was injected twice daily for 5 days prior to exposure to noise alone and rats were assessed
4 weeks later, PBN reduced the magnitude of hearing loss. When PBN was injected daily prior
to the injection of AN and again following noise exposure, the magnitude of hearing loss was
equivalent to that exhibited by rats treated by noise alone. (No significant difference was found
between rats receiving noise alone and those receiving PBN plus noise.) These results indicated
that reactive oxygen species (ROS) were responsible for the ototoxic effects of AN. These
findings are consistent with the suggestion that mitochondria injury within cochlear cells,
primarily or secondarily through oxidative stress, may be a common feature of ototoxicity
induced by chemicals and noise (Fischel-Ghodsian et al., 2004).
Results from a recent study indicated that hearing loss from AN and noise exposure
involves histologic damage to hair cells on the surface of the organ of Corti (Pouyatos et al.,
2005). Groups of five male Long-Evans rats were given s.c. injections of 0 or 50 mg/kg-day AN
for 5 consecutive days, with or without exposure 30 minutes later to noise for 4 hours/day (95 or
97 dB octave-band noise at 8 kHz). Hearing dysfunction of these rats were then assessed by:
(1) distortion product otoacoustic emissions (DPOAEs) before exposure, as well as 1 hour and
4 weeks postexposure; (2) CAP for auditory threshold sensitivity 4 weeks after the last treatment;
and (3) number of hair cells on surface preparation of the organ of Corti 4 weeks after treatment.
Permanent effects on these endpoints (i.e., effects observed 4 weeks following the last treatment)
were only observed in rats exposed to both AN and noise and not in rats exposed to AN or noise
alone. Permanent effects from combined exposure to AN and noise included auditory threshold
shifts (13-16 dB between 7 and 40 kHz), a decrease in DPOAE amplitudes (up to 25 dB at
19 kHz), and significant outer hair cell (OHC) loss in the cochleae. With the AN plus 97 dB
treatment, average OHC loss was 20, 16, and 9% in the first, second, and third rows,
respectively, in the areas corresponding to frequencies ranging from 13 to 47 kHz. Similar
effects were found in the AN plus 95 dB treatment group. This study demonstrated AN could
potentiate NIHL at noise levels that are relevant to human exposure.
Pouyatos et al. (2007) further proposed that AN exacerbated NIHL by decreasing
antioxidant defenses of hair cell. This hypothesis was tested in a study in which the capability of
specific antioxidants in the protection of the cochlea of male Long-Evans rats treated with
50 mg/kg AN s.c. 30 minutes prior to the daily noise exposure of 97 dB sound pressure level
4 hours/day for 5 days. Sixty-five Long-Evans rats (2-12/group) were exposed to different
combinations of noise, AN, and antioxidants; AN alone or AN + STS (150 mg/kg i.p.), a CN
inhibitor; AN + 4-methylpyrazole (4MP,100 mg/kg i.p.), a drug that blocks CN generation by
competing with CYP2E1; AN + L-N-acetylcysteine (L-NAC) (4 x 400 mg/kg, orally), a
pro-GSH drug; noise (97 dB octave band of noise [OBNJ/8 kHz) alone; noise and STS, 4MP, or
L-NAC; or noise plus AN and antioxidants. To evaluate auditory impairment, DPOAEs and
CAPs were measured prior to experimental treatment and 3 days and 4 weeks after treatment. At
190
DRAFT - DO NOT CITE OR QUOTE
-------
the end of exposure, cochleae were harvested for histologic examination. Additional rats (n =
64) were used to measure cochlear and liver GSH and blood CN levels at different time points
after treatment.
At 3 days postexposure, similar auditory loss was found in animals exposed to AN +
noise, STS + AN + noise, 4MP +AN + noise, and L-NAC + AN + noise. The maximum shifts
averaged 25-30 dB between 12 and 32 kHz. At 4 weeks postexposure, animals exposed to
L-NAC + AN + noise recovered to baseline levels above 25 kHz. However, at lower
frequencies, L-NAC did not prevent auditory loss caused by AN + noise exposure. Animals
received combined exposure to AN + noise, and AN + noise + STS or AN + noise + 4MP
showed little change in producing auditory loss.
In addition, the cochleae from rats exposed to AN and noise demonstrated substantial
damage in the basal half of the organ of Corti. Mean OHC loss averaged 35% in the three rows
in the region corresponding to frequencies above 12 kHz. Neither STS nor 4MP pretreatment
protected against OHC loss caused by AN + noise. However, pretreatment with L-NAC reduced
the OHC loss caused by AN + noise in the region corresponding to 25 kHz and above.
Liver GSH level was depleted by 63% 1 hour after AN injection. Cotreatment with STS
or 4MP reduced GSH level about 80% at 1 hour and 52% at 3 hours. However, with L-NAC
pretreatment, GSH level was reduced only by 23 and 20% at 1 and 3 hours, respectively.
Similarly, whereas AN treatment depleted cochlear GSH levels to undetectable levels at 1 hour,
STS pretreatment had no effect on GSH depletion. 4MP pretreatment only slightly reduced GSH
depletion. However, L-NAC pretreatment not only protected but induced an increase in GSH
levels above control levels. Pouyatos et al. (2007) concluded that, since L-NAC cotreatment
reduced auditory loss and OHC loss from AN + noise treatment, GSH is involved in the
protection of the cochlea against ROS generated by moderate noise levels. However, CN did not
appear to be involved in this potentiation.
4.4.2. Immunological Effects of AN
A case study by Balda (1975) described a human subject who developed contact
dermatitis following the use of a Plexidur finger splint. Investigators obtained a positive result in
a patch test with AN, which is one of the constituents of the polymer Plexidur. This observation
suggested that AN might induce immunological response in exposed subjects.
The immunotoxicity of AN was evaluated in groups of six male CD-I mice given a
single oral dose in water or repeated oral doses for 5 or 14 days (Ahmed et al., 1993). In each of
these experiments, a positive control group received a single immunosuppressive dose of
225 mg/kg of cyclophosphamide i.p. In the first experiment, mice given a single gavage dose of
0 or 13.5 mg/kg AN in water were evaluated 3 or 5 days later for BW, relative organ weights
(thymus, spleen, liver, and kidney), number of viable splenocytes, and total and differential
WBC counts in spleen cells suspension (Ahmed et al., 1993). AN slightly reduced BWs in mice
191
DRAFT - DO NOT CITE OR QUOTE
-------
(6-7%) after 3 and 5 days, but the difference was not biologically significant. In treated mice,
relative spleen weights were increased by 50% after 5 days, whereas relative thymus weights
were decreased by 42% after 5 days; relative liver weight was decreased by 11% on day 3 but
was 9% higher than controls on day 5. AN treatment reduced total leucocytes by 42-57%,
lymphocytes by 39 and 65%, monocytes by 38 and 50%, and neutrophils by 67 and 73% on
days 5 and 3, respectively.
In the second experiment, mice received 0 or 6.75 mg/kg-day of AN in water on
5 consecutive days and were evaluated on day 6 (Ahmed et al., 1993). In mice treated for
5 days, BWs were reduced by 27%, relative liver weights were decreased by 10%, and relative
spleen weights were increased by 25%. Total splenocytes were decreased by 51%, with total
blood leucocytes and lymphocytes reduced by 48 and 68%, respectively; circulating monocytes
were increased by 88%. The immunotoxic effects of AN in both studies were comparable to
effects elicited by the known immunosuppressant cyclophosphamide.
In a third experiment, groups of male CD-I mice (six/group) were treated with AN by
gavage in water for 14 days (Ahmed et al., 1993). There were a total of nine groups: three
groups dosed with AN at 1.35, 2.7, or 5.4 mg/kg-day only, three groups dosed the same way but
immunized with sheep red blood cells (SRBCs) on day 9 of exposure, and three concurrent
control groups (normal, SRBC-immunized, and a positive control group treated intraperitoneally
with cyclophosphamide [225 mg/kg] 1 day before immunization). Mice were evaluated for total
BW, relative organ weights (thymus, spleen, liver, and lung), number of splenocytes, total and
differential WBC counts in spleen cell suspensions, and histopathology of lymph nodes, lung,
and intestinal Peyer's patches. Subsets of spleen lymphocytes (T and B cells) were enumerated
by flow cytometry. Spleen cells from mice immunized with SRBCs were evaluated in a plaque-
forming cell assay.
In nonimmunized mice, AN decreased BW by 12-15% and increased relative thymus
weights by 43% at 2.7 mg/kg-day and relative lung weights by 39% at 5.4 mg/kg-day. AN
increased splenocyte viability by 49-264%) in a dose-independent manner. AN also caused dose-
independent increases of 126-293%) in relative spleen weight at all doses. Total leukocyte
counts/spleen were increased by 171, 119, and 107% at the low to high doses; lymphocyte
counts/spleen were significantly reduced by 25—45% at all doses, while monocyte and neutrophil
counts/spleen were increased concomitantly by as much as 78-fold. Reductions in lymphocyte
subsets were observed at all doses: T-cells by 40—53%, B-cells by 32—36%, T-helper cells by
40-59%), and T-suppressor cells by 49-62%). Histological examination revealed severe
enlargement of mesenteric lymph nodes and intestinal Peyer's patches, abscesses and massive
necrotic damage in the lung, and swellings in the brachial lymph nodes in mice treated with AN
at all doses but more aggressive in the 5.4 mg/kg group. No incidence data were reported for
these lesions. Some of the mice treated with 5.4 mg/kg-day died rapidly. Microbiological
192
DRAFT - DO NOT CITE OR QUOTE
-------
examination revealed the swelling in the lymph nodes was related to migration of normal
intestinal flora, which the study authors attributed to immunosuppressive effects of AN.
In the SRBC-immunized mice, AN treatment did not affect the BW. The relative weight
of lung, liver, and thymus showed inconsistent and dose-independent increases. The viable
spleen cells showed 77% increases only in the 2.7 mg/kg group. The lymphocytic count showed
dose-independent decreases. However, the neutrophilic and monocytic counts showed large
increases at all doses. Decreases in lymphocyte subsets were found in all three doses of AN:
T-cells by 42-45%, B-cells by 37-50%), T-helper by 43-54%), and T-suppressors by 40-48%).
The IgM antibody plaque forming cell response was decreased by 55, 23, and 65 after treatment
with 1.35, 2.7, and 5.4 mg/kg AN, respectively. The lowest dose used in this study, 1.35 mg/kg-
day, was a LOAEL for immunotoxicity (suppression of humoral and cell-mediated immunity) in
mice treated for 14 days.
Hamada et al. (1998) further investigated the immunotoxicity of AN by administering
2.7 mg/kg-day AN (1/10 the LD50) to male CD-I mice orally for either 5, 10, or 15 days. All
mice were injected with 100 mg/kg bromodeoxyuridine (BrdU) i.p. 1 hour before sacrifice. An
immunohistochemical assessment of the number of cells capable of producing IgA in different
intestinal compartments as a result of AN administration was conducted. Uptake of
-3
H-thymidine into splenocytes derived from treated animals and stimulated with different
mitogens—phytohemagglutinin (PHA), concanavalin-A (con-A), or lipopolysaccharide (LPS)—
was measured. The rate of proliferation of gut epithelial cells of different intestinal
compartments was determined by the incorporation of BrdU in newly synthesized DNA of
S-phase cells.
The mitogenic response of mouse splenocytes to PHA, con-A, and LPS was affected by
AN exposure as shown by a significant decrease in [ H]-thymidine incorporation (Table 4-49).
The decreases were 68-79%> with con-A and 35-51% with LPS, depending on the time intervals.
"3
However, uptake of [ H]-thymidine by splenocytes after stimulation with PHA was markedly
reduced only after 15 days of exposure. These results suggested that AN induced systemic
suppression of humoral immunity (by the decrease in mitogen response to LPS) and cell-
mediated immunity (by the inhibition of mitogen response to con-A and PHA).
193
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-49. Time course of the effect of AN administration on
[3H]-thymidine uptake into mouse splenocytes under the influence of
different mitogens in vitro
Mitogens
Control
D of AN treatment
5
10
15
PHAa
1,331 ± 163
1,203 ±265
1,434 ±35
898 ± 32b
con-Aa
14,927 ±972
4,756 ± 532b
4,792 ± 946b
3,198 ±448b
LPS3
1,225 ± 112
803 ± 87b
486 ± 2b
522 ± 31b
"Values are counts/min, mean ± standard error of the mean; n = 4.
bSignificantly different from controls (p < 0.05) as calculated by the authors.
Source: Hamada et al. (1998).
"3
Inhibition of [ H]-thymidine uptake by stimulated spleen lymphocytes may indicate a
systemic immunosuppression by AN. Such an effect could also occur locally, as indicated by a
reduction in the number of IgA-producing cells in all intestinal compartments following AN
administration. The counts of IgA-producing cells were reduced by 56-77% in the duodenum,
44-67% in the jejunum, and 60-62% in the ileum. Another local effect of AN was demonstrated
by the increased incorporation of BrdU into epithelial cells of the duodenum (threefold) and
ileum (1.6-fold) of AN-treated animals. This result indicated that the rate of cell proliferation
was markedly increased following oral AN administration, even as a result of short-term
treatment. The study authors considered this to be the result of a regenerative response to
chemically induced intestinal injury and suggested that AN-induced immunosuppressive effect
systemically and locally in the gut, as well as increases in the rate of cell proliferation, may
contribute to carcinogenicity of AN in the gut.
Summary
AN caused contact dermatitis in a human subject exposed via the use of Plexidur finger
splint. In mice, AN suppressed cell-mediated and humoral immunity systemically and locally in
the intestine. Table 4-50 summarizes immunotoxicity studies of AN.
194
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-50. Summary of immunotoxicity studies of AN
Test
species
Endpoint/effect
Exposure
concentration/condition
Exposure
duration
Results
References
Human
subject
Contact dermatitis
Use of a Plexidur finger splint
ND
Positive patch test
Balda, 1975
Male CD-I
mice
(n = 6)
Immunosuppressive
Effect
Study 1:
Single dose of 0 or
13.5 mg/kg AN in water
(oral)
225 mg/kg cyclophosphamide
as positive control
3 or 5 d
Increased relative spleen
by 50%, decreased
relative thymus weight
by 42%, reduction in
total leucocytes,
lymphocytes,
monocytes, and
neutrophils.
Ahmed et al.,
1993
Study 2:
0 or 6.75 mg/kg-d AN in
water
225 mg/kg cyclophosphamide
as positive control.
5 d
Increased relative spleen
weights by 25%;
decreased total
spenocytes by 51%;
reduction in total blood
leucocytes and
lymphocytes by 48 and
68%, respectively.
Study 3:
a. Three groups dosed with
1.35, 2.7, or 5.4 mg/kg-d AN
b. Three groups dosed with
1.35, 2.7, or 5.4 mg/kg-d AN,
but immunized with SRBCs
on d 9 of exposure
c. Three concurrent control
groups (normal, SRBC-
immunized)
d. a positive control group
treated intraperitoneally with
cyclophosphamide
(225 mg/kg) 1 d before
immunization
14 d
In nonimmunized mice:
increased spleen weight
at all doses, increased
total leukocyte counts,
reduced lymphocyte
counts/spleen, and
increased monocyte and
neutrophil
counts/spleen.
Enlargement of
mesenteric lymph nodes
and intestinal Peyer's
patches, swellings in the
brachial lymph nodes for
all dose groups.
In the SRBC-immunized
mice: viable spleen cells
showed 77% increase in
the 2.7 mg/kg group.
Decreases in
lymphocyte count,
increases in neutrophilic
and monocytic counts,
and decreases in plaque
forming cell response
were observed. The
LOAEL for
immunotoxicity was
1.35 mg/kg-d.
195
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-50. Summary of immunotoxicity studies of AN
Test
species
Endpoint/effect
Exposure
concentration/condition
Exposure
duration
Results
References
Male
CD-mice
Systemic and local
immunosuppression
2.7 mg/kg-d AN orally
5, 10, or
15 d
Decrease 3H-thymidine
incorporation into
splenocytes in: (l)mice
treated with mitogens
con-A or LPS,
suggesting systemic
suppression of cell-
mediated immunity, and
(2) mice treated with
mitogen PHA,
suggesting systemic
suppression of humoral
immunity.
Local
immunosuppression in
the gut, as indicated by:
(1) reduction in the
number of IgA-
producing cells in all
intestinal compartments,
and (2) increased
proliferation of
epithelial cells of the
duodenum.
Hamada et al.,
1998.
4.5. MECHANISTIC DATA AND OTHER STUDIES IN SUPPORT OF THE MODE OF
ACTION
4.5.1. Mode-of-Action Studies
Studies have been conducted with the primary purpose of evaluating the potential
mechanisms by which AN and/or its metabolites produce noncancer and cancer effects in
experimental animals. Studies that investigated potential mechanisms for noncancer effects
include GI hemorrhaging, effects on Hb and metabolism in RBCs, neurotoxicity, oxidative
stress, and immunotoxicity. Studies that investigated the potential mechanisms for the
carcinogenicity of AN include formation of DNA adducts, oxidative stress, intercellular
communication, and cell proliferation. In addition, genotoxicity studies are described in
Section 4.5.2.
4.5.1.1. Noncancer Endpoints
4.5.1.1.1. GI hemorrhaging. Ghanayem and Ahmed (1983) administered AN to male Sprague-
Dawley rats by both s.c. or oral routes, in each case observing significant gastric bleeding (see
Section 4.4.1.1). The response appeared to peak at 50 mg/kg, the effect being maximal at
2 hours after a 50 mg/kg s.c. dose. Different CYP450 inducers had different effects:
Aroclor 1254 more than doubled the effect, while PB reduced the bleeding by half. CYP450
196
DRAFT - DO NOT CITE OR QUOTE
-------
inhibitors SKF 525A and cobalt chloride were even more effective, with cobalt chloride reducing
gastric bleeding to almost untreated levels. GSH depletion by DEM did not prevent
hemorrhaging. The effect was not due to cyanide release from AN, since KCN administration
(6 mg/kg) did not induce bleeding. The results indicated that the gastric hemorrhaging was
caused by metabolic activation of AN to a reactive metabolite other than cyanide (probably
CEO).
As described in Section 4.4.1.1, the follow-up study by Ghanayem et al. (1985)
demonstrated that pretreatment of rats with sulfhydryl-containing compounds or atropine
protected the rats from AN-induced lesions. The findings suggested that modulation of
muscarinic receptors by AN increased gastric acid secretion and caused gastric mucosal erosions.
4.5.1.1.2. Effects on Hb and metabolism in RBCs. Farooqui and Ahmed (1983b) studied the
effect of AN on Hb and metabolism in RBCs both in vivo and in vitro. Male Sprague-Dawley
rats (three/group) were administered a single oral dose of 80 mg/kg aqueous AN, and blood was
collected 1 hour after treatment. Other groups of animals were sacrificed at 3, 6, and 24 hours
after dosing. Mean cell Hb concentration, hematocrit, and platelet counts were reduced to
78, 79, and 71% of the controls 1 hour after dosing. GSH levels were lowered significantly
within 1 hour. Among the metabolic intermediates measured, a reduction in the intracellular
concentration of 2,3-diphosphoglyceric acid and increases in intracellular levels of ATP,
pyruvate, and lactate were observed. The increase in the levels of pyruvate and lactate, which
were end products of glycolysis, suggested an increase in the metabolic rate in RBCs as a result
of exposure to AN (Table 4-51). A significant decrease in the activity of 2,3-diphosphoglycerate
mutase, an erythrocyte enzyme, was found in treated animals (3.61 ± 0.21 IU/g Hb in treated
animals after 1 hour vs. 4.57 ± 0.23 IU/g Hb in controls). In general, the intermediates studied
returned to normal values between 6 and 24 hours.
Table 4-51. Effect of AN on RBC metabolic intermediates following a single
oral dose
Intermediates
(nmol/mL blood)
Control
AN treatment
lhr
3 hrs
6 hrs
24 hrs
2,3-Diphosphoglycerate
3.6 ±0.3
2.8 ±0.2a
3.0 ±0.4
3.2 ± 0.3
3.7 ± 0.5
Adenosine triphosphate
0.49 ±0.04
0.60 ± 0.05a
0.64 ± 0.06a
0.66 ± 0.06a
0.58 ±0.05
Pyruvate
58 ±6
227 ± 21a
187 ± 13a
175 ± 16a
167 ± lla
Lactate
1,478 ±131
9,932 ± 21 la
1,889 ± 129a
1,955 ± 160a
2,079 ± 151a
GSH
1.39 ±0.11
0.28 ± 0.02a
0.34 ± 0.04a
0.47 ± 0.05a
1.27 ±0.15
"Significantly different from controls (p < 0.05) as calculated by the study authors.
Source: Farooqui and Ahmed (1983b).
197
DRAFT - DO NOT CITE OR QUOTE
-------
In another study, male Sprague-Dawley rats were treated with single oral doses of
46.5 mg/kg [2,3-14C]-AN (Farooqui and Ahmed, 1983b). RBC GSH concentrations were
lowered to 10% of controls in 1 hour, followed by a slow recovery of 10% in 5 hours. Extensive
covalent binding of AN to Hb (about 1.7 [j.mol equivalents of AN bound/mL RBCs) in 1 hour
was also observed.
For the in vitro study, isolated RBCs from male Sprague-Dawley rats were incubated
with 5 mmol/L AN at 37°C to investigate the ability of AN to interact covalently with Hb and
deplete GSH, and the effect of such depletion on Hb (Farooqui and Ahmed, 1983b). GSH in the
supernatant and in the RBCs was estimated by measuring nonprotein sulfhydryl groups. In
addition, the levels of Hb and MetHb and GSH conjugates with AN were monitored.
Incubation of rat RBCs with AN caused a depletion of more than 85% of intracellular
GSH within 1 hour. No hemolysis occurred during the incubation period. In addition, GSH was
not detected in the incubation media. Most of the GSH in the RBCs was converted to a
GSH-AN conjugate, S-cyanoethyl GSH. This conjugate was present in the RBCs after 24 hours,
suggesting that it was not metabolized further at least for 24 hours. Moreover, about 6% of Hb
was converted to MetHb in 1 hour and 8% in 3 hours (Farooqui and Ahmed, 1983b). MetHb
levels in controls during this time period were 0.61-0.87%).
The rate and extent of MetHb reduction to Hb in AN-treated RBCs was also determined
to investigate if this protective mechanism against such oxidative damage to Hb was affected.
Incubation of nitrite-treated RBCs (for conversion of Hb in RBCs to MetHb) with AN resulted in
a significant decrease in MetHb reduction, with a 70% decrease in RBCs treated with 10 mM
AN when compared with controls. In the same experiment, AN initiated hemolysis of RBCs at a
concentration of <0.1 M.
Farooqui and Ahmed (1983b) suggested that the effects of AN on RBC metabolism were
related to the availability of GSH. Oxidative stress induced by depletion of GSH as a result of
AN exposure may have stimulated the rate of RBC metabolism, based on increase in the end
products of glycolysis. Another possible explanation could be the impaired permeability of the
erythrocyte membrane due to extensive covalent binding of AN, resulting in the retention of
metabolic products. Since the levels of ATP and 2,3-diphosphoglycerate were altered and these
two intermediates regulate the oxygen dissociation curve, it was concluded that chronic exposure
to AN may lead to methemoglobinemia, damage to the RBC membrane, and impaired delivery
of oxygen to tissues.
Farooqui et al. (1990) provided in vitro data on the effect of AN on lipid metabolism in
RBCs. Three types of RBC preparations from male Sprague-Dawley rats were incubated with
AN: those containing oxyhemoglobin (HbO), those containing MetHb, and those containing
carboxyhemoglobin (HbCO). HbO-containing RBCs were taken directly from the RBC pellet;
MetHb-containing RBCs were produced by incubating packed RBCs with 0.5% sodium nitrite;
while RBCs containing HbCO were prepared by blowing carbon monoxide over a 20%
198
DRAFT - DO NOT CITE OR QUOTE
-------
suspension (volume/volume [v/v]) of RBCs until the visible spectrum of red cell lysate reached a
maximum at 570 nm. All preparations of RBCs were incubated for 1 hour with 10 mM AN and
variable additions of glucose. Following incubation, supernatants were removed and used to
estimate lipid peroxidation by measuring the concentration of conjugated dienes, while the red
cell pellets were used to determine Hb. In another set of experiments, isolated rat red cell
membranes were incubated with 25 mM AN at different temperatures. Total and Na+/K+-
ATPase activity were measured in control and AN-incubated red cell membranes.
Incubation of HbO-containing RBCs with AN resulted in the formation of MetHb, loss of
intact Hb, and membrane lipid peroxidation. The availability of glucose to HbO-RBC incubation
reduced the formation of MetHb by 33% and the loss of intact HbO by 33% but increased lipid
peroxidation by 35%. As a positive control, HbO-containing RBCs were incubated with 0.1 mM
t-butyl hydroperoxide, a strong GSH depleter. The formation of MetHb and non-intact Hb in the
positive control was 3 and 5 times higher than that in AN-incubated RBCs. Incubation of
MetHb-containing RBCs with AN also resulted in the loss of intact Hb and membrane lipid
peroxidation. However, availability of glucose resulted in only 13% increase in peroxidation of
membrane lipids. Lipid peroxidation was about 3 times higher in incubations of HbCO-
containing RBCs with AN, with or without glucose, than the other two RBC preparations. The
extent of lipid peroxidation in RBCs and isolated RBC membranes was dependent on the
concentrations of AN.
Farooqui et al. (1990) also demonstrated an inverse relationship between GSH
concentrations and lipid peroxidation in RBCs incubated with AN. A 75% reduction in GSH
levels in RBCs was observed as a result of AN incubation for 2 hours, with the half-life of GSH
depletion being less than 22 minutes. The concentration of lipid peroxides increased by 274%)
over control levels during the same period.
In addition, Farooqui et al. (1990) showed that Na+/K+-ATPase activity was reduced in
the isolated RBC membranes incubated with AN (release of inorganic phosphate: 87.9 ±9.1 vs.
145.6 ± 13.1 nmol/mg protein per hour at 37°C; and 4.3 ± 0.7 vs. 87.9 ±9.1 nmol/mg protein per
hour at 15°C). The degree of AN-induced inhibition of ATPase was temperature dependent.
The Km of Na+/K+-ATPase (3.5 mM at 37°C) was barely affected by the AN treatment, while the
Vmax was significantly lower (—36%) than that of controls. This noncompetitive inhibition of
Na+/K+-ATPase by AN was proposed to be the result of changes in the physicochemical
properties of RBC membrane macromolecules subsequent to irreversible binding of AN to
membrane proteins.
In summary, Farooqui et al. (1990) demonstrated that AN induced GSH depletion,
elevated lipid peroxidation, and enhanced concentration of MetHb in RBC, as well as inhibited
Na+/K+-ATPase activity in RBC membrane.
199
DRAFT - DO NOT CITE OR QUOTE
-------
4.5.1.1.3. Neurotoxicity. Campian et al. (2002) evaluated the capacity of AN to inactivate the
important glycolytic enzyme, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) in vitro in
order to elucidate the mechanism for the acute toxicity of AN. GAPDH was incubated with AN
(final concentrations ranged from 50 to 400 |iM), and the activity of GAPDH was assayed at
different time points. An irreversible inhibition of GAPDH activity was obtained. Incubation of
GAPDH with 200 |iM AN resulted in 90% loss of activity in about 60 minutes. The second-
order rate constant for inhibition of GAPDH activity was measured as 0.2 M"1 s"1 at pH 7.4 and
37°C.
The site of AN incorporation was identified by using matrix-assisted laser desorption
ionization time-of-flight mass spectrometry to analyze tryptic digests of control and AN-labeled
GAPDH (Campian et al., 2002). Inactivation of GAPDH was due to covalent binding of AN to
cysteine 149 at the active site of the enzyme. This finding demonstrated the specificity of AN
binding to cysteine residues and, more widely, implied the ability of AN to impair glycolytic
ATP production in vivo. Campian et al. (2002) speculated that the combination of glycolytic
ATP production impairment with inhibition of mitochondrial ATP synthesis by the AN
metabolite cyanide could result in metabolic arrest. This might have profound consequences for
the toxicity of AN in sensitive tissues such as the brain.
In a followup study (Campian et al., 2008), male Sprague-Dawley rats (3-5/group) were
injected subcutaneously with LD90 (115 mg/kg) AN and the brains of treated rats were frozen
when respiration ceased via head immersion (HI) in liquid nitrogen or funnel freezing (FF)
technique. Only minor decreases in ATP of 5% (FF) and 21% (HI) were found when respiration
ceased, although phosphocreatine was decreased by 74% (FF) and 80% (HI), possibly due to
inhibition of creatine kinase by AN. Campian et al. (2008) concluded that no toxicological
relevant depletion of ATP occurred when respiration ceased in AN treated rats. Hence, the acute
lethality of AN was not due to brain metabolic arrest.
Dorman et al. (1996) investigated the mechanisms by which AN and CEO exert
neurotoxic effects in primary dissociated cerebrocortical cell cultures prepared from GDI6-18
CD rats. Mature 7-day old cultures were exposed to 0-10 mM AN or 1-2 mM CEO for 8 h at
37°C. Cytotoxicity was evaluated by measuring leakage of LDH, GSH depletion, inhibition of
acetylcholinesterase (AChE) and histopathology.
Both AN and CEO induced dose-dependent increased in cytoxicity. Significant increase
in LDH-leakage was observed in neural cultures following 8 h exposure to 2.5-10 mM AN or
0.125 - 1 mM CEO. Significant reduction in GSH only occurred following exposure to 5 mM
AN. Thus, Dorman et al. (1996) concluded that GSH depletion probably played a limited role in
the development of AN toxicity. No change in AN-induced cytotoxicity was observed following
cotreatment with the cytochrome P-450 inhibitor 1-phenylimidazole, indicating minimal
metabolism of AN to CEO in the neural cells. AChE inhibition was not observed following 8 h
exposure up to 2.5 mM AN. Widespread necrosis of the small, round cholinergic neurons was
200
DRAFT - DO NOT CITE OR QUOTE
-------
present in the cultures treated with 2.5 mM AN or 0.125 mM CEO for 8 h. Astrocytes,
pyramidal neurons, and bipolar neurons were mostly unaffected. On the other hand, treatment
of cell cultures with >0.6 mM cyanide resulted in extensive loss of cytochrome oxidase activity
in the large pyramidal neurons without a concomitant increase in LDH leakage. Dorman et al.
(1996) concluded that AN may selectively destroy cholinergic neurons and induce neurotoxicity
in rats, and that AN metabolism to cyanide was not a prerequisite for the development of AN-
induced neurotoxicity.
Satayavivad et al. (1998) provided some evidence for the alterations of central
muscarinic functions from subchronic exposure to AN. Either 0, 1, or 25 mg/kg AN was
injected subcutaneously to male Wistar rats (10/group) 5 days/week for 8 weeks. The authors
studied the impact of AN on motor behavioral activities by using a computerized system to chart
the movements of rats within the cage. The system provided information on a total of 11 motor
behavioral parameters, such as distance traveled/time interval, resting time, time spent moving,
and number of clockwise and counterclockwise rotations. This test model can be used to detect
subtle changes of central muscarinic receptors during exposure to cholinomimetic agents.
Evaluations were carried out on each animal on day 5 of weeks 1, 2, 4, 6, and 8, 1 hour after
treatment with AN.
Both doses of AN were associated with marked decreases in all motor activities and a
concomitant increase in the resting time of treated rats compared with controls (Satayavivad et
al., 1998). There was a decrease in all motor parameter values at weeks 1 and 2, but most of
these effects were diminished by weeks 4 and 6. However, the effects of the high dose of AN
were more pronounced and longer lasting. Thus, the incidence of clockwise and
counterclockwise rotations in high-dose rats was reduced compared with controls throughout the
study (for example, 0.7 ± 0.3 [high dose], 2.6 ± 0.8 [low dose], and 2.8 ± 0.4 [controls]
counterclockwise rotations/10-minute study period after 8 weeks). The effects of intramuscular
injection of the muscarinic receptor antagonist, atropine, and the reversible acetylcholinesterase
inhibitor, physostigmine, with and without concurrent AN administration, also were evaluated in
this system. Atropine administration (10 mg/kg) was associated with increases in motor activity
that were enhanced by AN treatment at 25 mg/kg. Physostigmine (0.5 mg/kg) caused reductions
in motor activity irrespective of AN administration. Satayavivad et al. (1998) concluded that AN
possesses cholinomimetic effects, one of which might include the down-regulation of muscarinic
receptors. This would explain the marked increase in the response to atropine. Since AN did not
inhibit the activity of acetylcholinesterase, the cholinomimetic effect of AN might be mediated
by the release of acetylcholine from nerve endings.
Jacob and Ahmed (2003b) studied AN-induced neurotoxicity by exposing proliferating
normal human astrocytes (NHAs) in culture to 25-400 |iM AN for 12 hours. Assessment was
then made on cell viability, levels of endogenous antioxidants, GSH, catalase, levels of ROS, and
secretion of tumor necrosis factor (TNF-a), a cellular marker for oxidative stress and oxidative
201
DRAFT - DO NOT CITE OR QUOTE
-------
damage to nuclear DNA. Treatment with 25-50 [jM AN had no significant effect on viability of
the astrocytes. At 100-400 (j,M AN, 15-42% reduction in cell viability was observed (as
indicated by trypan blue exclusion). Reduced viability was further substantiated by 8-40%
increased cytotoxicity (as indicated by leakage of LDH). The morphology of astrocytes was
normal at concentrations up to 200 (j,M, but cells exposed to 400 (j,M AN showed a larger
number of swollen nuclei and enlarged membrane structures.
Intracellular levels of GSH were not affected at 25 and 50 |iM AN. However, a
significant dose-dependent decrease in GSH was observed at 100, 200, and 400 [xM AN (18, 28,
and 35% lower than controls, respectively). A concomitant increase in levels of oxidized GSH
(glutathione disulfide [GSSG]) was observed (Jacob and Ahmed, 2003b). The ratio of GSH to
GSSG was reduced from the control value of 37-18, 7, 3, and 2 at 50, 100, 200, and 400 [jM
AN, respectively. Compared to control level, catalase activity increased 21% at 100 |iM AN, but
declined to 37% below control levels at 400 [xM.
Significant increases in measures of oxidative stress (four- to sevenfold increase in the
generation of ROS) and oxidative DNA damage (greater than twofold increase in
8-oxodeoxyguanosine [8-oxodG]), were observed at 200-400 |iM AN. Treatment at 400 |iM
significantly increased the release of the inflammatory cytokine, TNF-a, by 30% compared with
controls. The observation that compromised antioxidant defense mechanisms (depletion of
glutathione, increase in GSSG, inhibition of catalase) occurred at the same exposure
concentration as reduced cell viability supported the hypothesis that oxidative stress in astrocytes
was a possible mechanism for neurotoxic effects of AN exposure.
In a translated Chinese study (Lu et al., 2005b), levels of monoamine neurotransmitters
and their metabolites were measured in the striatum and cerebellum of the brains of male
Sprague-Dawley rats (n = 10 per group, 7 selected randomly for measurement) exposed to 0, 50,
or 200 ppm AN in drinking water for 12 weeks. The study authors estimated the administered
doses to be 4.0 and 13.5 mg/kg-day for the 50 and 200 ppm groups, respectively. Monoamine
oxidase activity in the cerebral cortex was also measured.
Compared with control values, average dopamine levels in the striatum were decreased
by 76 and 64% in the 50 and 200 ppm groups, respectively, and by 46 and 18% in the
cerebellum. The decreases in dopamine levels were statistically significant only in the striatum.
No statistically significant exposure-related changes were observed in average levels of the
dopamine metabolite, 3,4-dihydroxyphenylacetic acid, in the striatum or the cerebellum except
for increased levels (by about 32%) for the 50 ppm group in the cerebellum compared with
controls. Average concentrations of serotonin were decreased by about 38 and 49% in the
striatum in the 50 and 200 ppm groups, respectively, and by about 41 and 68% in the cerebellum.
These changes were statistically significant only in the striatum. No statistically significant
exposure-related changes were observed in average levels of the serotonin metabolite,
5-hydroxyindoleacetic acid, in the striatum or the cerebellum, in average levels of
202
DRAFT - DO NOT CITE OR QUOTE
-------
norepinephrine in the two brain regions, or in average activities of monoamine oxidase. The
observed changes in the endpoints examined in this study are of uncertain biological relevance;
as such, the study does not identify NOAELs or LOAELs suitable for health hazard
identification or dose-response assessment. However, the striatum and cerebellum are major
centers for movement control, balance, and coordination. Thus, an abnormality in the
neurotransmitters will cause malfunction in the movement coordination of an organism. The
study authors suggested that a decrease in dopamine will reduce the environmental adaptability
of rats and provide neurotransmitter evidence for the effect of AN exposure on neurobehavior.
Serotonin plays a role in the maintenance of sleep and the emotional and psychological states of
the body. Thus, reduced serotonin level in the exposed groups may also have neurobehavioral
effect.
4.5.1.1.4. Oxidative stress. Farooqui and Ahmed (1983b) demonstrated that exposure of rats to
AN resulted in depletion of GSH and induction of oxidative stress in RBCs (see Section
4.5.1.1.2). Farooqui et al. (1990) provided in vitro data that AN induced GSH depletion, lipid
peroxidation, and enhanced concentration of MetHb in rat RBCs.
In another study, the cytotoxic effect of AN-potentiated oxidative stress in rat alveolar
macrophages was investigated (Bhooma and Venkataprasad, 1997). When alveolar macrophages
isolated from male Wistar rats were incubated with 200 nM to 20 [jM AN at 37°C for up to
4 hours, a dose-dependent loss of viability was observed. Incubation of alveolar macrophages
with 10 |iM AN increased the release of H2O2 by 44% when compared with controls. This effect
was abolished by the addition of antioxidant enzymes superoxide dismutase (SOD) or catalase to
the incubation medium. In addition, while exposure of alveolar macrophages to 10 |iM AN
resulted in 42% viability, addition of SOD exerted a marked protective effect with a resultant
viability of 79%. Thus, cell injury induced by AN is mediated by the production of highly toxic
OH radicals.
The role of oxidative stress and lipid peroxidation in the induction of the toxic effects of
AN was studied by a number of other research groups. In a study designed to evaluate the
possible role of free-radical-mediated lipid peroxidation in the etiology of AN-induced acute
adrenal necrosis, Silver and Szabo (1982) monitored the formation of MDA and conjugated
diene concentrations in the adrenal glands and other tissues (brain, liver, stomach, duodenum) of
female Sprague-Dawley rats treated with an i.v. dose of 150 mg/kg AN. Controls received i.v.
injections of aqueous 0.1% Tween 80. Rats were killed at 15, 30, 60, or 90 minutes after
injection of AN, and the liver, adrenal glands, brain, stomach, and duodenum were removed.
While no effects on the monitored parameters were observed in the mitochondria or microsomes
from adrenal gland, brain, duodenal mucosa, or glandular stomach mucosa of rats 30 minutes
after injection of AN, conjugated diene concentrations were elevated by 60% in hepatic
microsomes and 30 and 40% in gastric mitochondria and microsomes. Therefore, although lipid
203
DRAFT - DO NOT CITE OR QUOTE
-------
peroxidation is unlikely to be involved in the pathogenesis of AN-induced adrenal necrosis, AN
can cause lipid peroxidation in other target organs.
The cytotoxicity and oxidative stress induced by AN were examined in cultured
colonocytes from male and female Sprague-Dawley rats (Mohamadin et al., 2005). These cells
were exposed to AN in the concentration range of 0.1-2.0 mM for 60 minutes or incubated with
1.0 mM (the concentration that reduced viability by 50%) for different time intervals up to
180 minutes and then assayed for LDH leakage, cellular GSH levels, and lipid peroxidation.
Cell viability was reduced (assessed by trypan blue exclusion method) after 60-minute exposure
to 0.5 mM AN and higher; incubation with 1.0 mM reduced viability as early as 60 minutes and
by 72% at 180 minutes. Concentration-dependent increases in plasma membrane damage, as
assessed by leakage of LDH, and enhanced lipid peroxidation (production of thiobarbituric acid-
reactive substances [TBARS]) were observed following 60-minute exposures at all levels (0.1-2
mM); exposure to 1.0 mM AN produced a 2.5-fold increase in leakage after 60 minutes.
Significant reductions in glutathione levels resulted from 60-minute exposure to >1.0 mM
AN. The time course study revealed that, by 30 minutes, exposure to 1.0 mM AN caused a
significant decrease in GSH concentration that kept decreasing until 180 minutes, when the
experiment was terminated. In additional experiments, AN-induced membrane damage and lipid
peroxidation were reduced, but not totally abolished, by cotreatment with thiol-containing
compounds (GSH, N-acetyl-L-cysteine (NAC), and dithiothreitol (DTT)), antioxidant enzymes
(SOD and catalase), as well as the iron chelator desferoxamine (DFO) and the hydroxyl radical
scavenger DMSO. Mohamadin et al. (2005) suggested that the observed protective effects of
GSH, NAC, and DTT could be attributed to direct interaction with ROS, direct binding to toxic
metabolites, and/or enhancement of cellular GSH synthesis, while depletion of iron by DFO
could indirectly prevent cell damage by inhibiting the generation of hydroxyl radical.
Pretreatment of colonocytes with either SOD or catalase inhibited LDH leakage by about 23 and
54%), respectively, when compared with colonocytes treated with AN alone. These antioxidant
enzymes reduced TBARS production that was induced by AN by 17 and 45%>, respectively.
Since these antioxidants could not restore the normal level of LDH leakage or TBARS
production, Mohamadin et al. (2005) concluded that in addition to lipid peroxidation, other
factors contributed to AN-induced cytotoxicity.
Mahalakshmi et al. (2003) evaluated the potential protective effect of taurine (TAU)
against AN-induced oxidative stress in rat brain. Male Wistar rats (six/group) were exposed to
0 or 100 ppm AN (average intake 8-10 mg/kg-day) in drinking water for 14 or 28 days.
Additional groups of rats received TAU (10 g/kg in diet) alone or along with AN treatment for
14 or 28 days. AN had no effect on BW gain or weights of liver or brain. AN treatment for
14 days increased levels of TBARS by about 13%> in plasma and 30%> in brain and increased
levels of lipid hydroperoxides by 38%> in plasma and 31%> in brain; values after 28 days were
similar, except for a 45%> increase in lipid hydroperoxides in plasma. AN exposure also
204
DRAFT - DO NOT CITE OR QUOTE
-------
increased the percentage of DNA fragmentation detectable in brain by 60 and 81% after 14 and
28 days, respectively. TAU completely prevented the AN-induced increases in levels of TBARS
and lipid hydroperoxides in plasma and brain, but afforded only partial protection from
AN-induced increases in DNA fragmentation, which was significantly increased to 30 and 19%
after 14 and 28 days, respectively. These results demonstrated that, even as TAU completely
prevented the formation of ROS and oxidative stress, DNA fragmentation still occurred. Thus,
other factors in addition to oxidative stress caused DNA fragmentation as a result of AN
exposure.
AN significantly reduced activity levels of enzymatic antioxidants after 14 days (28-day
values were similar to 14-day values but slightly lower): SOD (50% lower in hemolysate, 30%
lower in brain), catalase (50% in hemolysate, 62% in brain), glutathione peroxidase (39% in
hemolysate, 60% in brain), and GST (20% in hemolysate and brain). TAU also partially
protected against these AN-induced reductions of enzymatic antioxidant activities, resulting in
reductions of less than 11% in rats treated with AN and TAU for 14 days when compared with
controls. For rats treated with AN and TAU for 28 days, there was insignificant reduction of
<5%. AN exposure for 14 days also lowered levels of nonenzymatic antioxidants, such as
ascorbic acid (by 30% in plasma and brain), a-tocopherol (by 40% in plasma and brain), and
GSH (by 45% in plasma and 28% in brain); reductions were slightly greater for the 28-day
exposure. In animals treated with TAU and AN for 14 days, the AN-related reductions in
nonenzymatic antioxidants were less than 10% compared with controls. These experiments
demonstrated that oral exposure to AN in drinking water increased oxidative stress in the brain
and that TAU partly protected against AN-induced oxidative stress by increasing the activities of
enzymatic antioxidants and replenishing nonenzymatic antioxidants.
However, Carrera et al. (2007) were not able to reproduce the results obtained by
Mahalakshimi et al. (2003) when oxidative stress parameters were measured in Wistar rats
treated with AN in vivo. Male Wistar rats (12/group) were treated with 0 or 200 ppm AN in
drinking water for 14 days. The estimated daily dose was 30 mg/kg-day, using water factor of
0.15 L/kg-day (USEPA, 1988). Brains were excised and homogenized, and lipid peroxidation
(as measured by nmol MDA/mg protein), catalase activity, GSH levels, and proteins in brain
tissue were measured. No differences were found in lipid peroxidation products, catalase
activity, and reduced and oxidized GSH levels in the control and treated rats. Carrera et al.
(2007) concluded that there was no evidence of oxidative damage in the brain of AN-treated rats
at the studied dose of AN.
In another study (El-Sayed et al., 2008), the effect of hesperidin (HES), an antioxidant
flavonoid, on AN-induced oxidative stress in rat brain was investigated. Male Swiss rats
(8/group) were treated with 50 mg/kg-day AN via oral gavage for 28 days (Group II). Control
rats received distilled water (Group I). Group III rats received 200 mg/kg-day HES i.p. for 28
days. Group IV rats received 200 mg/Kg-day HES i.p. 24 h before starting AN treatment and
205
DRAFT - DO NOT CITE OR QUOTE
-------
concomitantly with AN treatment. At study termination, brain GSH content, MDA content, and
enzymatic antioxidant parameters: SOD, CAT, glutathione peroxidase (GSH-Px), and GST were
measured. Histopathological examination was conducted on brain samples from two rats in each
group.
Brain lipid peroxides levels measured as MDA was increased by 107% in the AN-treated
rats, accompanied by a 63% decrease in brain GSH content as compared with controls (El-Sayed
et al., 2008). On the other hand, pretreatment of rats with HES prior to AN administration
resulted in 55% reduction in brain MDA content and 183%) increase in brain GSH content when
compared with the AN-treated group. In addition, significant decreases in enzymatic antioxidant
parameters were found in the brain of AN-treated rats, with SOD, CAT, GSH-Px, and GST
decreased by 43%, 64%, 52%, & 43%, respectively, when compared with controls. Pretreatment
with HES and coadministration with AN attenuated the reduction of enzymatic antioxidants
levels, resulting in elevations of SOD, CAT, GSH-Px, and GST by 73%, 169%, 197%, and 71%,
respectively, when compared with the AN-treated group.
Histopathological examination of brain sections indicated damage to neuronal cells of the
brain in AN-treated rats, as manifested by edema and interstitial neuronal atrophy with
perineuronal vacuolation. Pretreatment of the rats with HES nearly normalized the
histopathological changes induced by AN. This study indicated that treatment of rats with 50
mg/Kg-day AN via oral gavage induced oxidative stress in the brain, and that pretreatment with
the antioxidant HES might have protective role against AN-induced oxidative stress in the brain.
Zhang et al. (2002) studied the mechanisms by which AN induced oxidative stress and
found evidence that CYP450 metabolism is required for AN to affect the activities of antioxidant
enzymes, such as catalase, xanthine oxidase, or SOD. Syrian hamster embryo (SHE) cells were
incubated for 4, 24, and 48 hours with subcytolethal doses of AN (0, 25, 50, or 75 ng/mL), and
the effects of AN on enzymatic and nonenzymatic antioxidants were monitored. All three
concentrations of AN increased ROS (hydroxyl radicals, measured as 2,3-dihydroxybenzoic acid
formation from salicylic acid) levels in SHE cells at all time points, with concurrent depletion of
GSH after 4 hours of treatment, ranging from 80 to 66% reduction. GSH levels in all AN-
treatment groups returned to control values after 24 hours of treatment and increased only in the
SHE cells treated with 75 ng/mL AN after 48 hours of treatment.
Inhibition of the antioxidant enzymes was temporal. Decreased catalase activity was
observed following treatment with 50 and 75 [j,g/mL AN for 4 hours at 72 and 52%, respectively.
However, catalase activity was increased with 25, 50, and 75 ng/mL AN treatment for 24 hours
by 82, 138, and 182%), respectively. Similar increases were observed with 48 hours of treatment.
SOD activity was decreased only in SHE cells treated with 75 [j,g/mL AN for 4 hours.
The activity of xanthine oxidase, the enzyme that generates the superoxide radical and
hydrogen peroxide via the oxidation of hypoxanthine or xanthine by oxygen, was also
monitored. Xanthine oxidase activity was increased by 47% in SHE cells treated with 75 [j,g/mL
206
DRAFT - DO NOT CITE OR QUOTE
-------
AN for 24 hours. After 48 hours of treatment, both 50 and 75 p,g/mL AN significantly increased
xanthine oxidase activity. The addition of 0.5 mM of ABT (a suicide inhibitor of CYP450) to
the system prevented the AN-induced decrease in catalase activity after 4 and 24 hours.
Cotreatment of AN with ABT also blocked the increase in xanthine oxidase in SHE cells after
48 hours of treatment. ABT alone had no effect on catalase or xanthine oxidase activity. In
addition, AN had no effect on catalase or SOD activities in the absence of metabolic source
(SHE cell homogenate). Thus, AN-induced oxidative stress in SHE cells involved decrease of
antioxidants and activation of the oxidant enzyme xanthine oxidase. Taken together, these data
suggest that AN induced oxidative stress via its oxidative metabolism.
Nerland et al. (2003) proposed another mechanism by which AN might induce oxidative
stress in their study on the covalent binding site of AN to rat liver CAIII. Two-dimensional
polyacrylamide gel electrophoresis and autoradiography were used to locate proteins in male rat
liver cytosol that were radiolabeled after s.c. administration of 115 mg/kg of [2,3-14C]-AN to
male Sprague-Dawley rats. The intensely labeled spots in the autoradiogram were identified as
CAIII. Analysis of the trypic fragment established that only cysteine 186 in the CAIII was
labeled. Thus, AN selectively bound to the cysteine 186 residue of CAIII in rat liver. Nerland et
al. (2003) also showed that over 50% of rat CAIII had participated in scavenging the toxicant
AN. CAIII has been proposed to protect cells against oxidative stress by scavenging reactive
xenobiotics, thereby reducing covalent binding to more critical macromolecules. Thus, AN
exposure would impair this protective function by covalently binding to cysteine 186 of CAIII.
In another study, the cytotoxicity of AN was related to disturbances in intracellular ionic
homeostasis and induction of oxidative stress. Mikhutkina et al. (2004) investigated the role of
2+
disturbance in cell Ca homeostasis in AN-induced blebbing of thymocyte plasma membrane
and apotosis. Blebbing of cell membrane develops in the initial stage of cell damage and is a
sign of apoptosis and necrosis. A component of apoptogenic and necrogenic factors on the cell
2+ 9+
is Ca imbalance, a result of disturbed activity of ion channels and intracellular Ca stores.
2+ 2+
Exhaustion of intracellular Ca stores increases the activity of store-activated (capacitance) Ca
2+ 2+
channels that allow influx of Ca into cells. Exhaustion of Ca stores is a stimulus for
apoptosis.
When exposed to 5 mM AN in vitro for 1 hour, thymocytes isolated from male albino
mice exhibited twofold increases in the incidence of blebbing of the plasma membrane, followed
by apoptosis (as detected by the expression of phosphatidylserine on the outer membrane) and
necrosis (as indicated by membrane permeability to propidium iodide) (Mikhutkina et al., 2004).
The initial and terminal blebbing of the plasma membrane peaked by the 15th and 45th minutes of
incubation. The dynamics of terminal blebbing correlated with the accumulation of MDA in the
incubation medium. Cotreatment with compounds (e.g., caffeine and procaine) that regulate
2_|_
activity of intracellular Ca stores modulated the cytotoxic effect of AN, suggesting to
2_|_
Mikhutkina et al. (2004) that AN induced the release of Ca from the endoplasmic reticulum.
207
DRAFT - DO NOT CITE OR QUOTE
-------
Preincubation of thymoctyes with 10 [jM isoptin, a blocker of voltage-dependent ion
channels, decreased AN-induced apoptosis and necrosis but increased the number of cells in
AN-induced secondary necrosis and decreased the intensity of oxidative stress (as indicated by
the levels of MDA). This effect was related to the ability of isoptin (an antioxidant) to suppress
AN-induced generation of free radicals. Cotreatment of AN with 10 |iM SKF 96365, a blocker
of calcium release-activated channels that were dependent on the activity of voltage-dependent
ion channels, increased the apoptogenic, but not the necrogenic, activity of AN. (SKF96365 had
low prooxidant activity.) The cells treated with SKF 96365 and AN developed intensive
blebbing at various periods of incubation. Mikhutkina et al. (2004) interpreted the results to be
+ 9+
that isoptin, via K channels, only partially inhibited store-activated Ca entry into cells through
calcium release-activated channels and in turn, protected cells from apoptosis. However,
secondary necrosis (the process associated with cells unable to form apoptotic bodies) developed
in thymocytes with suppressed blebbing. On the other hand, direct inhibition of these channels
2+
with SKF 96365 completely blocked Ca entry and promoted progression of apoptosis. The
results of these experiments suggested that AN caused disturbances in intracellular calcium
balance that led to plasma membrane blebbing, apoptosis, and necrosis of thymocytes.
A recent study on the effects of AN on primary-cultured astrocytes from Wistar male rats
(Carrera et al., 2007) indicated that AN-induced cellular damage was not due to oxidative stress,
since antioxidants did not prevent AN-induced toxicity. In this study, primary cultured rat
astrocytes were treated for 1 hour with or without trolox (TRX) (100 |iM), TAU (5 mM), NAC
(20 mM), estradiol (10 [xM), or melatonin (MEL) (10 3 M, 10 5 M, or 10 7 M). They were then
exposed to 2.5 mM AN (which induced about 40% cell death) for 24 hours. Cell viability was
determined by measuring the activity of LDH released by damaged cells into the medium. GSH
levels were also measured.
Among the antioxidants included in this study, only NAC, a sulfhydryl donor, prevented
the decrease of the number of viable cells in the culture. None of the other antioxidants
prevented cell death induced by AN treatment. Moreover, 10 5 M MEL and TAU increased the
toxicity of AN in astrocytes. Treatment of astrocytes for 4 hours with AN partially depleted
intracellular GSH, whereas pretreatment with 20 mM NAC recovered GSH content to the control
levels. Carrera et al. (2007) concluded that protective effect of NAC was due to increase in the
intracellular pool of GSH, GSH conjugation with AN, decreasing the availability of AN for
being metabolized to CEO and cyanide, halting the toxicity of AN. Carrera et al. (2007) also
concluded cellular toxicity induced by AN could not be prevented only by antioxidants.
Oxidative stress summary
The role of oxidative stress in AN-induced cytotoxicity was evaluated in both in vitro and
in vivo studies (Mohamadin et al., 2005). AN was reported to induce oxidative stress and
decreased the viability of NHAs in cell cultures (Jacob and Ahmed, 2003b) and cultured rat
208
DRAFT - DO NOT CITE OR QUOTE
-------
colonocytes (Mohamadin et al., 2005). However, other causes contributed to AN-induced
cytotoxicity as antioxidants could not restore the normal level of LDH leakage or TBARS
production.
Zhang et al. (2002) studied the mechanisms by which AN induced oxidative stress in
SHE cells and suggested that AN induced oxidative stress via its oxidative metabolism. Other
proposed mechanisms by which AN might induce oxidative stress included covalent binding of
AN to cysteine 186 of rat liver CAIII (Nerland et al., 2003), and disturbances in intracellular
ionic homeostasis (Mikhutkina et al., 2004).
Drinking water studies in Wistar rats yielded contradictory results. Mahalakshmi et al.
(2003) reported exposure to 8-10 mg/kg-day AN significantly reduced enzymatic antioxidant
(SOD, glutathione peroxidase, catalase) levels in rat brain after 14 or 28 days, and that TAU
prevented the AN-induced increases in levels of TBARS and lipid hydroperoxides in plasma and
brain. However, Carrera et al. (2007) were unable to reproduce the results obtained by
Mahalakshimi et al. (2003) in another 14-day drinking water study in Wistar rats. No differences
were found in lipid peroxidation products, catalase activity, and reduced and oxidized GSH
levels in the control and treated rats. Carrera et al. (2007) concluded that there was no evidence
of oxidative damage in the brain of rats treated with 30 mg/kg-day AN for 14 days.
In an oral gavage study of male Swiss albino rats, treatment of rats with 50 mg/Kg-day
for 28 days induced increased levels of MDA in the brain, and decreased brain GSH content and
enzymatic antioxidant levels (El-Sayed et al., 2008). Pretreatment with HES (200 mg/Kg, i.p.)
attenuated the effects of AN treatment.
4.5.1.1.5. Immunotoxicity. Zabrodskii et al. (2000) studied the mechanisms for cell and
humoral immunosuppressive effect of AN. CBA mice were treated by a single s.c. injection at
one-half the LD50 (28 mg/kg) of AN. One day after treatment, mice were tested for delayed-type
hypersensitivity (DTH) reactions: a-naphthylbutyrate esterase activity in splenocytes, number of
antibody-producing cells in spleen, and the paw edema test. AN treatment suppressed primary
cell immune response, as demonstrated by a 61% reduction of the DTH reaction (edematous paw
weight). Secondary DTH response (the number of esterase-positive splenocytes) was reduced by
28%. Humoral immune response was also decreased, as shown by a 56% reduction of the
number of antibody-producing cells. The cholinesterase reactivator, bispyridinium dioxime
(dipyroxime), rehabilitated paw swelling completely but only partially restored the numbers of
antibody-producing cells and esterase-positive splenocytes.
The role of cytochrome c oxidase a3 in the mitochondrial respiration enzyme system of
immunocompetent cells was evaluated using hydrogen cyanide, a metabolite of AN, and its
antidote anticyan. Anticyan, an agent that converts Hb to MetHb, also improved the DTH effects
of AN but less efficiently than dipyroxime. A combination of both agents reversed the
immunotoxic effects of AN completely. Zabrodskii et al. (2000) suggested that AN-induced
209
DRAFT - DO NOT CITE OR QUOTE
-------
immunotoxic effects were the result of its anticholinesterase activity targeted at T lymphocytes
and general toxicity associated with inhibition of cytochrome c oxidase a3 of the immunocytes.
4.5.1.2. Cancer Effects
4.5.1.2.1. Formation ofDNA adducts. There is evidence that AN and its epoxide metabolite,
CEO, can form a number of different DNA adducts in vitro. For example, Solomon and Segal
(1985) demonstrated the direct, nonenzymatic alkylation of calf thymus DNA by AN. Following
a 40-day incubation of AN at 37°C and pH 7.4, the reaction products were identified as
cyanoethyl adducts of guanine and thymine and carboxyethyl adducts of adenine and cytosine.
The major adducts were 1-carboxyethyl adenosine (26%), 7-cyanoethyl-guanine (26%),
imidazole ring-opened 7,9-bis cyanoethyl guanine (19%), 3-cyanoethyl thymine (16%), and
N6-cyanoethyl adenine (8%).
CEO formed a number ofDNA adducts more quickly when incubated with DNA in vitro
(Solomon et al., 1993; Yates et al., 1993; Hogy and Guengerich, 1986). When calf thymus DNA
was incubated with CEO at pH 7.0-7.5 and 37°C for 3 hours (Solomon et al., 1993),
1 o
N -(oxoethyl)guanine (110 nmol/mg DNA), N -(2-hydroxy-2-carboxyethyl)deoxyuridine (80
nmol/mg DNA), and smaller amounts of adenine and thymine adducts were produced.
Therefore, the adducts formed from CEO were different than those formed from AN. The order
of reactivity with CEO was guanine > cytosine > adenine > thymine. In addition to reacting with
n
N7 of guanine, forming N -(oxoethyl)guanine, CEO reacted to a great extent with the cytosine
"3
residue, resulting in the detection of a major adduct, N -(2-hydroxy-2-
carboxyethyl)deoxyuridine. Solomon et al. (1993) proposed that the uracil adduct was formed
from an initial cytosine adduct from hydrolytic deamination of cytosine to uracil.
Yates et al. (1993) also characterized an adduct formed when calf thymus DNA was
"3
incubated with 150 mM CEO at 37°C for 3 hours. The adduct was identified as N -(2-cyano-
2-hydroxyethyl)deoxythymidine. This adduct was also formed when 150 mM [2,3-14C]-CEO
reacted with 10 mM deoxythymidine in vitro (Yates et al., 1993). Subsequent degradation of this
"3
adduct yielded N -(2,2-dihydroxyethyl)deoxythymidine.
Similarly, Guengerich et al. (1981) showed that when 1 mM [1-14C]-AN and
[2,3-14C]-AN were incubated at 37°C with 1.5 mg/mL calf thymus DNA in the presence of rat
liver microsomes or a reconstituted CYP450 system, DNA adducts were formed as measured by
irreversible binding of radioactive label to DNA. Formation ofDNA adducts was enhanced by
the presence of NADPH. Only trace levels ofDNA adducts were detected when incubated with
rat brain microsomes, and the reaction was not NADPH dependent, probably due to insignificant
metabolism of AN by brain microsomes. Guengerich et al. (1981) also showed nonenzymatic
irreversible binding of labeled CEO to calf thymus DNA. The extent of binding for
[2,3-14C]-labeled CEO was three- to fivefold greater than that for [l-14C]-labeled CEO.
Moreover, when 100 mM CEO was incubated with 50 mM adenosine for 24 hours at 37°C,
210
DRAFT - DO NOT CITE OR QUOTE
-------
l,N6-ethenoadenosine was formed. In another experiment, an unidentified product was formed
when CEO was incubated with cytidine.
Data from Peter et al. (1983a) showed the extent of irreversible binding of AN or its
metabolites to DNA to be lower after purification of isolated DNA and RNA by column
chromatography on hydroxyapatite. In their in vitro incubation experiment, [2,3-14C]-AN was
incubated with rat-liver microsomes, NADPH and DNA or RNA. Irreversible AN binding was
found to be 3 nmol/hour per mg DNA, when ethanol precipitation or phenol extraction of DNA
was used. When the isolated DNA was further purified by column chromatography on
hydroxyapatite, irreversible binding was found to be 0.15 nmol/hour per mg DNA. Peter et al.
(1983a) also administered [2,3-14C]-AN intraperitoneally to male Wistar rats and measured the
incorporation of radioactivity into hepatic RNA bases. The amount of radiolabel from AN
associated with hepatic DNA was lower than that from labeled vinyl chloride.
However, when hepatic DNA from these treated rats was isolated and hydrolyzed,
chromatography on PEI-cellulose showed two 14C peaks that did not correspond to known
standards. Thus, Peter et al. (1983a) concluded that AN or its metabolites could alkylate DNA,
although these DNA adducts had not yet been identified.
Yates et al. (1994) also characterized the products formed when 150 mM CEO reacted
with 50 mg/mL nucleotides for 3 hours at 37°C in vitro. The reaction of CEO with
5'-monophosphates of deoxyguanosine, deoxyadenosine, deoxycytidine, or deoxythymidine
resulted in the formation of at least one adduct for each nucleotide. These CEO-nucleotide
adducts were characterized as 2-cyano-2-hydroxyethyl phosphodiesters. The reaction of
n
deoxyguanosine-5'-monophosphate (dGMP) also produced a second adduct, N -(2-cyano-
2-hydroxyethyl)-dGMP. Yates et al. (1994) suggested that the cyano-hydroxy ethyl-
phosphodiester adduct could induce single and double DNA strand breaks (as observed when
CEO was incubated with pBR322 plasmid DNA) via interaction of the adduct's P-hydroxyl-
group with the DNA phosphate backbone.
Irreversible binding of AN or its metabolites to DNA in vivo has also been studied.
Farooqui and Ahmed (1983a) investigated the ability of [2,3-14C]-AN or its metabolites to bind
to macromolecules in rats in vivo. [2,3-14C]-AN was administered in a single dose to male
Sprague-Dawley rats (three to four per group) via gavage at a dose of 46.5 mg/kg in water.
Animals were sacrificed at 1, 6, 24, and 48 hours after dosing. Organs, including liver, kidney,
brain, spleen, and stomach, were dissected out and frozen rapidly. Nucleic acids extracted from
the homogenates were applied to the hydroxyapatite column for separation of RNA and DNA
fractions. Radioactivity was detected in the extracts of RNA and DNA from liver, stomach, and
brain. Bound radioactivity (as pmol equivalent AN/mg DNA) to DNA in the brain was the
highest at 24 hours (119 pmol/mg DNA), followed by stomach (81 pmol/mg DNA), and liver
(25 pmol/mg DNA). Bound radioactivity in all three organs plateaued at 24 hours and remained
unchanged thereafter. RNA from liver showed the highest amount of radioactivity, followed by
211
DRAFT - DO NOT CITE OR QUOTE
-------
brain and stomach. Bound radioactivity in liver RNA peaked at 6 hours after dosing, whereas
RNA from the brain and stomach showed maximal radioactivity at 24 hours. Binding to proteins
was extensive and time dependent. In the first hour after oral dosing, the highest protein binding
occurred in spleen and stomach, followed by liver, kidney, and brain. After 6 hours, protein
binding plateaued until 48 hours. At 6 hours, the highest protein binding occurred in the spleen,
followed by liver, stomach, and kidney, with brain having the lowest protein binding. The study
authors developed numerical indices for the AN-derived DNA alkylation; covalent binding
indices were 5.9, 51.9, and 65.3 for liver, stomach, and brain, respectively, after 24 hours.
Ahmed et al. (1992a) demonstrated the covalent binding of radiolabel from [2,3-14C]-AN
to testicular DNA after a single oral dose of 46.5 mg/kg of [2,3-14C]-AN to male Sprague-
Dawley rats. In a time course study, bound activity was shown to be greatest after 30 minutes
(8.93 ± 0.80 [j,mol AN bound/mol nucleotide), with covalent binding index of 10.15. Using an
identical experimental protocol, Ahmed et al. (1992b) demonstrated the capacity of AN to bind
covalently to DNA in the lung. Covalent binding of radioactivity to DNA increased with time
and was maximum at 12 hours after a single oral dose, with a covalent binding index of 3.48.
Binding was associated with a 55-72% decrease in replicative DNA synthesis at time points up
to 24 hours after dosing.
Hogy (1986) and Hogy and Guengerich (1986) studied the in vivo interaction of AN and
CEO with DNA. Three male F344 rats were administered AN (50 mg/kg, i.p.); three more rats
were administered CEO (6 mg/kg, i.p.). The rats were sacrificed after 2 hours, and the brain and
livers were removed and frozen. DNA and RNA were isolated from these tissues. For detection
1 "3
of N -(2-oxoethyl)guanine, DNA samples were reductively tritiated with NaB H4; the adduct
was released by neutral thermal hydrolysis and purified by thin-layer chromatography for
n
subsequent quantitation by liquid scintillation counting. N -(2-oxoethyl)guanine was detected in
1 1
liver DNA at 3.1 x 10 nucleotides per alkylation in AN-treated rats and 6.9 x 10 nucleotides
n
per alkylation in CEO-treated rats (Table 4-52). For brain DNA, N -(2-oxoethyl)guanine was
8 9
detected at 2.4 x 10 nucleotides per alkylation in AN-treated rats and at 1.1 x 10 nucleotides
per alkylation for CEO-treated rats. The values obtained from the brain DNA samples were near
the limit of detection. Thus, the presence of these adducts in the brain could not be
unequivocally verified.
212
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-52. Detection of N7-(2-oxoethyl)guanine after i.p. administration of
50 mg/kg AN or CEO to male F344 rats
Formation of N7-(2-oxoethyl)guanine
Liver
Brain
Compound
fmol/mg DNA
Nucleotides/alkylation
fmol/mg DNA
Nucleotides/alkylation
AN
109 ±71
3.1 x 107
14
2.4 x 108
CEO
48 ± 15
6.9 x 107
3
1.1 x 109
Source: Hogy(1986).
In the same study (Hogy, 1986; Hogy and Guengerich, 1986), rat DNA samples were
also analyzed by HPLC with a fluorescence detector for l,N6-ethenoadenosine and
l,N6-ethenodeoxyadenosine. These adducts were not detected with limits of detection estimated
to be 3 pmol/mg DNA and 1 pmol/mg RNA, respectively.
In another study, Prokopczyk et al. (1988) administered 50 or 100 mg/kg s.c. AN to male
F344 rats (10/group). DNA was isolated from liver and brain after 2 hours (50 mg/kg group) or
6 hours (100 mg/kg AN). An HPLC assay with fluorescence detector was used to detect
7-(2-cyanoethyl)guanine (detection limit: 1 per 5 x 104 guanine) and 06-cyanoethylguanine
(detection limit: 1 per 7 x 104 guanine). Neither adduct was detected. These two adducts were
not formed when CEO was incubated with calf thymus DNA in vitro (Solomon et al., 1993).
4.5.1.2.2. Oxidative stress. Jiang et al. (1998) evaluated the ability of AN to induce oxidative
stress in the brain cortex of rats. Male Sprague-Dawley rats (at least nine/group) were exposed
to 0, 5, 50, 100, or 200 ppm AN in drinking water for either 14, 28, or 90 days. These
concentrations were selected from those used in chronic bioassays (see Section 4.2.1.2). As
calculated by the study authors, average daily doses of 0, 0.6, 5.1, 8.9, or 15.0 mg/kg-day were
ingested over the 90-day treatment period by the control to high-dose groups, respectively. At
study termination, brains and livers were weighed. The livers and brain cortexes from six
rats/group were evaluated for the following oxidative endpoints: oxidative DNA damage
(8-hydroxy-2'-deoxyguanosine levels), lipid peroxidation (MDA levels), levels of nonenzymatic
antioxidants (glutathione and vitamin E), and the activities of enzymatic antioxidants (catalase,
SOD, and glutathione peroxidases). 8-Hydroxy-2'-deoxyguanosine has also been referred to as
8-oxo-7,8-dihyro-2'-guanosine (Murata et al., 2001) and 8-oxodeoxyguanosine (Whysner et al.,
1998a) and will be referred to as 8-oxodG in this document. ROS formation—as measured by
the formation of 2,3-dihydrobenzoic acid from salicylic acid in brain cortex and liver—were
evaluated in three rats/group injected with salicylic acid in saline 12 hours before termination.
In rats exposed to AN at concentrations as high as 200 ppm for up to 90 days, no effects
were noted on viability. A statistically significant reduction of 9% in BW was observed in the
200 ppm group after 90 days. No differences were observed in brain or liver weights nor were
213
DRAFT - DO NOT CITE OR QUOTE
-------
any of the measures of oxidative damage or antioxidant levels in the liver for all groups at any
sampling times. AN exposure resulted in statistically significant increases in oxidative stress
parameters and decreases in antioxidants levels in the brain cortex. Levels of hydroxy free
radicals (ROS) were significantly elevated in a dose-related manner in the 50-200 ppm groups,
beginning at 14 days and persisting until the 90-day time point; at 200 ppm, the increase was
approximately fivefold over controls. Levels of 8-oxodG in the cellular DNA of brain cortex
were significantly elevated (three- to fourfold compared with controls) in a dose-related manner
in the 100 and 200 ppm groups, beginning at 14 days of exposure, and two- to threefold in the
50 ppm group, beginning after 28 days of exposure. MDA levels were significantly increased
(1.5-fold) in the 200 ppm group after 14 days but at no other time point.
Slight but significant decreases in levels of vitamin E (about 20%) and glutathione (about
20%) in the brain cortex were observed in the 50-200 ppm groups but were statistically
significant only at the 14-day time point. Statistically significant, dose-dependent reductions (by
up to 60%) in catalase activity were observed in brain cortex at all exposure levels after 14 days,
at >100 ppm after 28 days, and at >50 ppm after 90 days; the transient effect at 5 ppm was not
considered to be biologically significant. Reductions in SOD levels in the brain (by up to 30% at
200 ppm) were dose related and statistically significant at >50 ppm after 14 days and at 200 ppm
after 28 or 90 days. In this study, a NOAEL of 5 ppm (0.6 mg/kg-day) and a LOAEL of 50 ppm
(5.1 mg/kg-day) were identified for increases in levels of oxidative damage (increased ROS and
DNA damage) and reductions in the antioxidant enzymes catalase and SOD in the brain of rats
exposed to AN in drinking water. Jiang et al. (1998) suggested that observed oxidative stress in
rat brain cortex following AN treatment could be induced by: (1) direct generation of free
radicals from AN or its metabolites, (2) binding of AN or its metabolites to free radical
scavengers (e.g., GSH, vitamin E), (3) modulation of the activity and/or synthesis of antioxidant
enzymes (e.g., SOD, catalase), and/or (4) interference with electron flow through the respiratory
chain via inhibition of cytochrome C oxidase by the cyanide ion, a metabolite of AN. However,
these potential mechanisms need to be further investigated.
In a follow up study, Pu et al. (2009) examined the potential for AN to induce oxidative
DNA damage in rats. These investigators also examined whether blood could serve as a
surrogate for the biomonitoring of oxidative stress induced by AN in target tissues (in particular
brain) of exposed populations. Male Sprague-Dawley rats (9/group) were treated with 0, 3, 30,
100 or 200 ppm AN in drinking water for 28 days. N-acetyl cysteine (NAC), an acetylated
precursor of glutathione, was coadministered at a dietary concentration of 0.3% to one group of
rats receiving 200 ppm AN in drinking water to evaluate its protective effect against potential
AN-induced oxidative stress. At the end of treatment, animals were sacrificed and blood
samples were collected immediately. The alkaline comet assay was used as a measure of direct
DNA damage in 3 rats/group, and the formamidopyrimidine DNA glycosylase (fpg)-modified
comet assay was used as a measure oxidative DNA damage in brain cortex and white blood cells
214
DRAFT - DO NOT CITE OR QUOTE
-------
(WBCs) of different treatment groups. 8-oxodG levels in brain tissues and WBCs were also
measured by HPLC with electrochemical detection. 2,3-DHBA was measured in WBCs and
brain tissues for the presence of ROS.
Pu et al. (2009) observed no increase in DNA damage in rat WBC and brain tissue
following exposure to AN with or without NAC coadministration using the alkaline comet assay.
Dose-dependent increases in DNA damage were observed in rat WBC and brain following
exposure to AN using the fpg-modified alkaline comet assay. These increases in DNA damage
were not observed in WBC and brain of rats exposed to 200 ppm AN and NAC. Pu et al. (2009)
interpreted the results as indicative of oxidative DNA damage in AN-exposed rats, and further
that the absence of oxidative DNA damage in rats coadministered NAC reflected the antioxidant
action of NAC. Significant increases in the level of 8-oxodG were found in WBC and brain of
rats treated with 100 and 200 ppm AN; no increase was observed in rats treated with 200 ppm
AN and NAC. 2,3-DHBA, a measure of ROS formation, was increased in the WBC of rats
treated with AN in a dose-dependent manner, but was not detectable in the brain of treated rats.
In addition, the ratios of reduced to oxidized glutathione (GSH/GSSG) were reported to be
significantly lower in rats treated with 30, 100 and 200 ppm AN, but not in a dose dependent
manner. Pu et al. (2009) concluded that AN induced oxidative stress and DNA damage in male
Sprague-Dawley rats and that the fpg-modified comet assay in WBC correlated with target tissue
oxidative DNA damage.
EPA identified certain issues with study design and choice of analytical methods that
raise questions about the interpretation of the findings by Pu et al. (2009). First, Pu et al. (2009)
stated that direct DNA damage was not detected in rat WBC and brain of AN-exposed rats based
on results in the standard alkaline comet assay. The standard comet assay detects single and
double strand breaks, cross link, oxidative DNA damage, apurinic/pyrimidinic sites and DNA
repair (Smith et al., 2006; Collins, 2007). The induction of other types of DNA damage not
detected by the comet assay cannot be precluded.
Second, Pu et al. (2009) utilized the fpg-modified comet assay to detect oxidative DNA
damage induced by AN; however, this assay is not specific for the detection of oxidative DNA
damage, as this assay also detects alkylation DNA damage (Speit et al., 2004; Smith et al., 2006).
The fpg-modified comet assay has been shown to be especially sensitive for the detection of
DNA damage by N-7 guanine alkylation, which was responsible for the observed DNA damage
by alkylating agents methylmethanesulfonate (MMS) and ethylmethanesulfonate (EMS) (Speit et
al., 2004). N-7 guanine adduct was detected in liver and brain (at the limit of detection) of AN-
treated rats (see Section 4.5.1.2.1). Thus, the observed DNA damage using the fpg-modified
comet assay in Pu et al. (2009) cannot necessarily be attributed to oxidative DNA damage, but
could also be due to N-7 guanine alkylation.
Third, NAC was used by Pu et al. (2009) to demonstrate the effect of an antioxidant to
protect against observed DNA damage. It should be noted that NAC is a precursor of
215
DRAFT - DO NOT CITE OR QUOTE
-------
glutathione. Glutathione conjugates with AN to N-acetyl-S-(2-cyanoethyl)cysteine, which can
then be excreted in the urine. Glutathione conjugation is therefore a detoxification pathway for
AN. Thus, the reduction in DNA damage in AN-treated rats coadministered with NAC observed
by Pu et al. (2009) may, in fact, reflect increased detoxification of AN and reduced availability
of AN for oxidation to CEO. The protective effect of NAC on AN toxicity was demonstrated by
Carrera et al. (2007) (see Section 4.5.1.1.4). Fourth, Pu et al. (2009) reported that the ratios of
reduced to oxidized glutathione (GSH/GSSH) were decreased in the brain cortex of rats treated
with AN in drinking water at concentrations of 30, 100, and 200 ppm for 28 days. This
reduction was not dose related; the GSH/GSSG ratio was similar in all three exposure groups
[Figure 8 in Pu et al. (2009)]. GSH/GSSG provided the redox status in the brain cortex. It
should be noted that in the 2-year drinking water study in Sprague-Dawley rats involving AN
drinking water concentrations of 30, 100, and 300 pmm (Quast, 2002), the incidence of tumors
increased with increasing exposure. The absence of a dose-related decrease in GSH/GSSG ratio
observed by Pu et al. (2009) does not parallel the dose-related increase in tumor incidence
observed in AN-exposed rats by Quast (2002). Thus, the GSH/GSSG ratio is not concordant
with an oxidative stress mode of action for the carcinogenicity of AN. Fifth, no ROS was
detected in the brain of AN exposed rats in Pu et al. (2009), as no 2,3-DHBA was detected in the
brain of AN-treated rats.
Whysner et al. (1998a) also evaluated oxidative DNA damage (increased levels of
8-oxodG) in the brains of male Sprague-Dawley and F344 rats exposed to AN in drinking water
for durations of either 21 or 94 days. In the first experiment, male Sprague-Dawley rats
(20/group) were exposed to 0, 3, 30, or 300 ppm AN in drinking water for 21 days (Whysner et
al.,1998a). Based on default values for the BW of male Sprague-Dawley rats in a subchronic
study (0.267 kg) and an allometric equation linking water consumption to BW (U.S. EPA, 1988),
the average intakes of AN were calculated as approximately 0, 0.43, 4.3, and 43 mg/kg-day for
the control to high-dose groups, respectively. At termination, brains, livers, and forestomachs
were excised from five rats/group each for DNA isolation. These tissues were analyzed for
GSH, cyst(e)ine, and 8-oxodG levels in nuclear DNA. TBARS levels in brains were determined
for another five rats/group. Brain homogenate from another five rats/group were analyzed for
cytochrome oxidase activity (in the mitochondria fraction), catalase (in supernatant), and
glutathione peroxidase (in homogenate). Tissues from another five rats/group were examined for
histopathology, but results were to be published separately.
In rats exposed to 30 or 300 ppm AN, statistically significant increases in 8-oxodG levels
in nuclear DNA were measured in brains (approximately twofold increase compared with control
values for both dose groups) and livers (approximately 1.4-fold increase for both dose groups
when compared with controls). Also observed in the 300 ppm dose group were approximately
twofold increases in the levels of glutathione and cyst(e)ine in the forestomach, and
approximately 50% increase in cyst(e)ine level in the brain. There were no exposure-related
216
DRAFT - DO NOT CITE OR QUOTE
-------
effects on levels of glutathione in the brain and liver or on levels of cytochrome oxidase,
catalase, glutathione peroxidase, or TBARS in the brain. However, glutathione level in the
forestomach was increased about 1.8-fold in the 300 ppm dose group compared with controls. In
this study, a NOAEL of 3 ppm (0.43 mg/kg-day) and a LOAEL of 30 ppm (4.3 mg/kg-day) were
identified for increased oxidative DNA damage in the brains and livers of Sprague-Dawley rats.
Whysner et al. (1998a) concluded that the formation of 8-oxodG from AN exposure did not
involve disruption of antioxidant defense or lipid peroxidation. In addition, the absence of effect
on brain cytochrome oxidase activity in exposed rats indicated lack of inhibition by cyanide, a
metabolite of AN. Thus, Whysner et al. (1998a) concluded that cyanide-induced metabolic
hypoxia did not appear to be involved in the generation of ROS by AN administered in drinking
water.
Whysner et al. (1998a) conducted a parallel 21-day drinking water study in male F344
rats (10/group). Concentrations of AN in drinking water were of 0, 1, 3, 10, 30, or 100 ppm.
Based on default values for the BW of male F344 rats in a subchronic study (0.18 kg) and an
allometric equation linking water consumption to BW (U.S. EPA, 1988), the average doses of
AN were estimated at 0, 0.16, 0.47, 1.6, 4.7, and 15.6 mg/kg-day in the control to high-dose
groups. An additional group of rats received 5 mg methylnitrosourea (MNU) per kg/week via
i.v. injection. At termination, rat brains were evaluated for levels of 8-oxodG in nuclear DNA,
cytochrome oxidase, glutathione, and cyst(e)ine. Levels of all these parameters in the brains of
AN-exposed rats and MNU-exposed rats were not significantly different from controls. The
levels of 8-oxodG in the brains of 3-100 ppm AN dose groups were the same, about 1.3-fold
higher than control values. However, the increases were not statistically significant. The highest
drinking water concentration of 100 ppm AN (15.6 mg/kg-day) was a NOAEL for oxidative
effects in the brains of F344 rats (i.e., oxidative stress was not demonstrated in the brains of F344
rats even at the highest dose of 15.6 mg/kg-day).
In the subchronic experiment by Whysner et al. (1998a), male Sprague-Dawley rats
(10 rats/group) were exposed to 0 or 100 ppm AN in drinking water for 3, 10, 31, or 94 days.
Two additional groups of Sprague-Dawley rats (six/group) were exposed to 5 mg MNU/kg/week
or 5 mg MNU/kg/week +100 ppm AN. Brains were assayed for levels of 8-oxodG, cytochrome
oxidase, and glutathione. Levels of 8-oxodG were also measured in the livers of rats exposed for
3, 10, or 94 days.
No effects on brain levels of cytochrome oxidase or glutathione were observed in any
dose group up to 94 days. The 8-oxodG levels were significantly increased by 77% following
exposure to 100 ppm AN for 3, 10, and 94 days. Administration of 5 mg/kg MNU (a DNA-
reactive carcinogen that produces glial cell tumors in rats) did not increase the level of 8-oxodG
in the brain of treated rats but increased 8-oxodG in the liver after 10 days. However,
coadministration of 100 ppm AN and MNU increased the 8-oxodG level in the brain after 31 and
94 days when compared with controls. Whysner et al. (1998a) proposed that AN-induced
217
DRAFT - DO NOT CITE OR QUOTE
-------
generation of ROS and resultant oxidative DNA damage represented one possible mode of action
for the neoplastic process in the rat brain. However, as discussed in Section 4.7.3.3.1, results
from this study did not support the proposed oxidative stress mode of action.
The ability of AN to induce oxidative stress and oxidative DNA damage was also studied
in a rat glial cell line and cultured rat hepatocytes in vitro (Kamendulis et al., 1999a). In parallel
experiments, DITNC1 rat astrocytes and rat hepatocytes were incubated with sublethal
concentrations (up to 1 mM) AN in vitro for 4 or 24 hours. The 8-oxodG levels in total cellular
(nuclear and mitochondrial) DNA, generation of ROS (measured as increase in generation of
2,3-dihydroxybenzoic acid in salicylic acid), levels of the lipid peroxidation product MDA, GSH,
and antioxidant enzymes activity were measured at the end of the incubation period.
In rat astrocytes, significant increases (up to 3.8-fold) in the production of 8-oxodG over
control was observed after 4 hours and up to 3.9-fold after 24 hours. No increase in 8-oxodG
formation was observed in rat hepatocytes at all AN concentrations and time examined. The
induction of 8-oxodG by AN in rat astrocytes was reversible. Following removal of AN after
24 hours of treatment, 8-oxodG levels returned to control values at all studied concentrations in
24 hours. Intercellular production of ROS was found to be increased 2- to 2.6-fold after 4 and
24 hours at 0.1 and 1.0 mM AN. No increase in ROS generation was found in rat hepatocytes at
any AN concentration or exposure duration. On the other hand, no significant change in MDA
formation (as indicator of lipid peroxidation) was found in either cell type following treatment
with AN.
A significant decrease in cellular GSH levels was observed in rat astrocytes treated with
0.1 and 1.0 mM AN for 4 hours (25-36% of control) and 24 hour (43-61% of control) and in
SOD activity in astrocytes treated for 4 hours with 1 mM AN (39% reduction over control) and
for 24 hours with 0.1 and 1 mM AN (38-40% reduction over control). In contrast, no significant
decrease in catalase and glutathione peroxidase activities was observed in treated astrocytes.
Cotreatment with L-2-oxothiazolidine-4-carboxylic acid (OTC), a precursor to GSH
biosynthesis, or with vitamin E, an antioxidant, reduced 8-oxodG and ROS formation induced by
AN treatment. These effects were not evident in isolated hepatocytes treated with AN. Results
from this study were in agreement with results from the in vivo study by Jiang et al. (1998) on rat
brain cortex. Kamendulis et al. (1999a) suggested that, since the formation of 8-oxodG and ROS
observed after AN treatment in this study was temporal, dose dependent, and reversible
following removal of AN in the culture medium, these are established properties of tumor-
promoting agents. Kamendulis et al. (1999a) proposed that AN-induced astrocytomas in rats
were produced via tumor promotion mechanisms.
Jacob and Ahmed (2003b) also demonstrated the ability of AN to induce oxidative stress
and oxidative DNA damage in NHA culture (see Section 4.5.1.1.3).
Pu et al. (2006) measured direct and oxidative DNA damage in cultured D1 TNC1 rat
astrocytes treated with 0-2.5 mM AN for 24 hours. Direct DNA damage was measured by the
218
DRAFT - DO NOT CITE OR QUOTE
-------
comet assay, and oxidative DNA damage was measured with a modified comet assay that
included enzymatic digestion with formamidopyrimidine glycosylase. 7-Ethoxyresorufin-O-
deethylase (EROD) and CYP2E1 activities in the astrocytes were measured.
At 2.5 mM AN, a 40% decrease in cell viability was observed. No increase in direct
DNA damage (measured as increase in tail moment) was observed at any AN concentration.
Hydrogen peroxide (20 |iM) was used as positive control, and significant increase in DNA
damage was found. On the other hand, a threefold increase in oxidative DNA damage was
observed in astrocytes treated with 1 mM AN. Supplementation of 1 mM AN with three
different antioxidants—vitamin E (150 (jM), TRX (water-soluble analog of vitamin E, 150 [xM),
and epigallocathechin-3 gallate (a polyphenol from green tea, 5 |iM)—reduced AN-induced
oxidative DNA damage by 65, 54, and 65%, respectively.
As expected, only low level CYP2E1 activity was measured in the astrocytes and was
about 10%) of that measured in mouse liver. EROD (an indicator of CYP1A) activity was
measured in rat astrocytes and was about 500-fold higher than that for CYP2E1. In addition, a
90% reduction in the activities of EROD and CYP2E1 was observed when the astrocytes were
cotreated with AN and 0.5 mM ABT for 24 hours. Cotreatment of astrocytes with 1 mM AN and
0.5 mM ABT also prevented the increase in oxidative DNA damage induced by AN, indicating
the metabolism of AN by P450 was required for the production of oxidative stress. GSH
depletion induced by 4 and 24 hours of treatments with DL-buthionine [S,R]-sulfoximine, a
selective inhibitor of y-glutamylcysteine synthetase, enhanced the oxidative DNA damage
induced by AN by 44-160%) over control). On the other hand, cotreatment with 2.5 mM OTC, a
precursor for glutathione biosynthesis, reduced AN-induced oxidative DNA damage by 63-85%).
Exposure to 0.1-0.5 mM cyanide also increased oxidative DNA damage (Pu et al., 2006).
EPA has identified issues in the study design, analytical methods, and data interpretation
in Pu et al. (2006). First, as mentioned previously in the discussion of Pu et al. (2009), the
alkaline comet assay measures DNA strand breaks and alkali-labile sites, and is not limited to the
detection of direct DNA damage only. For the alkaline comet assay, Pu et al. (2006) used
hydrogen peroxide as a positive control, demonstrating that this assay is not limited to detecting
direct DNA damage, but also oxidative DNA damage. Second, Pu et al. (2006) used the fpg-
modified comet assay for detection of oxidative DNA damage. As discussed previously, the fpg-
modified comet assay is not specific for the detection of oxidative DNA damage. Third,
CYP2E1 level in astrocytes was known to be low and was found to be about 10%> that found in
the mouse liver. Since bioactivation of AN to CEO via CYP2E1 is needed for direct DNA
damage to be detected, it is not surprising that direct DNA damage could not be detected in
astrocytes treated with AN (although oxidative DNA damage could have been detected in the
alkaline comet assay). As discussed previously in Section 3, CEO is a relatively stable reactive
metabolite, with a half-life of about 2 hours; therefore, most of the CEO detected in rat brain is
presumed to be formed in the liver of rats treated with AN and transported to the rat brain.
219
DRAFT - DO NOT CITE OR QUOTE
-------
Hence, direct DNA damage could not be detected in this in vitro study. Another weakness in
this study concerns the selection of the positive control. A carcinogen that needs bioactivation
via CYP2E1 to form a reactive metabolite and induce DNA damage would have served as a
better positive control than hydrogen peroxide, which does not require bioactivation. Therefore,
the findings reported in this in vitro study are not anticipated to predict what can occur in vivo.
Murata et al. (2001) investigated the enhancing effect of AN on the formation of
8-oxodG in calf thymus DNA, induced by hydrogen peroxide (H2O2) and Cu(II). Calf thymus
DNA was incubated with various concentrations of H2O2 plus CuCh in the presence or absence
of AN for 30 minutes. The level of Cu(II)-mediated 8-oxodG formation increased with
increasing concentration of H2O2. The addition of AN (0.1-0.5%) enhanced the formation of
8-oxodG by hydrogen peroxide and Cu(II) in a dose-dependent manner, whereas AN itself did
not cause DNA damage. The enhancing effect of AN was more marked in double-stranded than
32
in single-stranded DNA. Further experiments with [ P]-labeled DNA showed that addition of
AN enhanced the site-specific DNA damage at guanine residues, particularly at the 5'-site of the
GG and GGG sequences while H202/Cu(II) induced piperidine-labile sites at thymine, cytosine,
and guanine residues. Electron spin resonance spectroscopy showed that a nitrogen-centered
radical was generated from AN during incubation with hydrogen peroxide and Cu(II). Murata et
al. (2001) proposed that AN enhanced H^Ch-mediated DNA damage via nitrogen-centered
radical formation. Thus, AN may enhance endogenous oxidative stress, although AN itself does
not have the ability to induce oxidative DNA damage.
4.5.1.2.3. Intercellular communication. Kamendulis et al. (1999b) investigated the effect of
AN on gap junction intercellular communication (GJIC) in D1 TNC1 astrocytes (a rat astrocyte
transformed cell line) and primary rat hepatocytes in culture. Noncytolethal concentrations of
AN (0.01-1.0 mmol/L) were incubated with D1 TNC1 astrocytes or primary rat hepatocytes.
GJIC was determined by microinjection of lucifer yellow CH into cells. Dye coupling was
quantitated by determining the number of recipient cells in contact with microinjected cells that
showed communication. The reversibility of AN inhibition of GJIC was also evaluated by
replacement of AN with fresh medium.
Following 2 hours of treatment, AN at 0.1 and 1 mmol/L inhibited GJIC in D1 TNC1
astrocytes, which were putative target cells. After treatments for 4 and 24 hours, all
concentrations of AN (0.01-1.0 mmol/L) inhibited GJIC. The inhibitory effect of AN in this
system was dose dependent at all time points, reversible by removal of AN, and partially
suppressed by the presence of antioxidants such as vitamin E (0.1 mmol/L). After treatment with
both vitamin E and AN for 24 hours, inhibition of GJIC was reduced by 23%. On the other
hand, AN did not inhibit GJIC in primary cultured hepatocytes at all concentrations and
durations of treatment.
220
DRAFT - DO NOT CITE OR QUOTE
-------
AN also caused a concentration-dependent decrease in cellular GSH content in both
D1 TNC1 astrocytes and rat hepatocytes (Kamendulis et al., 1999b). Cotreatment with
5 mmol/L OTC, a precursor of GSH synthesis, reduced inhibition of GJIC by AN in D1 TNC1
astrocytes following 4 and 24 hours of exposure, with the greatest reduction observed at
1.0 mmol/L AN (up to 68%). However, depleting GSH by L-BSO (an inhibitor of intracellular
GSH synthesis) alone without AN did not affect GJIC in rat astrocytes. Thus, depletion of GSH
alone in astrocytes was not sufficient for the observed decrease in GJIC by AN.
Inhibition of intercellular communication by AN was implied in an inhibition of
metabolic cooperation assay. Two studies (Elmore et al., 1985; Umeda et al., 1985) evaluated
AN by using the Chinese hamster V79 cells, which are 6-thioguanine (6-TG) sensitive and
6-TG-resistant cloned cells that are the hypoxanthine guanine phosphoribosyl transferase (hprt~)
mutant of the V79 cell line. The hprf mutant cannot phosphorylate several purine analogues,
including 6-TG, and are therefore resistant to cell killing by the purine analogue. When the WT
cells are cultured at a density that permits frequent contact with the mutant cells, metabolic
cooperation (i.e., gap junction formed between these cells) allows the transfer of nutrients and
phosphorylated purine analogue from the WT cells to the mutant cells and decreases probability
of recovery of the mutant cells in purine analog selective medium. The principle of the assay is
based on the fact that recovery of 6-TG-resistant cells cocultivated with 6-TG-sensitive cells in
6-TG-containing medium increased by addition of compounds that inhibit metabolic
cooperation. This assay evaluates if the test agent can modulate gap junctional communication.
AN inhibited gap junction formation slightly and dose dependently (Umeda et al., 1985).
Average recovery of the mutant cells was 22% in the control, 33% at 1 mM AN, and 39% at
2 mM AN. Umeda et al. (1985) considered AN as positive in the metabolic cooperation assay.
In the study by Elmore et al. (1985), AN produced positive responses at noncytotoxic
concentrations of 10-50 p,g/mL after incubation for 3 days.
4.5.1.2.4. Cell proliferation. Ghanayem et al. (1997) examined the effects of AN on
forestomach cell proliferation and apoptosis in male F344 rats (12/group) administered either 0,
0.22, or 0.43 mmol/kg (0, 11.67, or 22.8 mg/kg) by gavage for 6 weeks. Six rats from each dose
group were used to assess BrdU incorporation in the stomach, the remaining six rats from each
group were used to assess BrdU incorporation in hepatocytes. Proliferation of forestomach
squamous epithelial cells was evaluated with light microscopy by determining the number of
cells (nuclei) per unit length muscularis and by quantitating BrdU-stained cells.
AN was shown to induce a dose-dependent increase in epithelial cell proliferation in the
forestomach, as determined by the incorporation of BrdU into S-phase DNA. The increase in
forestomach mucosal cell proliferation was significant at both the low and high doses of AN. No
cellular proliferation was detected in the liver and glandular stomach, which are not target organs
of AN carcinogenicity.
221
DRAFT - DO NOT CITE OR QUOTE
-------
There was an associated increase in the thickness (hyperplasia) of the forestomach
squamous mucosa, also possibly indicative of enhanced cell proliferation. Forestomach
hyperplasia (expressed as the increase in the total number of epithelial cells per mm muscularis)
was significant in the high-dose group and was about 60% above vehicle-treated controls.
Hyperplasia in the low-dose group was only 7% above controls and was statistically
insignificant. The effects of AN on cellular proliferation in the forestomach of treated rats were
also evaluated by quantitative determination of BrdU incorporation into S-phase DNA, using
immunohistochemical staining. The increase in forestomach mucosal cell proliferation was
significant at both the low- and high-dose groups. In addition, the effect of AN on apoptosis was
determined by in situ end labeling of tissue sections. Apoptotic bodies were observed in the
forestomach of rats treated with high-dose AN. No increase in apoptosis was detected in the
liver or glandular stomach of control or treated rats. Thus, AN induced a significant increase in
forestomach apoptosis at the high dose, coupled with an increase in hyperplasia.
Ghanayem et al. (1997) proposed that disruption of the normal balance between cell
proliferation and apoptosis in favor of enhanced forestomach cell proliferation (as reflected by
hyperplasia of the forestomach epitheilium) probably contributed to the pathogenesis of
AN-induced forestomach tumors. This suggestion was supported by the observations that cell
proliferation in forestomach squamous mucosa of treated rats occurred at doses that caused
forestomach tumors in rats and that cell proliferation was selective and only occurred in the
target organ forestomach but not in liver.
In a recent study, Chantara et al. (2006) evaluated whether AN induced extracellular
signal-regulated kinase (ERK) activation in human neuroblastoma SK-N-SH cells. The
activation of ERKs belonging to the mitogen-activated PK family has been implicated to play
crucial roles in cell proliferations and is involved in many steps of tumor progression (Fang and
Richardson, 2005; Platanias, 2003; Seger and Krebs, 1995). Dysfunction of the ERK signaling
pathways was shown to play a pivotal role in the development of many cancers, including
leukemia and colon cancer. Active forms of ERK 1/2 were found to be dually phosphorylated at
threonine and tyrosine. To investigate whether AN could activate ERK, the effect of AN on the
activation-association phosphorylation of ERK1/2 was measured in SK-N-SH cells. Treatment
with 400 [j,g/mL AN for 1 hour increased the activation-associated phosphorylation of ERK1/2.
Further increase in ERK1/2 phosphorylation was observed with 3 and 24 hours of incubation.
Furthermore, increase in ERK phosphorylation was found to be dependent on AN concentration.
When SK-N-SH cells were treated with specific mitogen-activated/ERK-activating kinase
(MEK) inhibitors, PD98059 (10 |iM) and U0126 (10 [xM), for 1 hour prior to treatment with
400 [ig/mL AN for 24 hours, activation of ERK by AN was significantly abolished. Thus,
Chantara et al. (2006) concluded that AN induced ERK1/2 phosphorylation in SK-N-SH cells via
activation of MEK.
222
DRAFT - DO NOT CITE OR QUOTE
-------
The role of muscarinic receptors in AN-mediated ERK activation was also investigated
by pretreating SK-N-SH cells with or without 10 [xM atropine, a muscarinic receptor antagonist,
followed by incubation with AN for 24 hours. Carbachol, a muscarinic receptor agonist, was
used as a positive control. Previous studies suggested that expression of muscarinic receptors
can induce cell proliferation by activating the ERK 1/2 pathway (Jimenez and Monti el, 2005).
The results showed that 1 mM carbachol induced ERK1/2 activation, and this effect was reduced
by atropine pretreatment. However, AN-induced ERK activation was not significantly altered by
atropine pretreatment, suggesting that muscarinic receptor stimulation may not be directly
involved in the observed AN-induced ERK activation (Chantara et al., 2006).
Chantara et al. (2006) also investigated whether oxidative stress generated by various
stimuli might result in activation of ERK by studying the effects of antioxidants on AN-induced
ERK activation. Three non-enzymatic antioxidants were used: NAC, ascorbic acid, and water-
soluble vitamin E (TRX) were used. When SK-N-SH cells were pretreated with 20 mM NAC
for 10 minutes or 1 mM ascorbic acid or 1 mM TRX for 1 hour prior to addition of 400 [j,g/mL
AN for 24 hours, no reduction in AN-induced ERK activation was found. Therefore, Chantara et
al. (2006) concluded that the activation of ERK by AN observed in SK-N-SH cells was not
mediated via an oxidative stress-dependent mechanism.
To determine if AN-induced ERK activation was mediated via PKC, PKC was inhibited
via several methods. In addition to applying the PKC inhibitors, GF109203X
(bisindolymalcimide) and rottlerin, PKC was also depleted by prolonged incubation of the cells
with phorbol 12-myristate 13-acetate (PMA). Inhibition of PKC by GF109203X significantly
reduced the increase in ERK1/2 phosphorylation to 26% of that caused by AN alone. Similarly,
rottlerin and prolonged treatment with PMA reduced the activation of ERK by AN. Therefore,
the study authors concluded that PKC played an important role in AN-induced ERK activation in
SK-N-SH cells. In summary, this study demonstrated that AN activated ERK1/2 in a
PKC-dependent manner and that oxidative stress and muscarinic receptor activation were
probably not involved in ERK1/2 activation by AN.
4.5.2. Genotoxicity Studies
The following subsections provide a summary of the results of AN and its reactive
metabolite, CEO, in in vitro and in vivo mutagenicity/genotoxicity assays. The overall
conclusion that can be drawn from these studies is that AN is genotoxic after metabolic
activation to CEO. Table 4-53, which provides a summary of studies on the mutagenicity/
genotoxicity of AN, is at the end of this section.
4.5.2.1. Studies in Humans
Seven studies have examined the genotoxicity of AN in vivo in humans who may have
been occupationally exposed to AN.
223
DRAFT - DO NOT CITE OR QUOTE
-------
Xu et al. (2003) measured the incidence of DNA strand breakage and sex chromosome
aneuploidy in the sperm of 30 workers exposed to AN for 2.8 years. The mean AN
"3
concentration at operation sites was 0.8 mg/m . Thirty sperm donors of the same approximate
age range from the general population were recruited as controls. Lower sperm density was
noted in exposed workers compared with controls (75 x 106/mL vs. 140 x 106/mL). DNA strand
breakage was detected in AN-exposed workers by using single-cell gel electrophoresis, with the
rate of comet sperm higher in the exposed workers than in the controls (28.7 vs. 15%). The
frequency of sex chromosome disomy was 0.69% in the exposed group and was >0.35%) in the
control group. There were also significant differences in the frequencies of XX-, YY-, and
XY-bearing sperm between the exposed and control groups.
Fan et al. (2006), in an article translated from Chinese, evaluated the application of a
micronucleus test by using buccal mucosal cells to detect genetic damages in AN-exposed
workers. The low concentration (average concentration 0.522 mg/m AN) exposed group
consisted of 41 healthy male workers with direct contact with AN in a chemical plant that
produced AN (by the oxidation of propylene, ammonia, and air) in Shanghai. Since the entire
propylene-ammonia oxidation process was carried out in a closed system of pipes and automated
technology, the chance of contact with AN was primarily at the time of on-site sampling and
pipe inspection. The average age of this exposed group was 37.4 years, and the range and
average exposure duration were 1-33 and 15.7 years, respectively.
-3
The intermediate (average concentration 1.998 mg/m ) exposed group consisted of
47 healthy male workers in an acrylic fiber factory in Shanghai. AN was used as a raw material
in the synthesis of polyacrylonitrile by polymerization. The average age of workers was
39.8 years, and the range and average exposure duration were 1-33 and 17.2 years, respectively.
The control group consisted of 31 healthy male workers with no exposure to any known mutagen
or AN and living in the same community. Their average age was 37.2 years, and the average
working duration was 16.7 years. The rates of alcohol consumption and cigarette smoking were
similar in the exposed and control groups.
Buccal mucosal cells were collected from second scrapings from the mucous membrane
on the inside of both cheeks of study subjects after rinsing their mouths with clean water. Blood
samples were also collected for measurement of micronuclei (MN) in peripheral blood
lymphocytes. The rates of occurrence of MN in buccal mucosal cells in the intermediate
concentration exposed, low concentration exposed, and control groups were 4, 3.68, and 2.03%,
respectively. The rates of MN in the low and intermediate concentration groups were
significantly higher than in the control group (p < 0.05). The rates of occurrence of MN in
peripheral blood lymphocytes in the intermediate and low concentration exposed and control
groups were 4.23, 2.44, and 2.48%>, respectively. The rate of occurrence of MN in blood
lymphocytes in the intermediate exposed group was significantly different from that in the
control group (p < 0.05).
224
DRAFT - DO NOT CITE OR QUOTE
-------
To investigate the relationship between AN exposure and the rate of occurrence of MN
and to eliminate the possibility of the presence of other confounding factors, a multivariate linear
regression analysis was conducted. The results indicated that the cumulative exposed amount of
AN, the recent exposed amount of AN, and the extent of cigarette smoking were important
factors in the rate of occurrence of MN in buccal mucosal cells and blood lymphocytes. Fan et
al. (2006) concluded that the micronucleus test of buccal mucosal cells could replace the
micronucleus test of lymphocytes in the peripheral blood as a screening test for genetic damage
in AN-exposed workers.
Borba et al. (1996) evaluated urinary genotoxicity of three groups of workers in an AN
fiber production plant (exposed group 1: 14 workers in the continuous polymerization section;
exposed group 2: 10 equipment maintenance workers; control group: 20 administrative workers
from the same plant). Urine extracts were used in the Ames test (using TA 98, +S9) to assess
gene reversion activity. No differences in urinary genotoxicity were found in the three groups.
Additionally, there were no significant differences in the incidence of SCE in peripheral
lymphocytes among the groups. However, the maintenance workers had a higher incidence of
CAs in lymphocytes, consisting of chromatid and chromosomal gaps and chromatid and
chromosomal breaks, than the controls (p < 0.003). These effects were also increased in
production workers but not to statistically significant levels, possibly due to the comparatively
low number of subjects in the study. The significantly higher incidence of CAs in maintenance
workers were in agreement with the highly significant levels of Hb adducts and CEVal in the
same population (see Section 4.1.2.2). The maintenance workers also had significantly higher
levels of erythrocyte MDA, an indicator of lipid peroxidation, than the other two group of
workers.
Ding et al. (2003) compared deletion frequencies of mitochondrial DNA in peripheral
lymphocytes in 47 workers exposed to a mean workplace concentration of 0.11 ppm AN and
47 nonexposed workers using PCR techniques (this study is described more fully in Section
4.1.2.2). No deletions were detected in the nonexposed group, but a deletion frequency of 17%
was detected in the exposed group. In a separate experiment on presumably nonexposed
individuals, no deletions in mitochondrial DNA were detected in samples from 12 high school
students, whereas the deletion frequency was 25% in samples from 12 elderly persons.
Consistent with the hypothesis that damage to mitochondrial DNA contributes to degenerative
diseases related to aging, the study authors suggested that occupational exposure to AN may
induce mitochondria DNA deletion in cells that are related to aging.
Using the FISH technique with probes for chromosomes 1 and 4, Beskid et al. (2006)
examined patterns of CAs in cultured lymphocytes from blood samples of 61 AN-exposed male
workers involved in the polymerization of Indian rubber and 49 nonexposed control subjects.
Stationary monitoring in the workplaces indicated AN air concentrations of 0.05-0.3 mg/m for a
group of 39 exposed workers sampled in 2000 and 0.05-0.7 mg/m for another group of
225
DRAFT - DO NOT CITE OR QUOTE
-------
22 exposed workers sampled in 2003. A 38% increase in frequency of aberrant cells in
AN-exposed workers was found to be statistically insignificant. However, the number of
reciprocal translocations increased by 53% (p < 0.05) in the AN-exposed group. In addition, a
significant increase in a relative number of insertions was found in the AN-exposed group.
Furthermore, chromosomal specificity was observed in lymphocytes with aberrations on
chromosome #1 and #4. In the AN-exposed group, the proportion of cells with aberrations on
chromosome #1 decreased significantly (58.8 vs. 73.8% in the control subjects, adjusted to age
and smoking), but aberrations on chromosome #4 increased (47.0 vs. 29.4% in the controls).
In an earlier study by the same research group, the frequency of CAs in peripheral blood
samples was studied in 45 male rubber polymerization workers exposed for the last 3 months to
"3
0.05-0.3 mg AN/m , 23 matched controls living in the same region (control group I), and
33 unexposed controls from Prague (control group II) (Sram et al., 2004). Subjects were
interviewed and completed questionnaires on demographic data, occupational and environmental
exposures, smoking habits, medications, X-ray examinations, viral infections, and alcohol
consumption within the 3 months before sampling. Cytogenetic analysis was conducted using
two methods. Conventional chromosomal analysis was used to quantify CAs (chromatid plus
chromosome breaks and chromatid plus chromosome exchanges), the number of aberrant cells
(those with breaks and exchanges, gaps not included), and the aberration frequency (number of
breaks per cell). The FISH technique, using probes for chromosomes 1 and 4, was employed to
quantify translocations. Conventional analysis did not detect any differences in the frequency of
CAs in exposed workers compared with either control group. FISH detected no differences in
the frequencies of aberrations or translocations in exposed workers compared with matched
controls (control group I), but the frequencies in both groups were significantly elevated
compared with unexposed controls from Prague (control group II). Sram et al. (2004) concluded
that occupational exposure to 0.05-0.3 mg/m AN did not present a significant genotoxic risk
and attributed higher frequencies in the exposed group and control group I to undetermined
factors present in the region in which the petrochemical industries are located but absent in
Prague.
Thiess and Fleig (1978) surveyed 18 workers at a plant that was used for manufacturing
copolymers of styrene and AN, styrene, AN, and butadiene and also for synthesis of organic
intermediates. The workers had been exposed to AN, on average, for 15.3 years and could have
been exposed simultaneously to styrene, ethylbenzene, butadiene, and other chemicals. The
control group consisted of 18 workers who had not been exposed to AN. Atmospheric
monitoring conducted between 1963 and 1974 revealed AN concentration of about 5 ppm, with
the possibility of higher peak values occurring in connection with special tasks. Between 1975
and 1977, exposure to AN had been reduced to an average of 1.5 ppm. The incidence of CAs in
the lymphocytes was measured in these 18 workers and in age-matched nonexposed controls.
The numbers of aberrant metaphases (gaps and iso-gaps included) were 5.5 ± 2.5% in exposed
226
DRAFT - DO NOT CITE OR QUOTE
-------
workers and 5.1 ± 2.3% in controls. When gaps were not included, the numbers dropped to
1.8 ± 1.3% in exposed workers and 2.0 ± 1.6% in controls. The differences were statistically not
significant. This study is of limited value because of mixed exposure in the exposed group,
small group size, and uncertainty in exposure levels.
4.5.2.2. In Vivo Tests in Mammals
Rats
Oral administration of up to 40 mg/kg AN or i.v. administration of up to 98 mg/kg AN to
male Sprague-Dawley rats did not induce MN in bone marrow and peripheral blood, respectively
(Morita et al., 1997). However, Wakata et al. (1998) demonstrated induction of MN in bone
marrow polychromatic erythrocytes of Sprague-Dawley rats (four/group) treated 2 times with
124.8 mg/kg AN i.v. and sampled 24 hours after treatment. A negative result was obtained in
peripheral blood. In another study (Rabello-Gay and Ahmed, 1980), male Sprague-Dawley rats
treated orally with 16 daily doses of 40 mg/kg-day AN showed no increase in CAs in the bone
marrow over controls.
Irreversible binding of radioactivity from [2,3-14C]-AN to DNA in brain, stomach, and
liver of male Sprague-Dawley rats was reported 24 hours after a single oral dose of 46.5 mg/kg
(Farooqui and Ahmed, 1983a). DNA alkylation was significantly higher in the target organs,
brain and stomach (119 and 81 pmol/mg DNA at 24 hours, respectively), than in the liver
(25 pmol/mg DNA). The covalent binding indices in the liver, stomach, and brain at 24 hours
after dosing were 5.9, 51.9, and 65.3, respectively. Similarly, covalent binding of [2,3-14C]-AN
or its metabolite to testicular (Ahmed et al., 1992a), lung (Ahmed et al., 1992b), and gastric
tissue DNA (Abdel-Rahman et al., 1994b) has been reported in male Sprague-Dawley rats
treated with a single oral dose of 46.5 mg/kg AN. Maximum covalent binding of radioactivity to
gastric DNA occurred at 15 minutes after dosing and occurred at 0.5 and 12 hours for testicular
and lung DNA, respectively. Alkylation of hepatic DNA was also reported when a single dose
of 0.2 mmol [2,3-14C]-AN was administered to male Wistar rats intraperitoneally (Peter et al.,
1983a). Two 14C peaks that did not cochromatograph with any known standards were observed
when the DNA hydrolysate from rat livers was chromatographed on PEI-cellulose column.
In another study, 0.6 mg/kg [2,3-14C]-CEO was administered to one F344 rat
intraperitoneally (Hogy and Guengerich, 1986), and the rat was sacrificed after 1 hour. Covalent
binding to both liver and brain protein was found, but no covalent binding to nucleic acids could
be detected at the level of 0.3 alkylations per 106 bases. In the same study, three male F344 rats
were administered 50 mg/kg AN i.p., and three other rats were administered 6 mg/kg CEO i.p.
n
(Hogy and Guengerich, 1986). The rats were sacrificed after 2 hours. N -(2-oxoethyl)guanine
was measured in liver DNA at the level of 0.032 and 0.014 alkylations/106 bases for CEO- and
n
AN-treated rats, respectively. In the brains of treated rats, the levels of N -(2-oxoethyl)guanine
were not above the limit of detection. Since DNA adduct was detected in liver DNA, covalent
227
DRAFT - DO NOT CITE OR QUOTE
-------
binding to DNA had to occur. The method used in the study was not sensitive enough to detect
low levels of alkylation of nucleic acids, probably because of the small amount of DNA sample
from one rat, and the DNA isolation protocol (DNA extract purified by hydroxylapatite column
chromatography and DNA fraction dialyzed against water and lyophilized then treated with
RNAase and proteinase K) and method for correction for contaminating protein via quantitative
amino acid analysis in the DNA sample may have allowed a more stringent determination of
DNA-bound material.
When the alkaline comet assay was used to detect DNA lesions, Sekihashi et al. (2002)
demonstrated DNA damage in the forestomach, colon, kidney, bladder, and lung of Wistar rats
treated with a single dose of 30 mg/kg AN i.p. but not in the brain or bone marrow.
Oral exposure to 100 ppm AN in drinking water for 14 or 28 days significantly increased
the level of cellular DNA fragmentation in the brain of male Wistar rats (Mahalakshmi et al.,
2003). Other aspects of this study are discussed in Section 4.5.1.1.4.
AN-induced unscheduled DNA synthesis (UDS) was demonstrated in several studies in
rats. In a study by Hogy and Guengerich (1986), male F344 rats (12/group) received a single
sublethal dose of 50 mg/kg AN in saline by gavage, followed by hydroxyurea to arrest
replicative DNA synthesis but allowing excision repair DNA synthesis. Two hours after dosing,
"3
the animals received s.c. methyl [ H]-labeled thymidine. This dose was repeated after 2 hours,
and half the dose was given again after 2 more hours for a total dose of 3.0 mCi/kg of BW. The
"3
animals were sacrificed 2 hours after the last methyl [ H]-labeled thymidine. A significant
occurrence of UDS was found in the livers but not in the brains of AN-treated rats.
Hogy and Guengerich (1986) also studied the effect of treatment on DNA synthesis over
4 hours in the liver and brain of male F344 rats 48 hours after an oral dose of 50 mg/kg AN.
DNA synthesis was decreased in the brain but not in the liver; replicative indices were 0.29 and
1.30, respectively. Thus, the carcinogenicity of AN in rat brain is not likely from cytotoxicity,
followed by an increased rate of DNA replication and leading to a greater chance of error during
the rapid DNA synthesis.
UDS was demonstrated in lung (Ahmed et al.,1992a), testis (Ahmed et al., 1992b), and
gastric tissue (Abdel-Rahman et al., 1994a) of AN-treated male Sprague-Dawley rats (12/group).
Animals received a single oral dose of 46.5 mg/kg AN in saline, with or without hydroxyurea
"3
cotreatment to block the endogenous deoxynucleotide pool. [ H]-Thymidine was administered
0.5, 6, or 24 hours after AN dosing, and animals were sacrificed 2 hours later. The replicative
index for DNA synthesis (i.e., the ratio of DNA synthesis in treated animals over controls) was
significantly reduced at all three time points in lung, while DNA repair in the lung was increased
by twofold at 0.5 hour and 1.6-fold at 6 hours following AN oral treatment. Similarly, DNA
synthesis was inhibited in testes at 0.5 and 24 hours after treatment (but increased at 6 hours),
whereas DNA repair increased at 1.5- and 33-fold at 0.5 and 24 hours after treatment. For
gastric tissue, DNA replicative synthesis was inhibited 6 hours after AN administration but was
228
DRAFT - DO NOT CITE OR QUOTE
-------
rebounded and followed by a twofold increase at 24 hours. A threefold increase in UDS was
observed at 24 hours after dosing.
On the other hand, Butterworth et al. (1992) followed the incorporations of
"3
[ H]-thymidine into the hepatocytes isolated from male F344 rats gavaged with either a single
dose of 75 mg/kg AN or five daily doses of 60 mg/kg AN. Single-dosed animals were sacrificed
2 or 12 hours after dosing. Multiple-dosed animals were sacrificed 4 hours after the last dose.
"3
Hepatocytes were isolated and plated on cover slips and incubated with [ H]-thymidine.
Autoradiography was used to detect UDS. No sign of AN-induced UDS was found in
hepatocytes from exposed rats. In addition, no UDS was found in the in vivo spermatocyte DNA
repair assay with AN, using cells isolated from the seminiferous tubules of the same treated rats
(Butterworth et al., 1992). The difference in results from Butterworth et al. (1992) and Hogy and
Guengerich (1986) regarding UDS in rat liver may be due to differences in methodology in that
"3
incorporation of [ H]-thymidine actually took place in vitro in the study by Butterworth et al.
(1992).
AN also produced negative results for dominant lethal assay in male F344 rats (Working
et al., 1987). In this assay, groups of 50 male F344 rats received AN by gavage at 0 or
60 mg/kg-day in 0.9% saline for 5 days. AN exposure in males had no effect on the incidence of
pre- or postimplantation losses, indicating a negative result to germ cells.
Mice
When the alkaline comet assay was used to detect DNA lesions, Sekihashi et al. (2002)
demonstrated DNA damage in the forestomach, colon, bladder, lung, and brain of male ddY
mice treated with 20 mg/kg AN i.p. DNA damage was not detected in the liver, kidney, or bone
marrow.
Sharief et al. (1986) determined that AN caused a slight increase on SCE frequencies in
bone marrow cells of male C57B1/6 mice (four/dose group). No increase in SCE frequencies
was observed in mice administered a single dose of up to 30 mg/kg AN intraperitoneally. Higher
doses were lethal to most of the animals. An increase in SCE frequency (2 x control) was
observed in the only surviving mouse at the 45 mg/kg dose group. However, Fahmy (1999)
reported AN-induced SCEs in bone marrow cells of male Swiss mice (five/group) treated with
7.5 mg/kg or 10 mg/kg AN i.p. 8 hours following BrdU treatment and with colchicines 2 hours
prior to sacrifice. The lowest dose of 5 mg/kg i.p. produced no significant effect on SCE
frequency.
Earlier CA studies in mice have been largely negative. Rabello-Gay and Ahmed (1980)
showed that AN did not produce increases in CAs in bone marrow cells of male Swiss mice
(six/group) when given orally for 4, 15, or 30 days at doses up to 21 mg/kg-day or by i.p.
injection at doses up to 20 mg/kg-day for the same duration. No increase in the incidence of
CAs compared with controls was observed in bone marrow cells of NMRI mice that were
229
DRAFT - DO NOT CITE OR QUOTE
-------
injected intraperitoneally with 20 or 30 mg/kg AN (Leonard et al., 1981). No increase in CAs in
-3
bone marrow cells and spermatogonia was observed in ICR mice treated with 20 or 100 mg/m
AN for 5 days (Zhurkov et al., 1983). However, Fahmy (1999) reported AN-induced CAs in
mouse spermatocytes after single oral doses of 15.5 or 31 mg/kg or three or five successive oral
doses of 7.75 mg/kg (1/8 LD50) in male Swiss mice. In addition, AN induced CAs in mouse
bone marrow cells and spleen cells after a single oral dose of 7.75 mg/kg or three or five
successive doses of 7.75 mg/kg. The aberrations were mainly of chromatid type (gaps, breaks,
fragments, and deletions), with metaphases carrying one aberration only being dominant.
Leonard et al. (1981) also reported AN did not induce MN in polychromatic erythrocytes
of male NMRI mice injected intraperitoneally with 20 or 30 mg/kg AN. Marginal, but
statistically significant, increases in MN were observed in bone marrow polychromatic
erythrocytes but not in peripheral blood when AN at doses of 5.6-45 mg/kg was administered to
male CD-I mice (five/group) via i.p. administration. Oral or i.v. injection yielded negative
results (Morita et al., 1997).
Treatment with AN produced negative results for dominant lethal assays in male mice.
The dominant lethal assay was used to detect CAs in meiotic and postmeiotic male germ cells.
Groups of five male NMRI mice were injected intraperitoneally with 0 or 30 mg/kg AN in saline
or isopropyl methanesulfonate (positive control) and then mated to untreated females (three per
male) for 5 weeks (Leonard et al., 1981); females were replaced after 7, 14, 21, and 28 days, and
the uterine contents were examined 17 days after mating. No evidence for dominant lethal
effects was observed.
4.5.2.3. Short-term Tests: Bacteria, Fungi, Drosophila, Others
There are a large number of reports of short-term genotoxicity test results on AN that
have been made available through the auspices of the International Programme on Chemical
Safety (IPCS). The IPCS coordinated the investigation of eight organic carcinogens known to be
either inactive or difficult to detect in the Salmonella assay, including AN, benzene,
diethylhexylphthalate, diethylstilbestrol, hexamethylphosphoramide (HMPA), PB, safrole, and
o-toluidine, as well as two noncarcinogens in rodent bioassays (benzoin and caprolactam). Most
of the available short-term genotoxicity tests were employed, and the work was carried out at
some of the major research and testing laboratories throughout the world. The purpose of this
endeavor was to evaluate the efficacy of these tests, to evaluate the strengths and weaknesses of
such tests, and to identify the assay systems to complement the widely used Salmonella assay.
All of the results from these studies have been compiled in a 750-page collection (Ashby et al.,
1985), and a 56-page synopsis of the results is available on the Internet (IPCS, 1985). The
conclusion from these evaluations was that AN, along with HMPA, o-toluidine, and safrole,
belonged to a group of genotoxins that were detected by most of the eukaryotic assays studied
and could easily be found to be nonmutagenic in the Salmonella assay because of protocol
230
DRAFT - DO NOT CITE OR QUOTE
-------
deficiencies associated with the overall metabolic capacity of the assay system. While several of
these studies have been cited in this section, an overview of the findings as they pertain to AN
follows.
AN, at concentrations not overtly toxic in a given assay, was found in 42 of 68 tests to
positively cause genotoxicity in bacteria, fungi, Drosophila melanogaster (mutation, gene
reversion, mitotic crossing over, aneuploidy), and mammalian cell culture assay systems, both
human and animal (single-strand breaks, UDS, CAs, SCEs, MN, and transformation). Among
the eight carcinogens studied, AN gave the most positive results, inducing genotoxicity
responses in 62% of all tests (42 out of 68 tests; 25 were negative and 1 was questionable). That
number of positive responses rose to 81% when a more stringent selection of assays was applied
(IPCS, 1985). Other studies, covering all assay types used in that collaborative study, are listed
in the two subsections on short-term assays (Sections 4.5.2.3 and 4.5.2.4).
4.5.2.3.1. Bacterial tests. A number of research groups have examined the capacity of AN to
induce gene reversion in the Ames test. One of the first published papers was Milvy and Wolff
(1977), which reported positive results of AN on Salmonella typhimurium strains TA 1535,
1538, and 1978 in the presence of S9 and an NADPH generating system. The first strain is
sensitive to base substitution mutagens, and the latter two strains are sensitive to frameshift
mutagens. Negative results were obtained when metabolic activation was excluded from the
system. Gene mutation by exposure of bacteria to an atmosphere containing AN was found to
occur at concentrations as low as 57 ppm (equivalent to 2 [xL of AN), lower than exposure in
solution or spotting AN to a "lawn" of bacteria on a plate. This could be because AN vaporized
readily in solution such that the actual exposure concentrations in solution or by spotting were
uncertain. The findings of Milvy and Wolff (1977) were criticized by Venitt (1978) in a letter to
the editor. While not disagreeing with the overall conclusion, Venitt (1978) considered the study
authors to have calculated mutation frequency incorrectly from the data. Venitt (1978)
confirmed AN to be mutagenic in TA 1535 but not in the frameshift strains TA 1538 and 1978.
Among other reports of AN activity in the Ames test, Lijinsky and Andrews (1980) found
AN at doses of 100-1,000 jag per plate to be positive in S. typhimurium TA 1535 in the presence
of S9 but negative in TA 98, 100, 1537, and 1538 in the plate incorporation assay. Similarly,
Brams et al. (1987) reported AN (50-750 (J,g/L) to be negative for gene reversion in TA 97, 98,
and 100 using plate incorporation assay, irrespective of metabolic activation. The study authors
accounted for these negative results with inadequate experimental conditions. Previously, they
had demonstrated that AN was mutagenic in TA 1530, 1535, and 1950 when 0.2% gaseous AN
was injected into the dessicator where the plates were incubated for 1 hour in the presence of S9
(the plate agar contained 200 |ig AN/plate) (de Meester et al., 1978). A lower mutagenic activity
was also detected with strains reversed by frameshift mutation (TA 98, 1978, 100), and assays
231
DRAFT - DO NOT CITE OR QUOTE
-------
conducted by the plate incorporation method gave negative results (de Meester et al., 1978; Dow
Chemical Co., 1976).
Jung et al. (1992) provided data from three laboratories that examined the ability of AN
to induce gene reversion in TA 102 in the plate incorporation assay, with uniformly negative
results. Hakura et al. (2005) reported that AN induced dose-dependent increases in the number
of revertants per plate in strain TA 100 exposed to concentrations ranging from 806 to 12,100 jag
per plate in the presence of rat or human liver S9 preparations, but the maximum response at
12,100 jag per plate was slightly less than twofold higher than the number of revertants per plate
observed in the negative controls.
Although the methodologies of the experiments may have been different, the positive
finding of mutagenicity by de Meester et al. (1978) for AN in TA 98 and 100 is in agreement
with data from Khudoley et al. (1987) that AN (concentration not provided) was positive in
TA 98 and 100 (with two- to fivefold increase in frequency of induced mutants), irrespective of
the presence of S9. Zhurkov et al. (1983) reported AN to dose-dependently induce mutations in
S. typhimurium TA 1535 but not in TA 1538. The presence of complete S9 fraction was required
for this effect, but a 9,000 x g microsomal supernatant without cofactors gave inconsistent
results. The concentrations tested were between 0.1 and 10,000 jag per dish. The highest
concentration was overtly toxic to the bacteria.
Other tests of the mutagenicity of AN in bacterial systems were negative. For example,
Nakamura et al. (1987) employed the SOS test to evaluate the capacity of AN to induce
expression of the umu gene in S. typhimurium TA 1535/pSK1002. This strain contains a
umuC-lacZ fused gene, such that a forward mutation results in umu gene expression and the
transcription of the lac operon and can be demonstrated phenotypically by a twofold increase in
P-galactosidase activity. AN, at concentrations up to 2,820 ng/mL, and seven other known
mutagens gave negative results in this system. Similarly, AN was negative in the SOS
chromotest in Escherichia coli PQ37 (Brams et al., 1987). It should be noted that only 4 out of
14 compounds that were positive in the Ames assay were positive in the SOS chromotest kit
used. Thus, some technical issues were responsible for the poor performance of the test kit
(Brams et al., 1987).
Venitt et al. (1977) tested AN for mutagenicity in a gene reversion assay (2-3 days at
37°C), using the tryptophan-dependent coli WP2 series of bacteria as indicator organisms.
Doses of 75 and 150 [j,mol per plate of AN produced a dose-related increase in the number of
revertant colonies (tip —> trp+) compared with untreated bacteria in WP2 (which is DNA repair
proficient), WP2 uvrA (which lacks excision repair), and WP2 uvrApolA (which lacks both
excision repair and DNA polymerase 1) without a need for S9 fraction. The study authors
confirmed these results by using a simplified "fluctuation" assay in which they obtained a dose-
dependent increase in mutation rate induced by 0.4-2 mM AN (20- to 40-fold lower than the
levels in plate test) in E. coli WP2. An exponential dose response was seen at lower
232
DRAFT - DO NOT CITE OR QUOTE
-------
concentrations, 0.1-0.4 mM, with mutant WP2 uvrApolA, which lacked both excision repair and
DNA polymerase-1. The effective mutagenic concentration range for AN was lowered by an
order of magnitude when an error-prone DNA-repair plasmid, pKMlOl, was introduced into
E. coli WP2. The study authors concluded from these results that AN caused non-excisable mis-
repair DNA damage that ultimately gave rise to DNA strand breaks. Venitt et al. (1977)
hypothesized that AN might react with thymine residues in DNA since AN has been shown to
cyanoethylate ring N atoms of minor tRNA nucleosides and ribothymidine and thymidine
(Ofengand, 1967).
Lambotte-Vandepaer et al. (1985, 1981, 1980) collected urine from male Wistar rats
(two/group) and NMRI mice (five/group) that had been administered a single dose of AN
(30 mg/kg intraperitoneally). The urine samples were evaluated for potential induction of gene
reversion in S. typhimurium TA 1530. Positive results were obtained in rats and mice in the
absence of S9. Mutagenic activity in urine was abolished by the presence of S9 in rats and
decreased in mice, a finding that suggests that the mutagenic agent in urine can be inactivated
metabolically. Pretreatment with PB (induces CYP450 monooxygenase), C0CI2 (inhibits
CYP450 monooxygenase), and DEM (depletes GSH), before AN treatment, slightly decreased
the mutagenic response in urine from mice and completely abolished the response in urine from
rats. However, the study authors were unable to identify the genotoxic compound in urine.
4.5.2.3.2. Fungi. Available studies that employed fungi in short-term assays on the
mutagenicity/genotoxicity of AN produced mostly positive results. AN induced mitotic gene
conversion in both stationary-phase and log-phase cultures at the his4 and trps loci of
Saccharomyces cerevisiae JD1, in the presence of metabolic activation by S9. Negative results
were obtained without metabolic activation (Brooks et al., 1985; Shell Oil, 1984a). AN did not
induce chromosome loss in S. cerevisiae D61.M (Whittaker et al., 1990) but elicited respiratory
deficiency, reflecting antimitochondrial activity.
4.5.2.3.3. Drosophila. A range of in vivo experimental systems used the fruit fly,
D. melanogaster, to examine the mutagenicity/genotoxicity of AN. Drosophila can biotransform
certain procarcinogens to their reactive metabolites and are used in short-term tests for
identifying carcinogens and in studies on the mechanism of mutagenesis of chemicals (Vogel et
al., 1999).
Osgood et al. (1991) used the Drosophila ZESTE system to monitor the potential for AN
to induce sex chromosome aneuploidy, following inhalation exposure of adult females to
2.7 ppm for up to 70 minutes. AN induced chromosome loss after exposure for 50 and
70 minutes. AN was mostly nontoxic at the tested dose, with only 13% killed after a 70-minute
exposure. Similarly, in a Drosophila somatic recombination and mutation assay, Drosophila
larvae were exposed to 5-20 mM AN in water for 9-11 days, and hatching females were scored
233
DRAFT - DO NOT CITE OR QUOTE
-------
for twin mosaic spots and single mosaic light spots in their eyes (Vogel, 1985). These genetic
markers might arise from many types of genetic alterations. Mitotic recombination and
chromosome breakage would result in mosaic twin spots, whereas deletions and gene mutations
would give rise to mosaic single light spots. AN was positive for the induction of mosaic single
spots at 5 mM (LC50 concentration was 10 mM) and negative for twin spots. Since the
classification of single spots or twin spots might be subjected to personal bias, the total of twin
and single spots was also reported, and AN gave positive result.
AN was shown to be mutagenic by having marginally positive effects in somatic
mutation/recombination assay on wing spots of Drosophila, with gas exposure of larva to 0.5-
1 [xL/1,150 mL for 0.5 or 1 hour (Wiirgler et al., 1985). AN gave negative results in a sex-linked
recessive lethal mutation test on postmeiotic and meiotic germ cells of male D. melanogaster,
exposed either by feeding of 420 ppm AN or injected with 3,500 ppm AN (Foureman et al.,
1994).
The in vitro effect of AN on taxol-purified microtubules from Drosophila and mouse
brain was evaluated by Sehgal et al. (1990). (Taxol promoted the formation and stability of
microtubules.) Microtubules assist in the movement of chromosomes in both mitosis and
meiosis. Polymerization and depolymerization of microtubules occur in cell division to separate
the chromosomes from the metaphase plate during anaphase. In this study, the assembly and
disassembly of microtubules was monitored spectrophotometrically in vitro. Previous results
from in vivo assays monitoring induced sex chromosome aneuploidy indicated that effective
aneuploidogens affected microtubule assembly. When taxol-purified D. melanogaster
microtubule was incubated at 37°C to allow polymerization, addition of 5 or 50 mM AN resulted
in 28 and 64% inhibition of microtubule assembly, respectively. When taxol-purified mouse
brain microtubules were incubated with 5 or 50 mM AN, 74 and 96% inhibitions of microtubule
assembly were observed, respectively. Thus, these results indicated that AN was an
aneuploidogen in vitro. On the other hand, taxol significantly affects microtubule
depolymerization assay, probably by stabilizing the formed microtubules. None of the tested
aneuploidogens, including colchicine, promoted disassembly to taxol-purified microtubules.
4.5.2.3.4. Other short-term tests. Yates et al. (1994) reported on the ability of CEO to induce
single- and double-strand DNA breaks in supercoiled DNA plasmid pBR322 DNA. Supercoiled
DNA (1 (j,g) was incubated for 3 hours at 37°C with >50 mM CEO and then subjected to agarose
gel electrophoresis. The study authors reported that CEO non-enzymatically induced DNA
strand breaks in a dose- and time-dependent manner, but detailed data were not provided. Peter
et al. (1983b) found that AN did not induce strand breaks in SV40 phage DNA in vitro, whereas
synthetic glycidonitrile (i.e., CEO) was effective in the same system (both agents were incubated
with DNA at 1 mmol/L for 17 hours in the dark, at 37°C, in a buffered solution without any
enzyme addition).
234
DRAFT - DO NOT CITE OR QUOTE
-------
4.5.2.4. Mammalian Cell Short-term Tests
4.5.2.4.1. Mutations. A number of research groups have examined the ability of AN to induce
forward mutations in human lymphoblast cell lines and the well-known mouse lymphoma
L5178Y Tk system.
Crespi et al. (1985) used two human lymphoblast cell lines, Tk6 (which does not contain
CYP450 activities) and AHH-1 (which is metabolically competent) to assess AN mutagenicity at
the thymidine kinase (Tk+/~) locus and the hprt locus, respectively. Tk6 cell cultures were treated
with 0-40 [j,g/mL AN for 3 hours with and without externally added metabolic activation (rat
liver S9). On the third day after treatment, the cultures were plated in a selective medium
containing 2 [j,g/mL trifluorothymidine. After incubation for 12 days, the plates were scored for
the presence of mutant colonies. Similarly, AHH-1 cell cultures were treated with AN for
28 hours, and the cultures were plated on the 6th and 7th day after treatment in a selective medium
containing 0.6 [ig/mL 6-TG. AN induced dose-dependent mutations at the Tk locus in Tk6
cells in the presence, but not in the absence, of S9. AN induced mutations at the hprt locus in
AHH-1 cells. The lowest AN concentrations that were mutagenic in these test systems were
40 [ig/mL with Tk6 cells +S9 and 25 ng/mL with AHH-1 cells.
Similarly, Recio and Skopek (1988a, b) assessed the mutagenicity of AN and its epoxide
metabolite, CEO, at the Tk+/~ locus in Tk6 human lymphoblast cells in the presence and absence
of rat liver S9. In the presence of S9, 2-hour incubations with 1.4 mM AN induced mutations at
the Tk+/~ locus as demonstrated by the presence of mutant clones when plated in trifluoro-
thymidine selection medium after treatment, but, in the absence of S9, no mutagenic activity was
observed over the concentration range of 0.4-1.5 mM (Recio and Skopek, 1988b). In contrast,
2-hour incubation with CEO at concentrations as low as 100 and 150 |iM induced a mutagenic
response (without metabolic activation) at the Tk+/~ locus (Recio and Skopek, 1988b). On a
molar basis, CEO was as mutagenic in this system as the well-known mutagen, ethyl
methanesulfonate (Recio and Skopek, 1988a).
Two classes of CEO-induced Tk mutant phenotypes were identified that differed in
their growth rates: Tkn with normal growth rate and Tks with slower growth rates. Southern blot
analysis of DNA of these two classes indicated that the phenotypes differed genotypically (Recio
and Skopek, 1988b). Ninety-six percent (25/26) of Tks mutants had lost a 14.8 kb DNA
fragment corresponding to the active Tk allele, whereas only 8% (1/12) of CEO induced Tkn
mutant clones and 22% (2/9) of spontaneous Tkn mutant clones had lost the 14.8 kb fragment
(Recio and Skopek, 1988a, b). Recio and Skopek (1988a) suggested that Tks mutants resulted
from large-scale DNA structural alterations involving the active Tk allele. CEO induced
predominantly Tkn mutants. Southern blot analysis of CEO-induced Tkn mutants indicated the
majority of these mutants were below the detection limit of <2 kb. Thus, CEO-induced
alterations are relatively small DNA alterations. Recio and Skopek (1988a) suggested that
235
DRAFT - DO NOT CITE OR QUOTE
-------
CEO-induced Tkn mutants resulted from point mutations or small insertion/deletions that
occurred during the replication or repair of CEO-modified DNA. Karyotypic analysis on two
Tkn mutants and 16 Tks mutants indicated that the majority of Tk mutants were not accompanied
by abnormalities of the chromosome on which the Tk gene resides (chromosome 17).
CEO also induced mutations at the hprt locus in Tk6 cells (Recio and Skopek, 1988a).
Characterization of the hprt mutations by cDNA sequencing analysis indicated that several hprt
mutations were formed. A major (8/14) type of CEO-induced mutation was the specific loss of
exons from the coding region of hprt. Remaining mutants (6/14) were single base substitutions
(point mutation) resulting from amino acid changes (A:T base pairs and G:C base pairs).
The mutagenicity of AN was evaluated in the mouse lymphoma L5178Y Tk+/~ forward
mutation assay by a number of laboratories. Oberly et al. (1996) demonstrated that AN
(activated with S9) was mutagenic at 40 ng/mL in this assay because of producing a more than
twofold increase in mutant frequency when compared to the mutant frequency of the solvent
controls. Earlier results in the mouse lymphoma L5178Y Tk+/~ system from several studies all
pointed to the capacity of AN at 12.5-200 ng/mL to induce forward mutations at the Tk with or
without S9 (Lee and Webber, 1985; Myhr et al., 1985; Amacher and Turner, 1985; Rudd, 1983).
Garner and Campbell (1985) also reported that AN induced mutations to ouabain and 6-TG
resistance in the mouse lymphoma L5178 Y cells in the presence of S9. However, negative
results were obtained by Styles et al. (1985) in the mouse lymphoma L5178Y Tk+/+ cell line
(Na+/K+ ATPase locus, ouabain was used for mutant selection) and Tk cell line
(trifluorothymidine was used for mutant selection). Using P388F mouse lymphoma Tk cell line
in the presence of S9, Anderson and Cross (1985) also obtained forward mutations to 5-iodo-
2-deoxyuridine resistance with AN. However, negative results were obtained by Lee and
Webber (1985) in Chinese hamster V79/HGPRT assay in which AN did not induce 8-azaguanine
resistance mutation with or without S9.
4.5.2.4.2. Other DNA effects
UDS
AN, at concentrations up to 2.5 mg/mL, was negative for induction of UDS in HeLa cells
with or without the presence of S9, using the scintillometric method (Martin and Campbell,
-3
1985). However, an increase in UDS, as measured by uptake of [ H]-thymidine, was reported in
cultured human lymphocytes treated with 5 x 10"1 M AN and S9 (Perocco et al., 1982). Negative
results were reported for the ability of AN to induce DNA repair synthesis (measured by
"3
incorporation of [ H]-thymidine and autoradiographic techniques) in rat hepatocyte primary
culture (Probst and Hill, 1985; Williams et al., 1985). However, no UDS, as measured by
autoradiography, was observed in any of the cultures treated with the eight tested carcinogens
(Probst and Hill, 1985). Similarly, only one of the eight carcinogens tested by Williams et al.
-3
(1985) induced DNA repair synthesis as determined by autoradiography of [ H]-thymidine
236
DRAFT - DO NOT CITE OR QUOTE
-------
incorporation (AN was negative in this assay). Thus, IPCS (1985) concluded the rat hepatocyte
autoradiographic UDS assay was an insensitive assay for determination of genotoxicity of these
chemicals and should be avoided for use as a complement to the Ames assay.
"3
Butterworth et al. (1992) used incorporation of [ H]-thymidine and autoradiographic
techniques to study the effect of AN and CEO on unscheduled DNA repair in vitro in rat
hepatocytes, and human mammary epithelial cells. AN and CEO were negative for the induction
of DNA repair in hepatocytes in vitro. There was some indication of a statistically insignificant
response at 0.1 mM CEO. However, CEO was toxic to the hepatocyte culture at 1 mM, the next
higher tested concentration. As noted previously, IPCS (1985) has determined that rat
hepatocyte autoradiographic UDS assay is insensitive for genotoxicity testing of AN and a group
of seven other carcinogens. However, CEO, but not AN, was positive for UDS in human
mammary epithelial cells in vitro (Butterworth et al., 1992).
DNA strand breaks
DNA single-strand breaks were measured by the alkaline elution method in the following
studies. DNA single-strand breaks were induced in [14C]-thymidine-labeled cultured adult
human bronchial epithelial cells treated with 200 or 500 ng/mL AN for 20 hours (Chang et al.,
1990). These concentrations were below the cytotoxic concentration of 600 [j,g/mL AN. DNA
single-strand breaks were also reported in cultured rat hepatocytes treated with 65.8 p,g/mL AN
for 3 hours (Bradley, 1985). Higher concentrations of 197 or 658 [ig/mL AN resulted in
cytotoxicity. In another study, AN induced DNA single-strand breaks in cultured Chinese
hamster ovary (CHO) cells treated with 3.7 x 103 or 5.3 x 104 [j,g/mL AN (7 x 10"2 or 1 x 10"1 M
AN) with or without S9 mix for 1 hour (Douglas et al., 1985). These concentrations were above
the cytotoxic concentrations of 5.31 and 53.1 [ig/mL AN (10"5 and 10"4 M) with and without
S9 mix, respectively. Thus, the observed DNA strand break effect was relatively weak in CHO
cells. Lakhanisky and Hendrickx (1985) reported that AN (concentrations not reported) did not
in induce DNA strand breaks in cultured CHO cells with or without S9. Therefore, cultured
CHO cells may not be a sensitive assay system to test for DNA strand breaks induced by AN
when compared with other cell cultures. On the other hand, DNA strand breaks was reported in
"3
SHE cells treated with AN. Parent and Casto (1979) observed incubation of [ H]-thymidine-
labeled primary SHE cells with 200 or 400 p,g/mL AN for 18 hours caused a shift in the
sedimentation pattern of the labeled cellular DNA when subjected to alkaline sucrose gradient.
4.5.2.4.3. Cytogenic effects. AN induced SCE in cultured adult human bronchial epithelial cells
treated with noncytotoxic concentrations of 150 or 300 [j,g/mL AN for 20 hours (Chang et al.,
1990). An increase in frequency of SCE was observed in human lymphocytes from two different
donors incubated for 1 hour with 5 x 10"4M AN and S9 mix. No increase in SCE was observed
without the S9 mix metabolizing system. (Perocco et al., 1982). AN was negative for the
237
DRAFT - DO NOT CITE OR QUOTE
-------
induction of SCEs in CHO cells (Ved Brat and Williams, 1982). However, when CHO cells
were cocultured with freshly isolated rat hepatocytes, Ved Brat and Williams (1982) observed
that AN at 10"4 M produced a greater than twofold increase in SCEs in the CHO cells, suggesting
that the rat hepatocytes metabolized AN to its reactive metabolite, which was then transported
into the CHO cells. Other cytogenetic findings in CHO cells included positive results for the
induction of SCEs at 2 mM AN with S9 (Natarajan et al., 1985) and CA at 4 mM AN with or
without S9, while Douglas et al. (1985) reported that AN at 10"1 M induced the formation of MN.
AN was reported to induce CAs in Chinese hamster lung (CHL) fibroblasts in culture without
metabolic activation at nontoxic concentrations of 12.5 ng/mL (Ishidate and Sofuni, 1985). AN
induced structural CAs in Chinese hamster liver fibroblast cell line (CH1-L) at the lowest
concentration of 2.5 [j.g/mL and higher (Danford, 1985).
Sasaki et al. (1980) found that 0.0053 mg/mL AN induced chromosome breaks in a
pseudodiploid Chinese hamster cell line (Don-6). AN was negative at concentrations up to
10 [ig/mL for the induction of CAs, SCEs, and polyploidy in cultured epithelial-like cells from
rat liver (RL4 cell line) (Priston and Dean, 1985; Shell Oil Co., 1984b). Mangir et al. (1991)
reported that CHO cells treated with AN demonstrated a growth and RNA synthesis rate that is
similar to that for agents that cause damage to nuclear DNA in cells. Growth and RNA synthesis
of CHO cells was inhibited with 0.001% (v/v) AN and completely inhibited at 0.005% (v/v) AN.
Kodama et al. (1989) conducted cytogenetic analyses on eight spontaneous and eight
CEO-induced Tks mutant clones in Tk6 human lymphoblastoid cultures that had lost the 14.8 kb
polymorphic band corresponding to the active Tk allele. These CEO-induced Tk'' mutants were
reported in the Recio and Skopek (1988a, b) studies. No chromosomal abnormalities were found
in the eight spontaneous mutants. On the other hand, a visible abnormality on chromosome 17
was found in one of the CEO-induced tks mutants and was marked by duplication of the long arm
of chromosome 17, with break points at ql 1 and q21. The latter break point was close to the
Tk locus, suggesting the observed aberration might be associated with Tk"/_ phenotype.
4.5.2.4.4. Transformation assays. Parent and Casto (1979) studied the capacity of AN to
induce transformations in primary SHE cells in culture by monitoring the incidence of
microscopically observed foci of morphologically transformed cells. In two experiments, SHE
cell cultures were exposed to 25-200 ng/mL AN for 18 hours, after which AN was removed and
cultures were subsequently inoculated with 200 focus-forming units of simian adenovirus SA7
and incubated for 3 hours. Colonies of surviving cells were counted after 8 days, and virus-
transformed foci were counted after 21 days. Pretreatment of cells with AN prior to viral
inoculation resulted in only slight enhancement of 1.8-fold in SA7 foci. In another experiment,
SHE cells were treated with the same concentrations of AN 5 hours after viral inoculation. This
resulted in an enhancement ratio that was markedly increased (8.9 at AN concentration of 200
[j,g/mL vs. 1.0 in controls).
238
DRAFT - DO NOT CITE OR QUOTE
-------
When SHE cells were treated for 6 days with 12-100 ng/mL AN without added
SA7 virus, foci of morphologically transformed cells were observed at 50 [j,g/mL AN (two
foci/six dishes) and 100 [j,g/mL AN (three foci/nine dishes).
The ability of AN to induce transformation in SHE cells was also evaluated by Barrett
and Lamb (1985). AN was considered to give a positive response according to the criterion of
inducing four or more transformed colonies per 2,000 surviving colonies. With a relative
survival of unity, the lowest concentration of AN (0.01 (j,g/mL) generated morphologically
transformed colonies at a rate of 4/1,149.
Lawrence and McGregor (1985) evaluated the ability of 10 potential carcinogens,
including AN, to induce morphological transformation in cultured embryonic mouse fibroblasts
(C3H/10T1/2, Clone 8) in the presence or absence of S9. Positive response was obtained at
16 [ig/mL AN in the presence of S9, while responses were uniformly negative in its absence.
Banerjee and Segal (1986) studied whether AN (0-200 ng/mL) could produce in vitro
transformation of C3H/10T1/2 and NIH/3T3 mouse fibroblast cells in culture. AN was
cytotoxic at the higher concentrations, with cell survival dropping below 75% at >50 |ag/m L in
C3H/10T1/2 and NIH/3T3 cells. Optimal transformation rates were obtained at AN
concentrations of 12.5 [ig/mL in C3H/10T1/2 cells. AN-induced transformation was observed at
concentrations between 3 and 100 ng/mL in NIH/3T3 cells.
Matthews et al. (1985) evaluated whether AN induced morphological transformation and
mutation to ouabain resistance in Balb/c-3T3 cells in culture with or without exogenous
metabolic activation. For activation transformation assay, Balb/c-3T3 cells were cocultured with
lethally X-irradiated primary F344 rat liver cells (RLCs). The RLC-3T3 cocultures were treated
with 0-25 [j,g/mL AN for 48 hours. The cocultures were then treated biweekly for 3 weeks with
0.05 [j,g/mL 12-O-tetradecenoyl-phorbol-13-acetate beginning 1-2 days after completion of AN
treatment. For nonactivation transformation assay, 3T3 cell cultures were incubated with 0-
20 [ig/mL AN for 72 hours. For RLC-3T3 cocultures, significant increase in relative
transformation activity was found in coculture treated with 8.8 ng/mL AN (noncytotoxic
concentration). At 16.7 [j,g/mL AN, relative cell survival was only 22%, and no significant
increase in transformation activity was found. No significant increase in transformation activity
was observed in AN-treated 3T3 cell cultures without RLCs.
For ouabain resistance (Ouar) mutation assay, 3T3 cell cultures were treated with 0-
150 [ig/mL AN with or without S9 mix for 4 and 24 hours, respectively. After the treatment, AN
was removed and the cultures were refed and maintained for 5-6 days for expression and
selection, using 2 mM cardiac glycoside ouabain. The appearance of ouabain-resistant (Qua1)
variants would indicate a mutation arose in the gene controlling the synthesis of cell membrane
Na+/K+ ATPase (Corsaro and Migeon, 1978). Significant increase in relative Ouar frequency
was observed at 50 [j,g/mL AN with S9 activation.
239
DRAFT - DO NOT CITE OR QUOTE
-------
Yuan and Wong (1991) used a nonfocus transfection-transformation assay to study the
capacity of the oxidative metabolite CEO to bring about functional changes in a plasmid that
would be indicative of a compound-induced mutation. A new plasmid that had been constructed
by ligating a human c-HA-ras-1 protooncogene to a pSV2neo mammalian vector was reacted
with CEO in vitro and then transfected into NIH3T3 cells. Cells were selected for neomycin
resistance and/or abnormal growth characteristics, the latter serving to discriminate between
colonies arising from ras mutations and those from cells that were not transfected (or that were
transfected with nonplasmid DNA). Although CEO-modified ras gave rise to two neomycin
resistant clones, they were probably not indicative of a ras mutation because their normal growth
rate and monolayer density were similar to negative control. Southern blot analysis of
transformant DNA also supported this conclusion. For example, when anti-benzo(a)pyrene-
7,8-dihydrodiol-9,10-epoxide transformant DNA was examined in this system as a positive
control, a fragment of 411 base pairs was revealed, indicating a ras mutation at codon 11 or 12.
However, both CEO-derived clones and untreated control showed the WT band of 355 base
pairs.
4.5.2.4.5. Genotoxicity summary. All identified studies concerning the mutagenicity or
genotoxicity of AN are compiled in Table 4-53. The overall weight of evidence from in vitro
and in vivo studies is adequate to support direct mutagenicity for the AN metabolite, CEO.
Table 4-53. Summary of studies on the mutagenicity/genotoxicity of AN
Test system
Endpoint/effect
Exposure
concentration
Exposure
duration
Result"
Reference
Humans
AN-exposed workers
n = 30
DNA strand breaks;
nondisjunction of sex
chromosomes in sperm
0.8 mg/m3
2.8 yrs
+
Xu et al.
(2003)
AN-exposed
workers:
31 controls, 41 low,
47 intermediate
Increase in MN in
buccal mucosal cells
(low and intermediate
groups) and blood
lymphocytes
(intermediate group)
of AN-exposed
workers.
Low =
0.522 mg/m3
Intermediate =
1.998 mg/m3
Low = 1-
33 yrs
(average
15.7 yrs)
Intermediate =
1-33 yrs
(average
17.2 yrs)
+
Fan et al.
(2006)
AN-exposed
polymerization
workers:
14 Maintenance
workers:
10 controls: 20
CAs in lymphocytes
NDb
ND
Maintenance
workers: +
Production
workers: (+)
Borba et al.
(1996)
AN-exposed
workers: 47;
47 controls
Deletion of
mitochondrial DNA in
lymphocytes
0.11 ppm
17.3 yrs
+
Ding et al.
(2003)
240
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-53. Summary of studies on the mutagenicity/genotoxicity of AN
Test system
Endpoint/effect
Exposure
concentration
Exposure
duration
Result"
Reference
AN-exposed workers
Group 1=39
Group 2 = 22
Unexposed controls
= 49
CAs (detected by
FISH) in cultured
lymphocytes from
peripheral blood
samples
Group 1 = 0.05-
0.3 mg/m3
Group 2 = 0.05-
0.7 mg/m3
3 mos
(+)
Significant
increase in the
number of
reciprocal
translocations and
relative number of
insertions.
Increase in
frequency of
aberrant cells not
significant.
Beskid et al.
(2006)
AN-exposed workers
(n = 45)
Matched controls =
23, unexposed
controls = 33
CAs in cultured
lymphocytes in
peripheral blood
samples
0.05-0.3 mg/m3
3 mos
Sram et al.
(2004)
AN-exposed workers
(n = 18);
controls = 18
CAs in lymphocytes
5 ppm, reduced
to 1.5 ppm
between 1975
and 1977
(possible
exposure to other
chemicals)
15.3 yrs
Thiess and
Fleig (1978)
Rats
Sprague-Dawley
(male) (4/group)
MN
125 mg/kg i.v.
Two
treatments
Bone marrow: +
Peripheral blood:
Wakata et al.
(1998)
Sprague-Dawley
(male)
MN in bone marrow
10-40 mg/kg oral
Single dose
-
Morita et al.
(1997)
MN in peripheral
blood
24.5-98 mg/kg
i.v.
Single dose
-
Morita et al.
(1997)
Sprague-Dawley
(male) n = 3
CAs in bone marrow
40 mg/kg oral
16 d
Rabello-Gay
and Ahmed
(1980)
Sprague-Dawley
(male) n = 3-4/group
Binding to DNA in
stomach, brain, and
liver
46.5 mg/kg oral
Single dose
+
Farooqui and
Ahmed (1983 a)
Sprague-Dawley
n = 4/group
Binding to testicular
DNA
46.5 mg/kg oral
Single dose
+
Ahmed et al.
(1992a)
Sprague-Dawley
(male) n = 4/group
Binding to lung DNA
46.5 mg/kg oral
Single dose
+
Ahmed et al.
(1992b)
Sprague-Dawley
(male) n = 3/group
Binding to gastric
tissue DNA
46.5 mg/kg oral
Single dose
+
Abdel-Rahman
et al. (1994b)
Wistar (male)
Alkylation of hepatic
DNA
0.2 mmol i.p.
Single dose
+
Peter et al.
(1983 a)
F344
n = 1
Binding to liver and
brain DNA
0.6 mg/kg CEO
1.p.
Single dose
Hogy and
Guengerich
(1986)
241 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-53. Summary of studies on the mutagenicity/genotoxicity of AN
Test system
Endpoint/effect
Exposure
concentration
Exposure
duration
Result"
Reference
F344
n = 3
DNA adduct formation
50 mg /kg AN
i.p. or 6 mg/kg
CEO i.p.
Single dose
Liver: +
Brain: (+)
Hogy (1986);
Hogy and
Guengerich
(1986)
Wistar
DNA damage in
forestomach, colon,
kidney, bladder, and
lung but not in brain or
bone marrow
30 mg/kg i.p.
Single dose
+
Sekihashi et al.
(2002)
Wistar, male
Fragmentation of brain
DNA
100 ppm AN in
drinking water
14 or 28 d
+
Mahal akshmi
et al. (2003)
F344
n = 12/group
UDS in liver but not
brain
50 mg/kg gavage
Single dose
+
Hogy and
Guengerich
(1986)
Sprague-Dawley
UDS in lung
46.5 mg/kg oral
Single dose
+
Ahmed et al.
(1992a)
UDS in gastric tissue
46.5 mg/kg oral
Single dose
Abdel-Rahman
et al. (1994a)
UDS in testis
46.5 mg/kg oral
Single dose
Ahmed et al.
(1992b)
F344
UDS in isolated
hepatocytes or
spermatocytes
a. 75 mg/kg
b. 60 mg/kg oral
a. Single dose
b. five daily
doses
Butterworth et
al. (1992)
F344
n = 50
Dominant lethal
mutations
60 mg/kg
5 d
-
Working et al.
(1987)
Mice
ddY
DNA damage in
forestomach, colon,
bladder, lung, and
brain
20 mg/kg i.p.
Single dose
+
Sekihashi et al.
(2002)
C57B1/6
n = 1-4/group
SCEs in bone marrow
10-45 mg/kg
Single dose
(+) positive at
toxic dose of
45 mg/kg
Sharief et al.
(1986)
Swiss
SCEs in bone marrow
7.5 or 10 mg/kg
i.p.
Single dose
+
Fahmy (1999)
CAs in spermatocytes,
bone marrow, and
spleen cells
a. 15.5 or
31 mg/kg oral
b. 7.75 mg/kg
a. Single dose
b. three or five
doses
+
Fahmy (1999)
Swiss
n = 3-6/group
CAs in bone marrow
7, 14, or
21 mg/kg-d oral
or 10, 15, or
20 mg/kg-d i.p.
4, 15, or 30 d
Rabello-Gay
and Ahmed
(1980)
NMRI male
n = 4/group
CAs in bone marrow
20 or 30 mg/kg
i.p.
Single dose
-
Leonard et al.
(1981)
C57B1/6
n = 1-4/group
CAs in bone marrow
10-45 mg/kg
Single dose
-
Sharief et al.
(1986)
NMRI male
n = 4-5/group
MN in erythrocytes
20 or 30 mg/kg
i.p.
Single dose
-
Leonard et al.
(1981)
242 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-53. Summary of studies on the mutagenicity/genotoxicity of AN
Test system
Endpoint/effect
Exposure
concentration
Exposure
duration
Result"
Reference
CD-I
MN in bone marrow
0-45 mg/kg i.p.
0-32 mg/kg oral
0-40 mg/kg i.v.
Single dose
i.p.: (+)
oral and i.v.: -
Morita et al.
(1997)
CD-I
MN in peripheral
blood
5.6-45 mg/kg i.p.
or 10-40 mg/kg
i.v.
Single dose
Morita et al.
(1997)
ICR
CAs in bone marrow
cells and
spermatogonia
100 or 20 mg/m3
5 d
Zhurkov et al.
(1983)
NMR1
n = 5/group
Dominant lethal
mutation
30 mg/kg i.p.
Single dose
-
Leonard et al.
(1981)
Short-term assays—bacteria
(-S9 / +S9)
S. typhimurium
TA 1535
Gene reversion (His+
revert ant)
5-20 |xL AN
solution
0.5 h
-/+
Milvy and
Wolff (1977)
TA 1535
Gene reversion
2-300 |xL AN
vapor
0.5-4 h
-/+
Milvy and
Wolff (1977)
TA 1538
Gene reversion
200 |xL AN vapor
2 h
-/+
Milvy and
Wolff (1977)
TA 1978
Gene reversion
5-10 |xL AN
solution
0.5 h
-/+
Milvy and
Wolff (1977)
TA 1535
Gene reversion
100-1,000 jxg
AN plate
incorporation
assay
ND
-/+
Lijinsky and
Andrews
(1980)
TA 1537, 1538, 98,
100
Gene reversion
Plate
incorporation
assay
ND
-/-
Lijinsky and
Andrews
(1980)
TA 1535
Gene reversion
0.1-
1,000 |Lig/dish
ND
-/+
Zhurkov et al.
(1983)
TA 1538
Gene reversion
0.1-
10,000 |Lig/dish
ND
-/-
Zhurkov et al.
(1983)
TA 97, 98, 100
Gene reversion
50-750 |j,g/mL
plate
incorporation
assay
48 h
-/-
Brams et al.
(1987)
TA 102
Gene reversion
Up to
5,000 |Lig/plate
ND
-/-
Jung et al.
(1992)
TA 98, 100, 1535,
1537, 1538
Gene reversion
0.1-5,000 |j,g/
plate
2d
-/-
Dow Chemical
Co. (1977)
TA 1530, 1535,
1950, 1538, 100, 98,
1978
Gene reversion
0.2% AN vapor
(200 |Lig/plate)
1 h
-/+ (weaker
response with
TA 100, 98, and
1978)
de Meester et
al. (1978);
Dow Chemical
Co. (1976)
TA 98, 100
Gene reversion
ND, plate
incorporation
assay
ND
+/+
Khudoley et al.
(1987)
TA 100
Gene reversion
806-12,100 |ig/
plate
48 h
-/(+)
Hakura et al.
(2005)
243 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-53. Summary of studies on the mutagenicity/genotoxicity of AN
Test system
Endpoint/effect
Exposure
concentration
Exposure
duration
Result"
Reference
TA 1535/pSK1002
umu gene expression
(increased
P-galactosidase
activity)
ND, 0.1 mL AN
in 2.5 mL culture
medium
2 h
-/-
Nakamura et
al. (1987)
TA 1530
Gene reversion
0.1 mL 24-hr
urine from rats
and mice treated
with a single dose
30 mg/kg AN i.p.
(plate
incorporation
assay)
48 h
+
Lambotte-
Vandepaer et
al. (1985, 1981,
1980)
E. coli WP2 series
Gene reversion
(trp~ —> trp+)
75 or 150 |jmol/
plate
2-3 d
+°
Venitt et al.
(1977)
E. coli PQ37
SOS chromotest
ND
2 h
-/-
Brams et al.
(1987)
Short-term assays—-fungi
S. cerevisiae JD1
Mitotic gene
conversion
250 or
500 |j.g/mL
18 h
-/+
Brooks et al.
(1985); Shell
Oil Co. (1984a)
S. cerevisiae D61.M
Chromosome loss
0.8 or
1.36 mg/mL
16 h
-/-
Whittaker et al.
(1990)
Short-term assays—-fruitfly
Adult female
Drosophila ZESTE
system
Sex chromosome loss
Inhalation
exposure to
2.7 ppm AN
50 min or
70 min
+
Osgood et al.
(1991)
Mosaic eye in
hatching females
Somatic recombination
and mutation
Treatment of
larvae with 5-
20 mM AN
Single dose,
incubate for
9-11 d
(+)
Vogel (1985)
Germ cells of male
D. melanogaster
Sex-linked recessive
lethal
Feeding:
420 ppm or
injection:
3,500 ppm
Feeding: 3 d
Foureman et al.
(1994)
Wing spots of
Drosophila
(mwh+l+jlr +/mei-9)
Somatic mutation and
recombination
Gas exposure of
larvae to 0.5-
1 (xL/1,150 mL
0.5 or 1 h
(+)
Wiirgler et al.
(1985)
D. melanogaster
ZESTE (inhibition of
taxol-purified
microtubule
assembly in vitro)
Aneuploidy
5 mM
Microtubule
assembly
monitored for
80 min
+
Sehgal et al.
(1990)
Other short-term assays
Supercoiled plasmid
DNA pBR322
CEO induced DNA
strand breaks
50 mM CEO
incubated with
1.1 ng
supercoiled
pBR322 plasmid
DNA
3 h
+
Yates et al.
(1994)
244 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-53. Summary of studies on the mutagenicity/genotoxicity of AN
Test system
Endpoint/effect
Exposure
concentration
Exposure
duration
Result"
Reference
SV40 phage DNA
CEO induced DNA
strand breaks, but not
AN
1 mM CEO
incubated with
5,000 dpm
[3H]-thymidine
labelled SV-40
phage DNA
17 h
+
Peter et al.
(1983b)
In vitro mammalian cell assays
Human lymphoblasts
716
Gene-locus mutations
at the Tk (thymidine
kinase) locus (71+A—>
Tk^)
40 (Xg/mL AN
3 h
-/+
Crespi et al.
(1985)
Mutation at Tk locus
1.4 mM AN
2 h
-/+
Recio and
Skopek (1988a,
b)
Mutation at Tk locus
100 (xM and
150 (xM CEO
2 h
+
Recio and
Skopek (1988a,
b)
Abnormality at
chromosome 17 in 1 of
8 CEO-induced tks
mutant clones
NAb
NA
+
Kodama et al.
(1989)
CEO-induced
mutations at the hprt
locus
NA
NA
+
Recio and
Skopek
(1988a)
Human lymphoblasts
AHH-1
Gene-locus mutations
at the hypoxanthine
guanine
phosphoribosyl
transferase locus
25 (xg/mL AN
28 h
+
Crespi et al.
(1985)
L5178Y mouse
lymphoma cells
Mutation at Tk+,~ locus
30 and 40 (xg/mL
AN
ND
+
Oberly et al.
(1996)
Mutation at Tk+/~ locus
10-40 (Xg/mL AN
4 h
+/+
Rudd (1983)
Mutation to ouabain or
6-TG resistance
12.5-200 (xg/mL
AN
2 h
+/+
Garner and
Campbell
(1985)
Induction of TkT^
mutants
80-225 (Xg/mL
AN
2 h
+/+
Lee and
Webber (1985)
Induction of TkT^
mutants
30 nL/mL AN
4 h
+/+
Myhr et al.
(1985)
Induction of TkT^
mutants
5-69 (Xg/mL AN
3 h
+/+
Amacher and
Turner (1985)
L5178Y 71+/+
Mutation to ouabain
resistance (Na+/K+
ATPase locus)
12.5-100 (Xg/mL
AN
2 h
Styles et al.
(1985)
L5178Y71+/~
Mutation to
trifluorothymidine
resistance
12.5-100 (Xg/mL
AN
2 h
Styles et al.
(1985)
245 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-53. Summary of studies on the mutagenicity/genotoxicity of AN
Test system
Endpoint/effect
Exposure
concentration
Exposure
duration
Result"
Reference
P388F mouse
lymphoma tk+/~
Mutation to 5-iodo-
2-deoxyuridine
resistance
80-160 (xg/mL
AN
24-48 h
-/+
Anderson and
Cross (1985)
V79/hprt
Induction of
8-azaguanine
resistance
50-200 (xg/mL
AN
2 h
-/-
Lee and
Webber (1985)
HeLa cells
UDS
2.5 mg/mL AN
2.5 h
-/-
Martin and
Campbell
(1985)
Human lymphocytes
UDS
5 x 10"1 M
4 h
-/+
Perocco et al.
(1982)
Primary cultures of
F344 rat hepatocytes
UDS
0.026-53 (xg/mL
AN
20 h
-
Probst and Hill
(1985)
Primary cultures of
F344 rat hepatocytes
UDS
10-1-102 (Xg/mL
AN
18-20 h
-
Williams et al.
(1985)
Primary F344 rat
hepatocyte
UDS
0.01-1 mM AN
17-19h
-
Butterworth et
al. (1992)
Primary F344 rat
hepatocyte
UDS
0.01-0.1 mM
CEO
17-19h
-
Butterworth et
al. (1992)
Human mammary
epithelial cell
UDS
0.1 mM CEO
24 h
+
Butterworth et
al. (1992)
Human bronchial
epithelial cells
DNA single-strand
breaks
200 and
500 (xg/mL AN
20 h
+
Chang et al.
(1990)
Rat hepatocytes
DNA single-strand
breaks
65.8 (xg/mL AN
3 h
+
Bradley (1985)
CHO cells
DNA single-strand
breaks
7 x 10"2-
1 x 10"1 M AN
1 h
+/+
Douglas et al.
(1985)
CHO cells
DNA single-strand
breaks
ND
ND
-/-
Lakhanisky
and Hendrickx
(1985)
SHE cells
DNA single-strand
breaks
200 or
400 (xg/mL AN
18 h
+
Parent and
Castro (1979)
Human bronchial
epithelial cells
SCEs
150 and
300 (xg/mL AN
20 h
+
Chang et al.
(1990)
Human lymphocytes
SCEs
5 x 10"4M
1 h
-/+
Perocco et al.
(1982)
CHO cells
SCEs
10"7-10"4M AN
3 h
d
Ved Brat and
Williams
(1982)
CHO cells cocultured
with freshly isolated
rat hepatocytes
SCEs
10-4M AN
3 h
+d
Ved Brat and
Williams
(1982)
CHO cells
SCEs
2mM AN
1 h
-/+
Natarajan et al.
(1985)
CHO cells
CAs
4mM AN
1 h
+/+
Natarajan et al.
(1985)
CHL fibroblast
CAs
12.5 (xg/mL AN
24 and 48 h
+d
Ishidate and
Sofuni (1985)
246 DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-53. Summary of studies on the mutagenicity/genotoxicity of AN
Test system
Endpoint/effect
Exposure
concentration
Exposure
duration
Result"
Reference
CH1-L liver
fibroblast
CAs
2.5 (xg/mL AN
36 h
+
Danford (1985)
Chinese hamster cell
line Don-6
CAs
1 x 10"4 M or
0.0053 mg/mL
26-30 h
+
Sasaki et al.
(1980)
CHO cells
MN
10-1 M AN
1 h
+/+
Douglas et al.
(1985)
Rat liver (RL4) cells
CAs, polyploidy,
SCEs
1.25,2.5,5.0, or
10 (ig/mL AN
2 h
d
Priston and
Dean (1985);
Shell Oil Co.
(1984b)
CHO cells
Inhibition of cell
growth and RNA
synthesis
0.001, 0.002, and
0.005% (v/v) AN
8 d
+d
Mangir et al.
(1991)
In vitro mammalian cell transformation
SHE cells
Cell transformation
50 or 100 (xg/mL
AN
6 d
+
Parent and
Castro (1979)
SHE cells
Enhancement of viral
transformation
100 or
200 (ig/mL AN
18 hrs before
SA7 or 5 hrs
after SA7
inoculation
+
Parent and
Castro (1979)
SHE cells
Cell transformation
0.01-1 ng/mL
AN
7 d
+
Barett and
Lamb (1985)
Mouse fibroblasts
NIH/3T3 cells
Cell transformation
12.5-200 ng/mL
AN
48 h
+
Banerjee and
Segal (1986)
Mouse fibroblasts
C3H/10T1/2 cells
Cell transformation
3-100 (ig/mL AN
48 h
+
Baneijee and
Segal (1986)
Mouse fibroblasts
C3H/10T1/2 cells
Cell transformation
16 (xg/mL AN
24 h
-/(+)
Lawrence and
McGregor
(1985)
Balb/c-3T3 cells
Cell transformation
8.8 (xg/mL AN
48 h
-/+'
Matthews et al.
(1985)
Balb/c-3T3 cells
Ouabain-resistant
mutants
50 (xg/mL AN
24 h
-/+
Matthews et al.
(1985)
NIH 3T3 cells
Cell transformation
after transfection with
CEO modified ras
DNA
NA
14 h
Yuan and
Wong (1991)
aNA = not applicable; ND = not determined.
b+ = Positive;- = negative; (+) = borderline positive.
°S9 activation not needed.
dS9 not included in the assay.
eWith or without coculture with lethally X-irradiated primary F344 RLCs.
In vitro evidence indicates that AN can be directly mutagenic, most likely through the
formation of CEO-DNA adducts. In short-term tests with bacteria, AN induced mutations in a
majority of test systems, most often requiring exogenous metabolic activation (Table 4-53). In
247
DRAFT - DO NOT CITE OR QUOTE
-------
mouse lymphoma cell assays, AN induced mutations at the Tk locus in most assays (Table 4-53).
In human lymphoblast Tk6 cells devoid of CYP450 activity, AN induced multiple mutations at
the Tk locus only in the presence of metabolic activation (Recio and Skopek, 1988a, b; Crespi et
al., 1985). CEO was mutagenic in Tk6 cells at 10-fold lower concentrations than AN itself and
was as mutagenic in this test system as the well-known mutagen, ethyl methanesulfonate (Recio
and Skopek, 1988a, b). CEO also induced multiple mutations at the hprt locus in Tk6 cells
(Recio and Skopek, 1988a). CEO induced UDS in cultured human mammary epithelial cells but
not in cultured rat hepatocytes (Butterworth, 1992).
Supporting evidence for the direct mutagenicity of AN and its metabolites comes from in
vivo studies (Table 4-53). Increases in the occurrence of MN in buccal mucosal cells and blood
lymphocytes were recently reported in AN exposed workers in China (Fan et al., 2006). Three
studies of AN workers reported genotoxic effects such as DNA strand breaks, nondisjunction of
sex chromosomes, and CAs (Beskid et al., 2006; Xu et al., 2003; Borba et al., 1996), whereas an
earlier study did not find elevated frequency of CAs in exposed workers (Thiess and Fleig,
1978). Increases in cytogenetic aberrations such as MN were found in assays of exposed rats
(Wakata et al., 1998) and a single assay of mice (Fahmy, 1999), but were not evident in other rat
and mouse assays (Morita et al., 1997; Zhurkov et al., 1983; Leonard et al., 1981; Rabello-Gay
and Ahmed, 1980). Comet assays found DNA damage in forestomach, colon, bladder, lung, and
brain in mice, following single i.p. injections of 20 mg/kg AN, and in forestomach, colon,
kidney, bladder, and lung of rats injected with 30 mg/kg (Sekihashi et al., 2002).
"3
UDS was detected by following the time course of H-thymidine incorporation into DNA
in lung, testis, and gastric tissue, following administration of single oral doses of 46.5 mg/kg AN
to Sprague-Dawley rats (Ahmed et al., 1996b, 1992a, b; Abdel-Rahman et al., 1994a).
Autoradiographic techniques did not detect UDS following incubation of primary cultures of
hepatocytes or spermatocytes from F344 rats given single oral doses of 75 mg/kg or five daily
doses of 60 mg/kg-day AN (Butterworth et al., 1992). Dominant lethal effects (from mutations
in germ cells) were not found in mice given single i.p. doses of 30 mg/kg AN (Leonard et al.,
1981) or rats given five oral doses of 60 mg/kg (Working et al., 1987). DNA binding by AN or
its metabolites was indicated by elevated levels of radioactivity in DNA from several tissues in
rats given single oral doses of 46.5 mg/kg radiolabeled AN (Abdel-Rahman et al., 1994b; Ahmed
et al., 1992a, b; Farooqui and Ahmed, 1983a).
In the European Union Risk Assessment Report on Acrylonitrile (EC 2004), AN was
regarded as "genotoxic or at least mutagenic, despite the recent publication of Whysner et al.
(1998a) which argues for a possible nongenotoxic mechanism for the tumour induction in
experimental animals."
4.5.2.5. Indirect Genotoxicity
As discussed in Section 4.5.1.2.2, studies are available that investigated the indirect
248
DRAFT - DO NOT CITE OR QUOTE
-------
genotoxicity of AN resulting from oxidative stress. These studies measured 8-oxodG levels in
DNA as biomarkers of oxidative DNA damage from AN exposure. 8-OxodG is mutagenic
(Kamiat et al., 1992; Moriya et al., 1991; Wood et al., 1990) and causes GC —~ TA transversions
during DNA replication.
In vitro studies
Zhang et al. (2000) studied AN-induced morphological transformation in SHE cells.
SHE cell culture was treated in vitro for up to 7 days with 0-75 [j,g/mL AN in 12.5 [j,g/mL
increments (75 [j,g/mL was cytotoxic, with 50% reduction in cell colony number). After 7 days
of exposure, there was a dose-dependent increase in morphological transformation at 50, 62.5,
and 75 (J,g/mL, reaching a transformation frequency of 1.3% at 75 [j,g/mL. After a 24-hour AN
exposure, an increase in transformation was not observed at all concentrations. Levels of
8-oxodG in DNA isolated from cells incubated with 75 |ig/mL were increased to 192 and 186%
of control after 2 and 3 days, indicating an association between cell transformation and oxidative
DNA damage. However, no increase in 8-oxodG was observed after 1 or 7 days. AN-induced
morphological transformation was inhibited by cotreatment with the antioxidants a-tocopherol
(100 |iM: up to 65%) inhibition) and (-)-epigallocathechin-3-gallate (5 [jM: up to 87%>
inhibition) for 7 days. Cotreatment with antioxidants also inhibited the formation of 8-oxodG in
SHE cell DNA from treatment with 75 [ig/mL AN.
In a later study by the same research group, Zhang et al. (2002) investigated the time
course of SHE cell morphological transformation in the presence of 75 [j,g/mL AN. The results
indicated that statistically significant increases in morphological transformation frequency could
not be observed until after 2 consecutive days of exposure; a plateau at a transformation
frequency of about 2%> was reached after 4-5 days of exposure. In another experiment,
coadministration of 0.5 mmol/L ABT, a nonspecific suicidal CYP450 inhibitor, for 7 days
significantly reduced the rate of cell transformation from about 1.25% to about 0.3% (shown
graphically), demonstrating the need for metabolic activation of AN in this test system.
Zhang et al. (2002) also showed that AN (25, 50, and 75 (j,g/mL) increased the amount of
ROS (measured by 2,3-dihydroxybenzoic acid production) in the SHE cells after 4, 24, and
48 hours of treatment. At the same time, xanthine oxidase (which generates the superoxide
radical and hydrogen peroxide via oxidation of hypoxanthine or xanthine by oxygen) activity
was increased by 47% in SHE cells after 24 hours of treatment with 75 [ig/mL AN. After
48 hours of treatment, xanthine oxidase activity was increased in both 50 and 75 [j,g/mL (80%)
AN groups. This increase in xanthine oxidase activity was blocked by cotreatment with 0.5 mM
ABT. On the other hand, antioxidant GSH was depleted 66-80% by all doses of AN (25, 50, and
75 ng/mL) after 4 hours of treatment and returned to control levels after 24 hours. At 48 hours, a
significant increase in GSH was observed with the 75 ng/mL group but not in other dose groups.
Antioxidant enzyme catalase activity was significantly decreased after 4 hours of treatment with
249
DRAFT - DO NOT CITE OR QUOTE
-------
50 and 75 ng/mL AN but increased after 24 and 48 hours of treatment. Cotreatment with ABT
prevented the decrease and increase in activity of catalase at 4 and 24 hours after treatment,
respectively. A transient decrease in SOD activity was also observed after 4 hours of treatment
with 75 [j,g/mL AN.
The effect of AN on catalase and SOD activities was also studied in a cell-free system
(Zhang et al., 2002). Catalase and SOD activities were determined in samples containing
purified catalase or SOD incubated with 75 ng/mL AN in the presence or absence of SHE cell
homogenate. In the absence of a metabolic source of SHE cell homogenate, no inhibition of
catalase activity was seen following incubation up to 60 minutes with AN. In the presence of
SHE cell homogenate, AN significantly decreased catalase activity in a time-dependent manner
after 10 minutes of incubation. Similarly, a significant time-dependent decrease in SOD activity
was observed following 30 minutes incubation with AN. The study authors concluded that
morphological transformation of SHE cells is caused by oxidative stress as a result of oxidative
metabolism of AN.
Kamendulis et al. (1999a) also investigated oxidative DNA damage induced by AN in
DITNC1, a rat glial astrocyte cell line, and primary rat hepatocytes exposed to AN in vitro (see
Section 4.5.1.2.2). AN was cytotoxic at concentrations >2.5 mmol/L (133 (j.g/mL) (as measured
by the release of LDH from the cells) to both cell lines, following >4 hours of exposure.
Concentrations of 0.01, 0.1, and 1.0 mmol/L AN (0.53, 5.3, and 53 (J,g/mL, respectively) caused
a dose-dependent increase in formation of 8-oxodG in astrocytes but not in hepatocytes.
Corresponding increases in ROS formation were also observed in astrocytes only. However, no
oxidative lipid damage (as evaluated by formation of MDA, a product of lipid peroxidation) was
found in either cell type following treatment with AN at all exposure concentrations or durations.
The formation of 8-oxodG in rat astrocytes was reversible. Following treatment with AN for
24 hours and removal of AN for 24 hours afterwards, 8-oxodG levels returned to control values
in all concentrations examined. Kamendulis et al. (1999a) concluded that this demonstrated
property was consistent with tumor promoting agents.
Pu et al. (2006) investigated oxidative DNA damage induced by AN in D1TNC1 rat
astrocyte cell line using the fpg-modified comet assay. Increase in oxidative DNA damage was
observed in astrocytes treated with 1 mM AN for 24 h. No increase in oxidative DNA damage
was observed in astrocytes treated with 0.005 - 0.75 mM AN for 24 hours.
Jacob and Ahmed (2003b) investigated oxidative stress in cultured NHA (4631) treated
with up to 400 |iM AN for 12 hours. AN was cytotoxic at concentrations >50 [jM. Cell viability
was 85, 78, and 58%, respectively, at AN concentrations of 100, 200, and 400 [xM. Significant
increases in measures of oxidative stress were observed at cytotoxic concentrations of 200-400
|iM AN: the production of ROS was increased four- to sevenfold, whereas 8-oxodG levels were
increased more than twofold.
250
DRAFT - DO NOT CITE OR QUOTE
-------
Esmat et al. (2007) investigated cytoxicity in rat (strain not known) primary glial cells
exposed to 0-5.0 mM AN up to 12 hours. Cell membrane integrity was evaluated by trypan blue
exclusion and LDH leakage. About 50% membrane damage in primary glial cells was observed
in incubations containing 1.0 mM AN for 3 hours. Thus, subsequent studies on AN-induced
oxidative stress were performed using 1 mM AN for 3 hours incubation. AN increased MDA
levels (indicator of lipid peroxidation) to about ninefold compared with control incubations and
depleted GSH level to about 7% of controls., while no change in total glutathione level was
observed. AN induced CN formation by glial cells and decreased ATP level by about 90% as
compared with the control. Pretreatment with 5 mM NAC (an acetylated precursor of cysteine
and GSH, and an antioxidant), reduced MDA level by 40% as compared with glial cells treated
with AN alone and raised GSH and total glutathione level in cell extract to about 2.5-fold of
control and AN alone treated group. Pretreatment with NAC also caused reduction of CN~
induced by AN to about 15% as compared wth the AN alone treated group, and raised ATP level
to about sixfold, as compared with the AN alone treated group.
It should be noted that NAC stimulated GSH synthesis, enhanced glutathione-S-
transferase activity, and was a powerful nucleophile capable of scavenging free radicals (De
Vries and De Flora, 1993). Observed increases in GSH and total glutathione and decrease in CN"
formation with NAC pretreatment in Esmat et al. (2007) would suggest that most of the
administered AN could be detoxified via conjugation via the GSH pathway (consistent with
findings by Carerra et al., 2007), as less AN was available for oxidation to CN". Esmat et al.
(2007) concluded that AN toxicity was at least partly mediated by oxidative stress.
In vivo studies
As described in Section 4.5.1.2.2, Jiang et al. (1998) reported increases in measures of
oxidative stress in the brain cortices but not the livers of male Sprague-Dawley rats exposed to
AN in drinking water at concentrations of 50-200 ppm for up to 90 days. Increases in ROS and
oxidative DNA damage (increased levels of 8-oxodG) and concomitant decreases in antioxidant
enzymes (catalase, SOD) were observed in the brain cortices of exposed rats. Transient small
decreases in the antioxidants vitamin E and glutathione were also observed in the brain cortices
of AN-exposed rats only at 14 days. A transient increase in MDA level was observed only at the
highest dose group after 14 days but not at other time points.
Pu et al. (2009) reported oxidative DNA damage in WBC and brain of male Sprague-
Dawley rats treated with 100-200 ppm AN; however, EPA has identified issues with this study
(see Section 4.5.1.2.2).
In another study, Whysner et al. (1998a) examined the ability of AN to induce oxidative
DNA damage in the brain, liver, and forestomach of rats by exposing male Sprague-Dawley and
F344 rats up to 300 ppm AN in drinking water for 21 days (see Section 4.5.1.2.2). As shown in
Table 4-54, elevated levels of 8-oxodG were found in DNA from the brain and liver of Sprague-
251
DRAFT - DO NOT CITE OR QUOTE
-------
Dawley rats. However, 8-oxodG levels in the forestomach of exposed Sprague-Dawley rats and
the brain of exposed F344 rats were not statistically significantly elevated compared with
controls. Significant increase in 8-oxodG levels was found in the liver of exposed Sprague-
Dawley rats, although the liver is not a target organ for carcinogenicity in adult rats.
Table 4-54. Formation of 8-oxodG in DNA from tissues of male Sprague-
Dawley and F344 rats exposed to AN in drinking water for 21 days
Dose group
Formation of 8-oxodG (mol/105 mol dG)
Sprague-Dawley rats
F344 rats
Liver
Forestomach
Brain
Brain
Control
0.67 ±0.22
0.68 ±0.14
0.62 ±0.08
0.79 ±0.37
3 ppm
0.72 ±0.06
1.77 ±0.55
0.86 ±0.41
1.07 ±0.41
30 ppm
0.95 ± 0.19a
1.59 ± 0.30
1.35 ± 0.49a
1.03 ±0.38
300 ppm
0.96 ± 0.15a
1.44 ± 1.22
1.29 ± 0.10a
1.06 ± 0.48b
"Significantly different from controls (p < 0.05) as calculated by the study authors.
bExposure was at 100 ppm.
Source: Whysner et al. (1998a).
Whysner et al. (1998a) also exposed male Sprague-Dawley rats to 100 ppm AN in
drinking water for up to 94 days. (This dose was carcinogenic in Sprague-Dawley rats in a
chronic study [Johannsen and Levinskas, 2002a; Biodynamics, 1980b].) The rats were divided
into four groups and were exposed to distilled water, 100 ppm AN, 5 mg MNU (a DNA-reactive
carcinogen that produces glial cell tumors in rats) per week, or 100 ppm AN plus 5 mg MNU per
week. Levels of 8-oxodG in brain and liver of rats exposed to 100 ppm AN were significantly
greater than those in controls after 10 days (1.31 ± 0.52 in brains of exposed rats vs. 0.65 ±
0.22 mol per 105 mol dGin controls and 0.70 ± 0.20 in livers of exposed rats vs. 0.49 ±0.10 mol
per 105 mol dG in controls). Administration of 5 mg/kg MNU alone did not increase the level of
8-oxodG in the brain of treated rats but increased 8-oxodG in the liver after 10 days. However,
coadministration of 100 ppm AN and MNU increased 8-oxodG level in the brain after 31 days
and 94 days when compared with the MNU-only group.
Whysner et al. (1998a) suggested that AN-induced tumors may be produced by a mode of
action involving 8-oxodG. However, several findings in this study did not support this proposed
mode of action. First, no significant increase in 8-oxodG levels were found in the brain DNA of
F344 rats exposed to AN in the 21-day study whereas AN was carcinogenic to F344 rats in a
chronic drinking water study (Johannsen and Levinskas, 2002b). Second, although a significant
increase in 8-oxodG levels was found in the brain DNA of Sprague-Dawley rats exposed to AN
for 21 days, no dose-dependent increase was observed above 30 ppm, which was not the dose
producing the highest occurrence of tumors in the chronic bioassay. Third, no increase in
252
DRAFT - DO NOT CITE OR QUOTE
-------
8-oxodG levels were found in the forestomach DNA of exposed rats. The forestomach was a
target organ for AN carcinogenicity. Finally, increase in 8-oxodG levels was found in liver
DNA of exposed rats. Liver was not a target organ for AN carcinogenicity in adult rats.
Therefore, there was no association between 8-oxodG levels and tumorigenicity in target organs.
Results of indirect mutagenicity/genotoxicity studies of AN are summarized in
Table 4-55.
253
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-55. Summary of studies on the indirect mutagenicity or genotoxicity
of AN
Test system
Endpoint/effect
Exposure
concentration
Exposure
duration
Result
Reference
In vitro mammalian cell assays
SHE cells
Cell
transformation
50-75 (ig/mL AN
7 d
+
Zhang et al.
(2000)
SHE cells
8-oxodG in DNA
75 (xg/mL AN
Increased after
2 and 3 d but not
after 1 or 7 d
(+)
Zhang et al.
(2000)
SHE cells
ROS
25-75 (ig/mL AN
4-48 h
+
Zhang et al.
(2002)
DITNClrat glial
astrocytes
8-oxodG in DNA
0.01-1 mM AN
(0.53-53 ng/mL)
4 or 24 h
+a
Kamendulis et al.
(1999a)
D1TNC1 rat
astrocytes
DNA damage in
fpg-comet assay
1 mM
24 h
+
Pu et al. (2006)
Rat hepatocytes
8-oxodG in DNA
0.01-1 mM AN
4 or 24 h
-
Kamendulis et al.
(1999a)
NHAs
8-oxodG in DNA
200-400 (xM AN
12 h
+
Jacob and Ahmed
(2003b)
NHAs
ROS
200^100 (xM AN
12 h
+
Jacob and Ahmed
(2003b)
In vivo studies in rats
Male Sprague-
Dawley rats
8-oxodG in brain
cortex DNA
0-200 ppm AN in
drinking water
100 and 200 ppm:
+ after 14-90 d
50 ppm: + after
28 and 90 d
+
Jiang et al. (1998)
Male Sprague-
Dawley rats
8-oxodG in liver
DNA
0-200 ppm AN in
drinking water
14-90 d
-
Jiang et al. (1998)
Male Sprague-
Dawley rats
8-oxodG in WBC
and brain DNA
100 or 200 ppm
AN in drinking
water
28 days
+
Pu et al. (2009)
Male Sprague-
Dawley rats
8-oxodG in brain
DNA
30 or 300 ppm
AN in drinking
water
21 d
+
Whysner et al.
(1998a)
Male Sprague-
Dawley rats
8-oxodG in liver
DNA
30 or 300 ppm
AN in drinking
water
21 d
+
Whysner et al.
(1998a)
Male Sprague-
Dawley rats
8-oxodG in
forestomach DNA
0-300 ppm AN in
drinking water
21 d
-
Whysner et al.
(1998a)
Male F344 rats
8-oxodG in brain
DNA
0-100 ppm AN in
drinking water
21 d
-
Whysner et al.
(1998a)
Male Sprague-
Dawley rats
8-oxodG in brain
DNA
100 ppm AN in
drinking water
3-94 d
+
Whysner et al.
(1998a)
aThe formation of 8-oxodG was reversible. Following treatment with AN for 24 hrs and removal of AN for 24 hrs
afterwards, 8-oxodG levels returned to control values.
+ = Positive;- = negative; (+) = borderline positive.
254
DRAFT - DO NOT CITE OR QUOTE
-------
4.6. SYNTHESIS AND EVALUATION OF MAJOR NONCANCER EFFECTS AND
MODE OF ACTION—ORAL AND INHALATION
4.6.1. Oral
No studies are available regarding noncancer health effects in humans following acute,
subchronic, or chronic oral exposure to AN.
The chronic oral noncancer toxicity database for AN consists of results from four rat
toxicity and cancer bioassays (one with F344 rats and three with Sprague-Dawley rats) (see
Table 4-56 for references), one toxicity and cancer bioassay with B6C3Fi mice (NTP, 2001),
and a three-generation developmental/reproductive toxicity study with Sprague-Dawley rats
(Friedman and Beliles, 2002; Litton Bionetics, 1992). NOAELs and LOAELs for noncancer
effects in these studies are summarized in Table 4-56. Also included in Table 4-56 are
summaries of results from a 12-week gavage study of nerve conduction velocities in male
Sprague-Dawley rats (Gagnaire et al., 1998), a 60-day gavage study of testicular toxicity in male
CD-I mice (Tandon et al., 1988), a 14-week gavage toxicity bioassay in B6C3Fi mice (NTP,
2001), a developmental toxicity study in Sprague-Dawley rats exposed by gavage during GDs 6-
15 (Murray et al., 1978), and a neurobehavioral study of Sprague-Dawley rats exposed to AN in
drinking water for up to 12 weeks (Rongzhu et al., 2007).
Table 4-56. Noncancer effects in animals repeatedly exposed to AN by the
oral route
Reference
Exposure conditions,
NOAEL
LOAEL
Species
mg/kg-d
(mg/kg-d)
Effect
Johannsen and
Levinskas (2002b);
Biodynamics (1980c)
F344 rat
0,0.1,0.3,0.8,2.5, 8.4 (M)
0,0.1,0.4, 1.3,3.7, 10.9(F)
DW, 2 yrs; sacrifices at 6, 12,
and 18 mos and termination
(99 wks or 699-706 d [F];
107 wks or 770-777 d [M])
0.1 (M)
0.1(F)
2.5 (M)
1.3(F)
0.3a (M)
0.4a (F)
8.4b(M)
3.7b (F)
Forestomach squamous cell
hyperplasia and hyperkeratosis,
increased incidence
Decreased survival after 18 mos
2.5 (M)
3.7 (F)
8.4a(M)
10.9a (F)
Decreased BW (>10% compared
with control)
0.4 (F)
1.3b(F)
Increase in serum alkaline
phosphatase (F only)
2.5 (M)
8.4b (M)
Increase in epidermal inclusion
cysts (M only)
Johannsen and
Levinskas (2002a);
Biodynamics (1980a)
Sprague-Dawley rat
0, 0.09, 8.0 (M)
0,0.15, 10.7(F)
DW, 22 mos (M); 19 mos
(F); sacrifices at 6, 12, and
18 mos and termination
ND(M)
0.15 (F)
8.0 (M)
0.15 (F)
0.09b(M)
10.7a (F)
ND(M)
10.7a (F)
Forestomach squamous cell
hyperplasia, increased severity
Renal transitional cell hyperplasia
(F only)
0.09 (M)
0.15 (F)
8.0b (M)
10.7a (F)
Decreased survival at 10 mos (M +
F); decreased BW (10/8% [M/F])
255
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-56. Noncancer effects in animals repeatedly exposed to AN by the
oral route
Reference
Exposure conditions,
NOAEL
LOAEL
Species
mg/kg-d
(mg/kg-d)
Effect
Johannsen and
Levinskas (2002a);
Biodynamics (1980b)
Sprague-Dawley rat
0,0.1, 10 (M + F)
Gavage, 7 d/wk for 20 mos;
sacrifices at 6, 12, and
18 mos and termination
0.1 (M + F)
0.1
10a (M + F)
10a
Forestomach squamous cell
hyperplasia, increased severity
Renal transitional cell hyperplasia
at some but not all sacrifices
0.1
10a
Decreased survival after 14 mos
(M + F); decreased BW (6-13%
compared with controls [M only]).
Quast (2002); Quast
etal. (1980a)
Sprague-Dawley rat
0,3.4, 8.5, 21.3 (M)
0, 4.4, 10.8, 25.0 (F)
DW, 2 yrs
3.4 (M)
ND(F)
8.5b(M)
4.4b(F)
Forestomach squamous cell
hyperplasia and hyperkeratosis,
increased incidence
ND(F)
4.4b(F)
Gliosis in brain with or without
perivascular cuffing in females;
not significant in males
ND(M)
4.4 (F)
3.4b (M)
10.8b (F)
Chronic nephropathy
8.5 (M)
ND(F)
21.3b(M)
4.4b (F)
Decreased survival after 300 (F) or
480 (M) d
8.5 (M)
10.8(F)
21.3b (M)
25.0b (F)
Decreased BW (>10% compared
with control); clinical signs of
nervous system dysfunction
NTP (2001)
B6C3Fl mouse
0,2.5, 10, 20 (M + F)
Gavage, 5 d/wk for 2 yrs
Adjusted doses: 0,1.8,7.1,
14.3 mg/kg-d (M + F)
1.8 (M)
7.1(F)
1.8 (M)
14.3 (F)
7.1b (M)
14.3b (F)
7.1b (M)
ND(F)
Forestomach squamous cell
hyperplasia and hyperkeratosis,
increased incidence
Increased incidence of Harderian
gland hyperplasia
ND(F)
1.8b(F)
Increased incidence of ovarian
cysts or ovarian atrophy
7.1 (M + F)
14.3a (M +
F)
Decreased survival after 15 wks
NTP (2001)
B6C3Fi mouse
0, 5, 10, 20, 40 60 (M + F)
Gavage, 5 d/wk for 14 wks.
Adjusted doses: 0, 3.6, 7.1,
14.3,28.6, and 42.9 (M + F)
28.6 (M)
14.3 (F)
ND(M)
28.6a (F)
Forestomach inflammation,
ulceration, and epithelial
hyperplasia, increased incidence in
F only
No histopathology data for
60 mg/kg groups
14.3 (M +
F)
28.6b (M +
F)
Increased mortality
Tandon et al. (1988)
CD-I mouse
0, 1, 10 (M only)
Gavage, for 60 d
1
10b
45% decreased sperm count; 40%
of seminiferous tubules examined
degenerated
256
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-56. Noncancer effects in animals repeatedly exposed to AN by the
oral route
Reference
Exposure conditions,
NOAEL
LOAEL
Species
mg/kg-d
(mg/kg-d)
Effect
Friedman and Beliles
(2002); Litton
Bionetics (1992)
Sprague-Dawley rat
0, 11, 37 (M)
0, 20, 40 (F)
DW, three-generation
reproduction study
ND
1 lb (M)
20b (F)
Decreased viability for Fib pups;
no changes in fertility index or
pregnancy outcome for F1 a, Flb,
F2a, F2b, F3a, orF3b generations
ND
1 lb (M)
20b (F)
LOAEL for decreased lactation
index in Flb parents for F2a pups;
small deficits in postnatal pup
weights (10-40%, variable across
generations) and postnatal survival
(about 10%, variable across
generations) in higher dosed
groups
Murray et al. (1978)
Sprague-Dawley rat
0, 10, 25, 65 (F only)
Gavage GDs 6-15
10
25
Maternal effects: forestomach
hyperplasia, 11% increased liver
weight, 22% decrease in BW gain
10
25
Fetal effects: missing vertebrae in
6/17 litters, 7% decreased BW
Gagnaire et al. (1998)
Sprague-Dawley rat
0, 12.5,25, 50 (M only)
Gavage, 5 d/wk for 12 wks
25
50a
SCV decreased beginning wk 6
(-7.5%),-17.8% by wk 12;
-10.6% after 8 wks recovery;
decreased BW after 4 wks (-17%
at wk 12); hind limb weakness in
5/11 surviving high-dose rats after
wk 9
Rongzhu et al. (2007)
Sprague-Dawley rat
0, 4, 13.5 (M only)
DW for 0, 4, 8, 12 wks
ND
4b
Neurobehavioral alterations, as
indicated by decreased motor
coordination, increased training
duration, head twitching,
trembling, circling, backwards
pedaling, and decreased home-
cage activities
Szabo et al., (1984)
Sprague-Dawley rats
0, 0.2, 4, 20, and 100 mg/kg-
day (F only)
DW for up to 60 days
4(F)
20 (F)
Enlarged kidneys, increase in
regional hyperplasia of the gastric
mucosa
Szabo et al. (1984)
Sprague-Dawley rats
0, 0,2, 4, 20 or 100 mg/kg-
day (F only)
Daily gavage for up to 60
days
ND
0.2 (F)
Adrenocortical hyperplasia
Quast et al. (1975)
Beagle dog
0, 10, 16, 17 (M)
0, 8, 17, 18 (F)
DW for 6 months
10 (M)
8(F)
16 (M)
17(F)
Early mortality, histopathological
lesions of the esophagus and
tongue.
"significantly different from controls (p < 0.01).
Significantly different from controls (p < 0.05).
DW = drinking water; F = female; M = male; ND = not determined
257
DRAFT - DO NOT CITE OR QUOTE
-------
Forestomach lesions (epithelial hyperplasia and hyperkeratosis) were the most
consistently observed noncancer effect associated with chronic oral exposure of rats and mice to
AN and were associated with the lowest LOAELs in AN-exposed rats (Table 4-56). The lowest
LOAELs were 0.09 and 10.7 mg/kg-day for increased severity of forestomach lesions in male
and female Sprague-Dawley rats exposed by drinking water (Johannsen and Levinskas, 2002a;
Biodynamics, 1980a), 0.3 and 0.4 mg/kg-day for increased incidence of forestomach lesions in
male and female F344 rats exposed in drinking water (Johannsen and Levinskas, 2002b;
Biodynamics, 1980c), and 8.5 and 4.4 mg/kg-day for increased incidence of forestomach lesions
in male and female Sprague-Dawley rats exposed in drinking water (Quast, 2002; Quast et al.,
1980a). Increased incidences of animals with hyperplasia or hyperkeratosis of the forestomach
were also observed in B6C3Fi mice exposed by gavage (5 days/week) for 2 years to 10 mg/kg-
day (males) or 20 mg/kg-day (females) (NTP, 2001), female B6C3Fi mice exposed to 40 mg/kg-
day by gavage (5 days/week) for 14 weeks (NTP, 2001), and pregnant Sprague-Dawley rats
exposed to 25 mg/kg-day by gavage on GDs 6-15 (Murray et al., 1978). Other effects observed
in orally exposed animals generally were observed at higher doses (Table 4-56).
Kidney effects included increased incidence of renal transitional cell hyperplasia in
Sprague-Dawley rats exposed by gavage or in drinking water to about 10 mg/kg-day but not
0.1 mg/kg-day (Johannsen and Levinskas, 2002a; Biodynamics, 1980a) and increased incidence
of chronic nephropathy in Sprague-Dawley rats exposed to drinking water doses of 3.4 mg/kg-
day (males) or 10.8 mg/kg-day (females) (Quast 2002; Quast et al., 1980a).
Decreased survival was observed after 18 months in F344 rats exposed to drinking water
at 8.4 mg/kg-day (males) and 3.7 mg/kg-day (females) (Johnannsen and Levinskas, 2002b) and
in Sprague-Dawley rats exposed to drinking water at 21.2 mg/kg-day (males) and 4.4 mg/kg-day
(females) (Quast, 2002). In another drinking water study of Sprague-Dawley rats, decreased
survival was observed at 8.0 mg/kg-day (males) and 10.7 mg/kg-day (females).
Increased incidences of ovarian lesions (cysts or atrophy) were observed in female mice
exposed to gavage doses of 2.5, 10, or 20 mg/kg-day (NTP, 2001), but exposure-related ovarian
lesions were not observed in the chronic rat bioassays. In a reproductive toxicity study in
Sprague-Dawley rats exposed to 0, 100, or 500 ppm AN in drinking water (11 or 37 mg/kg-day
[males]; 20 or 40 mg/kg-day [females]), small deficits in postnatal pup weights and pup survival
without effects on fertility or pregnancy success were observed in three generations exposed to
either 100 or 500 ppm (Friedman and Beliles, 2002; Litton Bionetics, 1992) (see Table 4-56).
Other reproductive effects observed in animals were a reduction of epididymal sperm counts and
degeneration of seminiferous tubules in CD-I mice exposed to 10 mg/kg-day by gavage for
60 days (Tandon et al., 1988), but no exposure-related lesions in male reproductive organs were
found in B6C3Fi mice exposed by gavage to up to 40 mg/kg-day for 14 weeks or up to
20 mg/kg-day for 2 years (NTP, 2001). Likewise, no effects on sperm motility parameters were
found in male B6C3Fi mice exposed to up to 20 mg/kg-day for 14 weeks (NTP, 2001).
258
DRAFT - DO NOT CITE OR QUOTE
-------
The only oral developmental toxicity study (Murray et al., 1978) identified 10 and
25 mg/kg-day as a NOAEL and LOAEL for maternal effects (increased incidence of
forestomach hyperplasia, 10% increased absolute liver weight, and 22% decreased BW gain) and
developmental effects (7% decrease in fetal BW and missing vertebrae associated with short tails
in 6/17 litters, compared with 1/38 in the controls). The study involved daily gavage exposure
on GDs 6-15.
Neurological effects in animals associated with chronic oral exposure to AN were
observed in: (1) a report of gross clinical signs of neurological impairment in about 10% of
Sprague-Dawley rats (paralysis, head tilt, circling, and seizures) exposed to drinking water
concentrations of 500 ppm (providing doses of about 65 and 74 mg/kg-day for males and
females) but not in rats exposed to 100 ppm (13 and 15 mg/kg-day, males and females) (Bigner
et al., 1986); (2) decreases (8—15%) in SCV in tail nerves and hind-limb weakness in male
Sprague-Dawley rats exposed by gavage to 50 mg/kg-day (5 days/week) for 6-12 weeks but not
in rats exposed to 25 mg/kg-day (Gagnaire et al., 1998); and (3) neurobehavioral alterations in
male Sprague-Dawley rats exposed to 4 and 13.46 mg/kg-day AN for 4, 8, or 12 weeks (head
twitching, trembling, circling, backwards pedaling, decreased home-cage activities, decreased
motor coordination, and learning and memory) (Rongshu et al., 2007). These studies indicated
that neurological impairment in AN-exposed rats occurred at higher administered doses than
doses associated with hyperplasia and hyperkeratosis in forestomach squamous epithelial cells.
4.6.2. Inhalation
Table 4-57 lists the NOAELs and LOAELs from four cross-sectional health examinations
and surveys of subjective symptoms in Japanese and Chinese acrylic fiber workers (Chen et al.,
2000; Kaneko and Omae, 1992; Muto et al., 1992; Sakurai et al., 1978), a cross-sectional
examination of performance of acrylic fiber workers in a battery of neurobehavioral tests (Lu et
al., 2005a), a study on the effect of AN exposure on the sperm cells of AN production plant
workers (Xu et al., 2003), and three cross-sectional surveys of reproductive outcomes in Chinese
acrylic fiber or AN manufacturing workers (Dong et al., 2000b; Li, 2000; Dong and Pan, 1995).
More detailed descriptions of these studies are in Section 4.1.2.2. Each of these epidemiologic
studies measured AN concentrations in workplace air samples and compared results of the
examinations and surveys in exposed workers with unexposed workers of similar age. The
human LOAELs and NOAELs in Table 4-57 are mean or midpoint values of the range of
concentrations determined for various groups of exposed workers. Table 4-57 also lists
NOAELs and LOAELs identified in the animal toxicity studies involving repeated inhalation
exposure to AN.
259
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-57. Noncancer effects in human workers and animals repeatedly exposed
to AN by inhalation
Reference/
Study subjects
Exposure conditions"
NOAEL"
LOAEL
Effect
ppm
Muto et al. (1992)
Acrylic fiber workers
Male workers with average
17 yrs of occupational
exposure compared with
unexposed workers
0.19
1.13
Statistically significantly increased
prevalences of subjective symptoms (e.g.,
heaviness of stomach, poor memory,
irritability) but no increases in prevalences of
physical signs or abnormal values in
urinalytic, hematological, liver function, or
blood pressure variables compared with
controls.
Sakurai et al. (1978)
Acrylic fiber workers
Male workers with average
10-12 yrs of occupational
exposure compared with
unexposed workers
ND
4.2
Statistically insignificant increase in incidence
of palpable liver, reddening of the conjunctiva
and pharynx, and the occurrence of skin
rashes compared with unexposed controls.
No survey of subjective symptoms.
Kaneko and Omae
(1992)
Acrylic fiber workers
Male workers with average
5.6, 7.0, or 8.6 yrs of
occupational exposure in
three groups of factories
(mean AN concentrations
at 1.8,7.4, and 14.1 ppm)
were compared with
unexposed workers
ND
1.8
Statistically significantly increased
prevalences of subjective symptoms (e.g.,
headaches, tongue trouble, choking lump in
chest, fatigue) were found in workers from all
three groups of factories when compared with
controls.
Chen et al. (2000)
Acrylic fiber workers
Male and female workers
with average 13 yrs of
occupational exposure
compared with unexposed
workers
ND
0.48
Statistically significantly increased
prevalences of subjective symptoms (e.g.,
headache, dizziness, poor memory, choking
feeling in chest, loss of appetite) compared
with controls.
Lu et al. (2005a)
Acrylic fiber workers
Male and female workers
with >1 yr of occupational
exposure (average duration
not available) compared
with unexposed workers
ND
0.11
Small, but statistically significant, deficits in
tests of neurobehavior in monomer workers
and fiber workers (average air concentrations
of 0.11 ppm for monomer workers and
0.91 ppm for fiber workers) compared with
controls.
Xu et al. (2003)
AN production workers
Male workers with 2.8 yrs
of occupational exposure
in a AN production plant,
compared with controls of
approximate age range in
the general population.
ND
0.37
Statistically significant decrease in sperm
density and sperm number in exposed
workers, statistically significant increase in
DNA strand breakage and sex chromosome
aneuploidy in sperm cells of exposed workers
compared with controls
Dong and Pan (1995)
Acrylic fiber workers
Male and female workers
with average 3.2 and
10.2 yrs of occupational
exposure compared with
unexposed controls
ND
3.7
Statistically significant increased prevalences
of adverse reproductive outcomes (premature
delivery [10.7 vs. 3.5%, in wives of exposed
males], stillbirths [4.5 vs. 0%, in exposed
females], and sterility [5.0 vs. 1.8%, in
exposed males]) compared with controls.
Dong et al. (2000b)
Acrylic fiber workers
Male and female workers
with average 11 and
10.4 yrs of occupational
exposure compared with
unexposed controls
ND
3.6
Statistically significantly increased
prevalences of adverse reproductive outcomes
(increased stillbirths [2.7 vs. 0%], birth
defects [21.3 vs. 4.8%], and premature
deliveries [8.2 vs. 3.9%]) in female workers
compared with controls.
260
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-57. Noncancer effects in human workers and animals repeatedly exposed
to AN by inhalation
Reference/
Study subjects
Exposure conditions"
NOAEL"
LOAEL
Effect
ppm
Li (2000)
AN manufacturing
workers
Female workers with
average 14 yrs of
occupational exposure
compared with unexposed
controls
ND
7.5
Statistically significant increased prevalences
of adverse reproductive outcomes (sterility
[2.6 vs. 0.8%], pregnancy complications
[20.8 vs. 7.1%], premature deliveries [11.6 vs.
4.7%], and congenital defects [25.4 vs. 4.2%])
in female exposed workers compared with
controls.
Quast et al. (1980b)
Sprague-Dawley rats
(M + F)
0, 20, 80 ppm, 6 hrs/d,
5 d/wk for 2 yrs
ND
20
Statistically significant increase in incidence
of lesions in nasal epithelia at 20 ppm in
males and females; focal necrosis in liver of
females. At 80 ppm, other lesions occurred at
increased incidences—gliosis and
perivascular cuffing in brain in both sexes,
hepatic necrosis in females, focal nephrosis
and thyroid cysts in males, and hyperplasia in
nonglandular stomach.
Gagnaire et al. (1998)
Sprague-Dawley rats
(M only)
0, 25, 50, 100 ppm,
6 hrs/d, 5 d/wk for 24 wks
ND
25
Statistically significant deficits (5% decreased
compared with controls) in sensory
conduction velocity of the tail nerve.
Murray et al. (1978)
Pregnant Sprague-
Dawley rats
0, 40, or 80 ppm, 6 hrs/d
on GDs 6-15
ND
40
40
80
Maternal weight gain decreased by >20%
compared with controls.
No increased incidence of litters with a single
malformation; 6/35 litters with short tail, short
trunk, missing vertebrae, or missing ribs vs.
1/33 in controls.
Saillenfait et al. (1993)
Pregnant Sprague-
Dawley rats
0, 12, 25, 50, 100 ppm,
6 hrs/d on GDs 6-20
12
12
25
25
Statistically significant decreased maternal
weight gain compared with controls.
Statistically significant decreased fetal BW,
>5% decreased compared with controls. No
exposure-related increased incidences of
litters with fetal anomalies.
aND = cannot be determined
In acute AN poisoning cases involving humans (Chen et al., 1999), first aid treatment of
victims generally included antidotes for cyanide poisoning and pure oxygen to overcome
respiratory distress caused by damage to the lung. Other clinical symptoms following such
poisoning included blood chemistry changes indicative of slight liver damage. In an analysis of
144 case reports of acute AN poisoning, Chen et al. (1999) estimated that exposure levels were
in the 18-258 ppm range (40-560 mg/m3) for 60 cases and >460 ppm (>1,000 mg/m3) for the
remaining 84 cases. Subjective symptoms reported for 92-100% of the cases included dizziness,
headache, chest tightness, feebleness, and hyperactive knee jerk. Sore throat, dyspnea, vomiting,
abdominal pain, fainting, and congestion of the pharynx were reported in 60-87% of cases.
Other less frequently reported symptoms or effects (5—32% of cases) included numbness of
261
DRAFT - DO NOT CITE OR QUOTE
-------
limbs, convulsion, rapid heart rate, cough, hoarseness, rough breathing sound, coma, and
abnormal liver function (Chen et al., 1999).
In the cross-sectional studies of AN-exposed workers, subjective symptoms that were
reported with increased prevalences compared with unexposed workers included dizziness,
headache, chest tightness, and poor memory, indicating respiratory irritation and neurological
effects. Average workplace air concentrations associated with increased prevalences of these
subjective symptoms were 1.13 ppm (Muto et al., 1992), 1.8 ppm (Kaneko and Omae, 1992), and
0.48 ppm (Chen et al., 2000) (see Table 4-57). No statistically significant increases in the
prevalences of subjective symptoms were found in a group of acrylic fiber workers whose
average workplace air concentration was 0.19 ppm (Muto et al., 1992). The studies by Muto et
al. (1992) and Sakurai et al. (1978) included clinical physical examinations, but no statistically
significant increases in prevalences of physical signs (such as reddened conjunctiva or pharynx)
or abnormal values in clinical chemistry variables (including activities of liver enzymes) were
found in exposed workers (more details of results from these studies can be found in Section
4.1.2.2.2 and Table 4-22).
To provide support that neurological effects are the most sensitive identified class of
effects from repeated inhalation exposure to AN, statistically significant deficits in several
neurobehavioral tests were measured in exposed workers in a Chinese acrylic fiber
manufacturing plant with mean workplace air concentrations of 0.11 ppm (range 0.00-1.70 ppm)
and 0.91 ppm (range 0.00-8.34 ppm) in two different process areas. Deficits in exposed workers
compared with nonexposed workers were noted in a Profile of Mood States test (20-68% higher
for negative moods such as anger and confusion), a Simple Reaction Time test of attention and
response speed (10-16% deficits), and the backward sequence of the Digit Span test of auditory
memory (21-24% deficits).
Statistically significant increases in the prevalences of adverse reproductive outcomes
were associated with mean workplace air concentrations of 3.7 ppm (Dong and Pan, 1995),
3.6 ppm (Dong et al., 2000a), and 7.5 ppm (Li, 2000), indicating that reproductive effects from
occupational exposure may occur at higher exposure levels than those associated with mild
neurobehavioral effects. However, statistically significant decrease in sperm density and
number, statistically significant increase in DNA strand breakage and sex chromosome
aneuploidy in sperm cells were reported in workers exposed to 0.37 ppm AN with average 2.8
years of exposure (Xu et al., 2003). An epidemiological study conducted in the area surrounding
an AN-using plant in Hungary reported increased ORs for several congenital malformations
(e.g., undescended testis), but no exposure data were provided in this study (Czeizel et al., 2000,
1999).
Exposure levels associated with adverse effects in Sprague-Dawley rats were higher than
the workplace air concentrations associated with adverse effects in AN-exposed workers
(Table 4-57). The lowest exposure level in the 2-year inhalation bioassay, 20 ppm, produced
262
DRAFT - DO NOT CITE OR QUOTE
-------
increased incidence of lesions in nasal epithelia (hyperplasia of mucus-secreting cells in males
and flattening of the respiratory epithelium in females). At 80 ppm, further increases in
incidences of nasal lesions of a wider variety were observed as well as statistically significant
increases in the incidences of histopathological lesions at other sites, including gliosis and
perivascular cuffing in the brain of males and females, focal nephrosis and thyroid cysts in
males, hepatic necrosis in females, and hyperplasia and hyperkeratosis of the nonglandular
portion of the stomach in both sexes combined (Quast et al., 1980b). Other effects in Sprague-
Dawley rats associated with repeated inhalation exposure to AN included deficits in sensory
nerve conduction in the tail nerve of rats exposed to 25 ppm AN (Gagnaire et al., 1998) and
developmental effects. Statistically significantly increased incidence of litters with missing
vertebrae, missing ribs, or anteriorly displaced ovaries were observed in rats exposed to 80 ppm,
but not 40 ppm, 6 hours/day on GDs 6-15 (Murray et al., 1978). In another study involving
6 hours/day inhalation exposure of Sprague-Dawley rats on GDs 6-20, developmental effects
were restricted to a statistically significant decrease in fetal weight gain per litter, compared with
controls, associated with exposure to 25 ppm (5% decrease), 50 ppm (8% decrease), or 100 ppm
(15% decrease); no effects on fetal weight gain were observed at 12 ppm (Saillenfait et al.,
1993).
4.6.3. Mode-of-Action Information
The precise modes of action whereby AN induces noncancer effects are unknown.
However, a general understanding of the processes by which AN is metabolized within the body
allows some conclusions to be drawn about the range of processes that might be involved in
bringing about one or more of its toxic responses. Relevant metabolic processes are likely to
include partitioning between detoxification and oxidative activation sub-pathways, conversion of
AN to one or more toxic metabolites, depletion of GSH, the association of AN metabolism with
the onset of oxidative stress, and the ability of CEO, the reactive metabolite of AN, to covalently
bind to macromolecules, such as proteins. In addition, other processes related to the parent
compound AN, such as its cholinomimetic effects, may also be involved.
4.6.3.1. GI Effects
The relationship between AN metabolism and GI hemorrhage in rats was suggested by
Ghanayem and Ahmed (1983). GI bleeding was observed 3 hours after a single dose of
50 mg/kg AN was administered to Sprague-Dawley rats orally or subcutaneously, with no
significant difference in the amount of GI blood loss resulting from either route of
administration. Thus, AN-induced GI bleeding was not a result of direct irritation of AN on the
GI tract.
Pretreatment of rats with CYP450 enzyme inducer Aroclor 1254 increased blood loss by
240% (Ghanayem and Ahmed, 1983). In contrast, pretreatment of rats with CYP450 inhibitors
263
DRAFT - DO NOT CITE OR QUOTE
-------
cobalt chloride or SKF 525A prior to AN administration produced significant decreases in blood
loss of 10 and 40%, respectively. Pretreatment of rats with DEM, a known depletor of GSH,
prior to AN administration produced no significant change in GI bleeding. In addition,
administration of a sublethal dose (6 mg/kg s.c.) of KCN did not induce GI bleeding when
compared with controls. Therefore, Ghanayem and Ahmed (1983) concluded that metabolic
activation of AN by CYP450 to a reactive metabolite other than cyanide (probably CEO) was a
prerequisite for AN to induce gastric hemorrhage.
Ahmed et al. (1996a) showed irreversibly bound AN-derived radioactivity in intestinal
mucosa following a single i.v. injection of 2-[14C]-AN to male F344 rats. A recent study by
Jacob and Ahmed (2003a) also demonstrated that AN and or its metabolites accumulated and
covalently interacted in GI mucosa of male F344 rats treated either by i.v. or orally with
2-[14C]-AN. These studies supported the hypothesis that AN-induced injury of the GI mucosa is
not due to direct irritation by AN but by metabolic activation and macromolecular interaction of
AN metabolite in these tissues.
Ghanayem et al. (1985) also studied the mechanism of AN-induced gastric mucosal
necrosis in the glandular stomach in male Sprague-Dawley rats. Subcutaneous administration of
40 or 50 mg/kg AN caused a significant decrease in hepatic and gastric GSH concentration
3 hours after treatment, and induced gastric necrosis. Pretreatment of rats with various metabolic
modulators (CYP450 monooxygenase and GSH) before administration showed that there was a
significant inverse relationship between gastric GSH concentration and AN-induced gastric
erosions. Pretreatment of rats with sulfhydryl-containing compounds (cysteine or cysteamine)
protected against AN-induced gastric necrosis and blocked the depletion of gastric GSH.
In addition, AN-induced gastric erosions could be prevented by pretreatment with
atropine, a muscarinic receptor blocker, suggesting the involvement of muscarinic receptors in
the AN-induced gastric mucosal necrosis (Ghanayem et al., 1985). Activation of acetylcholine
muscarinic receptors is known to increase gastric acid secretion and cause gastric erosions.
Because muscarinic receptors are known to contain sulfhydryl groups in their active site (Ikeda
et al., 1980; Aronstam et al., 1978), Ghanayem et al. (1985) hypothesized that AN inactivated
critical sulfhydryl groups and caused gastric erosions by locally modulating muscarinic
acetylcholine receptors in the stomach.
4.6.3.2. Neurological Effects
Increased prevalence of subjective symptoms of neurological effects was associated with
average workplace air concentrations of 1.13 ppm (Muto et al., 1992), 1.8 ppm (Kaneko and
Omae, 1992), and 0.48 ppm (Chen et al., 2000), and small deficits in performance in a battery of
neurobehavioral tests were observed in workers from factories with average workplace air
concentrations of 0.11 and 0.91 ppm (Lu et al., 2005a). In a case of acute severe accidental AN
poisoning (AN concentration at 62 mg/m ) of a worker (Fei and Xu, 2006), impairment of the
264
DRAFT - DO NOT CITE OR QUOTE
-------
CNS, including diffused damage to the cerebral cortex layer (pyramidal system) and damage to
the subcortical layer (extrapyramidal system), was observed. In addition to poisoning symptoms
(dizziness, headache, difficulty breathing, confusion, convulsion, etc.), cerebral focal damage
was also detected in the patient, suggesting Parkinson's syndrome (static tremor, muscle rigidity,
increased muscle tension and lead-pipe rigidity, slow motor activity). The clinical symptoms
were related to those induced by adverse effects on the cholinergic system and dopaminergic
system.
Neurotoxicological effects of AN in animals included cholinomimetic effects on the
peripheral and central muscarinic systems in Sprague-Dawley rats after administration of
nonlethal oral doses of 20, 40, or 80 mg/kg (Ghanayem et al., 1991) and the development of
brain lesions in Sprague-Dawley rats (gliosis and perivascular cuffing) following chronic
inhalation exposure to 80 ppm (Quast et al., 1980b) or chronic drinking water exposure to
4.4 mg/kg-day (females only).
The neurobehavioral effect of AN was suggested to be related to changes in brain
monoamine neurotransmitter levels (Lu et al., 2005b). In a 12-week drinking water study of
male Sprague-Dawley rats (Lu et al., 2005b), dopamine levels were decreased by 76 and 46% in
rats exposed to 50 ppm AN, in the striatum and cerebellum, respectively. Serotonin levels were
decreased by 38 and 41% in the striatum and cerebellum, respectively, for rats exposed to
50 ppm AN. In the case of acute severe AN poisoning, the development of Parkinson's
syndrome and other CNS impairment in the exposed worker would suggest involvement of the
cholinergic system and dopaminergic system (Fei and Xu, 2006).
The cholinomimetic effects were thought to be due to an effect of AN on muscarinic
receptors, since atropine sulfate (which blocks both central and peripheral muscarinic receptors)
protected animals against these effects (Ghanayem et al., 1991). In an earlier study on
AN-induced gastric mucosal necrosis (Ghanayem et al., 1985), it was reported that pretreatment
with atropine and sulfhydryl-containing chemicals protected against such lesions. Since
muscarinic receptors contain sulfhydryls in their active sites (Aronstam et al., 1978) and
sulfhydryl-depleting chemicals are known to potentiate chemically induced activation of
muscarinic receptors (Hedlund and Bartfai, 1979), Ghanayem et al. (1991) speculated that
depletion and/or inactivation of endogenous sulfhydryls by AN and/or its metabolite may cause
configurational changes of muscarinic receptor binding affinity that, in turn, lead to the
development of acetylcholine-like (cholinomimetic) toxic effects. It is unlikely that these
cholinomimetic effects are due to inhibition of acetylcholinesterase activity. Rajendran and
Muthu (1981) reported that AN did not inhibit the activity of acetylcholinesterase. In addition,
Satayavivad et al. (1998) reported that AN had no effect on decreased motor activity induced by
physostigmine (an inhibitor of acetylcholinesterase). Thus, Satayavivad et al. (1998) proposed
that the cholinomimetic effect of AN might be mediated by the release of acetylcholine from
nerve endings.
265
DRAFT - DO NOT CITE OR QUOTE
-------
In addition, Ghanayem et al. (1991) proposed that lipid peroxidation may at least partly
be involved in AN-induced cholinergic overstimulation. In noting that AN enhanced lipid
peroxidation and inhibited Na+,K+-ATPase in RBCs in vitro, Farooqui et al. (1990) speculated
that disruption of the lipid microenvironment in membranes by either or both of these processes
might impact the muscarinic receptor function and induce cholinergic overstimulation.
For acute CNS effects, the signs were similar to those produced by cyanide. Ghanayem
et al. (1991) proposed that the free cyanide liberated from AN during its metabolism may
contribute to these effects. It is well established that cyanide causes CNS dysfunction by
inhibition of cellular respiration via inactivation of tissue cytochrome c oxidase, which is the
terminal electron acceptor in cellular energy production (Klaassen, 2001). Intraperitoneal
injection of 30 mg/kg AN in Chinese hamsters decreased cerebral succinate dehydrogenase and
cytochrome oxidase activities (Zitting et al., 1981). These authors suggested that the observed
biochemical effects were likely due to the formation of cyanide from AN.
The neurotoxicity of AN may also result from the covalent binding of AN or its
metabolites to enzymes. There is abundant evidence that important proteins bearing cysteine
residues (e.g., enzymes such as GSTM1) can bind AN, thereby possibly impairing their functions
and creating metabolic imbalances leading to toxicity. For example, AN was shown to
covalently bind to the important glycolytic enzyme GAPDH in vitro (Campian et al., 2002). AN
specifically targeted and bound to cysteine 149 in the active center of this enzyme, causing
irreversible inhibition of its activity. This suggested that AN might impair glycolytic ATP
production in vivo. Campian et al. (2002) speculated that the combination of glycolytic
impairment with inhibition of mitochondrial ATP synthesis by cyanide released from AN could
result in metabolic arrest. However, Campian et al. (2008) demonstrated in male Sprague-
Dawley rats that acute lethality of AN was not due to brain metabolic arrest (see Section
4.5.1.1.3),
4.6.3.3. Reproductive/Developmental Effects
As discussed in Section 4.5.2.1, lower sperm density was noted in workers exposed to
"3
0.8 mg/m AN compared with controls of approximate age range from the general population
(75 x 106/mL vs. 140 x 106/mL) (Xu et al., 2003). DNA strand breakage was also detected in
AN-exposed workers using single-cell gel electrophoresis, with the rate of comet sperm higher in
the exposed workers than in the control (28.7 vs. 15%). The frequency of sex chromosome
disomy was 0.69% in exposed groups and was higher than 0.35% in the control group. There
were also significant differences in the frequencies of XX-, YY-, and XY-bearing sperm between
exposed and control groups.
AN-induced effects on the male reproductive system are restricted to a report that
treatment of CD-I mice for 60 days with gavage doses of 10 mg/kg-day produced degeneration
of the seminiferous tubules and altered testicular activities of several enzymes (SDH, acid
266
DRAFT - DO NOT CITE OR QUOTE
-------
phosphatase, LDH, and P-glucuronidase) (Tandon et al., 1988). Exposure-related increases in
the incidence of lesions in male reproductive organs were not observed in 14-week or 2-year
gavage bioassays with B6C3Fi mice (NTP, 2001) or in the 2-year bioassays with Sprague-
Dawley or F344 rats (see Table 4-56).
The potential male reproductive effect of AN may involve the distribution and
metabolism of AN to CEO in the testis and interaction of CEO with tissue protein and DNA.
Radiolabeled AN distributed to the rat testis after oral and i.v. administration (Ahmed et al.,
1996a; Young et al., 1977). The testis has the capability to bioactivate AN (Abdel-Aziz et al.,
1997). Thus, CEO can either be formed in the liver and transported to the testis or be formed in
the testis in situ. AN has been reported to interact with testicular DNA in rats treated with a
single 46.5 mg/kg oral dose (Ahmed et al., 1992b). Covalent binding of [2,3-14C]-AN-derived
radioactivity to testicular DNA was maximal at 0.5 hours following administration, while DNA
synthesis in testicular tissue was decreased (80% of control). In addition, testicular DNA repair
was increased 1.5-fold at 0.5 hour and more than threefold at 24 hours after treatment (Ahmed et
al., 1992b). Alkylating agents have the potential to produce infertility via destruction of dividing
primary spermatogonia (Heinrichs and Juchau, 1980). Thus, any effect of AN on male
reproductive tissue may be due to its interference with testicular DNA synthesis and repair
processes. The consequence may be reproductive abnormalities as well as impact on altered
heritability in offspring.
No mechanistic studies are available to elucidate the mode of action whereby AN
induced ovarian atrophy and cysts in female mice chronically exposed by gavage to doses as low
as 2.5 mg/kg-day (NTP, 2001).
The mild developmental effects associated with gestational exposure to 80 ppm AN by
inhalation or 65 mg/kg-day by gavage (Murray et al., 1978) may be associated with the release of
cyanide during maternal metabolism of AN. Concurrent administration of thiosulfate, a cyanide
antagonist, was shown to protect against the malformations induced by i.p. injection of 80 mg/kg
AN in hamsters (Willhite et al., 1981). However, Saillenfait et al. (1993) tested for relative
developmental toxicities of eight aliphatic mononitriles and proposed that factors other than
cyanide liberation from the nitrile may be involved, since teratogenicity of inhaled aliphatic
mononitriles in rats could not be predicted based on the presence of a vinyl moiety in their
molecular structure.
4.6.3.4. Hematological Effects
Farooqui and Ahmed (1983b) conducted work to elucidate the mechanism of the
hematological effects of AN. Their findings pointed to substantial AN-induced covalent binding
of CEO to RBCs, GSH depletion, increase in rate of RBC metabolism, and increase in the
formation of two metabolic intermediates, ATP and 2,3-diphosphoglycerate, that regulate the
oxygen dissociation curve. These authors suggested that chronic exposure to AN may lead to
267
DRAFT - DO NOT CITE OR QUOTE
-------
methemoglobinemia, damage to RBC membranes, and impaired delivery of oxygen to the
tissues.
Oxidative stress and lipid peroxidation were also suggested as additional factors aiding in
the destruction of RBCs, evidenced by a significant decrease of Na+,K+-ATPase activity in
isolated RBC membranes (Farooqui et al., 1990).
AN-Hb adducts have been measured and used as a marker of exposure in humans. AN
can bind to amino acid residues other than cysteine. Thus, MacNeela et al. (1992) identified the
N-terminal cyanoethyl-valine adduct of Hb (CEVal), the formation of which may have
implications for the efficiency of oxygen transport to the tissues.
4.6.3.5. Immunological Effects
Zabrodskii et al. (2000) suggested a potential mechanism of action for the AN-induced
DTH. They found that, in mice, AN reduced the number of esterase-positive splenocytes, the
number of antibody-producing cells, and the inflammatory response induced by injection of
SRBCs in the paws of animals. Treatment of the animals with an esterase activity-restoring drug
restored the paw-response completely but not the numbers of immune-competent cells.
However, a combination of the esterase-restoring drug and a cyanide-trapping drug restored
immune function completely. Therefore, the study authors concluded that the immunotoxic
effect of AN was due to combined inhibition of esterase and cytochrome c oxidase a3 activities.
4.6.3.6. Covalent Binding to Sulfhydryl Groups
The capacity of AN to bind to proteins may be an important determinant of its
toxicological effects. AN has been shown to have high affinity for cysteine residues on proteins
and polypeptides. There is abundant evidence that AN will bind to the cysteine-bearing
tripeptide, GSH, even without the contribution of enzymes such as GST. As discussed in
Chapter 3, the formation of 2-(cyanoethyl)glutathione and the appearance of
2-(cyanoethyl)cysteine and N-acetyl 2-(cyanoethyl)cysteine in the urine have been taken as an
indication of the ready interaction of AN and GSH. However, the presence of excess AN can
cause GSH to become depleted. This will tend to channel AN into an oxidative reaction with
CYP2E1 and result in formation of CEO and other products. Perturbing the balance between
detoxification of AN with GSH and oxidation by CYP2E1 (for example, by blockade of
CYP2E1) may facilitate the binding of AN to other cysteine-bearing proteins when GSH levels
are low.
AN also has been shown to bind to the cysteine 186 residue of the enzyme CAIII in rat
liver in vivo. Nerland et al. (2003) pointed out that this enzyme may play a role in protecting
cells from oxidative stress. AN binding to this component, with possible conformational
changes in tertiary structure and functionality, might abolish this protective ability and increase
cell vulnerability to oxidative stress. In a similar study, Nerland et al. (2001) demonstrated that
268
DRAFT - DO NOT CITE OR QUOTE
-------
AN can bind also with high selectivity to cysteine 86 of GSTM1. However, in this case, the
binding did not exert an effect on the catalytic activity of the enzyme.
4.7. EVALUATION OF CARCINOGENICITY—SYNTHESIS OF HUMAN, ANIMAL,
AND OTHER SUPPORTING EVIDENCE, CONCLUSIONS ABOUT HUMAN
CARCINOGENICITY, AND LIKELY MODE OF ACTION
4.7.1. Summary of Overall Weight of Evidence
Following EPA's Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), AN is
"likely to be carcinogenic to humans," based predominantly on consistent results showing that
lifetime inhalation or oral exposure caused a statistically significantly increased incidence of
tumors at multiple tissue sites in rats and mice. The most consistently observed tissue sites with
tumors were the brain, forestomach, Zymbal gland in the ear canal in rats, and forestomach and
ocular Harderian gland in mice. In addition, rats exposed during gestation and throughout
adulthood showed higher incidences of brain tumors, Zymbal gland tumors, and extrahepatic
angiosarcomas than did rats with exposure throughout adulthood only; hepatomas, which were
not observed in exposed adult rats, were reported in rats exposed during gestation and throughout
adulthood. Epidemiological studies of AN-exposed workers provide no strong evidence that
mortality from any type of cancer is causally related to occupational exposure to AN, but limited
evidence from the largest and best-designed epidemiologic study (Blair et al., 1998) indicated
that workers with the longest duration and highest exposures to AN had a small, but statistically
significant, increased risk for dying from lung cancer.
The available weight of evidence is adequate for a direct mutagenic mode of action
involving DNA modification by reactive metabolite CEO. Other modes of action, including
oxidative stress and inhibition of intercellular communication, may also contribute. Currently,
no mechanistic explanations are available for the fact that AN induces brain tumors in rats but
not in mice, although this may be related to higher CEO levels in rat brains and the more rapid
clearance of CEO in mice than in rats (see Section 3.2). Mouse liver GST has a greater capacity
for conjugating AN or CEO than rat liver GST (see Section 3.3.2). The modes of action
involved in tumors induced by AN at other sites in animals (e.g., forestomach squamous
epithelium in rats and mice and Zymbal gland in rats) have not received adequate investigation.
The mutagenic mode of action for AN-induced tumors is considered to be relevant to
humans.The metabolism of AN in rats and humans is similar and there is evidence for
mutagenicity of AN in exposed humans.
4.7.2. Synthesis of Human, Animal, and Other Supporting Evidence
Section 4.1.2.2 presents an historical perspective and evaluation of the epidemiologic
studies of possible relationships between occupational exposure to AN and elevated risks for
cancer. The early studies were of small cohorts with limited follow-up periods and limited
269
DRAFT - DO NOT CITE OR QUOTE
-------
assessment of actual exposures. Elevations for incidence and death from lung cancer and
incidence of prostate cancer were sufficiently consistent among the early studies to provoke the
conduct of several additional studies over a 20-year period. The later studies had increased
power due to increased number of workers, longer periods of follow-up, increased sophistication
and quantification of exposure assessments, or inclusion of information on smoking habits.
Small excess risks for lung and prostate cancer were identified in a few of these studies, but
consistently and statistically significantly elevated risks were not observed across studies (see
Tables 4-3 and 4-9). Other studies reported small excess risks for other types of cancer (e.g.,
bladder, colon, and brain cancers), but these findings were even less consistent across studies
than the findings for lung and prostate cancer (see Tables 4-8, 4-11, 4-12 and 4-13).
The most informative study in the later generation of studies was the large, well-
documented study by Blair et al. (1998). This study was strong because it examined a large
cohort and attempted to adjust for known problems with earlier studies by quantifying exposures,
estimating the effect of smoking, and using an internal control group of unexposed workers. The
AN-exposed cohort as a whole experienced fewer deaths from lung cancer than those expected
from the experience of the general U.S. population (probably reflecting a healthy worker effect),
but when the exposed workers were grouped into quintiles of cumulative exposure, the lung
cancer rate in the highest quintile of exposure (>8 ppm-years) for those who were followed for
20 or more years was about two times the rate in the unexposed group of workers (RR = 2.1;
95% CI = 1.2-3.8). Conclusions regarding the association between AN exposure and other
cancer types (stomach, brain, breast) are limited due to the small number of site-specific cancer
deaths (Blair et al., 1998).
As concluded in Section 4.1.2.2, there is no strong evidence from the body of
epidemiologic studies that mortality from any type of cancer is causally related to exposure to
AN at the levels that have been measured in workplaces utilizing this chemical. However, based
on these epidemiology studies, there is suggestive evidence of a possible association between
occupational exposure to AN and increased risk of lung cancer.
In rat and mouse bioassays, AN has been demonstrated to be a multiple-site carcinogen.
Chronic oral exposure to AN induced tumors in the brain or spinal cord, Zymbal gland in the ear
canal, the forestomach, and, to a lesser degree and less consistently, the female mammary gland,
tongue, and intestine in several oral bioassays with F344 rats (Johanssen and Levinskas, 2002b;
Biodynamics, 1980c) and Sprague-Dawley rats (Johannsen and Levinskas, 2002a; Quast, 2002;
Biodynamics, 1980a, c; Quast et al., 1980a). In addition, lifetime inhalation cancer bioassays
with Sprague-Dawley rats found exposure-related increased incidences of brain tumors, Zymbal
gland tumors, intestinal tumors, malignant mammary gland tumors, and tongue tumors (Dow
Chemical Co., 1992a; Maltoni et al., 1988, 1977; Quast et al., 1980b). Strong evidence exists for
dose-response relationships for the carcinogenic responses in the CNS, Zymbal gland, and the
forestomach of rats, especially in the lifetime drinking water studies with multiple exposure
270
DRAFT - DO NOT CITE OR QUOTE
-------
levels. The lowest drinking water concentrations associated with significant increased incidence
of forestomach and brain tumors were 3 and 30 ppm, respectively, corresponding to daily doses
of about 0.3 and 2.5 mg/kg-day in F344 rats (Johanssen and Levinskas, 2002b; Biodynamics,
1980c).
With chronic inhalation exposure of Sprague-Dawley rats, increased incidences of brain
tumors occurred at air concentrations of 20 and 80 ppm, whereas the incidences of Zymbal gland
tumors, mammary gland adenocarcinomas and intestinal tumors were increased only at the
80 ppm level (Quast et al., 1980b). In mice, a single gavage lifetime bioassay identified the
forestomach and the Harderian gland as sites of tumor development, but elevated incidences of
brain, Zymbal gland, or mammary gland tumors were not found (NTP, 2001). A significant
increase in lung tumors was found in the female mid-dose group of exposed mice but not in the
female high-dose group or in any of the male exposed groups. NTP (2001) concluded that the
evidence for carcinogenicity in the mouse lung was equivocal; this conclusion is consistent with
the small magnitude of the increase, lack of a monotonic dose-response relationship, and lack of
a demonstrated carcinogenic response in exposed male mice. No consistent evidence was found
for carcinogenicity in the rat lung or the prostate or bladder tumors in rats or mice.
Results from two rat studies provide evidence that chronic exposure starting during early-
life periods may result in increased susceptibility to the carcinogenicity of AN, compared with
lifetime exposure during adulthood only (see Sections 4.2.2.2.2 and 4.2.1.2.8 for more details).
The possible human relevance of the rodent carcinogenic responses to AN is not fully
understood. The brain is the only organ for which there is a direct human counterpart among
rodent organs showing strong carcinogenic responses to AN, but the forestomach squamous
epithelium and Zymbal gland have analogous tissues in humans. Although humans do not have
a forestomach, the human esophagus is lined with squamous epithelial cells morphologically
similar to those in which tumors develop in rats, with the exception that the rat cells are
keratinized and the human cells are not (Cohen, 2004; Wester and Kroes, 1988). Likewise,
although humans do not have a Zymbal gland, a sebaceous gland in the ear canal, they do have
sebaceous glands. In contrast, humans and other primates do not have tissue analogous to the
Harderian gland, an ocular gland in rodents that secretes lipids and porphyrins (Cohen, 2004;
Sheldon, 1994; Albert et al., 1986). However, given that mutagenic modes of carcinogenic
action are plausible for AN or its metabolites (see Section 4.6.3), formation of tumors in these
organs may be indicative of a more generic carcinogenic hazard (Cohen, 2004). Chemicals with
a mutagenic mode of action are frequently observed to cause cancers in many sites in one
species, as well as to have different sites of tumor formation in different species. Therefore, it is
reasonable to consider the tumor responses in these rodent organs as indicators of AN-induced
carcinogenicity in humans. In addition, the National Academy of Sciences (NAS) (2008) in its
Science and Decisions: Advancing Risk Assessment, stated on page 134 that".. .the target organ
in a rodent species, such as the forestomach or Zymbal gland, may not have an exact human
271
DRAFT - DO NOT CITE OR QUOTE
-------
counterpart. However, the presence of carcinogenic action in tissues for which there is no
correspondence in humans or that may be regulated differently in humans does not mean that the
toxicity or tumor finding in animals is irrelevant. That the rodent tissue is sensitive to the
toxicant signifies that the toxicant MO As operate in a mammalian system that has characteristics
in common with similar or even not obviously related tissues in humans or human
subpopulations.. .it is typically impossible to rule out the relevance of an effect seen in a
particular rodent tissue unless there is detailed mechanistic information on why humans would
not be affected.. .In general, tissues that are responsive to a toxicant should be considered
relevant to human risk assessment unless mechanistic information demonstrates that the
processes occurring in the tissues could not occur in humans."
4.7.3. Mode-of-Action Information
The U.S. EPA (2005a) Guidelines for Carcinogen Risk Assessment defines mode of
action as a sequence of key events and processes, starting with the interaction of an agent with a
cell, proceeding through operational and anatomical changes, and resulting in cancer formation.
Mode of action is distinct from "mechanism of action," a term that implies greater understanding
and description of events, including those at the molecular level. Toxicokinetic processes
leading to the formation or distribution of the active agent (i.e., parent material or metabolite) to
the target tissue are not part of the mode of action. Examples of possible modes of carcinogenic
action include mutagenic, mitogenic, anti-apoptotic (inhibition of programmed cell death),
cytotoxic with reparative cell proliferation, and immunologic suppression.
AN has been demonstrated to be a multiple-site carcinogen in rats and mice, inducing
tumors in the forestomach, brain, and Zymbal gland in two strains of rats following chronic oral
exposure; in the forestomach and Harderian gland in a mouse strain following chronic oral
exposure; and in the forestomach, brain, Zymbal gland, intestinal tract, and tongue in one rat
strain following chronic inhalation exposure. Carcinogenic responses have also been reported
less consistently across studies in the female mammary gland and small intestine. Several
hypothesized modes of action by which AN causes cancer in the brain of rats have been
investigated to varying degrees and are discussed within the context of the modified Hill criteria
of causality as recommended in the most recent Agency guidelines (U.S. EPA, 2005a). These
discussions are followed by discussion of possible modes of action involved in the development
of forestomach tumors and hepatomas in rats and lung cancer in humans. Modes of action by
which AN may induce tumors in the Zymbal gland, Harderian gland, mammary gland, and
tongue have not been investigated. The generic mutagenic modes of action probably apply to
these organs also, unless there is information to indicate otherwise.
272
DRAFT - DO NOT CITE OR QUOTE
-------
4.7.3.1. Hypothesized Mode of Action: Direct Mutagenic Mode-of-Action
Key Events
This mode of action hypothesizes that AN is first activated by CYP2E1 (mostly in the
liver but also at other sites, such as intestinal mucosa or squamous epithelium of the
forestomach) to its reactive metabolite, CEO. CEO is then distributed to target organs (e.g., rat
brain) where it reacts with DNA, forming DNA adducts. DNA adduct formation results in
genetic damage, especially the formation of point mutations. Other types of DNA damage by
CEO are also observed in vivo, including DNA strand breaks, SCEs, and MN formation.
Mutagenicity is a well-established cause of carcinogenicity.
AN
\L CYP 2E1
CEO
i
Interaction with DNA
i
Gene mutation for tumor initiation
Following these key events, tumor growth may be promoted by any one or combination
of a number of cell-signaling pathways, leading to enhanced cell proliferation or inhibition of
programmed cell death.
4.7.3.2. Experimental Support for the Hypothesized Mode of Action
4.7.3.2.1. Strength, consistency, specificity of association. The evidence that supports a
mutagenic mode of action is strong and consistent (as summarized in Table 4-53). In addition to
positive findings in blood lymphocytes, buccal mucosal cells, and sperm in five epidemiologic
studies, DNA alkylation by AN was found in numerous tissues in rats or mice (brain, liver,
testes, forestomach, colon, kidney, bladder, and lung) treated with a single dose of AN. AN or
its reactive metabolite CEO yielded positive results in in vitro mutation assays using bacteria,
fungi, and insects, as well as animal and human cell cultures. The mutagenicity/genotoxicity is
specific and occurs in the absence of cytotoxicity or other overt toxicity. Although the temporal
relationship of adduct formation and mutagenicity with carcinogenicity has not been adequately
explored, these effects are seen in short-term assays (before tumor formation). Dose-response
concordance is observed between mutagenic doses in vivo and tumorigenic doses in rats and
mice. A mutagenic mode of action also comports with notions of biological plausibility and
coherence because AN is metabolized to an epoxide intermediate. Such agents are generally
capable of forming DNA adducts, which in turn have the potential to cause genetic damage,
273
DRAFT - DO NOT CITE OR QUOTE
-------
including mutations, and mutagenicity, in its turn, is a well-established cause of carcinogenicity.
This chain of key events is consistent with current understanding of the biology of cancer.
The possible association between a direct mutagenic action of CEO, the epoxide
metabolite of AN, and brain tumors in rats is supported by the detection of CEO in rat brains
after oral administration of 10 mg/kg AN to rats (Kedderis et al., 1993b), reported binding of
CEO to brain DNA after oral administration of [2,3-14C]-AN to rats (Farooqui and Ahmed,
1983 a), results from studies of in vitro DNA reactivity, positive results from in vitro tests of
mutagenicity and cytogenetic effects, and positive results from in vivo studies of DNA damage
and repair assays and other genotoxic endpoints in rats and mice. Studies that demonstrate the
mutagenicity of AN in occupational exposed workers are also available.
In the following subsections, experimental evidence supporting the mutagenic mode of
action of AN using different test systems is summarized.
In vitro DNA binding
AN itself reacts very slowly with DNA in vitro at very high, nonphysiological
concentrations (>1 M) (Solomon et al., 1984), but CEO forms different adducts in vitro with
DNA (Solomon et al., 1993; Yates et al., 1993) or nucleotides (Yates et al., 1994) more rapidly
(see Section 4.5.1.2.1). When calf thymus DNA was incubated with CEO for 3 hours (Solomon
1 "3
et al., 1993), the main adducts formed included N -(oxoethyl)guanine, N -(2-hydroxy-2-carboxy-
ethyl)deoxyuridine, and smaller amounts of adenine and thymine adducts. Yates et al. (1993)
"3
also identified the formation of N -(2-cyano-2-hydroxylethyl)deoxythymidine when CEO was
incubated with calf thymus DNA in vitro.
Mutations in bacteria
In short-term tests with bacteria, AN induced mutations in a majority of test systems,
often requiring the presence of exogenous metabolic systems (Table 4-53). AN induced
mutation in S. typhimurium strains TA 1530, 1535, 1538, 1937, and 1950 with strong response
when AN was tested in the vapor phase in the presence of S9. AN induced a weak response for
TA 98, 100, and 1978. Thus, AN induced gene reversion mainly by base substitution and
induced lower mutagenic activity with strains reversed by frameshift mutation. AN also
produced a dose-related increase in the number of revertant colonies compared with untreated
bacteria in E. coli WP2 (which is DNA repair proficient), WP2uvrA (which lacks excision
repair), and WP2 uvrApolA (which lacks both excision repair and DNA polymerase 1) without a
need for S9 fraction (Venitt et al., 1977).
Mutations in fungi and Drosophila
As shown in Table 4-53, AN induced mitotic gene conversion in S. cerevisiae JD1
(Brooks et al., 1985). AN also induced sex chromosome loss in the adult female Drosophila
274
DRAFT - DO NOT CITE OR QUOTE
-------
ZESTE system (Osgood et al., 1991), as well as somatic recombination and mutation in hatching
Drosophila with exposure in the larvae stage (Vogel, 1985; Wiirgler et al., 1985).
Mutations in mammalian cell culture
AN also induced mutations in mammalian cells in vitro. In the mouse lymphoma cell
assay, AN induced forward mutations at the Tk+j~ locus in most assays (Table 4-53). In human
lymphoblastoid Tk6 cells devoid of CYP450 activity, AN induced mutations at the Tk locus only
in the presence of an exogenous S9 metabolic system (Crespi et al., 1985), and CEO was
effective at 10-fold lower concentrations than AN (Recio and Skopek, 1988a, b).
Characterization of the Tk~ mutants in human lymphoblastoid Tk6 cultures by Recio and
Skopek (1988b) identified two classes of CEO-induced Tkr'~ mutant phenotypes that differed in
their growth rates. CEO-induced predominantly Tkn mutant with normal growth rate and Tks
with slower growth rate. Southern blot analysis of CEO-induced Tkn mutants indicated that the
majority of these mutants were below the detection limit of <2 kb. Thus, CEO-induced
alterations are relatively small DNA alterations. Recio and Skopek (1988a) suggested that CEO
induced Tkn mutants resulted from point mutations or small insertions/deletions that occurred
during the replication or repair of CEO-modified DNA. Kodama et al. (1989) conducted
cytogenetic analyses of eight CEO-induced Tks mutant clones reported in Recio and Skopek
(1988a, b). A visible abnormality on chromosome 17 was found in one of the CEO-induced Tks
mutants and was marked by duplication of the long arm of chromosome 17, with break points at
ql 1 and q21. The latter break point was close to the Tk locus, suggesting that the observed
aberration might be associated with Tkphenotype.
CEO also induced mutations at the hprt locus in Tk6 cells (Recio and Skopek, 1988a).
Characterization of the hprt mutations by cDNA sequencing analysis indicated that several hprt
mutations were formed. The majority of CEO-induced mutations were the specific loss of exons
from the coding region of hprt. Remaining mutants were single base substitutions (point
mutation) resulting from amino acid changes (A:T base pairs and G:C base pairs).
Cytogenetic effects in vitro
Additional evidence of mutagenicity included AN-induced cytogenetic changes, such as
SCEs or CAs in a majority of assays with CHO or CHL cells (Natarajan et al., 1985; VedBrat
and Williams, 1982; Ishidate et al., 1981) and human lymphocytes in vitro (Perocco et al., 1982)
but not in epithelial-like cells from rat liver (RL4 cell line) (see Table 4-53). AN also induced
DNA single-strand breaks in rat hepatocytes (Bradley, 1985) and CHO cells (Douglas et al.,
1985). Of note is that AN induced SCEs and DNA single-strand breaks in human bronchial
epithelial cells in culture without S9 (Chang et al., 1990), indicating that human bronchial
epithelial cells have the metabolic capabilities to activate AN to CEO.
275
DRAFT - DO NOT CITE OR QUOTE
-------
DNA repair in vitro
In DNA repair assays, neither AN nor CEO induced UDS in cultured rat hepatocytes
(Butterworth et al., 1992; Probst and Hill, 1985; Williams et al., 1985) (as measured by
"3
incorporation of [ H]-thymidine and autoradiographic techniques) or in HeLa cells with or
without the presence of S9 (using liquid scintillation spectrometry) (Martin and Campbell, 1985).
However, IPCS (1985) concluded that the rat hepatocyte autoradiographic UDS assay was too
insensitive for determination of genotoxicity of the eight tested carcinogens, including AN. On
the other hand, CEO, but not AN, induced UDS in human mammary epithelial cells in vitro
(Butterworth et al., 1992).
DNA repair in vivo
Unscheduled DNA repair activity was detected by following the time course of
"3
[ H]-thymidine incorporation into DNA isolated from lung (Ahmed et al., 1992a), testis (Ahmed
et al., 1992b), and gastric tissue (Ahmed et al., 1996b; Abdel-Rahman et al., 1994a) after
exposure of Sprague-Dawley rats to single oral doses of 46.5 mg/kg AN. In a study by Hogy and
Guengerich (1986), UDS was found in the liver of male F344 rats administered 50 mg/kg AN or
6 mg/kg CEO i.p. but was not found in the brain. The absence of detected UDS in the brain
could reflect the absence of DNA damage. Alternatively, this absence could reflect differences
in the repair rate of CEO-DNA adducts and could account for the observed target organ
specificity of AN in rat brain but not liver. Since the difference in DNA repair rates in brain and
liver is well known (Kleihues et al., 1977), the second explanation may be more sound. In
another study, UDS activity was not detected by autoradiographic techniques following
incubation of primary cultures of hepatocytes or spermatocytes from F344 rats given single oral
doses of 75 mg/kg or five daily doses of 60 mg/kg-day with [ H]-thymidine (Butterworth et al.,
1992). This difference in results from those by Hogy and Guengerich (1986) was probably due
to differences in methodology.
DNA binding in vivo
Several studies examining the amounts of radioactivity in DNA fractions of tissues
following acute in vivo exposure of rats to radiolabeled AN provide evidence of in vivo DNA
reactivity of AN or its metabolites. Maximal amounts of radioactivity covalently bound to
hydroxyapatite-purified DNA from the liver, stomach, and brain showed the following order
24 hours after male Sprague-Dawley rats were given single oral doses of 46.5 mg/kg
[2,3-14C]-AN: brain (-120 pmol AN equivalent/mg DNA) > stomach (-80 pmol/mg) > liver
(-25 pmol/mg) (Farooqui and Ahmed, 1983a). Other studies from the same group of
investigators found elevated covalent binding of radioactivity in gastric, testicular, and lung
DNA from similarly exposed rats (Abdel-Rahman et al., 1994b; Ahmed et al., 1992a, b). The
results from these studies provided some evidence of associations between covalent DNA
276
DRAFT - DO NOT CITE OR QUOTE
-------
binding following acute exposure and sites of tumor development following chronic exposure.
The methods to isolate DNA in these studies may not have been stringent enough to exclude
covalent binding of radiolabel to proteins (Whysner et al., 1998b; Geiger et al., 1983); however,
the degree to which these alternative processes may have contributed to the measured amounts of
radioactivity in the DNA fractions is unknown. When 0.6 mg/kg [2,3-14C]-CEO was
administered to one F344 rat i.p., covalent binding to both liver and brain protein was found, but
no covalent binding to both liver and brain nucleic acids could be detected at the level of
0.3 alkylations per 106 base (Hogy and Guengerich, 1986). However, the N7-(2-oxoethyl)
guanine adduct was detected in liver DNA of F344 rats treated with 6 mg/kg CEO or 50 mg/kg
AN i.p. in the same experiment. Thus, covalent binding of CEO to DNA had to occur. No DNA
binding was detected in the Hogy and Guengerich (1986) study, presumably due to the small
DNA sample from one rat, low CEO concentration (0.6 vs. 6 mg/kg CEO), stringent DNA
isolation procedure, and overcorrection of protein binding.
DNA adducts in vivo
Limited attempts to detect CEO-DNA adducts in brain tissues following acute in vivo
exposure protocols have been largely unsuccessful, but in vitro studies indicate that the
formation of other DNA adducts from AN and its metabolites are possible (Yates et al., 1994;
Solomon et al., 1993; Hogy and Guengerich, 1986; Solomon et al., 1984). N7-(2-oxoethyl)-
guanine, a CEO-DNA adduct formed in vitro following incubation of calf thymus DNA with
CEO, was detected in DNA isolated from the livers of male F344 rats, following single i.p.
administration of doses of 50 mg/kg AN or 6 mg/kg CEO, but was found only equivocally at
detection limit in DNA isolated from the brains of exposed rats (Hogy and Guengerich, 1986).
In another study, 7-(2-cyanoethyl)guanine and 06-(2-cyanoethyl)deoxyguanosine adducts were
not detected in DNA isolated from the liver or brain of male F344 rats given s.c. or i.v. injections
of 50 or 100 mg/kg AN (Prokopczyk et al., 1988). The DNA samples were analyzed for the
presence of the two adducts using HPLC with fluorescence detection. The detection limits were
20 [j,mol/mol guanine for 7-(2-cyanoethyl)guanine and 15 [j,mol/mol guanine for 06-(2-cyano-
ethyl)deoxyguanosine. These detection limits may be not sensitive enough. More importantly,
these two DNA adducts were not formed from incubation of CEO with calf thymus DNA in
vitro. Hence, the limitations of these studies were that not all the major adducts formed in in
vitro incubation of CEO with DNA were measured.
As discussed previously, the main adducts found after 3 hours of incubation of CEO with
1 "3
calf thymus DNA were N -(oxoethyl)guanine, N -(2-hydroxy-2-carboxyethyl)deoxyuridine
"3
(Solomon et al., 1993), andN -(2-cyano-2-hydroxylethyl)deoxythymidine (Yates et al., 1993).
Yet only N -(oxoethyl)guanine has been measured in available studies. Thus, it is highly likely
that the actual adducts formed from interaction of CEO with brain DNA have not yet been
looked for. It is also likely that the analytical methods used to measure DNA adducts in
277
DRAFT - DO NOT CITE OR QUOTE
-------
available studies were not sensitive enough for their detection. In addition, DNA adducts were
measured only in single-dose studies not repeated-dose studies.
DNA damage in rats and mice
Evidence of DNA damage after AN exposure in rats and mice is available. Comet assays
showed DNA damage in various tissues in both rats and mice exposed to AN (Sekihashi et al.,
2002). Single i.p. injections of 20 mg/kg AN induced DNA damage in forestomach, bladder, and
brain but not in colon, liver, kidney, or bone marrow of ddY mice, whereas single doses of
30 mg/kg induced DNA damage in forestomach, colon, kidney, bladder, and lung but not in
brain or bone marrow of Wistar rats (Sekihashi et al., 2002).
Increased frequencies of MN in bone marrow were found in Sprague-Dawley rats,
following i.v. injection of 98 or 124 mg/kg AN (Wakata et al., 1998) but were not found in
Sprague-Dawley rats following administration of oral doses up to 40 mg/kg (Morita et al., 1997)
or in male CD-I mice following administration of oral, i.p., or i.v. doses up to 45 mg/kg (Morita
et al., 1997). Increases in CAs in bone marrow cells were not found in Swiss albino mice
exposed to oral doses up to 20 mg/kg-day for 4, 15, or 30 days (Rabello-Gay and Ahmed, 1980);
in Sprague-Dawley rats given 16 daily doses of 40 mg/kg-day (Rabello-Gay and Ahmed, 1980);
in NMRI mice given single i.p. doses of 30 mg/kg (Leonard et al., 1981); or in ICR mice
"3
exposed in inhalation chambers to 20 or 100 mg/m for 5 days (Zhurkov et al., 1983). In
contrast, Fahmy (1999) reported that increased CAs occurred in spermatocytes of Swiss mice
following single oral doses of 15.5 or 31 mg/kg AN or five daily doses of 7.75 mg/kg and in
spleen cells and bone marrow cells after a single oral dose of 7.75 mg/kg.
Dominant lethal mutations in rats and mice
Dominant lethal mutations were not increased by treating male NMRI mice with single
i.p. doses of 30 mg/kg AN (Leonard et al., 1981) or male F344 rats with five oral doses of
60 mg/kg-day AN (Working et al., 1987), indicating no AN-induced germ cell mutations.
Chromosomal mutations in humans
There is sufficient evidence of AN-induced mutagenicity in occupationally exposed
workers. Fan et al. (2006) reported significant increase in the occurrence of MN in buccal
-3
mucosal cells of workers exposed to 0.52 or 1.99 mg/m AN for an average duration of 15.7—
17.2 years and significant increase in MN in peripheral blood lypmphocytes in workers exposed
-3
to 1.99 mg/m AN for an average of 17.2 years. The workers in the Fan et al. (2006) study were
exposed to higher concentrations than those in a previous study by Sram et al. (2004), in which
"3
the exposure concentration range was 0.05-0.3 mg/m AN, thus explaining the negative results
reported in Sram et al. (2004) or in other studies where exposure concentrations were not
reported. Evidence of CAs in AN-exposed workers was also reported by Borba et al. (1996), and
278
DRAFT - DO NOT CITE OR QUOTE
-------
DNA strand breakage and sex chromosome aneuploidy in the sperm of AN-exposed workers
were reported by Xu et al. (2003). In addition, Beskid et al. (2006) reported an increase in the
number of reciprocal translocations and the relative number of insertions in the chromosomes of
cultured lymphocytes of AN-exposed male workers.
Therefore, the overall in vitro and in vivo evidence in support of the direct mutagenicity
of CEO is strong, consistent, and specific.
4.7.3.2.2. Dose-response concordance. Chronic exposures of rats to AN in drinking water
concentrations >30 ppm (>2.5 mg/kg-day) or air concentrations >20 ppm were associated with
significantly increased incidences of brain tumors. No published studies are available that have
measured CEO-DNA adducts or other endpoints pertinent to mutagenicity in brain tissues
following acute-, subchronic-, or chronic-duration oral or inhalation exposures of rats to AN,
mainly because in vivo DNA adduct studies were conducted before the Solomon et al. (1993)
and Yates et al. (1993) studies that reported on DNA adducts formed from CEO in vitro. Hence,
these DNA adduct studies were not looking for the adducts formed in vitro from CEO. The
available studies that looked for CEO-DNA adducts in brain or other tissues involved single i.p.
(Hogy and Guengerich, 1986), s.c., or i.v. (Prokopczyk et al., 1988) administration protocols at
dose levels (50 or 100 mg/kg-day AN) that were higher than the chronic oral doses associated
with brain tumors in rats (0.3 to 40 mg/kg-day). However, no repeated-dose studies have been
conducted for detection of DNA adducts. The single oral dose of 46.5 mg/kg used in studies that
demonstrated UDS in the lung, gastric tissue, and testis of treated Sprague-Dawley rats (Abdel-
Rahman et al., 1994b; Ahmed et al., 1992a, b) was comparable to the high dose of 40 mg/kg-day
used in studies by Friedman and Beliles (2002), providing evidence that the dose that caused
DNA repair was carcinogenic.
A study in mice by Fahmy (1999) showed that increased CAs occurred in spermatocytes
of Swiss mice, following single oral doses of 15.5 or 31 mg/kg AN or five daily doses of
7.75 mg/kg, and in spleen cells, and bone marrow cells after a single oral dose of 7.75 mg/kg.
Fahmy (1999) also reported increases in SCEs in bone marrow cells of male Swiss mice after a
single i.p. dose of 7.5 or 10 mg/kg of AN. These doses were all in the range of, or comparable
to, the tumorigenic doses in B6C3Fi mice of 2.5-20 mg/kg-day in a 2-year bioassay.
The single doses employed by Sekihashi et al. (2002) that demonstrated DNA damage in
forestomach, colon, kidney, bladder, and lung of rats treated with 30 mg/kg i.p. and in colon,
bladder, lung, and brain of male ddY mice treated with 20 mg/kg i.p. were also within the range
of tumorigenic doses in the 2-year bioassay.
Therefore, the doses that demonstrated DNA damage or chromosome mutations in single-
dose studies are in concordance with tumorigenic doses in chronic bioassays.
279
DRAFT - DO NOT CITE OR QUOTE
-------
4.7.3.2.3. Temporal relationships. Currently available examinations of DNA damage,
chromosome mutations, SCEs, UDS, or CEO-DNA adducts in brain or other tissues are
restricted to single-dose acute administration protocols. Studies designed to examine temporal
relationships of key events, such as the presence of CEO-DNA adducts in brain tissue, are not
available. Nevertheless, results from the rat bioassays indicated that most AN tumors occur after
12-14 months of exposure or longer. Thus, the observed mutagenic effects of AN occurred
before tumor formation and provide support for a mutagenic mode of action.
4.7.3.2.4. Biological plausibility and coherence. The hypothesis that the key primary event in
AN induction of brain tumors is the formation of CEO-DNA adducts that lead to mutations that
initiate tumor formation is plausible based on in vitro and in vivo evidence for the direct
mutagenicity of the AN metabolite, CEO. However the available data do not establish that this is
the only mode of action by which AN may induce brain tumors in rats. Notably lacking in the
database are studies designed to detect a range of DNA adducts or mutations associated with
tumor initiation in brain tissues following prolonged oral or inhalation exposures at levels that
induced brain tumors in the chronic rat bioassays. In vitro studies by Solomon et al. (1993)
n
reported that CEO interacted with calf thymus DNA and, in addition to N -(oxoethyl)guanine,
formed N -(2-hydroxy-2-carboxyethyl)deoxyuridine from an initial cytosine adduct. Other
adenine and thymine adducts were also formed. Yates et al. (1993) demonstrated that CEO
"3
reacted calf thymus DNA formed N -(2-cyano-2-hydroxylethyl)-deoxythymidine. Yet, only
n
N -(oxoethyl)guanine has been measured in exposed rats in available studies (Hogy, 1986).
The available data do not provide an explanation of why AN induces brain tumors in
F344 and Sprague-Dawley rats but not in B6C3Fi mice. Kedderis et al. (1993b) reported that
when male F344 rats and male B6C3Fi mice were administered 10 mg/kg AN in water by
gavage, higher CEO concentrations were found in blood and brains of rats than in mice (13%
higher in blood, 23% higher in brain). Higher CEO concentrations in rat brain may at least
partially explain why tumors were only found in the brains of exposed rats but not in mice. In
addition, the clearance of CEO in mice was more rapid than in rats (Roberts et al., 1991). It may
also be that mice are more efficient in repairing DNA damage since DNA damage has been
detected in the brain of ddY mice exposed to 20 mg/kg AN, i.p. (Sekihashi et al., 2002). It has
been noted that the B6C3Fi mouse is generally insensitive to chemically induced neurogenesis in
NTP carcinogenesis bioassays (Radovsky & Mahler, 1999).
The difference in susceptibility to AN induction of brain tumors between mice and rats is
similar to that observed for glycidol, an aliphatic epoxide structurally similar to CEO (Irwin et
al., 1996). Glycidol is a direct acting alkylating agent, which induces tumors at a variety of sites
in both rats and mice. However, although significant induction of gliomas was observed in both
sexes of F344 rats after glycidol treatment, no brain tumors were induced in B6C3Fi mice.
Ethylene oxide also shows a similar pattern of tumorigenesis, inducing brain tumors in both
280
DRAFT - DO NOT CITE OR QUOTE
-------
sexes of exposed rats (Garmin et al., 1985; 1986), but no brain tumors in exposed mice (NTP,
1987). Thus, induction of brain tumors in rats but not in mice by known genotoxic carcinogenic
appears to reflect primarily a species difference in inherent susceptibility to brain tumorigenesis.
There have been no investigations to attempt to explain the apparent susceptibility of
early-life plus chronic adult exposures compared with adult-only exposures. It is possible that
mutations are established more effectively in rapidly dividing tissues such as the fetal brain,
since the higher rate of cell division increases the number of cells in S phase (Slikker et al.,
2004), which tends to make the cell population more vulnerable to the genotoxic effects of
chemicals. It is also possible that the lack of key DNA repair mechanisms in embryonic tissues
may be involved. However, DNA repair was not found in adult rat brain after administration of
6 mg/kg CEO i.p. (Hogy and Guengerich, 1986). It would appear that the rapid cell division in
earlier stages of development is a determining factor for the greater mutagenic and subsequent
carcinogenic sensitivity of perinatal life (Slikker et al., 2004).
Data for another chemical, ethylene oxide, that causes brain tumors (gliomas) in rats
n
show a different accumulation pattern for N -(2-oxoethyl)guanine adducts than that observed
following i.p. administration of AN or CEO. In rats exposed to 500 ppm ethylene oxide for
1 day, higher levels of N -(2-oxoethyl)guanine adducts were detected in DNA from brain than in
n
DNA from liver (Walker et al., 1990), whereas N -(2-oxoethyl)guanine adducts were detected in
brain at detection limit and only at low levels in the liver of rats given single i.p. injections of
50 mg/kg AN or 6 mg/kg CEO (Hogy and Guengerich, 1986). Whysner et al. (1998b) suggested
that these and other results indicated that glioma formation from chronic exposure to AN may
not involve the formation of N -(2-oxoethyl)guanine adducts; however, several alternative
explanations are possible. The lack of unequivocal detection of CEO-DNA adducts in rat brain
may indicate that the detection limits of the methods used were not sensitive enough. More
sensitive methods of detection, such as liquid chromatography-mass spectrometry (Poirier,
2004), have not been used in AN-induced DNA adduct studies. Alternatively, the stringent
methods used to isolate purified DNA (without associated proteins) in the experiments by Hogy
and Guengerich (1986) may have caused the loss of adducts or inhibited the recovery of
n
adducted DNA (Meek et al., 2003). Another possibility is that N -(2-oxoethyl)guanine adduct
may not be involved in mutations leading to AN-induced brain tumors, and other, as yet
uninvestigated, DNA adducts that were reported by Solomon et al. (1993) and Yates et al. (1993)
may be involved. Still another possibility is that CEO-DNA adducts and resultant mutations in
target tissues may occur only after prolonged exposure to AN. Notably absent from the available
mode-of-action database are experiments designed to detect a range of possible DNA adducts in
target tissues following repeated oral or inhalation exposure at exposure levels producing tumors
in the chronic bioassays.
Guengerich et al. (1986) discussed the findings in which AN was metabolized by liver
n
microsomes but not brain microsomes to form CEO. N -(2-oxoethyl)guanine DNA adducts were
281
DRAFT - DO NOT CITE OR QUOTE
-------
detected in liver but not the brain, yet AN induced tumors in the rat brain with chronic exposure
but not in the adult rat liver. In addition, AN-induced UDS was demonstrated in rat liver but not
brain (Hogy and Guengerich, 1986). Guengerich et al. (1986) proposed that AN was
metabolized in the liver to CEO. Since CEO formed by liver microsomes from AN has a half-
life of about 2 hours in neutral buffer, it can be transported easily via blood from liver to the
brain. Although DNA adducts have not yet been detected unequivocally in rat brains, CEO has
been measured in rat brains (Kedderis et al., 1993b) and shown to bind to brain DNA (Farooqui
and Ahmed, 1983b). Liver cells are efficient in repairing DNA damage via UDS, while brain
cells do not have this capability. This could help to explain why the rat brain is a target tissue of
AN carcinogenicity but the adult rat liver is not. It should be noted that hepatomas were found in
rats that were exposed to AN during gestation and extending through adulthood (Maltoni et al.,
1988). It is conceivable that the rat fetal and neonatal liver do not have sufficient capability to
repair DNA damage.
Meek et al. (2003) noted that there are several aspects of the development of AN-induced
tumors that are characteristic of tumors induced by compounds or metabolites that directly
interact with DNA. These comparisons add support to the evidence of a direct mutagenic mode
of carcinogenic action.
• Tumors are systemic and occur at multiple sites. Exposure-related increased incidences
were found for forestomach tumors, CNS tumors, and Zymbal gland tumors in chronic
oral exposure bioassays with F344 rats and Sprague-Dawley rats (which also showed
exposure-related increased incidence of tongue tumors); for forestomach and Harderian
gland tumors in a chronic oral exposure bioassay with B6C3Fi mice; for brain tumors,
Zymbal gland tumors, intestinal tumors, and tongue tumors in a chronic inhalation
exposure bioassay with Sprague-Dawley rats; and for brain tumors, Zymbal gland
tumors, hepatomas, and extrahepatic angiosarcomas in a chronic inhalation bioassay with
Sprague-Dawley rats exposed during gestation and extending throughout adulthood.
Particularly noteworthy is that Zymbal gland tumors in rats commonly occur with
carcinogens that are also mutagens. Of the 27 chemicals associated with site-specific
tumor induction in this gland found in NTP database, 23 were mutagenic in the
Salmonella assay. The NTP database indicated that these chemicals were multisite
carcinogens.
• Tumors sometimes occur at nontoxic doses or concentrations. Elevated incidences for
CNS tumors occurred in Sprague-Dawley (Quast, 2002; Quast et al., 1980a) and F344
(Johannsen and Levinskas, 2002b; Biodynamics, 1980c) rats chronically exposed to
drinking water concentrations (30 or 35 ppm) that did not induce elevated incidences of
nonneoplastic CNS lesions in interim sacrifices at 6, 12, or 18 months.
• Tumors sometimes occur after less-than-lifetime exposure durations. In an inhalation
study by Maltoni et al. (1977), Sprague-Dawley rats were exposed to 0-40 ppm AN via
282
DRAFT - DO NOT CITE OR QUOTE
-------
inhalation for 52 weeks. The animals were then allowed to complete their natural life
spans, with the final deaths occurring in week 136 (Maltoni et al., 1977). Gliomas were
developed in 20 and 40 ppm males at 84 and 63.5 weeks, respectively, and not in other
dose groups. In addition, forestomach papillomas and acanthomas (latency 103—
124 weeks) and Zymbal gland carcinomas (latency 77-102 weeks) were developed in
exposed rats. Thus, these tumors were developed long after exposure was terminated,
indicating the carcinogenic effect from AN exposure was not reversible. Although the
incidence of these tumors did not reach statistical significance, development of these
tumors does not support the hypothesis that AN-induced tumors are due to the promoting
action of AN alone.
• Tumors developed as early as 7-12 months following AN exposure. In the Sprague-
Dawley chronic drinking water bioassay, the number of high-dose (300 ppm) female rats
dying or sacrificed with CNS tumors were 1/1 during months 0-6, 5/13 during months
7-12, 14/23 during months 13-18, and 11/11 during months 19-24 (Quast, 2002). At the
lower exposure levels in this study (35 and 100 ppm), nearly all of the tumors in female
rats were detected after 13 months of exposure. In all groups of exposed Sprague-
Dawley male rats, all but one tumor were detected after 13 months of exposure (Quast,
2002). In the F344 chronic drinking water bioassay, the time of first detection of a rat
with a brain tumor was 481 days for male rats and 495 days for female rats,
approximately 16 months (Johannsen and Levinskas, 2002b; Biodynamics, 1980c). In
the three-generation reproductive toxicity study involving 46 weeks of exposure to AN in
drinking water, brain tumors were detected in 0/19, 1/20, and 2/24 F0 breeding females
exposed to 0, 100, or 500 ppm, respectively. In the other generations, incidences of brain
tumors were 0/20, 1/19, and 4/17 for the F1 breeding females and 0/20, 1/20, and 1/20 for
the F2 breeding females (Friedman and Beliles, 2002). Statistically significantly elevated
incidences of brain tumors occurred in Sprague-Dawley rats exposed by inhalation to
60 ppm AN during gestation and extending chronically throughout adulthood, but rats
exposed during gestation followed by a subchronicsubchronic-duration exposure during
adulthood showed an elevated incidence of brain tumors that was not statistically
significant when compared only with control (Maltoni et al., 1988). The weight of the
available evidence indicates that most AN-induced brain tumors require at least a half-
lifetime duration of exposure to develop, but associations with shorter durations of
exposure have been observed, specifically in female Sprague-Dawley rats exposed to 300
ppm in drinking water.
• The ratio of benign to malignant tumors is small. The brain tumors noted in AN-exposed
Sprague-Dawley or F344 rats were astrocytomas, which were malignant tumors. Most of
the Zymbal gland tumors were also malignant tumors.
283
DRAFT - DO NOT CITE OR QUOTE
-------
4.7.3.2.5. Human relevance. The metabolic scheme of AN in rats and humans is similar.
Humans are known to be able to activate AN to its reactive metabolite, CEO. As discussed in
Section 4.5.2.1, studies that demonstrate the mutagenicity of AN in humans are available. There
is sufficient evidence for mutagenicity of AN in exposed humans, and the direct mutagenic mode
of action of AN-induced carcinogenicity is considered to be relevant to humans.
4.7.3.3. Other Possible Modes of Action
Other modes of action may contribute, along with the direct mutagenic mode of action, to
tumorigenesis. These additional modes of action are evaluated in the following subsections. The
results of these evaluations indicate that these modes of action are not likely to be prinicipal
modes of action or to contribute to the carcinogenicity in a significant manner.
4.7.3.3.1. Oxidative stress
Key events
This mode of action hypothesizes that ROS are generated either directly from the oxidant
or are indirectly produced via activation of endogenous sources when the oxidant or its
metabolite is distributed to the target organ. Oxidative stress is induced. These free radical ROS
can interact with DNA and produce DNA damage leading to gene mutation for tumor initiation.
ROS can also interact with lipids via lipid peroxidation, resulting in cell damage.
One of the most prevalent biomarkers of oxidative DNA damage is 8-oxodG. This DNA
lesion has been found to produce mutations involving GC—>TA transversions due to base
mispairing and AT—>CG transversions due to misincorporation during DNA synthesis (Cheng et
al., 1992). G-C base pairs provide a common target for activating point mutations (e.g., in both
p53 and retinoblastoma tumor suppressor genes and in the ras family of oncogenes). Thus,
G-C base pairs in both tumor suppressor genes and oncogenes may represent a vulnerable target
for mutation by oxidative stress (Guyton and Kensler, 1993). The induction of base changes in
the DNA sequence of these genes may be the basis for tumor initiation by the oxidant.
Another role that oxidants may play in carcinogenesis is tumor promotion. ROS
generating systems are known to possess some of the biochemical actions of tumor promoters,
such as promoting a rapid and sustained decrease in antioxidant defenses, including SOD,
catalase, and glutathione peroxidase activities (O'Connell et al., 1986; Slaga et al., 1981).
Tumor growth may also be promoted by oxidative stress via modification of gene expression
through induction of gene transcription factors (e.g., NF/.-B or transcription factor protein [AP-1])
or change in DNA methylation status by ROS. Signal transduction pathways, including
AP-1 and NF/.-B, are known to be activated by ROS, and they lead to the transcription of genes
involved in cell growth regulatory pathway. Oxidative DNA damage can also result in DNA
hypomethylation by interfering with the ability of methyltransferases to interact with DNA,
allowing the expression of normally quiescent genes and promoting tumor growth. ROS can
284
DRAFT - DO NOT CITE OR QUOTE
-------
also induce the release of calcium from intracellular stores, resulting in the activation of kinases,
including PKC (Larsson and Cerrutti, 1989), which is known to regulate many intracellular
processes, including those related to growth and differentiation. Hence, any one or a
combination of these events may lead to enhanced cell proliferation or inhibition of programmed
cell death.
The plausibility of an indirect mutagenicity mode of action for AN-induced tumors is
evaluated in the following discussion.
Strength, consistency, specificity of association
Experimental data for indirect mutagenicity of AN are discussed in Section 4.5.2.5 and
summarized in Table 4-55. These studies are only briefly discussed here.
In vitro studies
There is some in vitro experimental evidence that supports an indirect mutagenic mode of
action for AN. When SHE cells were treated in vitro with 0-75 ng/mL AN in 12.5 ng/mL
increments, there was a dose-dependent increase in morphological transformation at 50, 62.5,
and 75 [j,g/mL after 7 days of exposure (Zhang et al., 2000). Levels of 8-oxodG isolated from
cells incubated with 75 ng/mL AN were increased to 192 and 186% of control after 2 and 3 days.
However, no increase in 8-oxodG was observed after 1 or 7 days.
AN-induced oxidative stress in SHE cells was confirmed in a later study by Zhang et al.
(2002). AN at 25, 50, or 75 [j,g/mL increased the amount of ROS in SHE cells after 4, 24, and
48 hours of treatment and increased xanthine oxidase activity 24 and 48 hours after treatment
with 75 [j,g/mL AN. AN also caused temporal changes in GSH levels and antioxidant enzyme
catalase and SOD activities. The involvement of CYP450 metabolism of AN in the production
of oxidative stress was indicated by the observation that inclusion of a nonspecific suicidal
inhibitor of CYP450 enzyme, ABT (0.5 mM), in the medium resulted in a significant reduction
(about 77%) in the cell transformation activity of 75 [j,g/mL AN (Zhang et al., 2002).
In another study, when cultured rat astrocytes were exposed for 4 or 24 hours to 0.01,0.1,
or 1 mM AN, up to a 3.9-fold increase in 8-oxodG was found in cellular DNA of the rat
astrocytes (Kamendulis et al., 1999a). No increase in 8-oxodG was found in rat hepatocytes
exposed to 0.01, 0.1, or 1 mM AN for 4 or 24 hours (Kamendulis et al., 1999a). Pu et al. (2006)
also reported a 3-fold increase in oxidative DNA damage (as measured by the fpg-modified
comet assay) in cultured D1TNC1 rat astrocytes treated with 1 mM AN for 24 hours. When
NHAs were treated with 200-400 [xM AN for 12 hours, a four- to sevenfold increase in the
generation of ROS and a greater than twofold increase in 8-oxodG were observed (Jacob and
Ahmed, 2003b).
In vivo studies
285
DRAFT - DO NOT CITE OR QUOTE
-------
Three studies on rats are available that investigated oxidative DNA damage in the brain
after exposure to AN in drinking water (Jiang et al., 1998; Whysner et al., 1998a). A two- to
threefold increase in 8-oxodG levels was found in cellular DNA from brain cortex of Sprague-
Dawley rats exposed to 50, 100, or 200 ppm AN in drinking water for 28 or 90 days (Jiang et al.,
1998). Levels of 8-oxodG in DNA increased with increasing exposure levels in this study. At
90 days, levels of 8-oxodG were more than twofold higher in DNA from 50 ppm rat brain cortex
compared with controls (Jiang et al., 1998). In addition, no increased levels of 8-oxodG were
found in the liver DNA of exposed rats at all dose levels following 14, 28, or 90 days of
exposure (Jiang et al., 1998). The liver is not a target organ for AN-induced carcinogenicity in
adult rats. In a follow-up study, Pu et al. (2009) reported an increase in 8-oxodG levels in brain
DNA of male Sprague-Dawley rats exposed to 100 or 200 ppm AN in drinking water for 28
days. However, EPA has identified issues with Pu et al. (2009) (see Section 4.5.1.2.1); therefore,
this study was not discussed separately here.
In another study, levels of 8-oxodG in brain nuclear DNA were increased by about
twofold in Sprague-Dawley rats exposed to 30 or 300 ppm AN in drinking water for 21 days and
by about 1.5-fold in rats exposed to 100 ppm AN for 94 days (Whysner et al., 1998a). However,
no significant increase in 8-oxodG levels was found in F344 rats exposed to 10, 30, or 100 ppm
for 21 days (Whysner et al., 1998a). (A statistically insignificant increase of about 30% was
found in the 3-100 ppm AN dose groups.) Levels of 8-oxodG in liver DNA were increased by
about 1.4-fold in Sprague-Dawley rats exposed to 30 or 300 ppm for 21 days and by about
1.3- and 2-fold following exposure to 100 ppm for 10 and 94 days, respectively (Whysner et al.,
1998a). Levels of 8-oxodG in liver DNA were not measured in F344 rats in this study. No
significant increase in 8-oxodG levels in DNA of forestomach (a target organ of AN
carcinogenicity) were found in Sprague-Dawley rats exposed to 3-300 ppm AN (see
Table 4-54).
Several inconsistencies can be found in the results of Jiang et al. (1998) and Whysner et
al. (1998a). While a dose-related increase in 8-oxodG levels in brain cortex DNA of Sprague-
Dawley rats exposed up to 200 ppm AN was reported by Jiang et al. (1998), a twofold increase
in 8-oxodG levels in brain DNA was found for Sprague-Dawley rats exposed to either 30 or
300 ppm in the study by Whysner et al. (1998a). Thus, there was no increase in 8-oxodG level
with a 10-fold increase in exposure concentration in Sprague-Dawley rats (Whysner et al.,
1998a). While 8-oxodG levels were measured via different sample preparation methods in these
two studies (i.e., 8-oxodG level was measured in nuclear DNA in whole brain in Whysner et al.
[1998a] and in cellular DNA from brain cortex in Jiang et al. [1998]), 8-oxodG level measured in
Sprague-Dawley rats exposed to 3 or 30 ppm AN for 21 days in Whysner et al. (1998a) was
comparable to that measured in Sprague-Dawley rats exposed to 5 or 50 ppm AN for 28 days in
Jiang et al. (1998) (0.86/105 and 1.35/105 dG vs. 1.8/105 and 2.5/105 dG). Thus, inconsistencies
286
DRAFT - DO NOT CITE OR QUOTE
-------
in the two studies regarding 8-oxodG levels in rat brain are not likely due to differences in
sample preparation.
Inconsistencies were also found in 8-oxodG levels in livers in the two rat studies. Jiang
et al. (1998) reported no increase in 8-oxodG levels in liver DNA of exposed rats, but Whysner
et al. (1998a) reported a 1.4-fold increase in 8-oxodG levels in liver DNA of Sprague-Dawley
rats exposed to 30 or 300 ppm for 21 days and a 1.3- and twofold increase following exposure to
100 ppm AN for 10 and 94 days, respectively. Moreover, the increase in 8-oxodG levels in brain
DNA was not much higher than that in liver DNA (see Table 4-55). The rat brain is a target
organ for AN-induced carcinogenicity but not adult rat liver. In addition no significant increase
in 8-oxodG levels in forestomach DNA was found. Thus, no specificity is indicated regarding
increased 8-oxodG levels (oxidative DNA damage) in target organ DNA and tumor formation.
Inconsistencies were also found regarding AN-induced disruption of antioxidant defense.
While Jiang et al. (1998) reported increases in ROS and concomitant persistent decreases in
antioxidant enzyme catalase activity in the brain cortex of Sprague-Dawley rats exposed to 50-
200 ppm AN in drinking water after 14, 28, and 90 days, as well as decrease in SOD activity and
GSH level in all dose groups after 14 days of treatment, Whysner et al. (1998a) reported no
changes in catalase and glutathione peroxidase activities and GSH levels in the brains of
Sprague-Dawley rats treated with 3, 30, or 300 ppm AN in drinking water for 21 days. Whysner
et al. (1998a) did report a dose-related increase in cysteine levels in the brains of Sprague-
Dawley rats exposed to AN for 21 days, and the increase was significant for the 300 ppm dose
groups. However, no changes in GSH and cysteine levels were found in the brains of F344 rats
exposed to 0, 1,3, 10, 30, or 300 ppm AN in the drinking water for 21 days (Whysner et al.,
1998a). Since F344 rats exposed to these dose levels of AN also developed brain tumors in
chronic bioassay, the formation of these tumors cannot be explained by oxidative stress resulting
from disruption of antioxidant defense. Whysner et al. (1998a) also found no changes in
cytochrome oxidase activities in the brain mitochondria of both exposed Sprague-Dawley rats
and F344 rats. Cyanide, a metabolite of AN, is a noncompetitive inhibitor of cytochrome
oxidase. No change in cytochrome oxidase activity indicated that no metabolic hypoxia occurred
in brain mitochondria as a result of inhibition of the enzyme by cyanide. Therefore, cyanide-
induced metabolic hypoxia did not appear to be involved in the mechanism of ROS generation
by AN.
Moreover, Whysner et al. (1998a) reported no changes in TBARS in the brains of all
groups of AN-exposed Sprague-Dawley rats, indicating the absence of lipid peroxidation,
another biomarker of oxidative stress and oxidative lipid damage. Jiang et al. (1998) reported
significant increase in MDA only in the brain cortex of rats exposed to 200 ppm AN for 14 days
and not in other dose groups at 14 days. No increase was found in all dose groups at 28 and
90 days. Since lipid peroxidation is another biomarker of oxidative stress, there is no strong
evidence for occurrence of significant oxidative stress in the brain of AN-exposed rats.
287
DRAFT - DO NOT CITE OR QUOTE
-------
In addition, Chantara et al. (2006) demonstrated that AN induced ERK activation via
PKC in SK-N-SH neuroblastoma cells. However, oxidative stress was found not to be involved
in AN-induced ERK1/2 activation, which played a crucial role in cell proliferation and tumor
progression. Thus, the potential tumor promotion effect of AN has not been related to oxidative
stress.
Dose-response concordance
Levels of 8-oxodG in DNA from brain cortex of Sprague-Dawley rats (Jiang et al., 1998)
or rat astrocytes (Kamendulis et al., 1999a) increased with increasing in vivo (50, 100, 200 ppm)
or in vitro (0.01, 0.1, 1.0 mM) exposure levels. However, 8-oxodG in DNA from brain of F344
rats exposed to 1-100 ppm AN were not significantly increased. In addition, 8-oxodG levels in
DNA from brain of F344 rats did not show a dose-response relationship (see Table 4-54), and no
correlation can be found between 8-oxodG levels in brain DNA of these rats and brain tumor
incidence of F344 rats exposed to the same concentration of AN in drinking water for 2 years
(Table 4-58). Thus, experimental data do not support the hypothesis that brain tumors from
F344 rats are the result of oxidative DNA damage.
Table 4-58. 8-OxodG in brain DNA and brain tumor incidence in male F344
rats exposed to AN in drinking water
Dose group (ppm)
8-oxodG (mol/105 mol dG)a
Incidence of brain astrocytomas'"
0
0.79 ±0.37
2/160
1
0.84 ±0.25
2/80
3
1.07 ±0.41
1/78
10
1.04 ±0.30
2/80
30
1.03 ±0.38
10/79
100
1.06 ±0.48
21/76
aData are from 21-d drinking water study by Whysner et al. (1998a).
bData are from 2-yr drinking water study by Biodynamics (1980b) and Johannsen and Levinskas (2002b). The
denominators for incidence of brain astrocytomas excluded rats from the 6- and 12-mo interim sacrifices and rats
that died before the appearance of the first tumor for this site.
Moreover, although Whysner et al. (1998a) demonstrated a significant increase in levels
of 8-oxodG in brain DNA of Sprague-Dawley rats exposed to AN for 21 days, no correlation can
be found between 8-oxodG levels in brain and tumor incidence in the 2-year bioassay
(Table 4-59). Therefore, dose-response data on Sprague-Dawley rats from Whysner et al.
(1998a) did not support oxidative DNA damage as the mode of action for AN-induced brain
tumor formation.
288
DRAFT - DO NOT CITE OR QUOTE
-------
Table 4-59. 8-OxodG in brain DNA and brain tumor incidence in male
Sprague-Dawley rats exposed to AN in drinking water
Dose group (ppm)
8-oxodG (mol/105 mol dG)a
Incidence of brain astrocytomas'"
0
0.62 ±0.08
1/80
1
0.86 ±0.41
ND
3
1.35 ±0.49
ND
10
ND
8/47
30
ND
19/48
100
1.29 ± 0.10
23/48
aData are from 21-d drinking water study by Whysner et al. (1998a).
bData are from 2-yr drinking water study by Quast (2002).
ND = cannot be determined
Temporal relationships
The detection of increased 8-oxodG levels in brains of Sprague-Dawley rats exposed for
subchronic durations (Jiang et al., 1998; Whysner et al., 1998a) is temporally consistent with
oxidative DNA damage being a plausible key precursor event in the development of later-
appearing tumors. However, no significant increase in 8-oxodG levels in the brain of exposed
F344 rats was found.
Biological plausibility and coherence
The demonstration of AN-induced oxidative DNA damage in rat brain cortexes following
sub chronic-duration exposure to AN at dose levels producing brain tumors with chronic
exposure would have provided support for the involvement of an indirect mutagenic mode of
action in AN carcinogenicity. However, no increase in oxidative DNA damage was found in
F344 rats exposed to AN in drinking water. Brain tumors were found in F344 rats at similar
frequencies as Sprague-Dawley rats in chronic bioassays. Thus, brain tumors in F344 rats cannot
be explained by oxidative DNA damage, and the predicted greater sensitivity of Sprague-Dawley
rats vs. F344 rats based on 8-oxodG levels measured in short-term studies is not reflected in
cancer bioassays. In addition, the presence of increased oxidative DNA damage in the livers of
Sprague-Dawley rats exposed to AN in drinking water in one study (Whysner et al., 1998a) but
not in another (Jiang et al., 1998) also raised the question regarding a significant association
between oxidative DNA damage and tumor formation.
The origin of oxidative stress is unclear. Proposed molecular mechanisms that may be
involved in AN induction of oxidative stress in the brain (or other toxicity targets) include direct
generation of free radicals by AN (or its metabolites), stimulation by AN or its metabolites of
systems that generate free radicals, binding of AN to free radical scavengers (e.g., GSH, vitamins
C or E) and depletion of stores of these antioxidants, inhibition of the expression or activities of
289
DRAFT - DO NOT CITE OR QUOTE
-------
antioxidant enzymes, and interference of mitochondrial respiratory electron flow via cyanide
inhibition of cytochrome c oxidase (Zhang et al., 2002; Jiang et al., 1998). As discussed
previously, some of these potential mechanisms have been ruled out.
As with the direct mutagenic mode of action, there are no data currently available to
indicate how or if this indirect mutagenic mode of action may explain the occurrence of brain
tumors in rats but not in mice, following chronic exposure to AN.
Human relevance
Humans possess the biochemical pathways for the key steps of this proposed mode of
action. This hypothetical mode of action, if it occurs, is considered to be relevant to humans.
Conclusion
While plausible, this postulated mode of action is not supported by in vivo studies in rats.
Studies on oxidative DNA damage or oxidative lipid damage in the brain of exposed rats do not
provide sufficient evidence to support this mode of action. In addition, Chantara et al. (2006)
demonstrated AN-induced ERK activation via PKC in SK-N-SH neuroblastoma cells. Oxidative
stress was found to be not involved in AN-induced ERK1/2 activation, which played a crucial
role in cell proliferation and tumor progression. Therefore, while this mode of action may play a
role, it is not likely to be the principal mode of action for AN-induced carcinogenicity.
4.7.3.3.2. Other modes of action for brain tumors
Key events
Other hypothesized modes involve actions by AN or its metabolites to directly or
indirectly alter the expression of genes (either at the level of translation, post-translation, or
protein activity), leading to the loss of control of cell growth and the ultimate promotion of
initiated cells into brain tumors. Possible modes of AN carcinogenic action include stimulation
of cell proliferation (mitogenic), cytotoxicity with subsequent reparative cell proliferation
(cytotoxic), inhibition of programmed cell death (anti-apoptotic), and inhibition of GJIC.
Studies specifically designed to examine these possible modes of action are restricted to a
study of GJIC in rat astrocytes (Kamendulis et al., 1999b).
Strength, consistency, and specificity of association
No studies are available that examine cellular proliferation indices in brain cells
following in vivo exposure to tumor-producing doses of AN. In in vitro studies with rat
astrocytes, indices of cytolethality after 24 hours of exposure occurred at higher AN
concentrations (2.5, 5.0, and 10.0 mM) than increased 8oxodG levels in DNA (0.01, 0.1, and
1.0 mM) (Kamendulis et al., 1999a). The available data do not support a prominent role for
cytotoxic or mitogenic modes of action in AN-induced rat brain tumors.
290
DRAFT - DO NOT CITE OR QUOTE
-------
Inhibition of GJIC has been shown to correlate with tumor promotion activity (i.e., the
loss of control of growth) and to be induced by various chemical agents thought to operate via
other modes of carcinogenic action, such as phorbol esters and PB (Kamendulis et al., 1999b).
Exposure of rat astrocytes to 0.01, 0.1, or 1 mM AN for 4-48 hours statistically significantly
inhibited GJIC compared with controls (Kamendulis et al., 1999b). The inhibition was reversible
and prevented by the presence of an antioxidant, a-tocopherol, or a precursor for the synthesis of
glutathione, OTC, in the culture medium. The results are consistent with the involvement of
AN-induced oxidative stress in the inhibition of GJIC. The inhibitory concentrations were the
same as those that produced oxidative DNA damage in a companion experiment (Kamendulis et
al., 1999a). The specificity of the inhibitory response to rat astrocytes was demonstrated by the
lack of inhibition of GJIC in rat hepatocytes exposed to 0.01, 0.1, or 1.0 mM AN (Kamendulis et
al., 1999b).
Dose-response concordance
The inhibition of GJIC in rat astrocytes increased with increasing sublethal
concentrations in the range of 0.01-1 mM (Kamendulis et al., 1999b). A concordance of these
concentrations to dose levels associated with brain tumors is not available.
Temporal relationships
The acute nature of the observed inhibition of GJIC is consistent with the hypothesis that
this action may be one of a number of precursor events involved in tumor promotion.
Biological plausibility and coherence
Other potential modes of action for AN-induced brain tumors have not been adequately
studied. The involvement of inhibition of GJIC by oxidative stress induced by AN or its
metabolites is plausible, based on the limited evidence with rat astrocytes (Kamendulis et al.,
1999b). However, the available histopathology data from interim sacrifices of the chronic rat
bioassays do not support the involvement of a mode of action involving brain cell cytotoxicity
followed by reparative cell proliferation.
4.7.3.3.3. Conclusions about modes of action for brain tumors. Data gaps still exist in the current
understanding of the mode of action for carcinogenicity of AN. However, there is sufficient
experimental evidence to support direct mutagenicity as the principal mode of action. Key
events are the generation of direct DNA damage by the AN metabolite, CEO, and interaction
with DNA. There is in vitro and in vivo evidence to support the occurrence of key events in the
brain following AN exposure. Other modes of action may contribute, along with direct
mutagenesis, to tumorigenesis. For example, AN-induced ERK activation via PKC may play a
role in cell proliferation and tumor progression. However, limited experimental evidence does
291
DRAFT - DO NOT CITE OR QUOTE
-------
not support them as alternatives. Available data are inadequate to establish an oxidative stress
mode of action for AN-induced carcinogenicity.
The available data do not provide an explanation for AN induction of brain tumors in
F344 and Sprague-Dawley rats but not in B6C3Fi mice. However, it should be noted that
generally mice are much less susceptible than rats in developing brain tumors resulting from
exposure to chemical carcinogens (Rice and Wilbourn, 2000; Radovsky and Mahler, 1999).
Hence, this species difference in response is not limited to AN alone. In addition, there have
been no investigations to attempt to explain the apparent susceptibility of early-life plus chronic
adult exposures compared with adult-only exposures. It is possible that a higher rate of cell
division in the fetal brain and the resultant greater sensitivity to direct mutagenic damage may be
involved (Slikker et al., 2004). Possible differences between early-life and adult stages of rat
development in pertinent physiologic processes, such as distribution of AN or its metabolites to
the brain, levels of antioxidants, activities of antioxidant enzymes, or repair of DNA damage,
have not been investigated to explain this apparent susceptibility of early-life stages to AN
carcinogenicity.
Possible differences between rats and humans in distribution of AN or its metabolites to
the brain, susceptibility to oxidative stress, or repair of DNA damage have not been investigated.
Identified differences between rats and humans in AN disposition are restricted to the finding of
higher rates of metabolic oxidation of AN in human vs. rat hepatic microsomes, presumably due
to a more active EH in humans (Kedderis and Batra, 1993; Kedderis et al., 1993c). The
relevance of this apparent difference in AN metabolism to possible species differences in
susceptibility to AN carcinogenicity in the brain is not understood. Within the framework of the
hypothesized direct mutagenic mode of action, the balance between the formation of CEO by
CYP2E1 and its hydrolysis by EH is thought to be important in determining the levels of CEO
that might be available to bind DNA or GSH. Experimental comparison of this balance in
human and rat brain tissues is not available.
4.7.3.4. Possible Modes of Action for Forestomach Tumors
There is evidence to suggest that, once delivered to forestomach epithelial cells, AN or its
metabolites shift the normal balance between cell proliferation and apoptosis, leading to
hyperplasia and eventually, with continued exposure and sustained net cellular proliferation, to
tumor formation. Sustained increased cellular proliferation is viewed as the key precursor event
in this hypothesized mode of carcinogenic action. Exposure of male F344 rats to gavage doses
of about 12 and 23 mg/kg-day AN for 6 weeks produced minimal to mild hyperplasia and
hyperkeratinization of the squamous mucosa of the forestomach but not the epithelium of the
glandular stomach or the liver (Ghanayem et al., 1997). The early induction of hyperplasia by
AN doses that resulted in forestomach tumors in chronic rat bioassays is temporally consistent
with the involvement of sustained cell proliferation in the development of these tumors. A dose-
292
DRAFT - DO NOT CITE OR QUOTE
-------
related increase in cell proliferation (as determined by BrdU incorporation into DNA) was
observed in the forestomach epithelium, but no increases were found in the glandular stomach or
the liver, which are not targets of AN toxicity or carcinogenicity from chronic exposure to AN
during adulthood. At the high-dose level, an increase in apoptotic cells was found. These results
suggest that AN stimulation of cellular proliferation overcame the apparent stimulation of
apoptosis at the higher dose, since hyperplasia was observed. Whether or not the stimulation of
cellular proliferation was due to a reparative response to cytotoxicity or to a direct mitogenic
action is uncertain. Acute administration of a higher dose of AN (50 mg/kg) caused gastric
mucosal necrosis in rats, which was shown to involve CYP-mediated metabolism of AN
(Ghanayem et al., 1985; Ghanayem and Ahmed, 1983). Whether or not these findings relate to
the stimulation of cellular proliferation and development of hyperplasia at the lower doses is
uncertain.
Possible direct or indirect mutagenic modes of action for forestomach tumors, such as
those investigated for brain tumors, have not been investigated. However, DNA damage, as
detected by the comet assay, was reported in the stomach of rats and mice exposed by i.p.
injection (Sekihashi et al., 2002). No significant oxidative DNA damage (levels of 8-oxodG)
was measured in the forestomach of rats exposed to 3, 30, or 300 ppm in drinking water for
21 days (Whysner et al., 1998a). In addition, AN was reported to bind to DNA in the stomach of
rats following a single oral dose of 46.5 mg/kg (Farooqui and Ahmed, 1983a). Given that a
mutagenic MO A is hypothesized for AN induced brain tumors in rats, that mutagenic
carcinogens usually cause tumors in multiple sites, and evidence of DNA damage in the
forestomach of AN treated rats, direct mutagenicity is a likely mode of action for AN-induced
forestomach tumors.
4.7.3.5. Possible Modes of Action for Hepatomas
Although adult female Sprague-Dawley rats exposed to 60 ppm AN by inhalation during
adulthood did not develop hepatomas, these tumors were found in 5/67 male offspring and
1/54 female offspring exposed during gestation and extending throughout adulthood (Maltoni et
al., 1988). In unexposed offspring in this study, hepatomas were found in 1/158 males and
0/149 females. This apparent susceptibility of early-life stages to the initiation of liver tumors
may be suggestive of a direct mutagenic mode of action in which embryonic DNA repair
mechanisms are inadequate to repair AN-induced DNA damage during this stage of development
n
when cells are rapidly dividing. N -(2-oxoethyl)guanine-DNA adducts have been measured in
the livers of adult rats following i.p. exposure to AN (Hogy and Guengerich, 1986). Increased
levels of 8-oxodG have also been measured in DNA from the livers of adult Sprague-Dawley
rats exposed to AN in drinking water for 21 or 94 days (Whysner et al., 1998a). Thus, indirect
mutagenic mode of action may also contribute. However, possible modes of action for the
293
DRAFT - DO NOT CITE OR QUOTE
-------
development or initiation of liver tumors during early stages of development have not been
investigated.
4.7.3.6. Possible Modes of Action for Tumors of Intestines, Tongue, Zymbal Gland, and
Harderian Gland
Based on analogy to other direct-acting mutagens that cause tumors at multiple sites in
animal bioassays, it is possible that a direct mutagenic mode of action may be involved in the
formation of AN-induced tumors at these sites in rats and mice. Although possible modes of
carcinogenic actions of AN at these sites have not been investigated, direct mutagenic mode of
action is the most likely mode of action. CYP2E1 enzymes occur in intestinal mucosa, and CEO
can form in intestine and bind to DNA. In addition, there is no evidence that AN would have a
tissue specificity that would lead to mutagenesis in brain but not other organs. Moreover, AN
appears to act systemically with DNA damage, and tumors occur in multiple tissues of treated
animals. The direct mutagenic mode of action is considered relevant to all tumor sites
4.7.3.7. Possible Modes of Action for Lung Cancer
Evidence of a possible association between occupational exposure to AN and lung cancer
is found in some studies, including the best available epidemiologic study that workers in the
highest exposure category (>8 ppm-years) with more than 20 years of employment displayed a
twofold increased risk for lung cancer compared with unexposed workers (Blair et al., 1998).
Human bronchial epithelium has CYP2E1 metabolic activities. In a short-term mutagenicity
assay, AN induced SCEs and DNA single-strand breaks in human bronchial epithelial cells
without the addition of S9 mix (Chang et al., 1990). Thus, a direct mutagenic mode of action is
supported for potential AN-induced lung cancer.
In contrast, there is no convincing evidence of lung cancer in rodents chronically exposed
by the oral or inhalation routes. The possible mode of action by which AN may induce lung
tumors in humans and not in rodents has not been investigated but may involve species
differences in inhalation rates and anatomical features of the respiratory tract.
4.8. SUSCEPTIBLE POPULATIONS AND LIFE STAGES
4.8.1. Possible Childhood Susceptibility
Evidence that children may be more susceptible to the acute or chronic toxicity of AN is
restricted to a report that children died after sleeping in rooms fumigated with a commercial form
of AN called Ventox, whereas adults sharing the same rooms only experienced skin or eye
irritation (Grunske, 1949; see Section 4.1.3.1).
Animal studies specifically designed to examine whether or not early-life stages of
development are more susceptible than adult stages to the induction of noncancer or cancer-
related effects by AN include a lifetime inhalation cancer bioassay in Sprague-Dawley rats
(Maltoni et al., 1988) and a three-generation drinking water reproductive toxicity bioassay in
294
DRAFT - DO NOT CITE OR QUOTE
-------
Sprague-Dawley rats that provides evidence of increased tumor incidences in F1 and F2 female
breeders after 46 weeks of exposure starting in utero (Friedman and Beliles, 2002). In addition,
a subacute study on Sprague-Dawley rats (Szabo et al., 1984) also found weanling rats to be
more susceptible than adult rats to the action of AN on the adrenals (see Section 4.2.1.1).
Results from the lifetime inhalation bioassay showed that rats that were chronically
exposed starting in utero had higher incidences of tumors at several sites (brain, Zymbal gland,
mammary gland, liver, and blood vessels) compared with rats chronically exposed during
adulthood only (Maltoni et al., 1988; see Section 4.2.2.2.2 for details). For example, brain
tumors occurred in 5.5% of rat dams chronically exposed to 60 ppm AN compared with no brain
tumors in unexposed controls (Table 4-42). In contrast, 16.4% of male and 18.5% of female
offspring of the exposed dams had brain tumors compared with 1.3% of male and 1.3% of
unexposed female rats exposed starting in utero, otherwise under the same exposure protocol as
the dams. The responses for brain tumors (male and female offspring), extrahepatic
angiosarcomas (female offspring), Zymbal gland (male offspring), and mammary gland (female
offspring) with chronic exposure were about threefold higher than those observed in female rats
exposed only during adulthood (Table 4-42). In addition, although none of the exposed or
unexposed dams had hepatomas, 5/67 male offspring with early-life plus chronic exposure had
hepatomas compared with 1/158 unexposed male offspring; hepatomas were observed in
1/54 exposed female offspring compared with 0/149 unexposed female offspring. Rats with
early-life plus subchronic (8-week) adult exposure did not show elevated incidences for these
tumors compared with adults with chronic adult exposure (Table 4-42).
Comparison of the results from the two chronic phases of this experiment indicated that,
compared with chronic exposure to 60 ppm during only adult stages of development, chronic
exposure to 60 ppm AN starting during gestation increased the risk for development of
extrahepatic angiosarcomas and malignant mammary gland tumors in females, hepatomas and
Zymbal gland carcinomas in males, and encephalic gliomas in males and females. Although
testing in adult male rats was not included in this study, the increase in hepatomas in male
offspring may indicate differences in detoxification and DNA repair mechanisms between fetal
and adult livers.
In the drinking water bioassay involving 46-week exposures of each of three generations
of female breeder Sprague-Dawley rats, there were slightly increased incidences of tumors in
F0 or F2b female breeders in the 100 or 500 ppm groups compared with controls, which were not
statistically significant, while Fib female breeders in the 500 ppm group showed statistically
significantly increased incidences of brain and Zymbal gland tumors compared with control, at
24 (4/17) vs. 0 and 18 (3/17) vs. 0%, respectively (Friedman and Beliles, 2002). Compared with
the 500 ppm F0 breeders that were exposed starting in adulthood, there was an approximate
twofold increase in Zymbal gland tumor incidence and a threefold increase in brain tumor
incidence in the 500 ppm Fib female breeders, exposed starting in utero (Table 4-35). The
295
DRAFT - DO NOT CITE OR QUOTE
-------
results provide evidence for the increased susceptibility to the carcinogenicity of AN during
early-life exposure, positive evidence being in the F1 generation. Increases in tumor incidence in
the F2 generation were not as pronounced and were not statistically significant.
As discussed in Sections 4.5.1 and 4.6.3, toxic effects from acute or chronic exposure to
AN have been associated with inhibition of glutathione-mediated detoxification of AN by
conjugation (e.g., glutathione depletion), transformations in the CYP2E1 metabolic pathway to
CEO and cyanide, and oxidative stress. Differences in enzymatic activities (e.g., CYP2E1, EH,
DNA repair enzymes, or antioxidant enzymes such as catalase or SOD) or pool sizes of reactive
oxygen scavengers (e.g., levels of vitamin E or C) or glutathione between early-life and adult
stages may result in life stage differences in susceptibility to AN toxicity.
For CYP2E1, immunoreactive CYP2E1 protein has been detected in human liver
microsomes as early as the second trimester (GDs 93-186) (Johnsrud et al., 2003). CYP2E1
enzyme activity increases shortly after birth but less in neonates than in older infants, children,
and adults (Johnson, 2003; Johnsrud et al., 2003; Vieira et al., 1996). However, in the fetal
brain, CYP2E1 activity is seen as early as 50 days gestation, with increasing levels seen to at
least the end of the first trimester (Brzezkinski et al., 1999). In rats, CYP2E1 protein is not
significantly expressed in fetal hepatic tissues, although an elevation of CYP2E1 mRNA was
seen within a few hours after birth, coincident with the transcriptional activation of the gene
(Borlakoglu et al., 1993). Moreover, only neonates expressed small quantities of the CYP2E1
protein, despite the large quantities of CYP2E1 mRNA found in perinatal and neonatal rats.
For EH, fetal immunoreactive enzyme content in human livers averages only 25% of that
found in adults with corresponding less enzyme activity (Cresteil et al., 1985). EH activity in
fetal liver and adrenal glands are about threefold higher than in kidney and lung with only a
weak correlation with gestational age between 10 and 25 weeks. Overall, EH activity in fetal
tissue is about 30-40% of that in adults (Pacifici and Rane, 1983b; Pacifici et al., 1983a). In
another study (Omiecinski et al., 1994), human microsomal EH (mEH) was detected in fetal liver
as early as 7.2 weeks, although at a much lower level of mEH protein, and demonstrated a linear
increase with gestational age to a level at 77 weeks that is about half of that observed in adult
liver. However, mEH activity in the fetal lung did not correlate with increasing age. Lung mEH
activity from days 85 to 130 of gestation was maintained at consistent levels, at about the same
level as fetal hepatic EH at 53 days, and varied only by threefold. In addition, Omiecinski et al.
(1994) reported that mEH activities in liver or lung, from either fetal or adult tissues, did not
correlate with corresponding mRNA levels.
The overall balance of pertinent enzymatic activities (AN metabolizing enzymes and
antioxidant enzymes) and pool sizes of reactive oxygen scavengers and glutathione will
determine the relative susceptibility of an individual or a life stage. The relatively high activity
of CYP2E1 in the brain compared to the liver of the developing human fetus, and low EH
activity for detoxification raise concern for increased susceptibility in early life to lung tumors,
296
DRAFT - DO NOT CITE OR QUOTE
-------
and brain tumors from AN exposure, as seen in rats. Additional research examining age-
dependent changes in these physiological variables and responses to AN may help to provide
explanations for the observed susceptibility of early-life stages in rats to the carcinogenicity of
AN, and is outside the scope of this assessment.
The importance of CYP2E1 metabolism in the acute toxicity and lethality of AN has been
demonstrated by the lack of lethality or gross signs of intoxication in CYP2El-null male mice
given single gavage doses up to 40 mg/kg, whereas all WT male mice given doses of 40 mg/kg
died within 3 hours of administration (Wang et al., 2002). This type of research approach (i.e.,
the comparison of pertinent endpoints in WT and CYP2El-null animals) may be useful in
helping to better understand the physiologic basis of the apparent susceptibility of early-life
stages to the carcinogenicity of AN, which was demonstrated in the rat study by Maltoni et al.
(1988). Subchronic Subchronic
4.8.2. Possible Geriatric Susceptibility
Age-related reductions in antioxidant or glutathione pool sizes may increase
susceptibility of elderly people to the tissue-damaging actions of AN and reactive metabolites,
but specific studies of the possible increased susceptibility of aged people or rats to AN are not
available. A 35% decrease in glutathione levels in the liver of aged F344 rats compared with
younger animals was associated with an age-related decrease in the levels and activity of
y-glutamylcysteine ligase, a key enzyme in the synthesis of glutathione (Suh et al., 2004). In
Wistar rats, the liver and kidney of 22-month-old rats showed significant decreases, compared
with 10-week-old rats, in glutathione and glutathione peroxidase and increased levels of
biomarkers of lipid peroxidation (Martin et al., 2003). In another study with F344 rats (Tian et
al., 1998), activities of several antioxidant enzymes in several tissues displayed an age-dependent
decline. Enzymatic activities showing significant decline with age included SOD in the heart,
kidney, and serum; glutathione peroxidase in the serum and kidney; and catalase activities in the
brain, liver, and kidney. These changes indicated a lower resistance to oxidative stress in older
animals.
The possible toxicological impact of an age-related decline in glutathione could be offset
by an age-related decline in CYP2E1-mediated metabolism of AN leading to reactive
metabolites (CEO), cyanide, or ROS. Many studies have reported that hepatic enzymatic
activities of CYP2E1 are lower in elderly human subjects (>65 years) compared with younger
adults (see Tanaka, 1998, for review). Decreased hepatic CYP2E1 enzyme activities have also
been reported in aged rats. For example, hepatic CYP2E1 enzyme activities were decreased by
46% in 18-month-old rats compared with 8-month-old rats, whereas no age-related changes in
CYP2E1 mRNA or protein content were evident (Wauthier et al., 2004). Such age-related
declines in CYP2E1 enzyme activities may lead to lower tissue levels of reactive metabolites or
cyanide but also may lead to elevated levels or increased residence time of AN in aged tissues
compared with younger tissues. No definitive conclusions about the possible susceptibility of
297
DRAFT - DO NOT CITE OR QUOTE
-------
the elderly to AN toxicity can be drawn without more specific studies designed to examine the
effects of age on susceptibility to AN toxicity.
As studied in a small number of individuals (n = 47), long-term occupational exposure to
AN increased the deletion rate in mitochondrial DNA to a level equivalent to that seen in a group
of elderly nonexposed subjects (n = 12) (Ding et al., 2003) (see Section 4.1.2.2). The study
authors suggested that AN may have an effect on the molecular processes of aging.
4.8.3. Possible Gender Differences
No reports of gender differences in susceptibility of humans to AN toxicity are available.
No consistent gender-related differences in carcinogenic responses were observed in rats or
mice, following chronic oral exposure to AN, or in rats, following chronic inhalation exposure.
With oral exposure, male and female groups showed similarly increased incidences of brain
astrocytomas (rats: Quast, 2002; Bigner et al., 1986; Biodynamics, 1980a, b, c; Quast et al.,
1980a), forestomach tumors (rats: Quast, 2002; Bigner et al., 1986; Biodynamics, 1980a, b, c;
Quast et al., 1980a; mice: NTP, 2001), Zymbal's gland tumors (rats: Quast, 2002; Bigner et al.,
1986; Biodynamics, 1980a, b, c; Quast et al., 1980a), and Harderian gland tumors (mice: NTP,
2001). Following inhalation exposure to 80 ppm AN for 2 years, male and female exposed
groups of rats showed similarly increased incidences of brain/CNS tumors and Zymbal's gland
tumors compared with the respective control groups (Quast et al., 1980b).
No marked gender-related differences in noncarcinogenic responses were observed in the
rat chronic oral toxicity study reported by Quast (2002). Groups of male and female rats were
exposed to AN in drinking water at concentrations of 0, 35, 100, or 300 ppm. At the 1-year
interim sacrifice, the only exposure-related noncancer histopathologic finding was an increase in
the incidence of forestomach squamous cell hyperplasia in rats exposed to concentrations of
100 ppm (4/10 males, 7/10 females) or 300 ppm (10/10 males, 9/10 females). At the 2-year
sacrifice, incidences of stomach lesions (nonglandular hyperplasia and/or hyperkeratosis) were
similar in male (15/80, 15/47, 44/48, and 45/80 for the control through high-concentration
groups, respectively) and female (20/80, 23/48, 41/48, and 47/48) rats at the same exposure level
(Quast, 2002). These data, however, give some indication that the forestomach epithelium of
female rats may have been slightly more susceptible than that of male rats: at the 1-year
sacrifice, 7/10 100 ppm females (70%) had lesions compared with 4/10 100 ppm males (40%); at
2 years, 23/48 35 ppm females (48%) had lesions compared with 15/47 35 ppm males (32%).
In contrast to the lack of apparent gender differences in carcinogenic or noncarcinogenic
responses in rodents with chronic exposure, male mice appear to be more susceptible to the acute
oral toxicity of AN than female mice (Chanas et al., 2003; NTP, 2001).
In a 14-week study, groups of 10 male and 10 female mice were given 0, 5, 10, 20, 40, or
60 mg/kg AN in deionized water by gavage 5 days/week (NTP, 2001). Exposure-related
mortalities were restricted to the first week of the study and occurred in the 40- and 60-mg/kg
298
DRAFT - DO NOT CITE OR QUOTE
-------
groups. Both male and female groups showed mortality, but only at 40 mg/kg was the incidence
of deaths higher in male mice compared with females (9/10 and 10/10, males, and 3/10 and
10/10, females, at 40 and 60 mg/kg, respectively).
In a subsequent study, groups of three to four male and three to four female WT or
CYP2El-null mice (mixed 129/Sv and C57BL) were given single gavage doses of 0, 2.5, 10, 20,
or 40 mg/kg in tap water and sacrificed 1 or 3 hours later (Chanas et al., 2003). All male WT
mice exposed to 40 mg/kg died within 3 hours, showing gross signs typical of cyanide poisoning
(rapid shallow breathing, cyanosis, trembling, and convulsions). In contrast, exposed female WT
mice showed milder gross signs of poisoning, and none died within the 3-hour period. One hour
after dose administration, concentrations of cyanide in blood were statistically significantly
elevated, compared with vehicle controls, in WT male mice given doses >2.5 mg/kg. At doses
>10 mg/kg, female WT mice also showed elevated cyanide levels in blood, compared with
controls, but cyanide levels were <50% that in male WT mice. Exposed CYP2El-null mice of
both genders showed no elevation in blood cyanide concentrations compared with vehicle
controls. Cyanide levels in brain and kidney tissues were also higher in WT males compared
with females; cyanide levels in liver and lung tissues showed less distinct differences between
male and female mice.
Expression of hepatic, renal, and pulmonary CYP2E1, soluble EH, and microsomal EH
were measured in male and female WT mice using Western blot analysis (Chanas et al., 2003).
In the liver, WT males showed greater expression of EH, both soluble and microsomal, than did
females; CYP2E1 levels were similar in males and females. In the kidney, male WT mice
showed markedly higher levels of CYP2E1 (about fourfold), moderately higher levels of soluble
EH (about twofold), and comparable levels of microsomal EH compared with female mice.
Higher CYP2E1 and soluble EH in the kidney of male mice provided explanation for higher
blood cyanide levels in the kidney and acute lethality in male mice. In the lung, no gender
differences in the expression of these enzymes were apparent. The results indicated that male
mice were more susceptible than female mice to the acute toxicity and lethality of AN and that
this difference was associated with higher blood, kidney, and brain levels of cyanide in males
shortly after dose administration. Chanas et al. (2003) noted that another possible explanation
was that female mice had greater detoxification capability of converting cyanide to thiocyanate,
as demonstrated by excretion of greater amount of thiocyanate in urine than males after repeated
administration of equal doses of AN (NTP, 2001).
In a subsequent study, male mice (mixed 129/Sv and C57BL) showed higher blood levels
of cyanide than female mice 1 hour after gavage administration of 0.047, 0.095, 0.19, or
0.38 mmol/kg (2.5, 5, 10, or 20 mg/kg) AN (El Hadri et al., 2005). This difference between
genders was evident in WT mice and in microsomal EH-null (mEH-null) mice, but blood levels
of cyanide were lower in exposed mEH-null mice, compared with comparably exposed WT
mice. As in the previous experiments reported by Chanas et al. (2003), exposure to AN induced
299
DRAFT - DO NOT CITE OR QUOTE
-------
no elevation of blood levels of cyanide in CYP2El-null mice of either gender. Western blot
analysis revealed no gender differences in expression of CYP2E1 in the liver of WT or mEH-
null mice or in expression of mEH in WT or CYP2El-null mice. However, expression of
soluble EH in the liver was greater in WT males, compared with females. No gender differences
in expression of this enzyme was observed in mEH-null mice or in CYP2El-null mice.
The effect of gender on the expression of CYP2E1 and EH have been investigated in
humans and rats. Gender was reported to have no influence on the level of CYP2E1 in human
liver (George et al., 1995). Sex-related patterns in the activity of hepatic EH activity was studied
in Sprague-Dawley rats (Chengelis, 1988). At week 4, epoxide activity in both male and female
rats was equivalent (3.5-4.0 nmol/minute per mg protein, or 85-100 nmol/minute per g liver).
However, there were consistent increases in activity in males from week 4 to 78, while activity in
females actually decreased, but returned to week 4 levels during the later stages (week 78-103).
EH activity in male liver peaked at week 78 (about 10 nmol/minute per mg protein, or
350 nmol/minute per g liver). There was a sharp decline in EH activity in aged male rats to
about 6 nmol/minute per mg protein, or 200 nmol/minute per g liver at 104 weeks. Cornet et al.
(1994) studied gender-related changes in microsomal and cytosolic EH activity in male and
female Brown Norway rats. At 15 week, microsomal EH activity was about the same for male
and female rats at 4.5 nmol/mg protein/minute. The microsomal EH activity decreased strongly
as a function of age in female rats, and in the 125-week-old females, the activity was only half of
that found in 15-week-old rats. However, there was no age-related change in males, although the
activity in the 83- and 125-week age groups was significantly lower than that in 28-week-old
males. The activity of EH activity was higher in males than in females in these aged animals. In
another study that compared hepatic microsomal EH in different strains of adult rats (170-250 g)
(Oesch et al., 1983), EH activities in females were found to be 71-88% of those in males in all
strains. Denlinger and Vesell (1989) studied the hormonal regulation on the developmental
pattern of EH in F344 rats, and found that EH activities in males increased gradually until
puberty, when activities in males rose rapidly to be from 1.5- to twofold higher than those in
females. The higher activity in males was not seen if the males were castrated 24 hours after
birth. When castrated males and females were injected with testosterone propionate (0.5 mg s.c.)
on days 1, 3, and 5 postpartum, increased mEH an cEH activities were observed at adulthood.
Thus, Denlinger and Vesell (1989) concluded that full adult expression of EH activities depends
on hormonal influences exerted neonatally.
Gender-related differences were also observed in cytosolic EH activity of Brown Norway
rats (Cornet et al., 1994). Significant gender-related differences in the cytosolic EH activity
were found in 15-, 28-, and 83-week-old rats with the male animals showing the highest values.
In the oldest animals and in 56-week-old rats comparable cytosolic EH activities were found in
both genders.
300
DRAFT - DO NOT CITE OR QUOTE
-------
In summary, the available data from animal studies provide no evidence of consistent or
marked gender differences in susceptibility to noncancer or cancer-related effects from chronic
exposure to AN. There is evidence that male mice are more susceptible than female mice to the
acute, cyanide-induced toxicity and lethality of AN, but whether or not this apparent gender
dimorphism extends to other species is uncertain at the present time.
4.8.4. Genetic Polymorphisms
4.8.4.1. CYP450
In humans, CYP2E1 exists in several modifications that differ in amino acid sequence
(alleles). Human variability in their susceptibility to the toxic effects of AN likely exists since
CYP2E1 activities may fluctuate between one person and another.
Microsomal CYP2E1 activities varied from 6- to 20-fold in human livers (Lucas et al.,
1993). Environmental factors, diet habits, and/or genetic factors may account for the observed
interindividual variations observed. In addition, CYP2E1 is elevated in obese overfed rats
(Salazar et al., 1988) and diabetic rats (Song et al., 1987), suggesting induction by increased
plasma levels of ketone bodies (Bellward et al., 1988). Moreover, CYP2E1 is elevated in
lymphocytes from poorly controlled insulin-dependent diabetics (Song et al., 1990). Thus, obese
and diabetic individuals may have elevated levels of CYP2E1.
Besides induction, polymorphism of the human CYP2E1 gene may have an impact on
AN metabolism in humans. Stephens et al. (1994) compared two restriction fragment length
polymorphic sites of the CYP2E1 gene (Rsa 1 and Dra 1), in 695 African-American, European-
American, and Taiwanese subjects. Rare alleles at these two loci have been associated with a
reduced risk for lung cancer in Japanese and Swedish populations. Stephens et al. (1994)
demonstrated that rare alleles (c2 and C) at the Rsa 1 and Dra 1 sites were at least twice as
frequent in Taiwanese populations (28 and 24%, respectively) compared with African-
Americans (1-8%) or European-Americans (4-11%), raising the possibility of differential
susceptibility to chemically induced cancers across ethnic groups. However, when Carriere et al.
(1996) measured the allele frequencies for Ras 1, Dra 1, and Taq 1 polymorphic sites in liver
CYP2E1 from kidney donors (n = 93) in Geneva, they failed to find a correlation between
frequencies of rare alleles and CYP2E1 activity. This implied that observed differences in
enzyme activity in humans were more likely to be the result of different levels of induction by
environmental factors or other genetic factors.
McCarver et al. (1998) identified a 100-bp insertion mutation in the regulatory region of
CYP2E1 gene. Associated with an elevated CYP2E1 metabolic activity, this insertion mutation
appears to be present only in obese people or persons who had recently consumed alcohol. The
incidence of this mutation was seen in 31% of 65 African-American samples but only in 6.9% of
58 Caucasian samples (McCarver et al., 1998). If tumor formation is determined by the activity
301
DRAFT - DO NOT CITE OR QUOTE
-------
of CYP2E1, this mutation might put affected individuals at a higher risk of cancer from
tumorigenic agents that are metabolized by CYP2E1.
Thier et al. (2002) were unable to detect any influence of six genetic CYP2E1
polymorphisms (G.1259C, A_3i6G, T.297A, G.35T, G4804A, and T7668A) on the formation of
N-(cyanoethyl)valine Hb adducts of AN. Conversely, in a study confined to a cohort of
individuals from ethnic minorities (African-Americans and Mexican-Americans), Wu et al.
(1998) observed an increased incidence of the CYP2E1 Dral DD genotype in peripheral WBC
DNA in 126 patients with untreated lung cancer compared with 193 unaffected controls. This
could imply an etiological association between the CYP2E1 Dral polymorphism and tumor
formation in the lung. Altered toxicokinetic characteristics of CYP2E1 in persons exposed to
AN might result in some persons being more vulnerable than others to the tumorigenic effects of
the compound, with concomitant changes to the rate and amount of formation of the toxic
metabolite and associated changes in susceptible individuals.
Kim et al. (1996), investigating the differences in CYP2E1 activities between
20 Caucasian and 20 Japanese men, pointed out that the significantly lower activity of CYP2E1
in Japanese men might account for the lower rate of some cancers in Japanese compared to
Caucasian men.
4.8.4.2. Glutathione S-transferases
Thier et al. (1999) studied the formation of Hb adducts of AN (N-[cyanoethyl]valine,
N-[methyl]valine, and N-[hydroxyethyl]valine) in a group of 59 people occupationally exposed
to the chemical. They reported their findings in relation to subjects' smoking habits and their
genetic status with respect to the GST isozymes GSTM1 and GSTT1. Included in the study was
an evaluation of smoking habits, since elevated adduct levels of AN in Hb have been reported in
smokers. There was no correlation between adduct levels and either the subjects' status of GST
isozymes or smoking habit. Thus, neither GSTM1 nor GSTT1 appears as a major AN-
metabolizing isoenzyme in humans. However, in a follow-up study with the same group of
59 workers, Thier et al. (2001) reported that polymorphism of the GSTP1 gene at codon 104 was
associated with a higher level of N-(cyanoethyl) valine adducts, while GSTM3 variants had no
effect on Hb adduct formation. Any potential effect arising from such polymorphisms on the
health risk to AN-exposed humans awaits to be evaluated. However, Zielinska et al. (2004)
reported an increase in the frequency of the GSTPlb/b genotype in children with cancer (OR =
5.7, CI = 2.4-13.8).
4.8.4.3. EH
Two major polymorphisms of mEH have been identified in human population (Hasset et
al., 1997). One is a polymorphism in exon 3 that changes the tyrosine residue 113 (Tyrl 13) to
histidine (His 113), and the other is an A—>G substitution in exon 4 that changes the histidine
302
DRAFT - DO NOT CITE OR QUOTE
-------
residue 139 (His 139) to arginine (Argl39). In vitro expression studies demonstrated that with
the Tyrl 13His polymorphism, the corresponding mEH enzymatic activity is decreased by about
40% (Hasset et al., 1997). On the other hand, the Hisl39Arg polymorphism results in increased
enzyme activity. Population studies have demonstrated that the low activity 113His allele
correlates with an increased risk for lung cancer (Benhamou et al., 1998), colon cancer (Harrison
et al., 1999), hepatocellular carcinoma developed after aflatoxin exposure (McGlynn et al.,
1995), and chronic obstructive pulmonary disease (Smith and Harrison, 1997).
The mEH genotypes was shown to play a significant role in human sensitivity to the
genotoxic effects of exposure to 1,3-butadiene (Abdel-Rahman et al., 2003, 2001). The
carcinogenic and mutagenic effects of 1,3-butadiene are thought to be due to its epoxide
metabolites, and the hydrolytic pathway involving mEH is the main detoxification pathway for
1,3-butadiene-reactive intermediates in humans (Jackson et al., 2000). In a study of
49 nonsmoking workers from two styrene-butadiene rubber plants, the hprt gene mutation assay
was used as a biomarker of genotoxic effect of BD (Abdel-Rahman et al., 2003, 2001). Abdel-
Rahman et al. (2003, 2001) evaluated the effect of polymorphisms in both exon 3 and exon 4 of
the mEH gene as modifiers of individual susceptibility to the mutagenic response associated with
exposure to 1,3-butadiene, and found a progressive increase in hprt mutant frequency with
declining mEH activity in the high exposure group (>150 ppb). The highest frequency of hprt
mutant lymphocytes occurred in the group with the mEH low-activity genotype. Individuals
with low mEH activity had three- and twofold increases in hprt mutant frequency compared to
individuals with high and intermediate mEH activity, respectively. In the low exposure group,
there was no difference in hprt mutant frequency between high-, intermediate-, and low-activity
individuals. Although there are no studies that evaluate the role played by mEH genotypes in
human sensitivity to the toxicity of AN, since EH is involved in the hydrolysis of CEO and its
elimination, polymorphisms in exon 3 and exon 4 of the mEH gene are likely to have a role in
human susceptibility.
303
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
As previously discussed in Section 4, no human studies currently exist that involve oral
exposures to AN. The primary route of AN exposure in humans is via inhalation. The available
animal oral toxicity studies, however, identify forestomach lesions (i.e., squamous cell epithelial
hyperplasia and hyperkeratosis) as the most sensitive, prevalent, and consistent noncancer effect
associated with chronic oral exposure to AN (see Table 4-56 and Figure 5-1). Although,
benchmark dose (BMD) modeling of dose-response data is preferred, several of the endpoints
under consideration (i.e., incidences of chronic nephropathy, ovarian cysts, and gliosis in the
brain in rodents) were not amenable to modeling because either there was no evidence of a dose
response or the response was not a monotonically increasing function of dose. A
NOAEL/LOAEL approach was used as a common basis for comparison. Using this approach,
forestomach lesions were selected as the critical effect on which to base the derivation of the
RfD. Although, anatomically, humans do not possess a forestomach, they do have comparable
squamous cell epithelial tissues in their oral cavity and in the upper two-thirds of their esophagus
(IARC, 1999). Moreover, the forestomach lesions observed in animals were not likely due to the
direct irritating effect of AN on gastric tissue. GI bleeding has been observed with single s.c. or
oral administration of AN in rats (Ghanayem and Ahmed, 1983) that likely resulted from the
distribution of AN metabolites from blood into the GI mucosa. Jacob and Ahmed (2003a) have
also demonstrated that AN or its metabolites accumulated and covalently interacted with the GI
mucosa of F344 rats treated either orally or intravenously with 2-[14C]-AN, supporting the theory
of metabolic incorporation and macromolecular interaction of AN or its metabolites with gastric
tissue.
Two 2-year drinking water studies, one in Sprague-Dawley rats (Quast, 2002; Quast et
al., 1980a) and the other in F344 rats (Johannsen and Levinskas, 2002b; Biodynamics 1980c),
and a 2-year gavage study in B6C3Fi mice (NTP, 2001) provided the best available dose-
response data on which to base the RfD (Table 5-1 for rats and Table 5-2 for mice). In Sprague-
Dawley rats, significantly elevated incidences of hyperplasia and hyperkeratosis in squamous
epithelium of the forestomach occurred in both males and females exposed to AN in drinking
water at the two highest concentrations administered (i.e., 100 and 300 ppm) with incidences
approaching 100% at the highest dose. Female Sprague-Dawley rats also exhibited statistically
significantly elevated incidences at the lowest concentration of AN administered (i.e., 35 ppm).
In F344 rats, both males and females exposed to 3, 10, and 30 ppm AN in drinking water
exhibited statistically significantly elevated incidences of forestomach lesions with incidences of
approximately 20-30%. Male and female F344 rats exposed to the lowest and highest
304
DRAFT - DO NOT CITE OR QUOTE
-------
concentrations of AN in drinking water (i.e., 1 and 100 ppm) did not show statistically
significantly elevated incidences of forestomach lesions.
305
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-1. Incidences of forestomach lesions (hyperplasia or
hyperkeratosis) in Sprague-Dawley and F344 rats exposed to AN in
drinking water for 2 years
Sex
Administered
concentration
(ppm in drinking
water)
Administered
dose3
(mg/kg-d)
Predicted internal dose metricsb
Incidence of
forestomach
lesions0
AN-AUC in rat
blood
(mg/L)
CEO-AUC in rat
blood
(mg/L)
Sprague-Dawley rats
(Sources: Quast, 2002; Quast et al., 1980a)
Male
0
0
0
0
15/80(19%)
35
3.4
2.06 x 10-2
1.83 x 10-3
15/47 (32%)
100
8.5
5.36 x 10-2
4.36 x 10-3
44/48 (92%)°
300
21.3
1.46 x 10"1
9.70 x 10-3
45/48 (94%)°
Female
0
0
0
0
20/80 (25%)
35
4.4
2.37 x 10-2
2.07 x 10-3
23/48 (48%)°
100
10.8
6.18 x 10-2
4.87 x 10-3
41/48 (85%)°
300
25.0
1.56 x 10"1
1.01 x 10-2
47/48 (98%)°
F344 ratsd
(Sources: Johannsen and Levinskas, 2002b; Biodynamics, 1980c)
Male
0
0
0
0
11/159(7%)
1
0.08
4.33 x 10-4
4.06 x 10-5
3/80 (4%)
3
0.25
1.35 x 10-3
1.27 x 10-4
18/75 (24%)°
10
0.83
4.52 x 10-3
4.19 x 10-4
13/80 (16%)°
30
2.48
1.37 x 10-2
1.23 x 10-3
17/80 (22%)°
100
8.37
4.85 x 10-2
3.97 x 10-3
9/77 (12%)
Female
0
0
0
0
4/156 (3%)
1
0.12
5.73 x 10-4
5.32 x 10-5
2/80 (3%)
3
0.36
1.72 x 10-3
1.59 x 10-4
16/80 (20%)°
10
1.25
6.02 x 10-3
5.49 x 10-4
23/74 (31 %)°
30
3.65
1.79 x 10-2
1.58 x 10-3
13/80 (16%)°
100
10.90
5.63 x 10-2
4.46 x 10-3
5/74 (7%)
aAdministered doses were averages calculated by the study authors based on animal BW and drinking water
intake.
bThe EPA-modified rat physiologically based pharmacokinetic (PBPK) model of Keddaris et al. (1996) was
employed to predict a rat internal dose (i.e., either AN-AUC or CEO-AUC concentration in blood, where AUC =
area under the curve) resulting from the ingestion of the specified administered dose of AN consumed in six bolus
episodes/d.
"Indicates significantly different (atp < 0.05) from control incidence by Fisher's exact test,
incidences for F344 rats do not include animals from the 6- and 12-mo sacrifices and were further adjusted to
exclude (from the denominators) rats that died between 0 and 12 mos in the study. Rats dying during this time
period were determined from page 6 of Appendix H and Table 1 in Biodynamics (1980c) and Table 8 in
Johannsen and Levinskas (2002b). Unscheduled deaths between 0 and 12 mos in the study occurred in two female
controls, two males at 3 ppm, three females at 10 ppm, and three males and three females at 100 ppm.
306
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-2. Incidences of forestomach lesions (hyperplasia or hyperkeratosis)
in male and female B6C3Fi mice administered AN via gavage for 2 years
Dose (mg/kg-d)a
Lesion site and type
0
2.5
10
20
Males
Forestomach hyperplasia or hyperkeratosis
2/50
4/50
10/50a
13/50b
(4%)
(8%)
(20%)
(26%)
Females
Forestomach hyperplasia or hyperkeratosis
2/50
2/50
5/50
8/50a
(4%)
(4%)
(10%)
(16%)
aSignificantly elevated above vehicle control as determined by EPA using Fisher's exact test (p < 0.05).
bSignificantly elevated above vehicle control as determined by EPA using Fisher's exact test (p < 0.01).
Source: NTP(2001).
In B6C3Fi mice, males exhibited statistically significantly elevated incidences of
forestomach lesions (hyperplasia or hyperkeratosis) at the two highest doses administered (i.e.,
10 and 20 mg/kg-day), while females showed statistically significantly elevated incidences of
forestomach lesions at the highest dose only.
While the Johannsen and Levinskas (2002b) drinking water study with F344 rats was
selected as the principal study on which to base the RfD, candidate RfDs were also developed
from studies in Sprague-Dawley rats (Quast, 2002) and B6C3Fi mice (NTP, 2001) for
comparison purposes. The Johannsen and Levinskas (2002b) study was selected as the principal
study primarily because it employed five doses of AN, ranging from 1 to 100 ppm, and thus
tested a more complete range of doses, especially in the low-dose region, than did either the
Quast (2002) study in Sprague-Dawley rats or the NTP (2001) study in B6C3Fi mice, which
both employed three doses (i.e., 35, 100, and 300 ppm in rats and 2.5, 10, and 20 mg/kg-day in
mice). In addition, Johannsen and Levinskas (2002b) and Quast (2002) are both drinking water
studies, which are preferred over the NTP (2001) gavage study in B6C3Fi mice because drinking
water exposure is more relevant to humans. Finally, of these three studies1, the Johannsen and
Levinskas (2002b) study identified the lowest LOAELs based on an increased incidence of
nonneoplastic stomach lesions (i.e., 0.3 mg/kg-day for males and 0.4 mg/kg-day for females).
Therefore, hyperplasia or hyperkeratosis in F344 rats was selected as critical effect for derivation
of RfD.
Although SD male rats in the Johannsen and Levinskas (2002a) study exhibited the lowest LOAEL (0.1 mg/kg-d)
for all endpoints across all studies considered for RfD derivation (see Figure 5-1), this study experienced significant
mortality, especially in the high-dose groups, rendering the results of this study highly suspect, and thus, it was
deemed not suitable for RfD derivation.
307
DRAFT - DO NOT CITE OR QUOTE
-------
5.1.2. Methods of Analysis—Including Use of PBTK Modeling
Candidate RfDs for AN were derived from the incidence of forestomach lesions in
Sprague-Dawley rats, F344 rats, and B6C3Fi mice using a BMD approach. Incidences of
forestomach lesions from chronic drinking water studies in male and female Sprague-Dawley
rats (Quast, 2002) and F344 rats (Johannsen and Levinskas, 2002b) provided four sets of dose-
response data from which to derive candidate RfDs (see Table 5-1), while incidences of
forestomach lesions from a chronic gavage study in male and female B6C3Fi mice provided an
additional two sets of dose-response data (see Table 5-2). The incidence data from rats in Table
5-1 were not combined across strains for purposes of dose-response modeling because response
levels were different across the two strains. As described further below, dose-response modeling
of the incidence of forestomach lesions after 2 years of AN exposure was carried out using BMD
modeling. For rats, in addition to administered dose, two alternative internal dose metrics (AN
in blood and CEO in blood), as estimated by PBTK modeling, were employed in BMD
modeling. In mice, only administered dose was employed as a dose metric for BMD modeling
purposes.
5.1.2.1. PBTK Modeling
In deriving candidate RfDs from the rat studies, internal dose metrics generated using
PBTK models developed by EPA (see Section 3.5; Appendix C) based on the rat and human
PBTK models of Kedderis et al. (1996) and Sweeney et al. (2003), respectively, were employed.
The EPA-modified PBTK models included many realistic features, each of them leading to a
different dose metric. The primary features of interest were estimated daily average internal
concentrations of AN or CEO in blood (area under the curve [AUC] expressed on a 24-hour
basis) and estimates based on continuous versus episodic exposure, with episodic exposure more
realistically reflecting how rats (and humans) actually consume drinking water.
As indicated above, two potentially useful chemical markers of internal exposure were
selected (i.e., the concentration in blood of the parent compound, AN, and its reactive metabolite,
CEO). These two internal dose metrics were evaluated under an episodic exposure pattern
because rats (and humans) consume drinking water in an episodic manner. In addition, for rats
and mice, the externally administered AN dose was also used for deriving candidate RfDs for
purposes of comparison with the candidate RfDs derived based on internal doses of AN and
CEO estimated from the PBTK model.
For this assessment, internal dose metrics were evaluated for use in cross-species
extrapolation from rats because of the following:
• The rat and human PBTK models incorporated species differences in physiological
processes influencing the disposition of AN and CEO and were developed with rat in
vivo toxicokinetic data, human in vitro metabolic data, and rat-to-human allometric
308
DRAFT - DO NOT CITE OR QUOTE
-------
scaling (Section 3.5; Appendix C; Sweeney et al., 2003; Kedderis et al., 1996). As such,
approaches using these models were expected to provide more accurate bases for
extrapolation of dose-response relationships from rats to humans than approaches based
on administered animal dose alone.
• For oral exposures, PBTK model predictions of AN or CEO concentrations in blood
appear to provide a more accurate basis for extrapolation than predictions based on
forestomach AN-AUC or forestomach CEO-AUC concentrations because the stomach
compartment in the PBTK model was not calibrated with measurements of AN or CEO in
the stomach epithelium, the site of the forestomach lesions. Thus, more research is
needed before a more physiologically meaningful stomach compartment can be
incorporated into the model. Blood AN concentrations predicted by the rat PBTK model
were fairly close to measured AN concentrations in rats following oral exposure, but the
model consistently predicted higher CEO concentrations in blood than was reported in
studies of orally exposed rats (see Figure 3 in Kedderis et al., 1996).
Given the relative uncertainties in the PBTK predictions noted above, especially for
CEO, both internal dose metrics (AN and CEO in blood) were used in deriving candidate RfDs
based on the rat data, in addition to administered dose. The concentration of AN administered in
the bioassay, in terms of mg/kg-day, was used as input into the EPA-modified rat PBTK model
of Kedderis et al. (1996) in order to predict a rat internal dose (either AN-AUC or CEO-AUC
concentration in blood) resulting from the ingestion of the total daily administered dose of AN
consumed in six bolus episodes per day. The resulting predicted AN-AUC or CEO-AUC
concentrations in rat blood are shown in Table 5-1 for male and female SD and F344 rats.
5.1.2.2. BMD Modeling
The incidences of forestomach lesions observed following 2 years of AN exposure in
male and female SD and F344 rats were modeled using AN and CEO in blood, expressed in
mg/L, as internal dose metrics. In addition, incidences of these same lesions were modeled in
male and female SD and F344 rats, as well as male and female B6C3Fi mice, employing
administered dose. For B6C3Fi mice, administered doses were multiplied by 5/7 to convert 5
day/week gavage exposures to 7 day/week continuous exposures. In all cases, all of the
dichotomous dose-response models available in EPA's BMD Software (BMDS, version 2.0)
(i.e., the gamma, logistic, log-logistic, probit, log-probit, multistage, Weibull, and quantal-linear
models) were fit to these incidence data. Because the incidence of forestomach lesions in male
and female F344 rats did not increase monotonically across all administered concentrations,
however, only incidence data from the three lowest concentrations (i.e., 0, 1, and 3 ppm) were
ultimately used in dose-response modeling. Although the incidence of forestomach lesions did
309
DRAFT - DO NOT CITE OR QUOTE
-------
increase from 20 to 31% at 3 and 10 ppm, respectively, in female F344 rats, this dose-response
relationship could still not be adequately described by any of the existing dichotomous dose-
response models in BMDS, and thus, the 10 ppm dose group was dropped in females prior to
modeling. For the same reason, the incidence data from the highest dose group in male Sprague-
Dawley rats needed to be dropped prior to BMD modeling.
In most cases, several models fit the data equally well (i.e., exhibited % goodness-of-fit
p values >0.1). Of those models exhibiting adequate fit, the selected model was the one with the
lowest Akaike's Information Criterion (AIC) value. BMDLio and BMDL05 estimates were then
derived from this selected model. If more than one model shared the lowest AIC, the mean
BMDLio and BMDL05 were calculated, as per the EPA's Benchmark Dose Technical Guidance
Document (U.S. EPA, 2000b). Appendix B-l provides additional details regarding the BMD
modeling results used in the derivation of the RfD.
Although BMDLio values were derived for comparison purposes, a benchmark response
(BMR) of 5% extra risk was selected as the point of departure (POD) for deriving the RfD based
on both biological and statistical considerations. Biologically, the endpoint selected on which to
base the RfD derivation (i.e., forestomach lesions—hyperplasia or hyperkeratosis) is considered
severe. Although the severity of the hyperplasia observed was generally not reported, Johannsen
and Levinskas (2002b) concluded that, "This [forestomach] lesion has been described as
progressing from hyperplasia and hyperkeratosis to papilloma, and ultimately to carcinoma,
following chronic oral exposure...." Statistically, the dose at 5% extra risk identifies a POD
near the lower end of the observed data. Consequently, 5% extra risk appears to be a reasonable
choice for the POD for RfD derivation.
For the two internal dose metrics selected (AN and CEO in blood), once the BMDLio and
BMDL05 estimates (or their means) were derived from the selected model(s), these estimates,
expressed as rat internal AUCs (in mg/L), were then input into the EPA-modified human PBTK
model of Sweeney et al. (2003) in order to predict the human equivalent administered dose of
AN that would result in a human 24-hour blood AN-AUC or CEO-AUC equivalent to the
corresponding rat AUC, again assuming six bolus ingestion episodes per day. The resulting
predicted 95% lower bounds on the human equivalent administered dose of AN, expressed in
mg/kg-day, represent potential PODs for deriving candidate RfDs. In the case of administered
dose, the BMDLio and BMDL05 estimates (or their means) are already expressed as human
equivalent doses (HEDs) in mg/kg-day, and are thus used directly as potential PODs for deriving
candidate RfDs. The BMDLio and BMDL05 estimates and their corresponding PODs across the
three dose metrics are presented in Tables 5-3, 5-4, and 5-5 for Sprague-Dawley rats, F344 rats,
and B6C3Fi mice, respectively.
310
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-3. Candidate RfDs based on BMD modeling of the incidence of forestomach lesions (hyperplasia or
hyperkeratosis) in male and female Sprague-Dawley rats exposed to AN in drinking water for 2 years
Sex
Endpoint
Dose metric
BMRa
BMDLb
PODc
(mg/kg-d)
UF
Candidate RFDd
(mg/kg-d)
Male
Forestomach
lesions
Administered
dose
(mg/kg-d)
5%
0.72 mg/kg-d
0.72
100
7.20 x 10-3
10%
1.27 mg/kg-d
1.27
100
1.27 x 10-2
Predicted AN in
blood
(mg/L)
5%
4.32 x 10-3 mg/L
1.27
30
4.23 x 10-2
10%
7.72 x 10-3 mg/L
2.17
30
7.23 x 10-2
Predicted CEO
in blood
(mg/L)
5%
3.90 x 10"4 mg/L
3.86 x 10-2
30
1.29 x 10-3
10%
6.82 x 10-4 mg/L
6.75 x 10-2
30
2.25 x 10-3
Female
Forestomach
lesions
Administered
dose
(mg/kg-d)
5%
0.64 mg/kg-d
0.64
100
6.39 x 10-3
10%
1.24 mg/kg-d
1.24
100
1.24 x 10-2
Predicted AN in
blood
(mg/L)
5%
1.76 x 10-3 mg/L
5.35 x 10-1
30
1.78 x 10-2
10%
3.62 x 10-3 mg/L
1.07
30
3.57 x 10-2
Predicted CEO
in blood
(mg/L)
5%
2.88 x 10-4 mg/L
2.84 x 10-2
30
9.48 x 10-4
10%
5.58 x 10-4 mg/L
5.51 x 10-2
30
1.84 x 10-3
aBMR refers to the 95% lower confidence limit on the administered or PBPK-predicted internal dose in the rat associated with a 5 or 10% extra risk for the incidence of
forestomach lesions.
bAll dichotomous models in EPA's BMDS (version 2.0) were fit to the incidence of forestomach lesions (hyperplasia or hyperkeratosis) in Sprague-Dawley rats using the
data presented in Table 5-1. For BMD modeling, three different dose metrics were employed: (1) administered animal dose expressed in mg/kg-d, (2) AN in blood
(predicted) expressed in mg/L, and (3) CEO in blood (predicted) expressed in mg/L. Adequate fit of a model was achieved if the %2 goodness-of-fit statistic yielded a p-
value >0.1. Of those models exhibiting adequate fit, the selected model was the model with the lowest AIC value. BMDLio and BMDL05 estimates were derived from
the selected model. If more than one model shared the lowest AIC, the mean BMDLio and BMDL05 were calculated and are presented, as per the EPA's Benchmark
Dose Technical Guidance Document (U.S. EPA, 2000b). Appendix B-l provides additional details regarding these BMD modeling results.
Tor administered dose, the POD is the BMDL05 or BMDL10 based on BMD modeling using administered animal dose as the dose metric. For the internal dose metrics
(AN and CEO in blood), the PODs are PBPK model-derived human equivalent administered doses of AN that would result in a human 24-hr blood AN-AUC or CEO-
AUC equivalent to the corresponding rat BMDL05 or BMDL10 values, assuming AN ingestion in six bolus episodes/d.
dRfD = POD/UF.
Sources: Quast (2002); Quast et al. (1980a).
311
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-4. Candidate RfDs based on BMD modeling of the incidence of forestomach lesions (hyperplasia or
hyperkeratosis) in male and female F344 rats exposed to AN in drinking water for 2 years
Sex
Endpoint
Dose metric
BMRa
BMDLb
PODc
(mg/kg-d)
UF
Candidate RFDd
(mg/kg-d)
Male
Forestomach
lesions
Administered
dose
(mg/kg-d)
5%
9.97 x 10-2 mg/kg-d
9.97 x 10-2
100
9.97 x 10-4
10%
0.151 mg/kg-d
0.151
100
1.51 x 10-3
Predicted AN in
blood
(mg/L)
5%
5.39 x 10-4 mg/L
1.66 x 101
30
5.55 x 10-3
10%
8.14 x 10"4 mg/L
2.50 x 10-1
30
8.35 x 10-3
Predicted CEO
in blood
(mg/L)
5%
5.06 x 10"5 mg/L
5.00 x 10"3
30
1.67 x 10-4
10%
7.65 x 10-5 mg/L
7.56 x 10"3
30
2.52 x 10-4
Female
Forestomach
lesions
Administered
dose
(mg/kg-d)
5%
0.117 mg/kg-d
0.117
100
1.17 x 10-3
10%
0.209 mg/kg-d
0.209
100
2.09 x 10-3
Predicted AN in
blood
(mg/L)
5%
5.58 x 10-4 mg/L
1.72 x 10-1
30
5.74 x 10-3
10%
9.97 x 10-4 mg/L
3.06 x 10-1
30
1.02 x 10-2
Predicted CEO
in blood
(mg/L)
5%
5.17 x 10"5 mg/L
5.11 x 10-3
30
1.70 x 10-4
10%
9.23 x 10-5 mg/L
9.12 x 10-3
30
3.04 x 10-4
aBMR refers to the 95% lower confidence limit on the administered or PBPK-predicted internal dose in the rat associated with a 5 or 10% extra risk for the incidence of
forestomach lesions.
bAll dichotomous models in EPA's BMDS (version 2.0) were fit to the incidence of forestomach lesions (hyperplasia or hyperkeratosis) in F344 rats using the data
presented in Table 5-1. For BMD modeling, three different dose metrics were employed: (1) administered animal dose expressed in mg/kg-d, (2) AN in blood
(predicted) expressed in mg/L, and (3) CEO in blood (predicted) expressed in mg/L. Adequate fit of a model was achieved if the %2 goodness-of-fit statistic yielded a
p-valuc >0.1. Of those models exhibiting adequate fit, the selected model was the model with the lowest AIC value. BMDLio and BMDL05 estimates were derived from
the selected model. If more than one model shared the lowest AIC, the mean BMDL10 and BMDL05 were calculated and are presented, as per the EPA's Benchmark
Dose Technical Guidance Document (U.S. EPA, 2000b). Appendix B-l provides additional details regarding these BMD modeling results.
Tor administered dose, the POD is the BMDL05 or BMDL10 based on BMD modeling using administered animal dose as the dose metric. For the internal dose metrics
(AN and CEO in blood), the PODs are PBPK model-derived human equivalent administered doses of AN that would result in a human 24-hr blood AN-AUC or CEO-
AUC equivalent to the corresponding rat BMDL05 or BMDL10 values, assuming AN ingestion in six bolus episodes/d.
dRfD = POD/UF.
Sources: Johannsen and Levinskas (2002b); Biodynamics (1980c).
312
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-5. Candidate RfDs based on BMD modeling of the incidence of forestomach lesions (hyperplasia or
hyperkeratosis) in male and female B6C3Fi mice exposed to AN via gavage for 2 years
Sex
Endpoint
Dose metric
BMRa
BMDLb
(mg/kg-d)
PODc
(mg/kg-d)
UF
Candidate RfDd
(mg/kg-d)
Male
Forestomach
lesions
Administered
dose
(mg/kg-d)
5%
1.43
1.43
100
1.43 x 10-2
10%
3.01
3.01
100
3.01 x 10-2
Female
Forestomach
lesions
Administered
dose
(mg/kg-d)
5%
3.02
3.02
100
3.02 x 10-2
10%
6.20
6.20
100
6.20 x 10-2
aBMR refers to the 95% lower confidence limit on the administered dose in the mouse associated with either a 5 or 10% extra risk for the incidence of forestomach
lesions.
bAll dichotomous models in EPA's BMDS (version 2.0) were fit to the incidence of forestomach lesions (hyperplasia or hyperkeratosis) in B6C3Fi mice using the data
presented in Table 5-2. For BMD modeling, administered animal dose, expressed in mg/kg-d, was employed. Adequate fit of a model was achieved if the %2 goodness-
of-fit statistic yielded a p-valuc >0.1. Of those models exhibiting adequate fit, the selected model was the model with the lowest AIC value. BMDLio and BMDL05
estimates were derived from the selected model. If more than one model shared the lowest AIC, the mean BMDLio and BMDL05 were calculated and are presented, as
per the EPA's Benchmark Dose Technical Guidance Document (U.S. EPA, 2000b). Appendix B-l provides additional details regarding these BMD modeling results.
°The POD is the BMDL05 or BMDLio based on BMD modeling using administered animal dose as the dose metric. No internal dose metrics were employed for mice
because of the absence of a PBPK model for this species.
dRfD = POD/UF.
Source: NTP(2001).
313
DRAFT - DO NOT CITE OR QUOTE
-------
5.1.3. RfD Derivation—Including Application of Uncertainty Factors (UFs)
In Tables 5-3 and 5-4, the candidate PODs based on the internal dose metrics (i.e., CEO-
AUC in blood and AN-AUC in blood) were divided by a composite UF of 30 to account for two
areas of uncertainty. An UF of 3 (i.e., 10°5) was used to account for uncertainty associated with
extrapolating from rats to humans. A factor of 3 was chosen instead of the default value of 10,
because toxicokinetic uncertainty was decreased by the application of the rat and human PBTK
models. The applied factor of 3 was selected to account for any remaining toxicokinetic
uncertainties in the dosimetric extrapolation, and possible differences in the response of rat and
human target tissues to AN or its metabolites (i.e., toxicodynamic differences). A default UF of
10 was then used to account for uncertainty associated with extrapolation to individuals or
groups within the human population who may be especially sensitive or susceptible to the
noncancer toxicity of AN.
The candidate PODs based on administered dose in Tables 5-3, 5-4, and 5-5 were divided
by a composite UF of 100 to again account for two areas of uncertainty. First, a UF of 10 was
used to account for uncertainty associated with extrapolating from rats to humans because no
PBTK model was employed, and thus, a default factor of 10 was needed to account for both
toxicokinetic and toxicodynamic uncertainties. Secondly, as above, a default UF of 10 was used
to account for uncertainty associated with extrapolation to individuals or groups within the
human population who may be especially sensitive or susceptible to the noncancer toxicity of
AN. Considerations in selecting individual UFs are discussed in greater detail below.
A default UF of 10 was used to account for uncertainty associated with extrapolation to
individuals or groups within the human population who may be especially sensitive or
susceptible to the noncancer toxicity of AN. As discussed in Section 4.8, available data on AN
are inadequate to reliably identify populations that may be especially susceptible to the
noncancer toxicity of AN. There is one human case report indicating that children are more
susceptible to AN toxicity. These data at least suggest that juvenile populations may be
especially susceptible to the noncancerous effects of AN exposure.
An UF to account for extrapolation from subchronic to chronic exposure was not
necessary because of the availability of chronic oral studies in both rats and mice.
An UF to account for LOAEL to NOAEL extrapolation was not applied because the
BMR selected for BMD modeling, a 5% extra risk of forestomach lesions, was judged to have
minimal biological significance.
A database deficiency UF was not used in the development of the RfD. Although studies
of health effects in humans exposed to AN by the oral route are not available, the animal oral
toxicity database is particularly robust. As discussed in Section 4.6.1, there are nine rat toxicity
and cancer bioassays, one toxicity and cancer bioassay with B6C3Fi mice, a three-generation
(46-week) developmental/reproductive rat toxicity study with Sprague-Dawley rats, a 12-week
gavage study of nerve conduction velocities in male Sprague-Dawley rats, a 14-week gavage
314
DRAFT - DO NOT CITE OR QUOTE
-------
toxicity bioassay in B6C3Fi mice, a developmental toxicity study in Sprague-Dawley rats
exposed by gavage during GDs 6-15, and a 90-day study of oxidative stress indicators in the
brain and liver of F344 rats. These animal data identify forestomach lesions as the most
sensitive, prevalent, and consistent noncancer effect in animals associated with repeated oral
exposure to AN, and thus provide evidence that an RfD based on this effect will be protective for
other potential noncancer effects, including neurological, reproductive, and developmental
effects. In addition, the data for forestomach lesions in rats and mice provide an adequate
characterization of the dose-response relationship for the development of these lesions with
chronic exposure, which, coupled with the application of the PBTK models for cross-species
dosimetric extrapolation, adds to the confidence in the estimation of a chronic oral human dose
expected to be without risk in the general population for the development of gastric lesions.
For the purpose of this assessment, CEO-AUC in blood is considered to be the most
appropriate dose metric to use in deriving an RfD. CEO is believed to be the reactive metabolite
most likely responsible for the noncancer (and cancer) effects observed following AN exposure.
Results from acute exposure studies in rats indicate that CEO plays a key role in the mode of
action by which AN elicits stomach lesions. Research has shown that pretreatment with
inhibitors of CYP2E1 inhibited the development of GI ulceration, following acute exposure of
rats to AN, and doses of KCN equivalent to toxic doses of AN did not induce GI bleeding in a
similar manner to AN (Ghanayem et al., 1985; Ghanayem and Ahmed, 1983). For oral
exposures, then, the use of CEO-AUC as the internal dose metric for cross-species extrapolation
is recommended. Therefore, while the PBTK model does not predict CEO in blood as accurately
as it predicts AN in blood, CEO in blood provides the most biologically relevant dose metric for
modeling the incidence of forestomach lesions and for extrapolating from orally exposed rats to
humans.
In comparing the candidate RfDs for the preferred dose metric of predicted CEO in blood
across Tables 5-3 and 5-4, the candidate RfDs based on the forestomach lesion incidence data in
F344 rats are about an order of magnitude lower than those candidate RfDs based on
forestomach lesion incidence data in Sprague-Dawley rats. This comparison confirms what had
previously been predicted (i.e., F344 rats are more sensitive to AN exposure than Sprague-
Dawley rats). Moreover, these same candidate RfDs in F344 rats are approximately two orders
of magnitude lower than the candidate RfDs based on administered dose in B6C3Fi mice.
Therefore, for the critical endpoint (i.e., forestomach lesions), F344 rats are the most sensitive
species and strain. Consequently, the RfD is 2 x 10"4 mg/kg-day. This RfD is consistent with
the candidate RfDs derived for both male (1.67 x 10"4 mg/kg-day) and female (1.70 x 10"4
mg/kg-day) F344 rats using predicted CEO in blood as the dose metric and based on the
incidence of forestomach lesions following chronic oral exposure to AN.
5.1.4. Oral Data Array for Noncancer Endpoints
315
DRAFT - DO NOT CITE OR QUOTE
-------
LOAELs based on selected animal studies presented in Table 4-56 are arrayed for
comparison in Figure 5-1, and provide a perspective on the RfD developed in the previous
section from data in F344 rats (Johannsen and Levinskas, 2002b). Figure 5-1 should be
interpreted with caution, however, because the LOAELs across studies are not necessarily
comparable due to the lack of any indication regarding the confidence in the data sets from
which the LOAELs were derived. In addition, the nature, severity, and incidence of effects at a
LOAEL are also likely to vary. For example, the incidence of forestomach squamous cell
hyperplasia in male and female F344 rats at the LOAEL were both 17% (see Figure 5-1), while
in Sprague-Dawley rats, the incidence of the same lesions at the LOAEL in males and females
were 73 and 23%, respectively. Figure 5-1 provides a graphical comparison of organ-specific
LOAELs from chronic studies in experimental animals. The predominant noncancer effect of
chronic oral exposure to AN is hyperplasia and hyperkeratosis of squamous cell epithelial tissue
in the forestomach. The LOAELs for this endpoint are lower than those observed for chronic
nephropathy, gliosis, ovarian cysts, or developmental effects. Therefore, the RfD based on
hyperplasia and hyperkeratosis of squamous cell epithelial tissue in the forestomach should be
protective of other effects resulting from oral exposure to AN.
316
DRAFT - DO NOT CITE OR QUOTE
-------
Figure 5-1. Comparison of LOAELs for noncancer effects across target organs following oral exposure to AN in animals.
20
16
14
a>
(A
0 _ 12
a >
= "?
1 jj? 10
o o>
t- ¦
<
1 1 1 1
[ ] = Percent of animals with effects at
the LOAEL (corrected for background)
J1Q%L
¦
14.3
~
¦
[5%]
[31%]
¦
[73%]
10.7
~
10.8
~
¦
8.5
~
[10%]
7 A
¦
£23%]
~
[_14%]
¦
4.4
~
[14%]
3.4
4.4
~
17%1
J17%]
MQfl/J
~
r 16%i
1.8
0.3
~
0.4
~
0.1
~
Squamous
cell
hyperplasia
in F344 male
rats
[Johannsen
& Levinskas,
2002b]
Squamous
cell
hyperplasia
in F344
female rats
[Johannsen
& Levinskas,
2002b]
Squamous
cell
hyperplasia
in SD male
rats [Quast,
2002]
Squamous
cell
hyperplasia
in SD female
rats [Quast,
2002]
Squamous
cell
hyperplasia
in SD male
rats
[Johannsen
& Levinskas,
2002a]
Squamous
cell
hyperplasia
in SD female
rats
[Johannsen
& Levinskas,
2002a]
Squamous
cell
hyperplasia
in B6C3F1
male mice
[NTP, 2001]
Squamous
cell
hyperplasia
in B6C3F1
female mice
[NTP, 2001]
Chronic
nephropathy
(severe) in
SD male rats
[Quast, 2002]
Chronic
nephropathy
(minimal) in
SD female
rats [Quast,
2002]
Ovarian
cysts in
B6C3F1
female mice
[NTP, 2001]
Gliosis in the
brain of SD
female rats
[Quast, 2002]
Target Organ Effects
317
DRAFT - DO NOT CITE OR QUOTE
-------
5.1.5. Previous RfD Assessment
No RfD was derived in the previous IRIS assessment of AN.
5.2. INHALATION RfC
5.2.1. Choice of Principal Study and Critical Effect
As discussed in Section 4.1.2, results from several cross-sectional health examinations
and surveys of AN-exposed workers (Chen et al., 2000; Kaneko and Omae, 1992; Muto et al.,
1992; Sakurai et al., 1978) and a study of the performance on a battery of neurobehavioral tests
by exposed workers (Lu et al., 2005a) identified increased prevalence of subjective neurological
symptoms (e.g., headache, poor memory, and irritability) and small performance deficits in
neurobehavioral tests as the critical effects from chronic occupational inhalation exposure to AN.
An increased prevalence of subjective symptoms was associated with average workplace air
concentrations of 1.13 ppm (Muto et al., 1992), 1.8 ppm (Kaneko and Omae, 1992), and
0.48 ppm (Chen et al., 2000). Deficits in the neurobehavioral tests were associated with average
workplace air concentrations of 0.11 ppm for a group of workers designated as monomer
workers and 0.91 ppm for a group of workers designated as fiber workers (Lu et al., 2005a).
Cross-sectional epidemiologic surveys of reproductive outcomes in AN-exposed workers
found increased prevalence of adverse reproductive outcomes associated with somewhat higher
average workplace air concentrations of 3.7 ppm (Dong and Pan, 1995), 3.6 ppm (Dong et al.,
2000a), and 7.5 ppm (Li, 2000). Adverse outcomes with statistically significantly increased
prevalence compared with unexposed workers included the following:
• Premature deliveries—10.7% in unexposed wives of exposed males vs. 3.5% in
controls (implying the sperm of exposed males may be affected) (Dong and Pan,
1995); 8.2%) in exposed females vs. 3.9%> in controls (Dong et al., 2000b); and 11.6%
in exposed females vs. 4.7% in controls (Li, 2000)
• Stillbirths—4.5% in exposed females vs. 0% in controls (Dong and Pan, 1995); 2.7%
in exposed females vs. 0% in controls (Dong et al., 2000b)
• Sterility—5.0% in wives of exposed males vs. 1.8% in wives of controls (Dong and
Pan, 1995); 2.6% in exposed females vs. 0.8% in controls (Li, 2000)
• Birth defects—21.3% in exposed females vs. 4.8% in controls (Dong et al., 2000b);
25.4% in exposed females vs. 4.2% in controls (Li, 2000)
• Pregnancy complications—20.8% in exposed females vs. 7.1% in controls (Li, 2000)
318
DRAFT - DO NOT CITE OR QUOTE
-------
Exposure levels associated with adverse effects in the available animal inhalation toxicity
studies were higher than the workplace air concentrations associated with adverse effects in
AN-exposed workers (see Table 4-57). With repeated inhalation exposure to AN, effects noted
with the lowest exposure levels were as follows:
• Increased incidence of nasal epithelial lesions in Sprague-Dawley rats exposed for
2 years to 20 ppm AN (Quast et al., 1980b)
• Deficits in sensory nerve conduction in the tail nerve of Sprague-Dawley rats exposed for
12 weeks to AN concentrations >50 ppm, but not 25 ppm (Gagnaire et al., 1998)
• Increased incidence of litters with short tail, short trunk, missing vertebrae, or anteriorly
displaced ovaries (combined) in Sprague-Dawley rats exposed to AN concentrations of
80 ppm, but not 40 ppm, on GDs 6-15 (Murray et al., 1978)
• Decreased fetal weight gain per litter in Sprague-Dawley rats exposed to AN
concentrations >25 ppm, but not to 12 ppm, on GDs 6-20 (Saillenfait et al., 1993)
The cross-sectional study of neurobehavioral performance in acrylic fiber workers by Lu
et al. (2005a) was selected as the principal study for deriving the RfC because it is the best
available study that identified neurobehavioral effects in workers exposed to AN. Previous
occupational studies by Kaneko and Omae (1992) and Muto et al. (1992) reported subjective
neurological symptoms (e.g., poor memory and irritability) in exposed workers. Lu et al.
(2005a) utilized the WHO-recommended NCTB administered by trained physicians to evaluate
these neurobehavioral effects systematically. Hence, the results were more reliable when
compared with those based on self reporting. In addition, neurobehavioral effects were also
reported in AN-treated rats (Rongzhu et al., 2007; Ghanayem et al. 1991). Confounding by other
workplace exposures is not considered likely.
The distribution of employment duration for the monomer workers was 1-10 years, 23%;
11-20 years, 42%; and >20 years, 35%. For fiber workers, the distribution was 1-10 years,
47%; 11-20 years, 23%; and >20 years, 30%. Geometric mean workplace AN air concentrations
were 0.11 ppm for the monomer operations areas (range 0-1.70 ppm based on 390 stationary air
samples collected between 1997 and 1999) and 0.91 ppm for the acrylic fiber operations areas
(range 0.00-8.34 ppm based on 570 samples). For monomer workers, the following statistically
significant deficits, compared with unexposed controls, were measured:
• 41-68%) higher scores for negative moods (i.e., anger, confusion, depression, fatigue, and
tension) in the Profile of Mood States Test.
319
DRAFT - DO NOT CITE OR QUOTE
-------
• 16% longer times in the Simple Reaction Time Test of attention and visual response
speed.
• 21% lower scores in the backward sequence of the Digit Span Test of auditory memory.
• 4% lower scores in the Benton Visual Retention Test, a measure of visual perception and
memory.
• 14% lower scores in the Pursuit Aiming II Test, a measure of motor steadiness.
Statistically significant deficits were not found in the Santa Ana Test for manual dexterity
or in the Digital Symbol Test for perceptual motor speed. Fiber workers showed deficits of a
similar magnitude in many of the same tests (20-44% higher in Profile of Mood States Test,
10%) longer in the Simple Reaction Time Test, 24% lower score in the backward sequence of the
Digit Span Test, 4% lower scores in the Benton Visual Retention Test, and 10% lower scores in
the Pursuit Aiming II Test), with the exception that scores in the forward sequence of the Digit
Span Test were significantly better than those of unexposed workers.
Lu et al. (2005a) reported that air in these workplaces also presented potential exposures
to cyanide in the monomer operations areas and methyl methacrylate in the fiber areas, but
measurements of these chemicals in air samples were not made. Cyanide is a breakdown product
of AN. Its concentration in the monomer plant should be low compared with AN. Methyl
methacrylate was generally used as a minor component in the production of acrylic fiber. With
an RfC of 0.7 mg/m3 (U.S. EPA, 1998b), its toxicity is much lower than that for AN. Therefore,
potential exposure to these confounders was determined to be an insufficient limitation to
preclude using Lu et al. (2005a) to identify neurological effects as potential health hazards from
occupational exposure to AN and derive an RfC for chronic inhalation exposure to AN. Both
groups of workers showed deficits, despite being exposed to different potentially confounding
chemicals, and the results from this study are consistent with the increased prevalence of
subjective symptoms in other studies of AN-exposed workers.
An RfC based on the results from the chronic inhalation bioassay with Sprague-Dawley
rats (Quast et al., 1980b) was also derived for comparison purposes. As discussed in
Section 4.6.2, statistically significant increased incidence of inflammatory and degenerative
nasal lesions (i.e., hyperplasia of mucus-secreting cells in males and flattening of respiratory
epithelium in females) occurred in rats exposed to the lowest level of AN in this two-year
bioassay, 20 ppm (6 hours/day, 5 days/week), and represent the critical effects in animals
exposed to AN chronically by inhalation. At the higher exposure level, 80 ppm, other nasal
lesions with elevated incidences were suppurative rhinitis and focal erosion of the mucous lining
320
DRAFT - DO NOT CITE OR QUOTE
-------
in females and hyperplasia of respiratory epithelium in males and females. Other lesions with
elevated incidences at the 80 ppm exposure level were gliosis and perivascular cuffing in the
brain of males and females, focal nephrosis and thyroid cysts in males, and hepatic necrosis in
females (which was also elevated at 20 ppm).
5.2.2. Methods of Analysis
A NOAEL/LOAEL approach to the human data was used to derive the RfC. The
performance deficits measured in monomer and fiber workers were judged to be adverse and the
geometric mean of the range of air concentrations measured for the monomer work areas,
0.11 ppm (0.24 mg/m ), was selected as the POD for deriving the RfC.
For the animal data, a BMD approach was used. Dose-response models available in
EPA's BMDS (version 1.3.2) were fit to incidence data for flattening of the respiratory
epithelium in female rats and hyperplasia of mucus-secreting cells in male rats. Prior to
modeling, animal exposure data were converted to human equivalent concentrations (HECs)
using U.S. EPA (1994) methods for extrathoracic respiratory effects from a category 1 gas
(Table 5-6). A BMR of 10% extra risk was selected as the POD for deriving the RfC based on
both biological and statistical considerations. Biologically, the endpoints selected on which to
derive the RfC (i.e., flattening of the respiratory epithelium in female rats and hyperplasia of
mucus-secreting cells in male rats) are relatively benign. Statistically, the dose at 10% extra risk
identifies a POD near the lower end of the data range. Thus, a BMR of 10% extra risk was
selected for this analysis consistent with the U.S. EPA (. Benchmark concentrations (BMCs) and
the 95% lower bounds on the BMCs (BMCLios) for the best-fitting models are shown in Table 5-
6. Potential PODs for the animal-based RfC are the BMCLios of 0.082 and 0.059 mg/m3 for
nasal effects in male and female rats, respectively. More detailed information on these BMD
modeling results is presented in Appendix B-2.
321
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-6. Results of dose-response analyses of incidence data for selected
nasal lesions in male and female Sprague-Dawley rats exposed by inhalation
to AN for 2 years
Nasal lesion
Administered
concentration
(mg/m3)
HFC
(mg/m3)a
Lesion
incidence
BMC10
(mg/m3)b
BMCL10
(mg/m3)b
Candidate
RfC
(mg/m3)c
Hyperplasia of mucus-
secreting cells in males
0
0
0/11 (0%)
0.187
0.082
3 x 10-3
43.4
2.1
7/12 (58%)d
173.6
8.5
8/10 (80%)d
Flattening of respiratory
epithelium in females
0
0
1/11 (9.1%)
0.162
0.059
2 x 10-3
43.4
2.1
7/10 (70%)d
173.6
8.5
8/10 (80%)d
aHEC as per U.S. EPA (1994) methods for a category 1 gas producing an upper respiratory effect.
Sample calculation: 43.4 mg/m3 x 6h/24h x 5d/7d x RGDRET = 2.1 mg/m3, where RGDRET = 0.275 = [VE/SAet] rat
^ [VE/SAet] human; VE = minute volume = 0.281 L/min rat, 13.8 L/min human; SAET = extrathoracic surface area =
15 cm2 rat, 200 cm2 human.
^MCio and BMCLio refer to the BMD model-predicted air concentration and its 95% lower confidence limit,
associated with a 10% extra risk for having nonneoplastic nasal lesions. BMCi0s and BMCL10s are estimated from
the best-fitting model among those fit to the data. More detailed information on the BMD modeling results is
presented in Appendix B-2.
°RfC = BMCL10/UF, where the UF is 30.
Statistically significantly different from control value as reported by Quast et al., 1980b.
Source: Quast et al. (1980b).
5.2.3. RfC Derivation—Including Application of UFs
The LOAEL of 0.11 ppm (0.24 mg/m3) from Lu et al. (2005a) for statistically significant
performance deficits in neurobehavioral tests of mood, attention and speed, auditory memory,
visual perception and memory, and motor steadiness in humans occupationally exposed to AN
via inhalation was used for derivation of the RfC. Since the LOAEL was from an occupational
study, the adjusted LOAEL for continuous exposure was obtained by multiplying the study
LOAEL by a factor of 0.36 (5 days/7 days x 10 m3/day ^ 20 m3/day). This adjusted LOAEL
"3
(LOAELadj) of 0.03 ppm (0.086 mg/m ) was divided by a composite UF of 100 to arrive at an
RfC of 9 x 10"4 mg/m3. This two-step calculation is illustrated below:
1. LOAELadj = 0.24 mg/m3 x 0.36 = 0.086 mg/m3
2. RfC = 0.086 mg/m3 100 = 0.00086 mg/m3 or 0.9 (J,g/m3
An UF of 10 was used for extrapolating from a LOAEL to a NOAEL. The RfC is
derived from an occupational study. A default UF of 10 was used to account for sensitive or
susceptible members of the general population, including children. A database UF was not
322
DRAFT - DO NOT CITE OR QUOTE
-------
applied since the database for AN is robust, and occupational exposure studies exist that
evaluated reproductive effects in AN-exposed workers.
3 3 3 3 3
Comparative animal-based RfCs of 3 x 10" mg/m (or 3 (J,g/m ) and 2x10" mg/m (or
3 3 3
2 (J,g/m ) were derived by dividing the BMDLi0s of 0.082 mg/m and 0.059 mg/m for nasal
lesions in male and female rats, respectively, by UFs of 30 (3 for extrapolating from rats to
humans using the default U.S. EPA [1994] dosimetric adjustment and 10 to protect sensitive
subpopulations). These candidate RfCs are displayed in Table 5-6. Extrapolating the results
from animal toxicity studies to derive an RfC has inherently greater uncertainty than using the
results from the cross-sectional studies of health effects in human workers; however, the human-
based and animal-based RfCs differ by about twofold. For this assessment, the human-based
4 3 3
RfC of 9 x 10" mg/m or 0.9 (J,g/m is the recommended value.
5.2.4. Inhalation Data Array for Noncancer Endpoints
LOAELs based on selected studies included in Table 4-57 are arrayed in Figure 5-2 and
provide perspective on the RfC derived from Lu et al. (2005a). Figure 5-2 should be interpreted
with caution because the LOAELs across studies are not necessarily comparable, due to inherent
limitations in NOAEL/LOAEL determination (U.S. EPA, 2000b), nor is the confidence in the
data sets from which the LOAELs were derived the same. The nature, severity, and incidence of
effects occurring at a LOAEL are likely to vary. The text in Section 5.2.1 should be consulted
for a more complete understanding of the issues associated with each data set and the rationale
for the selection of the critical effect and principal study used to derive the RfC. The most
sensitive endpoint is the neurobehavioral effects identified in Lu et al. (2005a) and this endpoint
provides the basis for the derivation of the RfC.
323
DRAFT - DO NOT CITE OR QUOTE
-------
10
7.5
~
3.7
3.6
~
~
1.8
~
1.1
+, . . . .
JQ.5
~
0.1
~
~
Subjective Subjective Subjective
Symptoms in Males Symptoms in Males Symptoms in Males
[Kaneko & Omae, [Muto et al., 1992] and Females
1992] [Chen et al., 2000]
Neurobehavioral
Deficits in Males
and Females
[Lu et al., 2005a]
Decrease in Sperm
Density and
Number and Sex
Chromosome
Adverse
Reproductive
Outcomes in Males
and Females
Aneuploidy in Males [Dong & Pan, 1995]
[Xu et al., 2003]
Adverse
Reproductive
Outcomes in Males
and Females
[Dong et al., 2000b]
Adverse
Reproductive
Outcomes in
Females
[Li et al., 2000]
Observed Effects
Figure 5-2. Comparison of LOAELs for noncancer effects in human workers following inhalation
exposure to AN.
324 DRAFT - DO NOT CITE OR QUOTE
-------
5.2.5. Previous Inhalation Assessment
3 3
Previously, EPA derived an RfC of 2 x 10" mg/m from the low exposure level (i.e.,
"3
20 ppm AN, duration-adjusted to a LOAELhec of 1.9 mg/m ) in the Quast et al. (1980b) animal
study, which was identified as a LOAEL for the onset of hyperplasia of the mucus-secreting
cells. The LOAELhec was adjusted with a combined UF of 1,000 to derive the RfC. This
composite UF was made up of 10 for intraspecies variability; 3 for interspecies variability, where
dosimetric adjustments had already been applied to account for part of this area of uncertainty;
3 for extrapolation from a minimally adverse LOAEL to a NOAEL; and 10 for database
deficiencies. The latter UF was applied because of the lack of an inhalation bioassay in a second
species and the absence of reproductive data by the inhalation route where an oral study existed
that showed reproductive effects.
New pertinent information, available since the previous RfC was developed, includes:
(1) several cross-sectional health examinations and surveys of subjective symptoms and
reproductive outcomes in AN-exposed workers; (2) a published cross-sectional study of
performance in a battery of neurobehavioral tests by AN-exposed workers (Lu et al., 2005a) (the
principal study for the current RfC); (3) toxicokinetic information and the development of PBTK
models for AN; and (4) two inhalation developmental toxicity studies in rats.
5.3. UNCERTAINTIES IN THE ORAL RfD AND INHALATION RfC
The following discussion identifies uncertainties associated with the RfD or RfC for AN.
As presented earlier in this chapter (Sections 5.1.2 and 5.1.3; and 5.2.2 and 5.2.3), the UF
approach, following EPA methodology for RfC and RfD development (U.S. EPA, 2002, 1994),
was applied to a POD. For the RfD, the POD was determined as BMDL05 of internal dose
estimated from rats and subsequently converted to a FLED. For the RfC, a LOAEL was derived
from an epidemiologic study for neurobehavioral effects. Factors accounting for uncertainties
associated with a number of steps in the analyses were adopted to account for extrapolating from
the POD, the starting point in the analysis, to a no-adverse-effect level (LOAEL to NOAEL)
given insufficient data in the principal study for BMD modeling, to a diverse population of
varying susceptibilities. These extrapolations are carried out with default approaches instead of
data on AN, given the paucity of experimental AN data to inform individual steps.
Selection ofprincipal study and critical effect for reference value determination
Hyperplasia and hyperkeratosis of squamous epithelium of the forestomach was selected
as the critical effect for RfD. This effect is the most prevalent, consistent, and most sensitive
effect in rats and mice. Since GI bleeding occurs after s.c. injection of AN in rats, this effect is
probably not due to local irritation on gastric tissues, but is likely due to binding of CEO to GI
mucosa. Although humans do not possess a forestomach, humans do have comparable
squamous cell epithelial tissues in their oral cavity and the upper two-thirds of their esophagus
(IARC, 1999). Thus, there is little uncertainty that this effect is relevant to humans.
325
DRAFT - DO NOT CITE OR QUOTE
-------
For derivation of the RfC, both a 2-year rat inhalation study and epidemiologic studies
are available. To reduce uncertainty in extrapolating from animals to humans, epidemiologic
studies are preferred. Several earlier cross-sectional health surveys of AN-exposed workers
reported subjective neurological symptoms. In addition, adverse reproductive outcomes were
also found in cross-sectional epidemiologic surveys of AN-exposed workers in China.
Neurobehavioral effects of AN-exposed workers were selected as the critical effect since this
effect was observed from several occupational studies with workers chronically exposed to AN
via the inhalation route.
Lu et al. (2005a) mentioned several limitations of this study. One was that the cited
exposure data represented estimates of previous exposure levels and no contemporaneous
personal monitoring data were available. Furthermore, as the NCTB was developed for
populations in Europe and North America, it is not known to what extent cultural differences
may have affected results of the Profile of Mood States Test, shown to be sensitive to cultural
differences. In addition, the study authors could not rule out the possibility that examiner drift
may have affected the results. Selection bias among volunteers in different groups was
mentioned as a possible confounding effect, but was discounted on the basis that the
participation among exposed workers was so high. Moreover, the largest measures of
neurobehavioral effects occurred in the acrylic fiber workers, who had lower average AN
exposure levels.
Animal to human extrapolation
No human oral exposure studies are available for derivation of the RfD. For derivation of
the RfD, extrapolating dose-response data from animals to humans is a source of uncertainty. A
PBTK model, which has its own associated uncertainties, was used to address toxicokinetic
differences between animals and humans. Uncertainties of the PBTK model are discussed as
part of the overall uncertainty discussion (Section 5.4.4.5), with quantitative details given in
Appendix D. Residual uncertainties pertaining to unknown interspecies differences in
pharmacodynamics were addressed by application of a UF of 3.
A human occupational exposure study (Lu et al., 2005a) was used for derivation of the
RfC, eliminating uncertainty associated with extrapolation from animals to humans. An RfC
was also derived from a two-year inhalation study of rats based on increased incidence of
inflammatory and degenerative nasal lesions. This alternative RfC was, at most, threefold higher
than the RfC derived from Lu et al. (2005a).
Dose-response modeling
BMD modeling was used to estimate the POD for the RfD. While models with better
biological support may exist, the selected models provided adequate mathematical fits to the
experimental data sets. BMD modeling has advantages over a POD based on a NOAEL or
326
DRAFT - DO NOT CITE OR QUOTE
-------
LOAEL because NOAELs/LOAELs are a reflection of the particular exposure concentration or
dose at which a study was conducted, they do not make use of the dose-response curve, and they
do not address the variability of the study population. NOAELs and LOAELs also are less
amenable to quantitative uncertainty analysis.
The RfC was based on a LOAEL identified from an occupational epidemiology study.
As stated above, there are several reasons to prefer a POD obtained from BMD modeling.
However, the available data only supported a LOAEL. A UF to address extrapolation to a less
adverse response level was applied.
Intrahuman variability
Heterogeneity among humans is another source of uncertainty. Uncertainty related to
human variation needs consideration, also, in extrapolation from a small subset of presumably
healthy humans (i.e., workers) to a larger, more diverse general population. Available data from
animal studies provide no evidence of gender differences in susceptibility to toxicity of AN,
although no data are available regarding possible gender differences in susceptibility. Human
genetic polymorphisms in CYP2E1 activities likely contribute to variability in human
susceptibility to the toxic effects of AN (see Section 4.8.4.1). A UF of 10 was used to account
for intrahuman variability for derivation of RfD and RfC. A factor of 10 has been found to be
generally sufficient to account for human variability in response to chemical exposure (Renwick
and Lazarus, 1998).
5.4. CANCER ASSESSMENT
5.4.1. Choice of Study/Data—with Rationale and Justification
As previously discussed in Section 4.1.2.2, evidence of a possible association between
exposure to AN and cancer in humans has been found in some studies. The best available
occupational epidemiologic study (Blair et al., 1998) reported that workers exposed to AN via
inhalation in the highest cumulative exposure category (i.e., >8 ppm-years) with more than
20 years of employment displayed a twofold increased risk of death from lung cancer compared
with unexposed workers (RR = 2.1, 95% CI = 1.2-3.8). To date, Blair et al. (1998) is largest
cohort study to assess the relationship between AN and cancer, following 25,460 workers in
eight AN-producing facilities, with two-thirds of this cohort having a follow-up period of over
20 years. The large sample size and the duration of follow-up time in this study provides a good
opportunity to detect any substantial elevation of case-specific cancer deaths. The findings from
Blair et al. (1998) are not inconsistent with other studies evaluating carcinogenic endpoints from
AN exposure. However, Blair et al. (1998) addressed known problems with earlier studies by
quantifying exposures, examining potential confounding from smoking, and employing an
internal control group of unexposed workers. And although only 5% of the cohort had died by
the time of analysis, this study has a large number of observed deaths that can be used to assess
327
DRAFT - DO NOT CITE OR QUOTE
-------
the relationship between AN-exposure and cancer based on cancer mortality. Also, Blair et al.
(1998) utilized a detail job-exposure matrix as part of the exposure assessment. For these
reasons, data from Blair et al. (1998) were chosen to derive an inhalation unit risk (IUR) for AN.
Additionally, the use of epidemiology data for the derivation of the IUR, versus using data from
an animal bioassay, reduces uncertainty inherent to animal to human extrapolation.
Sections 4.6.1 and 4.6.2 summarized the current animal data on AN indicating that it is a
multiple-site carcinogen in chronic oral and inhalation bioassays with rats and mice. The best
available animal studies for evaluating the dose-response relationship between AN exposures
and forestomach, CNS, Zymbal gland, tongue, and mammary gland tumors were two chronic
drinking water studies, one with Sprague-Dawley rats exposed to 0, 35, 100, or 300 ppm AN in
drinking water for 2 years (Quast, 2002; Quast et al., 1980a) and the other with F344 rats
exposed to 0, 1,3, 10, 30, or 100 ppm AN in drinking water for 2 years (Johannsen and
Levinskas, 2002b; Biodynamics, 1980c). In addition, data from another bioassay with Sprague-
Dawley rats exposed to AN in drinking water at concentrations of 0, 1, or 100 ppm for 2 years
were also considered (Johannsen and Levinskas, 2002a; Biodynamics, 1980a). In the absence of
definitive human studies demonstrating the carcinogenicity of chronic oral exposure to AN,
developing an oral cancer slope factor (CSF) from these animal studies is reasonable, especially
because the application of the AN PBTK models described in Chapter 3 can decrease the
toxicokinetic uncertainty in the interspecies extrapolation and in the mode of action of
carcinogenicity of AN in rodents versus humans.
Animal data from the only available chronic inhalation cancer bioassay with multiple
exposure levels (Dow Chemical Co., 1992a; Quast et al., 1980b) were also used to develop an
IUR to compare with the one derived from AN-exposed workers (Blair et al., 1998). In this
animal study, male and female Sprague-Dawley rats were exposed to 0, 20, or 80 ppm AN in air,
6 hours/day, 5 days/week, for 2 years. At 80 ppm, significantly increased incidences of
astrocytomas and glial cell proliferation and Zymbal gland tumors in males and females,
malignant mammary gland tumors (adenocarcinomas) in females, as well as intestinal and tongue
tumors in males, were found. At 20 ppm, male and female rats showed increased incidences of
astrocytomas and glial cell proliferation and Zymbal gland tumors. IUR estimates were derived
based on dose-response data for these tumors using the AN PBTK models previously described to
dosimetrically extrapolate from rats to humans.
5.4.2. Dose-Response Data
5.4.2.1. Human Occupational Data
An analysis of the lung cancer mortality and AN exposure data from the Blair et al.
(1998) study was conducted to derive an IUR estimate for AN. This analysis used the approach
described by Starr et al. (2004) in which the risk of death from lung cancer in AN-exposed
workers was characterized using a semi-parametric Cox regression model with time-dependent
328
DRAFT - DO NOT CITE OR QUOTE
-------
covariates. The Cox regression model has several advantages in that it allows inclusion of
individual exposure histories and utilizes internal controls, thus avoiding confounding by the
"healthy worker" effect. In contrast to the analysis done by Starr et al. (2004), the analysis
conducted for this assessment included the entire cohort (not just white male workers), and the
final model only had cumulative exposure as a covariate. Starr et al. (2004) included a second
covariate in their final model, plant of employment. This analysis is further described in
Appendix B-7.
5.4.2.2. Rat Oral Data
Incidence data for forestomach, CNS, Zymbal gland, tongue, and mammary gland tumors
in Sprague-Dawley and F344 rats exposed to AN in drinking water for 2 years were used to
develop site-specific oral CSFs for AN (Tables 5-7-5-9). Tumor incidences in F344 rats were
adjusted to exclude animals dying before the first appearance of each tumor type, which ranged
from day 419 to 495. No early mortality adjustments were made to the cumulative tumor
incidences from the Sprague-Dawley bioassay (Quast, 2002) because CNS and Zymbal gland
tumors were seen in some high-dose female rats as early as 7 months (210 days) after exposure
began, and no differences in survival were observed across the dose groups during the first
6 months of the study.
329
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-7. Incidence of CNS tumors in Sprague-Dawley and F344 rats
exposed to AN in drinking water for 2 years
Sprague-Dawley rats
(Quast, 2002; Quast et al., 1980a)a
Sex
AN drinking water
concentration (ppm)
Glial cell proliferation1"
Astrocytomas
Overall CNS tumor
Male
0
0/80
1/80
1/80
35
4/47°
8/47°
12/47°
100
3/48
19/48
22/48°
300
7/48°
23/48°
30/48°
Female
0
0/80
1/80
1/80
35
3/48
17/48°
20/48°
100
3/48
22/48°
25/48°
300
7/48°
24/48°
31/48°
F344 rats
(Johannsen and Levinskas, 2002b; Biodynamics, 1980c)d
Sex
AN drinking water
concentration (ppm)
Glial cell proliferation1"
Astrocytomas
Overall CNS
tumors
Brain
Spinal cord
Brain
Spinal cord
Male
0
0/160
0/156
2/160
1/156
3/160
1
0/80
0/79
2/80
0/79
2/80
3
0/78
0/70
1/78
0/70
1/78
10
0/80
0/78
2/80
0/78
2/80
30
0/79
0/79
10/79°
0/79
10/79°
100
0/76
0/70
21/76°
4/70°
25/76°
Female
0
0/157
0/155
1/157
0/155
1/157
1
0/80
0/78
1/80
0/78
1/80
3
0/80
0/79
2/80
0/79
2/80
10
0/77
0/72
4/75
1/72
5/75
30
0/80
0/87
6/80°
0/77
6/80°
100
0/76
0/69
23/76°
1/69
24/76°
"Incidence denominators were calculated from the total number of animals examined from the beginning of the
study. Numerators for CNS tumor incidences (glial cell proliferation, astrocytomas, or overall incidence of glial
cell proliferation or astrocytomas) for these Sprague-Dawley rats were reported as combined brain and spinal cord
lesions by Quast et al. (1980a).
bGlial cell proliferation is a smaller-sized lesion, either focal or multifocal, than astrocytomas, suggestive of an early
tumor.
"Statistically significantly different from controls (p < 0.05) as calculated by the study authors.
dThe denominators for incidences in these F344 rats exclude rats from the 6- and 12-mo sacrifices and unscheduled
deaths prior to the 12-mo sacrifice. Numerators for the overall incidences were the number of rats with astrocytomas
in brain or spinal cord. Reviews of summaries of individual animal pathology reports for this study (Appendix H,
Biodynamics, 1980c) indicated that five of the seven F344 rats showing spinal cord astrocytomas also showed a
brain astrocytoma; thus, the number was not always as great as the sum of the numerators for the incidences of these
lesions in the two tissues (e.g., 21/76 male 100-ppm brain tissues had astrocytomas and 4/70 male 100-ppm spinal
cord tissues had astrocytomas, but all four rats with astrocytomas also had brain astrocytomas). Because the
response to AN was predominately in brain tissue and a few spinal cord tissue samples were missing in each
exposure group, denominators for the overall incidences were taken as the number of rats examined for brain lesions
330
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-8. Incidence of mammary gland tumors in F344 and Sprague-
Dawley rats exposed to AN in drinking water for 2 years
Sex
AN drinking water
concentration
(ppm)
Benign mammary gland
tumors"
Malignant
mammary
gland tumorsb
Benign and/or
malignant mammary
gland tumors
F344 rats
(Johannsen and Levinskas, 2002b; Biodynamics, 1980c)°'d
Male
0
No statistically significantly elevated incidences of mammary gland
tumors were found in exposed groups compared with controls.
1
3
10
30
100
Female
0
12/156
3/156
14/156
1
5/80
4/80
8/80
3
6/80
0/80
6/80
10
8/79
1/78
9/80
30
9/80
3/80
12/80
100
9/73
6/7 3e
14/7 3e
Sprague-Dawley rats
(Quast, 2002; Quast et al., 1980a)f
Male
0
Not
reported
Not
reported
Not
reported
35
100
300
Female
0
52/80
1/80
58/80
35
35/48
1/48
42/48e
100
33/48
3/48
42/48e
300
22/48
10/48e
35/48
"Incidence includes fibroadenomas for F344 rats and fibroadenomas/adenofibromas/adenomas for Sprague-Dawley
rats.
incidence includes adenocarcinomas and carcinomas.
°The denominators for tumor incidences in F344 rats excluded rats from the 6- and 12-mo sacrifices and rats that
died prior to 12 mos. Mammary gland tumor incidences are for animals scheduled for the 18-mo and terminal
sacrifices. Microscopic examinations were only conducted on mammary glands showing gross signs of tumors—
the inclusion of all rats living for >52 wks in the denominators assumes that rats without gross signs of tumors were
also without microscopic neoplastic changes.
dAnimals with multiple tumor types within a tissue were counted only once.
"Statistically significantly different from controls (at p < 0.05) via Fisher's exact test.
denominators were calculated from the total number of animals examined from the beginning of the study.
Incidences were reported as total number of rats with benign-only, malignant-only, or benign and/or malignant
tumors.
331
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-9. Tumor incidences Sprague-Dawley and F344 in rats exposed to
AN in drinking water for 2 years
Sex
AN drinking
water
concentration
(ppm)
Forestomach
tumors
CNS
tumors
Zymbal gland
tumors
Tongue
tumors
Intestinal
tumors
Mammary gland
tumors
Sprague-Dawley rats
(Quast, 2002; Quast et al., 1980a)
Male
0
0/80
1/80
3/80
1/80
Not
reported
Not reported
35
2/47
12/47a
4/47
2/47
100
23/48a
22/48a
3/48
4/48
300
39/48a
30/48a
16/48a
5/48a
Female
0
1/80
1/80
1/80
0/80
0/80
58/80
35
1/48
20/48a
5/48a
1/48
1/48
42/48a
100
12/48a
25/48a
9/48a
2/48
4/48a
42/48a
300
30/48a
31/48a
18/48a
12/48a
4/48a
35/48
F344 rats
(Johannsen and Levinskas, 2002b; Biodynamics, 1980c)b
Male
0
0/159
3/160
1/147
Not examined
3/159
1
1/80
2/80
1/76
2/80
3
4/7 8a
1/78
0/73
2/78
10
3/80a
2/80
0/67
2/80
30
4/80a
10/79°
2/71°
0/80
100
1/77
25/76°
14/68°
2/77
Female
0
1/157
1/157
0/157
Not examined
14/156
1
1/80
1/80
0/73
8/80
3
2/79
2/80
0/73
6/80
10
2/77
5/75
0/70
9/80
30
4/80a
6/80°
2/73°
12/80
100
2/75
24/76°
8/62°
14/7 3a
aSignificantly different from controls (p < 0.05).
bThe denominators for tumor incidences in F344 rats excluded rats from the 6- and 12-mo sacrifices and
unscheduled deaths prior to the 12-mo sacrifice. Numerators for the incidences of CNS tumors were derived by
adding the number of rats with brain or spinal astrocytomas; denominators were taken as the greater of the number
of rats examined for brain or spinal cord lesions after the 12-mo sacrifice. Numerators for Zymbal gland tumor
incidences included squamous cell papillomas and carcinomas designated to occur in the ear canal. Mammary
gland tumor incidences are for fibroadenomas and adenocarcinomas in animals sacrificed or found dead after
12 mos. Tongues were not routinely histopathologically examined for tumors in this bioassay.
Significantly different from controls (p < 0.01).
332
DRAFT - DO NOT CITE OR QUOTE
-------
As previously noted, a second 2-year AN drinking water study employing Sprague-
Dawley rats has been conducted (Johannsen and Levinskas, 2002a; Biodynamics, 1980a). To
determine whether these data should also be employed in the cancer assessment, an analysis was
performed to evaluate the statistical validity of pooling tumor incidence data from the two
Sprague-Dawley rat chronic AN drinking water studies. The first study (Quast, 2002; Quast et
al., 1980a) exposed animals to AN drinking water concentrations of 0, 35, 100, and 300 ppm,
while the second study (Johannsen and Levinskas, 2002a) employed AN drinking water
concentrations of 0, 1, and 100 ppm. The dichotomous multistage model in BMDS was fit to the
tumor incidence data from three sites (i.e., forestomach, CNS, and Zymbal gland) in each sex
across both studies, using administered animal dose expressed in mg/kg-day. A statistical test
described by Stiteler et al. (1993), which employs a maximum likelihood ratio statistic
distributed as a % , was then used to test the null hypothesis that the two data sets are compatible
with a common dose-response model. If the null hypothesis is not rejected, this provides
evidence that the results from the two studies may be pooled.
As discussed in more detail in Appendix B-5, the statistical tests indicated that some, but
not all, of the data sets from the two studies were consistent with a common dose-response
model. More specifically, the results of this analysis showed that forestomach and Zymbal gland
tumors in both male and female Sprague-Dawley rats were not compatible with a common dose-
response model, while CNS tumors in male and female Sprague-Dawley rats were compatible
with a common dose-response model. Because of these conflicting results, it was decided that
the results from the two Sprague-Dawley rat drinking water studies would not be pooled.
Therefore, the final dose-response analysis for deriving the oral slope factor for AN focused on
the two rat drinking water studies containing the most dose groups (i.e., the Sprague-Dawley rat
bioassay reported by Quast [2002] and the F344 rat bioassay reported by Johannsen and
Levinskas [2002b]).
5.4.2.3. Rat Inhalation Data
Incidence data for intestinal, CNS, Zymbal gland, tongue, and mammary gland tumors in
Sprague-Dawley rats exposed to AN via inhalation were used for deriving site-specific IURs for
AN (Table 5-10). With the exception of one male in the 80 ppm exposure group that died with a
CNS tumor after 7-12 months on study, all of the remaining tumors occurred in rats that died or
were sacrificed after at least 12 months of AN exposure. Denominators for the incidences in
Table 5-10 excluded animals that died without a tumor before 12 months on study because these
animals were not exposed long enough to be at risk for tumor development. Although a
statistically significantly elevated incidence in tongue tumors was observed in male rats at 80
ppm, tongues from only 14 of the males in the 20 ppm exposure group were examined. No data
on the incidence of tongue tumors in female rats were presented in the original study report by
Quast et al. (1980b).
333
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-10. Incidences of intestinal, CNS, Zymbal gland, tongue, and
mammary gland tumors in Sprague-Dawley rats exposed to AN via
inhalation for 2 years
Sex
AN air
concentration
(ppm)
Intestinal
tumors"
CNS
tumorsa'b
Zymbal gland
tumors"
Tongue
tumors"
Mammary gland
adenocarcinomas"
Male
0
4/96
0/96
2/96
1/95
-
20
3/93
4/93
4/93
0/14
-
80
17/82°
22/82°
11/82°
7/82°
-
Female
0
-
0/93
0/93
-
9/93
20
-
8/99°
1/98
-
8/98
80
-
20/89°
11/89°
-
20/99°
Tor all incidence data, the denominators excluded rats dying earlier than 12 mos in the study. These data were
ascertained from Tables 22, 25, 31, 34, and 35 in the original study report by Quast et al. (1980b).
incidences for CNS tumors (brain and spinal cord) in Sprague-Dawley rats listed in this table, as reported by
Quast et al. (1980b), include both glial cell proliferation and astrocytomas.
"Statistically significantly different from controls (p < 0.05) as calculated by the study authors.
Sources: Dow Chemical (1992a); Quast et al. (1980b).
5.4.3. Dose-Response Modeling
5.4.3.1. Human Occupational Data
A Cox regression model was developed based on an analysis of the Blair et al. (1998)
lung cancer mortality data for AN-exposed workers. This model yielded a regression coefficient
3 3
(fi) estimate of 1.2 x 10" with an associated estimated standard error of 2.47 x 10" and a
correspondingp value of 0.61. The 95% upper confidence limit (UCL) on the parameter
"3
estimate, /?, was 5.3 x 10" . The parameter estimate, /?, in this model was about threefold greater
than the same parameter estimate in the final model developed by Starr et al. (2004), but the
UCL was similar. The Cox regression model described above was used to predict an AN
exposure concentration (EC) and its associated 95% lower confidence limit (LEC), associated
with a 10" (1%) risk of dying from lung cancer at age 80 (ECoi and LECoi, respectively).
Conversions of occupational exposures to continuous environmental exposures were performed
to account for differences in the number of days exposed per year (240 vs. 365 days) and in the
amount of air inhaled during an 8-hour workday versus a 24-hour day (10 and 20 m /day,
respectively). The resulting AN exposure estimates (by age 80) were ECoi = 0.992 ppm (or 2.2
3 3
mg/m ) and LECoi = 0.238 ppm (or 0.524 mg/m ). The IUR estimate was then derived by linear
2 1
extrapolation from the LECoi. The corresponding unit risk was calculated to be 4.2 x 10" ppm"
2 3
or 2 x 10" per mg/m . As previously noted, Appendix B-7 provides more details on this Cox
regression analysis.
334
DRAFT - DO NOT CITE OR QUOTE
-------
5.4.3.2. Rat Oral Data
Within each rat strain, sex, and tumor site, the multistage model, employing EPA's
BMDS (version 1.4.1), was fit to the tumor incidence data from the bioassays in Sprague-
Dawley (Quast, 2002) and F344 (Johannsen and Levinskas, 2002b) rats shown in Table 5-9
using two internal dose metrics, CEO concentration in blood (AUC/24 hours) and AN
concentration in blood (AUC/24 hours). As already discussed, CEO is a DNA-reactive epoxide
metabolite thought to play a key role in the carcinogenic mode of action of AN. The
EPA-modified PBTK model employed consistently predicted higher CEO concentrations in
blood and brain than were reported in studies of orally exposed rats (see Figures 3 and 4 in
Kedderis et al., 1996). In contrast, AN concentrations in blood, brain, and liver predicted by the
same PBTK model were fairly close to measured AN concentrations in rats following oral
exposure. Ultimately, CEO levels in blood were chosen as the internal dose metric of choice for
the oral cancer dose-response assessment because CEO is believed to be the most biologically
relevant dose metric, and it is also consistent with the dose metric employed in derivation of the
RfD. As with the RfD, CEO concentration in blood represents a reasonable internal dose metric
for extrapolating from orally exposed rats to humans.
The AN concentrations in drinking water (in ppm) employed in the Quast (2002) and
Johannsen and Levinskas (2002b) rat studies were converted to AN administered doses (in
mg/kg-day) by the study authors, using water intake data recorded during the study. These
administered doses of AN were then used as input into the EPA-modified rat PBTK model of
Kedderis et al. (1996) in order to predict a rat internal dose (either CEO-AUC or AN-AUC
concentration in blood) resulting from the ingestion of a total daily dose of AN equivalent to the
administered dose consumed in six bolus episodes per day that reflect the daily drinking water
consumption pattern of rats. The resulting predicted CEO-AUC or AN-AUC concentrations in
rat blood were then employed in dose-response modeling. Table 5-11 displays the relationship
between AN drinking water concentrations (in ppm), administered animal doses (in mg/kg-day),
and two internal dose metrics (i.e., CEO in blood and AN in blood, both expressed in mg/L)
predicted from the PBTK model.
335
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-11. Four different dose metrics, two external and two internal, based
on doses employed in studies of Sprague-Dawley and F344 rats exposed to
AN in drinking water for 2 years
Species, strain
Sex
External dose
Predicted internal dose"
(reference)
Concentration in
drinking water
(ppm)
Administered
doseb
(mg/kg-d)
CEO in blood
(mg/L)
AN in blood
(mg/L)
0
0
0
0
Male
35
3.4
1.83 x 10-3
2.06 x 10-2
Rat,
Sprague-Dawley
(Quast, 2002)
100
8.5
4.36 x 10-3
5.36 x 10-2
300
21.3
9.70 x 10-3
1.46 x 10"1
0
0
0
0
Female
35
4.4
2.07 x 10-3
2.37 x 10-2
100
10.8
4.87 x 10-3
6.18 x 10-2
300
25.0
1.01 x 10-2
1.56 x 10"1
0
0
0
0
1
0.08
4.06 x 10-5
4.33 x 10-4
Male
3
0.25
1.27 x 10-4
1.35 x 10-3
10
0.83
4.19 x 10-4
4.52 x 10-3
Rat, F344
(Johannsen and
Levinskas, 2002b)
30
2.48
1.23 x 10-3
1.37 x 10-2
100
8.37
3.97 x 10-3
4.85 x 10-2
0
0
0
0
1
0.12
5.32 x 10-5
5.73 x 10-4
Female
3
0.36
1.59 x 10-4
1.72 x 10-3
10
1.25
5.49 x 10-4
6.02 x 10-3
30
3.65
1.58 x 10-3
1.79 x 10-2
100
10.90
4.46 x 10-3
5.63 x 10-2
The EPA-modified rat PBPK model of Keddaris et al. (1996) was employed to predict a rat internal dose (i.e.,
either AN-AUC or CEO-AUC concentration in blood) resulting from the ingestion of an administered dose of AN
consumed in six bolus episodes/d.
bAdministered doses were averages calculated by the study authors based on animal BW and drinking water intake.
Employing the internal dose metrics in Table 5-11 (i.e., CEO in blood and AN in blood),
successive stages of the multistage model, starting with stage 1 and ending with the stage equal
to the number of dose groups minus one, were fit to the tumor incidence data at a particular site
for each rat strain and sex. Then, all stages of the multistage model that did not show a
significant lack of fit (i.e. ,p>0. 1) were compared using AIC. The stage of the multistage model
with the lowest AIC was selected as the "best-fit" model. For most tumor sites, the one-stage
model exhibited the best fit.
A BMR of 10% extra risk was selected for all tumor sites, consistent with the Guidelines
for Carcinogen Risk Assessment (U.S. EPA, 2005a), which recommend identifying the POD near
the lower end of the observed data. Using the best-fit model, the resulting BMDi0s and
336
DRAFT - DO NOT CITE OR QUOTE
-------
BMDL10S were estimated for each tumor site within each rat strain and sex. A summary of the
results of this BMD modeling is shown in Table 5-12. Additional details regarding dose-
response modeling used in the derivation of the oral CSFs are provided in Appendix B-3.
Table 5-12. BMD modeling results using tumor incidence data from male
and female Sprague-Dawley and F344 rat studies in which animals were
exposed to AN in drinking water for 2 years
BMD modeling results3
Strain/species
Sex
Tumor site
Dose metric
BMD10
(mg/L)
BMDL10
(mg/L)
Forestomach
CEO
1.87 x 10-3
1.44 x 10-3
AN
2.26 x 10-2
1.70 x 10-2
CNS
CEO
8.87 x 10-4
7.16 x 10-4
Male
AN
8.82 x 10-3
6.64 x 10-3
Zymbal gland
CEO
5.46 x 10-3
3.15 x 10-3
AN
8.94 x 10-2
4.26 x 10-2
Tongue
CEO
8.78 x 10-3
4.90 x 10-3
AN
1.29 x 10"1
6.97 x 10-2
Sprague-Dawley rats
Forestomach
CEO
3.29 x 10-3
2.38 x 10-3
AN
4.22 x 10-2
2.49 x 10-2
CNS
CEO
5.79 x 10-4
4.51 x 10-4
AN
7.12 x 10-3
5.55 x 10-3
Female
Zymbal gland
CEO
2.40 x 10-3
1.78 x 10-3
AN
3.41 x lO-2
2.52 x 10-2
Tongue
CEO
6.70 x 10-3
4.74 x 10-3
AN
9.67 x 10-2
6.10 x 10-2
Mammary gland
CEO
5.50 x 10-4
2.98 x 10-4
AN
7.05 x 10-3
3.77 x 10-3
337 DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-12. BMD modeling results using tumor incidence data from male
and female Sprague-Dawley and F344 rat studies in which animals were
exposed to AN in drinking water for 2 years
BMD modeling results3
Strain/species
Sex
Tumor site
Dose metric
BMD10
(mg/L)
BMDL10
(mg/L)
Forestomach
CEO
6.03 x 10-4
3.55 x 10-4
AN
6.48 x 10-3
3.81 x 10-3
Male
CNS
CEO
1.16 x 10-3
8.74 x 10-4
AN
1.37 x 10-2
1.03 x 10-2
Zymbal gland
CEO
2.73 x 10-3
2.19 x 10-3
AN
3.31 x 10-2
2.59x 10-2
F344 rats
Forestomach
CEO
3.65 x 10-3
1.69 x 10-3
AN
4.13 x 10-2
1.90 x 10-2
CNS
CEO
1.39 x 10-3
1.05 x 10-3
Female
AN
1.70 x 10-2
1.28 x 10-2
Zymbal gland
CEO
3.78 x 10-3
2.97 x 10-3
AN
5.41 x 10-2
3.35 x 10-2
Mammary gland
CEO
3.58 x 10-3
1.97 x 10-3
AN
4.51 x 10-2
2.45 x 10-2
aThe multistage model in EPA's BMDS (version 1.4.1) was fit to each set of tumor incidence data from the
Sprague-Dawley and F344 rat bioassays, as shown in Table 5-9, using the two internal dose metrics, CEO in blood
and AN in blood, expressed in mg/L. An adequate fit of the MS model was achieved if the %2 goodness-of-fit
statistic yielded p> 0.1. In the case of CNS tumors in Sprague-Dawley female rats, an adequate fit of the
multistage model to the data could not be achieved; therefore, the best fitting of the other models available in
BMDS (assessed by AIC), the log-logistic model, was used. Numbers in parentheses indicate the number of dose
groups dropped in order to obtain an adequate fit (starting with the highest dose group). Appendix B-3 provides
further details on these BMD modeling results.
Sources: Johannsen and Levinskas (2002b) (F344); Quast (2002) (Sprague-Dawley).
After completion of the dose-response modeling, the BMDLios estimated for each tumor
site within each rat strain and sex for each internal dose metric were input into the EPA-modified
human PBTK model of Sweeney et al. (2003) in order to predict the human equivalent
administered dose of AN that corresponds to the estimated BMDLio, assuming six bolus
ingestion episodes of AN per day. The resulting predicted human equivalent administered dose
of AN, expressed in mg/kg-day, are shown in the third column of Table 5-13 for the internal
dose metric CEO in blood and Table 5-14 for the internal dose metric AN in blood. Finally, for
each rat strain, sex, and tumor site, the site-specific oral CSFs shown in the last column of Table
5-13 (based on CEO concentration in blood) and Table 5-14 (based on AN concentration in
blood) were derived by dividing the BMR (i.e., 10% or 0.1) by the human equivalent
administered dose of AN (in mg/kg-day), displayed in the third column of Tables 5-13 and 5-14.
338
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-13. Site-specific oral CSFs for AN based on BMD modeling of
tumor incidence data in rats and predicted CEO levels in blood
(AUC/24 hours) of rats and humans assuming episodic exposure to AN
Rat strain, gender
Tumor site
Rat BMDL10a
(mg/L)
BMDL10/HEDb
(mg/kg-d)
CSFC
(mg/kg-d)1
Sprague-Dawley males
Forestomach
1.44 x 10-3
0.142
0.704
CNS
7.16 x 10-4
0.071
1.410
Zymbal gland
3.15 x 10-3
0.312
0.320
Tongue
4.90 x 10-3
0.486
0.206
Sprague-Dawley females
Forestomach
2.38 x 10-3
0.236
0.424
CNS
4.51 x 10-4
0.045
2.222
Zymbal gland
1.78 x 10-3
0.176
0.567
Tongue
4.74 x 10-3
0.470
0.213
Mammary gland
2.98 x 10-4
0.029
3.390
F344 males
Forestomach
3.55 x 10-4
0.035
2.850
CNS
8.74 x 10-4
0.086
1.160
Zymbal gland
2.19 x 10-3
0.217
0.462
F344females
Forestomach
1.69 x 10-3
0.167
0.599
CNS
1.05 x 10-3
0.104
0.963
Zymbal gland
2.97 x 10-3
0.294
0.340
Mammary gland
1.97 x 10-3
0.195
0.514
aRat BMDL10 refers to the estimated 95% lower confidence limit on the internal dose of CEO in blood in
the rat associated with a 10% extra risk for the incidence of tumors at the specified site in the associated
strain and sex. This value is taken from the last column of Table 5-12.
bPBTK model-derived HED of AN that would result in a human 24-hr blood CEO-AUC equivalent to the
rat CEO-AUC presented in the previous column of the table, assuming AN ingestion in six bolus
episodes/d.
CCSF — BMR/BMDLio/hed or 0. 1/BMDLio/hed-
Sources: Johannsen and Levinskas (2002b) (F344); Quast (2002) (Sprague-Dawley).
339
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-14. Site-specific oral CSFs for AN based on BMD modeling of
tumor incidence data in rats and predicted AN levels in blood
(AUC/24 hours) of rats and humans assuming episodic exposure to AN
Rat strain, gender
Tumor site
Rat BMDL10a
(mg/L)
BMDL10/HEDb
(mg/kg-d)
CSFC
(mg/kg-d)1
Sprague-Dawley males
Forestomach
1.70 x 10-2
4.30
0.023
CNS
6.64 x 10-3
1.89
0.053
Zymbal gland
4.26 x 10-2
8.59
0.012
Tongue
6.97 x 10-2
11.85
0.008
Sprague-Dawley Females
Forestomach
2.49 x 10-2
5.82
0.017
CNS
5.55 x 10-3
1.60
0.062
Zymbal gland
2.52 x 10-2
5.87
0.017
Tongue
6.10 x 10-2
10.90
0.009
Mammary gland
3.77 x 10-3
1.11
0.090
F344 males
Forestomach
3.81 x 10-3
1.12
0.089
CNS
1.03 x 10-2
2.82
0.036
Zymbal gland
2.59 x 10-2
5.99
0.017
F344females
Forestomach
1.90 x 10-2
4.71
0.021
CNS
1.28 x 10-2
3.38
0.030
Zymbal gland
3.35 x 10-2
7.26
0.014
Mammary gland
2.45 x 10-2
5.75
0.017
aRat BMDLio refers to the estimated 95% lower confidence limit on the internal dose of AN in blood in the
rat associated with a 10% extra risk for the incidence of tumors at the specified site in the associated strain
and sex. This value is taken from the last column of Table 5-12.
bPBTK model-derived HED of AN that would result in a human 24-hr blood AN-AUC equivalent to the
rat AN-AUC presented in the previous column of the table, assuming AN ingestion in six bolus
episodes/d.
°CSF = BMR/BMDL10/hed or 0. 1/BMDL10/Hed-
Sources: Johannsen and Levinskas (2002b) (F344); Quast (2002) (Sprague-Dawley).
In comparing Tables 5-13 and 5-14, the CEO-based site-specific oral CSF estimates in
Table 5-13 are higher than those based on AN blood levels in Table 5-14. This occurs because
the VmaxC/Km for AN oxidation to CEO is estimated to be about 10 times higher in humans than
in the rat. AN enzymatic GSH conjugation is estimated to be 1.5 times higher in the rat, but the
2nd-order removal constant for non-enzymatic reaction of AN and GSH is assumed to be the
same and the GSH tissue levels are roughly the same (and lower in the larger tissue groups). The
result, for example, with a steady oral "infusion" of 30 mg/kg-day, is that AN is removed
somewhat faster overall, leading to physiologically based pharmacokinetic (PBPK) predicted
340
DRAFT - DO NOT CITE OR QUOTE
-------
steady-state blood level of 0.114 mg/L AN in humans vs. 0.151 in the rat, but a much higher
portion of this goes to CEO in the human yielding a CEO blood level of 0.297 vs. 0.013 mg/L in
the rat. Thus, AN:CEO = 11.4 in the rat and 0.385 in the human, and so the CEO-based risk
estimates are approximately 50 times higher than the AN-based estimates.
For comparative purposes, in Table 5-15, the BMDLio estimates generated from the
BMD modeling of the incidence of tumors in male and female Sprague-Dawley and F344 rats
using administered animal dose (modeling results not shown) were converted to human
equivalent administered doses of AN using BW scaling to the 3/4 power. Then, as in Tables 5-13
and 5-14, oral CSFs were derived by dividing the BMR (i.e., 10% or 0.1) by the human
equivalent administered dose of AN (in mg/kg-day).
Table 5-15. Site-specific oral CSFs for AN based on BMD modeling of
tumor incidence data in rats and BW scaling to the % power to convert from
rat to human administered doses
Rat strain, gender
Tumor site
Rat BMDL10a
(mg/kg-d)
BIYIDL,o,m.;i,b
(mg/kg-d)
CSFC
(mg/kg-d)1
Sprague-Dawley males
Forestomach
2.76
0.812
0.123
CNS
1.48
0.436
0.229
Zymbal gland
5.17
1.52
0.066
Tongue
10.4
3.04
0.033
Sprague-Dawley females
Forestomach
4.81
1.27
0.079
CNS
0.99
0.262
0.382
Zymbal gland
4.19
1.11
0.090
Tongue
8.30
2.19
0.046
Mammary gland
0.66
0.174
0.575
F344 males
Forestomach
0.70
0.193
0.517
CNS
1.81
0.493
0.203
Zymbal gland
3.43
0.932
0.107
F344females
Forestomach
3.89
0.926
0.108
CNS
2.52
0.602
0.166
Zymbal gland
6.64
1.59
0.063
Mammary gland
4.77
1.14
0.088
aRat BMDL10 is the estimated 95% lower confidence limit on the administered dose of AN in the rat
associated with a 10% extra risk for the incidence of tumors at the specified site in the associated strain
and sex. This value is generated from the "best-fit" dose-response model in BMDS (version 1.4.1).
bThe HED of AN equal to the rat BMDLio in the previous column converted through use of BW scaling to
the 3/i power.
CCSF — BMR/BMDLio/hed or 0. 1/BMDLio/hed-
Sources: Johannsen and Levinskas (2002b) (F344); Quast (2002) (Sprague-Dawley).
341
DRAFT - DO NOT CITE OR QUOTE
-------
5.4.3.3. Rat Inhalation Data
Within each sex and for each tumor site, the multistage model, employing EPA's BMDS
(version 1.4.1), was fit to the tumor incidence data from the AN inhalation bioassay in Sprague-
Dawley rats (Quast et al., 1980b) shown in Table 5-10 using the internal dose metric, CEO
concentration in blood (AUC/24 hours). In contrast to the approach used for oral exposure, only
one internal dose metric was selected on which to base site-specific IURs (i.e., CEO in blood),
because, in contrast to oral exposure, the EPA-modified PBTK model adequately predicted
measured blood and brain concentrations of CEO in rats exposed to AN via inhalation (Kedderis
et al., 1996). Furthermore, as mentioned previously, CEO is the DNA-reactive metabolite
thought to be key in the carcinogenic mode of action of AN.
Prior to dose-response modeling, the AN concentrations in air (in ppm) administered in
the Quast et al. (1980b) rat study were input into the EPA-modified rat PBTK model of Kedderis
et al. (1996) in order to predict a rat internal dose (CEO-AUC concentration in blood) resulting
from inhalation exposure to the administered air concentration of AN. The resulting predicted
CEO-AUC concentrations in rat blood were then employed in dose-response modeling.
Table 5-16 displays the relationship between AN concentrations in air (expressed in ppm) and
the internal dose metric, CEO in blood (expressed in mg/L), predicted from the PBTK model.
Table 5-16. Two different dose metrics, one external and one internal, based
on administered air concentrations of AN employed in a 2-year bioassay in
Sprague-Dawley rats
Species, strain
Sex
Exposed AN concentration in air
(ppm)
Predicted CEO
concentration in blood"
(mg/L)
0
0
Male
20
2.17 x 10-3
Rat, Sprague-Dawley
80
8.20 x 10-3
0
0
Female
20
2.18 x 10-3
80
8.24 x 10-3
aSee Table 5-11.
Source: Quast et al. (1980b).
After completion of the dose-response modeling, the BMCios and BMCLi0s estimated
within each sex for each tumor site using CEO in blood as the dose metric from Table 5-17 were
input into the EPA-modified human PBTK model of Sweeney et al. (2003) in order to predict the
HEC of AN in air that corresponds to the estimated BMCio and BMCLio in animals. The
resulting predicted 95% lower bounds of the HECs (BMCLi0/hec) of AN in air, expressed in
342
DRAFT - DO NOT CITE OR QUOTE
-------
"3
mg/m , are shown in the third column of Table 5-18 for the internal dose metric CEO in blood.
Finally, for each sex and tumor site, the site-specific IURs shown in the last column of
Table 5-19 (based on CEO concentration in blood) were derived by dividing the BMR (i.e., 10%
or 0.1) by the BMCLi0/hec-
Table 5-17. BMD modeling results using tumor incidence data from male
and female Sprague-Dawley rats exposed to AN via inhalation for 2 years
and CEO concentration in blood predicted from an EPA-modified PBTK
model
Strain, species
Sex
BMD modeling results
Tumor site
Dose metric
BMC10
(mg/L)
BMCL10
(mg/L)
Rat, Sprague-
Dawley
Male
Intestine
CEO
6.06 x 10-3
4.47 x 10-3
CNS
CEO
3.14 x 10-3
2.31 x 10-3
Zymbal gland
CEO
7.26 x 10-3
4.40 x 10-3
Female
Tongue
CEO
9.48 x 10-3
6.39 x 10-3
CNS
CEO
3.21 x 10-3
2.39 x 10-3
Zymbal gland
CEO
7.90 x 10-3
5.09 x 10-3
Mammary gland
CEO
7.31 x 10-3
4.33 x 10-3
aThe multistage model in EPA's BMDS (version 1.4.1) was fit to each set of tumor incidence data from the
Sprague-Dawley rat bioassay, as shown in Table 5-10, using the internal dose metric, CEO in blood, expressed in
mg/L. An adequate fit of the multistage model was achieved if the %2 goodness-of-fit statistic yielded p> 0.1.
Appendix B-4 provides further details on these BMD modeling results.
Source: Quast et al. (1980b).
343
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-18. Site-specific IURs for AN based on BMD modeling of tumor
incidence data in Sprague-Dawley rats and PBTK modeling of CEO levels in
blood (AUC/24 hours) of rats and humans
Gender Rat BMCL10a BMCL10/HEcb IURC
Tumor site (mg/L) (mg/m3) (mg/m3)1
Males
Intestine
4.47 x 10-3
6.00
0.017
CNS
2.31 x 10-3
3.10
0.032
Zymbal gland
4.40 x 10-3
5.91
0.017
Tongue
6.39 x 10-3
8.58
0.012
Females
CNS
2.39 x 10-3
3.21
0.031
Zymbal gland
5.09 x 10-3
6.83
0.015
Mammary gland
4.33 x 10-3
5.81
0.017
aRat BMCLio refers to the estimated 95% lower confidence limit on the concentration of CEO in blood of
the rat associated with a 10% extra risk for the incidence of tumors at the specified site based on BMD
modeling. This value is taken from the last column of Table 5-17.
VBTK model-derived HEC of AN in air that would result in a human 24-hr blood CEO-AUC
concentration equivalent to the rat BMCLio in the previous column of the table assuming continuous
exposure to AN. The human PBTK model employed did not contain the human in vitro to in vivo
modifying factor for CEO hydrolysis proposed by Sweeney et al. (2003).
°IUR = BMR/BMCL10/hec or 0.1/BMCL10/hec-
Table 5-19. Site-specific IURs for AN based on BMD modeling of tumor
incidence data in Sprague-Dawley rats exposed to AN via inhalation
Gender
Rat BMCL10a
BMCL10/hec
IURC
Tumor site
(ppm)
(mg/m3)
(mg/m3)1
Males
Intestine
42.68
93.58
1.07 x 10-3
CNS
22.23
48.74
2.05 x 10-3
Zymbal gland
42.53
93.25
1.07 x 10-3
Tongue
59.41
130.26
7.68 x 10-4
Females
CNS
22.89
50.19
1.99 x 10-3
Zymbal gland
48.74
106.87
9.36 x 10-4
Mammary gland
37.82
82.92
1.21 x 10-3
aRat BMCL10 refers to the estimated 95% lower confidence limit on the concentration of AN in air
associated with a 10% extra risk for the incidence of tumors at the specified site based on BMD modeling.
bThe HEC of AN in air equivalent to the rat BMCL10 in the previous column of the table generated through
use of the following conversion equation: mg/m3 = [ppm x molecular weight]/24.20, where molecular
weight = 53.06.
°IUR = BMR/BMCLio/hec or 0. 1/BMCL10/hec-
344
DRAFT - DO NOT CITE OR QUOTE
-------
5.4.4. Cancer Risk Values
5.4.4.1. Oral CSFs
Because AN has been demonstrated to produce tumors at multiple sites in rats, the
estimation of risk based on only one tumor site may underestimate the overall carcinogenic
potential of AN. Under the assumption that AN causes tumors by a mutagenic mode of action,
estimates of the composite risk of etiologically distinct tumor types in each rat strain/sex
combination considered in Section 5.4.3.2 were derived employing the Bayesian approach
described in Appendix B-6. This approach is consistent with the recommendations of the NRC
(1994) and the Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), which
recommends "adding risk estimates derived from different tumor sites (NRC, 1994) as an option
for "how best to represent the human cancer risk." This composite risk is associated with the
potential for developing tumors, not at all of the sites, but at any combination of the sites
observed in male or female rats. For this analysis, the etiologically distinct tumor sites
associated with AN exposure were assumed to be statistically independent. This assumption
cannot currently be verified, but if not correct could lead to an overestimate of risk. NRC (1994)
concluded that a general assumption of statistical independence of tumors within animals was
not likely to introduce substantial error in assessing carcinogenic potency from rodent bioassay
data.
Table 5-20 presents the estimated human equivalent oral CSFs for both site-specific risks
and composite risks from multiple tumor types from each rat strain and sex exposed to AN via
drinking water based on the internal dose metric, CEO concentration in blood (AUC over
24 hours).
345
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-20. Estimated human equivalent oral CSFs for composite risk based
on tumor incidence in AN-exposed rats and predicted CEO-AUC levels in
blood
Rat strain and gender
Tumor sites included
Site-specific oral CSFa
(mg/kg-d)1
Oral CSF for composite risk
across sites
(mg/kg-d)1
Sprague-Dawley Male
Forestomach
0.704
2.8
CNS
1.410
Zymbal gland
0.320
Tongue
0.206
Sprague-Dawley Female
Forestomach
0.424
5.0
CNS
2.222
Zymbal gland
0.567
Tongue
0.213
Mammary gland
3.390
F344 Male
Forestomach
2.850
3.1
CNS
1.160
Zymbal gland
0.462
F344 Female
Forestomach
0.599
1.9
CNS
0.963
Zymbal gland
0.340
Mammary gland
0.514
aThese site-specific oral CSFs are taken from the last column of Table 5-13.
In Sprague-Dawley rats, males showed statistically significantly increased incidences of
tumors at four sites (i.e., forestomach, CNS, Zymbal gland, and tongue), while females showed
statistically significantly increased incidences of tumors at five sites (i.e., forestomach, CNS,
Zymbal gland, tongue, and mammary gland), following chronic oral exposure to AN. The
composite oral CSF based on tumor incidences in male Sprague-Dawley rats chronically exposed
to AN (using CEO concentration in blood) is estimated to be 2.8 per mg/kg-day, about twice as
high as the estimated slope factor resulting from the most sensitive single site in this strain and
sex. For female Sprague-Dawley rats, the composite oral CSF (based on CEO concentration in
blood) was 5.0 per mg/kg-day, approximately 50% higher than the estimated slope resulting
from the most sensitive single site in this strain and sex.
In F344 rats, males showed statistically significant increased incidences of tumors at
three sites (i.e., forestomach, CNS, and Zymbal gland), while females showed statistically
significant increased incidences of tumors at four sites (i.e., forestomach, CNS, Zymbal gland,
and mammary gland), following oral exposure to AN in drinking water for 2 years. Again, under
346
DRAFT - DO NOT CITE OR QUOTE
-------
the assumption that the tumors at these sites are statistically independent, the composite oral CSF
in F344 males (based on CEO concentration in blood) was 3.1 per mg/kg-day, slightly higher
than the estimated slope factor of 2.8 per mg/kg-day based on tumors in the forestomach alone,
the most sensitive single site in this strain and sex. For female F344 rats, the composite oral CSF
(based on CEO concentration in blood) was 1.9 per mg/kg-day, about twice the estimated slope
factor of 0.96 per mg/kg-day from tumors of the CNS, the most sensitive single site in this strain
and sex.
The slope factors obtained from male and female Sprague-Dawley rats and F344 rats
ranged from 2 to 5 (mg/kg-day)"1. While female Sprague-Dawley rats had the highest slope
factor due to increase in mammary gland tumor risk, the increase in mammary gland tumor risk
in female F344 rats was not as high. There is no information as to which rat strain may be most
similar to humans. Both strains of rats developed same tumors in response to AN exposure. For
the purpose of this assessment, the slope factor of 5 per mg/kg-day is recommended for use in
humans because it is the value obtained from the most sensitive species, strain, and sex (i.e.,
female Sprague-Dawley rats). This slope factor should not be used with exposures greater than
0.04 mg/kg-day (the lowest POD supporting the composite risk) because above this level the
slope factor cannot be expected to be an adequate approximation to the dose-response
relationship. The fitted dose-response relationship and pharmacokinetic models should be used
to estimate risk above this exposure level. See Section 5.6 regarding the application of age-
dependent adjustment factors (ADAFs).
Previous IRIS Assessment
In the previous IRIS assessment completed in 1991, EPA derived an oral CSF of
5 x 10"1 (mg/kg-day)"1 by application of the linearized multistage model to incidence data from
rats with astrocytomas, Zymbal gland carcinomas, or forestomach papillomas or carcinomas
from the three 2-year drinking water studies in rats considered under the current assessment
(Biodynamics, 1980a,c; Quast et al., 1980a). At that time, no PBTK models were available for
generating internal dose metrics or informing interspecies extrapolation; therefore, interspecies
2/3 1
extrapolation was accomplished using BW scaling. The final slope factor of 5 x 10" (mg/kg-
day)"1 was the geometric mean of the slope factors derived from each of these three studies, i.e.,
4 x 10"1 (mg/kg-day)"1 (Biodynamics, 1980a), 4 x 10"1 (mg/kg-day)"1 (Biodynamics, 1980c), and
10 x 10"1 (mg/kg-day)"1 (Quast et al., 1980a).
5.4.4.2. Inhalation Unit Risk
Employing human data, the ECoi and LECoi, defined as the AN exposure concentration
and its associated 95% lower confidence limit, respectively, that result in a 1% increase in the
risk of dying from lung cancer at age 80 were estimated from the Cox regression model derived
from the Blair et al. (1998) occupational epidemiology study. These ECoi and LECoi values
347
DRAFT - DO NOT CITE OR QUOTE
-------
"3
were 0.992 and 0.238 ppm, respectively (or 2,187 and 524 (J,g/m , respectively). From the
LECoi, an IUR of 4.2 x 10"2 ppm"1 or 2 x 10"2 (mg/m3)"1 was derived.
For inhalation exposures to AN in animals, as with oral exposures, estimates of composite
risk addressing multiple tumor sites (within strain and sex) based on BMD modeling results from
a chronic inhalation rodent bioassay in Sprague-Dawley rats (Dow Chemical 1992a; Quast et al.,
1980b) were generated. These estimates were derived employing the same Bayesian approach
described in Appendix B-6 for oral exposures.
Table 5-21 presents the human equivalent IURs for both site-specific risks and composite
risks from multiple tumor types observed in each sex of Sprague-Dawley rat exposed to AN
using the internal dose metric, CEO concentration in blood (AUC for 24 hours).
Table 5-21. Estimated human equivalent composite IURs based on tumor
incidence in AN-exposed rats and predicted CEO-AUC levels in blood
Rat strain and gender
Tumor sites included
Site-specific IURa
(mg/m3)"1
Composite IUR
(mg/m3)"1
Sprague-Dawley male
Intestine
0.017
6.8 x 10"2
CNS
0.032
Zymbal gland
0.017
Tongue
0.012
Sprague-Dawley female
CNS
0.031
5.7 x 10"2
Zymbal gland
0.015
Mammary gland
0.017
"These site-specific IURs are taken from the last column of Table 5-18.
From the inhalation bioassay, male rats showed statistically significant elevated tumor
incidences at four sites (i.e., intestine, CNS, Zymbal gland, and tongue), while female rats
showed statistically significant elevations at three tumor sites (i.e., CNS, Zymbal gland, and
mammary gland), when exposed to AN via inhalation for 2 years. Under the assumption that
tumors at these sites are statistically independent, the estimated composite IUR in male rats was
2 3
6.8 x 10" per mg/m , approximately 2 times higher than the site-specific IUR estimate of 3.2 x
2 3
10" per mg/m based on the most sensitive single site in males (i.e., CNS). For female rats, the
2 3
composite IUR was 5.7 x 10" per mg/m , about 2 times higher than the site-specific estimate of
2 3
3.1x10" per mg/m from the most sensitive single site in females (i.e., CNS). These unit risks
should not be used with exposures greater than 3 mg/m (the lowest POD supporting the
composite risk), because above this level the unit risk cannot be expected to be an adequate
approximation of the dose-response relationship. The fitted dose-response relationship and
pharmacokinetic models should be used to estimate risk above this exposure level.
348
DRAFT - DO NOT CITE OR QUOTE
-------
The IUR recommended for use in estimating cancer risks associated with chronic
2 3
inhalation exposures to AN during adult stages of development is 2 x 10" per mg/m (or
5 3
2 x 10" per (j,g/m )—the value based on human data (i.e., the occupational epidemiology study
by Blair et al., 1998). See Section 5.4.4.3 regarding the application of ADAFs.
Previous IRIS assessment
In the previous IRIS assessment, an IUR was also derived based on human data. This
IUR was based on an increase in lung cancer incidence in humans occupationally exposed to
5 3
AN, as described by O'Berg (1980). A value of 6.8 x 10" per (j,g/m was derived by the
application of an RR model, which was adjusted for smoking and was based on a continuous
lifetime equivalent of occupational exposure to AN.
5.4.4.3. Quantitative Analysis of Early Life Exposure Scenarios
The mode of action for carcinogenicity of AN is determined to be mutagenic
(Section 4.7.3.1), which raises concern for increased early life cancer susceptibility. Consistent
with this possibility, two studies in Sprague-Dawley rats provide evidence of increased cancer
susceptibility associated with chronic AN exposure that begins in early periods of
development—a chronic-duration inhalation cancer bioassay (Maltoni et al., 1988) and a three-
generation drinking water reproductive toxicity study (Friedman and Beliles, 2002). This
evidence was summarized in Section 4.8.1. According to the Supplemental Guidance for
Assessing Susceptibility from Early-Life Exposure to Carcinogens (U.S. EPA, 2005b), "when
developing quantitative estimates of cancer risk, the Agency recommends integration of age-
specific values for both exposure and toxicity/potency where such data are available and
appropriate" (also see Box 5 in Figure 3 in U.S. EPA [2005b]). The available evidence is
considered below.
5.4.4.3.1. Summary of Maltoni et al. (1988). Maltoni et al. (1988) exposed pregnant Sprague-
Dawley rats to 60 ppm AN by inhalation for 104 weeks (see Section 4.2.2.2.2 for a complete
description). In addition, the offspring of these dams were exposed to AN starting at GD 12 for a
total of either 15 or 104 weeks. All animals were observed until natural death.
As noted in Section 4.2.2.2.2, Maltoni et al. (1988) reported statistically significant
increases in tumor incidences for both male and female offspring exposed for 104 weeks
compared with their controls. Female offspring showed statistically significant increases in
malignant mammary tumors, encephalic gliomas, and extra-hepatic angiosarcomas, while male
offspring showed statistically significant increases in encephalic gliomas, Zymbal gland
carcinomas, and hepatomas. While there were no statistically significant increases for the
exposed dams, their responses were slightly increased over background for most of these sites—
mammary tumors at 2.3% extra risk, encephalic gliomas at 5.6%, extra-hepatic angiosarcomas at
349
DRAFT - DO NOT CITE OR QUOTE
-------
1.9%, and Zymbal gland tumors at 3.8%. No hepatomas were observed in the dams. These
results are summarized in Table 5-22. Note that there were no concurrently exposed or control
adult male rats with which to compare the male offspring incidences; comparisons with the
female adult data involve assuming that adult male responses would have been concordant.
Although this profile of responses is similar to that seen for the female rats in Quast et al.
(1980b), the Quast study was not considered to be a suitable baseline for analyzing the male
offspring responses.
For offspring exposed for a total of 15 weeks, although there were no statistically
significantly elevated tumor incidences at any of the evaluated sites, the responses at these sites
were elevated over respective controls and were consistent with a linear or even supralinear
increase in response with increasing length of exposure to 60 ppm. Figure 5-3 shows the
responses for both sets of offspring and the dams for the sites with statistically significant
responses after chronic exposure. Note that while the extra-hepatic angiosarcoma response in the
male offspring was not statistically significant for either exposure group, the responses were of a
similar magnitude to the response of the female offspring exposed for 104 weeks. Although the
study did not have sufficient power to find individual group responses of about 5% and below
statistically significant, the mostly statistically significant trends and their consistency across
sites, as well as across sexes for the angiosarcomas, add weight to the overall conclusion of
increased early-life susceptibility to the carcinogenicity of AN.
2Assuming a continuous equivalent lifetime exposure concentration of 12.1 ppm for the administered concentration
in the Maltoni et al. (1988) study, dose-response relationships developed from the Quast et al. (1980b) inhalation
study (without considering pharmacokinetics) predict a similar incidence profile for the adult female rats, at about
2% extra risk predicted vs. 2.3% observed for mammary tumors, 10 vs. 5.6% observed for gliomas, and 0%
predicted vs. 1.9% observed for extra-hepatic angiosarcomas. The differences between the female adult rats in the
two studies are not large and can be explained by differences between animal colonies and husbandry practices,
including diet. This analysis provides some assurance that any early-life susceptibility estimates derived from
Maltoni et al. (1988) are compatible with the unit risks derived from the Quast et al. (1980b) bioassay.
350
DRAFT - DO NOT CITE OR QUOTE
-------
Table 5-22. Tumor incidences in Sprague-Dawley rats following inhalation exposure to 60 ppm AN for 104 weeks,
starting either in utero (GD 12) or as adults at 13 wks of age
Tumor site
Age at start of
exposure
Length of exposure to 60 ppm
Trend test
/>-valucb
Ratio of
GD 12 and 13-week
groups'
extra risks
at 104 weeks
0 weeks (control)
15 weeks
104 weeks
x/na
%
x/na
%
Extra risk
x/na
%
Extra risk
Female rats
Malignant mammary
tumors
GD 12
8/149
5.4
4/60
6.7
1.4
9/54
16.7
11.9
0.008
5.1
13 wks
2/60
3.3
-
-
-
3/54
5.6
2.3
>0.05
Encephalic gliomas
GD 12
2/149
1.3
2/60
3.3
2.0
10/54
18.5
17.4
<0.001
3.1
13 wks
0/60
0.0
-
-
-
3/54
5.6
5.6
>0.05
Extrahepatic
angiosarcomas
GD 12
0/149
0.0
1/60
1.7
1.7
3/54
5.6
5.6
0.005
3.0
13 wks
0/60
0.0
-
-
-
1/54
1.9
1.9
>0.05
Male rats
Encephalic gliomas
GD 12
2/158
1.3
3/60
5.0
3.8
11/67
16.4
15.3
<0.001
2.8
13 wksc
0/60
0.0
-
-
-
3/54
5.6
5.6
>0.05
Zymbal gland
carcinomas
GD 12
2/158
1.3
4/60
6.7
5.5
10/67
14.9
13.8
<0.001
3.5
13 wks
1/60
1.7
-
-
-
3/54
5.6
3.8
>0.05
Hepatomas
GD 12
1/158
0.6
1/60
1.7
1.0
5/67
7.5
6.9
0.002
NA
13 wks
0/60
0.0
-
-
-
0/54
0.0
0.0
>0.05
Extrahepatic
angiosarcomas
GD 12
1/158
0.6
3/60
5.0
4.4
3/67
4.5
3.9
>0.05
2.1
13 wks
0/60
0.0
-
-
-
1/54
1.9
1.9
>0.05
11 x = number of animals with specified tumor, n = number of animals at start of exposure.
bCochran-Armitage trend test for datasets with 3 or more groups (early-life exposures); Fisher's exact test otherwise (adult exposures).
°Values in italics rely on adult female rat data; no male rats were exposed starting at age 13 wks in this study.
Source: Maltoni et al. (1988).
351
DRAFT - DO NOT CITE OR QUOTE
-------
0.30
0.30
Malignant mammary tumors
—~—female offspring
—A— female adults
Gliomas
—~—female offspring
-A—female adults
-¦—male offspring
Weeks of exposure
Weeks of exposure
—¦—male offspring
- A-- female adults
-female offspring
-female adults
-male offspring
Weeks of exposure
Extra-hepatic angiosarcomas
Zymbal's gland tumors
Weeks of exposure
.30
Hepatomas
.25
—¦— male offspring
--A-• female adults
.E 0.20
.15
.10
.05
.00
¦10
0
10
20
30
40
50
60
70
80
90
100 110
Weeks of exposure
Error bars denote 95% CIs for each response.
Figure 5-3. Comparison of tumor responses to inhalation exposure of
acrylonitrile, by age at start of exposure (offspring at GD 12, adults at
13 weeks), sex, and length of exposure, for Sprague-Dawley rats (Maltoni et
al., 1988).
352
DRAFT- DO NOT CITE OR QUOTE
-------
Figure 5-3 shows linear trends with increasing length of exposure among the offspring
for malignant mammary tumors (females), gliomas (males and females), extrahepatic
angiosarcomas (females) and hepatomas (males), while supralinear trends are suggested for
extrahepatic angiosarcomas and Zymbal gland tumors in males. Linear trends suggest that
offspring responses to the 15-week exposure are proportional to offspring responses resulting
from the chronic exposure; that is, that the expected response is about 1/7 (15 weeks/104 weeks)
of the response to chronic exposure starting in utero. The overall response after 104 weeks of
exposure is markedly higher than that of the chronically exposed adult rats, suggesting that once
the susceptible process (or processes) has started it continues well into adulthood. For each site,
the ratios of the extra risk associated with chronic exposure starting at GD 12 relative to starting
at age week 13 ranged from about 3 to 5 for individual sites in female rats, supported by a factor
of 3 for gliomas in male rats (see Table 5-22).
In contrast, the supralinear trends suggest greater than proportional susceptibility in the
early-life period which decreases over time, or which does not occur after the early-life period,
as possibly illustrated by the male offspring angiosarcomas. On the other hand, the 15-week
period used in this study includes a nontrivial period that cannot be considered early-life, given
that rats are generally considered sexually mature by about 8 weeks of age. Consequently the
magnitude of the early-life sensitivity, at least for two tumor sites, may be underestimated by
these data without further characterization of the most sensitive timeframe. The low numbers of
responses from the 15-week exposures (1-4 offspring and 0-3 adults responding per site) likely
impact the resolution of early-life susceptibility estimates for these sites. From an overall point
of view, however, note that the 15-week responses considered collectively are generally similar
to the female adult responses following 104 weeks of exposure. Because these early-life
responses result from sevenfold less exposure, they demonstrate approximately 7 times (or
approximately one order of magnitude) more sensitivity than the adult responses.
5.4.4.3.2. Summary of Friedman and Beliles (2002). Supporting evidence of increased early-
life susceptibility was also seen in the three-generation reproduction study reported by Friedman
and Beliles (2002). There were statistically significant increases in Zymbal gland carcinomas
and astrocytomas after approximately 51 weeks of exposure to 500 ppm AN in drinking water
for F1 Sprague-Dawley female rats relative to F1 controls, while the incidences for F0 female
rats were not significantly elevated at 500 ppm. The incidences for F2 female rats, exposed in
utero like the F1 females, were not considered to be significantly elevated. Section 4.2.1.2.8
provides details of the study design, and Table 5-23 summarizes the results.
353
DRAFT- DO NOT CITE OR QUOTE
-------
Table 5-23. Tumor incidences in female Sprague-Dawley rats exposed to
AN in drinking water for approximately 46 weeks, starting either in utero
or at 5 weeks of age
Tumor site
Generation
Tumor incidence, and percent, at AN in
drinking water, ppm
Trend
test
/j-valuc"
Ratio of
early life +
adult to adult
extra risks at
500 ppm
0
100
500
Brain
astrocytomas
F0 (adult exposure)
0/19(0%)
1/20 (5%)
2/24 (8%)
0.247
-
Fib
0/20 (0%)
1/19 (5%)
4/17 (24%)
0.005
2.8
F2b
0/20 (0%)
1/20 (5%)
1/20 (5%)
0.506
-
Fib +F2b
0/40 (0%)
2/39 (5%)
5/37 (14%)
0.007
1.8
Zymbal gland
F0 (adult exposure)
0/19(0%)
0/20 (0%)
2/24 (8%)
0.036
-
Fib
0/20 (0%)
2/19(11%)
3/17 (18%)
0.088
2.1
F2b
0/20 (0%)
0/20 (0%)
3/20 (15%)
0.007
1.9
Fib +F2b
0/40 (0%)
2/39 (5%)
6/37 (16%)
0.005
2.0
aCochran-Armitage test for linear trend.
Sources: Friedman and Beliles (2002); Litton Bionetics (1992).
Ratios of extra risks of astrocytomas and Zymbal gland tumors were estimated for the
500 ppm Fib and F2b groups if there was a statistically significant linear trend. These results
suggest an increased early-life cancer susceptibility of about two- to threefold at the high dose.
However, these results do not support a quantitative estimate of relative susceptibility because
several factors contribute to underestimating any susceptibility. First, with only 20 animals per
treatment group, this study had limited power to detect tumor increases, relative to the Maltoni et
al. (1988) study and most other carcinogencity studies, with the consequence that any increased
susceptibility at the low dose (100 ppm) may be less readily apparent. For example, an apparent
effect of 0/20 is consistent with a tumor response as high as 17%, the 95% UCL on a response of
0/20.
Second, while it is notable that tumors were observed as early as they were, after no more
than 51 weeks without another year of follow-up, these data do not inform lifetime risk estimates
adequately. Although cancer incidence is expected to increase with increasing age (Doll, 1971),
dose-response relationships starting at different ages may not be parallel. Given the short
exposure and follow-up in this study, the data from this group are better considered as supporting
early-life susceptibility qualitatively.
5.4.4.3.3 Analysis to support estimation of early-life susceptibility adjustment factors. In
summary, two data sets support the conclusion of increased susceptibility to carcinogenicity with
AN exposure in early life. One data set conducted by Maltoni et al. (1988), in female rats with
354
DRAFT- DO NOT CITE OR QUOTE
-------
chronic exposure starting in early life or starting as adults, provides a sufficient basis for
developing quantitative estimates of susceptibility. Due to the lack of adult male referent groups
(control or exposed), the male offspring data provide support for early-life susceptibility. The
female data supported point estimates for susceptibility adjustment factors of 5, based on
malignant mammary tumors, and 3, based on gliomas or extra-hepatic angiosarcomas. In other
words, cancer potency for specific tumor sites following chronic exposure to 60 ppm AN appears
to be three- to fivefold higher if exposure starts in early-life rather than during adulthood.
Before recommending ADAFs based on these data, important considerations include the
impact of experimental variability and overall risk encompassing multiple tumors. Given the
incomplete understanding of the operant modes of action, it is unknown whether the tumor types
observed in the Maltoni et al. (1988) study are independent of each other. They were not
reported as metastases, which would clearly not be independent. Under the assumption that
these tumor sites are independent, an analysis was undertaken focusing on overall risk as well as
characterizing the impact of experimental variability.
A Monte Carlo analysis was conducted to obtain the distribution of the ratio of combined
extra risk from the three tumor sites for the chronically exposed female offspring to the
analogous combined extra risk from adult-only exposure. For each simulation, the outcome in
each control and dose group was simulated by drawing a binomial B(n,p) random variable with
n equal to the number of animals in the group and p obtained by dividing number of animals with
tumor in the original dataset by the total number of animals. Since data for only one dose group
was available for each tumor site, the individual binomial probabilities for this simulation run
were estimated by dividing the number of animals with tumors by the total number of animals,
rather than by fitting a dose-response model to the responses for each pair of control and exposed
group. Then the risks from the three individual tumor sites were combined assuming
independence of different tumor outcomes, as noted above. The standard formula for the
probability of the union of multiple sets was used to compute the combined probability of tumor
(dose d = 60 ppm was the same for all three tumor types):
PAorBorc(d) = PA(d) + PB(d) + Pc(d) - PA(d) * PB(d) - PA(d) * Pc(d) - (1)
pB(d) * pc(d)+pA(d) * pB(d) * pc(d).
Then the combined extra risk was computed as:
ER-AorBorc(d) = [PAorBorc(d) ^Aorliorc''^)]^ [1 " PAorBoic(O)]- @)
Only the individual extra risks where probabilities in the control group did not exceed the
probability in dose group were retained, in order not to create negative estimates of extra risk. If
355
DRAFT- DO NOT CITE OR QUOTE
-------
only two tumor results were combined in a particular realization, the combined probability in
equation (1) was adapted for two tumors
PAorB(d) = PA(d) + PB(d) - PA(d) * PB(d).
Finally, the ratio of the extra risks is given by:
ERAorBorc(d I adult & early-life)/ ERAorBorC(d | adult only) (3)
was calculated and retained for each simulation. The simulation was repeated 10,000 times and
results are summarized in the first row of 5-24. Simulations were done in R language
(www. r-proj ect. org).
Because some of the control groups did not have any observed tumors, the approach
above assigns no variability to such groups. As historic background rates are not available, a
sensitivity analysis was conducted to evaluate if assigning a small probability of tumor (l/2n) to
these control groups and the resulting variability influences the distribution of the ratio (eq 3).
The simulations were performed as described above. Results are summarized in the second row
of Table 5-24.
Table 5-24. Overall extra risks of multiple tumor incidence, for chronic
exposure beginning either in early-life or in adulthood; based on Sprague-
Dawley rats exposed to AN
Treatment of 0%
tumors in
referent groups
Start of
exposure period
Number of simulations
with k tumors retained
Percentiles of the distribution of the ratio of
extra risks
k= 3
k= 2
k= 1
5th
Median
Mean3
95th
Zero background
as observed
Adult
7,578
2,422
0
0.037
0.095
0.100
0.178
Early-life
9,903
97
0
0.204
0.314
0.314
0.423
Ratio
1.543
3.232
4.219
(9,957)
9.628
Zero background
replaced by l/2nb
Adult
5,483
3,923
574
0.021
0.085
0.089
0.166
Early-life
9,685
314
1
0.199
0.311
0.311
0.421
Ratio
1.610
3.594
6.373
(9,929)
15.627
aFor calculation of means, the figure in parentheses indicates the number of finite ratios; the other percentiles were
based on 10,000 simulations.
bForn = 60, l/2n = 1/120 = 0.0083.
Source: Maltoni et al. (1988).
Extra risks for chronic exposure starting in early-life are not sensitive to the assumption
regarding background probabilities, as shown by the similar percentiles for both cases. But the
356
DRAFT- DO NOT CITE OR QUOTE
-------
extra risks for adult-only exposure decrease slightly when some variability at 0% response is
assumed, resulting in slightly higher ratios. Because the second analysis is more realistic,
allowing a background response in two of the sites and addressing plausible variability, it is used
to derive the data-specific adjustment factors. Ninety-five percent of the values fell between
1.6 and 15.6.
Note that the analogous composite adjustment factor resulting from the default factors in
the SG for chronic exposure beginning in early life is 1.7; the analysis provides a distribution of
adjustment factors with about 95% of the possible values being larger than the default adjustment
factors. The median and mean ratios, at 3.6 and 6.4, respectively, are far enough apart to
indicate a nonsymmetrical distribution. The median of 3.6 is a better estimate of central
tendency in this case.
Recommended early life susceptibility factors. The derived early-life susceptibility factor
of 3.6 is to be used in assessing oral or inhalation exposures to AN that begin at birth (or earlier)
and extend to or beyond 70 years of age. When exposures involving shorter periods of early-life
exposure to AN are assessed, a modification consistent with the SG framework is recommended,
as outlined in this section.
The available information for identifying appropriate ADAF(s) to apply to specific early-
life periods is limited. While U.S. EPA's SG (U.S. EPA, 2005b) separates the pre-adult period
into two age ranges, 0-<2 and 2-<16 years, the Maltoni et al. (1988) study design did not
address age periods corresponding to these windows of human susceptibility. The offspring
exposed for 15 weeks were about 14 weeks old at the time their exposure stopped (estimated by
subtracting 9 days (GDs 21-12) from 15 weeks); female rats reach sexual maturity at about 8
weeks (35 days), so 14 weeks in rats would correspond to a higher age in humans than 16 years.
Therefore, the Maltoni et al. (1988) data can support only one ADAF for the entire pre-adult
window.
An ADAF for the pre-adult window can be derived from the derived lifetime estimate by
treating it as a weighted sum of ADAFs for separate windows of susceptibility, consistent with
the general framework in the SG for applying the default ADAFs, using a simplification of the
SG framework for estimating human risk from exposure including early-life periods:
ADAF for lifetime exposure starting at birth = 3.6
= [ADAFp x tP + ADAFa x (70 yr - tP)]/70, (4)
where the subscripts P and A denote pre-adult and adult, respectively, and tp = duration (in years)
of exposure during the pre-adult period. By convention, ADAFa must be 1. With two
unknowns, ADAFp and tp, one value must be inferred so that the other can be estimated.
Because the Maltoni et al. (1988) data from offspring exposed for 15 weeks are consistent with a
10-fold increased sensitivity relative to rats exposed only as adults (see Figure 5-3 and
accompanying text), and the default ADAF of 10 in the SG for the most sensitive period (0-
357
DRAFT- DO NOT CITE OR QUOTE
-------
<2 years of age) is also 10, an ADAFP of 10 seems a reasonable starting point. Substituting these
two ADAFs in (eq 4) and solving for tp leads to an empirical age cutpoint of 20 years.
This cutpoint is necessarily somewhat arbitrary—it is not an assertion that 14 weeks of
age in a rat is equivalent to 20 years in a human—but appears to be a straightforward
interpretation of the available data. Other interpretations are possible, such as that sensitivity at
the earliest ages (i.e., before 2 years of age) could be relatively higher than later pre-adult ages,
but further manipulation of the AN-specific data to align more closely with the SG early-life
periods and relative sensitivities (i.e., that the earlier period might be about 3 times more
sensitive that the second pre-adult period) is judged to be too speculative.
Illustrations of applying the derived ADAFs for estimating human cancer risk from
lifetime exposures starting at birth are provided in Table 5-25.
Table 5-25. Application of ADAFs for estimating human cancer risk from
70-year oral or inhalation exposures to AN from ages 0 to 70
Age Group
(yrs)
Duration adjustment
ADAF
Age group risk
For oral exposure to 0.01 mg/kg-d, with oral slope factor of 5 (mg/kg-day)"1 for adult lifetime exposure:
0-<20
20 yrs/70 yrs = 0.29
10
0.14a
20-70
50 yrs/70 yrs = 0.71
1
0.04
Total: 0-70
70 yrs/70 yrs = 1
3.6
0.18
For inhalation exposure to 1 mg/m3, with IUR of 2 x 10"2 (mg/m3)"1 [or 2 x 10"5 (jig/m3)-1] for adult lifetime
exposure:
0-<20
0.29
10
0.057
20-70
0.71
1
0.014
Total: 0-70
1
3.6
0.072
aAge group risk = duration adjustment x ADAF x exposure x CSF or unit risk.
Equation (4) provides the framework for estimating risks from exposures involving shorter early-
life periods. Two examples are provided:
• If calculating the cancer risk for a 30-year exposure to a constant oral exposure level of
0.01 mg/kg-day from ages 0 to 30, the duration adjustments would be 20/70 and 10/70,
and the age group risks would be 0.14 and 0.01, resulting in a total risk estimate of 0.15.
• If calculating the cancer risk for a 30-year exposure to a constant oral exposure level of
0.01 mg/kg-day from ages 20-50, the duration adjustments would be 0/70 and 30/70, and
the age group risks would be 0 and 0.02, resulting in a total risk estimate of 0.02.
358
DRAFT- DO NOT CITE OR QUOTE
-------
5.4.4.3.4. Uncertainties in early-life susceptibility estimates. Unlike many of the studies
supporting the default adjustment factors recommended in the SG, the available early-life studies
for AN include gestational exposure. The pharmacodynamics and pharmacokinetics of AN
during the perinatal period are not well understood, nor are the time frame and nature of the
response leading to the apparent susceptibility; it is possible that gestational exposure is a
necessary component of the increased susceptibility.
Regarding the level of exposure, it is not known whether perinatal exposure was
comparable to adult exposure. If perinatal exposure was greater than adult exposure, the
apparent susceptibility would be at least partly attributable to higher exposure; if lower, the
apparent susceptibility would be underestimated.
Other uncertainties associated with the recommended data-derived early-life
susceptibility adjustment factor include: (1) whether the relative susceptibility is dose-
dependent, leading to different values for other parts of the dose-response range; (2) whether the
relative susceptibility estimated from female data characterizes the relative susceptibility for
males; note particularly the observation of hepatomas in male offspring exposed chronically
beginning in early life; (3) the relevance of extrapolating these results to humans; and
(4) possible underestimation of risk associated with the lack of knowledge of individual times of
death if tumors in the exposed animals tended to occur earlier than in control animals or if
tumors in the animals exposed perinatally tended to occur earlier than in those exposed starting
at age 13 weeks.
Some uncertainty in cancer risk estimates can generally be characterized by fitting a
dose-response model to the incidence data in order to provide a CI on the risk estimate. In this
case, however, there was only one exposed group for each exposure pattern. Other methods for
characterizing the statistical uncertainty in these data are currently being investigated.
5.4.5. Uncertainties in Cancer Risk ValuesRisk estimates have inherent uncertainties. This
subsection discusses the uncertainties that may be associated with cancer risk values for oral and
inhalation cancer assessments.
5.4.5.1. Oral Cancer Assessment
Uncertainties related to the oral cancer assessment are discussed below and summarized
in Table 5-26 and Figure 5-4.
359
DRAFT- DO NOT CITE OR QUOTE
-------
Table 5-26. Summary of uncertainty in the AN oral cancer risk assessment
Consideration/
approach
Impact on cancer risk
estimate
Decision
Justification
Low-dose
extrapolation
procedure
The selected model does not
represent all possible models
one might fit, and other
models could conceivably be
selected to yield more
extreme results consistent
with the observed data, both
higher and lower than those
included in this assessment.
Multistage model to
determine POD, linear
low-dose extrapolation
from POD.
Available mode of action data support
direct mutagenicity as the key mode of
action and low-dose linear extrapolation
approach. EPA's Guidelines for
Carcinogen Risk Assessment (U.S. EPA,
2005a): mutagens "are assessed with a
linear approach." Mutagenic mode of
action functions systemically at multiple
tumor sites.
PBTK model
Oral slope factor based on
PBTK modeling and internal
CEO levels is about sixfold
higher than slope factor
based on the default
approach.
EPA-revised model was
used.
The revised PBTK model includes EH
activity in rats and provides better
estimates of internal dose.
Dose metric
Alternatives could decrease
oral slope factor (e.g., use of
AN-AUC instead of CEO-
AUC could lower slope
factor by 30-fold).
CEO-AUC was used.
CEO is the reactive metabolite that binds
to DNA and initiates tumor formation.
AN is not the causal agent for cancer.
Statistical
uncertainty at
POD
Oral slope factor will be
reduced 1.3-fold ifBMD
used as POD instead of
BMDL.
BMDL (approach for
calculating reasonable
upper bound).
Limited size of bioassay results in
sampling variability; BMDL is lower
95% confidence limit of BMD.
Bioassay
Oral slope factor would be
the same if study on F 3 44
rats were used (Johannsen
and Levinskas, 2002b).
Quast (2002) study on
Sprague-Dawley rats
was used.
Quast (2002) has separate interim
sacrifice groups and examined more
endpoints than the Biodynamics (1980c)
data sets.
Species/gender
combination
Human risk could decrease
or increase, depending on
relative sensitivity.
Oral slope factor
derived from study on
female Sprague-
Dawley rats.
It was assumed that humans are as
sensitive as the most sensitive rodent
gender/species tested; true
correspondence is unknown. The
carcinogenic response occurs across
species. Generally, direct site
concordance is not assumed; consistent
with this view, some rodent tumors are
not found in humans (e.g., Zymbal's
gland tumors, Harderian gland tumors)
and rat and mouse tumor types also
differ.
Human population
variability in
metabolism and
response/sensitive
subpopulations
Low-dose risk increase to an
unknown extent.
Considered
qualitatively.
Increased early life susceptibility
demonstrated in rat studies. Variability
in human susceptibility likely exists due
to differences in microsomal CYP2E1
activities and possibly GST activity.
360 DRAFT- DO NOT CITE OR QUOTE
-------
6
S 4-
o
ro
O ra
¦c
dn 3 ¦
O)
E
.2 8.
~
0.40
5.0
1.03
~ CSF - CEO in Blood
• CSF - Administered Dose
3.1
0.68 -
1.9
0.36
Male Sprague-Dawley
Rats
Female Sprague-
Dawley Rats
Male F344 Rats Female F344 Rats
Sex/Strain/Species
Figure 5-4. Comparison of composite oral CSFs derived from tumor
incidence data in four different sex/strain/species of rats exposed chronically
to AN. For each sex/strain/species combination, two different dose metrics
were employed: (1) CEO concentration in blood, and (2) human equivalent
administered dose.
Choice of study
Several bioassays by the oral route are available. The 2-year drinking water studies of
Sprague-Dawley rats (Quast, 2002; Quast et al., 1980a) and F344 rats (Johannsen and Levinskas,
2002a; Biodynamics, 1980a) were selected as principal studies. These two studies have
sufficient numbers of animals, investigate a thorough set of endpoints, and are considered well
conducted. The tumor types observed and incidence of treatment-related tumors were similar in
these two studies. Moreover, cancer risk estimates from these two studies differed only by a
factor of 1.5. Similarities in these results increase the level of confidence of estimated cancer
risk. The Quast (2002) study of Sprague-Dawley rats investigated more endpoints and is
therefore a slightly stronger study.
361
DRAFT- DO NOT CITE OR QUOTE
-------
PBTK model
Use of the PBTK model to extrapolate from rodents to humans also introduces some
uncertainty. An extensive quantitative analysis of the PBPK model uncertainty is presented in
Appendix D. Briefly, PBPK models are computational tools used to predict chemical/drug
disposition and are comprised of three distinct types of information: physiological,
physicochemical, and biochemical. The physiological data are independent of chemical-specific
data, and describe such parameters as organ volumes and blood flows. Physicochemical
parameters are chemical-specific and specify parameters, such as PCs or permeability.
Biochemical parameters define the rates of chemical transformation or binding. Because the
physiological data are considered to be well-characterized, analysis focused on uncertainties in
the chemical-specific parameters employed in the PBPK model used to predict AN dosimetry in
humans, which is adapted from that of Sweeney et al. (2003). (Only a small number of
(metabolic) parameters were changed in the EPA's adaptation.) Uncertainty analysis and
characterization of the human model lead to the following conclusions.
• Blood:air and tissue:air PCs as measured directly (in rodents) vs. estimated by a
computational tool were compared, and the impact on model predictions of peak AN and
peak CEO levels in brain and blood was found to be less than 30% (difference between
predictions using the alternate sets of PCs).
• Comparison of model predictions to the limited human data of Jakubowski et al. (1987)
shows that the model over-predicts the inhalation respiratory retention measured
(predicted ~ 70%; measured 44-58%). This over-prediction may be due to the fact that
the model does not fully describe the exposure apparatus (which can introduce additional
airway "dead-space") or that the model does not describe gas absorption/desorption in the
conducting airways that can reduce uptake rates. However, since the error (if any) is in
the direction of over-prediction of uptake, the error is health-protective.
• Once absorbed, the model predicts that 8.4% of AN is converted to CEO and then
hydrolyzed, while 30% is converted to CEO and then conjugated with GSH. These
values bracket the observation of 16.3% of retained AN being excreted in urine as "CEO"
(after acid extraction). Thus, this limited observation is in line with model predictions
subsequent to AN predictions, though the exact level of agreement or error cannot be
determined due to the distinction between observed quantity (CEO in urine) and what the
model predicts (CEO metabolic rates).
• The PBTK model over-predicts CEO levels in the rat at early time-points after i.v.
injection (Figure 3-3) and in blood and brain at all time points measured after oral
exposure (Figure 3-5a). Kedderis et al. (1996) suggest that the overestimation of CEO by
the rat model at the early time points (which also occurs in the EPA's revision) may be
due to an intrahepatic first pass effect, as occurs with other epoxides formed in situ from
362
DRAFT- DO NOT CITE OR QUOTE
-------
their parent olefins (Filser and Bolt, 1984). However, this explanation is unlikely and a
more plausible explanation is that the model does not account for time-dependence in
GSH levels. Inclusion of GSH dynamics would be a much more significant and intensive
change to the model structure than single variation implemented here (addition of EH to
the rat model with parameter re-estimation) and so has not been considered.
• Parametric sensitivity analysis of human predictions showed a shift in importance
between the EH and GSH pathways for CEO elimination from being approximately equal
in the implementation of Sweeney et al. (2003) to lower significance for EH (now low
but significant) and high significance for GSH with the revised parameters. The
oxidation of AN to CEO has a significant effect on AN predictions but only a small effect
on CEO in both model versions, reflecting the high dependence of CEO concentrations
on CEO metabolic removal. Not surprisingly, brain:tissue PCs significantly affect
predictions in brain tissue, and the rate constant for oral absorption affects peak AN
concentrations in blood, but not the AUC. While the human parameters are largely
derived from human in vitro data, which gives a much higher confidence than would
occur had they only been extrapolated from rats, the high dependence of CEO
concentrations on GSH conjugation rates, together with the hypothesis above regarding
the lack of GSH dynamics in the model, point to that pathway description in particular as
being the greatest source of quantitative uncertainty.
• Estimated coefficients of variation for predicted concentrations in brain and blood after
inhalation exposure are -0.6-0.7 for AN and 0.9-1.2 for CEO; after oral exposure these
are 0.8-0.9 for AN and 0.7-1.0 for CEO. Based on these values, the human model
predictions are expected to be accurate to within a factor of approximately 3, which is the
standard assumption for pharmacokinetic variability among humans.
It can be noted that the oral CSFs are larger when based on blood levels of CEO versus
blood levels of AN. The VmaxC/Km for AN oxidation in humans is estimated to be about
10 times higher than in the rat. AN enzymatic GSH conjugation is estimated to be 1.5 times
higher in the rat, but the 2nd-order removal constant is the same and the GSH tissue levels are
roughly the same (and lower in the larger tissue groups). The result, for example, with a steady
oral "infusion" of 30 mg/kg-day, is that AN is removed somewhat faster overall, leading to
steady-state blood level of 0.114 mg/L in the human vs. 0.151 in the rat. However, a much
higher portion of this goes to CEO in the human yielding a CEO blood level of 0.297 vs.
0.013 mg/L in the rat. Thus, CEO:AN = 0.088 in the rat and 2.6 in the human so that the
CEO-based risk estimate is higher than the AN-based estimate.
363
DRAFT- DO NOT CITE OR QUOTE
-------
Dose metric
AN is activated by CYP2E1 into its reactive metabolite, CEO, which binds to DNA and
initiates tumor formation. Therefore, AUC CEO concentration in blood was selected as the
internal dosimetric for PBTK modeling. AN is not anticipated to be the causal agent for
carcinogenesis. Uncertainty in the risk estimate related to the dose metric is primarily associated
with PBTK model estimation.
Statistical uncertainty at the POD
Parameter uncertainty within the chosen model reflects the sample size of the cancer
bioassay. For the multistage model applied to this data set, there is a relatively small degree of
uncertainty at the BMDLio (the POD for linear low-dose extrapolation), which is approximately
1.4-fold lower than the BMDio (Appendix B, Table B-37). The highest value was selected in
order to provide a reasonable upper-bound risk estimate.
With regard to the combined risk estimate, under the assumption of independence of the
tumor type/site considered, no additional uncertainty is added to the estimated POD. Each
combined estimate is a statistically rigorous restatement of the statistical uncertainty associated
with the risk estimates derived from the individual sites. The only assumption in the combining
tumors approach is independence of tumors. This assumption is consistent with NRC (1994)
recommendations.
Choice of low dose extrapolation approach
The mode of action is a key consideration in deciding how risks should be estimated for
low-dose exposure. The mode of action for cancer effects of AN is discussed extensively in
Section 4.7.3.1 and is determined to be mainly due to direct mutagenicity. Other modes of action
are plausible, but the evidence suggests they may not be key to the carcinogenicity of AN. The
pattern of tumors is consistent with DNA-reactive chemicals. When the mode of action is
determined to be direct mutagenicity, a linear approach is used to estimate low-exposure risk, in
accordance with EPA's Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a).
Choice of species/gender
No human cancer epidemiology study via the oral route is available. Cancer risk is
estimated for both male and female rats, and cancer risk for the strain and gender with the
highest risk, Sprague-Dawley females, was selected. Mammary gland tumors were observed in
female rats of both studies. Although a cancer bioassay in B6C3Fi mice is also available, the
exposure was via gavage and no PBTK model is available for mice. It was assumed that humans
are as sensitive overall as rats, although the true correspondence is unknown. Site concordance
is not assumed, nor is it necessary (U.S. EPA, 2005a).
364
DRAFT- DO NOT CITE OR QUOTE
-------
Human population variability
Heterogeneity among humans is another source of uncertainty, in that human risk can be
higher or lower than that estimated from rats, depending on relative sensitivity. Human genetic
polymorphisms in CYP2E1 activities likely contribute to variability in susceptibility to the toxic
effects of AN (see Section 4.8.4.1). In addition, increased early-life susceptibility was
demonstrated in two studies in rats exposed gestationally (see Section 5.4.4.3).
5.4.5.2. Inhalation Cancer Assessment
Uncertainties related to the IUR assessment are discussed below and summarized in
Table 5-27 and Figure 5-5.
Table 5-27. Summary of uncertainty in the AN inhalation cancer risk
assessment
Consideration/
approach
Impact on cancer risk
estimate3
Decision
Justification
Choice of study
Cancer risk estimate
could increase because
only lung cancer deaths
evaluated.
Lung cancer mortality and
exposure data from Blair et al.
(1998) study was used to derive
IUR.
IUR derived from best available
cancer epidemiological study of AN
exposed workers has fewer inherent
uncertainties than IUR derived from
rat study (Quast et al., 1980b).
Low-dose
extrapolation
procedure
The selected model does
not represent all possible
models one might fit, and
other models could
conceivably be selected
to yield more extreme
results consistent with the
observed data, both
higher and lower than
those included in this
assessment.
Low-dose linear.
EPA's Guidelines for Carcinogen
Risk Assessment (U.S. EPA, 2005a):
mutagens "are assessed with a linear
approach."
Statistical
uncertainty at
POD
IUR would be reduced
fourfold if BMD used as
POD instead of BMDL.
BMDL, approach for
calculating reasonable upper
bound.
Limited size of study results in
sampling variability; BMCL is
lower 95% confidence limit on
ECoi.
Choice of
species/gender
IUR estimated from
exposed male and female
workers.
IUR estimated from entire
cohort was used.
No apparent gender difference
between male and female workers;
male and female rats were also
similar.
Human population
variability in
metabolism and
response/sensitive
subpopulations
Low-dose risk for general
population can increase to
an unknown extent due to
underestimation of human
variability (healthy
worker effect).
Semi-parametric Cox
regression model with time-
dependent covariates was used.
The Cox model allows
inclusion of individual
exposure histories and utilizes
internal controls, thus avoiding
confounding by the healthy
worker effect.
Increased early-life susceptibility
demonstrated in rat studies. Human
variability in susceptibility likely
exists due to differences in
microsomal CYP2E1 activities, and
possibly GST activity.
365
DRAFT- DO NOT CITE OR QUOTE
-------
0.100
0.090
0.080
0.070
£ 0.060
*¦%
.2 E
? c
0.050
JH 0.040
c
0.030
0.020
0.010
0.000
r
•
~ IUR - CEO in Blood
• IUR - Human Equivalent Administered Concentration
0.070
~
w
0.060
A
~
0.020
¦
U.0U40
•
0.0035
•
Male Sprague-Dawley Rats Female Sprague-Dawley Rats
Humans Exposed
Occupationally
Sex/Strain/Species
Figure 5-5. Comparison of composite IURs derived from: (1) tumor
incidence data in male and female Sprague-Dawley rats exposed chronically
to AN, and (2) neurological effects in humans exposed to AN occupationally.
In deriving the animal-based IURs, two different dose metrics were
employed: (1) predicted CEO concentration in blood, and (2) human
equivalent administered AN concentration in air.
Choice of study
As mentioned previously, the Blair et al. (1998) cohort study is the largest cohort
assessment of the relationship between AN exposure and cancer. This study had the distinct
advantage of quantifying exposure and using an internal control group of unexposed workers, all
factors identified as shortcomings in previous studies. Information on smoking history, a
potential confounder when lung cancer is the outcome of interest, was available, but only for a
small subset within the Blair et al. (1998) cohort. Blair et al. (1998) limited their analysis to the
white male population and noted that adjustment for smoking reduced the risk of lung cancer
slightly. The data used in this assessment to derive the IUR included the entire cohort;
consequently, the smoking history data were incomplete for this analysis, leading to an area of
uncertainty surrounding this risk estimate.
366
DRAFT- DO NOT CITE OR QUOTE
-------
Another source of uncertainty stems from the use of only lung cancer mortality for the
derivation of the IUR. As previously described, other studies reported small, but not consistent,
excess risks for other types of cancer (prostate, bladder, brain cancers) associated with AN
exposure. Thus, basing the IUR on only lung cancer mortality may underestimate the
carcinogenic potential of AN. Further, as the IUR is based on mortality rather than lung cancer
incidence, though in the case of lung cancer mortality is a good surrogate of lung cancer
incidence, there may be potential for underestimation of carcinogenic potential with this
approach. Additional uncertainties related to the statistical analysis of these data are further
discussed in Appendix B-7. Briefly, uncertainties of the statistical approach include: (1) Cox
model that was fit to the data is not a biologically based model; (2) the estimator of the
cumulative hazard does not account for the covariate path and hence is only an approximation;
and (3) the estimate of risk is obtained using the first-order "linearized" approximation.
However, obtained results are consistent with assumptions of the first-order approximation
validity.
Other uncertainties associated with the NCI/NIOSH cohort include nondifferential
exposure misclassification, lung cancer misdiagnosis among the internal controls, the relatively
short follow-up period (reflected by the relatively small proportion of mortality within the
cohort), and the extrapolation of continuous environmental exposure from 8-hour occupational
exposure without consideration of potential recovery mechanisms between daily exposures. The
nondifferential exposure misclassification contributes to an underestimation of risk, while the
impacts of short follow-up and extrapolation to a continuous exposure scenario are unclear.
Nonetheless, the NCI/NIOSH cohort provides the most robust data set in terms of sample size
and exposure assessment for the derivation of the IUR.
Statistical uncertainty at the POD
Parameter uncertainty within the chosen model reflects the limited sample size of the
cancer bioassay. For the results relying on the cohort study, the ECoi is approximately fourfold
higher than the LECoi.
For the multistage model applied to the rat data, in support of the human-based results,
there is a reasonably small degree of uncertainty at the 10% incidence level (the POD for linear
low-dose extrapolation). Composite BMCios for overall cancer risk are approximately 1.3-fold
higher than their corresponding BMCLi0s for both male and female rats.
With regard to the combined risk estimate, under the assumption of independence of the
tumor type/site considered no additional uncertainty is added to the estimated POD. Each
combined estimate is a statistically rigorous restatement of the statistical uncertainty associated
with the risk estimates derived from the individual sites. This assumption is consistent with NRC
(1994) recommendations.
367
DRAFT- DO NOT CITE OR QUOTE
-------
Low-dose extrapolation procedure
As discussed previously, the key mode of action for carcinogenicity of AN is determined
to be direct mutagenicity. A linear approach is used to estimate low-exposure risk, in accordance
with EPA's Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a). Nonlinear low-dose
extrapolation approach is not used since data do not support other modes of action at this time.
Choice of species/gender
A human study was selected for quantification of IUR. Both male and female workers
were evaluated. A 2-year inhalation study of male and female rats was used for comparison.
Estimates of IURs from the human study and from the rat studies were similar. IURs estimated
from male and female rats were also similar, although there was a suggestion of higher cancer
risk in females associated with mammary gland tumors.
Human population variability
Heterogeneity among humans is another source of uncertainty. Uncertainty related to
human variation needs consideration, also, in extrapolation from a small subset of human
population to a larger, more diverse population. The study population in Blair et al. (1998) was
AN-exposed workers in the United States. To counter the healthy worker effect, internal
controls were used in the study. Available data from animal studies provide no evidence of
gender differences in susceptibility to toxicity of AN, although no data is available regarding
possible gender differences in susceptibility. Polymorphisms in the CYP2E1 gene that affect
enzyme activity likely contribute to the variability of human susceptibility towards the toxic
effect of AN (see Section 4.8.4.1).
368
DRAFT- DO NOT CITE OR QUOTE
-------
6. MAJOR CONCLUSIONS IN THE CHARACTERIZATION
OF HAZARD AND DOSE-RESPONSE
6.1. HUMAN HAZARD POTENTIAL
AN (CASRN 107-13-1) is a colorless, flammable, and volatile liquid with a weakly pungent
onion- or garlic-like odor. It is a commercially important chemical used in the manufacture of
acrylic and modacrylic fibers, plastics (ABS and AN-styrene resins), and nitrile rubbers and as
an intermediate in the synthesis of other chemicals, such as adiponitrile and acrylamide.
Exposure to airborne AN is possible for people living in the vicinity of emission sources such as
acrylic fiber or chemical manufacturing plants or waste sites.
AN is rapidly and nearly completely absorbed, widely distributed to tissues, and
biochemically transformed into metabolites that are excreted in the urine and, to a much lesser
extent, in feces and expired air. Two major metabolic pathways for AN have been identified:
detoxification of AN by conjugation with GSH, forming the urinary metabolite N-acetyl-S-
(2-cyanoethyl)cysteine, and oxidation of AN to its epoxide metabolite, CEO, via CYP2E1. CEO
can bind to tissue macromolecules, such as proteins and DNA. CEO can be hydrolyzed to
cyanide with further transformation to thiocyanate, which is excreted in urine. Alternatively,
CEO can interact with GSH, resulting in the formation of a number of other urinary metabolites.
There are no studies directly identifying health hazards in humans following oral
exposures of any duration, but results from a robust array of studies in rats and mice identify
noncancer lesions in the gastric squamous epithelium and tumors in multiple tissues as potential
health hazards to humans exposed to AN by the oral route for chronic durations. Results from
cross-sectional epidemiologic studies of AN-exposed workers identify increased prevalences of
neurological symptoms and adverse reproductive outcomes in occupationally exposed workers as
potential health hazards from chronic inhalation exposure. Cancer is another potential human
health hazard from chronic inhalation exposure, as indicated by increased incidences of tumors at
several tissue sites in rat chronic inhalation bioassays. Studies of AN-exposed workers have
provided no strong evidence that mortality from any type of cancer is casually related to
occupational exposure, but limited evidence from the best-designed epidemiologic study found a
small, but statistically significant, increased risk for dying from lung cancer in workers with the
longest durations and highest exposures to AN.
The animal toxicity database identifies hyperplasia and hyperkeratosis of the squamous
epithelium of the forestomach as the most sensitive noncancer effects associated with repeated
oral exposure to AN. Following chronic oral exposure, these lesions have been observed in rats
at drinking water concentrations as low as 1-3 ppm (0.09-0.3 mg/kg-day) (Johannsen and
Levinskas, 2002a, b; Biodynamics, 1980a, b). Other effects have been observed in repeatedly
exposed animals, generally at higher exposure levels. These include ovarian atrophy in female
369
DRAFT- DO NOT CITE OR QUOTE
-------
mice exposed to doses >2.5 mg/kg-day for 2 years (NTP, 2001); chronic nephropathy in male
and female rats exposed for 2 years to 3.4 and 10.8 mg/kg-day AN, respectively (Quast et al.,
1980a); gliosis in the brain of female Sprague-Dawley rats at 4.4 mg/kg-day (Quast et al.,
1980a); decreased sperm count in male mice exposed to 10 mg/kg-day for 60 days (Tandon et
al., 1988); hind-limb weakness and decreased sensory nerve conduction velocity in male rats
exposed to 50 mg/kg-day AN for 12 weeks (Gagnaire et al., 1998); and neurobehavioral effects
in male rats exposed to 4 mg/kg-day AN in drinking water for 8 or 12 weeks (Rongzhu et al.,
2007). No changes in fertility index or pregnancy success were found in a three-generation study
of rats exposed to drinking water doses as high as 39 mg/kg-day (Friedman and Beliles, 2002;
Litton Bionetics, 1992). Mild developmental effects were observed at 11 and 20 mg/kg-day
(small deficits in postnatal pup weight or survival) in this three-generation rat study and at 25
mg/kg-day (increased litters with pups with missing vertebrae) but not at 10 mg/kg-day in rat
fetuses exposed on GDs 6-15 (Murray et al., 1978).
Repeated inhalation exposure to AN in the workplace has been associated with increased
prevalence of subjective neurological symptoms, such as headache, poor memory, and irritability
(Chen et al., 2000; Kaneko and Omae, 1992; Muto et al., 1992; Sakurai et al., 1978) and small
performance deficits in neurobehavioral tests of mood, attention and speed, auditory memory,
visual perception and memory, and motor steadiness (Lu et al., 2005a). Such effects have been
associated with average workplace air concentrations of 0.1 or 0.9 ppm and appear to be the most
sensitive noncancer effects from repeated inhalation exposure to AN. Adverse reproductive
outcomes, such as increased prevalences of premature deliveries, stillbirths, sterility, birth
defects, and pregnancy complications, have been associated with occupational exposure to
average workplace concentrations ranging from 3.6 to 7.5 ppm (Dong et al., 2000a; Li, 2000;
Dong and Pan, 1995). Sub chronic Sub chronic and chronic inhalation toxicity studies in rats
identified other noncancer effects at higher exposure levels, including nasal epithelial lesions in
rats exposed to concentrations of 20 or 80 ppm for 2 years (Quast et al., 1980b), decreased nerve
conduction velocity and hind-limb weakness in rats exposed to >25 ppm AN (Gagnaire et al.,
1998), increased incidence of rat fetuses with missing vertebrae, missing ribs, or anteriorly
displaced ovaries following exposure of pregnant Sprague-Dawley rats to 80 ppm on GDs 6-15
(Murray et al., 1978), and decreased rat fetal weight gain following exposure of pregnant
Sprague-Dawley rats to concentrations >25 ppm on GDs 6-20 (Saillenfait et al., 1993).
The genotoxicity of AN has been evaluated in multiple systems in vitro and in vivo. In
addition to positive findings in blood lymphocytes, buccal mucosal cells, and sperm in five
epidemiologic studies, DNA alkylation by AN was found in numerous tissues in rats or mice
(brain, liver, testes, forestomach, colon, kidney, bladder, and lung) treated with a single dose of
AN. AN or its reactive metabolite, CEO, yielded positive results in in vitro mutation assays
using bacteria, fungi, and insects, as well as animal and human cell cultures. The weight of
evidence from these studies suggests that AN is mutagenic after metabolic activation to CEO.
370
DRAFT- DO NOT CITE OR QUOTE
-------
Following EPA Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), AN is
"likely to be carcinogenic to humans," based predominantly on consistent results showing that
lifetime inhalation or oral exposure caused statistically significantly increased incidence of
tumors at multiple tissue sites in rats and mice. Lifetime oral exposure to AN caused increased
incidences of tumors at multiple tissue sites, including the brain, forestomach, and Zymbal gland,
in several rat studies and the forestomach and Harderian gland in a gavage study in mice (NTP,
2001). Lifetime inhalation bioassays with Sprague-Dawley rats found exposure-related increases
in the incidences of brain tumors, Zymbal gland tumors, intestinal tumors, tongue tumors, and
malignant mammary gland tumors (Dow Chemical Co., 1992a; Maltoni et al., 1988, 1977; Quast
et al., 1980b). Also, there is some association between AN exposure and lung cancer deaths in
occupationally exposed workers. Rats exposed during gestation and through adulthood
displayed higher incidences of tumors at several tissue sites than did rats exposed throughout
adulthood only (Friedman and Beliles, 2002; Maltoni et al., 1988). The latter results suggest that
exposure to AN during gestation and childhood may present increased risk of developing cancer
compared with exposure during adulthood alone.
Although data gaps still exist in the current understanding of the mode of action for
carcinogenicity of AN, there is adequate experimental evidence to support a direct mutagenic
mode of action as the key mode of action for AN-induced tumors. Other modes of action may
contribute, but limited data do not appear supportive at this time. Indirect mutagenicity via
oxidative DNA damage is plausible, but oxidative stress was not supported by experimental
evidence as a key mode of action.
6.2. DOSE-RESPONSE
6.2.1. Oral RfD
The available oral toxicity studies in animals identify nonneoplastic forestomach lesions
(i.e., squamous cell epithelial hyperplasia and hyperkeratosis) as the most sensitive noncancer
effect associated with chronic oral exposure to AN. These lesions are expected to be relevant to
humans because, although humans do not possess a forestomach, they do have comparable
squamous cell epithelial tissue in their oral cavity and in the upper two-thirds of their esophagus.
A 2-year drinking water study with F344 rats (Johanssen and Levinskas, 2002b) was
selected as the principal study on which to base the RfD because it provided the best available
dose-response data. This study included five drinking water exposure levels ranging from 1 to
100 ppm, and it also identified the lowest administered-dose LOAEL of the available chronic
animal toxicity studies based on an increased incidence of forestomach lesions (i.e., LOAELs of
0.3 and 0.4 mg/kg-day for males and females, respectively, with associated NOAELs of 0.1
mg/kg-day for both sexes).
An RfD of 2 x 10"4 mg/kg-day (or 0.2 |ig/kg-day) is derived for use in humans because it
is the lowest value derived based on the observed critical effect in the selected principal study
371
DRAFT- DO NOT CITE OR QUOTE
-------
(i.e., incidence of nonneoplastic forestomach lesions in F344 rats in a 2-year drinking water
study). This RfD was derived using a BMD approach based on incidence data for forestomach
lesions in male and female F344 rats exposed chronically to AN via drinking water and
pharmacokinetic data from EPA-modified rat and human PBTK models (Section 3.5; Appendix
C; Sweeney et al., 2003; Kedderis et al., 1996).
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, separately, employing two
different internal dose metrics (i.e., AN in blood and CEO in blood). These two internal dose
metrics were derived by converting administered rat doses of AN to internal rat doses (either AN
or CEO in blood) using the rat PBTK model of Kedderis et al. (1996), as modified by EPA
(Section 3.5; Appendix C). Then, for each sex, the BMDL associated with both a 5 and 10%
extra risk for gastric epithelial lesions was derived based on the best-fitting model. This BMDL
based on rat internal dose was then converted to the human equivalent administered dose of AN
by using the human PBTK model of Sweeney et al. (2003) (with EPA-modified parameters;
Section 3.5; Appendix C). As discussed in more detail in Section 5.1, CEO in blood was
selected as the best available internal dose metric for cross-species extrapolation with oral
exposure, while 5% extra risk was chosen as the most appropriate BMR level. The human
equivalent administered dose of AN represents a POD for noncancer effects and was then
divided by a UF of 30 (3 to account for uncertainty in extrapolating from rats to humans with
dosimetric adjustment and 10 to account for variation in response from average humans to
sensitive humans) to arrive at an RfD.
Confidence in the principal study selected for the RfD is high. The principal study,
Johannsen and Levinskas (2002b), was selected from eight chronic rat studies and one gavage
study in exposed mice. The study employed five exposure levels of AN, ranging from 1 to
100 ppm, and thus provided a more complete description of the dose-response relationship in the
low-dose region than did either the Quast (2002) study in Sprague-Dawley rats or the NTP
(2001) study in B6C3Fi mice, which started at higher dose levels, and both employed only three
dose groups. Johannsen and Levinskas (2002b) identified the lowest LOAELs based on an
increased incidence of hyperplasia and hyperkeratosis in squamous epithelium of forestomach
(0.4 mg/kg-day for males and 0.4 mg/kg-day for females). No human exposure study is
available for derivation of the RfD. Overall confidence in the RfD is medium, reflecting these
considerations.
6.2.2. Inhalation RfC
Results from several cross-sectional epidemiologic studies of AN-exposed workers
identified increased prevalence of neurological symptoms and small performance deficits in
neurobehavioral tests as the critical effects resulting from chronic inhalation exposure to AN.
The cross-sectional study of neurobehavioral performance measures in acrylic fiber workers by
372
DRAFT- DO NOT CITE OR QUOTE
-------
Lu et al. (2005a) was selected as the principal study for RfC derivation because it identified the
lowest reliable exposure level in humans associated with adverse neurological effects.
A NOAEL/LOAEL approach was used to derive the RfC from the human data. The
average workplace AN air concentration of 0.11 ppm for workers in the monomer work areas of
the acrylic fiber plant was selected as the LOAEL or POD for RfC derivation. This LOAEL of
"3
0.11 ppm (0.24 mg/m ) for small, but statistically significant, performance deficits in
neurobehavioral tests of mood, attention and speed, auditory memory, visual perception and
memory, and motor steadiness was divided by a UF of 100 (10 for extrapolating from a LOAEL
to a NOAEL and 10 to account for extrapolating from healthy workers to sensitive humans),
"3
after conversion to an equivalent continuous exposure of 0.086 mg/m , to arrive at an RfC for
AN of 0.9 (J,g/m3.
As discussed in more detail in Section 5.2, comparative animal-based RfCs for AN of 3
x 10"3 mg/m3 (or 3 (J,g/m3) and 2 x 10"3 mg/m3 (or 2 (J,g/m3) were derived based on PODs from
BMD modeling of nasal lesions observed in male and female rats, respectively, exposed to AN
via inhalation for 2 years (Quast et al., 1980b). In deriving these RfCs, the PODs were divided
by a composite UF of 30 (3 for extrapolating from rats to humans using the default U.S. EPA
(1994) dosimetric adjustment and 10 to account for variation from average humans to sensitive
humans). These animal-based RfCs are quite consistent with the human-based value. However,
"3
the human-based value of 0.9 (j,g/m is selected as the RfC, because extrapolating from animals
has greater associated uncertainty than extrapolating from humans.
The principal study is given medium confidence because it is the best available study
that identified neurobehavioral effects of AN in occupationally exposed workers. Previous
occupational studies by Kaneko and Omae (1992) and Muto et al. (1992) reported subjective
neurological symptoms in exposed workers. Lu et al. (2005a) utilized the WHO-recommended
NCTB administered by trained physicians to evaluate these neurobehavioral effects systemically.
Hence, the results were more reliable when compared with those based on self reporting. The
confidence in the principal study is medium because there are several limitations in the study.
One was that the cited exposure data represented estimates of previous exposure levels and no
contemporaneous personal monitoring data were available. In addition, the study authors could
not rule out the possibility that examiner drift may have affected the results. Moreover, largest
measures of neurobehavioral effect occurred on the acrylic fiber workers, who had lower average
exposure level.
Confidence in the database is medium to high. An RfC based on the results from a
chronic inhalation study with Sprague-Dawley rats was also derived for comparison.
Statistically significant increased incidence of inflammatory and degenerative nasal lesions
occurred in rats exposed to the lowest level in this 2-year bioassay. The alternative RfC derived
from the rat study is only about threefold higher than the RfC derived from the occupational
exposure study. Overall confidence in the RfC is medium to high, reflecting these
373
DRAFT- DO NOT CITE OR QUOTE
-------
considerations.
6.2.3. Oral CSF
Incidence data for forestomach, CNS, Zymbal gland, tongue, and mammary gland tumors
in male and female Sprague-Dawley (Quast, 2002) and F344 (Johanssen and Levinskas, 2002a)
rats were used to develop site-specific oral CSFs for AN, employing EPA-modified rat and
human PBTK models for cross-species dosimetric extrapolation (Section 3.5; Appendix C;
Sweeney et al., 2003; Kedderis et al., 1996). These animal studies were selected for the
development of oral CSFs because human data are not available, and these studies are the best
available chronic bioassays for characterizing the dose-response relationships for these
AN-induced tumors. Weight-of-evidence evaluation following U.S. EPA (2005a) guidelines
determined that a direct mutagenic mode of action, most likely via the reactive AN metabolite
CEO, was the principal mode of action. Consequently, a linear low-dose extrapolation approach
was used in the development of the oral CSFs.
As discussed in more detail in Section 5.4, CEO in blood and AN in blood were both
evaluated as internal dose metrics for use in cross-species extrapolation with oral exposure. The
multistage dose-response model in BMDS (version 1.4.1) was fit to the male and female rat
tumor incidence data, using internal animal dose expressed as either CEO or AN in blood. Rat
administered doses were converted to rat internal doses (either CEO or AN in blood) by using
the rat PBTK model of Kedderis et al. (1996), as modified by EPA (Section 3.5; Appendix C).
For each of these two dose metrics, the resulting best-fit model for each endpoint was then used
to derive a 95% lower confidence limit on the dose associated with 10% extra risk (i.e., a
BMDLio). These BMDLs, based on internal rat doses, were then converted to human equivalent
administered doses of AN using the human PBTK model of Sweeney et al. (2003), as modified
by EPA (Section 3.5; Appendix C). The site-specific oral CSFs based on incidence data from
each tumor site were derived by linear extrapolation from the human equivalent administered
doses down to the origin (oral CSF = 0. 1/BMDLio/hed)- Within rat strain and sex, an oral slope
factor for the composite risk across these tumor sites, based on CEO in blood, was then estimated
by employing the procedure described in Section 5.4.4.1.
The highest composite slope factor of 5 per mg/kg-day is recommended for use in
humans because it has the highest potency based on the observed critical effect in the selected
principal study (i.e., incidence of tumors in male and female Sprague-Dawley rats). This slope
factor is recommended for estimating composite cancer risks in humans who experience chronic
oral exposures to AN during adulthood. This slope factor should not be used with exposures
greater than 0.04 mg/kg-day (the lowest POD supporting the composite risk) because above this
level, the slope factor cannot be expected to be an adequate approximation to the dose-response
relationship. The fitted dose-response relationship and pharmacokinetic models should be used
to estimate risk above this exposure level.
374
DRAFT- DO NOT CITE OR QUOTE
-------
Results from a chronic rat inhalation cancer bioassay (Maltoni et al., 1988) and a three-
generation rat drinking water reproductive toxicity study (Friedman and Beliles, 2002) indicate
that AN exposure during gestation and early periods of development may increase risks for
cancer by at least threefold compared with chronic AN exposures during adulthood only. A
chemical-specific early life susceptibility factor of 3.6 was developed for assessing cancer risks
associated with lifetime exposure beginning at birth, based on the incidence data in the Maltoni
et al. (1988) bioassay for mammary tumors, brain tumors, and extra-hepatic angiosarcomas in
female rats.
6.2.4. Cancer Inhalation Unit Risk
An analysis of the human lung cancer mortality data from the Blair et al. (1998) cohort
study of AN-exposed workers was conducted to derive an IUR estimate for AN based on human
data. This analysis employed the approach presented by Starr et al. (2004) and is further
described in Appendix B-7. In brief, the risk of death from lung cancer in AN-exposed workers
was characterized by using a semi-parametric Cox regression model with a cumulative exposure
metric (i.e., ppm-working years) as the only time-dependent covariate. In contrast to the analysis
by Starr et al. (2004), the entire cohort was included in the analysis conducted for this assessment
(not just white male workers), and the final model only included cumulative exposure as the
covariate.
The Cox regression model described above was used to estimate an AN exposure level
and its associated 95% lower confidence limit corresponding to a 1% risk of dying from lung
cancer by age 80 (i.e., ECoi and LECoi, respectively). Conversion of occupational exposures to
continuous environmental exposures was accomplished by adjusting for differences in the
"3
amount of air inhaled during an 8-hour workday versus a 24-hour day (10 and 20 m /day,
respectively). The IUR estimate was derived by linear extrapolation from the LECoi The
predicted ECoi and LECoi derived from the Cox regression model based on the Blair et al. (1998)
data were 0.992 and 0.238 ppm (2,168 and 524 (J,g/m3) AN, respectively. From the LECoi, an
5 3
IUR estimate of 0.042 per ppm (2 x 10" per (j,g/m ) was derived.
Some uncertainty is associated with this IUR estimate because of the study on which it is
based. More specifically, no adjustment for smoking was applied in the Blair et al. (1998) study.
The investigators collected smoking information on only about 10% of the exposed and
unexposed members of the cohort and found similar incidences of smoking in the two groups.
Because their statistical analysis compared exposed and unexposed groups of workers, Blair et
al. (1998) noted that the adjustment for smoking made only a slight difference in the results of
their analysis. Other uncertainties associated with the Blair et al. (1998) data set included
nondifferential exposure misclassification and a relatively short follow-up period, an average of
21 years. Additionally, the outcome of the study was lung cancer mortality, not lung cancer
375
DRAFT- DO NOT CITE OR QUOTE
-------
incidence. Additional uncertainties related to the statistical analysis of these data are discussed
further in Appendix B-7.
Tumor incidence data from the best available animal inhalation study (Dow Chemical
Co., 1992a) were selected for describing the dose-response relationship between intestinal, CNS,
Zymbal's gland, tongue, and mammary gland tumors and AN exposure. These dose-response
relationships were used to derive animal-based IURs for AN for comparative purposes. As with
the animal-based oral slope factor, a direct mutagenic mode of action, most likely via the
reactive AN metabolite CEO, was assumed, because available information was inadequate to
establish AN's mode of carcinogenic action in these tissues in animals. Consequently, a linear
low-dose extrapolation approach was used in the development of the animal-based IURs.
Based on the assumption that the epoxide metabolite, CEO, is critical to AN's
carcinogenic mode of action, the selected internal dose metric was CEO in blood. In contrast to
oral exposure, the EPA-modified PBTK model adequately predicted measured blood and brain
concentrations of CEO in rats exposed to AN by inhalation (Kedderis et al., 1996). The
multistage model was fit to the rat tumor incidence and CEO concentration in blood predicted by
the PBTK model of Kedderis et al. (1996), as modified by EPA (Section 3.5; Appendix C). The
best-fitting stage of the model was used to derive 95% lower bounds on rat internal blood
concentrations of CEO associated with 10% extra risk (BMCLios). The human PBTK model of
Sweeney et al. (2003) (parameters modified; Section 3.5; Appendix C), was then used to
calculate human equivalent administered concentrations of AN, corresponding to the rat
BMCLios. These human equivalent administered concentrations of AN were used as the POD
for the IUR estimates via linear extrapolation down to the origin (i.e., IUR = 0.1/BMCLio/hec).
The IUR estimates for multiple tumor sites were derived within each rat sex by
employing the procedure described in Section 5.4.4.1. The resulting composite IUR estimates
2 2 3 5 5 3
were 7x10" and 6x10" per mg/m (7x10" and 6x10" per (j,g/m ) and were derived based on
tumor incidence data from male and female Sprague-Dawley rats, respectively. These unit risks
"3
should not be used with exposures greater than 3 mg/m (the lowest POD supporting the
composite risk), because above this level, the slope factor cannot be expected to be an adequate
approximation of the dose-response relationship. The fitted dose-response relationship and
pharmacokinetic models should be used to estimate risk above this exposure level.
5 3
Given that the IUR of 2 x 10" per (j,g/m is based on human data and is consistent with
the IUR derived from the animal data, it is the IUR recommended for estimating cancer risks
associated with chronic inhalation exposures to AN during adult stages of development.
As described for the oral slope factor, a chemical-specific early-life adjustment factor of
3.6 was developed for assessing cancer risks associated with lifetime exposure beginning at
birth, based on the multisite tumor incidence data for female rats in the Maltoni et al. (1988)
bioassay.
7. REFERENCES
376
DRAFT- DO NOT CITE OR QUOTE
-------
Abdel-Aziz, AH; Abdel-Naim, AB; Hamada, FM; et al. (1997) In-vitro testicular bioactivation of acrylonitrile.
Pharmacol Res 35:129-134.
Abdel-Rahman, SZ; Nouraldeen, AM; Abo-Elwafa, AA; et al. (1994a) Acrylonitrile-induced reversible inhibition of
uridine uptake by isolated rat intestinal epithelial cells. Toxicol Vitro 8:139-143.
Abdel-Rahman, SZ; Nouraldeen, AM; Ahmed, AE. (1994b) Molecular interaction of [2,3-14C]acrylonitrile with
DNA in gastric tissue of rat. J Biochem Toxicol 9:191-198.
Abdel-Rahman, SZ; Ammenheuser, MM; Ward, JB. (2001) Human sensitivity to 1,3-butadiene: role of microsomal
epoxide hydrolase polymorphisms. Carcinogenesis 22:415^23.
Abdel-Rahman, SZ; El-Zein, RA; Ammenheuser, MM; et al. (2003) Variability in human sensitivity to
1,3 butadiene: influence of the allelic variants of the microsomal epoxide hydrolase gene. Environ Mole
Mutagen41:40-146.
Abreu, ME; Ahmed, AE. (1980) Metabolism of acrylonitrile to cyanide. In vitro studies. Drug Metab Dispos 8:376-
379.
Acrylonitrile Group. (2000) Initial submission: letter from Acrylonitrile Group to USEPA re 4 unpublished surveys
of Chinese chemical industry worker exposure to acrylonitrile, w/attachments and dated 4/26/00. Submitted under
TSCA Section 8E; EPA Document No. 88-000000150; NTIS No. OTS0559911.
ACS (American Chemical Society). (2003) Online information sheet. Available online at
http://pubs.acs.org/hotartcl.cenear. 960624/prod.html.
Ahmed, AE; Abreu, ME. (1981) Microsomal metabolism of acrylonitrile in liver and brain. Adv Exp Med Biol
136(Pt. B):1229-1238.
Ahmed, AE; Patel, K. (1981) Acrylonitrile: in vivo metabolism in rats and mice. Drug Metab Dispos 9:219-222.
Ahmed, AE; Farooqui, MYH; Upreti, RK; et al. (1982) Distribution and covalent interactions of [1 -14C]acrylonitrile
in the rat. Toxicology 23:159-175.
Ahmed, AE; Farooqui, MYH; Upreti, RK; et al. (1983) Comparative toxicokinetics of 2,3-14C-and
1 14C acrylonitrile in the rat. J Appl Toxicol 3:39-47.
Ahmed, AE; Abdel-Aziz, AH; Abdel-Rahman, SZ; et al. (1992a) Pulmonary toxicity of acrylonitrile: covalent
interaction and effect on replicative and unscheduled DNA synthesis in the lung. Toxicology 76:1-14.
Ahmed, AE; Abdel-Rahman, SZ; Nouraldeen, AM. (1992b) Acrylonitrile interaction with testicular DNA in rats. J
Biochem Toxicol 7:5-11.
Ahmed, AE; Hamada, FMA; Abdel-Aziz, AH; et al. (1993) Immunotoxicity of acrylonitrile: flow cytometric study
of spleen lymphocyte subsets and the plaque forming cells response in immunized mice. J Biomed Sci Ther 9:139—
161.
Ahmed, AE; Jacob, S; Ghanayem, BI. (1996a) Comparative disposition of acrylonitrile and methacrylonitrile:
quantitative whole-body autoradiographic studies in rats. Fundam Appl Toxicol 33:49-59.
Ahmed, AE; Nouraldeen, AM; Abdel-Rahman, SZ; et al. (1996b) Role of glutathione modulation in acrylonitrile-
induced gastric DNA damage in rats. Arch Toxicol 70:620-627.
Albert, DM; Frayer, WC; Black, HE; et al. (1986) The Harderian gland: its tumors and its relevance to humans.
Trans Am Ophthalmol Soc 84:321-341.
377
DRAFT- DO NOT CITE OR QUOTE
-------
Amacher, E; Turner, G. (1985) Tests for gene mutational activity in the L5158Y/TK assay system. Prog Mutat Res
5:487-196.
Andersen, ME. (1991) Physiological modeling of organic compounds. Ann Occup Hyg 35:309-321.
Anderson, D; Cross, MF. (1985) Suitability of the P388F mouse lymphoma system for detecting potential
carcinogens and mutagens. Food Chem Toxicol 23:115-118.
Appel, KE; Peter, H; Bolt, HM. (1981) Effect of potential antidotes on the acute toxicity of acrylonitrile. Int Arch
Occup Environ Health 49:157-163.
Aronstam, RS; Abood, LG; Hoss, W. (1978) Influence of sulfhydryl reagents and heavy metals on the functional
state of the muscarinic acetylcholine receptor in rat brain. Mol Pharmacol 14:575-586.
Ashby, J; de Serres, F; Draper, M; et al. (1985) Evaluation of short-term tests for carcinogens. Prog Mutat Res 5:1-
752
ATSDR (Agency for Toxic Substances and Disease Registry). (1990) Toxicological profile for acrylonitrile. Public
Health Service, U.S. Department of Health and Human Services, Atlanta, GA. Available online at
http://www.atsdr.cdc.gov/toxprofiles. (accessed April 21, 2009).
Babanov, GP; Kiiuchikov, VN; Kaaraieva, NI; et al. (1959) Clinical symptoms of chronic poisoning by acrylonitrile.
Vrach Delo 8:833-836.
Bachofen, M; Weibel, ER. (1977) Alterations of the gas exchange apparatus in adult respiratory insufficiency
associated with septicemia. Am Rev Respir Dis 116:589-615.
Bader, M; Wrbitzky, R. (2006) Follow-up biomonitoring after accidental exposure to acrylonitrile—implications for
protein adducts as a dose monitorfor short-term exposures. Toxicol Lett 162(2—3): 125—131.
Bakker, JG; Jongen, SM; Van Neer, FC; et al. (1991) Occupational contact dermatitis due to acrylonitrile. Contact
Dermatitis 24:50-53.
Balda, BR. (1975) Acrylonitrile as a contact allergen. Hautarzt 26:599-601 (German).
Banerjee, S; Segal, A. (1986) In vitro transformation of C3H/10T1/2 and NIH/3T3 cells by acrylonitrile and
acrylamide. Cancer Lett 32:293-304.
Barnes, JM. (1970) Observations on the effects on rats of compounds related to acrylamide. Brit J Ind Med 27:147-
149.
Barrett, JC; Lamb, PW. (1985) Tests with the Syrian hamster embryo cell transformation assay. Prog Mutat Res
5:623-628.
Bellward, GD; Chang, T; Rodrigues, B; et al. (1988) Hepatic cytochrome P-450j induction in the spontaneously
diabetic BB rat. Mol Pharmacol 33(2): 140-143.
Benesh, V; Cherna, V. (1959) Acrylonitrile: acute toxicity and mechanism of action. J Hyg Epidemiol Microbiol
Immunol 3:106-116.
Benhamou, S; Reinikainen, M; Bouchardy, C; et al. (1998) Association between lung cancer and microsomal
epoxide hydrolase genotypes. Cancer Res 58:5291-5293.
Benn, T; Osborne, K. (1998) Mortality of United Kingdom acrylonitrile workers—an extended and updated study.
Scand J Work Environ Health 24:17-24.
Benz, FW; Nerland, DE. (2005) Effect of cytochrome P450 inhibitors and anticonvulsants on the acute toxicity of
acrylonitrile. Arch Toxicol 79:610-614.
378
DRAFT- DO NOT CITE OR QUOTE
-------
Benz, FW; Nerland, DE; Pierce, WM; et al. (1990) Acute acrylonitrile toxicity: studies on the mechanism of the
antidotal effect of D- and L-cysteine and their N-acetyl derivatives in the rat. Toxicol Appl Pharmacol 102:142-
150.
Benz, FW; Nerland, DE; Li, J; et al. (1997a) Dose dependence of covalent binding of acrylonitrile to tissue protein
and globin in rats. Fundam Appl Toxicol 36:149-156.
Benz, FW; Nerland, DE; Corbett, D; et al. (1997b) Biological markers of acute acrylonitrile intoxication in rats as a
function of dose and time. Fundam Appl Toxicol 36:141-148.
Bergmark, E. (1997) Hemoglobin adducts of acrylamide and acrylonitrile in laboratory workers, smokers and
nonsmokers. Chem Res Toxicol 10:78-84.
Beskid, 0; Dusek, Z; Solansky, I; et al. (2006) The effects of exposure to different clastogens on the pattern of
chromosomal aberrations detected by FISH whole chromosome painting in occupationally exposed individuals.
Mutat Res 594(l-2):20-29.
Bhooma, T; Venkataprasad, N. (1997) Acrylonitrile potentiates oxidative stress in rat alveolar macrophages. Bull
Environ Contam Toxicol 58:71-78.
Bhooma, T; Padmavathi, B; Devaraj, SN. (1992) Effect of acrylonitrile on the procoagulant activity of rat lung. Bull
Environ Contam Toxicol 48:321-326.
Bigner, DD; Bigner, SH; Burger, PC; et al. (1986) Primary brain tumours in F344 rats chronically exposed to
acrylonitrile in their drinking water. Food Chem Toxicol 24:129-137.
Biodynamics. (1980a) Initial submission: 24-month oral toxicity/carcinogenicity study with acrylonitrile
administered to Spartan rats in drinking water (final report) with cover letter dated 072492. Submitted under TSCA
Section 8E; EPA Document No. 88-920005779; NTIS No. OTS0544562.
Biodynamics. (1980b) Initial submission: Twenty-four month oral toxicity/carcinogenicity study of acrylonitrile
administered by intubation to Spartan rats (final report) with cover letter dated 072492. Submitted under TSCA
Section 8E; EPA Document No. 88-920005775; NTIS No. OTS0544558.
Biodynamics. (1980c) Initial submission: 24-month oral toxicity/carcinogenicity study with acrylonitrile
administered in drinking water to F344 rats (final report) with cover letter dated 072492. Submitted under TSCA
Section 8E; EPA Document No. 88-920005777; NTIS No. OTS0544560 (summary only). Full report available from
National Institute for Occupational Safety and Health, Centers for Disease Control and Prevention, U.S. Department
of Health and Human Services, Cincinnati, OH, 1-800-356-4674.
Blair, A; Stewart, PA; Zaebst, DD; et al. (1998) Mortality of industrial workers exposed to acrylonitrile. Scand J
Work Environ Health 24(Suppl 2):25-41.
Borak, J. (1992) Acute acrylonitrile toxicity: reconsideration of mechanisms and antidotes. The OEM Report 6:19—
21.
Borba, H; Monteiro, M; Proenca, MJ; et al. (1996) Evaluation of some biomonitoring markers in occupationally
exposed populations to acrylonitrile. Teratog Carcinog Mutagen 16:205-218.
Borlakoglu, JT; Scott, A; Henderson, CJ et al. (1993) Expression of P450 isoenzymes during rat liver organogenesis.
Int J Biochem 25:1659-1668.
Bradley, MO. (1985) Measurement of DNA single-strand breaks by alkaline elution in rat hepatocytes. Prog Mutat
Res 5:353-357.
Brams, A; Buchet, JP; Crutzen-Fayt, MC; et al. (1987) A comparative study, with 40 chemicals, of the efficiency of
the Salmonella assay and the SOS chromotest (kit procedure). Toxicol Lett 38:123-133.
379
DRAFT- DO NOT CITE OR QUOTE
-------
Breslow, NE. (1974) Covariance analysis of censored survival data. Biometrics 30:89-99.
Brooks, TM; Gonzalez, LP; Calvert, R; et al. (1985) The induction of mitotic gene conversion in the yeast
Saccharomyces cerevisiae strain JD1 Prog Mutat Res 5:225-228.
Brzezinski, MR; Boutelet-Bochan, H; Person, RE; et al. (1999) Catalytic activity and quantitation of cytochrome
P450 2E1 in prenatal human brain. J Pharm Exp Therap 289(3): 1648-1653.
Buckpitt, AR; Rollins, DE; Mitchell, JR. (1979) Varying effects of sulfhydryl nucleophiles on acetaminophen
oxidation and sulfhydryl adduct formation. Biochem Pharmacol 28:2941-2946.
Burka, LT; Sanchez, IM; Ahmed, AE; et al. (1994) Comparative metabolism and disposition of acrylonitrile and
methacrylonitrile in rats. Arch Toxicol 68:611-618.
Butterworth, BE; Eldridge, SR; Sprankle, CS; et al. (1992) Tissue-specific genotoxic effects of acrylamide and
acrylonitrile. Environ Mol Mutagen 20:148-155.
Campian, EC; Cai, J; Benz, FW. (2002) Acrylonitrile irreversibly inactivates glyceraldehyde-3-phosphate
dehydrogenase by alkylating the catalytically active cysteine-149. Chem Biol Interact 140:279-291.
Campian, EC; Benz, FW. (2008) The acute lethality of acrylonitrile is not due to brain metabolic arrest. Toxicology
253: 104-109.
Carrera, MP; Antolin, I; Martin, V; et al. (2007) Antioxidants do not prevent acrylonitrile-induced toxicity. Toxicol
Lett 169:236-244.
Carriere, V; Berthou, F; Baird, S; et al. (1996) Human cytochrome P450 2E1 (CYP2E1) from genotype to
phenotype. Pharmacogenetics 6:203-211.
Chanas, B; Wang, H; Ghanayem, BI. (2003) Differential metabolism of acrylonitrile to cyanide is responsible for
the greater sensitivity of male vs female mice: role of CYP2E1 and epoxide hydrolases. Toxicol Appl Pharmacol
193:293-302.
Chang, CM; Hsia, MT; Stoner, GD; et al. (1990) Acrylonitrile-induced sister-chromatid exchanges and DNA single-
strand breaks in adult human bronchial epithelial cells. Mutat Res 241:355-360.
Chantara, W; Watcharasit, P; Thiantanawat, A; et al. (2006) Acrylonitrile-induced extracellular signal-regulated
kinase (ERK) activation via protein kinase C (PKC) in SK-N-SH neuroblastoma cells. J Appl Toxicol 26:517-523.
Checkoway, H; Pearce, NE; Crawford-Brown, D. (1989) Research methods in occupational epidemiology. New
York, NY: Oxford University Press.
Chen, JL; Walrath, J; O'Berg, MT; et al. (1987) Cancer incidence and mortality among workers exposed to
acrylonitrile. Am J Ind Med 11:157-163.
Chen, Y; Chen, C; Jin, S; et al. (1999) The diagnosis and treatment of acute acrylonitrile poisoning: a clinical study
of 144 cases. J Occup Health 41:172-176.
Chen, Y; Chen, C; Zhu, P. (2000) Study on the effects of occupational exposure to acrylonitrile in workers. China
Occup Med J 18(3). Available online at http://www.iHb2.com/A-ISSN~1006-9070(2007)03-0001-04.htm (accessed
date April 22, 2009).
Cheng, KC; Cahill, DS; Kasai, H; et al. (1992) 8-Hydroxyguanine, an abundant form of oxidative DNA damage,
causes G—T and A—C substitutions. J Biol Chem 267(1): 166-172.
Chengelis, CP. (1988) Age- and sex-related changes in epoxide hydrolase, UDP-glucuronosyl transferase,
glutathione S-transferase, and PAPS sulphotransferase in Sprague-Dawley rats. Xenobiotica 18:1225-1237.
380
DRAFT- DO NOT CITE OR QUOTE
-------
Cohen, SM. (2004) Human carcinogenic risk evaluation: an alternative approach to the two-year rodent bioassay.
Toxicol Sci 80:225-229.
Cole, CE; Tran, HT; Schlosser, PM (2001) Physiologically based pharmacokinetic modeling of benzene metabolism
in mice through extrapolation from in vitro to in vivo. J Toxicol Environ Health 62:439-465.
Collins, AR. (2007) Investigating oxidative DNA damage and its repair using the comet assay. Mutation
Research/Reviews in Mutation Research 681: 24-32.
Collins, JJ; Page, LC; Caporossi, JC; et al. (1989) Mortality patterns among employees exposed to acrylonitrile. J
Occup Med 31:368-371.
Cornet, M; Mertens, K; Callaerts, A; et al. (1994) Age- and gender-related changes in the hepatic metabolism of
2-methylpropene and relationship to epoxide metabolizing enzymes. Mech Ageing Develop74:103-115.
Corsaro, CM; Migeon, BR. (1978) Gene expression in euploid human hybrid cells: ouabain resistance is
codominant. Somatic Cell Genet 4:531-540.
Cox, DR. (1972) Regression models and life-tables (with discussion). J R Stat Soc [Ser B] 34:187-220.
Crespi, CL; Ryan, CG; Seixas, GM; et al. (1985) Tests for mutagenic activity using mutation assays at two loci in
the human lymphoblast cell lines YK6 and AHH-1. Prog Mutat Res 5:497-516.
Cresteil, T; Beaune, P; Kremers, P; Celier, C; et al. (1985) Immunoquantification of epoxide hydrolase and
cytochrome P-450 isozymes in fetal and adult human liver microsomes. Eur J Biochem 151:343—350.
Cuterman, A; Manwaring, D; Curreri, PW. (1977) The role of fibrinogen degradation products in the pathogenesis
of respiratory distress syndrome. Surgery 82:703-709.
Czeizel, AE; Hegedus, S; Timar, L. (1999) Congenital abnormalities and indicators of germinal mutations in the
vicinity of an acrylonitrile producing factory. Mutat Res 427:105-123.
Czeizel, AE; Hegedus, S; Timar, L. (2000) Corrigendum to "Congenital abnormalities and indicators of germinal
mutations in the vicinity of an acrylonitrile producing factory." Mutat Res 453:105-106.
Czeizel, AE; Szilvasi, R; Timar, L; et al. (2004) Occupational epidemiological study of workers in an acrylonitrile
using factory with particular attention to cancers and birth defects. Mutat Res 547:79-89.
Danford, N. (1985) Tests for chromosomal aberrations and aneuploidy in the Chinese hamster fibroblast cell line
CH1 -L. Prog Mutat Res 5:3 97-411.
de Waziers, I; Cugnenc, PH; Yang, CS; et al. (1990) Cytochrome P 450 isoenzymes, epoxide hydrolase and
glutathione transferases in rat and human hepatic and extrahepatic tissues. J Pharmacol Exp Ther 253(l):387-94.
de Meester, C; Poncelet, F; Roberfroid, M; et al. (1978) Mutagenicity of acrylonitrile. Toxicology 11:19-27.
Delivanova, S; Popovski, P; Orusev, T. (1978) Blepaharoconjunctivitis in workers in the manufacture of synthetic
polyacrylonitrile fibers. God Zb Med Fak Skopje 24:279-282.
Delzell, E; Monson, RR. (1982) Mortality among rubber workers: VI. Men with potential exposure to acrylonitrile.
J Occup Med 24:767-769.
Denlinger, CL; Vesell, ES (1989) Hormonal regulation of the developmental pattern of epoxide hydrolases.
Biochem Pharmacol 38:603-610.
De Vries, N ; De Flora, S (1993) N-Acetyl-L-cysteine. J. Cell Biochem 17F : S270-S277.
381
DRAFT- DO NOT CITE OR QUOTE
-------
Ding, S; Lai-ji, MA; Fan, W; et al. (2003) Study on mitochondrial DNA damage in peripheral blood nucleate cells
of the workers exposed to acrylonitrile. Chinese J Indust Hyg Occup Disease 21:99—101. (Chinese).
Doll, R. (1971) The age distribution of cancer: implications for models of carcinogenesis. J R Stat Soc [Ser A]
134:133-166.
Dong, D; Pan, J. (1995) Acrylonitrile effect on worker's reproductive system. Petrochem Safe Technol Mag 5:30-
31
Dong, D; Tao, D; Yang, Y. (2000a) Study of occupational harmfulness to acrylonitrile workers. Industrial Health
Department of Safety and Technology, Daqing Petrochemical General Plant. Submitted under TSCA Section 8E;
EPA Document No. 89-000000313; NTIS No. OTS0559911.
Dong, D; Wang, D; Ai, X; et al. (2000b) Study of acrylonitrile hazardous effects on workers' reproductive system.
Submitted under TSCA Section 8E; EPA Document No. 89-000000313; NTIS No. OTS0559911.
Douglas, GR; Blakey, DH; Liu-Lee, VW; et al. (1985) Alkaline sucrose sedimentation, sister chromatid exchange
and micronucleus assays in CHO cells. Prog Mutat Res 5:359-366.
Dow Chemical Co. (1976) Interim Report. In vitro microbiological mutagenicity studies of Dow Chemical
Company Compounds. Submitted under TSCA Section FYI; EPA Document No. FYI-OTS-0677-0150; NTIS No.
OTS0000150-0.
Dow Chemical Co. (1977) Teratologic evaluation of acrylonitrile monomer given to rats by gavage. Submitted
under TSCA SectionFYI; EPA Document No. FYI-OTS-0677-0150; NTIS No. OTS0000150-0.
Dow Chemical Co. (1992a) Initial submission: 2-year toxicity and oncogenicity study with acrylonitrile following
inhalation exposure in rats (final report) with cover letter dated 080392. Submitted under TSCA Section 8E; EPA
Document No. 88-9200006574; NTIS No. OTS0545173.
Dow Chemical Co. (1992b) Initial submission: teratologic evaluation of gavaged acrylonitrile monomer in rats with
cover letter dated 08/10/92. Submitted under TSCA Section 8E; EPA Document No. 88-920010410; NTIS No.
OTS0555785.
Dudley, HC; Neal, PA. (1942) Toxicology of acrylonitrile (vinyl cyanide). I. A study of the acute toxicity. J Ind
Hyg Toxicol 24:27-36.
Bureau, Institute for Health and Protection, European Commission and Joint Research Centre, Dublin, Ireland;
EINECS (European Inventory of Existing Commerical Chemical Substances) No. 203-466-5. Available online at
http://ecb.jrc.ec.europa.eu/DOCUMENTS/Existing- EC (European Communities). (2004) European Union risk
assessment report: acrylonitrile. European Chemicals
Chemicals/RISK_ASSESSMENT/REPORT/acrylonitrilereport029.pdf (accessed April 20, 2009).
El-Sayed, EM; Abo-Salem, OM; Abd-Ellah, MF et al. (2008) Hesperidin, an antioxidant flavonoid, prevents
acrylonitrile-induced oxidative stress in rat brain. J Biochem Molecular Toxicol 22: 268-273.
El Hadri, L; Chanas, B.; Ghanayem, BI. (2005) Comparative metabolism of methacrylonitrile and acrylonitrile to
cyanide using cytochrome P4502E1 and microsomal epoxide hydrolase-null mice. Toxicol Appl Pharmacol
205:116-125.
Elmore, E; Korytynski, EA; Smith, MP. (1985) Tests with the Chinese hamster V79 inhibition of metabolic
cooperation assay. Prog Mutat Res 5:597-612.
Esmat, M; El-Demerdash, E; El-Mesallamy, H; et al. (2007) Toxicity and oxidative stress of acrylonitrile in rat
primary glial cells: preventive effects of N-acetylcysteine. Toxicol Lett 171:111-118.
Fahmy, MA. (1999) Evaluation of the genotoxicity of acrylonitrile in different tissues of male mice. Cytologia
64:1-9.
382
DRAFT- DO NOT CITE OR QUOTE
-------
Fan, W; Wang, WL; Ding, S; et al. (2006) [Application of micronucleus test of buccal mucosal cells in assessing the
genetic damage of workers exposed to acrylonitrile.] Zhonghua Lao Dong Wei Sheng Zhi Ye Bing Za Zhi
24(2): 106-108. (Chinese)
Fang, JY; Richardson, BC. (2005) The MAPK signaling pathways and colorectal cancer. Lancet Oncol 6:322-327.
Farooqui, MY; Ahmed, AE. (1982) Molecular interactions of acrylonitrile and potassium cyanide with rat blood.
ChemBiol Interact 38:145-159.
Farooqui, MY; Ahmed, AE. (1983a) In vivo interactions of acrylonitrile with macromolecules in rats. Chem Biol
Interact 47:363-371.
Farooqui, MY; Ahmed, AE. (1983b) The effects of acrylonitrile on hemoglobin and red cell metabolism. J Toxicol
Environ Health 12:695-707.
Farooqui, MY; Mumtaz, MM; Ghanayem, BI; et al. (1990) Hemoglobin degradation, lipid peroxidation, and
inhibition of Na+/K(+)-ATPase in rat erythrocytes exposed to acrylonitrile. J Biochem Toxicol 5:221-227.
Fechter, LD. (2004) Promotion of noise-induced hearing loss by chemical contaminants. J Toxicol Environ Health
A 67:727-740.
Fechter, LD; Sjaak, FL; Najeeb, K; et al. (2003) Acrylonitrile produces transient cochlear function loss and
potentiates permanent noise-induced hearing loss. Toxicol Sci 75:117-123.
Fechter, LD; Gearhart, C; Shirwany, NA. (2004) Acrylonitrile potentiates noise-induced hearing loss in rat. JARO
5:90-98.
Fei, N; Xu, LZ. (2006) [Parkinson's syndrome after acute severe acrylonitrile poisoning in one patient.] Zhonghua
Lao Dong Wei Sheng Zhi Ye Bing Za Zhi 24(5):309-310. (Chinese)
Fennell, TR; Sumner, SCJ. (1994) Identification of metabolites of carcinogens by 13C NMR spectroscopy. Drug
Metab Rev 26:469-481.
Fennell, TR; Kedderis, GL; Sumner, SCJ. (1991) Urinary metabolites of [l,2,3-13C]acrylonitrile in rats and mice
detected by carbon-13 nuclear magnetic resonance spectroscopy. Chem Res Toxicol 4:678-687.
Fennell, TR; MacNeela, JP; Morris, RW; et al. (2000) Hemoglobin adducts from acrylonitrile and ethylene oxide in
cigarette smokers: effects of glutathione S-transferase Tl-null and Ml-null genotypes. Cancer Epidemiol
Biomarkers Prev 9:705-712.
Filser, JG; Bolt, HM. (1984) Inhalation pharmacokinetics based on gas uptake studies. VI. Comparative evaluation
of ethylene oxide and butadiene monoxide as exhaled reactive metabolites of ethylene and 1,3-butadiene in rats.
Arch Toxicol 55:219-223.
Fischel-Ghodsian, N; Kopke, RD; Ge, X. (2004) Mitochondrial dysfunction in hearing loss. Mitochondrion 4:675-
694.
Foureman, P; Mason, JM; Valencia, R; et al. (1994) Chemical mutagenesis testing in drosophila. IX. Results of
50 coded compounds tested for the National Toxicology Program. Environ Mol Mutagen 23:51-63.
Friedman, MA; Beliles, RP. (2002) Three-generation reproduction study of rats receiving acrylonitrile in drinking
water. Toxicol Lett 132:249-261.
Gagnaire, F; Marignac, B; Bonnet, P. (1998) Relative neurotoxicological properties of five unsaturated aliphatic
nitriles in rats. J Appl Toxicol 18:25-31.
Gallagher, GT; Maull, EA; Kovacs, K; et al. (1988) Neoplasms in rats ingesting acrylonitrile for two years. J Am
Coll Toxicol 7:603-615.
383
DRAFT- DO NOT CITE OR QUOTE
-------
Gargas, ML; Andersen, ME; Teo, SK; et al. (1995) A physiologically based dosimetry description of acrylonitrile
and cyanoethylene oxide in the rat. Toxicol Appl Pharmacol 134:185-194.
Garner, RC; Campbell, J. (1985) Tests for the induction of mutations to ouabain or 6-thioguanine resistance in
mouse lymphoma L5178Y cells. Prog Mutat Res 5:525-529.
Garmin, RH; Snellings, WM; Maronpot, RR. (1985) Brain tumors in F344 rats associated with chronic inhalation
exposure to ethylene oxide. Neurotoxicol 6: 117-138.
Garmin, RH; Snellings, WM; Maronpot, RR (1986) Frequency, size and location of brain tumours in F344 rats
chronically exposed to ethylene oxide. Food Chem Toxicol 24: 145-153.
Geiger, LE; Hogy, LL; Guengerich, FP. (1983) Metabolism of acrylonitrile by isolated rat hepatocytes. Cancer Res
43:3080-3087.
Geng, J; Strobel, HW. (1993) Identification of cytochromes P.450 1A2, 2A1, 2C7, 2E1 in rat glioma C6 cell line by
RP PCR and specific restriction enzyme digestion. Biochem Biophys Res Commun 197:1179-1184.
George, J; Byth, K; Farrell, GC. (1995) Age but not agender selectively affects expression of individual cytochrome
P450 proteins in human liver. Biochemical Pharmacol 50:727-730.
Ghanayem, BI; Ahmed, AE. (1982) In vivo biotransformation and biliary excretion of l-14C-acrylonitrile in rats.
Arch Toxicol 50:175-185.
Ghanayem, BI; Ahmed, AE. (1983) Acrylonitrile-induced gastrointestinal hemorrhage and the effects of metabolism
modulation in rats. Toxicol Appl Pharmacol 68:290-296.
Ghanayem, BI; Ahmed, AE. (1986) Prevention of acrylonitrile-induced gastrointestinal bleeding by sulfhydryl
compounds, atropine and cimetidine. Res Commun Chem Pathol Pharmacol 53:141-144.
Ghanayem, BI; Boor, PJ; Ahmed, AE. (1985) Acrylonitrile-induced gastric mucosal necrosis: role of gastric
glutathione. J Pharmacol Exp Ther 232:570-577.
Ghanayem, BI; Farooqui, MY; Elshabrawy, O; et al. (1991) Assessment of the acute acrylonitrile-induced
neurotoxicity in rats. Neurotoxicol Teratol 13:499-502.
Ghanayem, BI; Elwell, MR; Eldridge, SR. (1997) Effects of the carcinogen, acrylonitrile, on forestomach cell
proliferation and apoptosis in the rat: comparison with methacrylonitrile. Carcinogenesis 18:675-680.
Grunske, F. (1949) Health care and occupational medicine: ventox and ventox intoxication. Dtsch Med
Wochenscher 74:1081-1083. (as cited in ATSDR, 1990).
Guengerich, FP; Wang, P; Mason, PS; et al. (1979) Rat and human microsomal epoxide hydratase. Immunological
characterization of various forms of the enzyme. J Biol Chem 254(23): 12255-12259.
Guengerich, FP; Geiger, LE; Hogy, LL; et al. (1981) In vitro metabolism of acrylonitrile to 2-cyanoethylene oxide,
reaction with glutathione, and irreversible binding to proteins and nucleic acids. Cancer Res 41:4925-4933.
Guengerich, FP; Hogy, LL; Inskeep, PB; et al. (1986) Metabolism and covalent binding of vic-dihaloalkanes, vinyl
halides and acrylonitrile. IARC Sci Publ 70:255-260.
Gut, I: Kopecky, J; Filip, J. (1981) Acrylonitrile-14C metabolism in rats: effect of the route of administration on the
elimination of thiocyanate and other radioactive metabolites in urine and feces. J Hyg Epidemiol Microbiol
Immunol 25:12-16.
Gut, I; Nerudova, J; Frantik, E; et al. (1984) Acrylonitrile inhalation in rats: I. Effect on intermediary metabolism.
J Hyg Epidemiol Microbiol Immunol 28:369-376.
384
DRAFT- DO NOT CITE OR QUOTE
-------
Gut, I; Nerudova, J; Stiborova, A; et al. (1985) Acrylonitrile inhalation in rats: II. Excretion of thioethers and
thiocyanate in urine. J Hyg Epidemiol Microbiol Immunol 29:9-13.
Guyton, KZ; Kensler, TW. (1993) Oxidative mechanisms in carcinogenesis. Brit Med Bull 49:523-544.
Hakura, A; Shimada, H; Nakajima, M; et al. (2005) Salmonella/human S9 mutagenicity test: a collaborative study
with 58 compounds. Mutagenesis 20:217-228.
Hamada, FM; Abdel-Aziz, AH; Abd-Allah, AR; et al. (1998) Possible functional immunotoxicity of acrylonitrile
(VCN). Pharmacol Res 37:123-129.
Harrison, DJ; Hubbard, AL; MacMillan, J; et al. (1999) Microsomal epoxide hydrolase gene polymorphism and
susceptibility to colon cancer. Br J Cancer 79:168-171.
Hashimoto, K; Kobayasi, T. (1961) A case of acute dermatitis caused by contact with acrylonitrile. Q J Labor Res
9:21-24.
Haskell Laboratory. (1992a) Initial submission: teratological evaluation of inhaled acrylonitrile monomer in rats
with cover letter dated 090192. Submitted under TSCA Section 8E; EPA Document No. 88-920008569; NTIS No.
OTS0570857.
Haskell Laboratory. (1992b) Initial submission: acute inhalation toxicity in rats with acrylonitrile (inhibited),
methyacrylonitrile (inhibited), and acetonitrile with cover letter dated 101592. Submitted under TSCA Section 8E;
EPA Document No. 88-920009947; NTIS No. OTS0571605.
Hassett, C; Lin, J; Carty, CL; et al. (1997). Human hepatic microsomal epoxide hydrolase: comparative analysis of
polymorphic expression. Arch Biochem 337:275-283.
Hedlund, B; Bartfai, T. (1979) The importance of thiol- and disulfide groups in agonist and antagonist binding to the
muscarinic receptor. Mol Pharmacol 15:531-544.
Heinrichs, WL; Juchau, MR. (1980) Extrahepatic drug metabolism: the gonads. In: Gram, TE, ed. Extrahepatic
metabolism of drugs and other foreign compounds. New York, NY: SP Medical and Scientific Books, pp. 319-332.
Hogy, LL. (1986) Metabolism of acrylonitrile and interactions with DNA (adduct, carcinogen, epoxide). Ph.D.
thesis submitted to Vanderbilt University, Nashville, TN. Available from ProQuest, Ann Arbor, MI, Document No.
8616356.
Hogy, LL; Guengerich, FP. (1986) In vivo interaction of acrylonitrile and 2-cyanoethylene oxide with DNA in rats.
Cancer Res 46:3932-3938.
IARC (International Agency for Research on Cancer). (1999) Predictive value of rodent forestomach and gastric
neuroendocrine tumours in evaluating carcinogenic risks to humans, views and expert opinions of an IARC Working
Group. Technical Publication No. 39. Lyon, France: International Agency for Research on Cancer.
Ikeda, SR; Aronstam, RS; Eldefrawi, ME. (1980) Nature of regional and chemically induced differences in the
binding properties of muscarinic acetylcholine receptors from rat brain. Neuropharmacology 19:575-585.
IPCS (International Programme on Chemical Safety). (1983) Acrylonitrile. Environmental health criteria. Vol. 28.
Geneva, Switzerland: World Health Organization. Available online at
http://www.inchem.org/documents/ehc/ehc/ehc28.htm (accessed April 21, 2009)
IPCS (International Programme on Chemical Safety). (1985) Summary report of the evaluation of short-term tests
from carcinogens (collaborative study on in vitro tests). Environmental health criteria. Vol. 47. Available online at
http://www.inchem.org/documents/ehc/ehc/ehc47.htm (accessed April 21, 2009)
385
DRAFT- DO NOT CITE OR QUOTE
-------
IPCS (International Programme on Chemical Safety). (2002) Acrylonitrile. Concise international chemical
assessment document. Vol. 39. Geneva, Switzerland: World Health Organization. Available online at
http://www.inchem.org/documents/cicads/cicads/cicad39.htm (accessed April 21, 2009).
Irwin, RD; Eustis, SL; Stefanski, S et al. (1996). Carcinogenicity of glycidol in F344 rats and B6C3F1 mice. J. of
Applied Toxicology 16: 201-209.
Ishidate, M, Jr; Sofuni, T. (1985) The in vitro chromosomal aberration test using Chinese hamster lung (CHL)
fibroblast cells in culture. Prog Mutat Res 5:427-432.
Ishidate, M, Jr; Sofuni, T; Yoshikawa, K. (1981) Chromosomal aberration tests in vitro as a primary screening tool
for environmental mutagens and/or carcinogens. Gann 27:95-108.
Ivanescu, M; Berinde, M; Simionescu, L. (1990) Testosterone in sera of workers exposed to acrylonitrile. Revue
Roum Med - Ser Endocrinol 28:187-192.
Jackson, MA; Stack, FH; Rice, JM; et al. (2000) A review of the genetic and related effects of 1,3-butadiene in
rodents and humans. Mutat Res 463:181 -213.
Jacob, S; Ahmed, AE. (2003a) Effect of route of administration on the disposition of acrylonitrile: quantitative
whole-body autoradiographic study in rats. Pharmacol Res 48:479-488.
Jacob, S; Ahmed, AE. (2003b) Acrylonitrile-induced neurotoxicity in normal human astrocytes: oxidative stress and
8-hydroxy-2'-deoxyguanosine formation. Toxicol MechMethods 13:169-179.
Jakubowski, M; Linhart, I; Pielas, G; et al. (1987) 2-Cyanoethylmercapturic acid (CEMA) in the urine as a possible
indicator of exposure to acrylonitrile. Br J Ind Med 44:834-840.
Jedlicka, V; Pasek, A; Gola, J. (1958) Pesticides in foods. III. Acrylonitrile as a food insecticide. Hyg Epidemiol
Microbiol Immunol. 1:116-125.
Jiang, J; Xu, Y; Klaunig, J E. (1998) Induction of oxidative stress in rat brain by acrylonitrile (ACN). Toxicol Sci
46:333-341.
Jimenez, E; Montiel, M. (2005) Activation of MAP kinase by muscarinic cholinergic receptors induces cell
proliferation and protein synthesis in human breast cancer cells. J Cell Physiol 204:678-686.
Johannsen, FR; Levinskas, GJ. (2002a) Comparative chronic toxicity and carcinogenicity of acrylonitrile by
drinking water and oral intubation to Spartan Sprague-Dawley rats. Toxicol Lett 132:197-219.
Johannsen, FR; Levinskas, GJ. (2002b) Chronic toxicity and oncogenic dose-response effects of lifetime oral
acrylonitrile exposure to F344 rats. Toxicol Lett 132:221-247.
Johnson, TN. (2003) The development of drug metabolising enzymes and their influence on the susceptibility to
adverse drug reactions in children. Toxicology 192:37-48.
Johnsrud, EK; Koukouritaki, SB; Divrkaran, K; et al. (2003) Human hepatic CYP2E1 expression during
development. J Pharmacol Exp Ther 307:402-407.
Jung, R; Engelhart, G; Herbolt, B; et al. (1992) Collaborative study of mutagenicity with Salmonella typhimurium
TA102. Mutat Res 278:265-270.
Kamendulis, LM; Jiang, J; Xu, Y; et al. (1999a) Induction of oxidative stress and oxidative damage in rat glial cells
by acrylonitrile. Carcinogenesis 20:1555-1560.
Kamendulis, LM; Jiang, J; Zhang, H; et al. (1999b) The effect of acrylonitrile on gap junctional intercellular
communication in rat astrocytes. Cell Biol Toxicol 15:173-183.
386
DRAFT- DO NOT CITE OR QUOTE
-------
Kamiat, H; Miura, K; Ishikawa, H; et al. (1992) C-Ha-ras containing 8-hydroxyguanine at codon 12 induces point
mutations at the modified and adjacent positions. Cancer Res 52:3483-3485.
Kaneko, Y; Omae, K. (1992) Effect of chronic exposure to acrylonitrile on subjective symptoms. Keio J Med
41:25-32.
Kedderis, GL. (1997) Development of a physiologically based dosimetry description for acrylonitrile (ACN) in
humans. Toxicologist 36(1, Pt. 2):31.
Kedderis, GL; Batra, R. (1993) Species differences in the hydrolysis of 2-cyanoethylene oxide, the epoxide
metabolite of acrylonitrile. Carcinogenesis 14:685-689.
Kedderis, GL; Held, SD. (1998) Refinement of the human dosimetry description for acrylonitrile (ACN).
Toxicologist 42(1-S):142.
Kedderis, GL; Sumner, SC; Held, SD; et al. (1993a) Dose-dependent urinary excretion of acrylonitrile metabolites
by rats and mice. Toxicol Appl Pharmacol 120:288-297.
Kedderis, GL; Batra, R; Held, SD; et al. (1993b) Rodent tissue distribution of 2-cyanoethylene oxide, the epoxide
metabolite of acrylonitrile. Toxicol Lett 69:25-30.
Kedderis, GL; Batra, R; Koop, DR. (1993c) Epoxidation of acrylonitrile by rat and human cytochromes P450.
Chem Res Toxicol 6:866-871.
Kedderis, GL; Batra, R; Turner, MJ, Jr. (1995) Conjugation of acrylonitrile and 2-cyanoethylene oxide with hepatic
glutathione. Toxicol Appl Pharmacol 135:9-17.
Kedderis, GL; Teo, SK; Batra, R; et al. (1996) Refinement and verification of the physiologically based dosimetry
description for acrylonitrile in rats. Toxicol Appl Pharmacol 140:422-435.
Kemper, RA; Krause, RJ; Elfarra, AA. (2001) Metabolism of butadiene monoxide by freshly isolated hepatocytes
from mice and rats: different partitioning between oxidative, hydrolytic, and conjugation pathways. Drug Metab
Dispos 29:830-836.
Khudoley, VV; Mizgireuv, I; Pliss, GB. (1987) The study of mutagenic activity of carcinogens and other chemical
agents with Salmonella typhimurium assays: testing of 126 compounds. Arch Geschwulstforsch 57:453-462.
Kim, RB; Yamazaki, H; Chiba, K; et al. (1996) In vivo and in vitro characterization of CYP2E1 activity in Japanese
and Caucasians. J Pharmacol Exp Ther 279:4-11.
Klaasen, CD, ed. (2001) Casarett and Doull's toxicology: the basic science of poisons. 6th edition. New York, NY:
McGraw-Hill Companies, Inc.
Kleihues, P; Bucheler, J. (1977) Long-term persistence of 06-methylguanine in rat brain DNA. Nature
269(5629):625-626.
Knobloch, K; Szendzikowski, S; Czajkowska, T; et al. (1971) Experimental studies on acute and subacute toxicity
of acrylonitrile. Med Pr 22:257-269.
Kodama, Y; Boreiko, CJ; Skopek, TR; et al. (1989) Cytogenetic analysis of spontaneous and 2-cyanoethylene oxide-
induced tk-/- mutants in TK6 human lymphoblastoid cultures. Environ Mol Mutagen 14:149-154.
Kohn, MC; Melnick, RL. (2000) The privileged access model of 1,3-butadiene disposition. Environ Health Perspect
108(Suppl 5):911-917.
Kopecky, J; Gut, I; Nerudova, J; et al. (1980) Two routes of acrylonitrile metabolism. J Hyg Epidemiol Microbiol
Immunol 24:356-362.
387
DRAFT- DO NOT CITE OR QUOTE
-------
Kopylev, L; Chen, C; White, P. (2007) Towards quantitative uncertainty assessment for cancer risks: central
estimates and probability distributions of risk in dose-response modeling. Regul Toxicol Pharmacol 49(3):203-207.
Krause, RJ; Sharer, JE; Elfarra, AA. (1997) Epoxide hydrolase-dependent metabolism of butadiene monoxide to
3-butene-l,2-diol in mouse, rat and human liver. Drug Metab Dispos 25:1013-1015.
Lakhanisky, Th; Hendrickx, B. (1985) Induction of DNA single-strand breaks in CHO cells in culture. Prog Mutat
Res 5:367-370.
Lambotte-Vandepaer, M; Duverger-van Bogaert, M; de Meester, C; et al. (1980) Mutagenicity of urine from rats
and mice treated with acrylonitrile. Toxicology 16:67-71.
Lambotte-Vandepaer, M; Duverger-van Bogaert, M; de Meester, C; et al. (1981) Identification of two urinary
metabolites of rats treated with acrylonitrile; influence of several inhibitors on the mutagenicity of those urines.
Toxicol Lett 7:321-327.
Lambotte-Vandepaer, M; Duverger-van Bogaert, M; Rollmann, B. (1985) Metabolism and mutagenicity of
acrylonitrile: an in vivo study. Environ Mutagen 7:655-662.
Langvardt, PW; Putzig, CL; Young, JD; et al. (1979) Isolation and identification of urinary metabolites of vinyl-type
compounds: application to metabolites of acrylonitrile and acrylamide. Toxicol Appl Pharmacol 48: A161.
Langvardt, PW; Putzig, CL; Braun, WH; et al. (1980) Identification of the major urinary metabolites of acrylonitrile
in the rat. J Toxicol Environ Health 6:273-282.
Larsson, R; Cerutti, P. (1989) Translocation and enhancement of phosphotransferase activity of protein kinase C
following exposure in mouse epidermal cells to oxidants. Cancer Res 49:5627-5632.
Lawrence, N; McGregor, DB. (1985) Assays for the induction of morphological transformation in C3H/10T1/2 cells
in culture with and without S9-mediated metabolic activation. Prog Mutat Res 5:651-658.
Lee, CG; Webber, TD. (1985) The induction of gene mutations in the mouse lymphoma L5178Y/TK+/" assay and the
Chinese hamster V79/HGPRT assay. Prog Mutat Res 5:547-554.
Leonard, A; Garny, V; Poncelet, F; et al. (1981) Mutagenicity of acrylonitrile in mouse. Toxicol Lett 7:329-334.
Li, Z. (2000) Study on reproductive organs in female workers exposed to acrylonitrile. Lanzhou Medical College.
Submitted under TSCA Section 8E; EPA Document No. 88-000000150; NTIS No. OTS0559911.
Lijinsky, W; Andrews, AW. (1980) Mutagenicity of vinyl compounds in Salmonella typhimurium. Teratog
Carcinog Mutagen 1:259-267.
Lipscomb, JC; Fisher, JW; Confer, PD; et al. (1998) In vitro extrapolation for trichloroethylene metabolism in
humans. Toxicol Appl Pharmacol 152(2):376-387.
Lipscomb, JC; Teuschler, LK; Swartout, J; et al. (2003) The impact of cytochrome P450 2E1-dependent metabolic
variance on a risk-relevant pharmacokinetic outcome in humans. Risk Anal 23(6): 1221-1238.
Litton Bionetics. (1992) Initial submission: three-generation reproduction study of rats receiving acrylonitrile in
drinking water (final report) with attachments and cover letter dated 043092. Submitted under TSCA Section 8E;
EPA Document No. 88-920002178; NTIS No. OTS0536313.
Liu, X; Xiao, W; Wang, Z; et al. (2004) Effect of acrylonitrile on spermatogenesis in mice. J Hyg Res 33:345-346.
Lorz, H. (1950) On the percutaneous poisoning by acrylonitrile (Ventox). Dtsch Med Wochenschr 75:1087-1088.
Lu, R; Ziqiang C; Fusheng J; et al. (2005a) Neurobehavioral effects of occupational exposure to acrylonitrile in
Chinese workers. Environ Toxicol Pharmacol 19:695-700.
388
DRAFT- DO NOT CITE OR QUOTE
-------
Lu, R; Chen, Z; Fusheng, J. (2005b) [Effect of acrylonitrile on monoamine neurotransmitters and their metabolites
in the brains of rats.] Chinese J Prev Med 39:122-126. (Chinese)
Lucas, D; Berthou, F; Dreano, Y; et al. (1993) Comparison of levels of cytochromes P-450, CYP1A2, CYP2E1, and
their related monooxygenase activities in human surgical liver samples. Alcohol Clin Exp Res 17(4): 900-905.
MacNeela, JP; Osterman-Golkar, SM; Turner, MJ; et al. (1992) 2-Cyanoethylvaline in hemoglobin as a dosimeter
for exposure to acrylonitrile. Proc Am Assoc Cancer Res 33:147.
Mahalakshmi, K; Pushpakiran, G; Anuradha, CV. (2003) Taurine prevents acrylonitrile-induced oxidative stress in
rat brain. Pol J Pharmacol 55:1037-1043.
Malik, AB; Lee, BC; van der Zee, H; et al. (1979) The role of fibrin in the genesis of pulmonary edema after
embolization in dogs. Circ Res 45(1): 120-125.
Maltoni, C; Ciliberti, A; Di, MV. (1977) Carcinogenicity bioassays on rats of acrylonitrile administered by
inhalation and by ingestion. Med Lav 68:401-411.
Maltoni, C; Ciliberti, A; Cotti, G; et al. (1988) Long-term carcinogenicity bioassays on acrylonitrile administered by
inhalation and by ingestion to Sprague-Dawley rats. Ann NY Acad Sci 534:179-202.
Mangir, M; Al, BI; Lochmann, E-R; et al. (1991) A test system for the rapid detection of nuclear and cytoplasmic
damage in Chinese hamster ovary cells. Chemosphere 23:777-784.
Marsh, GM; Gula, MJ; Youk, AO; et al. (1999) Mortality among chemical plant workers exposed to acrylonitrile
and other substances. Am J Ind Med 36:423-436.
Marsh, GM; Youk, AO; Collins, JJ. (2001) Reevaluation of lung cancer risk in the acrylonitrile cohort study of the
National Cancer Institute and the National Institute for Occupational Safety and Health. Scand J Work Environ
Health 27:5-13.
Martin, CN; Campbell, J. (1985) Tests for the induction of unscheduled DNA repair synthesis in HeLa cells. Prog
Mutat Res 5:375-379.
Martin, C; Dutertre-Catella, H; Radionoff, M; et al. (2003) Effect of age and photoperiodic conditions on
metabolism and oxidative stress related markers at different circadian stages in rat liver and kidney. Life Sci
73(3):327-335.
Mastrangelo, G; Serena, R; Marzia, V. (1993) Mortality from tumours in workers in an acrylic fibre factory. Occup
Med 43:155-158.
Matthews, EJ; DelBalzo, T; Rundell, JO. (1985) Assays for morphological transformation and mutation to ouabain
resistance of Balb/c-3T3 cells in culture. Prog Mutat Res 5:639-650.
McCarver, DG; Byun, R; Hines, RN; et al. (1998) A genetic polymorphism in the regulatory sequences of human
CYP2E1: association with increased chlorzoxazone hydroxylation in the presence of obesity and ethanol intake.
Toxicol Appl Pharmacol 152:276-281.
McGlynn, KA; Rosvold, EA; Lustbader, ED; et al. (1995) Susceptibility to hepatocellular carcinoma is associated
with genetic variation in the enzymatic detoxification of aflatoxin Bl. Proc Natl Acad Sci USA 92:2384-2387.
Meek, ME; Bucher, JR; Cohen, SM; et al. (2003) A framework for human relevance analysis of information on
carcinogenic modes of action. Crit Rev Toxicol 33:591-653.
Mehrotra, J; Khanna, VK; Husain, R; et al. (1988) Biochemical and developmental effects in rats following in utero
exposure to acrylonitrile: a preliminary report. Ind Health 26:251-255.
389
DRAFT- DO NOT CITE OR QUOTE
-------
Mikhutkina, SV; Salmina, AB; Sychev, AV; et al. (2004) Blebbing of thymocyte plasma membrane and apoptosis
are related to impairment of capacitance Ca2+ entry into cells. Bull Exp Biol Med 6:551-555.
Milvy, P; Wolff, M. (1977) Mutagenic studies with acrylonitrile. Mutat Res 48:271-278.
Mohamadin, AH; El-Demerdash, E; El-Beshbishy, HA; et al. (2005) Acrylonitrile-induced toxicity and oxidative
stress in isolated rat colonocytes. Environ Toxicol Pharmacol 19:371-377.
Morita, T; Asano, N; Awogi, T; et al. (1997) Evaluation of the rodent micronucleus assay in the screening of IARC
carcinogens (groups 1,2A and 2B): The summary report of the 6th collaborative study by CSGMT/JEMS MMS.
Collaborative study of the micronucleus group test. Mammalian Mutagenicity Study Group. Mutat Res 389:3-122.
Moriya, M; Du, C; Bodepudi, V; et al. (1991) Site-specific mutagenesis using a gapped duplex vector: a study of
translesion synthesis past 8-oxodeoxyguanine in E. coli. Mutat Res 254:281-288.
Mostafa, AM; Abdel-Naim, AB; Abo-Salem, 0; et al. (1999) Renal metabolism of acrylonitrile to cyanide: in vitro
studies. Pharmacol Res 40:195-200.
Muller, G; Verkoyen, C; Soton, N; et al. (1987) Urinary excretion of acrylonitrile and its metabolites in rats. Arch
Toxicol 60:464-466.
Mulvihill, J; Cazenave, JP; Mazzucotelli, JP; et al. (1992) Minimodule dialyser for quantitative ex vivo evaluation
of membrane haemocompatibility in humans: comparison of acrylonitrile copolymer cuprophan and polysulphone
hollow fibres. Biomaterials 13:527-536.
Murata, M; Ohnishi, S; Kawanishi, S. (2001) Acrylonitrile enhances H202-mediated DNA damage via nitrogen-
centered radical formation. Chem Res Toxicol 14:1421-1427.
Murray, FJ; Schwetz, BA; Nitschke, KD; et al. (1978) Teratogenicity of acrylonitrile given to rats by gavage or by
inhalation. Food Cosmet Toxicol 16:547-551.
Muto, T; Sakurai, H; Omae, K; et al. (1992) Health profiles of workers exposed to acrylonitrile. Keio J Med
41:154-160.
Myhr, B; Bowers, L; Kaspary, WJ. (1985) Assays for the induction of gene mutations at the thymidine kinase locus
in L5178Y mouse lymphoma cells in culture. Prog Mutat Res 5:555-568.
Nakamura, S; Oda, Y; Shimada, T; et al. (1987) SOS-inducing activity of chemical carcinogens and mutagens in
Salmonella typhimurium TA1535-psk-1002: examination with 151 chemicals. Mutat Res 192:239-246.
NAS (2008) Science and decisions: advancing risk assessment. National Academy of Sciences.
Natarajan, AT; Bussmann, CJM; van Kesteren-van Leeuwen, AC; et al. (1985) Tests for chromosomal aberrations
and sister chromatid exchanges in cultured Chinese hamster ovary (CHO) cells. Prog Mutat Res 5:433-437.
Nerland, DE; Cai, J; Pierce, WM, Jr; et al. (2001) Covalent binding of acrylonitrile to specific rat liver glutathione
S-transferases in vivo. Chem Res Toxicol 14:799-806.
Nerland, DE; Cai, J; Benz, FW. (2003) Selective covalent binding of acrylonitrile to Cys 186 in rat liver carbonic
anhydrase III in vivo. Chem Res Toxicol 16:583-589.
Nilsen, OG; Toftgard, R; Eneroth, P. (1980) Effects of arylonitrile on rat liver cytochrome P-450,
benzo(A)pyrenemetabolism and serum hormone levels. Toxicol Lett 6:399-404.
NLM (National Library of Medicine ). (2003) Acrylonitrile. HSDB (Hazardous Substances Data Bank). National
Institutes of Health, Bethesda, MD. Available on-line at http://toxnet.nlm.nih.gov (accessed April 21, 2009).
390
DRAFT- DO NOT CITE OR QUOTE
-------
NRC (National Research Council). (1983) Risk assessment in the federal government: managing the process.
Washington, DC: National Academy Press.
NRC (National Research Council). (1994) Science and judgment in risk assessment. Washington, DC: National
Academy Press.
NTP (National Toxicology Program). (1987) Toxicology and carcinogenesis studies of ethylene oxide (CAS No.
75-21-8) in mice (inhalation studies). Tech. report Series No. 326. Available from the National Institute of
Environmental Health Sciences, Research Triangle Park, NC
NTP (National Toxicology Program). (2001) Toxicology and carcinogenesis studies of acrylonitrile (CAS No. 107-
13-1) in B6C3F1 mice (gavage studies). Public Health Service, U.S. Department of Health and Human Services;
NTP TR 506. Available from the National Institute of Environmental Health Sciences, Research Triangle Park, NC
Available online at http://ntp.niehs.nih.gov/ntpweb/index. cfm?objectid=D16D6C59-FlF6-975E-
7D23D1519B8CD7A5#tr500 (accessed April 21, 2009).
O'Berg, MT. (1980) Epidemiologic study of workers exposed to acrylonitrile. J Occup Med 22:245-252.
O'Berg, MT; Chen, JL; Burke, CA; et al. (1985) Epidemiologic study of workers exposed to acrylonitrile: an
update. J Occup Med 27:835-840.
Oberly, TJ; Hoffman, WP; Garriott, ML. (1996) An evaluation of the twofold rule for assessing a positive response
in the L5178Y TK(+/-) mouse lymphoma assay. Mutat Res 369:221-232.
O'Connell, IF; Klein-Szanto, AJP; DiGiovanni, DM; et al. (1986) Enhanced malignant progression of mouse skin
tumors by the free-radical generator benzoyl peroxide. Cancer Res 46:2863-2865.
Oesch, F; Zimmer, A; Glatt, HR. (1983) Microsomal epoxide hydrolase in different rat strains. Biochem Pharmacol
32:1783-1788.
Ofengand, J. (1967) The function of pseudouridylic acid in transfer ribonucleic acid. I. The specific cyanoethylation
of pseudouridine, inosine, and 4-thiouridine by acrylonitrile. J Biol Chem 242(21):5034-5045.
Ohio Department of Health. (2006) Cancer incidence among residents of Addyston Village, Hamilton County, Ohio
1993-2003. Hamilton County General Health District, Cincinnati, OH and the Ohio Department of Health,
Columbus, OH. Final report. May 25, 2006.
Omiecinski, CJ; Alcher, L; Swenson, L. (1994) Developmental expression of human microsomal epoxide hydrolase.
J Pharmacol Exp Ther 269:417-23.
Osgood, C; Bloomfield, M; Zimmering, S. (1991) Aneuploidy in drosophila. IV. Inhalation studies on the induction
of aneuploidy by nitriles. Mutat Res 259:165-176.
Osterman-Golkar, SM; MacNeela, JP; Turner, MJ; et al. (1994) Monitoring exposure to acrylonitrile using adducts
with N-terminal valine in hemoglobin. Carcinogenesis 15:2701-2707.
Pacifici, GM; Peng, D; Rane, A (1983) Epoxide hydrolase and aryl hydrocarbon hydroxylase in human fetal tissues:
activities in nuclear and microsomal fractions and in isolated hepatocytes. Pediatr Pharmacol 3:189-197.
Pacifici, GM; and Rane, A (1983b) Epoxide hydrolase in human fetal liver. Pharmacology 26: 241-248.
Parent, RA; Casto, BC. (1979) Effect of acrylonitrile on primary Syrian golden hamster embryo cells in culture:
transformation and DNA fragmentation. J Natl Cancer Inst 62:1025-1029.
Paulet, G; Desnos, J. (1961) Acrylonitrile, toxicity, mechanism of action, therapeutic uses. Arch Int Pharmacodyn
131:54-83.
391
DRAFT- DO NOT CITE OR QUOTE
-------
Perez, HL; Segerback, D; Osterman-Golkar, S. (1999) Adducts of acrylonitrile with hemoglobin in nonsmokers and
in participants in a smoking cessation program. Chem Res Toxicol 12:869-873.
Perocco, P; Pane, G; Bolognesi, S; et al. (1982) Increase of sister chromatid exchange and unscheduled synthesis of
deoxyribonucleic acid by acrylonitrile in human lymphocytes in vitro. Scand J Work Environ Health 8:290-293.
Peter, H; Appel, KE; Berg, R; et al. (1983a) Irreversible binding of acrylonitrile to nucleic acids. Xenobiotica
13:19-25.
Peter, H; Schwarz, M; Mathiasch, B; et al. (1983b) A note on synthesis and reactivity towards DNA of
glycidonitrile, the epoxide of acrylonitrile. Carcinogenesis 4:235-237.
Pilon, D; Roberts, AE; Rickert, DE. (1988a) Effect of glutathione depletion on the irreversible association of
acrylonitrile with tissue macromolcules after oral administration to rats. Toxicol Appl Pharmacol 95:311-320.
Pilon, D; Roberts, AE; Rickert, DE. (1988b) Effect of glutathione depletion on the uptake of acrylonitrile vapors and
on its irreversible association with tissue macromolcules. Toxicol Appl Pharmacol 95:265-278.
Platanias, LC. (2003) Map kinase signaling pathways and hematologic malignancies. Blood 101(12):4667-4679.
Ploemen, JP; Wormhoudt, LW; Haenen, GR; et al. (1997) The use of human in vitro metabolic parameters to
explore the risk assessment of hazardous compounds: the case of ethylene dibromide. Toxicol Appl Pharmacol
143(1):56—69.
Poirier, MC. (2004) Chemical-induced DNA damage and human cancer risk. Nature 4:630-637.
Poulin, P; Theil, FP. (2002) Prediction of pharmacokinetics prior to in vivo studies. 1. Mechanism-based prediction
of volume of distribution. J Pharm Sci 91:129-156.
Pouyatos, B; Gearhart, CA; Fechter, LD. (2005) Acrylonitrile potentiates hearing loss and cochlear damage induced
by moderate noise exposure in rats. Toxicol Appl Pharmacol 204:46-56.
Pouyatos, B; Gearhart, C; Nelson-Miller, A; et al. (2007) Oxidative stress pathways in the potentiation of noise-
induced hearing loss by acrylonitrile. Hear Res 224:61-74.
Priston, RAJ; Dean, BJ. (1985) Tests for the induction of chromosomal aberrations, polyploidy and sister chromatid
exchanges in rat liver (RL4) cells. Prog Mutat Res 5:387-395.
Probst, GS; Hill, LE. (1985) Tests for the induction of DNA-repair synthesis in primary cultures of adult rat
hepatocytes. Prog Mutat Res 5:381-386.
Prokopczyk, B; Bertinato, P; Hoffman, D. (1988) Cyanoethylation of DNA in vivo in 3-(methylnitrosamino)-
propionitrile, an Areca-derived carcinogen. Cancer Res 48:6780-6784.
Pu, X; Kamendulis, LM; Klaunig, JE. (2006) Acrylonitrile-induced oxidative DNA damage in rat astrocytes.
Environ Mol Mutagen 47:631-638.
Pu, X; Kamendulis, LM; Klaunig, JE (2009) Acrylonitrile-induced oxidative stress and oxidative DNA damage in
male Sprague-Dawley rats. Toxicological Sciences 111: 64-71.
Quast, JF. (2002) Two-year toxicity and oncogenicity study with acrylonitrile incorporated in the drinking water of
rats. Toxicol Lett 132:153-196.
Quast, JF; Humiston, CG; Schwetz, BA; et al. (1975) A six-month oral toxicity study incorporating acrylonitrile in
the drinking water of purebred beagle dogs. Dow Toxicology Research Laboratory, Midland, MI.
Quast, JF; Wade, C; Humiston, C; et al. (1980a) Two-year toxicity and oncogenicity study with acrylonitrile
incorporated in the drinking water of rats. Dow Chemical Co., Toxicology Research Laboratory, Midland, MI.
392
DRAFT- DO NOT CITE OR QUOTE
-------
Quast, JF; Schuetz, MF; Balmer, TS; et al. (1980b) A two-year toxicity and oncogenicity study with acrylonitrile
following inhalation exposure of rats. Dow Chemical Co., Toxicology Research Laboratory, Midland, MI.
Rabello-Gay, MN; Ahmed, AE. (1980) Acrylonitrile: in vivo cytogenetic studies in mice and rats. Mutat Res
79:249-255.
Radovsky, A & Mahler, JF (1999) Nervous system. In: Maronpot, RR, Boorman, GA, and Gaul, BW ed. Pathology
of the Mouse - Reference and Atlas, St. Louis, MO, Cache River Press: 445-470.
Rajendran, S; Muthu, M. (1981) Effect of acrylonitrile on trehalase, phosphorylase and acetylcholinesterase
activities in Tribolium castaneum Herbst and Trogoderma granarium Everts. Experientia 37(8):886-887.
Recio, L; Skopek, TR. (1988a) The cellular and molcular analysis of acrylonitrile-induced mutations in human cells.
CUT Activities 8:1-6.
Recio, L; Skopek, TR. (1988b) Mutagenicity of acrylonitrile and its metabolite 2-cyanoethylene oxide in human
lymphoblasts in vitro. Mutat Res 206:297-305.
Rice, JM; Wilbourn, JD. (2000) Tumors of the nervous system in carcinogenic hazard identification. Toxicol Pathol
28:202-214.
Renwick, AG ; Lazarus, NR. (1998) Human variability and noncancer risk assessment-an analysis of the default
uncertainty factor. Regul Toxicol Pharmacol 27(1, Pt. 1 ):3—20.
Roberts, AE; Lacy, SA; Pilon, D; et al. (1989) Metabolism of acrylonitrile to 2-cyanoethylene oxide in F344 rat
liver microsomes, lung microsomes, and lung cells. Drug Metab Dispos 17:481-486.
Roberts, AE; Kedderis, GL; Turner, MJ; et al. (1991) Species comparison of acrylonitrile epoxidation by
microsomes from mice, rats and humans: relationship to epoxide concentrations in mouse and rat blood.
Carcinogenesis 12:401-404.
Rogaczewska, T. (1975) Percutaneous absorption of acrylonitrile vapour in animals. Med Pr 19:349-353.
Rongzhu, L; Suhua, W; Guangwei, X; et al. (2007) Neurobehaviroal alterations in rats exposed in acrylonitrile in
drinking water. Hum Exp Toxicol 26:179-184.
Rouisse, L; Chakrabarti, S; Tuchweber, B. (1986) Acute nephrotoxic potential of acrylonitrile in F344 rats. Res
Commun Chem Pathol Pharmacol 53:347-360.
Rudd, CJ. (1983) L5178Y tk+/ mouse lymphoma forward mutation assay of acrylonitrile. Prepared by SRI
International, Menlo Park, CA, for the Environmental Research Center, U.S. Environmental Protection Agency,
Research Triangle Park, NC; SRI Project LSU-3447, EPA Contract No. 68-02-3703.
Saillenfait, AM; Sabate, JP. (2000) Comparative development toxicities of aliphatic nitriles: in vivo and in vitro
observations. Toxicol Appl Pharmacol 163:149-163.
Saillenfait, AM; Langonne, I; Sabate, JP; et al. (1992) Embryotoxicity of acrylonitrile in whole-embryo culture.
Toxicol In Vitro 6:253-260.
Saillenfait, AM; Bonnet, P; Guenier, JP; et al. (1993) Relative developmental toxicities of inhaled aliphatic
mononitriles in rats. Fundam Appl Toxicol 20:365-375.
Saillenfait, AM; Sabate, JP; Gaspard, C. (2004) Effects of aliphatic nitriles in micromass cultures of rat embryo limb
bud cells. Toxicol In Vitro 18:311-318.
Sakurai, H. (2000) Carcinogenicity and other health effects of acrylonitrile with reference to occupational exposure
limits. Ind Health 38:165-180.
393
DRAFT- DO NOT CITE OR QUOTE
-------
Sakurai, H; Kusumoto, M. (1972) Epidemiological study of health impairment among acrylonitrile workers. J Sci
Labor 48:273-282.
Sakurai, H; Onodera, M; Utsunomiya, T; et al. (1978) Health effects of acrylonitrile in acrylic fibre factories. Br J
Ind Med 35(3):219-225.
Salazar, DE; Sorge, CL; Corcoran, GB. (1988) Obesity as a risk factor for drug-induced organ injury. VI. Increased
hepatic P-450 concentration and microsomal ethanol oxidizing activity in the obese overfed rat. Biochem Biophys
Res Commun 157:315-320.
Sandberg, EC; Slanina, P. (1980) Distribution of [1 -14C] acrylonitrile in rat and monkey. Toxicol Lett 6:187-191.
Sapota, A. (1982) The disposition of [14C]acrylonitrile in rats. Xenobiotica 12:259-264.
Sasaki, M; Sugimura, K; Yoshida, MA; et al. (1980) Cytogenetic effects of 60 chemicals on cultured human and
Chinese hamster cells. La Kromosomo 11-20:574-584.
Satayavivad, J; Thiantanawat, A; Tuntawiroon, J; et al. (1998) Alterations of central muscarinic functions during
subchronic exposure to acrylonitrile in rats. Res Commun Biol Psychol Psychiat 23:29-42.
Scelo, G; Constantinescu, V; Csiki, I; et al. (2004) Occupational exposure to vinyl chloride, acrylonitrile and styrene
and lung cancer risk (Europe). Cancer Causes Control 15:445-452.
Seger, R; Krebs, EG. (1995) The MAPK signaling cascade. FASEB J 9:726-735.
Sehgal, A; Osgood, C; Zimmering, S. (1990) Aneuploidy in drosophila. Ill: Aneuploidogens inhibit in vitro
assembly of taxol-purified Drosophila microtubules. Environ Mol Mutagen 16:217-224.
Sekihashi, K; Yamamoto, A; Matsumura, Y; et al. (2002) Comparative investigation of multiple organs of mice and
rats in the comet assay. Mutat Res 517:53-75.
Selikoff, IJ; Seidman, H. (1992) Use of death certificates in epidemiological studies, including occupational hazards:
Variations in discordance of different asbestos-associated disease on best evidence ascertainment. Am J Ind Med
22:481-492
Sharief, Y; Brown, AM; Backer, LC; et al. (1986) Sister chromatid exchange and chromosome aberration analyses
in mice after in vivo exposure to acrylonitrile, styrene, or butadiene monoxide. Environ Mutagen 8:439-448.
Sheldon, W. (1994) Tumours of the Harderian gland. I ARC Sci Pub 111:101-103.
Shell Oil Co. (1984a) The induction of mitotic gene conversion in the yeast, Saccharomyces cerevisiae JD1, by ten
selected compounds in the IPCS collaborative study on short-term tests. Submitted under TSCA Section 8D; EPA
Document No. 86-870001635; NTIS No. OTS0516216.
Shell Oil Co. (1984b) Induction of chromosome aberrations, polyploidy and sister chromatid exchanges in rat liver
cells by chemical carcinogens. Submitted under TSCA Section 8D; EPA Document No. 86-870001636; NTIS No.
OTS0515712.
Shibata, M; Inoue, K; Yoshimura, Y; et al. (2004) Simultaneous determination of hydrogen cyanide and volatile
aliphatic nitriles by headspace gas chromatography, and its application to an in vivo study of the metabolism of
acrylonitrile in the rat. Arch Toxicol 78:301-305.
Shimizu, M; Lasker, JM; Tsutsumi, M; et al. (1990) Immunohistochemical localization of ethanol inducible
P-450IIE1 in the rat alimentary tract. Gastroenterology 99:1044-1053.
Silver, EH; Szabo, S. (1982) Possible role of lipid peroxidation in the actions of acrylonitrile on the adrenals, liver
and gastrointestinal tract. Res Commun Chem Pathol Pharmacol 36:33-43.
394
DRAFT- DO NOT CITE OR QUOTE
-------
Silver, EH; McComb, DJ; Kovacs, K; et al. (1982) Limited hepatotoxic potential of acrylonitrile in rats. Toxicol
Appl Pharmacol 64:131-139.
Silver, EH; Szabo, S; Cahill, M; et al. (1987) Time-course studies of the distribution of [1 -14C]acrylonitrile in rats
after intravenous administration. J Appl Toxicol 7:303-306.
Slaga, TJ; Klein-Szanto, AJP; Triplett, LL; et al. (1981) Skin tumor promoting activity of benzoyl peroxide, a
widely used free radical generating compound. Science 213:1023-1025.
Slikker, W; Mei, N; Chen, T. (2004) N-ethyl-N-nitrosourea (ENU) increased brain mutations in prenatal and
neonatal mice but not in the adults. Toxicol Sci 81:112-120.
Smith, CA; Harrison, DJ. (1997) Association between polymorphism in gene for microsomal epoxide hydrolase and
susceptibility to emphysema. Lancet 350:630-633.
Smith, CC; O'Donovan, MR; and Martin, EA (2006). hOGGl recognizes oxidative damage using the comet assay
with greater specificity than FPG or ENDOIII. Mutagenesis 21: 185-190.
Smyth, HF; Carpenter, CP. (1948) Further experience with the range-finding test in the industrial toxicology
laboratory. J Ind Hyg Toxicol 30:63-68.
Sohda, T; Shimizu, M; Kamimura, S; et al. (1993) Immunohistochemical demonstration of ethanol-inducible P450
2E1 in rat brain. Alcohol Alcohol Suppl. lB:69-75.
Solleveld, HK; Bigner, DA; Averill, DA et al. (1986) Brain tumors in man and animals: report of a workshop.
Environ Health Perspect 68: 155-173.
Solomon, JJ; Segal, A. (1985) Direct alkylation of calf thymus DNA by acrylonitrile. Isolation of cyanoethyl
adducts of guanine and thymine and carboxyethyl adducts of adenine and cytosine. Environ Health Perspect
62:227-230.
Solomon, JJ; Cote, IL; Wortman, M; et al. (1984) In vitro alkylation of calf thymus DNA by acrylonitrile. Isolation
of cyanoethyl-adducts of guanine and thymidine and carboxyethyl-adducts of adenine and cytosine. Chem-Biol
Interact 51:167-190.
Solomon, JJ; Singh, US; Segal, A. (1993) In vitro reactions of 2-cyanoethylene oxide with calf thymus DNA.
Chem-Biol Interact 88:115-135.
Song, BJ; Matsunaga, T; Hardwick, JP; et al. (1987) Stabilization of cytochrome P450j messenger ribonucleic acid
in the diabetic rat. Mol Endocrinol 1(8): 542-547.
Song, BJ; Veech, RL; Saenger, P. (1990) Cytochrome P-450IIE1 is elevated in lymphocytes from poorly controlled
insulin-dependent diabetics. J Clin Endocrinol Metab 71:1036-1040.
Speit, G; Schutz, P; Bonzheim, I. et al. (2004) Sensitivity of the FPG protein towards alkylation damage in the
comet assay. Toxicology letters 146: 151-158.
Spiegelhalter, D; Thomas, A; Best, N; et al. (2003) WinBUGS user manual. Version 1.4. Available online at
www.mrc-bsu.cam.ac.uk/bugs/winbugs/manuall4.pdf (accessed April 21, 2009).
Sram, RJ; Beskid, O; Binkova, B; et al. (2004) Cytogenetic analysis using fluorescence in situ hybridization (FISH)
to evaluate occupational exposure to carcinogens. Toxicol Lett 149:335-344.
Starr, TB; Gause, C; Youk, AD; et al. (2004) A risk assessment for occupational acrylonitrile exposure using
epidemiology data. Risk Anal 24:587-601.
Stephens, EA; Taylor, JA; Japlan, N; et al. (1994) Ethnic variation in the CYP2E1 gene: polymorphism analysis of
695 African-Americans, European-Americans and Taiwanese. Pharmacogenetics 4:185-192.
395
DRAFT- DO NOT CITE OR QUOTE
-------
Stewart, PA; Zaebst, D; Zey, JN; et al. (1998) Exposure assessment for a study of workers exposed to acrylonitrile.
Scand J Work Environ Health 24:42-53.
Stiteler, WM; Knauf, LA; Hertzberg, RC; et al. (1993) A statistical test of compatibility of data sets to a common
dose-response model. Regul Toxicol Pharmacol 18:392-402.
Strieker, EM; Hoffman, ML; Riccardi, CJ. (2003) Increased water intake by rats maintained on high NaCl diet:
analysis of ingestive behavior. Physiol Behav 79:621-631.
Styles, JA; Clay, P; Cross, MF. (1985) Assays for the induction of gene mutations at the thymidine kinase and the
Na+/K+ ATPase loci in two different mouse lymphoma cell lines in culture. Prog Mutat Res 5:587-596.
Subramanian, U; Ahmed, AE. (1995) Intestinal toxicity of acrylonitrile: in vivo metabolism by intestinal
cytochrome P450 2E1. Toxicol Appl Pharmacol 135:1-8.
Suh, JH; Shenvi, SV; Dixon, BM; et al. (2004) Decline in transcriptional activity of Nrf2 causes age-related loss of
glutathione synthesis, which is reversible with lipoic acid. Proc Natl Acad Sci USA 101:3381 —3 3 86.
Sumner, SC; Fennell, TR; Moore, TA; et al. (1999) Role of cytochrome P450 2E1 in the metabolism of acrylamide
and acrylonitrile in mice. Chem Res Toxicol 12:1110-1116.
Swaen, GM; Bloemen, LJ; Twisk, J; et al. (1992) Mortality of workers exposed to acrylonitrile. J Occup Med
34:801-809.
Swaen, GM; Bloemen, LJ; Twisk, J; et al. (1998) Mortality update of workers exposed to acrylonitrile in the
Netherlands. Scand J Work Environ Health 24(Suppl 2): 10-16.
Swaen, GM; Bloemen, LJ; Twisk, J; et al. (2004) Mortality update of workers exposed to acrylonitrile in the
Netherlands. J Occup Environ Med 46:691-698.
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.
Symons, JM; Kreckmann, KH; Sakr, CJ; et al. (2008) Mortality among workers exposed to acrylonitirled in fiber
production: an update. J Occup Environ Med 50:550-560.
Szabo, S; Reynolds, ES; Kovacs, K. (1976) Animal model of human disease. Waterhouse-Friderichsen syndrome.
Animal model: acrylonitrile-induced adrenal apoplexy. Am J Pathol 82:653-656.
Szabo, S; Bailey, KA; Boor, PJ; et al. (1977) Acrylonitrile and tissue glutathione: differential effect of acute and
chronic interactions. Biochem Biophys Res Commun 79:32-37.
Szabo, S; Huttner, I; Kovacs, K; et al. (1980) Pathogenesis of experimental adrenal hemorrhagic necrosis
("apoplexy"): ultrastructural, biochemical, neuropharmacologic, and blood coagulation studies with acrylonitrile in
the rat. Lab Invest 42:533-546.
Szabo, S; Gallagher, GT; Silver, EH; et al. (1984) Subacute and chronic action of acrylonitrile on adrenals and
gastrointestinal tract: biochemical, functional and ultrastructural studies in the rat. J Appl Toxicol 4:131-140.
Tanaka, E. (1998) In vivo age-related changes in hepatic drug-oxidizing capacity in humans. J Clin Pharm Ther
23:247-255.
Tandon, R; Saxena, DK; Chandra, SV; et al. (1988) Testicular effects of acrylonitrile in mice. Toxicol Lett 42:55-
63.
Tardiff, R; Talbot, D; Gerin, M; et al. (1987) Urinary excretion of mercapturic acids and thiocyanate in rats exposed
to acrylonitrile: influence of dose and route of administration. Toxicol Lett 39:255-261.
396
DRAFT- DO NOT CITE OR QUOTE
-------
Tavares, R; Borba, H; Monteiro, M; et al. (1996) Monitoring of exposure to acrylonitrile by determination of
N-(2-cyanoethyl)valine at the N-terminal position of haemoglobin. Carcinogenesis 17:2655-2660.
Teo, SK; Kedderis, GL; Gargas, ML. (1994) Determination of tissue partition coefficients for volatile tissue-reactive
chemicals: acrylonitrile and its metabolite 2-cyanoethylene oxide. Toxicol Appl Pharmacol 128:92-96.
Therneau, TM; Grambsch, PM. (2000) Modeling survival data: extending the Cox model. New York: Springer-
Verlag.
Thier, R; Lewalter, J; Kempkes, M; et al. (1999) Haemoglobin adducts of acrylonitrile and ethylene oxide in
acrylonitrile workers, dependent on polymorphisms of the glutathione transferases GSTT1 and GSTM1. Arch
Toxicol 97:197-202.
Thier, R; Balkenhol, H; Lewalter, J; et al. (2001) Influence of polymorphisms of the human glutathione transferases
and cytochrome P450 2E1 enzyme on the metabolism and toxicity of ethylene oxide and acrylonitrile. Mutat Res
482:41-46.
Thier, R; Lewalter, J; Selinske, S; et al. (2002) Possible impact of human CYP2E1 polymorphisms on the
metabolism of acrylonitrile. Toxicol Lett 128:249-255.
Thiess, AM; Fleig, I. (1978) Analysis of chromosomes of workers exposed to acrylonitrile. Arch Toxicol 41:149—
152.
Thiess, AM; Frentzel-Beyme, R; Link, R; et al. (1980) Mortality study in chemical personnel of various industries
exposed to acrylonitrile. Zentralbl Arbeitsmed Arbeitsschutz Prophyl Ergonomie 30:259-267.
Tian, L; Cai, Q; Wei, H. (1998) Alterations of antioxidant enzymes and oxidative damage to macromolecules in
different organs of rats during aging. Free Radic Biol Med 24(9): 1477-1484.
Turner, MJ, Jr; Held, SD; Kedderis, GL. (1989) Identification of mercapturic acid metabolites of acrylonitrile by
tandem mass spectrometry. In: Proceedings of the 37th ASMS conference on mass spectrometry and allied topics;
May 21-26, 1989; Miami Beach, FL. East Lansing, MI: American Society for Mass Spectrometry, pp. 1361-1362.
Umeda, M; Noda, K; Tanaka, K. (1985) Assays for inhibition of metabolic cooperation by a microassay method
Prog Mutat Res 5:619-622.
U.S. EPA (Environmental Protection Agency). (1983) Health assessment document for acrylonitrile [final report].
Environmental Criteria and Environmental Assessment Office, Office of Health and Environmental Assessment,
Office of Research and Development, Research Triangle Park, NC; EPA-600/8-82-007F. Available online at
http://nepis.epa.gov/Exe/ZyPurl.cgi?Dockey=3000170U.txt (accessed April 21, 2009).
U.S. EPA (Environmental Protection Agency). (1986a) Guidelines for the health risk assessment of chemical
mixtures. Federal Register 51(185):34014-34025.
U.S. EPA (Environmental Protection Agency). (1986b) Guidelines for mutagenicity risk assessment. Federal
Register 51(185):34006-34012.
U.S. EPA (Environmental Protection Agency). (1988) 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 April 21, 2009).
U.S. EPA (Environmental Protection Agency). (1991) Guidelines for developmental toxicity risk assessment.
Federal Register 56(234):63798-63826. Available online at http://www.epa.gov/ncea/raf/rafguid.htm (accessed
April 21, 2009).
397
DRAFT- DO NOT CITE OR QUOTE
-------
U.S. EPA (Environmental Protection Agency). (1994) 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 April 21, 2009).
U.S. EPA (Environmental Protection Agency). (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 April 21, 2009).
U.S. EPA (Environmental Protection Agency). (1996) Guidelines for reproductive toxicity risk assessment. Federal
Register 61(212):56274-56322.
U.S. EPA (Environmental Protection Agency). (1998a) Guidelines for neurotoxicity risk assessment. Federal
Register 63(93):26926-26954.
U.S. EPA (Environmental Protection Agency). (1998b) Toxicological review of methyl methacrylate. Integrated
Risk Information System (IRIS), National Center for Environmental Assessment, Washington, DC. Available online
at http://www.epa.gov/iris (accessed April 21, 2009).
U.S. EPA (Environmental Protection Agency). (2000a) Science policy council handbook: risk characterization.
Office of Science Policy, Office of Research and Development, Washington, DC. EPA100B00002.
U.S. EPA (Environmental Protection Agency). (2000b) Benchmark dose technical guidance document [external
review draft]. Risk Assessment Forum, Washington, DC; EPA/630/R-00/001. Available online at
http://oaspub.epa.gov/eims/eimscomm.getfile?p_download_id=4727 (accessed April 23, 2009).
U.S. EPA (Environmental Protection Agency). (2002) A review of the reference dose concentration and reference
concentration processess. 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 April 22, 2009).
U.S. EPA (Environmental Protection Agency). (2005a) Guidelines for carcinogen risk assessment. Federal Register
70(66):17765-18717. Available online at http://www.epa.gov/cancerguidelines (accessed April 22, 2009).
U.S. EPA (Environmental Protection Agency). (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 April 22, 2009).
U.S. EPA (Environmental Protection Agency). (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 April 22, 2009).
U.S. EPA (Environmental Protection Agency). (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 April 22, 2009).
Ved Brat, SV; Williams, GM. (1982) Hepatocyte-mediated production of sister chromatid exchange in co-cultured
cells by acrylonitrile: evidence for extra cellular transport of a stable reactive intermediate. Cancer Lett 17:213-216.
Venitt, S. (1978) Letter on acrylonitrile mutagenicity. Mutat Res 57:107-109.
Venitt, S; Bushell, CT; Osborne, M. (1977) Mutagenicity of acrylonitrile (cyanoethylene) in Escherichia coli.
Mutat Res 45:283-288.
Vernon, P; Dulak, L; Deskin, R. (1990) Acute toxicologic evaluation of acrylonitrile. J Am Coll Toxicol 1:114-
115.
398
DRAFT- DO NOT CITE OR QUOTE
-------
Vieira, I; Sonnier, M; Cresteil, T. (1996) Developmental expression of CYP2E1 in the human liver.
Hypermethylation control of gene expression during the neonatal period. Eur J Biochem 238:476-483.
Vogel, EW. (1985) The Drosophila somatic recombination and mutation assay (SRM) using the white-coral somatic
eye color system. Prog Mutat Res 5:313-317.
Vogel, RA; Kirkendall, WM. (1984) Acrylonitrile (vinyl cyanide) poisoning: a case report. Tex Med 80:48-51.
Vogel, EW; Graf, U; Frei, H; et al. (1999) The results of assays in Drosophila as indicators of exposure to
carcinogens. IARC SciPubl 146:427-470.
Wakata, A; Miyamae, Y; Sato, S; et al. (1998) Evaluation of the rat micronucleus test with bone marrow and
peripheral blood: summary of the 9th collaborative study by CSGMT/JEMS.MMS. Environ Mol Mutagen 32:84-
100.
Walker, VE; Fennell, TR; Boucheron, JA; et al. (1990) Macromelecular adducts of ethylene oxide: a literature
review and a time-course study on the formation of 7-(2-hydroxyethyl)guanine following exposures of rats by
inhalation. Mutat Res 233:151-164.
Wang, W; Xia, Z; Jin, F; et al. (2000) Investigation of prevalence rate in workers exposed to acrylonitrile. Chin J
Ind Hyg. Submitted under TSCA Section 8E; EPA Document No. 89-000000313; NTIS No. OTS0559911.
Wang, H; Chanas, B; Ghanayem, BI. (2002) Cytochrome P450 2E1 (CYP2E1) is essential for acrylonitrile
metabolism to cyanide: comparative studies using CYP2El-null and wild-type mice. Drug Metab Dispos 30:911-
917.
Wauthier, V; Verbeeck, RK; Calderon, PB. (2004) Age-related changes in the protein and mRNA levels of CYP2E1
and CYP3A isoforms as well as in their hepatic activities in Wistar rats. What role for oxidative stress? Arch
Toxicol 78:131-138.
Waxweiler, RJ; Stringer, W; Wagoner, J; et al. (1976) Neoplastic risk among workers exposed to vinyl chloride.
Ann N Y Acad Sci 271:40^18.
Waxweiler, RJ; Smith, AH; Falk, H; et al. (1981) Excess lung cancer risk in a synthetic chemicals plant. Environ
Health Perspect 41:159-165.
Werner, JB; Carter, JT. (1981) Mortality of United Kingdom acrylonitrile polymerization workers. Brit J Ind Med
38:247-253.
Wester, PW; Kroes, R. (1988) Forestomach carcinogens: pathology and relevance to man. Toxicol Pathol 16:165-
171.
Whittaker, SG; Zimmermann, FK; Dicus, B; et al. (1990) Detection of induced mitotic chromosome loss in
Saccharomyces cerevisiae—an interlaboratory assessment of 12 chemicals. Mutat Res 241:225-242.
Whysner, J; Steward, RE, III; Chen, D; et al. (1998a) Formation of 8-oxodeoxyguanosine in brain DNA of rats
exposed to acrylonitrile. Arch Toxicol 72:429-438.
Whysner, J; Ross, PM; Conaway, CC; et al. (1998b) Evaluation of possible genotoxic mechanisms for acrylonitrile
tumorigenicity. Regul Toxicol Pharmacol 27:217-239.
Willhite, CC; Ferm, VH; Smith, RP. (1981) Teratogenic effects of aliphatic nitriles. Teratology 17:317-323.
Williams, GM; Tong, C; Ved Brat, S. (1985) Tests with the rat hepatocyte primary culture/DNA-repair test. Prog
Mutat Res 5:341-345.
399
DRAFT- DO NOT CITE OR QUOTE
-------
WIL Research Laboratories. (2005) Acute inhalation toxicity study of acrylonitrile in albino rats. Prepared by WIL
Research Laboratories, LLC, Ashland, OH for Shanghai SECCO Petrochemical Company, Ltd., Shanghai, China.
Submitted under TSCA Section 8E; EPA Document No. 8EHQ-0805-16067.
Wilson, RH. (1944) Health hazards encountered in the manufacture of synthetic rubber. J Am Med Assoc 124:701—
703.
Wilson, RH; Hough, GV; McCormick, W. (1948) Medical problems encountered in the manufacture of American-
made rubber. Ind Med 17:199-207.
Wood, ML; Diadaroglu, M; Gajewski, E; et al. (1990) Mechanistic studies of ionizing radiation and oxidative
mutagenesis: genetic effects of single 8-hydroxyguanine (7-hydro-8-oxoguanine) residue inserted at a unique site in
a viral genome. Biochemistry 29:7024-7032.
Wood, SM; Buffler, PA; Burau, K; et al. (1998) Mortality and morbidity of workers exposed to acrylonitrile in fiber
production. Scand J Work Environ Health 24(Suppl 2):54-62.
Working, PK; Bentley, KS; Hurtt, ME; et al. (1987) Comparison of the dominant lethal effects of acrylonitrile and
acrylamide in male F344 rats. Mutagenesis 2:215-220.
Wu, W; Su, J; Huang, M. (1995) An epidemiological study on reproductive effects in female workers exposed to
acrylonitrile. Zhonghua Yu Fang Yi Xue Za Zhi (China Preventative Medicine Magazine) 29:83-85 (Chinese).
Wu, X; Amos, CI; Kemp, BL; et al. (1998) Cytochrome P450 2E1 Dral polymorphisms in lung cancer in minority
populations. Cancer Epidemiol Biomarkers Prev 7:13-18.
Wiirgler, FE; Graf, U; Frei, H. (1985) Somatic mutation and recombination test in wings of Drosophila
melanogaster. Prog Mutat Res 5:325-40.
Xiao, W. (2000a) Study of the toxic effects of acrylonitrile on liver. Lanzhou Medical College. Submitted under
TSCA Section 8E; EPA Document No. 89-000000313; NTIS No. OTS0559911.
Xiao, W. (2000b) Effects of acrylonitrile on activity of blood cholinesterase. Lanzhou Medical College. Submitted
under TSCA Section 8E; EPA Document No. 89-000000313; NTIS No. OTS0559911.
Xu, DX; Zhu, QX; Zheng, LK; et al. (2003) Exposure to acrylonitrile induced DNA strand breakage and sex
chromosome aneuploidy in human spermatozoa. Mutat Res 537:93-100.
Yates, JM; Sumner, SC; Turner, MJ; et al. (1993) Characterization of an adduct and its degradation product
produced by the reaction of cyanoethylene oxide with deoxythymidine and DNA. Carcinogenesis 14:1363-1369.
Yates, JM; Fennell, TR; Turner, MJ, Jr; et al. (1994) Characterization of phosphodiester adducts produced by the
reaction of cyanoethylene oxide with nucleotides. Carcinogenesis 15:277-283.
Younes, M; Sharma, SC; Siegers, CP. (1986) Glutathione depletion by phorone organ specificity and effect on
hepatic microsomal mixed-function oxidase system. Drug Chem Toxicol 9:67-73.
Young, JD; Slaurer, R; Karbowski, R. (1977) The pharmacokinetic and metabolism profile of 14C-acrylonitrile given
to rats by three routes. Prepared for the Manufacturing Chemical Association by the Toxicology Research Lab,
Dow Chemical, Midland, MI.
Younger Labs. (1992) Initial submission: toxicological investigation of acrylonitrile (AN) with cover letter dated
081992. Submitted under TSCA Section 8E; EPA Document No. 88-920008086; NTIS No. OTS0546081.
Yuan, B; Wong, JL. (1991) Inactivity of acrylonitrile epoxide to modify a Ha-ras DNA in a non-focus transfection-
transformation assay. Carcinogenesis 12:787-791.
400
DRAFT- DO NOT CITE OR QUOTE
-------
Zabrodskii, PF; Kirichuk, VF; Germanchuk, VG; et al. (2000) Mechanisms of immunotoxic effects of acrylonitrile.
Bull Exp Biol Med 5:463-465.
Zeller, H; Hoffman, H; Thiess, AM; et al. (1969) Toxicity of nitriles. Results of animal experiments and industrial
experiences during 15 years. Zbl Arbeitsmd Arbeitsschutz 19:225-237.
Zhang, H; Kamendulis, LM; Jiang, J; et al. (2000) Acrylonitrile-induced morphological transformation in Syrian
hamster embryo cells. Carcinogenesis 21:727-733.
Zhang, H; Kamendulis, LM; Klaunig, JE. (2002) Mechanisms for the induction of oxidative stress in Syrian hamster
embryo cells by acrylonitrile. Toxicol Sci 67:247-255.
Zhurkov, VS; Shram, RY; Dugan, AM. (1983) Analysis of the mutagenic activity of acrylonitrile. Gig Sanit (1):71—
72.
Zielinska, E; Zubowska, M; Bodalski, J. (2004) Polymorphism within the glutathione S-transferase P1 gene is
associated with increased susceptibility to childhood malignant diseases. Pediatr Blood Cancer 43:552-559.
Zitting, A; Tenhunen, R; Savolainen, H. (1981) Effect of intraperitoneally injected acrylonitrile on liver, kidney and
brain. Acta Pharmacol Toxicol 49:412-415.
Zotova, LV. (1975) Working conditions in the production of acrylonitrile and their influence on the workers'
organism. Gig Tr Prof Zabol 8:8-11.
401
DRAFT- DO NOT CITE OR QUOTE
-------
APPENDIX A. SUMMARY OF EXTERNAL PEER REVIEW AND PUBLIC
COMMENTS AND DISPOSITION
A-l
DRAFT - DO NOT CITE OR QUOTE
-------
APPENDIX B. BENCHMARK DOSE CALCULATIONS
APPENDIX B-l. NONCANCER ORAL DOSE-RESPONSE ASSESSMENT (RfD):
BENCHMARK DOSE MODELING RESULTS EMPLOYING THE INCIDENCE OF
FORESTOMACH LESIONS (HYPERPLASIA AND HYPERKERATOSIS) IN MALE
AND FEMALE SPRAGUE-DAWLEY RATS, F344 RATS, AND B6C3F! MICE
CHRONICALLY EXPOSED ORALLY TO AN FOR 2 YEARS
As previously discussed in Section 4, no human studies currently exist that involve oral
exposures to AN. The available animal oral toxicity data, however, identify forestomach lesions
(i.e., squamous cell epithelial hyperplasia and hyperkeratosis) as the most sensitive, prevalent,
and consistent noncancer effect associated with chronic oral exposure to AN. Therefore, this
endpoint was selected as the critical effect on which to base derivation of the RfD.
Two 2-year drinking water studies, one in Sprague-Dawley rats (Quast, 2002; Quast et
al., 1980a) and the other in F344 rats (Johannsen and Levinskas, 2002b; Biodynamics 1980c),
and a 2-year gavage study in B6C3Fi mice (NTP, 2001) provided the best available dose-
response data on which to base the RfD. Candidate RfDs for AN were derived from the
incidence of forestomach lesions in Sprague-Dawley rats, F344 rats, and B6C3Fi mice using a
BMD approach. Incidences of forestomach lesions from chronic drinking water studies in male
and female Sprague-Dawley rats (Quast, 2002) and F344 rats (Johannsen and Levinskas, 2002b)
provided four sets of dose-response data from which to derive candidate RfDs, while incidences
of forestomach lesions from a chronic gavage study in male and female B6C3Fi mice provided
an additional two sets of dose-response data. These six data sets are presented in Tables B-l (for
rats) and B-2 (for mice).
Table B-l. Incidences of forestomach lesions (hyperplasia or
hyperkeratosis) in Sprague-Dawley and F344 rats exposed to AN in drinking
water for 2 years
Sex
Administered
concentration
(ppm in drinking
water)
Administered
dose"
(mg/kg-d)
Predicted internal dose metricsb
Incidence of
forestomach
lesions0
AN-AUC in rat
blood
(mg/L)
CEO-AUC in rat
blood
(mg/L)
Sprague-Dawley rats
(Sources: Quast, 2002; Quast et al., 1980a)
Male
0
0
0
0
15/80(19%)
35
3.4
2.06 x 10-2
1.83 x 10-3
15/47 (32%)
100
8.5
5.36 x 10-2
4.36 x 10-3
44/48 (92%)°
300
21.3
1.46 x 10"1
9.70 x 10-3
45/48 (94%)°
Female
0
0
0
0
20/80 (25%)
35
4.4
2.37 x 10-2
2.07 x 10-3
23/48 (48%)°
B-l DRAFT- DO NOT CITE OR QUOTE
-------
Table B-l. Incidences of forestomach lesions (hyperplasia or
hyperkeratosis) in Sprague-Dawley and F344 rats exposed to AN in drinking
water for 2 years
Sex
Administered
concentration
(ppm in drinking
water)
Administered
dose3
(mg/kg-d)
Predicted internal dose metricsb
Incidence of
forestomach
lesions0
AN-AUC in rat
blood
(mg/L)
CEO-AUC in rat
blood
(mg/L)
100
10.8
6.18 x 10-2
4.87 x 10-3
41/48 (85%)°
300
25.0
1.56 x 10-1
1.01 x 10-2
47/48 (98%)°
F344 ratsd
(Sources: Johannsen and Levinskas, 2002b; Biodynamics, 1980c)d
Male
0
0
0
0
11/159(7%)
1
0.08
4.33 x 10-4
4.06 x 10-5
3/80 (4%)
3
0.25
1.35 x 10-3
1.27 x 10-4
18/75 (24%)°
10
0.83
4.52 x 10-3
4.19 x 10-4
13/80 (16%)°
30
2.48
1.37 x 10-2
1.23 x 10-3
17/80 (22%)°
100
8.37
4.85 x 10-2
3.97 x 10-3
9/77 (12%)
Female
0
0
0
0
4/156 (3%)
1
0.12
5.73 x 10-4
5.32 x 10-5
2/80 (3%)
3
0.36
1.72 x 10-3
1.59 x 10-4
16/80 (20%)°
10
1.25
6.02 x 10-3
5.49 x 10-4
23/74 (31 %)°
30
3.65
1.79 x 10-2
1.58 x 10-3
13/80 (16%)°
100
10.90
5.63 x 10-2
4.46 x 10-3
5/74 (7%)
"Administered doses were averages calculated by the study authors based on animal BW and drinking water intake.
bThe EPA-modified rat PBPK model of Keddaris et al. (1996) was employed to predict a rat internal dose (i.e.,
either AN-AUC or CEO-AUC concentration in blood) resulting from the ingestion of the specified administered
dose of AN consumed in six bolus episodes/d.
"Indicates significantly different (atp < 0.05) from control incidence by Fisher's exact test performed by Syracuse
Research Corporation.
incidences for F344 rats do not include animals from the 6- and 12-mo sacrifices and were further adjusted to
exclude (from the denominators) rats that died between 0 and 12 mos in the study. Rats dying during this time
period were determined from page 6 of Appendix H and Table 1 in Biodynamics (1980c) and Table 8 in Johannsen
and Levinskas (2002b). Unscheduled deaths between 0 and 12 mos in the study occurred in two female controls,
two males at 3 ppm, three females at 10 ppm, and three males and three females at 100 ppm.
B-2
DRAFT- DO NOT CITE OR QUOTE
-------
Table B-2. Incidences of forestomach lesions (hyperplasia or
hyperkeratosis) in male and female B6C3Fi mice administered AN via
gavage for 2 years
Dose (mg/kg-d)a
Lesion site and type
0
2.5
10
20
Males
Forestomach hyperplasia or hyperkeratosis
2/50
4/50
10/503
13/50b
(4%)
(8%)
(20%)
(26%)
Females
Forestomach hyperplasia or hyperkeratosis
2/50
2/50
5/50
8/50a
(4%)
(4%)
(10%)
(16%)
aSignificantly elevated above vehicle control as determined by EPA using Fisher's exact test (p < 0.05).
bSignificantly elevated above vehicle control as determined by EPA using Fisher's exact test (p < 0.01).
Source: NTP(2001).
The incidences of forestomach lesions observed following 2 years of AN exposure in
male and female SD and F344 rats were modeled using AN and CEO in blood, expressed in
mg/L, as internal dose metrics. In addition, incidences of these same lesions were modeled in
male and female SD and F344 rats, as well as male and female B6C3Fi mice, employing
administered dose. In all cases, all of the dichotomous dose-response models available in EPA's
BMDS software (version 2.0) (i.e., the gamma, logistic, log-logistic, probit, log-probit,
multistage, Weibull, and quantal-linear models) were fit to these incidence data. Because the
incidence of forestomach lesions in male and female F344 rats did not increase monotonically
across all administered concentrations, however, only incidence data from the three lowest
concentrations (i.e., 0, 1, and 3 ppm) were used in dose-response modeling. Similarly, in male
Sprague-Dawley rats, the incidence data from the highest dose group needed to be dropped prior
to BMD modeling. In most cases, several models fit the data equally well (i.e., exhibited %
goodness-of-fit p values greater than 0.1). Of those models exhibiting adequate fit, the selected
model was the one with the lowest AIC value. BMDLio and BMDL0s estimates were derived
from the selected model. If more than one model shared the lowest AIC, the mean BMDLio and
BMDL05 were calculated, as per the EPA's Benchmark Dose Technical Guidance Document
(U.S. EPA, 2000b).
In this appendix, detailed results of the dose-response modeling for forestomach lesions
in male and female SD and F344 rats, and B6C3Fi mice are presented. For each species, strain,
and sex, summaries of the dose-response data (i.e., animal administered and internal doses and
incidence data) are presented (Table B-l for rats and Table B-2 for mice). Then, again for each
species, strain, and sex, Tables B-3 (Sprague-Dawley rats), B-4 (F344 rats), and B-5 (B6C3Fi
mice) summarize the results of the dose-response modeling. Each of these tables is then
B-3
DRAFT- DO NOT CITE OR QUOTE
-------
followed by the standard output from EPA's BMDS, version 2.0, for the dose-response models
that did not exhibit statistically significant lack of fit.
Table B-3. Summary of the BMD modeling results based on the incidence of
forestomach lesions (hyperplasia or hyperkeratosis) in male and female
Sprague-Dawley rats exposed to AN in drinking water for 2 years
Sex
Endpoint
Dose metric
Selected model(s)a
x2
/j-valuc
AIC
BMDL10b
BMDL0Sb
Male
Forestomach
lesions
Estimated
administered
dose
(mg/kg-d)
Two-stage multistage
(1)
0.18
169.45
1.27
7.20 x 10"1
Predicted AN
in blood
(mg/L)
Two-stage multistage
(1)
0.24
169.01
7.72 x 10-3
4.32 x 10-3
Predicted CEO
in blood
(mg/L)
Two-stage multistage
(1)
0.12
170.09
6.82 x 10-4
3.90 x 10-4
B-4 DRAFT- DO NOT CITE OR QUOTE
-------
Table B-3. Summary of the BMD modeling results based on the incidence of
forestomach lesions (hyperplasia or hyperkeratosis) in male and female
Sprague-Dawley rats exposed to AN in drinking water for 2 years
Sex
Endpoint
Dose metric
Selected model(s)a
x2
/j-valuc
AIC
BMDL10b
BMDL0Sb
gamma
0.38
212.77
6.87 x 10-1
3.34 x 10"1
logistic
0.45
211.31
1.24
6.39 x 10 1
Estimated
administered
dose
log-logistic
0.98
212.03
1.41
9.17 x 10"1
log-probit
0.74
212.14
1.38
9.61 x 10-1
(mg/kg-d)
one-stage multistage
0.31
212.40
6.30 x 10-1
3.07 x 10"1
Weibull
0.31
213.07
6.74 x 10-1
3.28 x 10"1
quantal linear
0.31
212.40
6.30 x 10-1
3.07 x 10"1
gamma
0.32
212.96
3.76 x 10-3
1.83 x 10-3
Forestomach
lesions
logistic
0.23
212.00
7.07 x 10-3
3.64 x 10-3
Female
Predicted AN
log-logistic
0.98
212.03
6.93 x 10-3
4.36 x 10-3
in blood
log-probit
0.70
212.18
7.02 x 10-3
4.88 x 10-3
(mg/L)
one-stage multistage
0.42
211.79
3.62 x 10 3
1.76 x 10 3
Weibull
0.28
213.19
3.72 x 10-3
1.81 x 10-3
quantal linear
0.42
211.79
3.62 x 10 3
1.76 x 103
gamma
0.51
212.45
3.53 x 10-4
1.74 x 10-4
Predicted CEO
logistic
0.69
210.72
5.58 x 10 4
2.88 x 10 4
in blood
log-logistic
0.87
212.06
7.25 x 10-4
4.87 x 10-4
(mg/L)
probit
0.33
211.86
5.87 x 10-4
3.00 x 10-4
log-probit
0.86
212.06
7.17 x 10-4
5.15 x 10-4
aAll dichotomous models in EPA's BMDS (version 2.0) were fit to the incidence of forestomach lesions
(hyperplasia or hyperkeratosis) in Sprague-Dawley rats using the data presented in Table B-l. For BMD modeling,
three different dose metrics were employed: (1) administered animal dose (estimated) expressed in mg/kg-d, (2)
AN in blood (predicted) expressed in mg/L, and (3) CEO in blood (predicted) expressed in mg/L. Adequate fit of a
model was achieved if the %2 goodness-of-fit statistic yielded a />-value >0.1. The numbers in parentheses indicate
the number of dose groups dropped in order to obtain an adequate fit, starting with the highest dose group. Of those
models exhibiting adequate fit, the selected model was the model with the lowest AIC value, and is indicated in
bold in the table.
bBMDLio and BMDL05 estimates were derived from the selected model. If more than one model shared the lowest
AIC, the mean BMDLio and BMDL0s were calculated, as per the EPA's Benchmark Dose Technical Guidance
Document (U.S. EPA, 2000b).
Sources: Quast (2002); Quast et al. (1980a).
B-5
DRAFT- DO NOT CITE OR QUOTE
-------
SD Male Rats
"O
3
"§ 0.6
o
03
0.4
0.2
0 12 3 4
dose
16:25 10/03 2008
Multistage Model. (Version: 3.0; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpAB.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpAB.plt
Fri Oct 03 16:25:01 2008
BMDS Model Run
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2)]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = DOSE
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Multistage
BMDL
: Forestomach lesions (administered dose)
Multistage Model with 0.95 Confidence Level
B-6
DRAFT- DO NOT CITE OR QUOTE
-------
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
Background = 0.110075
Beta(l) = 0
Beta(2) = 0.0325391
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Beta(2)
Background 1 -0.37
Beta(2) -0.37 1
Parameter Estimates
Variable
Background
Beta(1)
Beta(2)
Estimate
0.170685
0
0 . 0278185
Std. Err.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model Log(likelihood) # Param's Deviance Test d.f. P-value
Full model -81.807 3
Fitted model -82.7268 2 1.83969 1 0.175
Reduced model -119.21 1 74.8052 2 <.0001
AIC:
169 .454
Goodness of Fit
Dose Est._Prob. Expected Observed Size
Scaled
Residual
0.0000
3.4000
8.5000
0.1707
0.3987
0.8889
13.655
18.741
42 .666
15.000
15 . 000
44 . 000
80
47
48
0.400
-1.114
0.613
Chi 2 =1.78
d.f. = 1
P-value = 0.1825
B-7
DRAFT- DO NOT CITE OR QUOTE
-------
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 1.35789
BMDL = 0.719865
BMDU = 1.59564
Taken together, (0.719865, 1.59564) is a 90 % two-sided confidence
interval for the BMD
B-8 DRAFT- DO NOT CITE OR QUOTE
-------
SD Male Rats
"O
3
"§ 0.6
o
03
0.4
0.2
0.01 0.02 0.03 0.04 0.05
dose
11:55 10/06 2008
Multistage Model. (Version: 3.0; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpBB.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpBB.plt
Mon Oct 06 11:55:17 2008
BMDS Model Run
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2)]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = DOSE
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Multistage
BMDL
: Forestomach lesions (AN in blood)
Multistage Model with 0.95 Confidence Level
B-9
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.121757
Beta(l) = 0
Beta(2) = 815.045
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Beta(2)
Background 1 -0.36
Beta(2) -0.36 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0.172109 * * *
Beta(1) 0 * * *
Beta(2) 714.086 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-81.807
-82 .5048
-119.21
# Param's
3
2
1
Deviance Test d.f.
1.39565
74.8052
P-value
0 .2375
<.0001
AIC:
169 . 01
Goodness of Fit
Dose Est._Prob. Expected Observed Size
Scaled
Residual
0.0000
0.0206
0.0536
0.1721
0.3885
0.8936
13 .769
18 .261
42.892
15.000
15 . 000
44 . 000
80
47
48
0.365
-0 . 976
0.519
Chi 2 = 1.35
d.f. = 1
P-value = 0.2445
B-10
DRAFT- DO NOT CITE OR QUOTE
-------
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.0084753
BMDL = 0.0043156
BMDU = 0.0099743
Taken together, (0.0043156, 0.0099743) is a 90 % two-sided confidence
interval for the BMD
B-l 1 DRAFT- DO NOT CITE OR QUOTE
-------
SD Male Rats
"O
3
"§ 0.6
o
03
0.4
0.2
0 0.0005 0.001 0.0015 0.002 0.0025 0.003 0.0035 0.004 0.0045
dose
12:07 10/06 2008
Multistage Model. (Version: 3.0; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpCC.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpCC.plt
Mon Oct 06 12:07:44 2008
BMDS Model Run
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2)]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = DOSE
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Multistage
BMDL
: Forestomach lesions (CEO in blood)
Multistage Model with 0.95 Confidence Level
B-12
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 = 0.0947166
Beta(l) = 0
Beta(2) = 124268
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Beta(2)
Background 1 -0.37
Beta(2) -0.37 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0.169148 * * *
Beta(1) 0 * * *
Beta(2) 102914 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-81.807
-83 . 0465
-119.21
# Param'
3
2
1
Deviance Test d.f.
2 .47897
74.8052
P-value
0.1154
<.0001
AIC:
170.093
Dose
Est. Prob.
Goodness of Fit
Expected
Observed
Size
Scaled
Residual
0.0000
0.0018
0.0044
0.1691
0.4114
0.8825
13 .532
19.334
42 .362
15 . 000
15 . 000
44 . 000
80
47
48
0.438
-1.285
0.734
Chi 2 = 2.3£
d.f. = 1
P-value = 0.1228
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.000705982
BMDL = 0.000390266
BMDU = 0.000828151
Taken together, (0.0003 90266, 0.000828151) is a 90 % two-sided confidence
interval for the BMD
B-13 DRAFT- DO NOT CITE OR QUOTE
-------
SD Female Rats: Forestomach lesions (administered dose)
Gamma Multi-Hit Model with 0.95 Confidence Level
~o
CD
-*—>
o
(D
=1
Total number of observations = 4
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
Default Initial (and Specified) Parameter Values
Background = 0.253086
Slope = 0.194377
Gamma Multi-Hit
BMDL BMD
0 5 10 15 20 25
dose
10/06 2008
B-14
DRAFT- DO NOT CITE OR QUOTE
-------
Power =
1.66787
Asymptotic Correlation Matrix of Parameter Estimates
Background
Slope
Power
Background
1
0.13
0.25
Slope
0.13
1
0 . 95
Power
0.25
0 . 95
1
Parameter Estimates
Variable
Background
Slope
Power
Estimate
0 .247085
0 .22327
1.72213
Std. Err.
0 . 0476211
0 . 0913548
0.704799
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
0.153749 0.340421
0.0442181 0 .402322
0.340745 3.10351
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-103 . 017
-103.386
-152.026
212 .772
# Param's
4
3
1
Deviance Test d.f.
0.738711
98 . 0188
P-value
0.3901
<.0001
Dose
Goodness of Fit
Est. Prob.
Expected
Observed
Size
Scaled
Residual
0.0000
4.4000
10.8000
25.0000
Chia2 = 0.78
0.2471
0.5044
0.8219
0.9879
d.f. = 1
19.767 20.000 80 0.060
24.212 23.000 48 -0.350
39.451 41.000 48 0.584
47.420 47.000 48 -0.555
P-value = 0.3785
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 1.12
BMDL = 0.334344
B-15
DRAFT- DO NOT CITE OR QUOTE
-------
Logistic Model with 0.95 Confidence Level
~o
CD
o
(D
o
03
0.6
0.4
0.2
14:44 10/06 2008
10 15 20 25
dose
Logistic Model. (Version: 2.12; Date: 05/16/2008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpDB.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpDB.plt
Mon Oct 06 14:44:22 2008
BMDS Model Run
The form of the probability function is:
P[response] = 1/[1+EXP(-intercept-slope*dose)]
Dependent variable = Response
Independent variable = DOSE
Slope parameter is not restricted
Total number of observations = 4
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
Default Initial Parameter Values
background = 0 Specified
intercept = -0.808992
slope = 0.180053
Asymptotic Correlation Matrix of Parameter Estimates
Logistic
-BMDL BMD
B-16
DRAFT- DO NOT CITE OR QUOTE
-------
( *** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
intercept slope
intercept 1 -0.66
slope -0.66 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
intercept -1.07104 0.225024 -1.51208 -0.630006
slope 0.23795 0.0368793 0.165668 0.310232
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-103.017
-103.655
-152.026
# Param'
4
2
1
Deviance Test d.f.
1.27604
98 . 0188
P-value
0.5283
<.0001
AIC:
211.31
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.2552 20.416 20.000 80 -0.107
4.4000 0.4940 23.711 23.000 48 -0.205
10.8000 0.8174 39.235 41.000 48 0.659
25.0000 0.9924 47.637 47.000 48 -1.062
Chia2 = 1.62 d.f. = 2 P-value = 0.4456
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.78799
BMDL = 0.639403
B-17
DRAFT- DO NOT CITE OR QUOTE
-------
Log-Logistic Model with 0.95 Confidence Level
0.8
"O
(D
O
CD
= 1
Total number of observations = 4
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 Parameter Values
background = 0.25
intercept = -4.55921
slope = 2.51883
Log-Logistic
BMDL
B-18
DRAFT- DO NOT CITE OR QUOTE
-------
Asymptotic Correlation Matrix of Parameter Estimates
background intercept
background 1 -0.3
intercept -0.3 1
slope 0.2 -0.96
slope
0.2
-0 . 96
1
Variable
background
intercept
slope
Parameter Estimates
Estimate
0 .250085
-4 .54432
2 .51011
Std. Err.
* - Indicates that this value is not calculated.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
#
Log(likelihood)
-103.017
-103.017
-152.026
Deviance Test d.f.
Param's
4
3 0.000749029
1 98.0188
P-value
0 . 9782
<.0001
AIC:
212.034
Dose
Est. Prob.
Goodness of Fit
Expected
Observed
Size
Scaled
Residual
0.0000
4.4000
10.8000
25.0000
0.2501
0.4785
0.8550
0.9788
20.007
22.970
41. 041
46.981
20.000
23 . 000
41. 000
47 . 000
80
48
48
48
-0 . 002
0 . 009
-0 . 017
0 . 019
Chia2 =0.00
d.f. = 1
P-value = 0.9782
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 1.89151
BMDL = 0.917407
B-19
DRAFT- DO NOT CITE OR QUOTE
-------
LogProbit Model with 0.95 Confidence Level
0.8
"O
(D
O
CD
= 1
Total number of observations = 4
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
LogProbit
BMDU BMD
B-20
DRAFT- DO NOT CITE OR QUOTE
-------
background =
intercept =
slope =
0.25
-2 .53925
1.39624
Asymptotic Correlation Matrix of Parameter Estimates
background intercept
background 1 -0.31
intercept -0.31 1
slope 0.21 -0.96
slope
0.21
-0 . 96
1
Parameter Estimates
Variable
background
intercept
slope
Estimate
0 .248842
-2.58881
1.42762
95.0% Wald Confidence Interval
Std. Err. Lower Conf. Limit Upper Conf. Limit
0.0480048 0.154755 0.34293
0.657292 -3.87708 -1.30054
0.288071 0.86301 1.99223
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
Log(likelihood)
-103.017
-103 . 07
-152.026
# Param'
4
3
1
Deviance Test d.f.
0.106572
98 . 0188
P-value
0.7441
<.0001
AIC:
212.14
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.2488 19.907 20.000 80 0.024
4.4000 0.4876 23.406 23.000 48 -0.117
10.8000 0.8427 40.448 41.000 48 0.219
25.0000 0.9832 47.192 47.000 48 -0.216
Chia2 = 0.11 d.f. = 1 P-value = 0.7415
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 1.93713
BMDL = 0.961058
B-21
DRAFT- DO NOT CITE OR QUOTE
-------
Multistage Model with 0.95 Confidence Level
~o
0
o
(D
-------
Default Initial Parameter Values
Background = 0.18073
Beta(1) = 0.14774
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.46
Beta(1) -0.46 1
Parameter Estimates
Variable
Background
Beta(1)
Estimate
0 .235056
0.131634
Std. Err.
* - Indicates that this value is not calculated.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-103 . 017
-104.199
-152.026
212.397
# Param's
4
2
1
Deviance Test d.f.
2 .36383
98 . 0188
P-value
0.3067
<.0001
Dose
Goodness of Fit
Est. Prob.
Expected
Observed
Size
Scaled
Residual
0.0000
4.4000
10.8000
25.0000
Chia2 = 2.35
0.2351
0.5714
0.8154
0.9715
d.f. = 2
18.804 20.000 80 0.315
27.425 23.000 48 -1.291
39.139 41.000 48 0.692
46.633 47.000 48 0.318
P-value = 0.3095
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0 . 05
Extra risk
0 . 95
0.389667
0.306761
0.505425
Taken together, (0.306761, 0.505425) is a 90
interval for the BMD
% two-sided confidence
B-23
DRAFT- DO NOT CITE OR QUOTE
-------
Weibull Model with 0.95 Confidence Level
~o
0
o
(D
=1
Total number of observations = 4
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
Default Initial (and Specified) Parameter Values
Background = 0.253086
Slope = 0.0679496
Power = 1.19621
Asymptotic Correlation Matrix of Parameter Estimates
Weibull
BMDL BMD
0 5 10 15 20 25
dose
10/06 2008
B-24
DRAFT- DO NOT CITE OR QUOTE
-------
Background
Background 1
Slope -0.31
Power 0.23
Slope
-0.31
1
-0 . 97
Power
0.23
-0 . 97
1
Parameter Estimates
Variable
Background
Slope
Power
Estimate
0 .245056
0 . 0673284
1.27711
Std. Err.
0 . 0471325
0 . 0436908
0 .255412
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
0.152678 0.337434
-0.018304 0.152961
0.776514 1.77771
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-103.017
-103 .534
-152.026
# Param'
4
3
1
Deviance Test d.f.
1. 03516
98 . 0188
P-value
0.308S
<.0001
AIC:
213 . 069
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.2451 19.604 20.000 80 0.103
4.4000 0.5170 24.816 23.000 48 -0.525
10.8000 0.8150 39.118 41.000 48 0.700
25.0000 0.9876 47.404 47.000 48 -0.526
Chia2 = 1.05 d.f. = 1 P-value = 0.3051
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.808158
BMDL = 0.327975
B-25
DRAFT- DO NOT CITE OR QUOTE
-------
Quantal Linear Model with 0.95 Confidence Level
~o
0
o
(D
-------
( *** The model parameter(s) -Power
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Slope
Background 1 -0.31
Slope -0.31 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0.235057 0.0449515 0.146954 0.323161
Slope 0.131633 0.019868 0.0926927 0.170574
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-103.017
-104.199
-152.026
# Param'
4
2
1
Deviance Test d.f.
2 .36383
98 . 0188
P-value
0.3067
<.0001
AIC:
212 .397
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.2351 18.805 20.000 80 0.315
4.4000 0.5714 27.425 23.000 48 -1.291
10.8000 0.8154 39.139 41.000 48 0.692
25.0000 0.9715 46.633 47.000 48 0.318
Chia2 = 2.35 d.f. = 2 P-value = 0.3095
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.389668
BMDL = 0.306761
B-27
DRAFT- DO NOT CITE OR QUOTE
-------
SD Female Rats: Nonneoplastic forestomach lesions (AN in blood)
Gamma Multi-Hit Model with 0.95 Confidence Level
~o
CD
-*—>
o
(D
=1
Total number of observations = 4
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
Default Initial (and Specified) Parameter Values
Background = 0.253086
Gamma Multi-Hit
ESMDL
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
dose
10/06 2008
B-28
DRAFT- DO NOT CITE OR QUOTE
-------
Slope =
Power =
28 . 0854
1.40734
Asymptotic Correlation Matrix of Parameter Estimates
Background
Slope
Power
Background
1
0.12
0 .24
Slope
0.12
1
0 . 94
Power
0 .24
0 . 94
1
Parameter Estimates
Variable
Background
Slope
Power
Estimate
0 .24677
32 . 9782
1.45925
Std. Err.
0.0475578
13 .6708
0.592225
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
0.153558 0.339981
6.18402 59.7725
0.298512 2.61999
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-103.017
-103.482
-152.026
# Param's
4
3
1
Deviance Test d.f.
0 . 931283
98 . 0188
P-value
0.3345
<.0001
AIC:
212 . 965
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.2468 19.742 20.000 80 0.067
0.0237 0.5082 24.395 23.000 48 -0.403
0.0618 0.8175 39.238 41.000 48 0.658
0.1560 0.9886 47.454 47.000 48 -0.617
Chia2 = 0.98 d.f. = 1 P-value = 0.3220
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00495567
BMDL = 0.00183293
B-29
DRAFT- DO NOT CITE OR QUOTE
-------
Logistic Model with 0.95 Confidence Level
0.8
"O
(D
O
CD
-------
( *** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
intercept slope
intercept 1 -0.65
slope -0.65 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
intercept -1.0292 0.221998 -1.46431 -0.594092
slope 40.4482 6.53178 27.6462 53.2503
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-103.017
-104 . 001
-152.026
# Param'
4
2
1
Deviance Test d.f.
1. 9683
98 . 0188
P-value
0.373S
<.0001
AIC:
212.002
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.2632 21.059 20.000 80 -0.269
0.0237 0.4824 23.153 23.000 48 -0.044
0.0618 0.8131 39.030 41.000 48 0.729
0.1560 0.9949 47.757 47.000 48 -1.539
Chia2 = 2.98 d.f. = 2 P-value = 0.2258
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00450626
BMDL = 0.00363943
B-31
DRAFT- DO NOT CITE OR QUOTE
-------
Log-Logistic Model with 0.95 Confidence Level
0.8
"O
(D
O
CD
= 1
Total number of observations = 4
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 Parameter Values
background = 0.25
intercept = 7.87595
slope = 2.32252
Log-Logistic
BMDL
B-32
DRAFT- DO NOT CITE OR QUOTE
-------
Asymptotic Correlation Matrix of Parameter Estimates
background
intercept
slope
background
1
0.11
0.2
intercept
0.11
1
0 . 98
slope
0.2
0 . 98
1
Variable
background
intercept
slope
Parameter Estimates
Estimate
0 .249911
7 . 9009
2 .32966
Std. Err.
* - Indicates that this value is not calculated.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
#
Log(likelihood)
-103.017
-103.017
-152.026
Deviance Test d.f.
Param's
4
3 0.000533199
1 98.0188
P-value
0.9816
<.0001
AIC:
212.034
Dose
Est. Prob.
Goodness of Fit
Expected
Observed
Size
Scaled
Residual
0.0000
0 . 0237
0.0618
0.1560
0.2499
0.4797
0.8535
0.9795
19.993
23 . 025
40 . 966
47.017
20.000
23 . 000
41. 000
47 . 000
80
48
48
48
0 . 002
-0 . 007
0 . 014
-0 . 017
Chia2 =0.00
d.f. = 1
P-value = 0.9816
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00951077
BMDL = 0.00435683
B-33
DRAFT- DO NOT CITE OR QUOTE
-------
LogProbit Model with 0.95 Confidence Level
1
0.8
~o
CD
O
(D
O
03
0.6
0.4
0.2
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
dose
16:04 10/06 2008
Probit Model. (Version: 3.1; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpF5.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpF5.plt
Mon Oct 06 16:04:48 2008
BMDS Model Run
The form of the probability function is:
P[response] = Background
+ (1-Background) * CuniNorm (Intercept+Slope*Log (Dose) ) ,
where CuniNorm(.) is the cumulative normal distribution function
Dependent variable = Response
Independent variable = DOSE
Slope parameter is restricted as slope >= 1
Total number of observations = 4
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
B-34 DRAFT- DO NOT CITE OR QUOTE
LogProbit
BMDL
-------
background = 0.2 5
intercept = 4.35171
slope = 1.28667
Asymptotic Correlation Matrix of Parameter Estimates
background intercept slope
background 1 0.12 0.21
intercept 0.12 1 0.97
slope 0.21 0.97 1
Parameter Estimates
Variable
background
intercept
slope
Estimate
0 .248635
4 .47618
1.32091
95.0% Wald Confidence Interval
Std. Err. Lower Conf. Limit Upper Conf. Limit
0.0479558 0.154644 0.342627
0.818254 2.87243 6.07993
0.267134 0.797338 1.84448
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-103.017
-103.09
-152.026
# Param'
4
3
1
Deviance Test d.f.
0.14675
98 . 0188
P-value
0.7017
<.0001
AIC:
212.18
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.2486 19.891 20.000 80 0.028
0.0237 0.4892 23.484 23.000 48 -0.140
0.0618 0.8406 40.348 41.000 48 0.257
0.1560 0.9838 47.222 47.000 48 -0.253
Chia2 = 0.15 d.f. = 1 P-value = 0.6982
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00971634
BMDL = 0.00488108
B-35
DRAFT- DO NOT CITE OR QUOTE
-------
Multistage Model with 0.95 Confidence Level
~o
0
o
(D
-------
Default Initial Parameter Values
Background = 0.2349
Beta(1) = 23.4597
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.46
Beta(1) -0.46 1
Parameter Estimates
Variable
Background
Beta(1)
Estimate
0 .237678
22.7713
Std. Err.
* - Indicates that this value is not calculated.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Deviance Test d.f.
Log(likelihood)
-103.017
-103.895
-152.026
211.789
# Param's
4
2
1
1.75601
98 . 0188
P-value
0 .4156
<.0001
Dose
Est. Prob.
Goodness of Fit
Expected Observed Size
Scaled
Residual
0.0000
0.2377
19.014
20.000
80
0.259
0 . 0237
0.5556
26.670
23 . 000
48
-1. 066
0.0618
0.8134
39 . 042
41. 000
48
0.725
0.1560
0.9782
46.951
47 . 000
48
0 . 048
Chia2 =1.73
d.f. = 2
P-value = 0.4207
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0 . 05
Extra risk
0 . 95
0.00225254
0.00176165
0.00294313
Taken together, (0.00176165, 0.00294313) is a 90
interval for the BMD
% two-sided confidence
B-37
DRAFT- DO NOT CITE OR QUOTE
-------
Weibull Model with 0.95 Confidence Level
~o
0
o
(D
=1
Total number of observations = 4
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
Default Initial (and Specified) Parameter Values
Background = 0.253086
Slope = 24.7803
Power = 1.10265
Asymptotic Correlation Matrix of Parameter Estimates
Weibull
EiMDL BMD
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
dose
10/06 2008
B-38
DRAFT- DO NOT CITE OR QUOTE
-------
Background
Slope
Power
Background
1
0.14
0.23
Slope
0.14
1
0 . 97
Power
0.23
0 . 97
1
Parameter Estimates
Variable
Background
Slope
Power
Estimate
0 .244833
36 .5794
1.17393
Std. Err.
0 . 0470971
23 .8993
0 .234182
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
0.152525 0.337142
-10.2624 83.4211
0.71494 1.63291
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-103.017
-103.595
-152.026
# Param'
4
3
1
Deviance Test d.f.
1.15727
98 . 0188
P-value
0 .282
<.0001
AIC:
213.191
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.2448 19.587 20.000 80 0.107
0.0237 0.5195 24.938 23.000 48 -0.560
0.0618 0.8125 38.998 41.000 48 0.740
0.1560 0.9879 47.417 47.000 48 -0.550
Chia2 = 1.18 d.f. = 1 P-value = 0.2783
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00371148
BMDL = 0.00181169
B-39
DRAFT- DO NOT CITE OR QUOTE
-------
Quantal Linear Model with 0.95 Confidence Level
~o
0
o
(D
-------
( *** The model parameter(s) -Power
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Slope
Background 1 -0.31
Slope -0.31 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0.237678 0.0452051 0.149078 0.326278
Slope 22.7713 3.5337 15.8454 29.6972
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-103.017
-103.895
-152.026
# Param'
4
2
1
Deviance Test d.f.
1.75601
98 . 0188
P-value
0 .4156
<.0001
AIC:
211.789
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.2377 19.014 20.000 80 0.259
0.0237 0.5556 26.670 23.000 48 -1.066
0.0618 0.8134 39.042 41.000 48 0.725
0.1560 0.9782 46.951 47.000 48 0.048
Chia2 = 1.73 d.f. = 2 P-value = 0.4207
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00225254
BMDL = 0.00176165
B-41
DRAFT- DO NOT CITE OR QUOTE
-------
SD Female Rats: Nonneoplastic forestomach lesions (CEO in blood)
Gamma Multi-Hit Model with 0.95 Confidence Level
"O
(D
-i—¦
o
CD
0.6
O
03
0.4
0.2
0.002 0.004 0.006 0.008 0.01
dose
09:34 10/07 2008
Gamma Model. (Version: 2.13; Date: 05/16/2008)
Input Data File: C:\USEPA\BMDS2\Temp\tmplE.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmplE.plt
Tue Oct 07 09:34:01 2008
BMDS Model Run
The form of the probability function is:
P[response]= background+(1-background)*CumGamma[slope*dose,power]
where CumGamma(.) is the cummulative Gamma distribution function
Dependent variable = Response
Independent variable = DOSE
Power parameter is restricted as power >=1
Total number of observations = 4
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
Default Initial (and Specified) Parameter Values
Background = 0.253086
Gamma Multi-Hit
3MDL
B-42
DRAFT- DO NOT CITE OR QUOTE
-------
Slope =
Power =
550.799
2.0706
Asymptotic Correlation Matrix of Parameter Estimates
Background
Slope
Power
Background
1
0.15
0.26
Slope
0.15
1
0 . 96
Power
0.26
0 . 96
1
Parameter Estimates
Variable
Background
Slope
Power
Estimate
0 .24778
600.839
2 . 07213
Std. Err.
0 . 0477681
241.065
0.851839
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
0.154157 0.341404
128.36 1073.32
0.402552 3.7417
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-103 . 017
-103.226
-152.026
212.453
# Param's
4
3
1
Deviance Test d.f.
0 .419043
98 . 0188
P-value
0.5174
<.0001
Dose
Goodness of Fit
Est. Prob.
Expected
Observed
Size
Scaled
Residual
0.0000
0 . 0021
0 . 0049
0.0101
Chia2 = 0.44
0.2478
0.4972
0.8304
0 . 9863
d.f. = 1
19.822 20.000 80 0.046
23.867 23.000 48 -0.250
39.860 41.000 48 0.438
47.341 47.000 48 -0.423
P-value = 0.5092
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.000640233
BMDL = 0.000173836
B-43
DRAFT- DO NOT CITE OR QUOTE
-------
Logistic Model with 0.95 Confidence Level
0.8
"O
(D
O
CD
-------
( *** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
intercept slope
intercept 1 -0.66
slope -0.66 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
intercept -1.11487 0.227873 -1.56149 -0.668246
slope 547.334 80.7582 389.05 705.617
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-103.017
-103.359
-152.026
# Param'
4
2
1
Deviance Test d.f.
0.684912
98 . 0188
P-value
0.71
<.0001
AIC:
210.718
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.2470 19.757 20.000 80 0.063
0.0021 0.5045 24.217 23.000 48 -0.351
0.0049 0.8250 39.600 41.000 48 0.532
0.0101 0.9880 47.425 47.000 48 -0.565
Chia2 = 0.73 d.f. = 2 P-value = 0.6946
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.000352967
BMDL = 0.000287581
B-45
DRAFT- DO NOT CITE OR QUOTE
-------
Log-Logistic Model with 0.95 Confidence Level
0.8
"O
(D
O
CD
= 1
Total number of observations = 4
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 Parameter Values
background = 0.25
intercept = 16.1818
slope = 2.75711
Log-Logistic
BMDL
B-46
DRAFT- DO NOT CITE OR QUOTE
-------
Asymptotic Correlation Matrix of Parameter Estimates
background
intercept
slope
background
1
0.17
0.21
intercept
0.17
1
0 . 99
slope
0.21
0 . 99
1
Variable
background
intercept
slope
Parameter Estimates
Estimate
0 .250617
15.839
2 .69969
Std. Err.
* - Indicates that this value is not calculated.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
Log(likelihood)
-103.017
-103 . 03
-152.026
# Param's
4
3
1
Deviance Test d.f.
0 . 027282
98 . 0188
P-value
0.8688
<.0001
AIC:
212 . 061
Dose
Est. Prob.
Goodness of Fit
Expected
Observed
Size
Scaled
Residual
0.0000
0 . 0021
0 . 0049
0.0101
0.2506
0.4757
0.8592
0.9766
20 . 049
22 .833
41.243
46.875
20.000
23 . 000
41. 000
47 . 000
80
48
48
48
-0 . 013
0 . 048
-0.101
0.119
Chia2 =0.03
d.f. = 1
P-value = 0.8699
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.000951374
BMDL = 0.000486765
B-47
DRAFT- DO NOT CITE OR QUOTE
-------
Probit Model with 0.95 Confidence Level
1
0.8
~o
CD
O
(D
O
03
0.6
0.4
0.2
0 0.002 0.004 0.006 0.008 0.01
dose
09:47 10/07 2008
Probit Model. (Version: 3.1; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmp21.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmp21.plt
Tue Oct 07 09:47:22 2008
BMDS Model Run
The form of the probability function is:
P[response] = CuniNorm(Intercept+Slope*Dose) ,
where CuniNorm(.) is the cumulative normal distribution function
Dependent variable = Response
Independent variable = DOSE
Slope parameter is not restricted
Total number of observations = 4
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
Default Initial (and Specified) Parameter Values
background = 0 Specified
intercept = -0.580731
slope = 266.006
B-48 DRAFT- DO NOT CITE OR QUOTE
Probit
BMDLBMD
-------
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
intercept
slope
intercept
1
-0.65
slope
-0.65
1
Variable
intercept
slope
Parameter Estimates
Estimate
-0.645576
306 .891
Std. Err.
0.132116
41. 902
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
-0.904519 -0.386633
224.765 389.018
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-103 . 017
-103.929
-152.026
211.858
# Param's
4
2
1
Deviance Test d.f.
1.82488
98 . 0188
P-value
0 .4015
<.0001
Dose
Goodness of Fit
Est. Prob.
Expected
Observed
Size
Scaled
Residual
0.0000
0 . 0021
0 . 0049
0.0101
Chia2 = 2.21
0.2593
0.4959
0.8021
0 . 9929
d.f. = 2
20.742 20.000 80 -0.189
23.803 23.000 48 -0.232
38.499 41.000 48 0.906
47.661 47.000 48 -1.139
P-value = 0.3315
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.000360199
BMDL = 0.000300191
B-49
DRAFT- DO NOT CITE OR QUOTE
-------
LogProbit Model with 0.95 Confidence Level
0.8
"O
(D
O
CD
= 1
Total number of observations = 4
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
-iii|ir
LogProbit
: BMDL BMD
J I I I |_L| l_l ¦ ¦
B-50
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial (and Specified) Parameter Values
background = 0.2 5
intercept = 8.97119
slope = 1.53079
Asymptotic Correlation Matrix of Parameter Estimates
background intercept slope
background 1 0.18 0.22
intercept 0.18 1 0.99
slope 0.22 0.99 1
Parameter Estimates
Variable
background
intercept
slope
Estimate
0 .249403
9 . 07823
1.54835
95.0% Wald Confidence Interval
Std. Err. Lower Conf. Limit Upper Conf. Limit
0.0481401 0.15505 0.343756
1.71894 5.70916 12.4473
0.310846 0.939107 2.1576
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-103.017
-103.031
-152.026
# Param'
4
3
1
Deviance Test d.f.
0 . 0292544
98 . 0188
P-value
0.8642
< . 0001
AIC:
212 . 063
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.2494 19.952 20.000 80 0.012
0.0021 0.4834 23.203 23.000 48 -0.059
0.0049 0.8482 40.715 41.000 48 0.115
0.0101 0.9814 47.106 47.000 48 -0.113
Chia2 = 0.03 d.f. = 1 P-value = 0.8634
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.000982443
BMDL = 0.000515017
B-51
DRAFT- DO NOT CITE OR QUOTE
-------
Table B-4. Summary of the BMD modeling results based on the incidence of
forestomach lesions (hyperplasia or hyperkeratosis) in male and female F344
rats exposed to AN in drinking water for 2 years
Sex
Endpoint
Dose metric
Selected model(s)a
x2
/j-valuc
AIC
BMDL10b
BMDL0Sb
Estimated
gamma (3)
0.32
193.27
1.51 x 10 1
9.97 x 10 2
administered
dose
(mg/kg-d)
two-stage multistage
(3)
0.14
194.76
1.40 x 10"1
8.02 x 10-2
Male
Forestomach
Predicted AN
gamma (3)
0.32
193.27
8.14x10 4
5.39 x 10 4
lesions
in blood
(mg/L)
two-stage multistage
(3)
0.14
194.77
7.56 x 10-4
4.34 x 10-4
Predicted CEO
gamma (3)
0.32
193.27
7.65 x 10 5
5.06 x 10 5
in blood
(mg/L)
two-stage multistage
(3)
0.14
194.76
7.10 x 10-5
4.07 x 10-5
Estimated
logistic (3)
0.32
141.06
2.31 x 10-1
1.54 x 10-1
administered
dose
two-stage multistage
(3)
0.39
140.80
2.09 x 10 1
1.17 x 10 1
(mg/kg-d)
probit (3)
0.26
141.38
2.18 x 10"1
1.40 x 10-1
Predicted AN
in blood
(mg/L)
logistic (3)
0.32
141.06
1.11 x 10-3
7.34 x 10-4
Female
Forestomach
lesions
two-stage multistage
(3)
0.39
140.80
9.97 x 10 4
5.58 x 10 4
probit (3)
0.26
141.38
1.04 x 10-3
6.69 x 10-4
Predicted CEO
in blood
(mg/L)
logistic (3)
0.32
141.07
1.02 x 10-4
6.78 x 10-5
two-stage multistage
(3)
0.39
140.81
9.23 x 10 5
5.17 x 10 5
probit (3)
0.26
141.40
9.63 x 10-5
6.19 x 10-5
aAll dichotomous models in EPA's BMDS (version 2.0) were fit to the incidence of forestomach lesions
(hyperplasia or hyperkeratosis) in F344 rats using the data presented in Table B-l. For BMD modeling, three
different dose metrics were employed: (1) administered animal dose (estimated) expressed in mg/kg-d, (2) AN in
blood (predicted) expressed in mg/L, and (3) CEO in blood (predicted) expressed in mg/L. Adequate fit of a model
was achieved if the %2 goodness-of-fit statistic yielded a />-value >0.1. The numbers in parentheses indicate the
number of dose groups dropped in order to obtain an adequate fit, starting with the highest dose group. Of those
models exhibiting adequate fit, the selected model was the model with the lowest AIC value, and is indicated in
bold in the table.
bBMDLio and BMDL05 estimates were derived from the selected model. If more than one model shared the lowest
AIC, the mean BMDLio and BMDL0s were calculated, as per the EPA's Benchmark Dose Technical Guidance
Document (U.S. EPA, 2000b).
Sources: Johannsen and Levinskas (2002b); Biodynamics (1980c).
B-52
DRAFT- DO NOT CITE OR QUOTE
-------
F344 Male Rats: Forestomach lesions (administered dose)
Gamma Multi-Hit Model with 0.95 Confidence Level
Gamma Multi-Hit
BMDL
0.05 0.1 0.15 0.2 0.25
dose
14:10 10/072008
Gamma Model. (Version: 2.13; Date: 05/16/2008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpl04.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpl04.plt
Tue Oct 07 14:10:45 2008
BMDS Model Run
The form of the probability function is:
P[response]= background+(1-background)*CumGamma[slope*dose,power],
where CumGamma(.) is the cummulative Gamma distribution function
Dependent variable = Response
Independent variable = DOSE
Power parameter is restricted as power >=1
Total number of observations = 3
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
Default Initial (and Specified) Parameter Values
Background = 0.071875
B-53
DRAFT- DO NOT CITE OR QUOTE
-------
Slope =
Power =
5.59733
2.91574
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Power
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background
Background 1
Slope -0.24
Slope
-0 .24
1
Variable
Background
Slope
Power
Parameter Estimates
Estimate
0 . 0585772
57 . 0862
18
Std. Err.
0 . 0151899
2 .86678
NA
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
0 . 0288055
51.4674
0 . 0883489
62 .705
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-94 .1158
-94.6364
-103.388
193.273
# Param's
3
2
1
Deviance Test d.f.
1. 04132
18.5446
P-value
0.3075
<.0001
Dose
Goodness of Fit
Est._Prob. Expected Observed Size
Scaled
Residual
0.0000
0.0800
0.2500
Chia2 =0.97
0.0586
0.0586
0.2400
d.f. = 1
9.314 11.000 159
4.686 3.000 80
18.000 18.000 75
P-value = 0.3250
0.569
-0.803
0 . 000
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.203802
BMDL = 0.0996986
B-54
DRAFT- DO NOT CITE OR QUOTE
-------
Multistage Model with 0.95 Confidence Level
0.35
0.3
0.25
"O
0
"G
£ 0.2
<
£=
O
-4—'
§ 0.15
LL
0.1
0.05
0
14:12 10/07 2008
Multistage Model. (Version: 3.0; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpl05.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpl05.plt
Tue Oct 07 14:12:49 2008
BMDS Model Run
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2)]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = DOSE
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Multistage
BMDL
BMD
0
0.05
0.1
0.15
0.2
0.25
dose
B-55
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.0438965
Beta(l) = 0
Beta(2) = 3.62384
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background
Beta(2)
Background
1
-0.49
Beta(2)
-0.49
1
Variable
Background
Beta(1)
Beta(2)
Parameter Estimates
Estimate
0 . 0563684
0
3 .20151
Std. Err.
* - Indicates that this value is not calculated.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
Analysis of Deviance Table
Model Log(likelihood) # Param's Deviance Test d.f.
Full model -94.1158 3
Fitted model -95.381 2 2.53038 1
Reduced model -103.388 1 18.5446 2
AIC: 194.762
Goodness of Fit
P-value
Dose
Est. Prob.
Expected
Observed
Size
0.1117
< . 0001
Scaled
Residual
0.0000
0.0800
0.2500
Chia2 = 2.21
0 . 0564
0.0755
0 .2275
d.f. = 1
8.963 11.000 159 0.701
6.041 3.000 80 -1.287
17.062 18.000 75 0.258
P-value = 0.1368
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0 . 05
Extra risk
0 . 95
0.126576
0 . 0802271
0.173506
Taken together, (0.0802271, 0.173506) is a 90
interval for the BMD
two-sided confidence
B-56
DRAFT- DO NOT CITE OR QUOTE
-------
F344 Male Rats: Forestomach lesions (AN in blood))
Gamma Multi-Hit Model with 0.95 Confidence Level
Gamma Multi-Hit
BMDL
0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012 0.0014
dose
14:46 10/07 2008
Gamma Model. (Version: 2.13; Date: 05/16/2008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpl07.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpl07.plt
Tue Oct 07 14:46:28 2008
BMDS Model Run
The form of the probability function is:
P[response]= background+(1-background)*CumGamma[slope*dose,power],
where CumGamma(.) is the cummulative Gamma distribution function
Dependent variable = Response
Independent variable = DOSE
Power parameter is restricted as power >=1
Total number of observations = 3
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
Default Initial (and Specified) Parameter Values
Background = 0.071875
Slope = 1039.5
B-57
DRAFT- DO NOT CITE OR QUOTE
-------
Power = 2 . 92132
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Power
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Slope
Background 1 -0.24
Slope -0.24 1
Parameter Estimates
Variable
Background
Slope
Power
Estimate
0 . 0585772
10571.5
18
Std. Err.
0 . 01519
530.886
NA
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
0 . 0288053
9531.01
0 . 088349
11612
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
AIC:
Log(likelihood)
-94.1158
-94 .6364
-103.388
193.273
# Param's
3
2
1
Deviance Test d.f.
P-value
1. 04133
18.5446
0.3075
<.0001
Dose
Goodness of Fit
Est._Prob. Expected Observed Size
Scaled
Residual
0.0000
0.0004
0.0014
Chia2 =0.97
0.0586
0.0586
0.2400
d.f. = 1
9.314 11.000 159
4.686 3.000 80
18.000 18.000 75
P-value = 0.3250
0.569
-0.803
0 . 000
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00110053
BMDL = 0.00053936
B-58
DRAFT- DO NOT CITE OR QUOTE
-------
Multistage Model with 0.95 Confidence Level
0.35
0.3
0.25
"O
0
"G
£ 0.2
<
£=
O
-4—'
§ 0.15
LL
0.1
0.05
0
0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012 0.0014
dose
14:54 10/07 2008
Multistage Model. (Version: 3.0; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmplOC.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmplOC.plt
Tue Oct 07 14:54:25 2008
BMDS Model Run
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2)]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = DOSE
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Multistage
BMDL
BMD
B-59
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.043854
Beta(l) = 0
Beta(2) = 124287
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Beta(2)
Background 1 -0.49
Beta(2) -0.49 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0.0563637 * * *
Beta(1) 0 * * *
Beta(2) 109747 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood) # Param's Deviance Test d.f. P-value
-94.1158 3
-95.3848 2 2.53803 1 0.1111
-103.388 1 18.5446 2 <.0001
AIC:
194.77
Goodness of Fit
Dose
Est. Prob.
Expected
Observed
Size
Scaled
Residual
0.0000 0.0564 8.962 11.000 159 0.701
0.0004 0.0756 6.047 3.000 80 -1.289
0.0014 0.2274 17.057 18.000 75 0.260
Chia2 = 2.22 d.f. = 1 P-value = 0.1363
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.000683652
BMDL = 0.000433678
BMDU = 0.000937168
Taken together, (0.000433678, 0.000937168) is a 90 % two-sided confidence
interval for the BMD
B-60
DRAFT- DO NOT CITE OR QUOTE
-------
F344 Male Rats: Forestomach lesions (CEO in blood)
Gamma Multi-Hit Model with 0.95 Confidence Level
Gamma Multi-Hit
BMDL
2e-005 4e-005 6e-005 8e-005 0.0001 0.00012
dose
16:32 10/07 2008
Gamma Model. (Version: 2.13; Date: 05/16/2008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpll5.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpll5.plt
Tue Oct 07 16:32:49 2008
BMDS Model Run
The form of the probability function is:
P[response]= background+(1-background)*CumGamma[slope*dose,power],
where CumGamma(.) is the cummulative Gamma distribution function
Dependent variable = Response
Independent variable = DOSE
Power parameter is restricted as power >=1
Total number of observations = 3
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
Default Initial (and Specified) Parameter Values
Background = 0.071875
B-61
DRAFT- DO NOT CITE OR QUOTE
-------
Slope = 11005
Power = 2 . 91337
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Power
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Slope
Background 1 -0.24
Slope -0.24 1
Parameter Estimates
Variable
Background
Slope
Power
Estimate
0 . 0585772
112374
18
Std. Err.
0 . 0151899
5643 .28
NA
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
0 . 0288055
101314
0 . 0883489
123435
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
AIC:
Log(likelihood)
-94.1158
-94 .6364
-103.388
193.273
# Param's
3
2
1
Deviance Test d.f.
P-value
1. 04132
18.5446
0.3075
<.0001
Dose
Goodness of Fit
Est._Prob. Expected Observed Size
Scaled
Residual
0.0000
0.0000
0.0001
Chia2 =0.97
0.0586
0.0586
0.2400
d.f. = 1
9.314 11.000 159
4.686 3.000 80
18.000 18.000 75
P-value = 0.3250
0.569
-0.803
0 . 000
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.000103532
BMDL = 5.06073e-005
B-62
DRAFT- DO NOT CITE OR QUOTE
-------
Multistage Model with 0.95 Confidence Level
0.35
0.3
0.25
"O
0
"G
£ 0.2
<
£=
O
-4—'
§ 0.15
LL
0.1
0.05
0
16:39 10/07 2008
Multistage Model. (Version: 3.0; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpllA.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpllA.plt
Tue Oct 07 16:39:04 2008
BMDS Model Run
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2)]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = DOSE
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Multistage
BMDL
BMD
0
2e-005
4e-005
6e-005
8e-005
0.0001
0.00012
dose
B-63
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.0439146
Beta(l) = 0
Beta(2) = 1.40418e+007
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Beta(2)
Background 1 -0.49
Beta(2) -0.49 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0.0563704 * * *
Beta(1) 0 * * *
Beta(2) 1.2408e+007 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood) # Param's Deviance Test d.f. P-value
-94.1158 3
-95.3793 2 2.52713 1 0.111£
-103.388 1 18.5446 2 <.0001
AIC:
194 .759
Dose
Est. Prob.
Goodness of Fit
Expected
Observed
Size
Scaled
Residual
0.0000
0.0000
0.0001
0 . 0564
0.0755
0 .2275
8 . 963
6 . 038
17 . 064
11.000
3 . 000
18 . 000
159
80
75
0.700
-1.286
0 .258
Chia2 = 2.21
d.f. = 1
P-value = 0.1371
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 6.42953e-005
BMDL = 4 . 07373e-005
BMDU = 8 . 81314e-005
Taken together, (4.07373e-005, 8.81314e-005) is a 90 % two-sided confidence interval for the
BMD
B-64
DRAFT- DO NOT CITE OR QUOTE
-------
F344 Female Rats: Forestomach lesions (administered dose)
Logistic Model with 0.95 Confidence Level
0.3
0.25
0.2
(D
O
CD
O
03
0.15
0.1
0.05
0.05 0.1 0.15 0.2 0.25 0.3 0.35
dose
13:49 10/08 2008
Logistic Model. (Version: 2.12; Date: 05/16/2008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpA62.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpA62.plt
Wed Oct 08 13:49:00 2008
BMDS Model Run
The form of the probability function is:
P[response] = 1/[1+EXP(-intercept-slope*dose)]
Dependent variable = Response
Independent variable = DOSE
Slope parameter is not restricted
Total number of observations = 3
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
Default Initial Parameter Values
background = 0 Specified
intercept = -3.79894
slope = 6 . 38252
Logistic
BMDL
B-65
DRAFT- DO NOT CITE OR QUOTE
-------
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
intercept slope
intercept 1 -0.88
slope -0.88 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
intercept -3.92594 0.477082 -4.86101 -2.99088
slope 6.94669 1.60094 3.80891 10.0845
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
AIC:
Log(likelihood)
-67 . 9873
-68.5288
-79.8392
141. 058
# Param's
3
2
1
Deviance Test d.f.
1. 08303
23 .7038
P-value
0 .298
<.0001
Goodness of Fit
Dose Est._Prob. Expected Observed Size
Scaled
Residual
0.0000
0.1200
0.3600
Chia2 =1.00
0.0193
0 . 0434
0.1939
d.f. = 1
3.017 4.000 156
3.474 2.000 80
15.509 16.000 80
P-value = 0.3175
0.571
-0.809
0.139
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0 . 05
Extra risk
0 . 95
0.189157
0.153543
B-66
DRAFT- DO NOT CITE OR QUOTE
-------
Multistage Model with 0.95 Confidence Level
~o
CD
o
(D
o
03
0.3
0.25
0.2
0.15
0.1
0.05
0.05 0.1 0.15 0.2 0.25 0.3 0.35
dose
13:58 10/08 2008
Multistage Model. (Version: 3.0; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpA65.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpA65.plt
Wed Oct 08 13:58:05 2008
BMDS Model Run
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2)]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = DOSE
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Multistage
BMDL
B-67
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.0147377
Beta(l) = 0
Beta(2) = 1.59649
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background
Beta(2)
Background
1
-0.51
Beta(2)
-0.51
1
Variable
Background
Beta(1)
Beta(2)
Parameter Estimates
Estimate
0 . 0216322
0
1.47012
Std. Err.
* - Indicates that this value is not calculated.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
Analysis of Deviance Table
Model Log(likelihood) # Param's Deviance Test d.f.
Full model -67.9873 3
Fitted model -68.401 2 0.827463 1
Reduced model -79.8392 1 23.7038 2
AIC: 140.802
Goodness of Fit
P-value
Dose
Est. Prob.
Expected
Observed
Size
0.363
<.0001
Scaled
Residual
0.0000
0.1200
0.3600
Chia2 = 0.74
0.0216
0 . 0421
0.1914
d.f. = 1
3.375 4.000 156
3.370 2.000 80
15.308 16.000 80
P-value = 0.3901
0.344
-0.763
0.197
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0 . 05
Extra risk
0 . 95
0.18679
0.116832
0 .243753
Taken together, (0.116832, 0.243753) is a 90
interval for the BMD
two-sided confidence
B-68
DRAFT- DO NOT CITE OR QUOTE
-------
Probit Model with 0.95 Confidence Level
0.3
0.25
0.2
(D
O
CD
O
03
0.15
0.1
0.05
0.05 0.1 0.15 0.2 0.25 0.3 0.35
dose
14:01 10/08 2008
Probit Model. (Version: 3.1; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpA66.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpA66.plt
Wed Oct 08 14:01:52 2008
BMDS Model Run
The form of the probability function is:
P[response] = CuniNorm(Intercept+Slope*Dose) ,
where CuniNorm(.) is the cumulative normal distribution function
Dependent variable = Response
Independent variable = DOSE
Slope parameter is not restricted
Total number of observations = 3
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
Default Initial (and Specified) Parameter Values
background = 0 Specified
intercept = -2.03888
slope = 3.08079
Probit
BMDL
B-69
DRAFT- DO NOT CITE OR QUOTE
-------
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
intercept slope
intercept 1 -0.81
slope -0.81 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
intercept -2.0746 0.1976 -2.46189 -1.68732
slope 3.33316 0.739981 1.88282 4.78349
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
AIC:
Log(likelihood)
-67 . 9873
-68.6922
-79.8392
141.384
# Param's
3
2
1
Deviance Test d.f.
P-value
1.40991
23 .7038
0 .2351
<.0001
Dose
Goodness of Fit
Est._Prob. Expected Observed Size
Scaled
Residual
0.0000
0.1200
0.3600
Chia2 = 1.2£
0.0190
0.0470
0.1909
d.f. = 1
2.966 4.000 156
3.760 2.000 80
15.270 16.000 80
P-value = 0.2588
0.606
-0 . 930
0 .208
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0 . 05
Extra risk
0 . 95
0.175274
0.140136
B-70
DRAFT- DO NOT CITE OR QUOTE
-------
F344 Female Rats: Forestomach lesions (AN in blood)
Logistic Model with 0.95 Confidence Level
0.3
0.25
0-2
(D
-i—1
O
CD
-------
slope =
1335 . 91
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
intercept slope
intercept 1 -0.88
slope -0.88 1
Parameter Estimates
Variable
intercept
slope
Estimate
-3 . 92577
1453 .89
Std. Err.
0 .477035
335 . 047
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
-4.86074 -2.9908
797.213 2110.57
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-67 . 9873
-68.5281
-79.8392
141. 056
# Param's
3
2
1
Deviance Test d.f.
1.0816
23 .7038
P-value
0 .2983
<.0001
Goodness of Fit
Dose Est._Prob. Expected Observed Size
Scaled
Residual
0.0000
0.0006
0.0017
Chia2 =1.00
0.0193
0 . 0434
0.1939
d.f. = 1
3.018 4.000 156
3.473 2.000 80
15.509 16.000 80
P-value = 0.3178
0.571
-0.808
0.139
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00090371
BMDL = 0.000733555
B-72
DRAFT- DO NOT CITE OR QUOTE
-------
Multistage Model with 0.95 Confidence Level
0.3
0.25
0.2
(D
O
CD
O
03
0.15
0.1
0.05
0.0002 0.0004 0.0006 0.0008 0.001 0.0012 0.0014 0.0016 0.0018
dose
16:07 10/08 2008
Multistage Model. (Version: 3.0; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpA72.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpA72.plt
Wed Oct 08 16:07:08 2008
BMDS Model Run
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2)]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = DOSE
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Multistage
BMDL
B-73
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.0147497
Beta(l) = 0
Beta(2) = 69935.2
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Beta(2)
Background 1 -0.51
Beta(2) -0.51 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0.021634 * * *
Beta(1) 0 * * *
Beta(2) 64407.8 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood) # Param's Deviance Test d.f. P-value
-67.9873 3
-68.4002 2 0.825863 1 0.3635
-79.8392 1 23.7038 2 <.0001
AIC:
140.8
Goodness of Fit
Dose
Est. Prob.
Expected
Observed
Size
Scaled
Residual
0.0000 0.0216 3.375 4.000 156 0.344
0.0006 0.0421 3.369 2.000 80 -0.762
0.0017 0.1914 15.310 16.000 80 0.196
Chia2 = 0.74 d.f. = 1 P-value = 0.3905
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.000892403
BMDL = 0.000558013
BMDU = 0.00116455
Taken together, (0.000558013, 0.00116455) is a 90 % two-sided confidence
interval for the BMD
B-74
DRAFT- DO NOT CITE OR QUOTE
-------
Probit Model with 0.95 Confidence Level
~o
CD
o
(D
o
03
0.3
0.25
0.2
0.15
0.1
0.05
0.0002 0.0004 0.0006 0.0008 0.001 0.0012 0.0014 0.0016 0.0018
dose
16:11 10/082008
Probit Model. (Version: 3.1; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpA73.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpA73.plt
Wed Oct 08 16:11:23 2008
BMDS Model Run
The form of the probability function is:
P[response] = CuniNorm(Intercept+Slope*Dose) ,
where CuniNorm(.) is the cumulative normal distribution function
Dependent variable = Response
Independent variable = DOSE
Slope parameter is not restricted
Total number of observations = 3
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
Default Initial (and Specified) Parameter Values
background = 0 Specified
intercept = -2.03885
slope = 644 . 864
B-75 DRAFT- DO NOT CITE OR QUOTE
Probit
BMDL
-------
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
intercept slope
intercept 1 -0.81
slope -0.81 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
intercept -2.07454 0.197583 -2.4618 -1.68729
slope 697.628 154.868 394.091 1001.16
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
AIC:
Log(likelihood)
-67 . 9873
-68.6913
-79.8392
141.383
# Param's
3
2
1
Deviance Test d.f.
P-value
1.40812
23 .7038
0 .2354
<.0001
Dose
Goodness of Fit
Est._Prob. Expected Observed Size
Scaled
Residual
0.0000
0.0006
0.0017
Chia2 = 1.27
0.0190
0.0470
0.1909
d.f. = 1
2.966 4.000 156
3.759 2.000 80
15.271 16.000 80
P-value = 0.2590
0.606
-0.929
0 .207
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.000837374
BMDL = 0.000669495
B-76
DRAFT- DO NOT CITE OR QUOTE
-------
F344 Female Rats: Forestomach lesions (CEO in blood)
Logistic Model with 0.95 Confidence Level
0.3
0.25
0-2
(D
-i—1
O
CD
-------
slope =
14448.5
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
intercept slope
intercept 1 -0.88
slope -0.88 1
Parameter Estimates
Variable
intercept
slope
Estimate
-3 . 92703
15733
Std. Err.
0 .4774
3627.17
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
-4.86272 -2.99135
8623.9 22842.1
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-67 . 9873
-68.5334
-79.8392
141. 067
# Param's
3
2
1
Deviance Test d.f.
1. 09233
23 .7038
P-value
0 .296
<.0001
Goodness of Fit
Dose Est._Prob. Expected Observed Size
Scaled
Residual
0.0000
0.0001
0.0002
Chia2 =1.01
0.0193
0 . 0435
0.1938
d.f. = 1
3.014 4.000 156
3.482 2.000 80
15.504 16.000 80
P-value = 0.3155
0.573
-0.812
0.140
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 8.35696e-005
BMDL = 6 .783 92e-005
B-78
DRAFT- DO NOT CITE OR QUOTE
-------
Multistage Model with 0.95 Confidence Level
~o
CD
o
(D
o
03
0.3
0.25
0.2
0.15
0.1
0.05
14:46 10/09 2008
2e-005 4e-005 6e-005 8e-005 0.0001 0.00012 0.00014 0.00016
dose
Multistage Model. (Version: 3.0; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpFC.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpFC.plt
Thu Oct 09 14:46:41 2008
BMDS Model Run
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2)]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = DOSE
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
BMDL
Multistage
B-79
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.0146597
Beta(l) = 0
Beta(2) = 8.18648e+006
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Beta(2)
Background 1 -0.51
Beta(2) -0.51 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0.0216202 * * *
Beta(1) 0 * * *
Beta(2) 7.53198e+006 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-67 . 9873
-68 .4062
-79.8392
# Param's
3
2
1
Deviance Test d.f.
0.837895
23 .7038
P-value
0.36
<.0001
AIC:
140.812
Dose
Est. Prob.
Goodness of Fit
Expected
Observed
Size
Scaled
Residual
0.0000
0.0001
0.0002
0.0216
0 . 0423
0.1913
3.373 4.000
3.380 2.000
15.301 16.000
156
80
80
0.345
-0.767
0.199
Chia2 = 0.75
d.f. = 1
P-value = 0.3873
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 8.25232e-005
BMDL = 5 .17106e-005
BMDU = 0.000107687
Taken together, (5.17106e-005, 0.000107687) is a 90 % two-sided confidence interval for the
BMD
B-80
DRAFT- DO NOT CITE OR QUOTE
-------
Probit Model with 0.95 Confidence Level
0.3
0.25
0.2
(D
O
CD
O
03
0.15
0.1
0.05
2e-005 4e-005 6e-005 8e-005 0.0001 0.00012 0.00014 0.00016
dose
14:50 10/09 2008
Probit Model. (Version: 3.1; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpFD.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpFD.plt
Thu Oct 09 14:50:46 2008
BMDS Model Run
The form of the probability function is:
P[response] = CuniNorm(Intercept+Slope*Dose) ,
where CuniNorm(.) is the cumulative normal distribution function
Dependent variable = Response
Independent variable = DOSE
Slope parameter is not restricted
Total number of observations = 3
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
Default Initial (and Specified) Parameter Values
background = 0 Specified
intercept = -2.03906
slope = 6972.11
Probit
BMDL
B-81
DRAFT- DO NOT CITE OR QUOTE
-------
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
intercept slope
intercept 1 -0.81
slope -0.81 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
intercept -2.075 0.197714 -2.46251 -1.68749
slope 7547.46 1676.24 4262.09 10832.8
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
AIC:
Log(likelihood)
-67 . 9873
-68.6981
-79.8392
141.396
# Param's
3
2
1
Deviance Test d.f.
P-value
1.42158
23 .7038
0.2331
<.0001
Dose
Goodness of Fit
Est._Prob. Expected Observed Size
Scaled
Residual
0.0000
0.0001
0.0002
Chia2 = 1.2S
0.0190
0 . 0471
0.1908
d.f. = 1
2.963 4.000 156 0.608
3.769 2.000 80 -0.934
15.264 16.000 80 0.209
P-value = 0.2569
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 7.74403e-005
BMDL = 6 .19187e-005
B-82
DRAFT- DO NOT CITE OR QUOTE
-------
Table B-5. Summary of the BMD modeling results based on the incidence of
forestomach lesions (hyperplasia or hyperkeratosis) in male and female
B6C3Fi mice exposed to AN via gavage for 2 years
Sex
Endpoint
Dose metric
Selected model(s)a
x2
/j-valuc
AIC
BMDL10b
BMDL0Sb
gamma
0.80
156.45
3.45
1.68
logistic
0.37
158.01
6.37
3.72
log-logistic
0.87
156.30
3.01
1.43
Forestomach
lesions
Estimated
probit
0.42
157.74
5.96
3.42
Male
administered dose
(mg/kg-d)
log-probit
0.33
158.23
5.53
3.85
one-stage
multistage
0.80
156.45
3.45
1.68
Weibull
0.80
156.45
3.45
1.68
quantal linear
0.80
156.45
3.45
1.68
gamma
0.74
116.17
6.24
3.04
logistic
0.87
114.33
8.95
5.60
log-logistic
0.75
116.17
5.98
2.83
Forestomach
lesions
Estimated
probit
0.90
114.28
8.51
5.19
Female
administered dose
(mg/kg-d)
log-probit
0.81
114.48
8.12
5.65
one-stage
multistage
0.92
114.24
6.20
3.02
Weibull
0.74
116.18
6.24
3.04
quantal linear
0.92
114.24
6.20
3.02
aAll dichotomous models in EPA's BMDS (version 2.0) were fit to the incidence of forestomach lesions
(hyperplasia or hyperkeratosis) in B6C3FJ mice using the data presented in Table B-2. For BMD modeling, animal
dose (estimated), expressed in mg/kg-d, was employed. Adequate fit of a model was achieved if the %2 goodness-
of-fit statistic yielded a />-value >0.1. Of those models exhibiting adequate fit, the selected model was the model
with the lowest AIC value, and is indicated in bold in the table.
bBMDL10 and BMDL05 estimates were derived from the selected model. If more than one model shared the lowest
AIC, the mean BMDL10 and BMDL05 were calculated, as per the EPA's Benchmark Dose Technical Guidance
Document (U.S. EPA, 2000b).
Source: NTP(2001).
B-83
DRAFT- DO NOT CITE OR QUOTE
-------
B6C3F! Male Mice: Forestomach lesions (administered dose)
Gamma Multi-Hit Model with 0.95 Confidence Level
~o
CD
-*—>
o
CD
=1
Total number of observations = 4
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
Default Initial (and Specified) Parameter Values
Background = 0.0490196
B-84
DRAFT- DO NOT CITE OR QUOTE
-------
Slope =
Power =
0 . 0449456
1.3
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Power
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background
Slope
Background
1
-0.53
Slope
-0.53
1
Variable
Background
Slope
Power
Parameter Estimates
Estimate
0 . 0444165
0 . 0201613
1
Std. Err.
0 . 025521
0 . 00593774
NA
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
-0 . 00560381
0 . 00852351
0 . 0944368
0 . 031799
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
Log(likelihood)
-76 . 0086
-76.2255
-82 .7874
# Param's
4
2
1
Deviance Test d.f.
0 .433714
13.5576
P-value
0.805
0 . 003574
AIC:
156.451
Dose
Est. Prob.
Goodness of Fit
Expected
Observed
Size
Scaled
Residual
0.0000
1.8000
7.1000
14.3000
0 . 0444
0.0785
0.1719
0 .2838
2 .221
3 . 924
8.593
14 .188
2 . 000
4 . 000
10 . 000
13 . 000
50
50
50
50
-0.152
0 . 040
0.527
-0.373
Chi 2 = 0.44
d.f. = 2
P-value = 0.8019
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0 . 05
Extra risk
0 . 95
2 .54415
1.68095
B-85
DRAFT- DO NOT CITE OR QUOTE
-------
Logistic Model with 0.95 Confidence Level
~o
CD
o
(D
o
03
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
Logistic
BMDL
10
12
14
dose
14:00 11/14 2008
Logistic Model. (Version: 2.12; Date: 05/16/2008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpll3.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpll3.plt
Fri Nov 14 14:00:33 2008
BMDS Model Run
The form of the probability function is:
P[response] = 1/[1+EXP(-intercept-slope*dose)]
Dependent variable = Response
Independent variable = DOSE
Slope parameter is not restricted
Total number of observations = 4
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
Default Initial Parameter Values
background = 0 Specified
intercept = -2.67373
slope = 0.130289
Asymptotic Correlation Matrix of Parameter Estimates
B-86
DRAFT- DO NOT CITE OR QUOTE
-------
( *** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
intercept slope
intercept 1 -0.83
slope -0.83 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
intercept -2.63925 0.372305 -3.36896 -1.90955
slope 0.121074 0.0365284 0.04948 0.192669
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-76 . 0086
-77.0042
-82 .7874
# Param'
4
2
1
Deviance Test d.f.
1. 99121
13.5576
P-value
0.3695
0 . 003574
AIC:
158.008
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0667 3.333 2.000 50 -0.756
1.8000 0.0816 4.078 4.000 50 -0.040
7.1000 0.1443 7.217 10.000 50 1.120
14.3000 0.2874 14.372 13.000 50 -0.429
Chia2 = 2.01 d.f. = 2 P-value = 0.3660
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 4 .80697
BMDL = 3 .72464
B-87
DRAFT- DO NOT CITE OR QUOTE
-------
Log-Logistic Model with 0.95 Confidence Level
~o
CD
o
(D
o
03
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
Log-Logistic
BMDL
10
12
14
dose
14:04 11/14 2008
Logistic Model. (Version: 2.12; Date: 05/16/2008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpll6.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpll6.plt
Fri Nov 14 14:04:51 2008
BMDS Model Run
The form of the probability function is:
P[response] = background+(1-background)/[1+EXP(-intercept-slope*Log(dose))]
Dependent variable = Response
Independent variable = DOSE
Slope parameter is restricted as slope >= 1
Total number of observations = 4
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 Parameter Values
background = 0.04
intercept = -3.73286
slope = 1
B-88
DRAFT- DO NOT CITE OR QUOTE
-------
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -slope
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
background intercept
background 1 -0.54
intercept -0.54 1
Parameter Estimates
Variable
background
intercept
slope
Estimate
0 . 0421528
-3 .76687
1
Std. Err.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-76 . 0086
-76.1489
-82 .7874
156.298
# Param'
4
2
1
Deviance Test d.f.
0 .280564
13.5576
P-value
0.8691
0 . 003574
Dose
Goodness of Fit
Est._Prob. Expected Observed
Size
Scaled
Residual
0.0000
1.8000
7.1000
14.3000
Chia2 = 0.2£
0 . 0422
0.0804
0.1772
0.2802
d.f. = 2
2.108 2.000 50 -0.076
4.021 4.000 50 -0.011
8.862 10.000 50 0.422
14.009 13.000 50 -0.318
P-value = 0.8674
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0 . 05
Extra risk
0 . 95
2 .27602
1.42584
B-89
DRAFT- DO NOT CITE OR QUOTE
-------
Probit Model with 0.95 Confidence Level
~o
CD
o
(D
o
03
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
Probit
BMDL
10
12
14
dose
14:27 11/14 2008
Probit Model. (Version: 3.1; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpl24.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpl24.plt
Fri Nov 14 14:27:57 2008
BMDS Model Run
The form of the probability function is:
P[response] = CuniNorm(Intercept+Slope*Dose) ,
where CuniNorm(.) is the cumulative normal distribution function
Dependent variable = Response
Independent variable = DOSE
Slope parameter is not restricted
Total number of observations = 4
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
Default Initial (and Specified) Parameter Values
background = 0 Specified
intercept = -1.57644
slope = 0.0736548
B-90
DRAFT- DO NOT CITE OR QUOTE
-------
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
intercept
slope
intercept
1
-0.8
slope
-0.8
1
Variable
intercept
slope
Parameter Estimates
Estimate
-1.52854
0.0679152
95.0% Wald Confidence Interval
Std. Err. Lower Conf. Limit Upper Conf. Limit
0.189536 -1.90002 -1.15706
0.0200887 0.028542 0.107288
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
Log(likelihood)
-76 . 0086
-76.8681
-82 .7874
# Param's
4
2
1
Deviance Test d.f.
1.71884
13.5576
P-value
0 .4234
0 . 003574
AIC:
157.736
Dose
Goodness of Fit
Est. Prob.
Expected
Observed
Size
Scaled
Residual
0.0000
1.8000
7.1000
14.3000
Chia2 =1.74
0.0632
0.0798
0.1477
0.2886
d.f. = 2
3.159 2.000 50 -0.674
3.991 4.000 50 0.005
7.385 10.000 50 1.042
14.432 13.000 50 -0.447
P-value = 0.4189
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0 . 05
Extra risk
0 . 95
4 .44923
3 .4216
B-91
DRAFT- DO NOT CITE OR QUOTE
-------
LogProbit Model with 0.95 Confidence Level
~o
CD
o
(D
o
03
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
LogProbit
BMDL
10
12
14
dose
14:16 11/14 2008
Probit Model. (Version: 3.1; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpllE.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpllE.plt
Fri Nov 14 14:16:36 2008
BMDS Model Run
The form of the probability function is:
P[response] = Background
+ (1-Background) * CuniNorm (Intercept+Slope*Log (Dose) ) ,
where CuniNorm(.) is the cumulative normal distribution function
Dependent variable = Response
Independent variable = DOSE
Slope parameter is restricted as slope >= 1
Total number of observations = 4
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
B-92
DRAFT- DO NOT CITE OR QUOTE
-------
background = 0.04
intercept = -3.00109
slope = 1
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -slope
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
background
intercept
background
1
-0.53
intercept
-0.53
1
Parameter Estimates
Variable
background
intercept
slope
Estimate
0 . 0650367
-3 .3131
1
Std. Err.
0 . 0264604
0 .219612
NA
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
0 . 0131753
-3 .74353
0.116898
-2 .88267
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
#
Log(likelihood)
-76 . 0086
-77.1132
-82 .7874
Param's
4
2
1
Deviance Test d.f.
2.20909
13.5576
P-value
0.3314
0 . 003574
AIC:
158 .226
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0650 3.252 2.000 50 -0.718
1.8000 0.0680 3.402 4.000 50 0.336
7.1000 0.1473 7.367 10.000 50 1.051
14.3000 0.3053 15.263 13.000 50 -0.695
Chia2 = 2.21 d.f. = 2 P-value = 0.3304
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 5.30287
BMDL = 3 .84888
B-93
DRAFT- DO NOT CITE OR QUOTE
-------
Multistage Model with 0.95 Confidence Level
~o
CD
o
(D
o
03
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
Multistage
BMDL
10
12
14
dose
14:22 11/14 2008
Multistage Model. (Version: 3.0; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpl21.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpl21.plt
Fri Nov 14 14:22:26 2008
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
B-94
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.0536508
Beta(1) = 0.018443
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.71
Beta(1) -0.71 1
Parameter Estimates
Variable
Background
Beta(1)
Estimate
0 . 0444165
0 . 0201613
Std. Err.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-76 . 0086
-76.2255
-82 .7874
# Param's
4
2
1
Deviance Test d.f.
0 .433714
13.5576
P-value
0.805
0 . 003574
AIC:
156.451
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000
0 . 0444
2 .221
2 . 000
50
-0.152
1.8000
0 . 0785
3 . 924
4 . 000
50
0 . 040
7.1000
0.1719
8.593
10 . 000
50
0.527
14.3000
0 .2838
14 .188
13 . 000
50
-0.373
Chia2 = 0.44 d.f. = 2 P-value = 0.8019
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 2 .54415
BMDL = 1.68095
BMDU = 4 .75191
Taken together, (1.68095, 4.75191) is a 90 % two-sided confidence
interval for the BMD
B-95 DRAFT- DO NOT CITE OR QUOTE
-------
Weibull Model with 0.95 Confidence Level
~o
CD
o
(D
o
03
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
Weibu
BMDL
10
12
14
dose
14:36 11/14 2008
Weibull Model using Weibull Model (Version: 2.12; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpl27.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpl27.plt
Fri Nov 14 14:36:13 2008
BMDS Model Run
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(-slope*doseApower)]
Dependent variable = Response
Independent variable = DOSE
Power parameter is restricted as power >=1
Total number of observations = 4
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
Default Initial (and Specified) Parameter Values
Background = 0.0490196
Slope = 0.0179876
Power = 1
Asymptotic Correlation Matrix of Parameter Estimates
B-96
DRAFT- DO NOT CITE OR QUOTE
-------
( *** The model parameter(s) -Power
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Slope
Background 1 -0.53
Slope -0.53 1
Parameter Estimates
Variable
Background
Slope
Power
Estimate
0 . 0444162
0 . 0201613
1
Std. Err.
0 . 0255203
0 . 00593763
NA
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
-0 . 00560266
0 . 00852374
0 . 0944351
0 . 0317988
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
# Param's Deviance Test d.f.
Log(likelihood) # Param'
-76.0086 4
-76.2255 2
-82.7874 1
0 .433714
13.5576
P-value
0.805
0 . 003574
AIC:
156.451
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0444 2.221 2.000 50 -0.152
1.8000 0.0785 3.924 4.000 50 0.040
7.1000 0.1719 8.593 10.000 50 0.527
14.3000 0.2838 14.188 13.000 50 -0.373
Chia2 = 0.44 d.f. = 2 P-value = 0.8019
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 2 .54415
BMDL = 1.68095
B-97
DRAFT- DO NOT CITE OR QUOTE
-------
Quantal Linear Model with 0.95 Confidence Level
~o
CD
o
(D
o
03
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
Quantal Linear
BMDL
10
12
14
dose
14:42 11/14 2008
Quantal Linear Model using Weibull Model (Version: 2.12; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpl2A.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpl2A.plt
Fri Nov 14 14:42:54 2008
BMDS Model Run
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(-slope*dose)]
Dependent variable = Response
Independent variable = DOSE
Total number of observations = 4
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
Default Initial (and Specified) Parameter Values
Background = 0.0490196
Slope = 0.0179876
Power = 1 Specified
Asymptotic Correlation Matrix of Parameter Estimates
B-98
DRAFT- DO NOT CITE OR QUOTE
-------
( *** The model parameter(s) -Power
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Slope
Background 1 -0.53
Slope -0.53 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0.0444162 0.0255203 -0.00560266 0.0944351
Slope 0.0201613 0.00593763 0.00852374 0.0317988
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
#
Log(likelihood) # Param'
-76.0086 4
-76.2255 2
-82.7874 1
Deviance Test d.f.
0 .433714
13.5576
P-value
0.805
0 . 003574
AIC:
156.451
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0444 2.221 2.000 50 -0.152
1.8000 0.0785 3.924 4.000 50 0.040
7.1000 0.1719 8.593 10.000 50 0.527
14.3000 0.2838 14.188 13.000 50 -0.373
Chia2 = 0.44 d.f. = 2 P-value = 0.8019
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 2 .54415
BMDL = 1.68095
B-99
DRAFT- DO NOT CITE OR QUOTE
-------
B6C3F! Female Mice: Forestomach lesions (administered dose)
Gamma Multi-Hit Model with 0.95 Confidence Level
~o
CD
o
(D
o
03
0.3
0.25
0.2
0.15 - -r-
0.1
0.05
Gamma Mu ti-Hit
BMDL
10
12
14
dose
15:44 11/14 2008
Gamma Model. (Version: 2.13; Date: 05/16/2008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpl31.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpl31.plt
Fri Nov 14 15:44:47 2008
BMDS Model Run
The form of the probability function is:
P[response]= background+(1-background)*CumGamma[slope*dose,power]
where CumGamma(.) is the cummulative Gamma distribution function
Dependent variable = Response
Independent variable = DOSE
Power parameter is restricted as power >=1
Total number of observations = 4
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
Default Initial (and Specified) Parameter Values
Background = 0.0490196
Slope = 0.0133015
B-100
DRAFT- DO NOT CITE OR QUOTE
-------
Power =
1.18093
Asymptotic Correlation Matrix of Parameter Estimates
Background
Background 1
Slope 0.39
Power 0.47
Slope
0.39
1
0 . 99
Power
0.47
0 . 99
1
Parameter Estimates
Variable
Background
Slope
Power
Estimate
0 . 0367793
0 . 0180646
1.28374
Std. Err.
0 . 0231276
0 . 0405793
1.22475
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
-0.00854987 0.0821085
-0 . 0614694 0.0975986
-1.11674 3.68421
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-55 . 0321
-55.0866
-58.1629
116.173
# Param's
4
3
1
Deviance Test d.f.
0.109064
6 .26165
P-value
0.7412
0.09955
Dose
Goodness of Fit
Est. Prob.
Expected
Observed
Size
Scaled
Residual
0.0000
1.8000
7.1000
14.3000
Chia2 = 0.11
0 . 0368
0 . 0468
0 . 0924
0.1639
d.f. = 1
1.839 2.000 50 0.121
2.342 2.000 50 -0.229
4.618 5.000 50 0.187
8.197 8.000 50 -0.075
P-value = 0.7429
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 6.31104
BMDL = 3.03872
B-101
DRAFT- DO NOT CITE OR QUOTE
-------
Logistic Model with 0.95 Confidence Level
Logistic
0.3
0.25
0.2
0.15
1
0.05
0
BMDL
BMD
0
2
4
6
8
10
12
14
dose
15:48 11/14 2008
Logistic Model. (Version: 2.12; Date: 05/16/2008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpl34.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpl34.plt
Fri Nov 14 15:48:45 2008
BMDS Model Run
The form of the probability function is:
P[response] = 1/[1+EXP(-intercept-slope*dose)]
Dependent variable = Response
Independent variable = DOSE
Slope parameter is not restricted
Total number of observations = 4
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
Default Initial Parameter Values
background = 0 Specified
intercept = -3.00612
slope = 0.102222
Asymptotic Correlation Matrix of Parameter Estimates
B-102
DRAFT- DO NOT CITE OR QUOTE
-------
( *** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
intercept slope
intercept 1 -0.84
slope -0.84 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
intercept -3.16594 0.471287 -4.08965 -2.24224
slope 0.108988 0.0453905 0.0200238 0.197951
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-55 . 0321
-55.1666
-58.1629
# Param'
4
2
1
Deviance Test d.f.
0.269
6 .26165
P-value
0.8742
0.09955
AIC:
114 .333
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0405 2.023 2.000 50 -0.017
1.8000 0.0488 2.441 2.000 50 -0.289
7.1000 0.0838 4.189 5.000 50 0.414
14.3000 0.1669 8.347 8.000 50 -0.132
Chia2 = 0.27 d.f. = 2 P-value = 0.8725
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 7.64457
BMDL = 5.59724
B-103
DRAFT- DO NOT CITE OR QUOTE
-------
Log-Logistic Model with 0.95 Confidence Level
~o
CD
o
(D
o
03
0.3
0.25
0.2
0.15 - -r-
0.1
0.05
Log-Logistic
BMDL
15:51 11/14 2008
10
12
14
dose
Logistic Model. (Version: 2.12; Date: 05/16/2008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpl37.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpl37.plt
Fri Nov 14 15:51:59 2008
BMDS Model Run
The form of the probability function is:
P[response] = background+(1-background)/[1+EXP(-intercept-slope*Log(dose))]
Dependent variable = Response
Independent variable = DOSE
Slope parameter is restricted as slope >= 1
Total number of observations = 4
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 Parameter Values
background = 0.04
intercept = -5.35004
slope = 1.30201
B-104
DRAFT- DO NOT CITE OR QUOTE
-------
Asymptotic Correlation Matrix of Parameter Estimates
background intercept
background 1 -0.53
intercept -0.53 1
slope 0.46 -0.98
slope
0.46
-0 . 98
1
Variable
background
intercept
slope
Parameter Estimates
Estimate
0 . 0366811
-5 .29891
1.2835
Std. Err.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
#
Log(likelihood)
-55 . 0321
-55 . 0846
-58.1629
Param's
4
3
1
Deviance Test d.f.
0.104995
6 .26165
P-value
0.7459
0.09955
AIC:
116 .169
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0367 1.834 2.000 50 0.125
1.8000 0.0468 2.340 2.000 50 -0.228
7.1000 0.0928 4.639 5.000 50 0.176
14.3000 0.1637 8.186 8.000 50 -0.071
Chia2 = 0.10 d.f. = 1 P-value = 0.7477
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 6 .26147
BMDL = 2 .83097
B-105
DRAFT- DO NOT CITE OR QUOTE
-------
Probit Model with 0.95 Confidence Level
Probit
0.3
0.25
0.2
0.15
1
0.05
0
BMDL
BMD
0
2
4
6
8
10
12
14
dose
16:05 11/14 2008
Probit Model. (Version: 3.1; Date: 05/16/2008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpl40.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpl40.plt
Fri Nov 14 16:05:36 2008
BMDS Model Run
The form of the probability function is:
P[response] = CuniNorm(Intercept+Slope*Dose) ,
where CuniNorm(.) is the cumulative normal distribution function
Dependent variable = Response
Independent variable = DOSE
Slope parameter is not restricted
Total number of observations = 4
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
Default Initial (and Specified) Parameter Values
background = 0 Specified
B-106
DRAFT- DO NOT CITE OR QUOTE
-------
intercept = -1.75537
slope = 0.0561083
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
intercept slope
intercept 1 -0.8
slope -0.8 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
intercept -1.7601 0.219358 -2.19003 -1.33017
slope 0.0553914 0.0228374 0.0106309 0.100152
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-55 . 0321
-55.1419
-58.1629
# Param'
4
2
1
Deviance Test d.f.
0 .219734
6 .26165
P-value
0.896
0.09955
AIC:
114.284
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0392 1.960 2.000 50 0.029
1.8000 0.0484 2.421 2.000 50 -0.277
7.1000 0.0858 4.292 5.000 50 0.357
14.3000 0.1665 8.326 8.000 50 -0.124
Chia2 = 0.22 d.f. = 2 P-value = 0.8955
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 7.2597
BMDL = 5.19213
B-107
DRAFT- DO NOT CITE OR QUOTE
-------
LogProbit Model with 0.95 Confidence Level
LogProbit
0.3
0.25
0.2
0.15
1
0.05
0
BMDL
BMD
0
2
4
6
8
10
12
14
dose
15:56 11/14 2008
Probit Model. (Version: 3.1; Date: 05/16/2008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpl3A.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpl3A.plt
Fri Nov 14 15:56:14 2008
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 = Response
Independent variable = DOSE
Slope parameter is restricted as slope >= 1
Total number of observations = 4
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
B-108
DRAFT- DO NOT CITE OR QUOTE
-------
background = 0.04
intercept = -3.50754
slope = 1
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -slope
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
background
intercept
background
1
-0.51
intercept
-0.51
1
Parameter Estimates
Variable
background
intercept
slope
Estimate
0 . 0436672
-3 .75124
1
Std. Err.
0 . 0205583
0 .270066
NA
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
0 . 00337363 0.0839607
-4.28056 -3.22192
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-55 . 0321
-55 .2386
-58.1629
# Param's
4
2
1
Deviance Test d.f.
0 .413147
6 .26165
P-value
0.8134
0.09955
AIC:
114 .477
Dose
Est. Prob.
Goodness of Fit
Expected
Observed
Size
Scaled
Residual
0.0000
1.8000
7.1000
14.3000
0 . 0437
0 . 0444
0.0787
0.1753
2 .183
2 .221
3 . 935
8.765
2 . 000
2 . 000
5 . 000
8 . 000
50
50
50
50
-0.127
-0.151
0.559
-0 .284
Chi 2 = 0.43
d.f. = 2
P-value = 0.8054
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 8.21852
BMDL = 5.64968
B-109
DRAFT- DO NOT CITE OR QUOTE
-------
Multistage Model with 0.95 Confidence Level
~o
CD
o
(D
o
03
0.3
0.25
0.2
0.15
0.1
0.05
8 10 12 14
dose
16:01 11/14 2008
Multistage Model. (Version: 3.0; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpl3D.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpl3D.plt
Fri Nov 14 16:01:58 2008
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
BMDL
Multistage
B-110
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.032589
Beta(1) = 0.00986338
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.7
Beta(1) -0.7 1
Parameter Estimates
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Variable
Background
Beta(1)
Estimate
0 . 0341442
0 . 00957473
Std. Err.
•k
•k
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-55 . 0321
-55.1226
-58.1629
# Param's
4
2
1
Deviance Test d.f.
0.181118
6 .26165
P-value
0 . 9134
0.09955
AIC:
114 .245
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000
0 . 0341
1.707
2 . 000
50
0 .228
1.8000
0 . 0506
2 .532
2 . 000
50
-0.343
7.1000
0.0976
4 .881
5 . 000
50
0 . 057
14.3000
0.1577
7.887
8 . 000
50
0 . 044
Chia2 = 0.18 d.f. = 2 P-value = 0.9162
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 5.35715
BMDL = 3 . 01931
BMDU = 16 .4514
Taken together, (3.01931, 16.4514) is a 90 % two-sided confidence
interval for the BMD
B-lll
DRAFT- DO NOT CITE OR QUOTE
-------
Weibull Model with 0.95 Confidence Level
Weibull
0.3
0.25
0.2
0.15
1
0.05
0
BMDL
BMD
0
2
4
6
8
10
12
14
dose
16:09 11/14 2008
Weibull Model using Weibull Model (Version: 2.12; Date: 05/16/2008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpl43.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpl43.plt
Fri Nov 14 16:09:11 2008
BMDS Model Run
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(-slope*dose^power)]
Dependent variable = Response
Independent variable = DOSE
Power parameter is restricted as power >=1
Total number of observations = 4
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
Default Initial (and Specified) Parameter Values
Background = 0.0490196
Slope = 0.00467913
Power = 1.25557
Asymptotic Correlation Matrix of Parameter Estimates
B-112
DRAFT- DO NOT CITE OR QUOTE
-------
Background
Background 1
Slope -0.54
Power 0.47
Slope
-0.54
1
-0.99
Power
0.47
-0.99
1
Parameter Estimates
Variable
Background
Slope
Power
Estimate
0 . 0366324
0 . 00530377
1.23448
Std. Err.
0 . 0230239
0 . 0138578
1.00512
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
-0 . 00849362 0.0817585
-0.021857 0.0324646
-0.735513 3.20447
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-55 . 0321
-55 . 0891
-58.1629
# Param'
4
3
1
Deviance Test d.f.
0.114024
6 .26165
P-value
0.7356
0.09955
AIC:
116 .178
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0366 1.832 2.000 50 0.127
1.8000 0.0471 2.357 2.000 50 -0.238
7.1000 0.0924 4.620 5.000 50 0.186
14.3000 0.1638 8.188 8.000 50 -0.072
Chia2 = 0.11 d.f. = 1 P-value = 0.7375
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 6 .28482
BMDL = 3 . 03737
B-113
DRAFT- DO NOT CITE OR QUOTE
-------
Quantal Linear Model with 0.95 Confidence Level
~o
CD
o
(D
o
03
0.3
0.25
0.2
0.15
0.1
0.05
Quantal Linear
BMDL
10
12
14
dose
16:13 11/142008
Quantal Linear Model using Weibull Model (Version: 2.12; Date: 05/16/2 008)
Input Data File: C:\USEPA\BMDS2\Temp\tmpl46.(d)
Gnuplot Plotting File: C:\USEPA\BMDS2\Temp\tmpl46.plt
Fri Nov 14 16:13:01 2008
BMDS Model Run
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(-slope*dose)]
Dependent variable = Response
Independent variable = DOSE
Total number of observations = 4
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
Default Initial (and Specified) Parameter Values
Background = 0.0490196
Slope = 0.00923495
Power = 1 Specified
Asymptotic Correlation Matrix of Parameter Estimates
B-114
DRAFT- DO NOT CITE OR QUOTE
-------
( *** The model parameter(s) -Power
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Slope
Background 1 -0.53
Slope -0.53 1
Parameter Estimates
Variable
Background
Slope
Estimate
0 . 0341442
0 . 00957473
95.0% Wald Confidence Interval
Std. Err. Lower Conf. Limit Upper Conf. Limit
0.0203958 -0.00583081 0.0741193
0.00414213 0.0014563 0.0176932
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
#
Log(likelihood) # Param'
-55.0321 4
-55.1226 2
-58.1629 1
Deviance Test d.f.
0.181118
6 .26165
P-value
0 . 9134
0.09955
AIC:
114 .245
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0341 1.707 2.000 50 0.228
1.8000 0.0506 2.532 2.000 50 -0.343
7.1000 0.0976 4.881 5.000 50 0.057
14.3000 0.1577 7.887 8.000 50 0.044
Chia2 = 0.18 d.f. = 2 P-value = 0.9162
Benchmark Dose Computation
Specified effect = 0.05
Risk Type = Extra risk
Confidence level = 0.95
BMD = 5.35715
BMDL = 3 . 01931
B-115
DRAFT- DO NOT CITE OR QUOTE
-------
APPENDIX B-2. NONCANCER INHALATION DOSE-RESPONSE ASSESSMENT
(RfC): BMD MODELING RESULTS EMPLOYING THE INCIDENCE DATA FOR
NONNEOPLASTIC NASAL LESIONS IN RATS EXPOSED TO AN BY INHALATION
FOR 2 YEARS (TABLES B-6 THROUGH B-9)
For these modeling exercises, a BMR of 10% extra risk was selected due to the limited
number of rats in each exposure group. BMC and BMCL refer to the model-predicted
concentration and its lower 95% confidence limit, respectively, associated with a 10% extra risk
for developing the lesion.
Table B-6. Incidence data for selected nasal lesions in Sprague-Dawley rats
exposed by inhalation to AN for 2 years
Nasal lesion
Exposure level (ppm)
Exposure level
(mg/m3)
HEC
(mg/m3)a
Incidence
Hyperplasia of mucus-
secreting cells in males
0
0
0
0/11 (0%)
20
43.4
2.1
7/12 (58%)b
80
173.6
8.5
8/10 (80%)b
Flattening of respiratory
epithelium in females
0
0
0
1/11 (9%)
20
43.4
2.1
7/10 (70%)b
80
173.6
8.5
8/10 (80%)b
aHEC as per U.S. EPA (1994b) methods for a category 1 gas producing an upper respiratory effect.
Sample calculation: 43.4 mg/m3 x 6h/24h x 5d/7d x RGDREt = 2.1 mg/m3, where RGDREt = 0.275 =
[VE/SAet] rat ^ [VE/SAet] human; VE = minute volume = 0.281 L/min rat; 13.8 L/min human; and SAET
= extrathoracic surface area = 5 cm2 rat, 200 cm2 human.
Statistically significantly different from control value as reported by the authors.
Source: Quast et al. (1980b).
B-116
DRAFT- DO NOT CITE OR QUOTE
-------
Table B-7. A summary of BMDS (version 1.3.2) modeling results based on
incidence of hyperplasia of mucus-secreting cells in male Sprague-Dawley
rats exposed to AN via inhalation for 2 years
Model
X2 /j-value"
AIC
BMC10b
(mg/m3)
BMCL10c
(mg/m3)
Gamma
0.34
30.27
0.396
0.252
Logistic
0.02
38.05
1.17
0.718
Log-logistic
0.94
28.43
0.187
0.082
Multistage
0.34
30.27
0.396
0.252
Probit
0.01
37.99
1.17
0.776
Log-probit
0.37
29.98
0.625
0.382
Weibull
0.34
30.27
0.396
0.252
a%2/>-value from the %2 test for lack of fit. Values <0.1 fail to meet conventional goodness-of-fit criteria.
bBMCio = BMC associated with 10% extra risk for nonneoplastic nasal lesions.
°BMCL10 = 95% lower confidence limit on the BMCi0 for nonneoplastic nasal lesions.
Source: Quast et al. (1980b).
B-117
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.3.2) model output for the best-fit model (i.e., log-logistic) based on
incidence of hyperplasia of mucus-secreting cells in male Sprague-Dawley rats exposed to
AN via inhalation for 2 years
Log-Logistic Model with 0.95 Confidence Level
Log-Logistic
0.8
0.6
it
<
* 0.4
o
CD
i_
Ll_
0.2
BMD
2
4
6
8
dose
13:38 05/16 2007
Logistic Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:20 $
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\NONCANCER\INHALATION\SD_MALE_NASAL_INHALATION.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\NONCANCER\INHALATION\SD_MALE_NASAL_INHALATION.pit
Wed May 16 13:38:57 2007
BMDS MODEL RUN
The form of the probability function is:
P[response] = background+(1-background)/[1+EXP(- intercept-slope*Log(dose))]
Dependent variable = Response
Independent variable = Dose
Slope parameter is restricted as slope >= 1
Total number of observations = 3
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
B-118
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
background = 0
intercept = -0.581093
slope = 1
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -background -slope
have been estimated at a boundary point, or have been specified by the user, and
do not appear in the correlation matrix )
intercept
intercept 1
Parameter Estimates
Variable
background
intercept
slope
Estimate
-0.522564
1
Std. Err.
NA
0 .479706
NA
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-13 .1543
-13 .2153
-22 .7373
Deviance Test DF
0.121833
19.1659
P-value
0 . 940S
<.0001
AIC:
28.4305
Goodness of Fit
Dose
Est. Prob.
Expected
Observed
Size
Scaled
Residual
0.0000
2.1000
8.5000
Chi-square =
0.0000
0.5546
0.8345
0.13
0 . 000
6 .655
8.345
DF = 2
11
12
10
P-value = 0.9390
0
0 .2001
-0 .2931
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.187372
BMDL = 0.0818673
B-119
DRAFT- DO NOT CITE OR QUOTE
-------
Table B-8. A summary of BMDS (version 1.3.2) modeling results based on
incidence of flattening of respiratory epithelium in female Sprague-Dawley
rats exposed to AN via inhalation for 2 years
Model
X2 /j-value"
AIC
BMC10b
(mg/m3)
BMCL10c
(mg/m3)
Gamma
0.08
35.78
0.419
0.245
Logistic
0.02
38.47
1.02
0.610
Log-logistic
0.43
33.50
0.162
0.059
Multistage
0.08
35.78
0.419
0.245
Probit
0.02
38.54
1.05
0.683
Log-probit
0.08
35.53
0.616
0.340
Weibull
0.08
35.78
0.419
0.245
a%2 /'-value from the %2 test for lack of fit. Values <0.1 fail to meet conventional goodness-of-fit criteria.
bBMCio = BMC associated with 10% extra risk for nonneoplastic nasal lesions.
°BMCLio = 95% lower confidence limit on the BMCio for nonneoplastic nasal lesions.
Source: Quast et al. (1980b).
B-120
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.3.2) model output for the best-fit model (log-logistic) based on incidence
of flattening of respiratory epithelium in female Sprague-Dawley rats exposed to AN via
inhalation for 2 years
Log-Logistic Model with 0.95 Confidence Level
Log-Logistic
0.8
T3
0)
1 0.6
it=
<
c
o
"C
CO
0.4
u_
0.2
BMDU BMD
0
2
4
6
8
dose
16:25 05/14 2007
Logistic Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:20 $
Input Data File: G:\ACN DOSE-RESPONSE MODELING\NONCANCER\SD_FEMALE_NASAL_INHALATION.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\NONCANCER\SD_FEMALE_NASAL_INHALATION.p1t
Mon May 14 16:25:24 2007
BMDS MODEL RUN
The form of the probability function is:
P[response] = background+(1-background)/[1+EXP(- intercept-slope*Log(dose))]
Dependent variable = Response
Independent variable = Dose
Slope parameter is restricted as slope >= 1
Total number of observations = 3
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 Parameter Values
background = 0.0909091
intercept = -0.457634
slope = 1
B-121
DRAFT- DO NOT CITE OR QUOTE
-------
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -slope
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
background intercept
background 1 -0.27
intercept -0.27 1
Variable
background
intercept
slope
Parameter Estimates
Estimate Std. Err.
0.0938695 0.089458
-0.374871 0.586482
1 NA
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-14 .4637
-14.751
-21.4714
33 .502
Deviance Test DF
0.574692
14 . 0155
P-value
0 .4484
0 . 0009048
Goodness of Fit
Dose
Est. Prob.
Expected
Observed
Size
Scaled
Residual
0.0000
2.1000
8.5000
Chi-square =
0 . 0939
0.6292
0.8676
0.61
1. 033
6 .292
8.676
DF = 1
11
10
10
P-value = 0.4334
-0 . 03367
0.4637
-0.6305
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.161645
BMDL = 0.0593975
B-122
DRAFT- DO NOT CITE OR QUOTE
-------
APPENDIX B-3. CANCER ORAL DOSE-RESPONSE ASSESSMENT: BMD DOSE
MODELING RESULTS FOR TUMOR INCIDENCE DATA FROM RATS
CHRONICALLY EXPOSED TO AN IN DRINKING WATER
As summarized in Section 4.6.1, AN is a multisite carcinogen in chronic oral rodent
bioassays. Oral carcinogenicity studies of animals chronically exposed to AN include drinking
water studies in two strains of rats, Sprague-Dawley (Johannsen and Levinskas, 2002a; Quast,
2002; Biodynamics, 1980a; Quast et al., 1980a) and F344 (Johannsen and Levinskas, 2002b;
Biodynamics, 1980b). In Sprague-Dawley rats, significantly increased incidences of
forestomach, CNS, Zymbal gland, tongue, and mammary gland (females only) tumors were
found. In F344 rats, significantly increased incidences of forestomach, CNS, Zymbal gland, and
mammary gland (females only) tumors were found.
Two of these chronic drinking water studies were selected for dose-response modeling
using the BMD approach and derivation of oral CSFs for AN. In one study by Quast (2002),
Sprague-Dawley rats were exposed to 0, 35, 100, or 300 ppm of AN in drinking water for
2 years. In the second study by Johannsen and Levinskas (2002b), F344 rats were exposed to 0,
1,3, 10, 30, or 100 ppm of AN in drinking water for 2 years.
In this appendix, detailed results of the dose-response modeling for each of the tumor
sites listed above are presented (Tables B-9 though B-26). For each tumor site, first a summary
of the dose-response data is presented, followed by a table summarizing the results of the dose-
response modeling. Finally, the standard output from EPA's BMDS, version 1.4.1, for the
selected dose-response model for each tumor site is presented.
In general, the multistage model was fit to all of the data sets with the BMR set at
0.1 (i.e., 10% extra risk). In fitting this model, successive stages of the multistage model,
starting with stage 1 and ending with the stage equal to the number of dose groups minus one,
were fit to the tumor incidence data at a particular site for each rat strain and sex employing the
internal dose metrics CEO in blood and AN in blood. Then, for each dose metric, all stages of
the multistage model that did not show a significant lack of fit (i.e. ,p> 0. 1) were compared
using AIC. The stage of the multistage model with the lowest AIC was selected as the best-fit
model. For most tumor sites, the one-stage model exhibited the best fit. For data sets that
exhibited a significant lack of fit for all stages of the multistage model, dose groups were
dropped (starting with the highest dose group) until an adequate fit was achieved.
B-123
DRAFT- DO NOT CITE OR QUOTE
-------
Sprague-Dawlev Rats (Quasi 2002; Quast et al., 1980a)
Tumor Site: Forestomach
Table B-9. Incidence of forestomach (nonglandular) tumors in Sprague-
Dawley rats exposed to AN in drinking water for 2 years
Sex
Administered
animal dose
(ppm in drinking
water)
Equivalent
administered
animal dose"
(mg/kg-d)
Predicted internal dose metrics
Incidence of
forestomach
tumorsb
AN-AUC in blood
(mg/L)
CEO-AUC in blood
(mg/L)
Male
0
0
0
0
0/80 (0%)
35
3.42
2.06 x 10-2
1.83 x 10"3
2/47 (4%)
100
8.53
5.36 x 10-2
4.36 x 10"3
23/48 (48%)°
300
21.2
1.46 x 10-1
9.70 x 10-3
39/48 (81%)°
Female
0
0
0
0
1/80 (1%)
35
4.36
2.37 x 10-2
2.07 x 10-3
1/48 (2%)
100
10.8
6.18 x 10-2
4.87 x 10-3
12/48 (25%)°
300
25.0
1.56 x 10-1
1.01 x 10-2
30/48 (62%)°
aAdministered doses were averages calculated by the study authors based on animal BW and drinking water intake,
incidences for Sprague-Dawley rats do not include animals from the 6- and 12-mo sacrifices and were further
adjusted to exclude (from the denominators) rats that died between 0 and 12 mos in the study.
Significantly different from controls (p < 0.05) as calculated by the study authors.
Sources: Quast (2002); Quast et al. (1980a).
Table B-10. Summary of BMD modeling results based on incidence of
forestomach (nonglandular) tumors in Sprague-Dawley rats exposed to AN
in drinking water for 2 years
Dose metric
Best-fit model"
X2 />-valucb
AIC
BMD10C
BMDL10d
Males
Administered dose
2°MSe
0.46
86.82
3.62 mg/kg-d
2.76 mg/kg-d
CEO
2°MSe
0.39
87.28
1.87 x 10-3 mg/L
1.44 x 10-3 mg/L
AN
2°MSe
0.54
86.45
2.26 x 10"2 mg/L
1.70 x 10"2 mg/L
Females
Administered dose
2°MS
0.17
145.97
7.76 mg/kg-d
4.81 mg/kg-d
CEO
2°MS
0.52
143.50
3.29 x 10-3 mg/L
2.38 x 10-3 mg/L
AN
2°MS
0.13
146.45
4.22 x 10"2 mg/L
2.49 x 10"2 mg/L
aDose-response models were fit using BMDS, version 1.4.1. "2°MS" indicates a two-stage multistage model.
hp value from the %2 goodness-of-fit test. Values <0.1 indicate a significant lack of fit.
°BMD10 = BMD at 10% extra risk.
dBMDLio = 95% lower confidence limit on the BMD at 10% extra risk.
"Highest dose dropped prior to model fitting.
B-124
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for forestomach tumors in Sprague-Dawley male rats
employing administered dose as a dose metric
Multistage Cancer Model with 0.95 Confidence Level
0.6
Multistage Cancer
Linear extrapolation
T -
0.5
0.4
I
0.3
_L
0.2
0.1
1 1
0
BMDL
±1
BMP , , , ,
012345678
dose
15:40 01/23 2009
Multistage Cancer Model. (Version: 1.5; Date: 02/20/2007)
Input Data File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_MALE_FORESTOMACH_DW.(d)
Gnuplot Plotting File: M:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_MALE_FORESTOMACH_DW.pit
Fri Jan 23 15:40:45 2009
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2) ]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
B-125
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0
Beta(l) = 0
Beta(2) = 0.00929613
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Background -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Beta(2)
Beta(2) 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0 * * *
Beta(1) 0 * * *
Beta(2) 0.00803192 * * *
* - Indicates that this value is not calculated.
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
#
Log(likelihood) # Param'
-41.5002 3
-42.4099 1
-71.7704 1
Deviance Test d.f.
P-value
1.81939
60.5403
0 .4026
<.0001
AIC:
86 .8198
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0000 0.000 0 80 0.000
3.4200 0.0897 4.214 2 47 -1.131
8.5300 0.4426 21.243 23 48 0.511
ChiA2 = 1.54 d.f. = 2 P-value = 0.4633
B-126
DRAFT- DO NOT CITE OR QUOTE
-------
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 3 .62184
BMDL = 2 .75694
BMDU = 4 .3186
Taken together, (2.75694, 4.3186 ) is a 90 % two-sided confidence
interval for the BMD
Multistage Cancer Slope Factor = 0.0362721
B-127 DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for forestomach tumors in Sprague-Dawley male rats
employing CEO in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
T3
0
ts
0
<
c
o
"8
0.6
0.5
0.4
0.3
0.2
0.1
0
Multistage
T -
. T ^
, BMDlJ ,
¦CD
^ \
o \
r 1 1 1 1
0 0.0005 0.001 0.0015 0.002 0.0025 0.003 0.0035 0.004 0.0045
dose
13:46 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_MALE_FORESTOMACH_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_MALE_FORESTOMACH_BLOOD_CEO.pit
Thu Sep 27 13:46:37 2007
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2) ]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
B-128
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0
Beta(l) = 0
Beta(2) = 35739.1
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Background -Beta(l)
have been estimated at a boundary point, or have been specified by the user, and
do not appear in the correlation matrix )
Beta(2)
Beta(2) 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0 * * *
Beta(1) 0 * * *
Beta(2) 30229.1 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-41.5002
-42 .6376
-71.7704
# Param'
3
1
1
Deviance Test d.f.
2 .27473
60.5403
P-value
0.3207
<.0001
AIC:
87 .2752
Dose
Goodness of Fit
Est._Prob. Expected Observed Size
Scaled
Residual
0.0000
0.0018
0.0044
0.0000
0.0963
0.4371
0 . 000
4 .525
20 . 981
0
2
23
80
47
48
0 . 000
-1.249
0.588
Chia2 =1.90
d.f. = 2
P-value = 0.385S
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0.1
Extra risk
0 . 95
0 . 00186692
0 . 00143604
0 . 00222594
Taken together, (0.00143604, 0.00222594) is a 90
% two-sided confidence interval for the BMD
B-129
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for forestomach tumors in Sprague-Dawley male rats
employing AN in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
Multistage
0.6
0.5
T3
0
0.4
<
c 0.3
o
"8
ro
i_
LL
0.2
BMDL
BMD
0
0.01
0.02
0.03
0.04
0.05
dose
10:32 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_MALE_FORESTOMACH_BLOOD_AN.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_MALE_FORESTOMACH_BLOOD_AN.pit
Thu Sep 27 10:32:01 2007
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2) ]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
B-130
DRAFT- DO NOT CITE OR QUOTE
-------
Background = 0
Beta(l) = 0
Beta(2) = 234.473
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Background -Beta(l)
have been estimated at a boundary point, or have been specified by the user, and
do not appear in the correlation matrix )
Beta(2)
Beta(2) 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0 * * *
Beta(1) 0 * * *
Beta(2) 206.388 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-41.5002
-42 .2261
-71.7704
# Param'
3
1
1
Deviance Test d.f.
1.45174
60.5403
P-value
0 .4835
<.0001
AIC:
86.4522
Dose
Est. Prob.
Goodness of Fit
Expected
Observed
Size
Scaled
Residual
0.0000
0.0206
0.0536
0.0000
0 . 0839
0.4473
0 . 000
3 . 941
21.470
0
2
23
80
47
48
0 . 000
-1. 022
0 .444
Chia2 = 1.24
d.f. = 2
P-value = 0.5377
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.0225942
BMDL = 0.0170019
BMDU = 0.0269423
Taken together, (0.0170019, 0.0269423) is a 90 % two-sided confidence
interval for the BMD
B-131
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for forestomach tumors in Sprague-Dawley female rats
employing administered dose as a dose metric
0.8
0.7
0.6
TD
3 0.5
£
< 0.4
c
o
tj 0.3
ro
L_
LL
0.2
0.1
0
0 5 10 15 20 25
dose
16:14 01/23 2009
Multistage Cancer Model. (Version: 1.5; Date: 02/20/2007)
Input Data File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_FEMALE_FORESTOMACH_DW.(d)
Gnuplot Plotting File: M:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_FORESTOMACH_DW.pit
Fri Jan 23 16:14:45 2009
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2) ]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 4
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
B-132 DRAFT- DO NOT CITE OR QUOTE
Multistage Cancer Model with 0.95 Confidence Level
Multistage Cancer
Linear extrapolation
T
-
i -
| ^ '
BMDlJ BMD ,
-------
Default Initial Parameter Values
Background = 0
Beta(1) = 0.0138175
Beta(2) = 0.00104033
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l) Beta(2)
Background 1 -0.57 0.3 9
Beta(1) -0.57 1 -0.93
Beta(2)
0.39
-0 . 93
Variable
Background
Beta(1)
Beta(2)
Parameter Estimates
Estimate
0 . 0107986
0 . 000833319
0 . 00164062
Std. Err.
* - Indicates that this value is not calculated.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
Log(likelihood)
-68 . 9836
-69 . 9834
-110.972
# Param's
4
3
1
Deviance Test d.f.
1. 99958
83 . 9771
P-value
0.1573
<.0001
AIC:
145 . 967
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0108 0.864 1 80 0.147
4.3600 0.0447 2.143 1 48 -0.799
10.8000 0.1904 9.139 12 48 1.052
25.0000 0.6525 31.321 30 48 -0.401
Chia2 = 1.93 d.f. = 1 P-value = 0.1652
B-133
DRAFT- DO NOT CITE OR QUOTE
-------
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 7.76379
BMDL = 4 .81488
BMDU = 9.11082
Taken together, (4.81488, 9.11082) is a 90 % two-sided confidence
interval for the BMD
Multistage Cancer Slope Factor = 0.0207689
B-134 DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for forestomach tumors in Sprague-Dawley female rats
employing CEO in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
0.8
Multistage
0.7
0.6
T3
ffi 0.5
< 0.4
c
o
tj 0.3
ro
LL
0.2
BMDlJ
BMD
0
0.002
0.004
0.006
0.008
0.01
dose
14:24 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_FORESTOMACH_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_FORESTOMACH_BLOOD_CEO.pit
Thu Sep 27 14:24:06 2007
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2) ]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 4
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
B-135
DRAFT- DO NOT CITE OR QUOTE
-------
Background = 0
Beta(1) = 14.7147
Beta(2) = 8253.52
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Beta(2)
Background 1 -0.48
Beta(2) -0.48 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0.00991437 * * *
Beta(1) 0 * * *
Beta(2) 9752.18 * * *
* - Indicates that this value is not calculated.
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
#
Log(likelihood)
-68 . 9836
-69.7499
-110.972
Param's
4
2
1
Deviance Test d.f.
P-value
1.53257
83 . 9771
0 .4647
<.0001
AIC:
143 .5
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0099 0.793 1 80 0.233
0.0021 0.0504 2.421 1 48 -0.937
0.0049 0.2144 10.289 12 48 0.602
0.0101 0.6339 30.426 30 48 -0.128
Chia2 = 1.31 d.f. = 2 P-value = 0.5192
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00328691
BMDL = 0.00238295
BMDU = 0.0037821
Taken together, (0.00238295, 0.0037821) is a 90 % two-sided confidence
interval for the BMD
B-136
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for forestomach tumors in Sprague-Dawley female rats
employing AN in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
0.8
Multistage
0.7
0.6
T3
ffi 0.5
< 0.4
c
o
tj 0.3
ro
LL
0.2
BMD
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
dose
10:49 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_FORESTOMACH_BLOOD_AN.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_FORESTOMACH_BLOOD_AN.pit
Thu Sep 27 10:49:49 2007
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2) ]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 4
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
B-137
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0
Beta(1) = 3.15264
Beta(2) = 20.8062
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l) Beta(2)
Background 1 -0.57 0.4
Beta(1) -0.57 1 -0.93
Beta(2)
0.4
-0 . 93
Variable
Background
Beta(1)
Beta(2)
Parameter Estimates
Estimate
0 . 0105127
0 . 96299
36 .4178
Std. Err.
* - Indicates that this value is not calculated.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
Log(likelihood)
-68 . 9836
-70 .2231
-110.972
# Param's
4
3
1
Deviance Test d.f.
2 .47911
83 . 9771
P-value
0.1154
<.0001
AIC:
146 .446
Dose
Goodness of Fit
Est. Prob.
Expected
Observed
Size
Scaled
Residual
0.0000
0 . 0237
0.0618
0.1560
Chia2 = 2.2c
0.0105
0 . 0524
0.1887
0.6490
d.f. = 1
0.841 1 80
2.516 1 48
9.060 12 48
31.154 30 48
P-value = 0.1300
0.174
-0.982
1. 085
-0.349
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0.1
Extra risk
0 . 95
0.0421673
0.0249138
0.055287
Taken together, (0.0249138, 0.055287) is a 90
interval for the BMD
% two-sided confidence
B-138
DRAFT- DO NOT CITE OR QUOTE
-------
Tumor Site: CNS (Quast, 2002; Quast et al., 1980a)
Table B-ll. Incidence of CNS tumors in Sprague-Dawley rats exposed to
AN in drinking water for 2 years
Sex
Administered
animal dose
(ppm in drinking
water)
Equivalent
administered
animal dose"
(mg/kg-d)
Predicted internal dose metrics
Incidence of CNS
tumorsb
AN-AUC in blood
(mg/L)
CEO-AUC in blood
(mg/L)
Male
0
0
0
0
1/80 (1%)
35
3.42
2.06 x 10-2
1.83 x 10-3
12/47 (26%)°
100
8.53
5.36 x 10-2
4.36 x 10-3
22/48 (46%)°
300
21.2
1.46 x 10-1
9.70 x 10-3
30/48 (62%)°
Female
0
0
0
0
1/80 (1%)
35
4.36
2.37 x 10-2
2.07 x 10-3
20/48 (42%)°
100
10.8
6.18 x 10-2
4.87 x 10-3
25/48 (52%)°
300
25.0
1.56 x 10-1
1.01 x 10-2
31/48 (65%)°
aAdministered doses were averages calculated by the study authors based on animal BW and drinking water
intake.
incidences for Sprague-Dawley rats do not include animals from the 6- and 12-mo sacrifices and were further
adjusted to exclude (from the denominators) rats that died between 0 and 12 mos in the study.
"Significantly different from controls (p < 0.05) as calculated by the study authors.
Table B-12. Summary of BMD modeling results based on incidence of CNS
tumors in Sprague-Dawley rats exposed to AN in drinking water for 2 years
Dose metric
Best-fit model3
X2 /j-valuc'
AIC
BMD10°
BMDL10d
Males
Administered dose
1°MS
0.16
201.38
1.84 mg/kg-d
1.48 mg/kg-d
CEO
1°MS
0.37
199.81
8.87 x 10"4 mg/L
7.16 x 10"4 mg/L
AN
l°MSe
0.59
134.64
8.82 x 10"3 mg/L
6.64 x 10"3 mg/L
Females
Administered dose
l°MSe
0.06
149.85
1.26 mg/kg-d
0.99 mg/kg-d
CEO
l°MSe
0.08
149.29
5.79 x 10"4 mg/L
4.51 x 10"4 mg/L
AN
l°MSe
0.04
150.45
7.12 x 10-3 mg/L
5.55 x 10"3 mg/L
aDose-response models were fit using BMDS, version 1.4.1. "1°MS" indicates a one-stage multistage model.
bp value from the %2 goodness of fit test. Values <0.1 indicate a significant lack of fit.
°BMD10 = BMD at 10% extra risk.
dBMDL10 = 95% lower confidence limit on the BMD at 10% extra risk.
"Highest dose dropped prior to model fitting.
B-139
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for CNS tumors in Sprague-Dawley male rats employing
administered dose as a dose metric
Multistage Cancer Model with 0.95 Confidence Level
0.8
0.7
Multistage Cancer
Linear extrapolation
—
0.6
T -
0.5
i -
0.4
0.3
0.2
0.1
I
0
BMDL BMD
0 5 10 15 20
dose
15:48 01/23 2009
Multistage Cancer Model. (Version: 1.5; Date: 02/20/2007)
Input Data File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_MALE_CNS_DW.(d)
Gnuplot Plotting File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_MALE_CNS_DW.pit
Fri Jan 23 15:48:14 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 = Response
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: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
B-140
DRAFT- DO NOT CITE OR QUOTE
-------
Background = 0.108455
Beta(1) = 0.0435027
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.62
Beta(1) -0.62 1
Parameter Estimates
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Variable
Background
Beta(1)
Estimate
0 . 0160781
0 . 0572137
Std. Err.
•k
•k
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-96 . 9359
-98.6909
-134 .574
# Param's
4
2
1
Deviance Test d.f.
3 .51012
75 .2765
P-value
0.172S
<.0001
AIC:
201.382
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0161 1.286 1 80 -0.254
3.4200 0.1909 8.974 12 47 1.123
8.5300 0.3960 19.010 22 48 0.882
21.2000 0.7075 33.958 30 48 -1.256
Chia2 = 3.68 d.f. = 2 P-value = 0.1587
B-141
DRAFT- DO NOT CITE OR QUOTE
-------
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 1.84153
BMDL = 1.4836
BMDU = 2 .34034
Taken together, (1.4836 , 2.34034) is a 90 % two-sided confidence
interval for the BMD
Multistage Cancer Slope Factor = 0.0674034
B-142 DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for CNS tumors in Sprague-Dawley male rats employing
CEO in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
Multistage
0.7
0.6
3 0.5
"G
| 0.4
* 0.3
o
CD
i_
Ll_
0.2
0 0.002 0.004 0.006 0.008 0.01
dose
09:11 09/25 2007
Multistage Model. $Revision: 2.1 $ $Date: 2000/08/21 03:38:21 $
Input Data File: G:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_MALE_CNS_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_MALE_CNS_BLOOD_CEO.pit
Tue Sep 25 09:11:57 2007
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
Background = 0.0859878
Beta(1) = 97.0212
B-143
DRAFT- DO NOT CITE OR QUOTE
-------
**** WARNING 2
**** WARNING 3
**** WARNING 4
**** WARNING 5
**** WARNING 6
WARNING 7
WARNING 8
WARNING 9
WARNING:
WARNING 0
WARNING 1
Completion code = -3. Optimum not found. Trying new starting point****
Completion code = -3 trying new start****
-3 trying new start****
-3 trying new start****
-3 trying new start****
-3 trying new start****
-3 trying new start****
-3 trying new start****
-3 trying new start****
-3 trying new start****
-3 trying new start****
Completion code
Completion code
Completion code
Completion code
Completion code
Completion code
Completion code
Completion code
Completion code
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.62
Beta(1) -0.62 1
Variable
Background
Beta(1)
Parameter Estimates
Estimate Std. Err.
0.0146577 0.100785
118.766 28.4627
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-96 . 9359
-97.9055
-134 .574
199.811
Deviance Test DF
1. 93933
75 .2765
P-value
0.3792
<.0001
Goodness of Fit
Dose Est._Prob. Expected Observed Size
Chi 2 Res.
i : 1
0.0000
i: 2
0.0018
i : 3
0 . 0044
i: 4
0.0097
Chi-square =
0 . 0147
0.2071
0 .4129
0.6886
2 . 01
1.173
9.736
19.820
33.055
DF = 2
1 80 -0.149
12 47 0.293
22 48 0.187
30 48 -0.297
P-value = 0.3669
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00088713
BMDL = 0.00071632
B-144
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for CNS tumors in Sprague-Dawley male rats employing AN
in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
0.6
0.5
1 04
£
C 0.3
0
"8
1 02
0.1
0
0 0.01 0.02 0.03 0.04 0.05
dose
10:40 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_MALE_CNS_BLOOD_AN.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_MALE_CNS_BLOOD_AN.pit
Thu Sep 27 10:40:46 2007
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
Multistage
BMDlJ B|VID
B-145
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.032766
Beta(1) = 11.0585
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.61
Beta(1) -0.61 1
Parameter Estimates
Variable
Background
Beta(1)
Estimate
0 . 0130586
11. 9435
Std. Err.
* - Indicates that this value is not calculated.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-65.1808
-65.3204
-87.5704
134 .641
# Param's
3
2
1
Deviance Test d.f.
0 .279194
44 .7792
P-value
0.5972
<.0001
Dose
Goodness of Fit
Est._Prob. Expected Observed Size
Scaled
Residual
0.0000
0.0206
0.0536
0 . 0131
0.2283
0.4797
Chia2 = 0.28 d.f. = 1
Benchmark Dose Computation
1.045 1
10.731 12
23.025 22
P-value = 0.5940
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0.1
Extra risk
0 . 95
0 . 00882158
0 . 00664413
0 . 0121508
80
47
48
-0 . 044
0 .441
-0.296
Taken together, (0.00664413, 0.0121508) is a 90
interval for the BMD
two-sided confidence
B-146
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for CNS tumors in Sprague-Dawley female rats employing
administered dose as a dose metric
Multistage Cancer Model with 0.95 Confidence Level
Multistage Cancer
Linear extrapolation
0.7
0.6
0.5
0.4
0.3
0.2
BMDlJ BMP
0 2 4 6 8 10
dose
16:19 01/23 2009
Multistage Cancer Model. (Version: 1.5; Date: 02/20/2007)
Input Data File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_FEMALE_CNS_DW.(d)
Gnuplot Plotting File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_FEMALE_CNS_DW.pit
Fri Jan 23 16:19:26 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 = Response
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: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
B-147
DRAFT- DO NOT CITE OR QUOTE
-------
Background = 0.0993663
Beta(1) = 0.0642026
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.62
Beta(1) -0.62 1
Parameter Estimates
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Variable
Background
Beta(1)
Estimate
0 . 0141417
0 . 0833707
Std. Err.
•k
•k
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-71.2064
-72 . 9251
-101.108
# Param's
3
2
1
Deviance Test d.f.
3 .43738
59.8036
P-value
0 . 06374
<.0001
AIC:
149.85
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0141 1.131 1 80 -0.124
4.3600 0.3146 15.100 20 48 1.523
10.8000 0.5993 28.768 25 48 -1.110
ChiA2 = 3.57 d.f. = 1 P-value = 0.0589
B-148
DRAFT- DO NOT CITE OR QUOTE
-------
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 1.26376
BMDL = 0.985061
BMDU = 1.66456
Taken together, (0.985061, 1.66456) is a 90 % two-sided confidence
interval for the BMD
Multistage Cancer Slope Factor = 0.101517
B-149 DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for CNS tumors in Sprague-Dawley female rats employing
CEO in blood as an internal dose metric
Multistage Cancer Model with 0.95 Confidence Level
0.7
0.6
0.5
T3
0
© 0.4
it
<
o 0.3
"8
t 0.2
0.1
0
0 0.001 0.002 0.003 0.004 0.005
dose
09:03 12/03 2008
Multistage Cancer Model. (Version: 1.5; Date: 02/20/2007)
Input Data File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_FEMALE_CNS_BLOOD_CEO.(d)
Gnuplot Plotting File: M:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_CNS_BLOOD_CEO.pit
Wed Dec 03 09:03:51 2008
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
Multistage Cancer
Linear extrapolation
BMDlJ BMP
B-150
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.0914603
Beta(1) = 144.025
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background
Beta(1)
1
-0.62
-0.62
1
Variable
Background
Beta(1)
Parameter Estimates
Estimate
0 . 0138803
182.002
Std. Err.
* - Indicates that this value is not calculated.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-71.2064
-72.6449
-101.108
149 .29
# Param's
3
2
1
Deviance Test d.f.
2 .87694
59.8036
P-value
0 . 08986
<.0001
Goodness of Fit
Dose Est._Prob. Expected Observed Size
Scaled
Residual
0.0000
0 . 0021
0 . 0049
Chia2 =2.97
0 . 0139
0.3234
0.5936
d.f. = 1
1.110 1 80 -0.106
15.525 20 48 1.381
28.491 25 48 -1.026
P-value = 0.0848
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0.1
Extra risk
0 . 95
0 . 000578899
0 . 000451438
0 . 000761804
Taken together, (0.000451438, 0.000761804) is a 90
interval for the BMD
% two-sided confidence
Multistage Cancer Slope Factor =
221.515
B-151
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for CNS tumors in Sprague-Dawley female rats employing
AN in blood as an internal dose metric
Multistage Cancer Model with 0.95 Confidence Level
0.7
0.6
0.5
T3
0
© 0.4
it
<
o 0.3
"8
t 0.2
0.1
0
0 0.01 0.02 0.03 0.04 0.05 0.06
dose
09:12 12/03 2008
Multistage Cancer Model. (Version: 1.5; Date: 02/20/2007)
Input Data File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_FEMALE_CNS_BLOOD_AN.(d)
Gnuplot Plotting File: M:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_CNS_BLOOD_AN.pit
Wed Dec 03 09:12:32 2008
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
Multistage Cancer
Linear extrapolation
BMDlJ BMP
B-152
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.106975
Beta(1) = 11.0861
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.62
Beta(1) -0.62 1
Parameter Estimates
Variable
Background
Beta(1)
Estimate
0 . 0144346
14 .7901
Std. Err.
* - Indicates that this value is not calculated.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Deviance Test d.f.
Log(likelihood)
-71.2064
-73.2271
-101.108
150 .454
# Param's
3
2
1
4 . 04128
59.8036
P-value
0 . 0444
<.0001
Dose
Est. Prob.
Goodness of Fit
Expected Observed Size
Scaled
Residual
0.0000
0 . 0144
1.155
1
80
-0.145
0 . 0237
0.3058
14 .681
20
48
1.666
0.0618
0.6049
29 . 034
25
48
-1.191
Chia2 = 4.22
d.f.
= 1 P-value
= 0.0400
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0.1
Extra risk
0 . 95
0 . 0071237
0 . 00554992
0 . 00939265
Taken together, (0.00554992, 0.00939265) is a 90
interval for the BMD
% two-sided confidence
Multistage Cancer Slope Factor =
18 . 0183
B-153
DRAFT- DO NOT CITE OR QUOTE
-------
Tumor Site: Zymbal Gland (Quast, 2002; Quast et al., 1980a)
Table B-13. Incidence of Zymbal gland tumors in Sprague-Dawley rats
exposed to AN in drinking water for 2 years
Sex
Administered
animal dose
(ppm in drinking
water)
Equivalent
administered
animal dose"
(mg/kg-d)
Predicted internal dose metrics
Incidence of
Zymbal gland
tumorsb
AN-AUC in blood
(mg/L)
CEO-AUC in blood
(mg/L)
Male
0
0
0
0
3/80 (4%)
35
3.42
2.06 x 10-2
1.83 x 10-3
4/47 (9%)
100
8.53
5.36 x 10-2
4.36 x 10-3
3/48 (6%)
300
21.2
1.46 x 10-1
9.70 x 10-3
16/48 (33%)°
Female
0
0
0
0
1/80 (1%)
35
4.36
2.37 x 10-2
2.07 x 10-3
5/48 (10%)c
100
10.8
6.18 x 10-2
4.87 x 10-3
9/48 (19%)°
300
25.0
1.56 x 10-1
1.01 x 10-2
18/48 (38%)°
aAdministered doses were averages calculated by the study authors based on animal BW and drinking water intake,
incidences for Sprague-Dawley rats do not include animals from the 6- and 12-mo sacrifices and were further
adjusted to exclude (from the denominators) rats that died between 0 and 12 mos in the study.
"Significantly different from controls (p < 0.05) as calculated by the study authors.
Table B-14. Summary of BMD modeling results based on incidence of
Zymbal gland tumors in Sprague-Dawley rats exposed to AN in drinking
water for 2 years
Dose metric
Best-fit model3
X2 /j-valuc'
AIC
BMD10°
BMDL10d
Males
Administered dose
2°MS
0.43
142.15
11.80 mg/kg-d
6.29 mg/kg-d
CEO
2°MS
0.38
142.46
5.46 x 10"3 mg/L
3.15 x 10-3 mg/L
AN
3°MS
0.28
143.58
8.94 x 10"2mg/L
4.26 x 10"2 mg/L
Females
Administered dose
1°MS
0.94
156.80
5.66 mg/kg-d
4.19 mg/kg-d
CEO
1°MS
0.95
156.77
2.40 x 10-3 mg/L
1.78 x 10-3 mg/L
AN
1°MS
0.85
156.98
3.41 x 10"2 mg/L
2.52 x 10"2 mg/L
aDose-response models were fit using BMDS, version 1.4.1. "1°MS" indicates a one-stage multistage model,
"2°MS" indicates a two-stage multistage model, and "3°MS" indicates a three-stage multistage model.
bp value from the %2 goodness of fit test. Values <0.1 indicate a significant lack of fit.
°BMD10 = BMD at 10% extra risk.
dBMDL10 = 95% lower confidence limit on the BMD at 10% extra risk.
B-154
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for Zymbal gland tumors in Sprague-Dawley male rats
employing administered dose as a dose metric
Multistage Cancer Model with 0.95 Confidence Level
Multistage Cancer
Linear extrapolation
0.5
0.4
T3
0
1 0.3
it
<
c
o
0.2
"8
ro
i_
LL
BMDL
5
BMD
0
10
15
20
dose
16:01 01/23 2009
Multistage Cancer Model. (Version: 1.5; Date: 02/20/2007)
Input Data File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_MALE_ZYMBAL_DW.(d)
Gnuplot Plotting File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_MALE_ZYMBAL_DW.pit
Fri Jan 23 16:01:39 2009
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2) ]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 4
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
B-155
DRAFT- DO NOT CITE OR QUOTE
-------
Background =
Beta(l) =
Beta(2) =
0 . 042253
0
0 . 00079507
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Beta(2)
Background 1 -0.49
Beta(2) -0.49 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0.0449236 * * *
Beta(1) 0 * * *
Beta(2) 0.000756128 * * *
* - Indicates that this value is not calculated.
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
#
Log(likelihood)
-68 .2481
-69 . 0738
-80.2977
Param's
4
2
1
Deviance Test d.f.
1.65141
24 . 0991
P-value
0 .4375
<.0001
AIC:
142 .148
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0449 3.594 3 80 -0.321
3.4200 0.0533 2.507 4 47 0.969
8.5300 0.0960 4.610 3 48 -0.789
21.2000 0.3201 15.364 16 48 0.197
Chia2 = 1.70 d.f. = 2 P-value = 0.4267
B-156
DRAFT- DO NOT CITE OR QUOTE
-------
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 11.8043
BMDL = 6 .28905
BMDU = 15.532
Taken together, (6.28905, 15.532 ) is a 90 % two-sided confidence
interval for the BMD
Multistage Cancer Slope Factor = 0.0159007
B-157 DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for Zymbal gland tumors in Sprague-Dawley male rats
employing CEO in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
Multistage
0.5
0.4
T3
<1)
"C
<1)
0.3
it
<
c
o
'¦B
CO
Ll_
0.2
BMD
BMQ
0.002
0.004
0.006
0.008
0.01
dose
14:09 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_MALE_ZYMBAL_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_MALE_ZYMBAL_BLOOD_CEO.pit
Thu Sep 27 14:09:41 2007
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2) ]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 4
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
Background = 0.0373755
B-158
DRAFT- DO NOT CITE OR QUOTE
-------
Beta(l) = 0
Beta(2) = 3819.76
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Beta(2)
Background 1 -0.51
Beta(2) -0.51 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0.0429942 * * *
Beta(1) 0 * * *
Beta(2) 3535.6 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-68 .2481
-69.231
-80 .2977
# Param'
4
2
1
Deviance Test d.f.
P-value
1. 96583
24 . 0991
0.3742
<.0001
AIC:
142 .462
B-159
DRAFT- DO NOT CITE OR QUOTE
-------
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0430 3.440 3 80 -0.242
0.0018 0.0543 2.550 4 47 0.934
0.0044 0.1052 5.050 3 48 -0.964
0.0097 0.3138 15.063 16 48 0.291
ChiA2 = 1.94 d.f. = 2 P-value = 0.3781
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0.1
Extra risk
0 . 95
0 . 00545893
0 . 00315099
0 . 00717256
Taken together, (0.00315099, 0.00717256) is a 90
interval for the BMD
two-sided confidence
B-160
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for Zymbal gland tumors in Sprague-Dawley male rats
employing AN in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
Multistage
0.5
0.4
T3
0
0.3
<
c
o
0.2
"8
ro
i_
LL
BMDL
BMD
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
dose
10:44 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_MALE_ZYMBAL_BLOOD_AN.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_MALE_ZYMBAL_BLOOD_AN.pit
Thu Sep 27 10:44:18 2007
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseA1-beta2 *doseA2-beta3 *doseA3)]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 4
Total number of records with missing values = 0
Total number of parameters in model = 4
Total number of specified parameters = 0
Degree of polynomial = 3
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
Background = 0.0542857
Beta(1) = 0.0987621
Beta(2) = 0
B-161
DRAFT- DO NOT CITE OR QUOTE
-------
Beta(3) =
107.536
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(2)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background
Beta(1)
Beta(3)
Background
1
-0.71
0.58
Beta(1)
-0.71
1
-0 . 95
Beta(3)
0.58
-0 . 95
1
Parameter Estimates
Variable
Background
Beta(1)
Beta(2)
Beta(3)
Estimate
0 . 0460186
0.432974
0
93 .2629
Std. Err.
* - Indicates that this value is not calculated.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
Log(likelihood) # Param's Deviance Test d.f. P-value
-68.2481 4
-68.7881 3 1.07988 1 0.2987
-80.2977 1 24.0991 3 <.0001
AIC:
143 .576
Dose
Goodness of Fit
Est. Prob.
Expected
Observed
Size
Scaled
Residual
0.0000
0.0206
0.0536
0.1460
Chia2 =1.16
0 . 0460
0.0553
0.0812
0.3301
d.f. = 1
3.681 3 80
2.597 4 47
3.897 3 48
15.843 16 48
P-value = 0.2812
-0.364
0.896
-0 .474
0 . 048
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0.1
Extra risk
0 . 95
0.0894055
0.0425525
0.117269
Taken together, (0.0425525, 0.117269) is a 90
interval for the BMD
% two-sided confidence
B-162
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for Zymbal gland tumors in Sprague-Dawley female rats
employing administered dose as a dose metric
Multistage Cancer Model with 0.95 Confidence Level
Multistage Cancer
Linear extrapolation
0.5
0.4
0.3
BMDL
BMD
0 5 10 15 20 25
dose
16:28 01/23 2009
Multistage Cancer Model. (Version: 1.5; Date: 02/20/2007)
Input Data File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_FEMALE_ZYMBAL_DW.(d)
Gnuplot Plotting File: M:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_ZYMBAL_DW.pit
Fri Jan 23 16:28:50 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 = Response
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: le-008
Parameter Convergence has been set to: le-008
B-163
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.0190401
Beta(1) = 0.0180112
Background
Beta(1)
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
1 -0.64
-0.64 1
Variable
Background
Beta(1)
Parameter Estimates
Estimate
0 . 013388
0 . 0186127
Std. Err.
* - Indicates that this value is not calculated.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Deviance Test d.f.
Log(likelihood)
-76 .3334
-76.3975
-93.6397
156.795
# Param's
4
2
1
0.128259
34 .6128
P-value
0 . 937S
<.0001
Dose
Est. Prob.
Goodness of Fit
Expected Observed Size
Scaled
Residual
0.0000
0.0134
1. 071
1
80
-0 . 069
4.3600
0.0903
4 .334
5
48
0.335
10.8000
0.1931
9.266
9
48
-0 . 097
25.0000
0.3805
18 .263
18
48
-0.078
Chia2 = 0.13
d.f. = 2
P-value = 0.9357
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0.1
Extra risk
0 . 95
5.66068
4.19199
8.10463
Taken together, (4.19199, 8.10463) is a 90
interval for the BMD
% two-sided confidence
Multistage Cancer Slope Factor =
0.023855
B-164
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for Zymbal gland tumors in Sprague-Dawley female rats
employing CEO in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
Multistage
0.5
0.4
T3
0
0.3
i..
LL
BMDL
BMD
0
0.002
0.004
0.006
0.008
0.01
dose
14:43 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_ZYMBAL_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_ZYMBAL_BLOOD_CEO.pit
Thu Sep 27 14:43:02 2007
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
B-165
DRAFT- DO NOT CITE OR QUOTE
-------
Background = 0.00883347
Beta(1) = 44.8786
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.65
Beta(1) -0.65 1
Parameter Estimates
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Variable
Background
Beta(1)
Estimate
0 . 0125628
43 . 9052
Std. Err.
•k
•k
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-76 .3334
-76.385
-93.6397
# Param's
4
2
1
Deviance Test d.f.
P-value
0.103357
34 .6128
0 . 9496
<.0001
AIC:
156 .77
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000
0 . 0126
1. 005
1
80
-0 . 005
0 . 0021
0.0983
4 .721
5
48
0.135
0 . 0049
0 .2026
9.727
9
48
-0.261
0.0101
0.3662
17.580
18
48
0.126
a2 = 0.10
d.f.
= 2 P-value
= 0.9501
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00239973
BMDL = 0.00178178
BMDU = 0.00341242
Taken together, (0.00178178, 0.00341242) is a 90 % two-sided confidence
interval for the BMD
B-166
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for Zymbal gland tumors in Sprague-Dawley female rats
employing AN in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
0.5
0.4
TD
CD
"8
§ 0.3
0.1
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
dose
10:55 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_ZYMBAL_BLOOD_AN.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_ZYMBAL_BLOOD_AN.pit
Thu Sep 27 10:55:27 2007
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
Multistage
B-167
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.0269441
Beta(1) = 2.86115
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.63
Beta(1) -0.63 1
Parameter Estimates
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Variable
Background
Beta(1)
Estimate
0 . 0143198
3 . 08705
Std. Err.
•k
•k
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-76 .3334
-76.4902
-93.6397
# Param's
4
2
1
Deviance Test d.f.
0.313733
34 .6128
P-value
0.854S
<.0001
AIC:
156 . <
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000
0 . 0143
1.146
1
80
-0.137
0 . 0237
0 . 0839
4 . 025
5
48
0.508
0.0618
0.1855
8 . 905
9
48
0 . 035
0.1560
0.3910
18.770
18
48
-0 .228
a2 = 0.33
d.f.
= 2 P-value
= 0.8481
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.0341298
BMDL = 0.0251994
BMDU = 0.0492353
Taken together, (0.0251994, 0.0492353) is a 90 % two-sided confidence
interval for the BMD
B-168
DRAFT- DO NOT CITE OR QUOTE
-------
Tumor Site: Tongue (Quast, 2002; Quast et al., 1980a)
Table B-15. Incidence of tongue tumors in Sprague-Dawley rats exposed to
AN in drinking water for 2 years
Sex
Administered
animal dose
(ppm in drinking
water)
Equivalent
administered
animal dose"
(mg/kg-d)
Predicted internal dose metrics
Incidence of
tongue tumorsb
AN-AUC in blood
(mg/L)
CEO-AUC in blood
(mg/L)
Male
0
0
0
0
1/80 (1%)
35
3.42
2.06 x 10-2
1.83 x 10-3
2/47 (4%)
100
8.53
5.36 x 10-2
4.36 x 10-3
4/48 (8%)
300
21.2
1.46 x 10-1
9.70 x 10-3
5/48 (10%)c
Female
0
0
0
0
0/80 (0%)
35
4.36
2.37 x 10-2
2.07 x 10-3
1/48 (2%)
100
10.8
6.18 x 10-2
4.87 x 10-3
2/48 (4%)
300
25.0
1.56 x 10-1
1.01 x 10-2
12/48 (25%)c
aAdministered doses were averages calculated by the study authors based on animal BW and drinking water intake,
incidences for Sprague-Dawley rats do not include animals from the 6- and 12-mo sacrifices and were further
adjusted to exclude (from the denominators) rats that died between 0 and 12 mos in the study.
"Significantly different from controls (p < 0.05) as calculated by the study authors.
Table B-16. Summary of BMD modeling results based on incidence of
tongue tumors in Sprague-Dawley rats exposed to AN in drinking water for
2 years
Dose metric
Best-fit model3
X2 /j-valuc'
AIC
BMD10C
BMDL10d
Males
Administered dose
1°MS
0.69
91.62
18.89 mg/kg-d
10.35 mg/kg-d
CEO
1°MS
0.78
91.40
8.78 x 10-3 mg/L
4.90 x 10"3 mg/L
AN
1°MS
0.62
91.83
1.29 x 10"1 mg/L
6.97 x 10"2 mg/L
Females
Administered dose
3°MS
0.92
84.51
15.99 mg/kg-d
10.64 mg/kg-d
CEO
3°MS
0.88
84.59
6.70 x 10-3 mg/L
4.74 x 10-3 mg/L
AN
3°MS
0.93
84.48
9.67 x 10-2 mg/L
6.10 x 10"2 mg/L
aDose-response models were fit using BMDS, version 1.4.1. "1°MS" indicates a one-stage multistage model and
"3°MS" indicates a three-stage multistage model.
bp value from the %2 goodness of fit test. Values <0.1 indicate a significant lack of fit.
°BMD10 = BMD at 10% extra risk.
dBMDL10 = 95% lower confidence limit on the BMD at 10% extra risk.
B-169
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for tongue tumors in Sprague-Dawley male rats employing
administered dose as a dose metric
Multistage Cancer Model with 0.95 Confidence Level
0.25
Multistage Cancer
Linear extrapolation
10
dose
0.15
15
20
15:53 01/23 2009
Multistage Cancer Model. (Version: 1.5; Date: 02/20/2007)
Input Data File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_MALE_TONGUE_DW.(d)
Gnuplot Plotting File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_MALE_TONGUE_DW.pit
Fri Jan 23 15:53:59 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 = Response
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: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
B-170
DRAFT- DO NOT CITE OR QUOTE
-------
Background = 0.0269073
Beta(1) = 0.00434306
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.69
Beta(1) -0.69 1
Parameter Estimates
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Variable
Background
Beta(1)
Estimate
0 . 0162768
0.00557883
Std. Err.
•k
•k
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-43 .4536
-43 .8112
-46 .7384
# Param's
4
2
1
Deviance Test d.f.
P-value
0.715245
6 .56959
0.6993
0 . 08696
AIC:
91.6224
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0163 1.302 1 80 -0.267
3.4200 0.0349 1.639 2 47 0.287
8.5300 0.0620 2.976 4 48 0.613
21.2000 0.1260 6.048 5 48 -0.456
Chia2 = 0.74 d.f. = 2 P-value = 0.6916
B-171
DRAFT- DO NOT CITE OR QUOTE
-------
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 18.8858
BMDL = 10.3512
BMDU = 62 .8005
Taken together, (10.3512, 62.8005) is a 90 % two-sided confidence
interval for the BMD
Multistage Cancer Slope Factor = 0.00966069
B-172 DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for tongue tumors in Sprague-Dawley male rats employing
CEO in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
0.25
0.2
"G 0.15
£
<
¦I 0.1
o
CO
!
Ll_
0.05
0
0 0.002 0.004 0.006 0.008 0.01
dose
14:19 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_MALE_TONGUE_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_MALE_TONGUE_BLOOD_CEO.pit
Thu Sep 27 14:19:57 2007
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
Multistage
B-173
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.0241886
Beta(1) = 9.76289
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.7
Beta(1) -0.7 1
Parameter Estimates
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Variable
Background
Beta(1)
Estimate
0 . 0150764
11. 9992
Std. Err.
•k
•k
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-43 .4536
-43 .6998
-46 .7384
# Param's
4
2
1
Deviance Test d.f.
P-value
0 .492394
6 .56959
0.7818
0.08696
AIC:
91.3995
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000
0.0151
1.206
1
80
-0.189
0.0018
0 . 0365
1.714
2
47
0 .223
0 . 0044
0.0653
3 .133
4
48
0.506
0.0097
0.1233
5 . 918
5
48
-0.403
Chia2 = 0.50 d.f. = 2 P-value = 0.7772
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00878063
BMDL = 0.00489829
BMDU = 0.0274259
Taken together, (0.00489829, 0.0274259) is a 90 % two-sided confidence
interval for the BMD
B-174
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for tongue tumors in Sprague-Dawley male rats employing
AN in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
0.25
0.2
"G 0.15
£
<
¦I 0.1
o
CO
!
Ll_
0.05
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
dose
10:47 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_MALE_TONGUE_BLOOD_AN.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_MALE_TONGUE_BLOOD_AN.pit
Thu Sep 27 10:47:13 2007
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
Multistage
B-175
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.0289604
Beta(1) = 0.615459
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.68
Beta(1) -0.68 1
Parameter Estimates
Variable
Background
Beta(1)
Estimate
0 . 0174494
0.814352
Std. Err.
* - Indicates that this value is not calculated.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Deviance Test d.f.
Log(likelihood)
-43 .4536
-43 . 9131
-46 .7384
91.8261
# Param's
4
2
1
0 . 919024
6 .56959
P-value
0.6316
0 . 08696
Dose
Est. Prob.
Goodness of Fit
Expected Observed Size
Scaled
Residual
0.0000
0 . 0174
1.396
1
80
-0.338
0.0206
0 . 0338
1.588
2
47
0.332
0 . 0536
0.0594
2 .852
4
48
0.701
0.1460
0.1276
6 .124
5
48
-0.486
a2 = 0.95
d.f.
= 2 P-value
= 0.6210
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0.1
Extra risk
0 . 95
0.12938
0.0696872
0 .4562
Taken together, (0.0696872, 0.4562 ) is a 90
interval for the BMD
% two-sided confidence
B-176
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for tongue tumors in Sprague-Dawley female rats employing
administered dose as a dose metric
Multistage Cancer Model with 0.95 Confidence Level
Multistage Cancer
Linear extrapolation
0.4
0.35
0.3
0.25
0.2
0.15
1
0.05
0
BMD
0 5 10 15 20 25
dose
16:26 01/23 2009
Multistage Cancer Model. (Version: 1.5; Date: 02/20/2007)
Input Data File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_FEMALE_TONGUE_DW.(d)
Gnuplot Plotting File: M:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_TONGUE_DW.pit
Fri Jan 23 16:26:08 2009
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseA1-beta2 *doseA2-beta3 *doseA3) ]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 4
Total number of records with missing values = 0
Total number of parameters in model = 4
Total number of specified parameters = 0
Degree of polynomial = 3
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
B-177
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.00365935
Beta(1) = 0.00218662
Beta(2) = 0
Beta(3) = 1.46642e-005
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Background -Beta(2)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Beta(l) Beta(3)
Beta(1) 1 -0.93
Beta(3) -0.93 1
Parameter Estimates
Variable
Background
Beta(1)
Beta(2)
Beta(3)
Estimate
0
0 . 00329024
0
1.28922e-005
Std. Err.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-40.1666
-40 .2528
-55 . 0401
# Param's
4
2
1
Deviance Test d.f.
P-value
0.172251
29.7469
0.9175
<.0001
AIC:
84 .5055
B-178
DRAFT- DO NOT CITE OR QUOTE
-------
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000
0.0000
0 . 000
0
80
0 . 000
4.3600
0 . 0153
0.734
1
48
0.313
10.8000
0.0505
2 .422
2
48
-0.278
25.0000
0.2470
11.856
12
48
0 . 048
Chia2 = 0.18 d.f. = 2 P-value = 0.9151
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 15.9932
BMDL = 10.6421
BMDU = 19.9332
Taken together, (10.6421, 19.9332) is a 90 % two-sided confidence
interval for the BMD
Multistage Cancer Slope Factor = 0.00939663
B-179
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for tongue tumors in Sprague-Dawley female rats employing
CEO in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
Multistage
0.4
0.35
0.3
T3
0
0.25
<
c
o
0.2
"8
ro
i_
LL
0.15
0.05
BMD
0
0.002
0.004
0.006
0.008
0.01
dose
14:49 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_TONGUE_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_TONGUE_BLOOD_CEO.pit
Thu Sep 27 14:49:16 2007
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseA1-beta2 *doseA2-beta3 *doseA3) ]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 4
Total number of records with missing values = 0
Total number of parameters in model = 4
Total number of specified parameters = 0
Degree of polynomial = 3
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
Background = 0.0043768
Beta(1) = 2.92561
Beta(2) = 0
Beta(3) = 245868
B-180
DRAFT- DO NOT CITE OR QUOTE
-------
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Background -Beta(2)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Beta(l) Beta(3)
Beta(1) 1 -0.92
Beta(3) -0.92 1
Parameter Estimates
Variable
Background
Beta(1)
Beta(2)
Beta(3)
Estimate
0
6 .26329
0
211165
Std. Err.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
# Param's Deviance Test d.f.
Log(likelihood) # Param'
-40.1666 4
-40.2953 2
-55.0401 1
0 .257381
29.7469
P-value
0.8792
<.0001
AIC:
84 .5907
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0000 0.000 0 80 0.000
0.0021 0.0147 0.707 1 48 0.351
0.0049 0.0534 2.564 2 48 -0.362
0.0101 0.2448 11.752 12 48 0.083
Chia2 = 0.26 d.f. = 2 P-value = 0.8776
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0.1
Extra risk
0 . 95
0 . 00669671
0 . 00473574
0 . 00818609
Taken together, (0.00473574, 0.00818609) is a 90
interval for the BMD
% two-sided confidence
B-181
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for tongue tumors in Sprague-Dawley female rats employing
AN in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
Multistage
0.4
0.35
0.3
T3
0
0.25
<
c
o
0.2
"8
ro
i_
LL
0.15
0.05
BMDL
BMD
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
dose
10:57 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_TONGUE_BLOOD_AN.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_TONGUE_BLOOD_AN.pit
Thu Sep 27 10:57:41 2007
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseA1-beta2 *doseA2-beta3 *doseA3) ]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 4
Total number of records with missing values = 0
Total number of parameters in model = 4
Total number of specified parameters = 0
Degree of polynomial = 3
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
Background = 0.0034253
Beta(1) = 0.462467
B-182
DRAFT- DO NOT CITE OR QUOTE
-------
Beta(2) = 0
Beta(3) = 55.8325
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Background -Beta(2)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Beta(l) Beta(3)
Beta(1) 1 -0.93
Beta(3) -0.93 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0 * * *
Beta(1) 0.629858 * * *
Beta(2) 0 * * *
Beta(3) 49.1667 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-40.1666
-40 .2395
-55 . 0401
# Param'
4
2
1
Deviance Test d.f.
0.145677
29.7469
P-value
0 . 9298
<.0001
AIC:
84 .479
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0000 0.000 0 80 0.000
0.0237 0.0155 0.742 1 48 0.302
0.0618 0.0493 2.365 2 48 -0.244
0.1560 0.2479 11.900 12 48 0.033
Chia2 = 0.15 d.f. = 2 P-value = 0.9271
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.0966976
BMDL = 0.0610292
BMDU = 0.122886
Taken together, (0.0610292, 0.122886) is a 90 % two-sided confidence
interval for the BMD
B-183
DRAFT- DO NOT CITE OR QUOTE
-------
Tumor Site: Mammary Gland (Quast, 2002; Quast et al., 1980a)
Table B-17. Incidence of mammary gland tumors in Sprague-Dawley rats
exposed to AN in drinking water for 2 years
Sex
Administered
animal dose
(ppm in drinking
water)
Equivalent
administered
animal dose"
(mg/kg-d)
Predicted internal dose metrics
Incidence of
mammary gland
tumorsb
AN-AUC in blood
(mg/L)
CEO-AUC in blood
(mg/L)
Female
0
0
0
0
58/80 (72%)
35
4.36
2.37 x 10-2
2.07 x 10-3
42/47 (89%)°
100
10.8
6.18 x 10-2
4.87 x 10-3
42/48 (88%)°
300
25.0
1.56 x 10-1
1.01 x 10-2
35/48 (73%)
"Administered doses were averages calculated by the study authors based on animal BW and drinking water intake,
incidences for Sprague-Dawley rats do not include animals from the 6- and 12-mo sacrifices and were further
adjusted to exclude (from the denominators) rats that died between 0 and 12 mos in the study.
"Significantly different from controls (p < 0.05) as calculated by the study authors.
Table B-18. Summary of BMD modeling results based on incidence of
mammary gland tumors in Sprague-Dawley rats exposed to AN in drinking
water for 2 years
Dose metric
Best-fit model3
X2 /j-valuc'
AIC
BMD10°
BMDL10d
Females
Administered dose
l°MSe
0.25
171.81
1.22 mg/kg-d
0.66 mg/kg-d
CEO
l°MSe
0.27
171.69
5.50 x 10"4 mg/L
2.98 x 10-4 mg/L
AN
l°MSe
0.24
171.93
7.05 x 10-3 mg/L
3.77 x 10-3 mg/L
aDose-response models were fit using BMDS, version 1.4.1. "1°MS" indicates a one-stage multistage model.
bp value from the %2 goodness of fit test. Values <0.1 indicate a significant lack of fit.
°BMD10 = BMD at 10% extra risk.
dBMDL10 = 95% lower confidence limit on the BMD at 10% extra risk.
"Highest dose dropped prior to model fitting.
B-184
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for mammary gland tumors in Sprague-Dawley female rats
employing administered dose as a dose metric
Multistage Cancer Model with 0.95 Confidence Level
Multistage Cancer
Linear extrapolation
0.95
0.9
-o 0.85
0
ts
0
0.8
<
° 0.75
"8
ro
i_
LL
0.7
0.65
0.6
BMD
0
2
4
6
8
10
dose
16:34 01/23 2009
Multistage Cancer Model. (Version: 1.5; Date: 02/20/2007)
Input Data File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\SD_FEMALE_MAMMARY_DW.(d)
Gnuplot Plotting File: M:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_MAMMARY_DW.pit
Fri Jan 23 16:34:25 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 = Response
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: le-008
Parameter Convergence has been set to: le-008
B-185
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.771359
Beta(1) = 0.0674842
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.58
Beta(1) -0.58 1
Parameter Estimates
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Variable
Background
Beta(1)
Estimate
0.741898
0 . 0860351
Std. Err.
•k
•k
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
#
Log(likelihood)
-83 .2234
-83 . 9062
-86 .3815
Param's
3
2
1
Deviance Test d.f.
P-value
1.36543
6 .31609
0 .2426
0 . 04251
AIC:
171.812
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.7419 59.352 58 80 -0.345
4.3600 0.8226 39.486 42 48 0.950
10.8000 0.8981 43.108 42 48 -0.529
Chia2 = 1.30 d.f. = 1 P-value = 0.2540
B-186
DRAFT- DO NOT CITE OR QUOTE
-------
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 1.22462
BMDL = 0.659237
BMDU = 4 . 94647
Taken together, (0.659237, 4.94647) is a 90 % two-sided confidence
interval for the BMD
Multistage Cancer Slope Factor = 0.151691
B-187 DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for mammary gland tumors in Sprague-Dawley female rats
employing CEO in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
Multistage
0.95
0.9
0.8
<
c
o
0.75
"8
ro
i_
LL
0.7
0.65
0.6
BMD
0
0.001
0.002
0.003
0.004
0.005
dose
14:53 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_MAMMARY_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_MAMMARY_BLOOD_CEO.pit
Thu Sep 27 14:53:05 2007
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
B-188
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.768564
Beta(1) = 152.668
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background
Beta(1)
1
-0.58
-0.58
1
Variable
Background
Beta(1)
Parameter Estimates
Estimate
0.740346
191.729
Std. Err.
* - Indicates that this value is not calculated.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-83 .2234
-83.8471
-86 .3815
171.694
# Param's
3
2
1
Deviance Test d.f.
1.24732
6 .31609
P-value
0 .2641
0 . 04251
Goodness of Fit
Dose Est._Prob. Expected Observed Size
Scaled
Residual
0.0000
0 . 0021
0 . 0049
Chia2 = 1.1S
0.7403
0.8254
0.8979
d.f. = 1
59.228 58 80 -0.313
39.619 42 48 0.905
43.101 42 48 -0.525
P-value = 0.2748
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0.1
Extra risk
0 . 95
0 . 000549528
0.000298184
0.00214159
Taken together, (0.0002 98184, 0.00214159) is a 90
interval for the BMD
two-sided confidence
B-189
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for mammary gland tumors in Sprague-Dawley female rats
employing AN in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
Multistage
0.95
0.9
0.8
<
c
o
0.75
"8
ro
i_
LL
0.7
0.65
0.6
BMD
0
0.02
0.03
0.04
0.05
0.06
dose
11:00 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_MAMMARY_BLOOD_AN.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\SD_FEMALE_MAMMARY_BLOOD_AN.pit
Thu Sep 27 11:00:59 2007
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
B-190
DRAFT- DO NOT CITE OR QUOTE
-------
Background = 0.773999
Beta(1) = 11.5581
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.58
Beta(1) -0.58 1
Parameter Estimates
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Variable
Background
Beta(1)
Estimate
0.743458
14 . 9428
Std. Err.
•k
•k
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-83 .2234
-83.9649
-86.3815
# Param's
3
2
1
Deviance Test d.f.
P-value
1.48294
6 .31609
0 .2233
0 . 04251
AIC:
171.93
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000
0.7435
59 .477
58
80
-0.378
0 . 0237
0.8200
39.358
42
48
0 . 992
0.0618
0.8981
43 .110
42
48
-0.529
a2 = 1.41
d.f.
= 1 P-value
= 0.2354
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00705091
BMDL = 0.00376506
BMDU = 0.0295793
Taken together, (0.00376506, 0.0295793) is a 90 % two-sided confidence
interval for the BMD
B-191
DRAFT- DO NOT CITE OR QUOTE
-------
F344 Rats (Johannsen and Levinskas, 2002b; Biodynamics, 1980b)
Tumor Site: Forestomach
Table B-19. Incidence of forestomach (nonglandular) tumors in F344 rats
exposed to AN in drinking water for 2 years
Sex
Administered
animal dose
(ppm in drinking
water)
Equivalent
administered
animal dose"
(mg/kg-d)
Predicted internal dose metrics
Incidence of
forestomach
tumorsb
AN-AUC in blood
(mg/L)
CEO-AUC in blood
(mg/L)
Male
0
0
0
0
0/159(0%)
1
0.08
4.33 x 10-4
4.06 x 10-5
1/80 (1%)
3
0.25
1.35 x 10"3
1.27 x 10-4
4/78 (5%)°
10
0.83
4.52 x 10"3
4.19 x 10-4
3/80 (4%)°
30
2.48
1.37 x 10-2
1.23 x 10-3
4/80 (5%)°
100
8.37
4.85 x 10-2
3.97 x 10-3
1/77 (1%)
Female
0
0
0
0
0/157 (0%)
1
0.12
5.73 x 10-4
5.32 x 10-5
1/80 (1%)
3
0.36
1.72 x 10-3
1.59 x 10-4
2/79 (3%)
10
1.25
6.02 x 10-3
5.49 x 10-4
2/77 (3%)
30
3.65
1.79 x 10-2
1.58 x 10-3
4/80 (5%)°
100
10.90
5.63 x 10-2
4.46 x 10-3
2/75 (3%)
aAdministered doses were averages calculated by the study authors based on animal BW and drinking water intake,
incidences for F344 rats do not include animals from the 6- and 12-mo sacrifices and were further adjusted to
exclude (from the denominators) rats that died between 0 and 12 mos in the study.
"Significantly different from controls (p < 0.05) as calculated by the study authors.
B-192
DRAFT- DO NOT CITE OR QUOTE
-------
Table B-20. Summary of BMD modeling results based on incidence of
forestomach (nonglandular) tumors in F344 rats exposed to AN in drinking
water for 2 years
Dose metric
Best-fit model3
X2 /j-valuc'
AIC
BMD10°
BMDL10d
Males
Administered
dose
l°MSe
0.18
74.08
1.19 mg/kg-d
0.70 mg/kg-d
CEO
l°MSe
0.19
74.04
6.03 x 10"4 mg/L
3.55 x 10"4 mg/L
AN
l°MSe
0.18
74.12
6.48 x 10"3 mg/L
3.81 x 10"3 mg/L
Females
Administered
dose
l°MSf
0.83
96.64
8.43 mg/kg-d
3.89 mg/kg-d
CEO
l°MSf
0.83
96.62
3.65 x 10-3 mg/L
1.69 x 10-3 mg/L
AN
l°MSf
0.82
96.65
4.13 x 10-2 mg/L
1.90 x 10"2 mg/L
aDose-response models were fit using BMDS, version 1.4.1. "1°MS" indicates a one-stage multistage model.
bp value from the %2 goodness of fit test. Values <0.1 indicate a significant lack of fit.
°BMD10 = BMD at 10% extra risk.
dBMDL10 = 95% lower confidence limit on the BMD at 10% extra risk.
eTwo highest doses dropped prior to model fitting.
fHighest dose dropped prior to model fitting.
B-193
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for forestomach tumors in F344 male rats employing
administered dose as a dose metric
Multistage Cancer Model with 0.95 Confidence Level
Multistage Cancer
Linear extrapolation
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
BMD
PMp
1.2
0
0.2
0.4
0.6
0.8
1
dose
14:29 01/26 2009
Multistage Cancer Model. (Version: 1.5; Date: 02/20/2007)
Input Data File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\F344_MALE_FORESTOMACH_DW.(d)
Gnuplot Plotting File: M:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F3 4 4_MALE_FORESTOMACH_DW.pit
Mon Jan 26 14:29:05 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 = Response
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: le-008
Parameter Convergence has been set to: le-008
B-194
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.0148134
Beta(1) = 0.0377128
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Beta(1)
Beta(1) 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0 * * *
Beta(1) 0.0883999 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-33 . 9463
-36 . 0377
-39.1548
# Param'
4
1
1
Deviance Test d.f.
P-value
4 .18282
10 .417
0 .2424
0 . 01533
AIC: 74.0755
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0000 0.000 0 159 0.000
0.0800 0.0070 0.564 1 80 0.583
0.2500 0.0219 1.705 4 78 1.777
0.8300 0.0707 5.660 3 80 -1.160
Chia2 = 4.84 d.f. = 3 P-value = 0.1836
B-195
DRAFT- DO NOT CITE OR QUOTE
-------
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 1.19186
BMDL = 0.701326
BMDU = 3 .81474
Taken together, (0.701326, 3.81474) is a 90 % two-sided confidence
interval for the BMD
Multistage Cancer Slope Factor = 0.142587
B-196 DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for forestomach tumors in F344 male rats employing CEO in
blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
Multistage
0.14
0.12
0.08
0.06
0.04
0.02
BMDL
BMP
0 0.0001 0.0002 0.0003 0.0004 0.0005 0.0006
dose
14:57 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F344_MALE_FORESTOMACH_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F344_MALE_FORESTOMACH_BLOOD_CEO.pit
Thu Sep 27 14:57:13 2007
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
B-197
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.0147622
Beta(1) = 74.9311
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Beta(1)
Beta(1) 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0 * * *
Beta(1) 174.815 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-33 . 9463
-36 . 0209
-39.1548
# Param'
4
1
1
Deviance Test d.f.
P-value
4 .14909
10 .417
0 .2458
0 . 01533
AIC: 74.0417
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0000 0.000 0 159 0.000
0.0000 0.0071 0.566 1 80 0.579
0.0001 0.0220 1.713 4 78 1.767
0.0004 0.0706 5.650 3 80 -1.157
Chia2 = 4.80 d.f. = 3 P-value = 0.1873
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.000602696
BMDL = 0.000354643
BMDU = 0.00191036
Taken together, (0.000354643, 0.00191036) is a 90 % two-sided confidence
interval for the BMD
B-198
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for forestomach tumors in F344 male rats employing AN in
blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
0.14
0.12
0.1
T3
0
© 0.08
c
o
"8
0.06
0.04
0.02
0
Multistage
BMDl]
0.001 0.002 0.003 0.004 0.005
dose
BMP
0.006
11:04 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F344_MALE_FORESTOMACH_BLOOD_AN.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F3 4 4_MALE_FORESTOMACH_BLOOD_AN.pit
Thu Sep 27 11:04:04 2007
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
B-199
DRAFT- DO NOT CITE OR QUOTE
-------
Background = 0.0148807
Beta(1) = 6.89727
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Beta(1)
Beta(1) 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0 * * *
Beta(1) 16.2684 * * *
* - Indicates that this value is not calculated.
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
# Param's Deviance Test d.f.
Log(likelihood) # Param'
-33.9463 4
-36.0601 1
-39.1548 1
P-value
4 .22747
10 .417
0.2379
0 . 01533
AIC:
74.1201
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0000 0.000 0 159 0.000
0.0004 0.0070 0.562 1 80 0.587
0.0014 0.0217 1.694 4 78 1.791
0.0045 0.0709 5.672 3 80 -1.164
Chia2 = 4.91 d.f. = 3 P-value = 0.1788
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00647638
BMDL = 0.00381089
BMDU = 0.0209986
Taken together, (0.00381089, 0.0209986) is a 90 % two-sided confidence
interval for the BMD
B-200
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for forestomach tumors in F344 female rats employing
administered dose as a dose metric
Multistage Cancer Model with 0.95 Confidence Level
0.14
0.12
0.1
TD
CD
I 0.08
it
<
o 0.06
"8
ro
lH 0.04
0.02
0
012345678
dose
14:49 01/26 2009
Multistage Cancer Model. (Version: 1.5; Date: 02/20/2007)
Input Data File: M:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F344_FEMALE_FORESTOMACH_DW.(d)
Gnuplot Plotting File: M:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F3 4 4_FEMALE_FORESTOMACH_DW.pit
Mon Jan 26 14:49:27 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 = Response
Independent variable = Dose
Total number of observations = 5
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: le-008
Parameter Convergence has been set to: le-008
Multistage Cancer
Linear extrapolation
BMDL
BMI
B-201
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.0127929
Beta(1) = 0.0107517
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.58
Beta(1) -0.58 1
Parameter Estimates
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Variable
Background
Beta(1)
Estimate
0 . 0101696
0 . 0125023
Std. Err.
•k
•k
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-45 . 9122
-46 .3178
-48 .4586
# Param'
5
2
1
Deviance Test d.f.
0.81118
5.09287
P-value
0.846S
0.2119
AIC:
96.6356
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000
0 . 0102
1.597
1
157
-0.475
0.1200
0.0117
0 . 932
1
80
0 . 071
0.3600
0 . 0146
1.155
2
79
0.793
1.2500
0.0255
1. 965
2
77
0 . 025
3.6500
0 . 0543
4 .346
4
80
-0.171
Chia2 = 0.89 d.f. = 3 P-value = 0.8283
B-202
DRAFT- DO NOT CITE OR QUOTE
-------
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 8.42727
BMDL = 3 .88972
BMDU = 48.9918
Taken together, (3.88972, 48.9918) is a 90 % two-sided confidence
interval for the BMD
Multistage Cancer Slope Factor = 0.0257088
B-203 DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for forestomach tumors in F344 female rats employing CEO
in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
0.14
Multistage
0.12
T3
0
0.08
<
§ 0.06
"8
LL
0.04
0.02
BMD
0
0.0005 0.001
0.0015 0.002 0.0025 0.003 0.0035
dose
15:07 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F344_FEMALE_FORESTOMACH_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F344_FEMALE_FORESTOMACH_BLOOD_CEO.pit
Thu Sep 27 15:07:15 2007
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 = Response
Independent variable = Dose
Total number of observations = 5
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: le-008
Parameter Convergence has been set to: le-008
B-204
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.0127198
Beta(1) = 24.8654
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.58
Beta(1) -0.58 1
Parameter Estimates
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Variable
Background
Beta(1)
Estimate
0 . 0101112
28.8833
Std. Err.
•k
•k
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-45 . 9122
-46 .3121
-48 .4586
# Param's
5
2
1
Deviance Test d.f.
P-value
0.799693
5 . 09287
0.8495
0 .2779
AIC:
96 .6241
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000
0 . 0101
1.587
1
157
-0.469
0.0001
0.0116
0 . 930
1
80
0 . 072
0.0002
0 . 0146
1.157
2
79
0.789
0.0005
0.0257
1. 978
2
77
0 . 016
0.0016
0 . 0543
4 .342
4
80
-0.169
Chia2 = 0.88 d.f. = 3 P-value = 0.8310
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.0036478
BMDL = 0.00168696
BMDU = 1361.26
Taken together, (0.00168696, 1361.26) is a 90 % two-sided confidence
interval for the BMD
B-205
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for forestomach tumors in F344 female rats employing AN in
blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
0.14
Multistage
0.12
T3
0
0.08
<
§ 0.06
"8
LL
0.04
0.02
BMDL
BMD
0
0.005 0.01
0.015 0.02 0.025 0.03 0.035 0.04
dose
11:30 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F344_FEMALE_FORESTOMACH_BLOOD_AN.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F344_FEMALE_FORESTOMACH_BLOOD_AN.pit
Thu Sep 27 11:30:39 2007
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 = Response
Independent variable = Dose
Total number of observations = 5
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: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
B-206
DRAFT- DO NOT CITE OR QUOTE
-------
Background = 0.0128833
Beta(1) = 2.18924
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.58
Beta(1) -0.58 1
Parameter Estimates
Variable
Background
Beta(1)
Estimate
0 . 0102427
2 .54885
Std. Err.
* - Indicates that this value is not calculated.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Deviance Test d.f.
Log(likelihood)
-45 . 9122
-46 .3252
-48 .4586
96 .6505
# Param's
5
2
1
0.826038
5 . 09287
P-value
0.8432
0 .2779
Dose
Est. Prob.
Goodness of Fit
Expected Observed Size
Scaled
Residual
0.0000
0.0102
1.608
1
157
-0 .482
0.0006
0.0117
0 . 935
1
80
0 . 068
0.0017
0 . 0146
1.151
2
79
0.797
0.0060
0.0253
1. 949
2
77
0 . 037
0.0179
0 . 0544
4 .351
4
80
-0.173
A2 = 0.
90 d.f.
= 3 P-value
= 0.8246
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0.1
Extra risk
0 . 95
0.0413364
0 . 0190324
0 .242436
Taken together, (0.0190324, 0.242436) is a 90
interval for the BMD
% two-sided confidence
B-207
DRAFT- DO NOT CITE OR QUOTE
-------
Tumor Site: CNS (Johannsen and Levinskas, 2002b; Biodynamics, 1980b)
Table B-21. Incidence of CNS tumors in F344 rats exposed to AN in
drinking water for 2 years
Sex
Administered
animal dose
(ppm in drinking
water)
Equivalent
administered
animal dose"
(mg/kg-d)
Predicted internal dose metrics
Incidence of CNS
tumorsb
AN-AUC in blood
(mg/L)
CEO-AUC in blood
(mg/L)
Male
0
0
0
0
0/160 (0%)
1
0.08
4.33 x 10-4
4.06 x 10-5
2/80 (3%)
3
0.25
1.35 x 10-3
1.27 x 10-4
1/78 (1%)
10
0.83
4.52 x 10-3
4.19 x 10-4
2/80 (3%)
30
2.48
1.37 x 10-2
1.23 x 10-3
10/79 (13%)c
100
8.37
4.85 x 10-2
3.97 x 10-3
25/76 (33%)°
Female
0
0
0
0
1/157 (1%)
1
0.12
5.73 x 10-4
5.32 x 10-5
1/80 (1%)
3
0.36
1.72 x 10-3
1.59 x 10-4
2/80 (3%)
10
1.25
6.02 x 10-3
5.49 x 10-4
5/75 (7%)
30
3.65
1.79 x 10-2
1.58 x 10-3
6/80 (8%)°
100
10.90
5.63 x 10-2
4.46 x 10-3
24/76 (32%)°
aAdministered doses were averages calculated by the study authors based on animal BW and drinking water intake,
incidences for F344 rats do not include animals from the 6- and 12-mo sacrifices and were further adjusted to
exclude (from the denominators) rats that died between 0 and 12 mos in the study.
"Significantly different from controls (p < 0.01) as calculated by the study authors.
Table B-22. Summary of BMD modeling results based on incidence of CNS
tumors in F344 rats exposed to AN in drinking water for 2 years
Dose metric
Best-fit model3
X2 /J-valuc'
AIC
BMD10C
BMDL10d
Males
Administered dose
1°MS
0.73
240.57
2.41 mg/kg-d
1.81 mg/kg-d
CEO
1°MS
0.70
240.75
1.16 x 10-3 mg/L
8.74 x 10"4 mg/L
AN
1°MS
0.74
240.44
1.37 x 10"2 mg/L
1.03 x 10"2 mg/L
Females
Administered dose
1°MS
0.68
222.13
3.34 mg/kg-d
2.52 mg/kg-d
CEO
1°MS
0.66
222.30
1.39 x 10-3 mg/L
1.05 x 10-3 mg/L
AN
1°MS
0.68
222.04
1.70 x 10"2 mg/L
1.28 x 10"2 mg/L
aDose-response models were fit using BMDS, version 1.4.1. "1°MS" indicates a one-stage multistage model.
bp value from the %2 goodness of fit test. Values <0.1 indicate a significant lack of fit.
°BMDio = BMD at 10% extra risk.
dBMDLio = 95% lower confidence limit on the BMD at 10% extra risk.
B-208
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for CNS tumors in F344 male rats employing administered
dose as a dose metric
Multistage Cancer Model with 0.95 Confidence Level
Multistage Cancer
Linear extrapolation
0.4
0.3
0.2
BMDL
BMD
012345678
dose
14:34 01/26 2009
Multistage Cancer Model. (Version: 1.5; Date: 02/20/2007)
Input Data File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\F344_MALE_CNS_DW.(d)
Gnuplot Plotting File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\F344_MALE_CNS_DW.pit
Mon Jan 26 14:34:10 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 = Response
Independent variable = Dose
Total number of observations = 6
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: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
B-209
DRAFT- DO NOT CITE OR QUOTE
-------
Background = 0.00946328
Beta(1) = 0.0466
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.47
Beta(1) -0.47 1
Parameter Estimates
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Variable
Background
Beta(1)
Estimate
0 . 0149509
0 . 0437933
Std. Err.
•k
•k
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-117.105
-118.287
-151.112
# Param's
6
2
1
Deviance Test d.f.
P-value
2 .36443
68 . 0145
0.6691
<.0001
AIC:
240.574
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000
0 . 0150
2 .392
3
160
0.396
0.0800
0.0184
1.472
2
80
0 .440
0.2500
0.0257
2 . 003
1
78
-0.718
0.8300
0 . 0501
4 . 009
2
80
-1. 030
2.4800
0.1163
9.190
10
79
0 .284
8.3700
0.3172
24 .110
25
76
0.219
a2 = 2.05
d.f.
= 4 P-value
= 0.7258
B-210
DRAFT- DO NOT CITE OR QUOTE
-------
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 2 .40586
BMDL = 1.8134
BMDU = 3 .32187
Taken together, (1.8134 , 3.32187) is a 90 % two-sided confidence
interval for the BMD
Multistage Cancer Slope Factor = 0.0551449
B-211 DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for CNS tumors in F344 male rats employing CEO in blood as
an internal dose metric
Multistage Model with 0.95 Confidence Level
Multistage
0.4
ffl 0.3
<
o 0.2
"8
!
Ll_
BMDL
BMD
0
0.0005 0.001 0.0015 0.002 0.0025 0.003 0.0035 0.004
dose
15:00 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\F344_MALE_CNS_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F3 4 4_MALE_CNS_BLOOD_CEO.pit
Thu Sep 27 15:00:12 2007
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 = Response
Independent variable = Dose
Total number of observations = 6
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: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
B-212
DRAFT- DO NOT CITE OR QUOTE
-------
Background = 0.00793445
Beta(1) = 98.3167
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.47
Beta(1) -0.47 1
Parameter Estimates
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Variable
Background
Beta(1)
Estimate
0 . 0147547
90 . 9591
Std. Err.
•k
•k
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-117.105
-118.375
-151.112
# Param's
6
2
1
Deviance Test d.f.
P-value
2 .54113
68 . 0145
0.6373
<.0001
AIC:
240.75
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000
0 . 0148
2 .361
3
160
0.419
0.0000
0.0184
1.471
2
80
0 .440
0.0001
0 . 0261
2 . 034
1
78
-0.734
0.0004
0 . 0516
4 .128
2
80
-1. 075
0.0012
0.1190
9 .404
10
79
0 .207
0 . 0040
0.3134
23 .817
25
76
0 .293
a2 = 2.19
d.f.
= 4 P-value
= 0.7002
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00115833
BMDL = 0.000873641
BMDU = 0.00159795
Taken together, (0.000873641, 0.00159795) is a 90 % two-sided confidence
interval for the BMD
B-213
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for CNS tumors in F344 male rats employing AN in blood as
an internal dose metric
Multistage Model with 0.95 Confidence Level
TD
0
"8
c
o
"8
0.4
0.3
0.2
0.1
Multistage
^ 1
<
BMDlJ
BMD
0.01
0.02
0.03
0.04
0.05
dose
11:07 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\F344_MALE_CNS_BLOOD_AN.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F3 4 4_MALE_CNS_BLOOD_AN.pit
Thu Sep 27 11:07:15 2007
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 = Response
Independent variable = Dose
Total number of observations = 6
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: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
B-214
DRAFT- DO NOT CITE OR QUOTE
-------
Background = 0.0109984
Beta(1) = 8.0341
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.46
Beta(1) -0.46 1
Parameter Estimates
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Variable
Background
Beta(1)
Estimate
0 . 0151616
7.67068
Std. Err.
•k
•k
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-117.105
-118.218
-151.112
# Param's
6
2
1
Deviance Test d.f.
P-value
2 .22757
68 . 0145
0.694
<.0001
AIC:
240 .437
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000
0.0152
2 .426
3
160
0.371
0.0004
0.0184
1.474
2
80
0.437
0.0014
0.0253
1. 974
1
78
-0.702
0.0045
0 . 0487
3 .898
2
80
-0 . 986
0.0137
0.1134
8 . 959
10
79
0.369
0.0485
0.3211
24 .405
25
76
0.146
a2 = 1.95
d.f.
= 4 P-value
= 0.7447
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.0137355
BMDL = 0.0103458
BMDU = 0.0189838
Taken together, (0.0103458, 0.0189838) is a 90 % two-sided confidence
interval for the BMD
B-215
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for CNS tumors in F344 female rats employing administered
dose as a dose metric
Multistage Cancer Model with 0.95 Confidence Level
0.4
Multistage Cancer
Linear extrapolation
T3
0
ts
0
<
c
o
"8
0.3
0.2
0.1
10
dose
14:55 01/26 2009
Multistage Cancer Model. (Version: 1.5; Date: 02/20/2007)
Input Data File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\F344_FEMALE_CNS_DW.(d)
Gnuplot Plotting File: M:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F3 4 4_FEMALE_CNS_DW.pit
Mon Jan 26 14:55:28 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 = Response
Independent variable = Dose
Total number of observations = 6
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: le-008
Parameter Convergence has been set to: le-008
B-216
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.00525608
Beta(1) = 0.0331149
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.49
Beta(1) -0.49 1
Parameter Estimates
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Variable
Background
Beta(1)
Estimate
0 . 00823648
0.0315054
Std. Err.
•k
•k
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-107.86
-109.065
-140.644
# Param's
6
2
1
Deviance Test d.f.
P-value
2 .41059
65.5686
0.6607
<.0001
AIC:
222 .13
Goodness of Fit
Scaled
Dose
Est. Prob.
Expected
Observed
Size
Residual
0.0000
0.0082
1.293
1
157
-0.259
0.1200
0 . 0120
0 . 958
1
80
0 . 043
0.3600
0.0194
1.554
2
80
0.362
1.2500
0 . 0465
3.490
5
75
0.828
3.6500
0.1160
9 .278
6
80
-1.144
10 . 9000
0 .2965
22 .534
24
76
0.368
Chia2 = 2.33 d.f. = 4 P-value = 0.6753
B-217
DRAFT- DO NOT CITE OR QUOTE
-------
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 3 .34421
BMDL = 2 .51554
BMDU = 4 .64166
Taken together, (2.51554, 4.64166) is a 90 % two-sided confidence
interval for the BMD
Multistage Cancer Slope Factor = 0.0397529
B-218 DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for CNS tumors in F344 female rats employing CEO in blood
as an internal dose metric
Multistage Model with 0.95 Confidence Level
Multistage
0.4
^ 0.3
<
§ 0.2
"8
LL
BMDU [BMP :
0.0005 0.001 0.0015 0.002 0.0025 0.003 0.0035 0.004 0.0045
0
dose
15:10 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F344_FEMALE_CNS_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F3 4 4_FEMALE_CNS_BLOOD_CEO.pit
Thu Sep 27 15:10:31 2007
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 = Response
Independent variable = Dose
Total number of observations = 6
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: le-008
Parameter Convergence has been set to: le-008
B-219
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.00362563
Beta(1) = 80.7118
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.5
Beta(1) -0.5 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0.00790184 * * *
Beta(1) 75.5979 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-107.86
-109.148
-140.644
# Param's
6
2
1
Deviance Test d.f.
P-value
2 .57752
65.5686
0.6308
< . 0001
AIC:
222 .297
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000
0.0079
1.241
1
157
-0.217
0.0001
0.0119
0 . 951
1
80
0 . 051
0.0002
0.0198
1.580
2
80
0.337
0.0005
0 . 0482
3 .618
5
75
0.745
0.0016
0.1196
9.568
6
80
-1.229
0.0045
0.2919
22.181
24
76
0.459
a2 = 2.44
d.f.
= 4 P-value
= 0.6554
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.0013937
BMDL = 0.00104985
BMDU = 0.00193063
Taken together, (0.00104985, 0.00193063) is a 90 % two-sided confidence
interval for the BMD
B-220
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for CNS tumors in F344 female rats employing AN in blood
as an internal dose metric
Multistage Model with 0.95 Confidence Level
Multistage
0.4
0.3
0.2
0
BIVIDlJ BMP
0.01 0.02
0
0.03
0.04
0.05
dose
11:34 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\F344_FEMALE_CNS_BLOOD_AN.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F3 4 4_FEMALE_CNS_BLOOD_AN.pit
Thu Sep 27 11:34:20 2007
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 = Response
Independent variable = Dose
Total number of observations = 6
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: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
B-221
DRAFT- DO NOT CITE OR QUOTE
-------
Background = 0.00679319
Beta(1) = 6.4212
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.49
Beta(1) -0.49 1
Parameter Estimates
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Variable
Background
Beta(1)
Estimate
0 . 00859669
6 .19715
Std. Err.
•k
•k
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-107.86
-109.017
-140.644
# Param's
6
2
1
Deviance Test d.f.
P-value
2 .31525
65.5686
0.678
<.0001
AIC:
222 . 035
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000
0.0086
1.350
1
157
-0.302
0.0006
0 . 0121
0 . 969
1
80
0 . 032
0.0017
0.0191
1.529
2
80
0.385
0.0060
0 . 0449
3 .368
5
75
0 . 910
0.0179
0.1127
9 . 015
6
80
-1. 066
0.0563
0.3006
22 .846
24
76
0.289
a2 = 2.29
d.f.
= 4 P-value
= 0.6828
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.0170014
BMDL = 0.0127694
BMDU = 0.0236456
Taken together, (0.0127694, 0.0236456) is a 90 % two-sided confidence
interval for the BMD
B-222
DRAFT- DO NOT CITE OR QUOTE
-------
Tumor Site: Zymbal's Gland (Johannsen and Levinskas, 2002b; Biodynamics, 1980b)
Table B-23. Incidence of Zymbal gland tumors in F344 rats exposed to AN
in drinking water for 2 years
Sex
Administered
animal dose
(ppm in drinking
water)
Equivalent
administered
animal dose"
(mg/kg-d)
Predicted internal dose metrics
Incidence of
Zymbal's gland
tumorsb
AN-AUC in blood
(mg/L)
CEO-AUC in blood
(mg/L)
Male
0
0
0
0
1/147 (1%)
1
0.08
4.33 x 10-4
4.06 x 10-5
1/76 (1%)
3
0.25
1.35 x 10-3
1.27 x 10-4
0/73 (0%)
10
0.83
4.52 x 10-3
4.19 x 10-4
0/67 (0%)
30
2.48
1.37 x 10-2
1.23 x 10-3
2/71 (3%)°
100
8.37
4.85 x 10-2
3.97 x 10-3
14/68 (21 %)°
Female
0
0
0
0
0/157 (0%)
1
0.12
5.73 x 10-4
5.32 x 10-5
0/73 (0%)
3
0.36
1.72 x 10-3
1.59 x 10-4
0/73 (0%)
10
1.25
6.02 x 10-3
5.49 x 10-4
0/70 (0%)
30
3.65
1.79 x 10-2
1.58 x 10-3
2/73 (3%)°
100
10.90
5.63 x 10-2
4.46 x 10-3
8/62 (13%)°
"Administered doses were averages calculated by the study authors based on animal BW and drinking water intake,
incidences for F344 rats do not include animals from the 6- and 12-mo sacrifices and were further adjusted to
exclude (from the denominators) rats that died between 0 and 12 mos in the study.
"Significantly different from controls (p < 0.01) as calculated by the study authors.
Table B-24. Summary of BMD modeling results based on incidence of
Zymbal gland tumors in F344 rats exposed to AN in drinking water for
2 years
Dose metric
Best-fit model3
X2 /j-valuc'
AIC
BMD10°
BMDL10d
Males
Administered dose
2°MS
0.78
116.51
5.73 mg/kg-d
4.55 mg/kg-d
CEO
2°MS
0.77
116.53
2.73 x 10-3 mg/L
2.19 x 10-3 mg/L
AN
2°MS
0.78
116.52
3.31 x 10"2 mg/L
2.59 x 10"2 mg/L
Females
Administered dose
2°MS
0.95
70.80
9.16 mg/kg-d
7.17 mg/kg-d
CEO
2°MS
0.99
68.68
3.78 x 10-3 mg/L
2.97 x 10-3 mg/L
AN
1°MS
0.89
70.82
5.41 x 10"2 mg/L
3.35 x 10"2 mg/L
aDose-response models were fit using BMDS, version 1.4.1. "1°MS" indicates a one-stage multistage model and
"2°MS" indicates a two-stage multistage model.
bp value from the %2 goodness of fit test. Values <0.1 indicate a significant lack of fit.
°BMD10 = BMD at 10% extra risk.
dBMDL10 = 95% lower confidence limit on the BMD at 10% extra risk.
B-223
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for Zymbal gland tumors in F344 male rats employing
administered dose as a dose metric
Multistage Cancer Model with 0.95 Confidence Level
0.35
0.3
0.25
TD
CD
I 0.2
it
<
o 015
"8
ro
ul 0.1
0.05
0
012345678
dose
14:41 01/26 2009
Multistage Cancer Model. (Version: 1.5; Date: 02/20/2007)
Input Data File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\F344_MALE_ZYMBAL_DW.(d)
Gnuplot Plotting File: M:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F3 4 4_MALE_ZYMBAL_DW.pit
Mon Jan 26 14:41:15 2009
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2) ]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 6
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Multistage Cancer
Linear extrapolation
B-224
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.00522088
Beta(l) = 0
Beta(2) = 0.00321913
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Beta(2)
Background 1 -0.37
Beta(2) -0.37 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0.00563419 * * *
Beta(1) 0 * * *
Beta(2) 0.00321262 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood) # Param's Deviance Test d.f.
-54.9964 6
-56.2568 2 2.52093 4
-77.5815 1 45.1702 5
P-value
0.640S
<.0001
AIC:
116 .514
Goodness of Fit
Dose
Est. Prob.
Expected
Observed
Size
Scaled
Residual
0.0000
0.0056
0.828
1
147
0.189
0.0800
0.0057
0.430
1
76
0.872
0.2500
0.0058
0 .426
0
73
-0.654
0.8300
0.0078
0.525
0
67
-0.727
2.4800
0.0251
1.781
2
71
0.166
8.3700
0 .2060
14 . 010
14
68
-0 . 003
a2 = 1.78
d.f.
= 4 P-value
= 0.7758
B-225
DRAFT- DO NOT CITE OR QUOTE
-------
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 5.72676
BMDL = 4 .54678
BMDU = 7.26717
Taken together, (4.54678, 7.26717) is a 90 % two-sided confidence
interval for the BMD
Multistage Cancer Slope Factor = 0.0219936
B-226 DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for Zymbal gland tumors in F344 male rats employing CEO
in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
0.35
Multistage
0.3
0.25
T3
0
0.2
<
§ 0.15
"8
ro
i_
LL
0.05
BMD
0
0.0005 0.001 0.0015 0.002 0.0025 0.003 0.0035 0.004
dose
15:02 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F344_MALE_ZYMBAL_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F3 4 4_MALE_ZYMBAL_BLOOD_CEO.pit
Thu Sep 27 15:02:53 2007
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2) ]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 6
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
B-227
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.00480523
Beta(l) = 0
Beta(2) = 14327.9
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Beta(2)
Background 1 -0.38
Beta(2) -0.38 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0.00556639 * * *
Beta(1) 0 * * *
Beta(2) 14170.3 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-54 . 9964
-56 .2667
-77.5815
# Param's
6
2
1
Deviance Test d.f.
P-value
2 .54061
45.1702
0.6374
<.0001
AIC: 116.533
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000
0.0056
0.818
1
147
0 .201
0.0000
0.0056
0 .425
1
76
0.885
0.0001
0.0058
0 .423
0
73
-0.652
0.0004
0.0080
0.538
0
67
-0.737
0.0012
0 . 0267
1.893
2
71
0 . 079
0 . 0040
0 .2046
13.913
14
68
0 . 026
A2 = 1.
80 d.f.
= 4 P-value
= 0.7727
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00272678
BMDL = 0.00218772
BMDU = 0.00345798
Taken together, (0.00218772, 0.00345798) is a 90 % two-sided confidence
interval for the BMD
B-228
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for Zymbal gland tumors in F344 male rats employing AN in
blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
0.35
Multistage
0.3
0.25
T3
0
0.2
<
§ 0.15
"8
ro
i_
LL
0.05
bmdlJ
BMD
0
0.01
0.02
0.03
0.04
0.05
dose
11:27 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F344_MALE_ZYMBAL_BLOOD_AN.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F3 4 4_MALE_ZYMBAL_BLOOD_AN.pit
Thu Sep 27 11:27:02 2007
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2) ]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 6
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
Background = 0.00503997
Beta(1) = 0.191539
B-229
DRAFT- DO NOT CITE OR QUOTE
-------
Beta(2) =
91. 9968
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Beta(2)
Background 1 -0.37
Beta(2) -0.37 1
Parameter Estimates
Variable
Background
Beta(1)
Beta(2)
Estimate
0 . 00571347
0
96 .3581
Std. Err.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Model
Full model
Fitted model
Analysis of Deviance Table
Deviance Test d.f.
Log(likelihood)
-54 . 9964
-56 .2582
# Param's
6
2
2 .52365
P-value
0.6404
Reduced model
•77.5815
1 45.
1702 5
<
AIC:
116.516
Goodness of Fit
Scaled
Dose
Est. Prob.
Expected
Observed
Size
Residual
0.0000
0.0057
0.840
1
147
0.175
0.0004
0.0057
0.436
1
76
0.858
0.0014
0.0059
0.430
0
73
-0.658
0 . 0045
0.0077
0.514
0
67
-0.720
0 . 0137
0 . 0235
1.671
2
71
0 .258
0.0485
0.2074
14 .101
14
68
-0 . 030
ChiA2 = 1
.78 d.f.
= 4 P-
¦value = 0.7755
Benchmark Dose Computation
Specified <
effect =
0.1
Risk Type
=
Extra risk
Confidence
level =
0 . 95
BMD =
0.033067
BMDL =
0.0258907
BMDU =
0.0419928
Taken together, (0.0258$
)07, 0 . 041992£
J) is a 90
% two-sided
conf idenc<
interval for the BMD
B-230
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for Zymbal gland tumors in F344 female rats employing
administered dose as a dose metric
Multistage Cancer Model with 0.95 Confidence Level
Multistage Cancer
Linear extrapolation
0.25
0.2
0.15
0.05
bmdlJ
BMD
0 2 4 6 8 10
dose
15:01 01/26 2009
Multistage Cancer Model. (Version: 1.5; Date: 02/20/2007)
Input Data File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\F344_FEMALE_ZYMBAL_DW.(d)
Gnuplot Plotting File: M:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F3 4 4_FEMALE_ZYMBAL_DW.pit
Mon Jan 26 15:01:22 2009
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2) ]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 6
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
B-231
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0
Beta(1) = 0.00476273
Beta(2) = 0.000742952
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Beta(l) Beta(2)
Beta(1) 1 -0.95
Beta(2) -0.95 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0 * * *
Beta(1) 0.000111429 * * *
Beta(2) 0.00124232 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-33.0086
-33 .4019
-49.1799
# Param's
6
2
1
Deviance Test d.f.
0.786523
32 .3425
P-value
0 . 9402
<.0001
AIC:
70.8038
Dose
Est. Prob.
Goodness of Fit
Expected
Observed
Size
Scaled
Residual
0.0000
0.0000
0 . 000
0
157
0 . 000
0.1200
0.0000
0 . 002
0
73
-0 . 048
0.3600
0.0002
0 . 015
0
73
-0.121
1.2500
0 . 0021
0.145
0
70
-0.382
3.6500
0 . 0168
1.227
2
73
0.703
10 . 9000
0.1383
8.573
8
62
-0.211
Chi 2 = 0.70
d.f. = 4
P-value = 0.9511
B-232
DRAFT- DO NOT CITE OR QUOTE
-------
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 9.16446
BMDL = 7.17181
BMDU = 13.5889
Taken together, (7.17181, 13.5889) is a 90 % two-sided confidence
interval for the BMD
Multistage Cancer Slope Factor = 0.0139435
B-233 DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for Zymbal gland tumors in F344 female rats employing CEO
in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
0.25
0.2
TD
CD
^ 0.15
<
% 0.1
ro
l_L
0.05
0
0 0.0005 0.001 0.0015 0.002 0.0025 0.003 0.0035 0.004 0.0045
dose
15:13 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F344_FEMALE_ZYMBAL_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F3 4 4_FEMALE_ZYMBAL_BLOOD_CEO.pit
Thu Sep 27 15:13:27 2007
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2) ]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 6
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Multistage
BMDL
BMP
B-234
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0
Beta(1) = 9.4165
Beta(2) = 4928.95
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Background -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Beta(2)
Beta(2) 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0 * * *
Beta(1) 0 * * *
Beta(2) 7382.1 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-33.0086
-33 .3423
-49.1799
# Param'
6
1
1
Deviance Test d.f.
0.667255
32 .3425
P-value
0.9847
< . 0001
AIC: 68.6845
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000
0.0000
0 . 000
0
157
0 . 000
0.0001
0.0000
0 . 002
0
73
-0 . 039
0.0002
0.0002
0 . 014
0
73
-0.117
0.0005
0 . 0022
0.156
0
70
-0.395
0.0016
0 . 0183
1.333
2
73
0.583
0.0045
0.1366
8.467
8
62
-0.173
Chia2 = 0.54 d.f. = 5 P-value = 0.9905
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00377789
BMDL = 0.00297198
BMDU = 0.00541558
Taken together, (0.00297198, 0.00541558) is a 90 % two-sided confidence
interval for the BMD
B-235
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for Zymbal gland tumors in F344 female rats employing AN
in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
Multistage
0.25
0.2
T3
0
0.15
<
c
o
"8
CO
i_
LL
0.05
BMDL
BMD
0
0.01
0.02
0.03
0.04
0.05
dose
11:37 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F344_FEMALE_ZYMBAL_BLOOD_AN.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F3 4 4_FEMALE_ZYMBAL_BLOOD_AN.pit
Thu Sep 27 11:37:11 2007
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 = Response
Independent variable = Dose
Total number of observations = 6
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: le-008
Parameter Convergence has been set to: le-008
B-236
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0
Beta(1) = 2.5025
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Beta(1)
Beta(1) 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0 * * *
Beta(1) 1.94599 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-33.0086
-34 .4088
-49.1799
# Param'
6
1
1
Deviance Test d.f.
P-value
2 .80031
32 .3425
0.7307
<.0001
AIC: 70.8176
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000
0.0000
0 . 000
0
157
0 . 000
0.0006
0.0011
0 . 081
0
73
-0.285
0.0017
0 . 0033
0 .244
0
73
-0.495
0.0060
0.0116
0.815
0
70
-0 . 908
0.0179
0 . 0342
2.499
2
73
-0.321
0.0563
0.1038
6 .434
8
62
0.652
Chia2 = 1.68 d.f. = 5 P-value = 0.8915
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.0541423
BMDL = 0.0335374
BMDU = 0.0957005
Taken together, (0.0335374, 0.0957005) is a 90 % two-sided confidence
interval for the BMD
B-237
DRAFT- DO NOT CITE OR QUOTE
-------
Tumor Site: Mammary Gland (Johannsen and Levinskas, 2002b; Biodynamics, 1980b)
Table B-25. Incidence of mammary gland tumors in F344 rats exposed to
AN in drinking water for 2 years
Administered
Equivalent
Predicted internal dose metrics
Incidence of
animal dose
administered
(ppm in drinking
animal dose"
AN-AUC in blood
CEO-AUC in blood
mammary gland
Sex
water)
(mg/kg-d)
(mg/L)
(mg/L)
tumorsb
0
0
0
0
14/156 (9%)
1
0.12
5.73 x 10-4
5.32 x 10-5
8/80 (10%)
Female
3
0.36
1.72 x 10-3
1.59 x 10-4
6/80 (8%)
10
1.25
6.02 x 10-3
5.49 x 10-4
9/80(11%)
30
3.65
1.79 x 10-2
1.58 x 10-3
12/80(15%)
100
10.90
5.63 x 10-2
4.46 x 10-3
14/73 (19%)
aAdministered doses were averages calculated by the study authors based on animal BW and drinking water intake,
incidences for F344 rats do not include animals from the 6- and 12-mo sacrifices and were further adjusted to
exclude (from the denominators) rats that died between 0 and 12 mos in the study.
Table B-26. Summary of BMD modeling results based on incidence of
mammary gland tumors in F344 rats exposed to AN in drinking water for 2
years
Dose metric
Best-fit model3
X2 /'-value1
AIC
BMD10C
BMDL10d
Females
Administered dose
1°MS
0.94
388.94
8.73 mg/kg-d
4.77 mg/kg-d
CEO
1°MS
0.94
388.88
3.58 x 10-3 mg/L
1.97 x 10-3 mg/L
AN
1°MS
0.93
389.00
4.51 x 10"2 mg/L
2.45 x 10"2 mg/L
aDose-response models were fit using BMDS, version 1.4.1. "1°MS" indicates a one-stage multistage model.
hp value from the %2 goodness of fit test. Values <0.1 indicate a significant lack of fit.
°BMDio = BMD at 10% extra risk.
dBMDLio = 95% lower confidence limit on the BMD at 10% extra risk.
B-238
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for mammary gland tumors in F344 female rats employing
administered dose as a dose metric
0.3
0.25
TD
CD
"G 0 2
it
<
.2 0.15
"8
ro
i_
LL
0.1
0.05
0 2 4 6 8 10
dose
15:04 01/26 2009
Multistage Cancer Model. (Version: 1.5; Date: 02/20/2007)
Input Data File: M:\ACN DOSE-RESPONSE MODELING\CANCER\ORAL\F344_FEMALE_MAMMARY_DW.(d)
Gnuplot Plotting File: M:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F344_FEMALE_MAMMARY_DW.pit
Mon Jan 26 15:04:09 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 = Response
Independent variable = Dose
Total number of observations = 6
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: le-008
Parameter Convergence has been set to: le-008
Multistage Cancer Model with 0.95 Confidence Level
Multistage Cancer
Linear extrapolation
Ji "T
T "
Vi|
. jl{ 1
bmp.lI
r
-
BMD
B-239
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.093497
Beta(1) = 0.0112514
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.52
Beta(1) -0.52 1
Parameter Estimates
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Variable
Background
Beta(1)
Estimate
0 . 0913697
0 . 0120689
Std. Err.
•k
•k
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-192.056
-192.471
-195.631
# Param's
6
2
1
Deviance Test d.f.
P-value
0.830208
7.14977
0 . 9344
0.2097
AIC:
5 . 943
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000
0.0914
14.254
14
156
-0 . 070
0.1200
0 . 0927
7 .415
8
80
0 .226
0.3600
0 . 0953
7.625
6
80
-0.619
1.2500
0.1050
8.398
9
80
0 .220
3.6500
0.1305
10 .442
12
80
0.517
10 . 9000
0 .2034
14 .846
14
73
-0 .246
Chia2 = 0.81 d.f. = 4 P-value = 0.9365
B-240
DRAFT- DO NOT CITE OR QUOTE
-------
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 8.72994
BMDL = 4 .77219
BMDU = 27.9615
Taken together, (4.77219, 27.9615) is a 90 % two-sided confidence
interval for the BMD
Multistage Cancer Slope Factor = 0.0209547
B-241 DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for mammary gland tumors in F344 female rats employing
CEO in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
0.3
0.25
TD
CD
"3 0 2
ij=
<
.9 0.15
o
CO
i_
LL
0.1
0.05
Multistage
0 0.0005 0.001 0.0015 0.002 0.0025 0.003 0.0035 0.004 0.0045
dose
15:16 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F344_FEMALE_MAMMARY_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F3 4 4_FEMALE_MAMMARY_BLOOD_CEO.pit
Thu Sep 27 15:16:49 2007
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 = Response
Independent variable = Dose
Total number of observations = 6
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: le-008
Parameter Convergence has been set to: le-008
B-242
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.0927789
Beta(1) = 27.631
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.53
Beta(1) -0.53 1
Parameter Estimates
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Variable
Background
Beta(1)
Estimate
0 . 0908696
29 .4458
Std. Err.
•k
•k
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-192.056
-192.441
-195.631
# Param's
6
2
1
Deviance Test d.f.
P-value
0.770201
7.14977
0 . 9424
0.2097
AIC:
388.883
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000
0.0909
14.176
14
156
-0.049
0.0001
0 . 0923
7.383
8
80
0.238
0.0002
0.0951
7.609
6
80
-0.613
0.0005
0.1054
8.436
9
80
0 .205
0.0016
0.1322
10.576
12
80
0.470
0.0045
0 .2028
14 .801
14
73
-0.233
a2 = 0.75
d.f.
= 4 P-value
= 0.9447
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0.1
Extra risk
0 . 95
0 . 00357812
0 . 00196521
0 . 011322
Taken together, (0.00196521, 0.011322) is a 90
interval for the BMD
% two-sided confidence
B-243
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for mammary gland tumors in F344 female rats employing
AN in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
Multistage
0.3
0.25
T3
0
0.2
<
.9 0.15
"8
ro
i_
LL
0.05
BMDL
BMD
0
0.01
0.02
0.03
0.04
0.05
dose
11:39 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F344_FEMALE_MAMMARY_BLOOD_AN.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\ORAL\F344_FEMALE_MAMMARY_BLOOD_AN.pit
Thu Sep 27 11:39:30 2007
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 = Response
Independent variable = Dose
Total number of observations = 6
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: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
B-244
DRAFT- DO NOT CITE OR QUOTE
-------
Background =
Beta(l) =
0 . 0941503
2 .16751
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background
Beta(1)
1
-0.52
-0.52
1
Variable
Background
Beta(1)
Parameter Estimates
Estimate
0 . 0918429
2 .33757
Std. Err.
* - Indicates that this value is not calculated.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-192 . 056
-192.5
-195.631
389.001
# Param's
6
2
1
Deviance Test d.f.
0.887943
7.14977
P-value
0 . 9263
0 .2097
Goodness of Fit
Dose
Est. Prob.
Expected
Observed
14
8
6
9
12
14
Size
Scaled
Residual
0.0000
0.0006
0.0017
0 . 0060
0.0179
0 . 0563
Chia2 = 0.87
0.0918
0 . 0931
0.0955
0.1045
0.1291
0 .2038
14 .327
7 .445
7.639
8.363
10.325
14 .880
156
80
80
80
80
73
-0 . 091
0 .214
-0.624
0.233
0.559
-0.256
d.f. = 4
P-value = 0.9282
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0.1
Extra risk
0 . 95
0.0450726
0.0245301
0.146059
Taken together, (0.0245301, 0.146059) is a 90
interval for the BMD
% two-sided confidence
B-245
DRAFT- DO NOT CITE OR QUOTE
-------
APPENDIX B-4. CANCER INHALATION DOSE-RESPONSE ASSESSMENT: BMD
MODELING RESULTS FOR TUMOR INCIDENCE DATA FROM RATS
CHRONICALLY EXPOSED TO AN VIA INHALATION
As summarized in Section 4.6.2, AN is a multisite carcinogen in chronic rodent
bioassays. Data from the only available chronic inhalation cancer bioassay with multiple
exposure levels to AN were used for dose-response assessment (Dow Chemical Co., 1992a;
Quast et al., 1980b). In this study, male and female Sprague-Dawley rats were exposed to AN in
air at concentrations of 0, 20, or 80 ppm 6 hours/day, 5 days/week for 2 years. At 80 ppm,
significantly increased incidences of CNS tumors (i.e., astrocytomas and glial cell proliferation)
and Zymbal's gland tumors were observed in males and females. Also at this concentration,
significantly increased incidences of malignant mammary gland tumors (i.e., adenocarcinomas)
in females, as well as intestinal and tongue tumors in males, were seen. At 20 ppm, male and
female rats exhibited increased incidences (although nonsignificant) of CNS tumors (i.e.,
astrocytomas and glial cell proliferation) and Zymbal's gland tumors.
In this appendix, detailed results of the dose-response modeling for each of the tumor
sites listed above are presented (Tables B-27 through B-36). For each tumor site, first a
summary of the dose-response data is presented, followed by a table summarizing the results of
the dose-response modeling. Finally, the standard output from EPA's BMDS, version 1.4.1, for
the selected dose-response model for each tumor site is presented.
In general, the multistage model was fit to all of the data sets with the BMR set at
0.1 (i.e., 10% extra risk). In fitting this model, successive stages of the multistage model,
starting with stage 1 and ending with the stage equal to the number of dose groups minus one,
were fit to the tumor incidence data at a particular site for each rat sex employing either
administered dose or the internal dose metric CEO in blood. Then, for each dose metric, all
stages of the multistage model that did not show a significant lack of fit (i.e. ,p> 0. 1) were
compared using AIC. The stage of the multistage model with the lowest AIC was selected as the
best-fit model. For most tumor sites, the one-stage model exhibited the best fit. For data sets
that exhibited a significant lack of fit for all stages of the multistage model, dose groups were
dropped (starting with the highest dose group) until an adequate fit was achieved.
B-246
DRAFT- DO NOT CITE OR QUOTE
-------
Sprague-Dawlev Rats (Dow Chemical Co., 1992a; Quast et al., 1980b)
Tumor Site: Intestine
Table B-27. Incidence of intestinal tumors in Sprague-Dawley rats exposed
to AN in air for 2 years
Sex
Administered AN concentration
(ppm in air)
Predicted CEO-AUC in blood
(mg/L)
Incidence of intestinal
tumors"
0
0
4/96 (4%)
Male
20
2.17 x 10-3
3/93 (3%)
80
8.20 x 10"3
17/82 (21 %)b
"Incidences for Sprague-Dawley rats do not include animals from the 6- and 12-mo sacrifices and were further
adjusted to exclude (from the denominators) rats that died between 0 and 12 mos in the study.
bSignificantly different from controls (p < 0.05) as calculated by the study authors.
Table B-28. Summary of BMD modeling results based on incidence of
intestinal tumors in Sprague-Dawley rats exposed to AN in air for 2 years
Dose metric
Best-fit model3
X2 p valueb
AIC
BMD10°
BMDL10d
Males
Administered dose
2°MS
0.45
148.05
59.04 ppm
42.68 ppm
CEO
2°MS
0.42
148.13
6.06 x 10"3 mg/L
4.47 x 10-3 mg/L
aDose-response models were fit using BMDS, version 1.4.1. "2°MS" indicates a two-stage multistage model
bp value from the %2 goodness of fit test. Values <0.1 indicate a significant lack of fit.
°BMD10 = BMD at 10% extra risk.
dBMDL10 = 95% lower confidence limit on the BMD at 10% extra risk.
B-247
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for intestinal tumors in Sprague-Dawley male rats employing
administered dose as a dose metric
Multistage Model with 0.95 Confidence Level
T3
0
ts
0
<
o
CO
0.3
Multistage
0.25
¦i
0.2
0.15
^ - ' 1 .
0.1
{
0.05
1 ——f
-
0
-1- X
BMDL
BMD , ,
0 10
13:10 04/13 2007
20
30
40
dose
50
60
70
80
Multistage Model. $Revision: 2.1 $ $Date: 2000/08/21 03:38:21 $
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_MALE_INTESTINE_INHALE_PPM.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_MALE_INTESTINE_INHALE_PPM.pit
Fri Apr 13 13:10:33 2007
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2) ]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
B-248
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.0312892
Beta(l) = 0
Beta(2) = 3.12226e-005
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Beta(2)
Background 1 -0.55
Beta(2) -0.55 1
Parameter Estimates
Variable Estimate Std. Err.
Background 0.033209 0.0736218
Beta(1) 0 NA
Beta(2) 3.02304e-005 2.28675e-005
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-71.7319
-72 . 0247
-81. 082
Deviance Test DF
0.585523
18.7001
P-value
0 .4442
<.0001
AIC:
148 . 049
B-249
DRAFT- DO NOT CITE OR QUOTE
-------
Goodness of Fit
Dose Est._Prob. Expected Observed Size ChiA2 Res.
i : 1
0.0000 0.0332 3.188 4 96 0.263
i: 2
20.0000 0.0448 4.169 3 93 -0.294
i : 3
80.0000 0.2033 16.669 17 82 0.025
Chi-square = 0.57 DF = 1 P-value = 0.4521
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 59.036
BMDL = 42 .6779
B-250
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for intestinal tumors in Sprague-Dawley male rats employing
CEO in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
0.3
TD
0
"8
c
o
"8
0.25
0.2
0.15
0.1
0.05
Multistage
0 0.001 0.002 0.003
BMPl]
0.004 0.005
dose
BMP
0.006
0.007 0.008
15:20 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_MALE_INTESTINE_INHALE_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_MALE_INTESTINE_INHALE_BLOOD_CEO.pit
Thu Sep 27 15:20:06 2007
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2) ]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
B-251
DRAFT- DO NOT CITE OR QUOTE
-------
Background = 0.0306135
Beta(l) = 0
Beta(2) = 2979.81
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Beta(2)
Background 1 -0.55
Beta(2) -0.55 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0.0328821 * * *
Beta(1) 0 * * *
Beta(2) 2868.6 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood) # Param's Deviance Test d.f. P-value
-71.7319 3
-72.0636 2 0.663417 1 0.4154
-81.082 1 18.7001 2 <.0001
AIC: 148.127
Goodness of Fit
Dose Est._Prob. Expected Observed Size
Scaled
Residual
0.0000 0.0329 3.157 4 96 0.483
0.0022 0.0459 4.265 3 93 -0.627
0.0082 0.2025 16.609 17 82 0.107
Chia2 = 0.64 d.f. = 1 P-value = 0.4246
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00606044
BMDL = 0.004465
BMDU = 0.00809222
Taken together, (0.004465, 0.00809222) is a 90 % two-sided confidence
interval for the BMD
B-252
DRAFT- DO NOT CITE OR QUOTE
-------
Tumor Site: CNS (Dow Chemical Co., 1992a; Quast et al., 1980b)
Table B-29. Incidence of CNS tumors in Sprague-Dawley rats exposed to
AN in air for 2 years
Sex
Administered AN concentration
(ppm in air)
Predicted CEO-AUC in blood
(mg/L)
Incidence of CNS tumors"
Male
0
0
0/96 (0%)
20
2.17 x 10-3
4/93 (4%)
80
8.20 x 10"3
22/82 (27%)b
Female
0
0
0/93 (0%)
20
2.18 x 10-3
8/99 (8%)b
80
8.24 x 10-3
20/89 (22%)b
"Incidences for Sprague-Dawley rats do not include animals from the 6- and 12-mo sacrifices and were further
adjusted to exclude (from the denominators) rats that died between 0 and 12 mos in the study.
bSignificantly different from controls (p < 0.05) as calculated by the study authors.
Table B-30. Summary of BMD modeling results based on incidence of CNS
tumors in Sprague-Dawley rats exposed to AN in air for 2 years
Dose metric
Best-fit model3
X2 /j-valuc'
AIC
BMD10°
BMDL10d
Males
Administered dose
1°MS
0.56
131.64
30.22 ppm
22.23 ppm
CEO
1°MS
0.50
131.92
3.14 x 10-3 mg/L
2.31 x 10-3 mg/L
Females
Administered dose
1°MS
0.80
152.86
30.79 ppm
22.89 ppm
CEO
1°MS
0.87
152.70
3.21 x 10-3 mg/L
2.39 x 10-3 mg/L
aDose-response models were fit using BMDS, version 1.4.1. "1°MS" indicates a one-stage multistage model.
hp value from the %2 goodness of fit test. Values <0.1 indicate a significant lack of fit.
°BMD10 = BMD at 10% extra risk.
dBMDL10 = 95% lower confidence limit on the BMD at 10% extra risk.
B-253
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for CNS tumors in Sprague-Dawley male rats employing
administered dose as a dose metric
Multistage Model with 0.95 Confidence Level
Multistage
0.35
0.3
0.25
0.2
0.15
1
0.05
0
BMD
0 10 20 30 40 50 60 70 80
dose
13:18 04/13 2007
Multistage Model. $Revision: 2.1 $ $Date: 2000/08/21 03:38:21 $
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_MALE_CNS_INHALE_PPM.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_MALE_CNS_INHALE_PPM.pit
Fri Apr 13 13:18:45 2007
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
B-254
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0
Beta(1) = 0.00403595
Asymptotic Correlation Matrix of Parameter Estimates
Beta(1)
( *** The model parameter(s) -Background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Beta(1)
1
Variable
Background
Beta(1)
Parameter Estimates
Estimate
0 . 00348587
Std. Err.
NA
0.00147454
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
Log(likelihood) Deviance Test DF
-64 .1853
-64 .8194
-85.6554
1.26826
42 . 9402
P-value
0.5304
<.0001
AIC: 131.639
Goodness of Fit
Dose Est._Prob. Expected Observed
Size
Chi 2 Res.
i : 1
0.0000
i: 2
20 . 0000
i : 3
80 . 0000
Chi-square =
0.0000
0.0673
0 .2434
1.15
0 . 000
6.263
19 . 956
DF = 2
0
4
22
96
93
82
0 . 000
-0.387
0.135
P-value = 0.5617
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0.1
Extra risk
0 . 95
30.225
22 .2323
B-255
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for CNS tumors in Sprague-Dawley male rats employing
CEO in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
0.4
0.35
0.3
| 0.25
£
< 0.2
c
tj 0.15
ro
L_
LL
0.1
0.05
0
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008
dose
15:22 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_MALE_CNS_INHALE_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_MALE_CNS_INHALE_BLOOD_CEO.pit
Thu Sep 27 15:22:30 2007
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
Multistage
BMDL
BMD
Default Initial Parameter Values
B-256
DRAFT- DO NOT CITE OR QUOTE
-------
Background = 0
Beta(1) = 39.4721
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Beta(1)
Beta(1) 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0 * * *
Beta(1) 33.5271 * * *
* - Indicates that this value is not calculated.
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
# Param's Deviance Test d.f.
Log(likelihood) # Param'
-64.1853 3
-64.9602 1
-85.6554 1
P-value
1.54975
42 . 9402
0 .4608
<.0001
AIC:
131.92
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0000 0.000 0 96 0.000
0.0022 0.0702 6.525 4 93 -1.025
0.0082 0.2404 19.711 22 82 0.592
Chia2 = 1.40 d.f. = 2 P-value = 0.4963
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00314255
BMDL = 0.00231159
BMDU = 0.00442253
Taken together, (0.00231159, 0.00442253) is a 90 % two-sided confidence
interval for the BMD
B-257
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for CNS tumors in Sprague-Dawley female rats
employing administered dose as a dose metric
Multistage Model with 0.95 Confidence Level
Multistage
0.3
0.25
T3
0 ~ ~
t5 0.2
0
it
<
c 0.15
o
"8
ro
i_
LL
0.05
BMD
0
10
30
40
50
60
70
80
dose
13:31 04/13 2007
Multistage Model. $Revision: 2.1 $ $Date: 2000/08/21 03:38:21 $
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_FEMALE_CNS_INHALE_PPM.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_FEMALE_CNS_INHALE_PPM.pit
Fri Apr 13 13:31:42 2007
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
B-258
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.00947538
Beta(1) = 0.00310229
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Beta(1)
Beta(1) 1
Parameter Estimates
Variable Estimate Std. Err.
Background 0 NA
Beta(1) 0.00342189 0.00148795
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Analysis of Deviance Table
Model Log(likelihood) Deviance Test DF P-value
Full model -75.2138
Fitted model -75.4293 0.431111 2 0.8061
Reduced model -91.1284 31.8293 2 <.0001
AIC: 152.859
Goodness of Fit
Dose Est._Prob. Expected Observed Size ChiA2 Res.
i : 1
0.0000 0.0000 0.000 0 93 0.000
i: 2
20.0000 0.0661 6.549 8 99 0.237
i : 3
80.0000 0.2395 21.314 20 89 -0.081
Chi-square = 0.45 DF = 2 P-value = 0.7982
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 30.7901
BMDL = 22 .8938
B-259
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for CNS tumors in Sprague-Dawley female rats
employing CEO in blood as an internal dose metric
0.35
0.3
0.25
TD
CD
% 0.2
<
O 0-15
"8
ro
u: 0.1
0.05
0
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008
dose
15:38 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_FEMALE_CNS_INHALE_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_FEMALE_CNS_INHALE_BLOOD_CEO.pit
Thu Sep 27 15:38:25 2007
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
Multistage Model with 0.95 Confidence Level
Multistage
BMD
B-260
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.00769725
Beta(1) = 30.2917
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Beta(1)
Beta(1) 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0 * * *
Beta(1) 32.7805 * * *
* - Indicates that this value is not calculated.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood) # Param's Deviance Test d.f. P-value
-75.2138 3
-75.3524 1 0.277338 2 0.8705
-91.1284 1 31.8293 2 <.0001
AIC:
152 .705
Goodness of Fit
Dose
Est. Prob.
Expected
Observed
Size
Scaled
Residual
0.0000 0.0000 0.000 0 93 0.000
0.0022 0.0690 6.827 8 99 0.465
0.0082 0.2367 21.065 20 89 -0.266
Chia2 = 0.29 d.f. = 2 P-value = 0.8663
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00321413
BMDL = 0.00238988
BMDU = 0.00446381
Taken together, (0.00238988, 0.00446381) is a 90 % two-sided confidence
interval for the BMD
B-261
DRAFT- DO NOT CITE OR QUOTE
-------
Tumor Site: Zymbal Gland (Dow Chemical Co., 1992a; Quast et al., 1980b)
Table B-31. Incidence of Zymbal gland tumors in Sprague-Dawley rats
exposed to AN in air for 2 years
Sex
Administered AN concentration
(ppm in air)
Predicted CEO-AUC in
blood
(mg/L)
Incidence of Zymbal gland
tumors"
Male
0
0
2/96 (2%)
20
2.17 x 10-3
4/93 (4%)
80
8.20 x 10-3
11/82 (13%)b
Female
0
0
0/93 (0%)
20
2.18 x 10-3
1/98 (1%)
80
8.24 x 10-3
11/89 (12%)b
"Incidences for Sprague-Dawley rats do not include animals from the 6- and 12-mo sacrifices and were further
adjusted to exclude (from the denominators) rats that died between 0 and 12 mos in the study.
bSignificantly different from controls (p < 0.05) as calculated by the study authors.
Table B-32. Summary of BMD modeling results based on incidence of
Zymbal's gland tumors in Sprague-Dawley rats exposed to AN in air for
2 years
Dose metric
Best-fit model3
X2 /J-valuc'
AIC
BMD10°
BMDL10d
Males
Administered dose
1°MS
0.78
121.17
70.29 ppm
42.53 ppm
CEO
1°MS
0.73
121.21
7.26 x 10-3 mg/L
4.40 x 10"3 mg/L
Females
Administered dose
1°MS
0.50
81.47
75.70 ppm
48.74 ppm
CEO
1°MS
0.46
81.67
7.90 x 10-3 mg/L
5.09 x 10-3 mg/L
aDose-response models were fit using BMDS, version 1.4.1. "1°MS" indicates a one-stage multistage model.
bp value from the %2 goodness of fit test. Values <0.1 indicate a significant lack of fit.
°BMD10 = BMD at 10% extra risk.
dBMDLio = 95% lower confidence limit on the BMD at 10% extra risk.
B-262
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for Zymbal gland tumors in Sprague-Dawley male rats
employing administered dose as a dose metric
Multistage Model with 0.95 Confidence Level
T3
0
ts
0
<
c
o
"8
0.2
Multistage
I
0.15
Ij
0.1
j
I ^
"j 1 -
0.05
T
-"T
^ I
-
0
BMDL
..., BMD ...,
0 10
20
30
40
dose
50
60
70
80
13:22 04/13 2007
Multistage Model. $Revision: 2.1 $ $Date: 2000/08/21 03:38:21 $
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_MALE_ZYMBAL_INHALE_PPM.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_MALE_ZYMBAL_INHALE_PPM.pit
Fri Apr 13 13:22:06 2007
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
B-263
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.0172853
Beta(1) = 0.00156747
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.66
Beta(1) -0.66 1
Variable
Background
Beta(1)
Parameter Estimates
Estimate
0 . 0193054
0 . 00149893
Std. Err.
0 . 0819561
0 . 00187956
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Deviance Test DF
Log(likelihood)
-58.5432
-58.5838
-63 .5267
121.168
0.0811571
9 . 9669
P-value
0.7757
0 . 00685
Dose
Goodness of Fit
Est._Prob. Expected Observed
Size
Chi 2 Res.
i : 1
0.0000
i: 2
20 . 0000
i : 3
80 . 0000
Chi-square =
0.0193
0 . 0483
0.1301
0 . 08
1.853
4.489
10.670
DF = 1
2
4
11
96
93
82
0 . 081
-0.114
0 . 036
P-value = 0.7780
B-264
DRAFT- DO NOT CITE OR QUOTE
-------
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 70.2907
BMDL = 42 .5347
B-265 DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for Zymbal gland tumors in Sprague-Dawley male rats
employing CEO in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
0.25
0.2
$ 0.15
£
<
¦I 0.1
o
CO
0.05
0
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008
dose
15:25 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_MALE_ZYMBAL_INHALE_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_MALE_ZYMBAL_INHALE_BLOOD_CEO.pit
Thu Sep 27 15:25:06 2007
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
Multistage
B-266
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.0165179
Beta(1) = 15.3411
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.66
Beta(1) -0.66 1
Parameter Estimates
Variable
Background
Beta(1)
Estimate
0 . 0190621
14.5086
Std. Err.
* - Indicates that this value is not calculated.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-58.5432
-58.6032
-63 .5267
121.206
# Param's
3
2
1
Deviance Test d.f.
0.120008
9 . 9669
P-value
0.729
0 . 00685
Goodness of Fit
Dose Est._Prob. Expected Observed Size
Scaled
Residual
0.0000
0 . 0022
0.0082
Chia2 = 0.12
0.0191
0.0495
0.1291
d.f. = 1
1.830 2 96 0.127
4.600 4 93 -0.287
10.586 11 82 0.136
P-value = 0.7323
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0.1
Extra risk
0 . 95
0 . 00726195
0 . 00440461
0 . 0159076
Taken together, (0.00440461, 0.0159076) is a 90
interval for the BMD
two-sided confidence
B-267
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for Zymbal gland tumors in Sprague-Dawley female rats
employing administered dose as a dose metric
Multistage Model with 0.95 Confidence Level
0.2
0.15
T3
0
"8
£
< 0.1
o
"8
ro
!
Ll_
0.05
0
0 10 20 30 40 50 60 70 80
dose
13:35 04/13 2007
Multistage Model. $Revision: 2.1 $ $Date: 2000/08/21 03:38:21 $
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_FEMALE_ZYMBAL_INHALE_PPM.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_FEMALE_ZYMBAL_INHALE_PPM.pit
Fri Apr 13 13:35:30 2007
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
Multistage
BMDL
BMP
B-268
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0
Beta(1) = 0.0017365
Asymptotic Correlation Matrix of Parameter Estimates
Beta(1)
( *** The model parameter(s) -Background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Beta(1)
1
Variable
Background
Beta(1)
Parameter Estimates
Estimate
0
0 . 00139182
Std. Err.
NA
0 . 00127839
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood) Deviance Test DF
-38.8683
-39.7334
-49.5377
81.4668
1.73007
21.3387
P-value
0 .421
<.0001
Dose
Goodness of Fit
Est._Prob. Expected Observed
Size
Chi 2 Res.
i : 1
0.0000
i: 2
20 . 0000
i : 3
80 . 0000
Chi-square =
0.0000
0.0275
0.1054
1.41
0 . 000
2 .690
9.378
DF = 2
0 93 0.000
1 98 -0.646
11 89 0.193
P-value = 0.4952
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0.1
Extra risk
0 . 95
75.6997
48.7384
B-269
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for Zymbal gland tumors in Sprague-Dawley female rats
employing CEO in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
TD
0
"8
c
o
"8
Multistage
T _
l
_ T
. i
BMDL
^1 1
BMD
0.2
0.15
0.1
0.05
0
0 0.001
15:41 09/27 2007
0.002 0.003
0.004
dose
0.005 0.006 0.007 0.008
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_FEMALE_ZYMBAL_INHALE_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_FEMALE_ZYMBAL_INHALE_BLOOD_CEO.pit
Thu Sep 27 15:41:40 2007
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
B-270
DRAFT- DO NOT CITE OR QUOTE
-------
Background = 0
Beta(1) = 16.8864
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Background
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Beta(1)
Beta(1) 1
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
Background 0 * * *
Beta(1) 13.3383 * * *
* - Indicates that this value is not calculated.
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
# Param's Deviance Test d.f.
Log(likelihood) # Param'
-38.8683 3
-39.8337 1
-49.5377 1
P-value
1. 93077
21.3387
0.3808
<.0001
AIC:
81.6675
Goodness of Fit
Scaled
Dose Est._Prob. Expected Observed Size Residual
0.0000 0.0000 0.000 0 93 0.000
0.0022 0.0287 2.808 1 98 -1.095
0.0082 0.1041 9.263 11 89 0.603
Chia2 = 1.56 d.f. = 2 P-value = 0.4579
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 0.00789912
BMDL = 0.00508579
BMDU = 0.0132303
Taken together, (0.00508579, 0.0132303) is a 90 % two-sided confidence
interval for the BMD
B-271
DRAFT- DO NOT CITE OR QUOTE
-------
Tumor Site: Tongue (Dow Chemical Co., 1992a; Quast et al., 1980b)
Table B-33. Incidence of tongue tumors in Sprague-Dawley rats exposed to
AN in air for 2 years
Sex
Administered AN concentration
(ppm in air)
Predicted CEO-AUC in blood
(mg/L)
Incidence of tongue tumors3
0
0
1/95 (1%)
Male
20
2.17 x 10-3
0/14 (0%)
80
8.20 x 10-3
7/82 (9%)b
"Incidences for Sprague-Dawley rats do not include animals from the 6- and 12-mo sacrifices and were further
adjusted to exclude (from the denominators) rats that died between 0 and 12 mos in the study.
bSignificantly different from controls (p < 0.05) as calculated by the study authors.
Table B-34. Summary of BMD modeling results based on incidence of
tongue tumors in Sprague-Dawley rats exposed to AN in air for 2 years
Dose metric
Best-fit model3
X2 /j-valuc'
AIC
BMD10°
BMDL10d
Males
Administered
dose
1°MS
0.51
63.76
111.06 ppm
59.41 ppm
CEO
2°MS
0.63
63.37
9.48 x 10"3 mg/L
6.39 x 10-3 mg/L
aDose-response models were fit using BMDS, version 1.4.1. "1°MS" indicates a one-stage multistage model.
"2°MS" indicates a two-stage multistage model.
bp value from the %2 goodness of fit test. Values <0.1 indicate a significant lack of fit.
°BMD10 = BMD at 10% extra risk.
dBMDL10 = 95% lower confidence limit on the BMD at 10% extra risk.
B-272
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for tongue tumors in Sprague-Dawley male rats employing
administered dose as a dose metric
Multistage Model with 0.95 Confidence Level
Multistage
0.2
0.15
1
0.05
0
BMDlJ
BMD
0 20 40 60 80 100 120
dose
13:25 04/13 2007
Multistage Model. $Revision: 2.1 $ $Date: 2000/08/21 03:38:21 $
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_MALE_TONGUE_INHALE_PPM.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_MALE_TONGUE_INHALE_PPM.pit
Fri Apr 13 13:25:25 2007
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
B-273
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0
Beta(1) = 0.00109944
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background 1 -0.65
Beta(1) -0.65 1
Parameter Estimates
Variable Estimate Std. Err.
Background 0.00947867 0.0968926
Beta(1) 0.000948673 0.00186696
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
AIC:
Log(likelihood)
-29.4666
-29.8775
-33 .2127
63.755
Deviance Test DF
0.821799
7 .49227
P-value
0.3647
0 . 02361
Dose
Goodness of Fit
Est._Prob. Expected Observed
Size
Chi 2 Res.
i : 1
0.0000
i: 2
20 . 0000
i : 3
80 . 0000
Chi-square =
0 . 0095
0.0281
0.0819
0.43
0 . 900
0.393
6 .713
DF = 1
95
14
82
0.112
-1.029
0 . 046
P-value = 0.5124
B-274
DRAFT- DO NOT CITE OR QUOTE
-------
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 111.061
BMDL = 5 9.4126
B-275 DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for tongue tumors in Sprague-Dawley male rats employing
CEO in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
TD
0
"8
c
o
"8
0.25
0.2
0.15
0.1
0.05
Multistage
-r-
1 J—
- -L O
BMDL
4-- ^ .
1 | -
BM
0.002
0.004
0.006
0.008
dose
15:28 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_MALE_TONGUE_INHALE_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_MALE_TONGUE_INHALE_BLOOD_CEO.pit
Thu Sep 27 15:28:11 2007
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2) ]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
B-276
DRAFT- DO NOT CITE OR QUOTE
-------
Background =
Beta(l) =
Beta(2) =
0 . 00257667
0
1279.63
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Beta(2)
Background 1 -0.64
Beta(2) -0.64 1
Parameter Estimates
Variable
Background
Beta(1)
Beta(2)
Estimate
0 . 00925756
0
1172 .26
Std. Err.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-29.4666
-29.6826
-33 .2127
63 .3652
# Param'
3
2
1
Deviance Test d.f.
0 .431994
7 .49227
P-value
0.511
0 . 02361
Goodness of Fit
Dose
Est. Prob.
Expected
Observed
Size
Scaled
Residual
0.0000
0 . 0022
0.0082
Chia2 = 0.23
0.0093
0 . 0147
0 . 0844
d.f. = 1
0.879 1 95
0.206 0 14
6.917 7 82
P-value = 0.6339
0.129
-0.457
0 . 033
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0.1
Extra risk
0 . 95
0.0094804
0.00638558
0.0279407
Taken together, (0.00638558, 0.0279407) is a 90
interval for the BMD
% two-sided confidence
B-277
DRAFT- DO NOT CITE OR QUOTE
-------
Tumor Site: Mammary Gland (Dow Chemical Co., 1992a; Quast et al., 1980b)
Table B-35. Incidence of mammary gland tumors in Sprague-Dawley rats
exposed to AN in air for 2 years
Sex
Administered AN concentration
(ppm in air)
Predicted CEO-AUC in
blood
(mg/L)
Incidence of mammary gland
tumors"
0
0
9/93 (10%)
Female
20
2.18 x 10-3
8/98 (8%)
80
8.24 x 10-3
20/99 (20%)b
"Incidences for Sprague-Dawley rats do not include animals from the 6- and 12-mo sacrifices and were further
adjusted to exclude (from the denominators) rats that died between 0 and 12 mos in the study.
bSignificantly different from controls (p < 0.05) as calculated by the study authors.
Table B-36. Summary of BMD modeling results based on incidence of
mammary gland tumors in Sprague-Dawley rats exposed to AN in air for
2 years
Dose metric
Best-fit model3
X2 /J-valuc'
AIC
BMD10°
BMDL10d
Females
Administered dose
1°MS
0.28
219.42
66.48 ppm
37.82 ppm
CEO
2°MS
0.57
218.52
7.31 x 10-3 mg/L
4.33 x 10-3 mg/L
aDose-response models were fit using BMDS, version 1.4.1. "1°MS" indicates a one-stage multistage model.
"2°MS" indicates a two-stage multistage model.
bp value from the %2 goodness of fit test. Values <0.1 indicate a significant lack of fit.
°BMD10 = BMD at 10% extra risk.
dBMDL10 = 95% lower confidence limit on the BMD at 10% extra risk.
B-278
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for mammary gland tumors in Sprague-Dawley female rats
employing administered dose as a dose metric
Multistage Model with 0.95 Confidence Level
Multistage
0 10 20
30 40 50
dose
60 70 80
13:38 04/13 2007
Multistage Model. $Revision: 2.1 $ $Date: 2000/08/21 03:38:21 $
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_FEMALE_MAMMARY_INHALE_PPM.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_FEMALE_MAMMARY_INHALE_PPM.pit
Fri Apr 13 13:38:03 2007
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 = Response
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: le-008
Parameter Convergence has been set to: le-008
B-279
DRAFT- DO NOT CITE OR QUOTE
-------
Default Initial Parameter Values
Background = 0.0767126
Beta(1) = 0.00173168
Asymptotic Correlation Matrix of Parameter Estimates
Background Beta(l)
Background
Beta(1)
1
-0.68
-0.68
1
Variable
Background
Beta(1)
Parameter Estimates
Estimate
0 . 0806351
0 . 00158474
Std. Err.
0 . 0783644
0 . 00179372
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-107 . 092
-107.71
-110.714
219.421
Deviance Test DF
1.23579
7 .24319
P-value
0.2663
0 . 02674
Dose
Goodness of Fit
Est._Prob. Expected Observed
Size
Chi 2 Res.
i : 1
0.0000
i: 2
20 . 0000
i : 3
80 . 0000
Chi-square =
0.0806
0.1093
0.1901
1.1S
7.499
10.713
18.820
DF = 1
93
20
P-value = 0.2754
0 .218
-0.284
0 . 077
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 66 .4843
BMDL = 37.8172
B-280
DRAFT- DO NOT CITE OR QUOTE
-------
BMDS (version 1.4.1) output for mammary gland tumors in Sprague-Dawley female rats
employing CEO in blood as an internal dose metric
Multistage Model with 0.95 Confidence Level
Multistage
0.3
0.25
T3
0
0.2
<
° 0.15
"8
ro
i_
LL
0.05
BMD
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008
dose
15:44 09/27 2007
Multistage Model. (Version: 2.8; Date: 02/20/2007)
Input Data File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_FEMALE_MAMMARY_INHALE_BLOOD_CEO.(d)
Gnuplot Plotting File: G:\ACN DOSE-RESPONSE
MODELING\CANCER\INHALATION\SD_FEMALE_MAMMARY_INHALE_BLOOD_CEO.pit
Thu Sep 27 15:44:26 2007
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*doseAl-beta2*doseA2) ]
The parameter betas are restricted to be positive
Dependent variable = Response
Independent variable = Dose
Total number of observations = 3
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
B-281
DRAFT- DO NOT CITE OR QUOTE
-------
Background =
Beta(l) =
Beta(2) =
0 . 0853385
0
1996 .34
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Background Beta(2)
Background 1 -0.6
Beta(2) -0.6 1
Parameter Estimates
Variable
Background
Beta(1)
Beta(2)
Estimate
0 . 0856208
0
1971.76
Std. Err.
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
* - Indicates that this value is not calculated.
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-107 . 092
-107.258
-110.714
218.517
# Param'
3
2
1
Deviance Test d.f.
0.332027
7 .24319
P-value
0.5645
0 . 02674
Goodness of Fit
Dose
Est. Prob.
Expected
Observed
Size
Scaled
Residual
0.0000
0 . 0022
0.0082
Chia2 = 0.33
0.0856
0 . 0941
0.2002
7 . 963
9 .226
19.818
93
20
d.f. = 1
P-value = 0.5658
0.384
-0 .424
0 . 046
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
BMDU =
0.1
Extra risk
0 . 95
0 . 00730992
0 . 00432882
0 . 0152838
Taken together, (0.00432882, 0.0152838) is a 90
interval for the BMD
% two-sided confidence
B-282
DRAFT- DO NOT CITE OR QUOTE
-------
APPENDIX B-5. ANALYSIS TO ASSESS COMBINING TUMOR INCIDENCE DATA
FROM TWO CANCER BIOASSAYS EMPLOYING SPRAGUE-DAWLEY RATS
A statistical analysis was conducted to determine whether tumor dose-response data from
two 2-year drinking water bioassays in Sprague-Dawley rats were similar enough to be
combined. The first study employed AN drinking water concentrations of 0, 35, 100, and
300 ppm (Quast, 2002; Quast et al., 1980a), while the second study used AN drinking water
concentrations of 0, 1, and 100 ppm (Johannsen and Levinskas, 2002a). To conduct the analysis,
the multistage model in BMDS (version 1.3.2) was fit to the tumor incidence data from three
sites (forestomach, CNS, and Zymbal gland) in each sex across both studies, using administered
animal dose expressed in mg/kg-day. Using the best-fit model, a statistical test described by
Stiteler et al. (1993), which employs a maximum likelihood ratio statistic distributed as a % , was
then used to test the null hypothesis that the corresponding data sets from the two studies are
compatible with a common dose-response model. If the null hypothesis is not rejected, this
analysis provides evidence that the results from the two studies may be pooled.
The results of this analysis showed that forestomach and Zymbal gland tumors in male
and female Sprague-Dawley rats were not compatible with a common dose-response model,
while CNS tumors in male and female Sprague-Dawley rats were compatible with a common
dose-response model. Because of these conflicting results, it was decided that the results from
the two Sprague-Dawley drinking water studies would not be pooled. Therefore, the final dose-
response analysis for deriving the oral slope factor for AN focused on the two rat drinking water
studies containing the most dose groups (i.e., the Sprague-Dawley rat bioassay reported by Quast
[2002] and the F344 rat bioassay reported by Johannsen and Levinskas [2002b]). For each tumor
site, summaries of the results of the statistical tests for compatibility are shown below.
Test 1: Forestomach tumors in male Sprague-Dawley rats
Tumor incidence data (Johannsen and Levinskas, 2002a):
• 0 ppm (0 mg/kg-day): 3/78
• 1 ppm (0.085 mg/kg-day): 3/78
• 100 ppm (8.53 mg/kg-day): 11/77
Best-fit model: 1-stage multistage
Log (likelihood) = -57.0113
Tumor incidence data (Quast, 2002):
• 0 ppm (0 mg/kg-day): 0/80
• 35 ppm (3.42 mg/kg-day): 2/47
B-283
DRAFT- DO NOT CITE OR QUOTE
-------
• 100 ppm (8.53 mg/kg-day): 23/48
• 300 ppm (21.20 mg/kg-day): 39/48
Best-fit model: 2-stage multistage (highest dose group dropped)
Log (likelihood) = -42.4099
Tumor incidence data (combined):
• 0 ppm (0 mg/kg-day): 3/158
• 1 ppm (0.085 mg/kg-day): 3/78
• 35 ppm (3.42 mg/kg-day): 2/47
• 100 ppm (8.53 mg/kg-day): 34/125
• 300 ppm (21.20 mg/kg-day): 39/48
Best-fit model: 2-stage multistage
Log (likelihood) = -132.865
Conclusion: Based on the likelihood ratio test statistic, -2 In A = 2[132.865 - (42.4099 +
57.0113)] = 2 x 33.4438 = 66.89, the data sets are not compatible with a common dose-response
model atp < 0.0001.
Test 2: Forestomach tumors in female Sprague-Dawley rats
Tumor incidence data (Johannsen and Levinskas, 2002a):
• 0 ppm (0 mg/kg-day): 1/80
• 1 ppm (0.11 mg/kg-day): 4/79
• 100 ppm (10.80 mg/kg-day): 7/79
Best-fit model: 1-stage multistage
Log (likelihood) = -45.8328
Tumor incidence data (Quast, 2002):
• 0 ppm (0 mg/kg-day): 1/80
• 35 ppm (4.36 mg/kg-day): 1/48
• 100 ppm (10.80 mg/kg-day): 12/48
• 300 ppm (25.00 mg/kg-day): 30/48
Best-fit model: 2-stage multistage
Log (likelihood) = -69.9834
B-284
DRAFT- DO NOT CITE OR QUOTE
-------
Tumor incidence data (combined):
• 0 ppm (0 mg/kg-day): 2/160
• 1 ppm (0.11 mg/kg-day): 4/79
• 35 ppm (4.36 mg/kg-day): 1/48
• 100 ppm (10.80 mg/kg-day): 19/127
• 300 ppm (25.00 mg/kg-day): 30/48
Best-fit model: 2-stage multistage
Log (likelihood) = -119.054
Conclusion: Based on the likelihood ratio test statistic, -2 In A = 2[119.054 - (69.9834 +
45.8328)] = 2 x 3.2378 = 6.48, the data sets are not compatible with a common dose-response
model at/? = 0.011.
Test 3: CNS tumors in male Sprague-Dawley rats
Tumor incidence data (Johannsen and Levinskas, 2002a):
• 0 ppm (0 mg/kg-day): 2/78
• 1 ppm (0.085 mg/kg-day): 3/75
• 100 ppm (8.53 mg/kg-day): 23/77
Best-fit model: 1-stage multistage
Log (likelihood) = -68.9254
Tumor incidence data (Quast, 2002):
• 0 ppm (0 mg/kg-day): 1/80
• 35 ppm (3.42 mg/kg-day): 12/47
• 100 ppm (8.53 mg/kg-day): 22/48
• 300 ppm (21.20 mg/kg-day): 30/48
Best-fit model: 1-stage multistage
Log (likelihood) = -98.6909
Tumor incidence data (combined):
• 0 ppm (0 mg/kg-day): 3/158
• 1 ppm (0.085 mg/kg-day): 3/75
B-285
DRAFT- DO NOT CITE OR QUOTE
-------
• 35 ppm (3.42 mg/kg-day): 12/47
• 100 ppm (8.53 mg/kg-day): 45/125
• 300 ppm (21.20 mg/kg-day): 30/48
Best-fit model: 1-stage multistage
Log (likelihood) = -168.873
Conclusion: Based on the likelihood ratio test statistic, -2 In A = 2[168.873 - (98.6909 +
68.9254)] = 2 x 1.2567 = 2.51, the data sets are compatible with a common dose-response model
at/? = 0.113.
Test 4: CNS Tumors in female Sprague-Dawley rats
Tumor incidence data (Johannsen and Levinskas, 2002a):
• 0 ppm (0 mg/kg-day): 0/79
• 1 ppm (0.11 mg/kg-day): 1/80
• 100 ppm (10.80 mg/kg-day): 39/78
Best-fit model: 1-stage multistage
Log (likelihood) = -59.5775
Tumor incidence data (Quast, 2002):
• 0 ppm (0 mg/kg-day): 1/80
• 35 ppm (4.36 mg/kg-day): 20/48
• 100 ppm (10.80 mg/kg-day): 25/48
• 300 ppm (25.00 mg/kg-day): 31/48
Best-fit model: Log-logistic
Log (likelihood) = -104.123
Tumor incidence data (combined):
• 0 ppm (0 mg/kg-day): 1/159
• 1 ppm (0.11 mg/kg-day): 1/80
• 35 ppm (4.36 mg/kg-day): 20/48
• 100 ppm (10.80 mg/kg-day): 64/126
• 300 ppm (25.00 mg/kg-day): 31/48
Best-fit model: Log-logistic
B-286
DRAFT- DO NOT CITE OR QUOTE
-------
Log (likelihood) = -164.485
Conclusion: Based on the likelihood ratio test statistic, -2 In A = 2[164.485 - (104.123 +
59.5775)] = 2 x 0.7845 = 1.57, the data sets are compatible with a common dose-response model
at p = 0.210.
Test 5: Zymbal gland tumors in male Sprague-Dawley rats
Tumor incidence data (Johannsen and Levinskas, 2002a):
• 0 ppm (0 mg/kg-day): 1/80
• 1 ppm (0.085 mg/kg-day): 0/71
• 100 ppm (8.53 mg/kg-day): 17/73
Best-fit model: 1-stage multistage
Log (likelihood) = -45.815
Tumor incidence data (Quast, 2002):
• 0 ppm (0 mg/kg-day): 3/80
• 35 ppm (3.42 mg/kg-day): 4/47
• 100 ppm (8.53 mg/kg-day): 3/48
• 300 ppm (21.20 mg/kg-day): 16/48
Best-fit model: 1-stage multistage
Log (likelihood) = -70.1142
Tumor incidence data (combined):
• 0 ppm (0 mg/kg-day): 4/160
• 1 ppm (0.085 mg/kg-day): 0/71
• 35 ppm (3.42 mg/kg-day): 4/47
• 100 ppm (8.53 mg/kg-day): 20/121
• 300 ppm (21.20 mg/kg-day): 16/48
Best-fit model: 1-stage multistage
Log (likelihood) = -118.806
Conclusion: Based on the likelihood ratio test statistic, -2 In A = 2[118.806 - (70.1142 +
45.815)] = 2 x 2.8768 = 5.75, the data sets are not compatible with a common dose-response
model atp = 0.016.
B-287
DRAFT- DO NOT CITE OR QUOTE
-------
Test 6: Zymbal gland tumors in female Sprague-Dawley rats
Tumor incidence data (Johannsen and Levinskas, 2002a):
• 0 ppm (0 mg/kg-day): 1/79
• 1 ppm (0.11 mg/kg-day): 0/75
• 100 ppm (10.80 mg/kg-day): 12/78
Best-fit model: 1-stage multistage
Log (likelihood) = -39.6428
Tumor incidence data (Quast, 2002):
• 0 ppm (0 mg/kg-day): 1/80
• 35 ppm (4.36 mg/kg-day): 5/48
• 100 ppm (10.80 mg/kg-day): 9/48
• 300 ppm (25.00 mg/kg-day): 18/48
Best-fit model: 1-stage multistage
Log (likelihood) = -76.3975
Tumor incidence data (combined):
• 0 ppm (0 mg/kg-day): 1/159
• 1 ppm (0.11 mg/kg-day): 0/75
• 35 ppm (4.36 mg/kg-day): 5/48
• 100 ppm (10.80 mg/kg-day): 21/126
• 300 ppm (25.00 mg/kg-day): 18/48
Best-fit model: 1-stage multistage
Log (likelihood) = -111.439
Conclusion: Based on the likelihood ratio test statistic, -2 In A = 2[111.439 - (76.3975 +
39.6428)] = 2 x (-4.6013) = -9.20 = 9.20, the data sets are not compatible with a common dose-
response model &tp = 0.002.
B-288
DRAFT- DO NOT CITE OR QUOTE
-------
APPENDIX B-6. ESTIMATION OF COMPOSITE CANCER RISK FROM EXPOSURE
TO AN BY COMBINING RISK ESTIMATES ACROSS MULTIPLE TUMOR SITES
Increased tumor incidences were observed at multiple sites in male and female rats
following exposure to AN both orally and by inhalation. With this multiplicity of tumors, the
concern is that a potency or risk estimate based solely on one tumor site (e.g., the most sensitive
site) may underestimate the overall cancer risk associated with exposure to this chemical. The
most recent EPA cancer guidelines (U.S. EPA, 2005a) identified two ways to approach this
issue: 1) analyze the incidence of tumor-bearing animals, or 2) combine the potencies associated
with significantly elevated tumors at each site. The NRC (1994) concluded that an approach
based on counts of animals with one or more tumors would tend to underestimate overall risk
when tumor types occur independently, and thus an approach based on combining the risk
estimates from each separate tumor type should be used.
Because potencies are typically upper bound estimates, summing such upper bound
estimates across tumor sites is likely to overstate the overall risk. Therefore, following the
recommendations of the NRC (1994) and the Guidelines for Carcinogen Risk Assessment (U.S.
EPA, 2005a), a statistically valid upper bound on combined risk was derived in order to gain
some understanding of the overall risk resulting from tumors occurring at multiple sites. It is
important to note that this estimate of overall potency describes the risk of developing tumors at
any combination of the sites considered and is not the risk of developing tumors at all three sites
simultaneously.
For modeling individual tumor data, the multistage model is specified as follows:
(1) P(d) = 1 - exp[-(qo + qid + q2d2 + ... + qmcT)].
The model for the combined (or composite) tumor risk is still multistage, with a
functional form that has the sum of stage-specific multistage coefficients as the corresponding
multistage coefficient.
(2) Pc(d) = 1 - exp[-(Zq0i + dlqh + d2Iq2l + ... + d"Iqfori=l,..., k,
where k = total number of sites
The resulting equation for fixed extra risk (BMR) is polynomial in dose (when logarithms
of both sides are taken) and can be straightforwardly solved for the combined BMD. However,
confidence bounds for this BMD are not able to be estimated by the current version of BMDS.
Therefore, a Bayesian approach to finding confidence bounds on the combined BMD was
implemented using WinBugs (Spiegelhalter et al., 2003). WinBugs software is freely available
and implements Markov chain Monte Carlo (MCMC) computations. Use of WinBugs has been
B-289
DRAFT- DO NOT CITE OR QUOTE
-------
demonstrated for derivation of a distribution of BMDs for a single multistage model (Kopylev et
al., 2007) and can be straightforwardly generalized to derive the distribution of BMDs for the
combined tumor load, following the NRC (1994) methodology described above. The advantage
of a Bayesian approach is that it produces a distribution of BMDs that allows better
characterization of statistical uncertainty. For the current analysis, a diffuse (high variance or
low tolerance) Gaussian prior restricted to be nonnegative was used. The posterior distribution
was based on three chains with 50,000 burn-in (i.e., the initial 50,000 simulations were dropped)
and a thinning rate of 20, resulting in 150,000 simulations total. The median and 5th percentile of
the posterior distribution provided the BMDio (central estimate) and BMDLio (lower bound) for
combined tumor load, respectively.
The methodology above was applied to the dose-response data for the male and female
Sprague-Dawley (Quast, 2002) and F344 (Johannsen and Levinskas, 2002b) rat drinking water
studies, as well as to the data from male and female Sprague-Dawley rats in the Quast et al.
(1980b) inhalation study. As with the risk estimates generated for individual tumor sites, the
combined analysis used the internal dose metric CEO in blood (see Appendices B-3 and B-4).
The human equivalent PODs are presented in Tables B-37 (episodic oral exposure) and B-38
(continuous inhalation exposure). Estimates of composite risk were estimated by dividing the
BMR of 10% extra risk by the composite BMDLioS (0.1/BMDLio). Human equivalent
composite CSFs are presented in Table B-39, and composite unit risks are presented in
Table B-40. The slopes derived from the composite BMDi0s (0.1/ BMDio) are also included in
these tables for comparison.
B-290
DRAFT- DO NOT CITE OR QUOTE
-------
Table B-37. Summary of PODs for composite cancer risk associated with
episodic oral exposure to AN, using CEO-AUC levels in blood as dose metric
and multiple tumor incidence data in rats
Rat strain, sex
PODs based on rat dose-response
using internal dose metric CEO-
AUC in blood (mg/L)
Human equivalent PODs"
(mg/kg-d)
Tumor site
BMD10
BMDL10
BMD10
BMDL10
Sprague-Dawley, male
Forestomach, CNS, Zymbal gland,
tongue
4.2 x 10-4
3.6 x 10-4
0.042
0.036
Sprague-Dawley, female
Forestomach, CNS, Zymbal gland,
tongue, mammary gland
2.6 x 10-4
2.0 x 10-4
0.026
0.020
F344, male
Forestomach, CNS, Zymbal gland
4.6 x 10-4
3.3 x 10-4
0.046
0.033
F344, female
Forestomach, CNS, Zymbal gland,
mammary gland
6.7 x 10-4
5.2 x 10-4
0.066
0.051
"Converted using human PBPK model.
Table B-38. Summary of PODs for composite cancer risk associated with
inhalation exposure to AN, based on multiple tumor incidence data in rats
and CEO-AUC levels in blood
Rat strain, sex
PODs for rat dose-response in
terms of CEO-AUC in blood
(mg/L)
Human equivalent PODs:
AN in air"
(mg/m3)
Tumor site
BMC10
BMCL10
BMC10
BMCL10
Sprague-Dawley, male
Intestine, CNS, Zymbal gland, tongue
1.4 x 10"3
1.1 x 10-3
1.9
1.5
Sprague-Dawley, female
CNS, Zymbal gland, mammary gland
1.7 x 10"3
1.3 x 10-3
2.3
1.8
aConverted using human PBPK model.
B-291
DRAFT- DO NOT CITE OR QUOTE
-------
Table B-39. Estimated human oral CSFs for AN based on multiple tumor
incidence data in rats and CEO-AUC levels in blood
Rat strain, sex
Tumor site
Slope derived from composite
BMD10a
(mg/kg-d)1
Composite CSFb
(mg/kg-d)1
Sprague-Dawley, male
Forestomach, CNS, Zymbal gland, tongue
2.4
2.8
Sprague-Dawley, female
Forestomach, CNS, Zymbal gland, tongue,
mammary gland
3.8
5.0
F344, male
Forestomach, CNS, Zymbal gland
2.2
3.1
F344, female
Forestomach, CNS, Zymbal gland,
mammary gland
1.5
1.9
aSlope estimated by 0.1/BMD10.
bCSF = 0.1/BMDL10.
Table B-40. Estimated human IURs for AN based on multiple tumor
incidence data in rats and CEO-AUC levels in blood
Rat strain, sex
Tumor site
Slope to background from
overall BMC10a
(mg/m3)1
Overall IURb
(mg/m3)1
Sprague-Dawley, male
Intestine, CNS, Zymbal gland,
tongue
5.4 x 10-2
6.8 x 10-2
Sprague-Dawley, female
CNS, Zymbal gland, mammary gland
4.4 x 10-2
5.7 x 10-2
aSlope estimated by 0.1/BMDi0.
bIUR = 0. 1/BMDLiq.
B-292
DRAFT- DO NOT CITE OR QUOTE
-------
APPENDIX B-7. STATISTICAL ANALYSIS OF BLAIR ET AL. (1998)
The Data
The data used in this analysis came from the best available epidemiological study (see
Section 5.4.1) by Blair et al. (1998). The raw data from the study were provided to EPA
courtesy of NCI. The process of data collection is described in Blair et al. (1998) and in more
detail in Stewart et al. (1998). The full raw data set contains 363 variables on 25,460 subjects.
The variables include demographic information, employment information, vital status, and
various measures of exposure per study year (1942-1983), including maximum and minimum
estimates, "best estimate," frequency of peaks, and several other ways of measuring exposure.
The raw data provided by NCI did not include the smoking data that were collected for
approximately 10% of the subjects. Therefore, smoking information was not part of the EPA's
statistical analysis. The uncertainty connected to lack of smoking information is discussed in
Section 5.4.5.2. The data on dates of exposure and corresponding exposure amounts, dates of
mortality due to lung cancer and other causes, a plant worked, and the birth year were retained
for the statistical analysis. Biological age was chosen as a time scale. Data handling and
statistical analysis were performed using S-Plus™ statistical software.
The Statistical Model
The semi-parametric Cox proportional hazards model (Cox, 1972) is a widely used model
in hazard regression. The Cox model with time-dependent covariates was chosen to model the
data for to two main reasons. Primarily, it allows taking into account individual covariate
history, allowing utilization of the extensive exposure data collected by Blair et al. (1998).
Additionally, the Cox model uses internal controls. Internal controls constitute an appropriate
comparison group, given the healthy worker effect observed by Blair et al. (1998) and further
demonstrated by Marsh et al. (2001).
In the Cox model, the conditional hazard function, given the covariate process Z(t), is
assumed to have the form:
X(t\Z(t)) = Ao(0exp(/?TZ(0)
where P is the vector of regression coefficients and Ao(t) denotes the baseline hazard function.
No particular shape is assumed for the baseline hazard; it is estimated nonparametrically. The
contributions of covariates to the hazard are multiplicative. When P Z(t) is small and Z(t)
represents exposure, the Cox proportional model is consistent with linearity of dose-response for
low doses. When no time-dependent covariates are present, the cumulative hazard function A0(t)
is estimated using a Breslow (Breslow, 1974) estimator:
B-293
DRAFT- DO NOT CITE OR QUOTE
-------
u-.n ZeXP(^Z;(0)
jeR(i)
The survival function and cumulative probability of death are then estimated using the
modified cumulative hazard estimator. The modified estimator only approximates the true
cumulative hazard since a covariate path is not encountered in the estimator. Therneau and
Grambsch (2000) discuss clinical examples when this estimator could lead to inappropriate
results, but the occupational case is very different from clinical examples, since at the baseline,
not all workers are exposed and exposure is nondecreasing and often ends well before death.
B-294
DRAFT- DO NOT CITE OR QUOTE
-------
Calculations of Excess Risk
Following Starr et al. (2004), EPA defined hazard functions for lung cancer mortality
Xlc(t) and other causes of death /Jic(t), with the corresponding cause-specific cumulative hazard
functions, Alc (t) and Aoc(t). The sum of Alc(t) and Aoc(t) is the all-causes mortality hazard
Aac(t) with corresponding survival function Sac(t). Then, the cumulative risk of lung cancer
mortality by age t, given covariate path Z(t)=z(t) is:
R'c(t\z(tJ) = (ii\z(t))Sac (ii\z(t))
u
-------
Mortality and Risk Estimates. The coxph function of S-Plus™ was used to obtain
estimates of p. The estimate of plc was equal to 1.24 x 10"3with the standard error of 2.47 x 10"2
(p = 0.61). The estimate of poc was equal to 6.7 x 10"4 with a standard error of 9.66 x 10"4
(p = 0.49). The estimates, covariates, and event times were then used to construct estimates of
cumulative hazard and risks, as described in the previous section. The exposures corresponding
to excess risks were divided by 2 to account for differences in volume of air inhaled during a
"3
working day and during a whole day (10 vs. 20 m /day). Resulting exposure estimates (by age
80): ECoi = 0.992 ppm and LECoi = 0.238 ppm. The corresponding unit risk was calculated to
equal 4.2 x 10"2 ppm"1.
Limitations of the Statistical Approach
The statistical approach followed the approach of Starr et al. (2004), and it shares the
limitations and uncertainties described there. The limitations and uncertainties include the
following:
• The Cox model fit the data adequately, but it was an empirical model fit rather than a
biologically based model.
• The estimator of the cumulative hazard does not account for the covariate path and hence,
is only an approximation.
• The estimate of risk is obtained using the first-order "linearized" approximation.
However, the results are consistent with assumptions of the first-order approximation
validity.
• The Cox model uses internal controls and hence, it is much more vulnerable than SMR
computation to contamination of pool of workers who are free of lung cancer by lung
cancer cases. That is caused by misdiagnosis of lung cancer as not a lung cancer.
Selikoff and Seidman (1992) showed based on histopathological analysis of a
comparably large industrial cohort (17,800 asbestos workers) that as much as 13.7% of
lung cancer cases are misdiagnosed on death certificates, both as neoplasms and
nonmalignant diseases, with about 3% as asbestos only diseases (mesothelioma and
asbestosis). Therefore, the Blair cohort may have as much as 10% of internal controls
that are actually lung cancers. That would cause a strong bias against finding a
statistically significant relationship.
B-296
DRAFT- DO NOT CITE OR QUOTE
-------
APPENDIX C. PBPK MODEL DESCRIPTIONS AND SOURCE CODE
Model Description
The PBPK models used to calculate rat and human internal doses for the development of
candidate RfDs, oral CSFs, and IURs were those revised from Kedderis et al. (1996) (rat, with
EH activity towards CEO added and other parameters revised) and Sweeney et al. (2003)
(human). Both models are flow limited, lumped compartment models that predict amounts and
concentrations of AN and its metabolite of toxicological interest, CEO, in blood and seven
tissues: lung, brain, fat, stomach, liver, and rapidly and slowly perfused tissues. Both models
include portals of entry for oral, inhalation, and i.v. routes of exposure. Drinking water is
modeled as a continuous infusion to the GI lumen. (The model code allows one to define
episodic oral exposure as a series of boluses up to 6 times/day, but the simpler approach of
treating the exposure as continuous was chosen.) For rats, the infusion rate is set such that the
total amount consumed equals the average total amount ingested during the bioassays as
determined from water consumption and administered AN concentration. Inhalation duration
and frequency can be explicitly defined. Simulated metabolism of AN and CEO occur only in
the liver. Elimination of both compounds is accomplished via second-order GSH conjugation in
various tissues, saturable hepatic metabolism, and pulmonary excretion into exhaled air.
The rat and human models are based on the same structural framework but have two
primary differences: (1) physiological and metabolic parameters for each are species specific,
and (2) the human model simulates saturable, enzyme-mediated hydrolysis of CEO, a metabolic
process not observed in rats. In the absence of human in vivo data for metabolism, human
metabolic parameters were estimated from in vitro values extrapolated to in vivo values by using
in vitro/in vivo ratios in rats, which is described fully in Sweeney et al. (2003). In the Sweeney
et al. (2003) model, a conversion factor for in vitro to in vivo extrapolation of human metabolic
constants is implicitly defined in the calculation of Vmax values for AN oxidation and CEO
hydrolysis. Model parameter values used for both species are given in Table C-l.
The acslXtreme model code as used to calculate the various dose metrics, followed by the
corresponding Matlab model code used to perform the optimizations for fitting of the revised
parameters, are given at the end of this appendix. The Matlab code is divided into three .m files:
the top-level "optACN2.m" file, a secondary "RunACN.m" file that is called by the opt file and
returns the value of the objective function, and an "EqACN.m" file that defines the set of
differential equations.
C-l
DRAFT - DO NOT CITE OR QUOTE
-------
Table C-l. Rat and human PBPK model parameter values
Parameter
Rat value"
Human value"
Total BW (kg)
0.25
70
Percentage of B W
Liver
4
2.57
Brain
0.6
2.00
Stomach
0.63
0.21
Fat
7
21.42
Rapidly perfused
3.77
5.22
Slowly perfused
75
58.58
Venous blood
3.9
4.35
Arterial blood
2.1
2.74
Blood flows (L/h/kg°74)
Cardiac output
14.0
13.4
Alveolar ventilation
14.0
12.9
Percentage of cardiac output
Liver
25
21.4
Brain
2.4
11.4
Stomach
1.3
1.3
Fat
9
5.2
Rapidly perfused
47.3
32.5
Slowly perfused
15
28.2
GSH content (mmol/L)
Liver
8.53
5.63
Brain
2.00
2.99
Stomach
4.59
3.61
Rapidly perfused
2.65
2.59
Slowly perfused
0.75
1.13
PCs
AN
CEO
AN
CEO
Blood: air
512
1,658
154
1,658
Livenblood
0.46
0.274
1.51
0.274
Brain:blood
0.40
1.407
1.34
1.407
Stomach:blood
0.46
0.274
1.51
0.274
Fat:blood
0.28
0.785
0.94
0.785
Rapidly perfused:blood
0.46
0.274
1.51
0.274
Slowly perfused:blood
0.35
1.853
1.16
1.853
Blood binding (h_1)
Hb (kec; kBC2)
1.245 (3.66)
1.134(3.33;
1.245 (3.66)
1.134 (3.33)
Blood sulfhydryls (kFBC, kFBC2)
2.54
0.68
0.0008
0.84
AN oxidation
VmaxC (mg/h/kg0 7)
5.0 (7.1)
15.6 (22.1)
Km (mg/L)
1.5 (2.76)
0.8
C-2 DRAFT - DO NOT CITE OR QUOTE
-------
Table C-l. Rat and human PBPK model parameter values
Parameter
Rat value"
Human value"
CEO hydrolysis
VmaxcC2 (mg/h/kg07)
...
841 (-)
Km2 (mg/L)
...
113 (—)
kEHC (L/h/kg07)
— (3.92)
— (2.20)
AN-GSH conjugation
Enzymatic: kFC (h/kg03)-1
73 (50)
113 (77)
Spontaneous: kso (mmol/h)"1
0.2584
0.2584
CEO-GSH conjugation
Liver: kFC2 (h/kg03)"1
500
197
Brain/liver (kFBRc/kFC2)
0.0234
0.0531
Stomach/liver (kFSTC/kFC2)
0.0538
0.0641
Rapidly perfused/liver (kFRC/kFC2)
0.0311
0.0460
Slowly perfused/liver (kFSC/kFC2)
0.00879
0.0201
Oral absorption: Ka (h_1)
8.0 (4.2)
8.0 (4.2)
aRevised values used in this assessment are italicized in parentheses.
Sources: Sweeney et al. (2003); Kedderis et al. (1996).
Estimation of Internal Doses Corresponding to Bioassay Exposures
The model code from Sweeney et al. (2003) was modified to enable the explicit
definition of desired daily oral intake of AN either as a continuous infusion or in an episodic
pattern of up to six episodes (or both continuous and episodic). The continuous-infusion oral
exposure rate (STDOSE) is simply calculated as being equal to the total daily exposure
(STEADYODOSE*BW, by continuous oral dosing) divided by 24 hours. The episodic exposure
is calculated as the drinking water concentration (DRCONC) times the standard drinking water
volume (DRVOL, based on BW) times the ratio of actual water consumption to standard
consumption (FRACVOL) times the fraction of daily water consumed at episode "I" (DRP(I)),
with that quantity added as a bolus to the GI lumen compartment at episode time "I" (DRT(I)).
The first episodic exposure time is assumed to be TIME = 0, the same time at which a daily
gavage dose (ODOSE*BW) is also administered.
To simulate inhalation exposures as described in Quast (1980b), the model was run to
provide 6-hour continuous inhalation exposures, repeating every 24 hours (by setting model
variable TCHNG1 = 6.0) for 5 days/week (by setting model variable TCHNG2 = 120, the
number of hours in 5 days). Inhalation simulations were run for 800 continuous simulated hours
so that the effect of the 5 day/week exposures would be included in the calculation of average
lifetime daily AUCs. Study-specific rat BWs were used. In this way the PBPK model accounted
for the dynamics of concentration rise at the beginning of each exposure and clearance at the end
C-3
DRAFT - DO NOT CITE OR QUOTE
-------
of each exposure. The resulting internal dose metrics for inhalation exposure of the rat are listed
in Table 5-16.
Rat to Human Extrapolations
For inhalation and oral cancer and oral noncancer effects, similar approaches were used
to extrapolate PODs from rats to humans. Following derivation of PODs for noncancer and
cancer endpoints, as weekly average concentrations of AN and CEO in rat blood tissue
(Ci!tiSsue!rat(BMCLio) in mg/L), HECs were then calculated using the human PBPK model and a
standard human BW of 70 kg, as the blood or brain concentration, assuming steady-state
exposure for a set of predefined exposure levels and that equal internal metrics in rats and
humans are associated with the same degree of response. The relationship between exposure
level and internal metric as predicted by the human PBPK model was then plotted in Excel and a
second-order polynomial of the form, HEC = a x Ci + b x Ci, where Ci is the internal dose
metric, was fit to the relationship for each metric (blood and brain AN and CEO).
Inhalation Exposure
The relationship between inhalation exposure level and internal metric as predicted by the
human PBPK model is shown in Figures C-l and C-2 for AN and CEO, respectively. This
polynomial form gave an excellent fit and was fit over a range of concentrations that bracketed
the range of Ci!tiSSue,rat(BMCLio) for various cases. The resulting polynomials were then used to
convert the rat Ci values to human equivalent exposures, reported in Chapter 5 for each toxicity
value.
C-4
DRAFT - DO NOT CITE OR QUOTE
-------
AN dosimetry
y = 38.6031 C + [4305.55C/(26.6520+
o ACN blood
~ ACN brain
— AN blood fit
— AN brain fit
0.0001 0.001 0.01 0.1 1 10 100
Internal average / steady state concentration (mg/L)
Points are steady-state internal AN concentrations predicted by the human PBPK
model for given inhalation exposure concentrations. Curves are polynomial
regressions.
Figure C-l. Human inhalation exposure level vs. internal AN concentration.
C-5
DRAFT - DO NOT CITE OR QUOTE
-------
100
E
Q.
a.
a)
i_
3
(0
o
Q.
X
a)
c
o
ro
(0
10
0.1
y = 41.103x2 + 618.54x
r2= 1
+ CEO blood
x CEO brain
— Poly. (CEO brain)
-Poly. (CEO blood)
y " 26.838X2 + 503.27x
r2 = 1
0.0001 0.001 0.01 0.1
Internal average / steady state concentration (mg/L)
Points are steady-state internal CEO concentrations predicted by the human PBPK
model for given inhalation exposure concentrations. Curves are polynomial
regressions.
Figure C-2. Human inhalation exposure level vs. internal CEO
concentrations.
Oral Cancer Risks
For extrapolation of oral cancer risks from exposure in drinking water, the exact pattern
of ingestion of drinking water by the rats in the bioassay was not measured and will be variable
among humans. Therefore, the rat PBPK model was used to estimate the internal doses of AN
and CEO at the exposure levels of the PODs as determined above (in mg/kg-day ingested), while
assuming two distinct exposure patterns that are expected to bound the truth: continuous
ingestion and ingestion in six boluses ("episodic").
The continuous pattern is simply described as continuous infusion to the stomach lumen
compartment (i.e., 24 hours/day) at a rate equal to the total daily ingestion (at the BMDLio).
This pattern leads to the lowest predicted peak concentration of any possible pattern that would
have the same total ingestion and, therefore, the least saturation of metabolism. The episodic
pattern assumes that rats consume their drinking water in six boluses, spaced at 4-hour intervals
during each day. The fraction of their total daily ingestion at each of these events is 23.3, 10, 10,
10, 23.3, and 23.4%, respectively. For continuous exposures, the steady-state blood and brain
concentrations of AN and CEO were calculated as internal dose metrics, while for the episodic
pattern, the daily average concentrations were determined.
C-6
DRAFT - DO NOT CITE OR QUOTE
-------
Strieker et al. (2003) measured water and food intake by Sprague-Dawley rats on diets
containing normal (1%) and high (8%) levels of NaCl. On days 5 and 10 of control exposure,
the rats were observed to consume water in 15.9 ± 2.0 and 19.3 ± 2.4 bouts, with amount
consumed per bout being 2.1 ± 0.3 and 1.8 ± 0.3 mL, respectively. Thus, actual consumption by
rats occurs in many more than six bouts, and the amount consumed per bout is much more
uniform than the episodic pattern. The data also show that most of the consumption occurs
during the 12-hour dark/waking period (represented by the first and final two boluses in the
EPA's pattern), with much less consumed during the light/sleeping period (represented by the
second to fourth boluses). Further, data on water consumption relative to controlled meal
delivery showed that drinking bouts can last as long as 80 seconds, rather than occurring in an
instantaneous bolus. Thus, actual consumption lies between an assumed continuous pattern and
the idealized episodic pattern. Because episodic consumption will lead to model prediction of
the greatest metabolic saturation, to obtain the desired prediction that represents an upper bound
of that behavior (including the possibility that water consumption in the bioassays, which
included F344 as well as Sprague-Dawley rats, may have differed from those observed by
Strieker at al. [2003]), this idealized episodic pattern rather than one designed to more closely
match that reported by Strieker et al. (2003) was chosen. The results of the PBPK simulations
for continuous oral infusion vs. an idealized episodic ingestion pattern for various oral BMDL
values differed by no more than 7%, indicating that for the rat internal dose metrics, the exact
exposure pattern was not significant.
To calculate the HEDs, similarly two alternate assumptions of continuous and episodic
exposure to the corresponding rat metrics were applied. For the episodic pattern, it was again
assumed that this occurred in six bolus doses but occurring at 0, 3, 5, 8, 11, and 15 hours
computation time (i.e., from the time of first ingestion in the morning, assumed to occur at
breakfast), with the amount consumed in each bolus being 25, 10, 25, 10, 25, and 5% of the total
daily intake, respectively. As with the inhalation extrapolation, the human PBPK model was run
using a range of fixed input dose rates (using both continuous and episodic regimens), the
resulting internal doses calculated, and then an empirical regression performed in Excel to
interpolate between those doses. The resulting simulation values (points) and regressions (lines)
are shown in Figures C-3 to C-6.
C-7
DRAFT - DO NOT CITE OR QUOTE
-------
100
Continuous Exposure Model
AN blood fits
o ACN blood
AN brain fits
~ ACN brain
CD
y = 27.0624 C + [161.3074-C/(0.569868 + C)]
S
I
1
0
3
= 35.6015 C + [138.101 C/(0.657976 + C)]
0
1
0.01
0.00001
0.0001
0.001
0.01
0.1
1
Internal steady state concentration (mg/L)
Points are PBPK model predictions; curves are polynomial regressions.
Figure C-3. Human oral exposure level vs. internal AN concentration for
continuous exposure.
100
Continuous Exposure Model
A CEO blood
x CEO brain
— Poly. (CEO blood)
— Poly. (CEO brain)
y = 10.185x + 98.2x
110
y = 6.7418x2 + 79.895x
0.01
0.0001
0.001
0.01
Internal steady state concentration (mg/L)
Points are PBPK model predictions; curves are polynomial regressions.
Figure C-4. Human oral exposure level vs. internal CEO concentration for
continuous exposure.
C-8
DRAFT - DO NOT CITE OR QUOTE
-------
Episodic Exposure Model — AN
100
AN blood fit
o AN blood
AN brain fit
~ AN brain
0
y = 51.3757 C + [15.2250 C/(0.0586390 + C)]
1
y = 40.6491 C + [1 5.2251 C/(0.074113 + C)]
1
0.01
0.00001
0.0001
0.001 0.01
Internal average concentration (mg/L)
Points are PBPK model predictions; curves are nonlinear regressions.
Figure C-5. Human oral exposure level vs. internal AN concentration for
episodic exposure.
Episodic Exposure Model — CEO
0
~ CEO blood
A CEO brain
Poly. (CEO blood)
Poly. (CEO brain)
1
y = 62.272X2 + 98.857x
y = 41.22X2 + 80.43x
1
0.01
0.0001
0.001
Internal
0.01
Internal average concentration (mg/L)
Points are PBPK model predictions; curves are polynomial regressions.
Figure C-6. Human oral exposure level vs. internal CEO concentration for
episodic exposure.
C-9
DRAFT - DO NOT CITE OR QUOTE
-------
acslXtreme Model Source Code
PROGRAM
ACRYLONITRILE and CYANOETHYLENE OXIDE MODEL w/ 1st order epoxide hydrolase
! Model of ACN administration and CEO production!
! Revised model of G.L. Kedderis!
! Original by M.L. Gargas 10/9/89, revised 5/6/92!
! Revised 2/21/94 by G.L. Kedderis!
! Refined 6/28-7/12/95, 8/8/95 by G.L. Kedderis!
Modified by SMH - 5/2/97!
Modified for continuous oral dose by LMS 3/7/00!
pst2, pr2, and pb2 revised to Kedderis et al 1996 values!
periodic drinking water reinstated!
modified for revised human scaling by LMS 7/26/01!
built blood flow mass balance into equations LMS 7/27/01!
built blood mass balance into equations, made blood binding!
rates equal in venous and arterial blood LMS 7/31/01!
modified to explicitly define DW cone. and daily ACN intake MHL 08/05!
statement added to calculate daily ACN intake (mg/kg/day) ml 08/05!
Modified 05/05/06 by Paul Schlosser and Allan Marcus to include!
Linear-range approximation to epoxide hydrolase model of ACN to CEO!
and to simplify the oral dosing calculations -- to avoid divide-by-zero!
errors and allow for either or both continuous and/or periodic exposure!
June '06 - Jan '07: Other modifications by P. Schlosser to allow both
continuous and episodic drinking water dosing, and to update default
parameters for rat
INTEGER I
REAL DRT(6
DIMENSION
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
!counter for
I, DRP(6)
AO(16)
AO = 0,0,0,0,
QPC = 14.
QCC = 14.
QLC = 0.25
QFC = 0.0 9
QBRC = 0.024
QSTC = 0.013
QSC =0.15
BW = 0.523
VLC = 0.04
VRC = 0.0377
VSC = 0.75
VBRC = 0.0 06
VFC = 0.0 7
VSTC = 0.0063
BVC =0.06
WBC = 0.6 5
PBR = 0.40
PBR2 = 1.4 07
PL = 0.46
PL2 = 0.274
PST = 0.46
PST2 = 0.274
PF = 0.28
PF2 = 0.785
PS = 0.35
PS2 = 1.853
drinking water arrays!
store drink water times, percents in array!
Set of zeros to force values to be >/= 0!
,0,0,0,0,0,0,0,0,0,0,0
Alveolar ventilation rate (L/hr)!
Cardiac output (L/hr)!
Fractional blood flow to liver!
Fractional blood flow to fat!
Fractional blood flow to brain!
Fractional blood flow to the stomach!
Fractional blood flow to slowly perfused tissue
Body weight (kg) male rat!
Fraction of liver tissue to total body!
Fraction of richly perfused tissue to total!
Fraction of slowly perfused tissue to total!
Fraction of brain tissue!
Fraction of fat tissue!
Fraction of stomach tissue!
Fraction of blood volume!
Fraction of venous blood volume!
AN brain/blood partition coefficient!
CEO brain/blood partition coefficient!
AN liver/blood partition coefficient!
CEO liver/blood partition coefficient!
AN stomach/blood partition coefficient!
CEO stomach/blood partition coefficient!
AN fat/blood partition coefficient!
CEO fat/blood partition coefficient!
AN slowly perfused tissue/blood partition!
CEO slowly perfused tissue/blood partition!
C-10
DRAFT - DO NOT CITE OR QUOTE
-------
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
CONSTANT
PR = 0.46
PR2 = 0.274
PB = 512
PB2 = 1658
MW = 53 . 0 6
MW2 = 6 9.05
KEHC = 3.92
VMAXC = 7.1
VMAXC2 = 0.0
KM = 2.76
KM2 = 113.
KFC =50
KFC2 = 50 0
KSO = 0.2584
GSHL = 8.53
GSHBR = 2 .
GSHST =4.59
GSHS = 0.75
GSHR = 2.65
KBC = 3 .66
KBC2 = 3.3 3
KFBC =2.54
KFBC2 = 0.6 8
KA = 4 .2
IVDOSE = 0.
CONC = 0.
CONC2 = 0.
AN rapidly perfused tissue/blood partition!
CEO rapidly perfused tissue/blood partition!
AN blood/air partition coefficient!
CEO blood/air partition coefficient!
ACN molecular weight (g/mol)!
CEO molecular weight (g/mol)!
Effective lst-order EH rate, liter/hr/kg^0.7
Maximum velocity of metabolism (mg/hr-lkg)!
Maximum velocity of metabolism, CEO (mg/hr-lkg)!
Michael is-Menten constant (mg/L) !
Michael is-Menten constant, CEO (mg/L) !
ACN first order metabolism rate const (/hr-lkg)!
CEO first order metabolism rate const (/hr-lkg)!
ACN second order reaction with GSH (L/mMol/hr)!
Liver GSH cone (mMol/L)!
Brain GSH (mMol/L)!
Stomach GSH (mMol/L)!
Slowly perfused (muscle) GSH (mMol/L)!
Rapidly perfused GSH (mMol/L)!
ACN 1ST order binding to blood hb (/hr)!
CEO 1ST order binding to blood hb (/hr)!
ACN 1ST order binding to blood RSH (/hr)!
CEO 1ST order binding to blood RSH (/hr)!
Oral uptake rate (/hr)!
IV dose (mg/kg)!
Inhaled concentration (ppm) ACN!
Inhaled concentration (ppm) CEO!
ODOSE= 0. ! Oral dose in mg/kg-day, assumed given at t=0, 24, etc.
! Timing commands!
CONSTANT TCHNG1 =24.0
CONSTANT TCHNG2 = 12 0.0
CONSTANT TINF = .003
! Length of exposure (hrs)!
! Allows for 5 day/week exposure!
! Length of IV infusion (hrs)!
CONSTANT STEADYODOSE = 0
Idaily dose mg/kg-day, if assuming continuous oral infusion
Iperiodic drinking water section!
CONSTANT DRCONC= 0.0 ! Cone, of ACN in drinking water (mg/L)
CONSTANT DRT=0,2,4,6,8,10
! Times for multiple oral drinks/day *after* 0
! Must be ascending, 0 <= times < 24 hr
! DRTIME(1) assumed = 0 and not used
CONSTANT DRP=1,0,0,0,0,0 !Percent consumed by drinking at those times
! Bolus of ODOSE*BW + DRPCT(1)*DRDOSE will be given at t=0,24,48, etc.
CONSTANT FRACVOL = 1.0
! Actual daily water consumption / nominal or control volume
! used when calculating DRDOSE
INITIAL
DRVOL = 0.102*BW**0.7 ! Water consumption (L/d)
! 2L/d water consumption for a 70kg human
DRDOSE = DRCONC*DRVOL*FRACVOL
DAYDOSE = ODOSE*BW + DRP(1)*DRDOSE
! Once daily oral dose; given at t=0 via initial condition
C-ll
DRAFT - DO NOT CITE OR QUOTE
-------
STDOSE = STEADYODOSE*BW/24 ! Steady oral dosing rate in mg/hr
IVR = IVDOSE*BW/TINF
NEWDAY =0; 1=2 ! First dose given as initial condition
SCHEDULE ORALDOSE.AT.DRT(2)
! Scaled parameters!
QC =
QP =
QL =
QF =
QBR =
QST =
QS =
QR =
VL =
VF =
VBR =
VST =
VS =
VR =
BV =
VAB =
WB =
GSHRB =
GSHRST =
GSHRS =
GSHRR =
VMAX =
VMAX2 =
KF =
KF2 =
KEH
KFBRC =
KFSTC =
KFSC =
KFRC =
KFBR =
KFST =
KFS =
KFR =
KFB =
KFB2 =
KB
KB 2 =
QCC*BW* * 0.74
QPC*BW* * 0.74
QLC*QC
QFC*QC
QBRC*QC
QSTC*QC
QSC*QC
QC-QL-QF-QBR-QS-
VLC*BW
VFC*BW
VBRC*BW
VSTC*BW
VSC*BW
VRC*BW
BVC*BW
BV* (1-WBC)
BV* WBC
GSHBR/GSHL
GSHST/GSHL
GSHS/GSHL
GSHR/GSHL
VMAXC*BW* * 0.7
VMAXC2 * BW* * 0.7
KFC/BW* * 0.3
KFC2/BW* * 0.3
= KEHC*BW* *0.7
KFC2*.1*GSHRB
KFC2*.1*GSHRST
KFC2*.1*GSHRS
KFC2*.1*GSHRR
KFBRC/BW* *.3
KFSTC/BW* *.3
KFSC/BW* *.3
KFRC/BW* *.3
KFBC
KFBC2
= KBC
KBC2
QST
[volume arterial blood!
ratio of brain GSH to liver GSH!
ratio of stomach GSH to liver GSH!
ratio of slowly perfused GSH to liver GSH!
ratio of richly perfused GSH to liver GSH!
Liver P450 ACN to CEO!
Liver CEO Hydrolysis!
Liver ACN-GSH rate!
Liver CEO-GSH/RSH rate!
Approx. first-order rate in liver!
CEO brain reaction - 10% GSH ratio!
CEO stomach reaction - 10% GSH ratio!
CEO slowly perfused reaction - 10% GSH ratio!
CEO richly perfused reaction - 10% GSH ratio!
CEO brain RSH rate!
CEO stomach RSH rate!
CEO SPT RSH rate!
CEO RPT RSH rate!
DAILYDOSE = ODOSE+STEADYODOSE+(DRDOSE/BW)+(AI/BW/(TSTOP/24))
Idaily dose mg/kg-day via oral and/or inhalation routes!
END
!of INITIAL
DYNAMIC
ALGORITHM IALG
MAXTERVAL MAXT
MINTERVAL MINT
CINTERVAL CINT
2 !Gear method for stiff systems!
1. Oe-3
1. Oe-9
0.5
DERIVATIVE
C-12
DRAFT - DO NOT CITE OR QUOTE
-------
i ACRYLONITRILE !
i i
! CI = Concentration in inhaled air (mg/L)!
CIZONE = PULSE(0.0,24.0,TCHNG1) * PULSE(0.0,168.0,TCHNG2)
CI = CIZONE*CONC*MW/24450.
! AI = Amount inhaled (mg)!
RAI = QP* CI
AI = INTEG(RAI,0)
! MR = Amount in stomach lumen (mg)!
RMR = STDOSE - KA*MR
AMR = INTEG(RMR,DAYDOSE)
MR = MAX(AMR, AO(1)) !no negative values!
! CAL = Concentration in arterial lung blood (mg/L)!
CAL = (QC*CVB+QP*CI)/(QC+(QP/PB))
! CA = Cone, in systemic arterial blood (mg/1)!
RAB = QC*(CAL-CA)-(KB+KFB)*AB
AB = INTEG(RAB,AO(2))
CA = AB/VAB
! AX
= Amount exhaled (mg)!
CX =
CAL/PB
CXPPM =
(0 . 7*CX+0.3*CI)*24450. /MW
RAX =
QP* CX
AX =
INTEG(RAX,AO(3))
! AST
= Amount in stomach tissue (mg)
RAST =
QST*(CA-CVST) - STGSH + KA*MR
AST =
INTEG(RAST,AO(4))
CVST =
AST/(VST* PST)
CST =
AST/VST
STGSH =
KSO* CVST*GSHST*VST
ASTG =
INTEG(STGSH,AO(5))
! AS = Amount in slowly perfused tissues (mg)!
RAS = QS*(CA-CVS) - SGSH
AS = INTEG(RAS,AO(6))
CVS = AS/(VS*PS)
CS = AS/VS
SGSH = KSO*CVS*GSHS*VS
ASG = INTEG(SGSH,AO(7))
! AR
= Amount in rapidly
RAR =
QR*(CA-CVR) - RGSH
AR =
INTEG(RAR,AO(8) )
CVR =
AR/(VR* PR)
CR =
AR/VR
RGSH =
KSO*CVR*GSHR*VR
ARG =
INTEG(RGSH,AO(9))
perfused tissues (mg)!
! AF = Amount in fat tissue (mg)!
RAF = QF*(CA-CVF)
AF = INTEG(RAF,AO(10))
CVF = AF/(VF*PF)
CF = AF/VF
C-13
DRAFT - DO NOT CITE OR QUOTE
-------
! ABR
= Amount in brain tissue
RABR =
QBR*(CA-CVBR) - BRGSH
ABR =
INTEG(RABR,AO(11))
CVBR =
ABR/(VBR* PBR)
CBR =
ABR/VBR
BRGSH =
KSO* CVBR*GSHBR*VBR
ABRG =
INTEG(BRGSH,AO(12))
! AL = Amount in liver tissue (mg)
RAL = QL*CA + QST*CVST - (QL+QST)*CVL - RAMI - RAM2 - LGSH
AL = INTEG(RAL,AO(13))
CVL = AL/(VL*PL)
CL = AL/VL
AUCL = INTEG(CL,0.)
LGSH = KSO*CVL*GSHL*VL
ALG = INTEG ( (LGSH+RAM2+(KB*CVB*WB) + (KFB*CVB*WB) ) , AO (14) )
! AMI = Amount metabolized, saturable (P450) and linear pathways (mg)!
RAMI = VMAX*CVL/(KM+CVL)
RAM1M = RAMI*1000/MW
AMI = INTEG(RAMI,AO(15))
! AM2 = Amount metabolized, first-order pathway (GST) (mg)!
RAM2 = KF* CVL*VL
AM2 = INTEG(RAM2,0)
! CV = Mixed venous blood concentration (mg/L)!
IV = IVR*(T<=TINF) ! PULSE(0,24,TINF)
CV = (QF*CVF + (QL+QST)*CVL + QS*CVS + QR*CVR + QBR*CVBR + IV)/QC
! CVB = Mixed venous ACN cone. after binding (mg/L)!
RVB = QC*(CV-CVB) - (KB+KFB)*VB
VB = INTEG(RVB,AO(16))
CVB = VB/WB
! TMASS = mass balance (mg)!
TMASS = ABR+AF+AL+AS+AR+AST+AM1+AM2+AX+MR+ASTG+ASG+ARG+ABRG+ALG
i i
i CE0 1
i i
! CI2 = Concentration in inhaled air (mg/L)!
CI2 = CIZONE*CONC2*MW2/24450.
! CAL2 = Concentration in arterial lung blood (mg/L)!
CAL2 = (QC*CVB2+QP*CI2)/(QC+(QP/PB2))
! CA2 = Cone, in systemic arterial blood (mg/1)!
RAB2 = QC*(CAL2-CA2)-(KB2+KFB2)*AB2
AB2 = INTEG(RAB2,0.)
CA2 = AB2/VAB
! AS2 = Amount in slowly perfused tissues (mg)!
RAS2 = QS*(CA2-CVS2) - KFS*CVS2*VS
AS2 = INTEG(RAS2,0.)
CVS2 = AS2 / (VS* PS2 )
CS2 = AS2/VS
C-14
DRAFT - DO NOT CITE OR QUOTE
-------
! AR2 = Amount in rapidly perfused tissues (mg)!
RAR2 = QR*(CA2-CVR2) - KFR*CVR2*VR
AR2 = INTEG(RAR2,0.)
CVR2 = AR2/(VR* PR2)
CR2 = AR2/VR
! AST2 = Amount in stomach (mg)!
RAST2 = QST*(CA2-CVST2) - KFST*CVST2*VST
AST2 = INTEG(RAST2,0.)
CVST2 = AST2/(VST*PST2)
CST2 = AST2/VST
! AF2 = Amount in fat tissue (mg)!
RAF2 = QF*(CA2-CVF2)
AF2 = INTEG(RAF2,0.)
CVF2 = AF2/(VF* PF2)
CF2 = AF2/VF
! ABR2 = Amount in brain tissue (mg)!
RABR2 = QBR* (CA2-CVBR2) - KFBR* CVBR2 *VBR
ABR2 = INTEG(RABR2,0.)
CVBR2 = ABR2/(VBR*PBR2)
CBR2 = ABR2/VBR
! CEO Hydrolysis!
! VMAX2 = maximum velocity of metabolism (mg/hr-lkg)!
! KM2 = Michael is-Menten constant (mg/L) !
! KEH is parameter for linear metabolism in liver to replace saturable
! AL2 = Amount CEO in liver tissue (mg)!
RAM = RAM1M*MW2/10 0 0.
RAL2ADD = QL*CA2 + QST*CVST2 + RAM
RAL2M = VMAX2*CVL2/(KM2+CVL2) + KEH*CVL2
RAL2 = RAL2ADD - (QL+QST)*CVL2 - KF2*CVL2*VL - RAL2M
AL2 = INTEG(RAL2,0.)
CVL2 = AL2/(VL* PL2)
CL2 = AL2/VL
! CV2 = Mixed venous blood concentration (mg/L)!
CV2 = (QF*CVF2 + (QL+QST)*CVL2 + QS*CVS2 + QR*CVR2 + QBR*CVBR2)/QC
! CVB2 = Mixed venous CEO cone. after binding!
RVB2 = QC*(CV2-CVB2) - (KB2+KFB2)*VB2
VB2 = INTEG(RVB2,0.)
CVB2 = VB2/WB
! Calculation of the AUC for ACN and CEO in the brain and blood!
AUCBR = INTEG(CBR,0.) ! AUC for ACN brain cone . !
AUCBR2 = INTEG(CBR2,0.)
AUCB = INTEG(CVB,0.)
AUCB2 = INTEG(CVB2,0.)
DAILYSTAUC = INTEG(CST,0
AUC for CEO brain cone.!
AUC for ACN blood cone.!
AUC for CEO blood cone. !
)/(TST0P/24) IMean daily stomach AN
DAILYST2AUC = INTEG(CST2,0.)/(TSTOP/24) IMean daily stomach CEO
END ! Derivative
! Code that is executed once at each communication interval goes here
CONSTANT TSTOP =24.0 ! Length of experiment (hrs)!
TERMT(T.GE.TSTOP, 'checked on communication interval: REACHED TSTOP')
DISCRETE ORALDOSE
IF (I.EQ.l) THEN
C-15
DRAFT - DO NOT CITE OR QUOTE
-------
AMR = AMR + DAYDOSE ! Initial dose of day/daily dose
ELSE
AMR = AMR + DRP(I)*DRDOSE ! Drinking percent
END IF
I = MOD((1 + 1) ,6)
IF (I.EQ.l) THEN
SCHEDULE ORALDOSE.AT.NEWDAY ! Go to start of the next day
NEWDAY = NEWDAY +24
ELSE
SCHEDULE ORALDOSE.AT.(NEWDAY+DRT(I)) ! Go to next drink time
END IF
END ! DISCRETE ORALDOSE
END
Dynamic
TERMINAL
DAILYBRAUC
DAILYBR2AUC
DAILYBAUC
DAILYB2AUC
Metrics calculated below!
AUCBR/(TSTOP/24) ! Mean daily stomach ACN
AUCBR2/(TSTOP/24) ! Mean daily stomach CEO
AUCB/(TSTOP/24) ! Mean daily blood ACN
AUCB2/(TSTOP/24) ! Mean daily blood CEO
PEAKPBR = MAX(0, CBR) ! Peak cone. parent compound in brain!
PEAKPB = MAX(0, CV) ! Peak cone. parent compound in blood!
PEAKMBR = MAX(0, CBR2) ! Peak cone. metabolite in brain!
PEAKMB = MAX(0, CV2) ! Peak cone. metabolite in blood!
Calc. of the avg. values of ACN and CEO in the brain and blood!
AVEBR = AUCBR/TSTOP
AVEBR2 = AUCBR2/TSTOP
AVEB = AUCB/TSTOP
AVEB2 = AUCB2/TSTOP
Average of ACN brain cone.!
Average of CEO brain cone.!
Average of ACN blood cone.!
Average of CEO blood cone.!
END ! TERMINAL
END ! PROGRAM
C-16
DRAFT - DO NOT CITE OR QUOTE
-------
Matlab source code - 3 programs (.m files)
% ACN / CEO PBPK model optimization file
O.
0
% Main program, or SHELL for PBPK model.
O.
0
% P = optACN2(pin, pvn, datfilename, switch)
O.
0
% swtch = optimization choice (see below)
% pin = values of *all* parameters (input)
% pvn = cell-vector of parameter names TO BE VARIED
% datfilename = name of data file (.csv) to use
% P = values of the output parameters (optimized)
o.
0
% where pin is the vector of paramters to input and swtch determines whether
% to run the simulation (swtch =0), or to optimize (swtch = 1 neld2
search).
o.
0
% pvn = {'namel' ; 'name2' ; ... } Note: use curly brackets and semi-colons
O.
Q
function P = optACN(pin,pvn,dfn,sw)
format long
% globalize data sets
global quan phys prs Ypt pv yname itr datfile pvar pbl jl jf ncv tsp pex
global ICs Ccafpt Ccafc nv Cost CO ps MW MW2 nv
% load dat set
quan.dat=load([dfn,'.csv1]);
swtch=sw;
%close all
% The sequence of columns in the data sets are as follows:
% tspan, CVB, CL, CBR, CVB2, CL2, CBR2, ODOSE(initial), IVDOSE(initial) ...
% % AN-GSH in urine, % CEO-GSH in urine, % Hb-bound material
yname={ 'CVB' ; 'CL' ; 'CBR' ; 'CVB2' ; 'CL2' ; 'CBR2' ; 'AGSHU' ; 'CGSHU' ; 'PBH ' ; 'MassBal' };
ncv=3 2;
% find indeces of parameters to be varied
pname={'VmaxC';'Km';'VmaxC2';'Km2';'kEHC';'kFC';'kFC2';'kA';'gammA';'gammC';
kH' ; 'kH2' } ;
%prs = [5.0; 1; l.e-18; 113.; 3.9; 73.; 500.; 8.; 1; 1] ;
%prs = [3.27; .000425; l.e-18; 113.; 3.9; 95.7; 500.; 8.; .0107; 1] ;
prs = [.102; 2.76; l.e-18; 113.; 0.98; 82.3; 14.1; 4.94; .0107; 0.941;
1.245; 1.134] ;
prs = [2.24; 2.76; l.e-18; 113.; 0.98; 82.3; 14.1; 4.94; .0107; 0.941;
1.245; 1.134] ;
%prs = [1.86; 1.5; l.e-18; 113.; 3.9; 15.7; 2240.; 8.;
%those above are default values
n'hl — / ' » . » » . » » . » » . » » . » » . » » . » » . » » . » » . » i
\ / ii ii ii ii ii ii ii ii ii ii
% KEHC = %Effective lst-order EH rate, liter/hr/kg^0.7
% VMAXC = %Maximum velocity of metabolism (mg/hr-lkg)%
% VMAXC2 = %Max. vel. of metabolism, CEO (mg/hr-lkg)%
% KM = %Michaelis-Menten constant (mg/L)%
% KM2 = %Michaelis-Menten constant, CEO (mg/L)%
% KFC = %ACN first order metab rate const (/hr-lkg)%
% KFC2 = %CEO first order metab rate const (/hr-lkg)%
% KA = %Oral uptake rate (/hr)%
.425; 1 . 10] ;
. T ' \ •
II I I
C-17
DRAFT - DO NOT CITE OR QUOTE
-------
fname='EqACN'; % file with set of model equations to use is fname.m
runfn='RunACN'; % Run file runfn.m
quan.ns = length(pname); pv = l:quan.ns; pvar = pname;
if length(pin)>quan.ns
error('input parameter vector is too long')
end
if sum(pin) % if pin =[] , then no optimization, use default params
popt = reshape(log(pin),length(pin) , 1) ;
if strmatch('allpvn,'exact')
if length(pin)~=quan.ns
error('length of pin is wrong; must be ',num2str(quan.ns),' to
match pvn = all')
end
prs=pin;
else
pvar = pvn;
quan.ns = length(pvn);
pv = zeros(1,quan.ns);
if (length(pin)~=quan.ns)&(length(pin)~=length(pname))
error(['Length of pin is wrong; must be ',num2str(quan.ns),...
' to match pvn or ', num2str(length(pname)), ' to match
pname.'])
else
for i = 1:quan.ns
if strmatch(pvn(i),pname,'exact')
pv(i) = strmatch(pvn(i),pname,'exact');
else
error([char(pvn{i}),' is not in the named list of
parameters.'])
end
end
end
if length(pin)==length(pname)
prs=pin; popt=popt(pv);
end
end
else
popt=log(prs); swtch=0; 'No pin so running simulation only with default
params.'
end
save pv pv
% find indeces of start times
quan.t = quan.dat (:,1) ;
tci = (quan.t > 0) ;
datc=quan.dat(:,[2:7,12:14]); % data columns used for fitting
nv = size(date, 2); % number of variables to plot, etc
Ccafc=zeros(length(quan.t),(nv+1));
nplt=0;
for i=l:6
quan.ic{i} = find((date(:,i)>0) & tci);
quan.datc{i} = date(quan.ic{i},i);
if quan.datc{i}
nplt=nplt+l;
end
quan.tc{i} = quan.t(quan.ic{i }) ;
quan.nc(i) = sum((date(:,i)>0) & tci);
end
nplt2=0;
for i=7:nv
C-18
DRAFT - DO NOT CITE OR QUOTE
-------
quan.ic{i} = find((date(:,i)>0) & tci);
quan.datc{i} = date(quan.ic{i},i);
if quan.datc{i}
nplt2=nplt2+l;
end
quan.tc{i} = quan.t(quan.ic{i }) ;
quan.nc(i) = sum((date(:,i)>0) & tci);
end
CO = sum(quan.nc)*(log(2*pi)+1) ;
quan.mc=datc>0;
sz = length(quan.t) ;
quan.yc=0*[date,date(:,1)];
quan.jt = find(quan.t == 0);
% jt = vector of indeces of rows in data file starting with "0"
quan.idf= length(quan.jt);
quan.jt = [quan.jt;(sz+1)];
% add a pseudo-index to mark the end of the data file
% constant parameters
%ACN second order rxn with GSH (L/mMol/hr)%
%Liver GSH cone (mMol/L)%
%Brain GSH (mMol/L)%
%Stomach GSH (mMol/L)%
%Slowly perfused (muscle) GSH (mMol/L)%
%Rapidly perfused GSH (mMol/L)%
ps.KSO = 0.2584;
ps.GSHL = 8.53;
ps.GSHBR = 2 . ;
ps.GSHST = 4.59;
ps.GSHS = 0.75;
ps.GSHR = 2.65;
%ps.KB
ps.KB0
ps.KB0
%ps.KH
ps.KFB
= 1.245 + 2.54;
£ACN 1ST order binding to blood hb + RSH (/hr)
= 2.54;
%ACN 1ST order binding to blood hb + RSH (/hr)%
1.245;
2 . 54 ;
%ps.KB2 = 1.134 + 0.68 + 0.413;
PS.KB2 0 = 0.6 8 + 0.413;
%ps.KH2 = 1.13 4;
phys.KHR = 1.134/1.245;
ps.KFB2 = 0.68;
%CEO 1ST order binding to blood hb + RSH + chem. hydrol. (/hr)
ps.tinf=.003; %blood infusion time (hr)
% ACN PCs
phys.PL = 0.46;
phys.PST = 0.46;
phys.PBR = 0.40;
phys.PF = 0.2 8
phys.PS = 0.3 5
phys.PR = 0.46
phys.PB = 512;
liver:blood PC
stomach tissue:blood PC, also used for GI
brain tissue:blood PC
fat tissue:blood PC
slowly tissue:blood PC
richly tissue:blodd PC
blood tissue:blodd PC
%CEO PCs
phys.PL2 = 0.2 74;
phys.PST2 = 0.2 74;
phys.PBR2 = 1.4 07;
phys.PF2 = 0.785,
phys.PS2 = 1.853
phys.PR2 = 0.2 74
phys.PB2 = 16 58;
's liver:blood PC
d stomach tissue:blood PC,
's brain tissue :blood PC
fat tissue:blood PC
slowly tissue:blood PC
's richly tissue :blodd PC
blood tissue:blodd PC
also used for GI
QPC = 14.; %Alveolar ventilation rate (L/hr)%
QCC = 14.; %Cardiac output (L/hr)%
C-19
DRAFT - DO NOT CITE OR
-------
QBRC
= 0.024;
% blood flow
to brain
QSTC
= 0.013;
% Blood flow
rate for stomach tissue (portal
QFC
0.09;
% blood flow
to fat tissue
QLC
0.25;
% Blood flow
rate for liver )
QSC
0.15;
% Blood flow
rate for slowly
phys
BW=0.25;
%body weight
VLC
0.04;
%Fraction of
liver tissue to total body%
VRC
0.0377;
%Fraction of
richly perf tiss to total%
VSC
0.75;
%Fraction of
sloly perf tiss to total%
VBRC
= 0.006;
%Fraction of
brain tissue%
VFC
0.07;
%Fraction of
fat tissue%
VSTC
= 0.0063;
%Fraction of
stomach tissue%
BV =
0 . 0 6*phys
.BW; %Blood vol%
WBC
= 0.65;
%Fraction of
venous blood vol%
MW = 53 . 0 6 ;
MW2 = 6 9.05;
%ACN molecular weight (g/mol)%
%CEO molecular weight (g/mol)%
phys
phys
phys
phys
vein
phys
phys
phys
phys
phys
phys
phys
phys
phys
phys
phys
phys
phys
phys
phys
phys
phys
phys
preset varying parameters
.QC = QCC*phys.BW^O.74;
.QP = QPC*phys.BW^O.74;
.QBR = QBRC*phys.QC; % blood flow to brain
.QST = QSTC*phys.QC; % Blood flow rate for stomach tissue (portal
QS = QSC*phys.QC
QL = QLC*phys.QC
QF = QFC*phys.QC
Blood flow rate for slowly
Blood flow rate for liver (hepatic artery)
blood flow to fat
QR = phys.QC-phys.QL-phys.QF-phys.QBR-phys.QS-phys.QST;
VBR = VBRC*phys.BW;
VST = VSTC*phys.BW;
VL = VLC*phys.BW;
VF = VFC*phys.BW;
VBR = VBRC*phys.BW;
VST = VSTC*phys.BW;
VS = VSC*phys.BW;
VR = VRC*phys.BW;
VAB = BV* (1 -WBC) ;
WB = BV* WBC ;
GSHRB = 0.l*ps.GSHBR/ps.GSHL;
GSHRST = 0.l*ps.GSHST/ps.GSHL;
GSHRS = 0.l*ps.GSHS/ps.GSHL;
GSHRR = 0.l*ps.GSHR/ps.GSHL
brain volume (L)
GI volume (use RP density)
Ivolume arterial blood%
%ratio of brain GSH*GST to liver GSH%
%ratio of stomach GSH*GST to liver GSH%
%ratio of spt GSH*GST to liver GSH%
>ratio of rpt GSH*GST to liver GSH%
ICs= [] ;
for i=l:quan.idf % idf is the number of simulations/initial conditions in
j1(i) = quan.jt(i); % index of initial data row for this
experiment/simulation
jf(i) = quan.jt(i+1)-1; % index of final data row for this
expt./simulation
tsp{i} = quan.t(j1(i):jf(i));
oralO=quan.dat(j1(i),8)*phys.BW;
ps.bloodO(i)=quan.dat(j1(i),9)*phys.BW/ps.tinf;
ps.cinh(i)=quan.dat(jl(i) , 10)*MW/24450; ps.tinh(i)=quan.dat(j1(i),11) ;
ICs=[ICs;[oralO,zeros(1,ncv-1)]]; % initial conditions
end
save ICs ICs
% call simulation or optimization procedure-
C-20
DRAFT - DO NOT CITE OR QUOTE
-------
quan.pt=0; itr=0; P=prs;
if swtch==0
quan.pt=l;
y=eval([runfn, ' (popt) ' ] ) ;
elseif swtch==l
pnew = exp(neld2(popt, runfn, le-6, 1000, 50000));
else
options = optimset('MaxFunEvals',8000,'Maxlter',8000);
pnew = exp(fminsearch(runfn,popt,options));
end
if swtch>0
P(pv) = pnew(l:quan.ns);
'Initial Parameter Values, Jinit ='
Cst=eval([runfn,'(popt)' ] ) ;
'Final Parameters, Jopt ='
quan.pt=l;
Cst=eval([runfn,'(log(P(pv)))']);
end
% final results/plots
co =
['rbkgmcrbkgmcrbkgmcrbkgmcrbkgmc rbkgmc rbkgmc rbkgmc rbkgmc rbkgmc rbkgmc rbkgmc rbk
gmcrbkgmcrbkgmcrbkgmc' ] ;
%co = ['kkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkk' ] ;
PC=„
['+o*xsdAv< >ph+o*xsd*v< >ph+o*xsdAv< >ph+o*xsd*v< >ph+o*xsdAv< >ph+o*xsd*v< >ph+o*
xsdAv< >ph+o*xsd*v< >ph']
linet = ['-
] ;
np=min(quan.idf, 96) ;
quan.jtp = [find(quan.tp == 0);(length(quan.tp)+1)];
% jt = vector of indeces of rows in data file starting with "0"; plus one
%close all
kf0 = 3 0 ;
kf=l+kf0;
figure(kf)
rows=ceil(nplt/2);
set(kf,'Units','inches','Position',[0.1, 0.5, 6.3, (0.1+2.5*rows)]);
set(gcf,'DefaultTextFontSize',10,'DefaultAxesFontSize',12,'DefaultTextFontNam
e', 'Ar i a1');
npt = -1;
for k = 1:6
if quan.datc{k}
npt=npt+2;
pos=mod(npt,(rows*2))+(npt>(rows*2));
h=subplot(rows,2,pos);
set(h, 'Units', 'inches', 'Position', [(3.5-
2.95*mod(pos, 2)) , (0.45 + 2.35*(rows-ceil(pos/2))) ,2.7,2.1] )
box on
jl=quan.jt (1) ;
j 2=quan.j t (2)-1 ;
pl=quan.jtp(1);
p2=quan.j tp(2)-1;
C-21
DRAFT - DO NOT CITE OR QUOTE
-------
dt=quan.t(j1:j 2) ;
dp = date(j1:j2,k);
dj = find(dp > 0) ;
dk=find(date(:,k)>0) ;
xm=ceil(1.01*max(quan.t( dk ))*10)/10;
xl=floor(0.99*min(quan.t( dk )*10)-1)/10;
if xl<0
xl = - round(2.5*xm)/100;
end
ym=max(loglO(1.1*[Ccafpt(find(Ccafpt(: , k)>0) , k) ;datc(dk,k)])) ;
if (lO^ceil(ym))/(lO^ym) > 2
ym = (lO^ceil(ym))/2;
else
ym = lO^ceil(ym);
end
set(gca,'YScale','log')
if k==l
axis([xl,xm,.03,ym])
end
if k==2
axis([xl,xm,.01,ym])
end
if k==3
axis([xl,xm,.01,ym])
end
if k==4
axis([xl,xm,.0 01,ym])
end
if k==5
axis([xl,xm,.0 01,ym])
end
if k==6
axis([xl,xm,.0 01,ym])
end
hold on
if dj
plot(dt(dj),dp(dj), ['k',pc (1)]) ; %, [co(1),pc (1)])
plot(quan.tp(pi:p2),Ccafpt(pi:p2,k), ['k',linet(1, :)]);
%[co(1) , linet(1, :) ] ) ;
end
i f np > 1
for i = 2:np
jl=quan.j t (i) ;
j2=quan.jt(i+1)-1;
dt=quan.t (j1: j 2) ;
pl=quan.jtp(i);
p2=quan.jtp(i+1)-1;
dp = date(j1:j2,k);
dj = find(dp > 0) ;
if dj
plot(dt(dj),dp(dj), ['k',pc (i)]) ; %, [co(1),pc (1)])
plot(quan.tp(pi:p2),Ccafpt(pi:p2,k),['k',linet(i,:)]);
%[co(1) , linet(1, :) ] ) ;
end
end
end
yl=0.l*get(gca,'YLim');
title(yname(k),'Position',[mean(get(gca,'XLim')),yl(2)],'FontSize',14,
'EdgeColor', 'k', 'BackgroundColor' , 'w' , 'Fontname' , 'Arial') ;
C-22 DRAFT - DO NOT CITE OR QUOTE
-------
if k= = 2
legend('show')
end
hold off
end
end
kf=2+kf0;
if quan.datc{7}
figure(kf);
set(kf, 'Units','inches','Position', [0.1, 0.5, 3.3, 7.6]);
set(gcf,'DefaultTextFontSize',14,'DefaultAxesFontSize',12,'DefaultTextFontNam
e', 'Ar i a1');
for k = 7:9
h=subplot(3,1,k-6);
set(h, 'Units', 'inches', 'Position', [.55, (0.45 + 2.35*(9-k)),2.7,2.1])
box on; hold on
dk=find(date(:,k)>0);
plot(quan.dat(dk-1,8),date(dk,k), ['k',pc(k-6)]) ; %, [co(1),pc (1)])
plot(quan.dat(dk-1,8),Ccafc(dk,k), ['k',linet(k-6, :) ] ) ;
%[co(1),linet (1, :)]) ;
yl=0.3*get(gca,'YLim');
title(yname(k),'Position',[mean(get(gca,'XLim')),yl(2)],'FontSize',14,
'EdgeColor', 'k', 'BackgroundColor' , 'w' , 'Fontname' , 'Arial') ;
hold off
end
end
k=nv+l; % show mass balance
figure(k)
set(k, 'Units', 'inches', 'Position', [1,1,4.1,4.1]) ;
hold on
title(yname(k));
pl=quan.jtp(1);
p2=quan.j tp(2)-1 ;
plot (quan.tp(pi:p2),Ccafpt(pl:p2,k) , [co(l),linet(1, :)]);
for i = 2:np
pl=quan.jtp (i) ;
p2=quan.jtp(i+1)-1;
plot(quan.tp(pi:p2),Ccafpt(pl:p2,k), [co(i),linet (i, :)]);
end
if swtch>0
'optimized params = '
[num2str([1:length(P)]'),char(pbl),char(pname),char(pbl),num2str(P)]
end
%RunACN file to return value of objective function, given input parameters
function Cost = RunACN(popt)
% globals
global phys prs yname quan itr pv pvar pbl jl jf ian ICs ncv tsp Ccafpt CO
global Cost Ccafc MW MW2 VMAX VMAX2 KF KF2 KEH KFBRC KFSTC KFSC KFRC KFBR
global KFST KFS KFR KA KM KM2 Ccafpt nv ps
tol = l.e-8 ;
options = odeset('RelTol',tol,'AbsTol',tol);
itr=itr+l;
C-23
DRAFT - DO NOT CITE OR QUOTE
-------
comp=0;
p = reshape(exp(popt),length(popt),1);
pu=prs;
pu(pv)=p;
save ptemp pu
[num2str(pv'),char(pbl(pv)),char(pvar),char(pbl(pv)),num2str(p)]
%quan.gam(1:5)=2./exp(exp(-popt((quan.ns+1):(quan.ns+5)) ));
%quan.gam(6:11)=quan.gam(5) ;
%quan.gam(1:5)
% display current values of parameters being valued (retransformed)
Ccafpt= [] ;Ccafw= [] ;
quan . tp= [] ;
comp=comp+l ;
VMAXC = pu(1);
KM = pu (2) ;
VMAXC2=pu(3)
KM2=pu (4) ;
KEHC = pu (5)
KFC = pu (6 ) ;
KFC2 = pu(7)
KA = pu(8) ;
gA = min(pu(9),2) ;
gC = min(pu (10),2);
%Max. vel. of metabolism, CEO (mg/hr-kg^0.7)%
%Michaelis-Menten constant (mg/L)%
%Max. vel of metabolism, CEO (mg/hr-jg^O.7)
%CEO M-M constant (mg/L)
%Effective lst-order EH rate, liter/hr/kg^O.7
%ACN first order metab rate const (/hr-lkg)%
%CEO first order metab rate const (/hr-lkg)%
%Oral uptake rate (/hr)%
% gammA
% gammC
ps.KH = pu(11);
ps.KHR2 = pu(12) ;
ps.KB = ps.KH + ps.KBO; %ACN 1ST order binding to blood hb + RSH (/hr)%
ps.KH2 = ps.KH*phys.KHR*ps.KHR2;
ps.KB2 = ps.KH2 + ps.KB2 0
% VMAX = VMAXC*(phys.BW^O. 7) ;
VMAX2 = VMAXC2*(phys.BW^O. 7) ;
KF = KFC/(phys.BW^O.3);
KF2 = KFC2/(phys.BW^O. 3 ) ;
KEH = KEHC*(phys.BW^O. 7) ;
%CEO rxn rates%
KFBR = KF2*phys.GSHRB; %
KFST = KF2*phys.GSHRST;
KFS = KF2*phys.GSHRS;
KFR = KF2*phys.GSHRR;
%Liver P450 ACN to CEO%
%Liver CEO Hydrolysis!
%Liver ACN-GSH rate%
%Liver CEO-GSH/RSH rate%
% Approx. first-order rate in liver
ICEO brain RSH rate%
ICEO stomach RSH rate%
ICEO SPT RSH rate%
ICEO RPT RSH rate%
% loop for running equations
% figure(15)
% set(15, 'Units', 'inches', 'Position', [1,1,4.1,4.1]) ;
% hold on
% title('RAMI');
for i=l:quan.idf % idf is the number of simulations/initial conditions in
% the data file this was computed in optACN.m
ian=i;
tsc = [0 :1050] '*quan.t(jf(i) )/1000 ;
%ICs (i, [1,16])
[ts,Ys] = odel5s(@EqACN2,tsc,ICs(i,:),options); %
% [tS (1:4) ,Ys(1:4,1:16) ] '
%[Ys(1:10,13) ,VMAX*Ys(1:10,13) ./ (KM+Ys (l:10,13)),(Ys(2:ll,15)-Ys(l:10,15))]
massb=sum(Ys(:,1:16),2) ;
% plot(tS,Ys(:,15)); %VMAX*Ys(:,13)./(KM+Ys(:,13)));
% The sequence of columns in the data sets are as follows:
C-24
DRAFT - DO NOT CITE OR QUOTE
-------
% tspan, CVB, CL, CBR, CVB2, CL2, CBR2, ODOSE(initial), IVDOSE(initial)
Yp= [Ys ( : , 16 )/phys . WB, Ys ( : , 13 ) /phys . VL, Ys ( : , 11)/phys . VBR, ...
Ys ( : , 24) /phys ,WB, Ys ( : , 23 )/phys . VL, Ys ( : , 22 )/phys . VBR, ...
Ys(: , 30:32)/phys.BW, massb];
Ccafc(j1(i) :j f(i), :)=interpl(ts,Yp,tsp{i}) ;
quan.tp=[quan.tp;ts];
Ccafpt=[Ccafpt;Yp];
Ccafw=[Ccafw; [tsp{i},Ccafc(j1(i) :j f (i), :) ] ] ;
% encj loop
end
o.
Q
if quan.pt==l
yname = {'AN in blood'; 'AN in liver'; 'AN in brain'; ...
'CEO in blood'; 'CEO in liver'; 'CEO in brain'; ...
'AN-GSH in urine'; 'CEO-GSH in urine'; 'Hb binding'; 'mass
balance' } ;
f id= fopen(['ACN.txt'] , 'w') ;
fprintf(fid, ['time' , '\t' , 'CVB', '\t' , 'CL' , '\t' , 'CBR' , '\t' , . . .
'CVB2', '\t' , 'CL2' , '\t' , 'CBR2 ' , ' \t' , 'mball', '\r'] ) ;
fprintf(fid,'%g\t%g\t%g\t%g\t%g\t%g\t%g\t%g\r',Ccafw');
%[quan.tp,Ccafpt]');
status=fclose(fid);
end
%Ccafpt=Ccafc;quan.tp=quan.t; % uncomment this line to plot only the fitted
lvalues
% cost function
Cost=C0+max(0,pu(9)-2)+max(0,pu(10)-2);gamm=[gA gA gA gC gC gC gC gC gC];
for i=l:nv
if quan.datc{i}
Cost = Cost + gamm(i)*sum(quan.date{i}) + ...
quan.nc(i)*log( sum( ((quan.date{i} - Ccafc(quan.ic{i},i)).*2)./...
(Ccafc(quan.ic{i},i).^gamm(i)) )/quan.nc(i) );
end
end
['itr = ',num2str(itr) , '; Cost = ',num2str(Cost)]
% ACN PBPK model equation file -- returns derivatve dx = dx/dt, given
% state variable x and parameters passed through global statements
o.
0
% function defined
function dx = eqs(t,x)
% globals
global prs phys ICs quan ps ian MW2 MW CINH TINH
global VMAX VMAX2 KF KF2 KEH KFBRC KFSTC KFSC KFRC KFBR KFST KFS KFR KA KM
KM 2
% assign parameters to names
dx=zeros(size(x));
% "dx" = dx/dt, where x is the state vector
ACRYLONITRILE
C-25
DRAFT - DO NOT CITE OR QUOTE
-------
% MR = x(l); %MR = Amount in stomach lumen (mg)!
% AB = x(2); %Amount in systemic arterial blood (mg)!
CA = x(2)/phys.VAB; %concentration
%AX = x(3); % ACN exhaled (mg)!
%AST = x(4); % ACN in stomach tissue tissue (mg)!
CST = x(4)/phys.VST; CVST = CST/phys.PST; % tissue/venous concn's
%ASTG = x(5); % ACN GSH conjugated in stomach tissue (mg)
%AS = x(6); % ACN in slowly perfused (mg)
CVS = x(6)/(phys.VS*phys.PS); % venous concentration
%ASG = x(7); % ACN-GSH conjugated in slowly (mg)
%AR = x(8); % ACN in richly (mg)
CVR = x(8)/(phys.VR*phys.PR); % venous concentration
%ARG = x(9); % ACN-GSH conjugated in richly
%AF = x(10); % ACN in fat (mg)
CVF = x(10)/(phys.VF*phys.PF); % venous concentration
%ABR = x(ll); % ACN in brain (mg)
CBR = x(11)/phys.VBR; CVBR = CBR/phys.PBR; % tissue/venous concn's
%ABRG = x(12); % ACN-GSH conjugated in brain
%AL = x(13); % ACN in liver (mg)
CL = x(13)/phys.VL; CVL = CL/phys.PL; % tissue/venous concn's
% ALG = x(14); ACN-GSH conjugated in liver (mg)
% AMI = x(15); ACN metabolized by P450 in liver (mg)
% AVB = x(16); ACN in mixed venous blood (mg)
CVB = x (16)/phys .WB; % concentration
RMR = KA*x(l); % rate of absorption from stomach
CI=ps.cinh(ian)*(t<=ps.tinh(ian)); %Inhaled concentration
CAL = (phys,QC*CVB+phys,QP*CI)/(phys.QC+(phys.QP/phys.PB));
%CAL = Concentration in arterial lung blood (mg/L)!
%CX = CAL/phys.PB; % concentration in exiting pulmonary air
% CXPPM = (0.7*CX+0,3*CI)*24450./ps.MW
%RAB = ps.KB*x(2)CA*phys.VAB; % ACN binding in arterial blood
STGSH = ps.KSO*CVST*ps.GSHST*phys.VST; % stomach GSH conjugation of ACN
SGSH = ps.KSO*CVS*ps.GSHS*phys.VS; % slowly GSH conjugation of ACN
RGSH = ps.KSO*CVR*ps.GSHR*phys.VR; % richly GSH conjugation of ACN
BRGSH = ps.KSO*CVBR*ps.GSHBR*phys.VBR; % brain GSH conjugation of ACN
LGSH = ps.KSO*CVL*ps.GSHL*phys.VL; % liver GSH conjugation of ACN
% ACN metabolized,saturable (P450) and linear pathways (mg)!
RAMI = VMAX*CVL/(KM+CVL); % (mg/hr)
%VMAX,KM,RAMI
%ACN metabolized,first-order pathway (GST) (mg)!
RAM2 = KF*CVL*phys.VL; % (mg/hr)
RTV =
phys,QF*CVF+(phys.QL+phys.QST)*CVL+phys.QS*CVS+phys.QR*CVR+phys.QBR*CVBR;
% mixed venous blood concentration
%RVB = ps.KB*CV*phys.WB; % ACN binding in venous blood
%[t,CVB/(RTV/phys.QC),CAL/CVB,CA/CAL,x(1),RMR]
ACN stomach lumen
ACN stomach tissue
dx (1)
= - RMR; %
dx (2)
= phys.QC*
dx (3)
= phys.QP*'
dx (4)
= phys.QST
dx (5)
= STGSH; %
dx (6)
= phys.QS*
dx (7)
= SGSH; % ,
dx (8)
= phys.QR*
dx (9)
= RGSH; % ,
dx(10)
= phys.QF*
dx(ll)
= phys.QBR
«3SH; % ACN in richly (mg)
jugated in richly
ACN in fat (mg)!
- BRGSH; % ACN in brain (mg)!
C-26
DRAFT - DO NOT CITE OR QUOTE
-------
dx(12) = BRGSH; % ACN-GSH conjugated in brain
dx(13) = phys.QL*CA + phys.QST*CVST -(phys.QL+phys.QST)*CVL -RAMI -RAM2 -
LGSH;
dx(14) = LGSH + RAM2 + ps.KB* (x(2)+x(16) ) ;
% ACN-GSH conjugated in liver + bound in blood
dx(15) = RAMI; % ACN metabolized by P450 in liver
dx(16) = RTV - phys.QC*CVB - ps.KB*x(16) + ps.bloodO(ian)*(t<=ps.tinf);
%Mixed ven. ACN cone after binding (mg/L)
dx(3 0) = STGSH + SGSH + RGSH + BRGSH + LGSH + RAM2 + ps.KFB* (x(2)+x(16) ) ;
% amount of AN binding to GSH
% [x(1 : 16) ,dx(1:16)] '
CEO
%AB2 = x(17); % CEO in arterial blood (mg)
CA2 = x(17)/phys.VAB; % concentration
%AS2 = x(18); % CEO in slowly perfused (mg)
CVS2 = x(18)/(phys.VS*phys.PS2); % venous concn
SGSH2 = KFS*CVS2*phys.VS;
%AR2 = x(19); % CEO in richly (mg)
CVR2 = x(19)/(phys.VR*phys.PR2); % venous concn
RGSH2 = KFR*CVR2*phys. VR;
%AST2 = x(20) ; % CEO in stomach tissue tissue (mg) !
CST2 = x(20)/phys.VST; CVST2 = CST2/phys.PST2; % tissue/venous concn's
STGSH2 = KFST*CVST2*phys.VST;
%AF2 = x(21); % CEO in fat (mg)
CVF2 = x(21)/(phys.VF*phys.PF2); % venous concn
%ABR2 = x(22); % CEO in brain (mg)
CBR2 = x(22)/phys.VBR; CVBR2 = CBR2/phys.PBR2; % tissue/venous concn's
BRGSH2 = KFBR*CVBR2*phys.VBR;
%AL2 = x(23); % CEO in liver (mg)
CL2 = x(23)/phys.VL; CVL2 = CL2/phys.PL2; % tissue/venous concn's
% AVB2 = x(24); % Mixed ven. CEO cone after binding!
CVB2 = x (24)/phys .WB; % concentration
RALG2 = (VMAX2*CVL2/(KM2+CVL2)) + KF2*CVL2*phys. VL;
% CEO hydrolysis, saturable + linear terms
% For fitting, either saturable or linear is set to zero
CV2 = (phys,QF*CVF2 + (phys.QL+phys.QST)*CVL2 + phys.QS*CVS2 + phys.QR*CVR2
+ phys,QBR*CVBR2)/phys.QC; % CEO mixed venous blood cone. (mg/L)
CAL2 = (phys.QC*CVB2)/(phys.QC+(phys.QP/phys.PB2)); % CEO in arterial lung
blood (mg/L)!
dx(17) = (phys.QC*CAL2)-(phys.QC*CA2)-(ps.KB2*x(17));
% CEO in systemic arterial after binding
dx(18) = phys.QS*(CA2-CVS2) - SGSH2; % CEO in slowly perfused tissues (mg)!
dx(19) = phys.QR*(CA2-CVR2) - RGSH2; % CEO in rapidly perfused tissues (mg)!
dx(2 0) = phys.QST*(CA2-CVST2) - STGSH2; % CEO in stomach (mg)!
dx(21) = phys.QF*(CA2-CVF2); % CEO in fat tissue (mg)!
dx(22) = phys.QBR*(CA2-CVBR2) - BRGSH2; % CEO in brain tissue (mg)!
dx(23) = phys.QL*(CA2-CVL2) + phys.QST*(CVST2-CVL2) - KEH*CVL2*phys.VL + ...
(RAM1*MW2/MW) - RALG2;
dx(24) = phys.QC*CV2 - phys.QC*CVB2 - ps.KB2*x(24);
% CEO mixed venous after binding
%Calculation of the AUC for ACN and CEO in liver, brain, and blood!
C-27
DRAFT - DO NOT CITE OR QUOTE
-------
dx(25) = CL; % AUC for ACN liver concentration
dx(26) = CBR; % AUC for ACN brain cone.!
dx(27) = CBR2; % AUC for CEO brain cone.!
dx(28) = CVB; % AUC for ACN blood cone.!
dx(29) = CVB2; % AUC for CEO blood cone.!
dx(31) = STGSH2 + SGSH2 + RGSH2 + BRGSH2 + RALG2 + ps.KFB2*(x(17)+x(24));
% amount of CEO binding to GSH
dx(3 2) = (ps.KH* (x(2)+x(16) ) ) + (ps.KH2* (x(17)+x(24) ) ) ;
% had problems with state variables going < 0 at one point (stiff system);
% the following is a fix for this.
%dx=(dx.*(x>=0)) + (abs (dx) .*(x<0)) ;
C-28
DRAFT - DO NOT CITE OR QUOTE
-------
APPENDIX D. UNCERTAINTIES ASSOCIATED WITH CHEMICAL-SPECIFIC
PARAMETERS EMPLOYED IN THE PBPK MODEL FOR AN DOSIMETRY
IN HUMANS
PBPK models are computational tools used to predict chemical/drug disposition. Models
are comprised of three distinct types of information: physiological, physicochemical, and
biochemical. The physiological data are chemically independent and describe such parameters
as organ volumes and blood flows. Physicochemical parameters are chemical specific and
specify parameters, such as PCs or permeability. Biochemical parameters define the rates of
chemical transformation or binding. In Appendix D, the EPA evaluates uncertainties in the
chemical-specific parameters employed in the PBPK model used to predict AN dosimetry in
humans, which is adapted from that of Sweeney et al. (2003). Only a small number of
(metabolic) parameters were changed in the EPA's adaptation, which will be noted below.
Otherwise, it should be understood that any discussion of the model of Sweeney et al. (2003)
applies to the one used in this assessment.
PCs
PCs describe the extent of distribution of chemical into the body, including target tissues
such as the brain and lungs. PCs employed in the model of Sweeney et al. (2003) were derived
from rat tissue:air PCs and human blood:air PC for AN or CEO in the following way:
Ptissue:blood (human) —
P tissue:air(l"at)/ P blood:air(human)
This equation assumes that the relative difference in PCs between tissues is the same in
rats as in humans. The systematic difference between the two is accounted for by the blood:air
PC; for humans, the experimentally determined value for AN (154) differed from the value for
the rat (512) by threefold. Due to the absence of experimental data, Sweeney et al. (2003) set the
human blood:air PC for CEO equal to the rat value (1,658).
It may be asked whether the estimates produced by the above equation are reasonable
considering the chemical characteristics of AN and CEO and the distribution of AN- and
CEO-soluble components in human tissue. To address these issues, the method of Poulin and
Theil (2002) was used to estimate the PtiSSUe:biood(human). The method is based on the additive
solubilities of a chemical in lipid and water, and the relative distribution of these substances in
tissues. Sensitivity analysis conducted by Sweeney et al. (2003) identified the Pst0mach:biood and
Pbrain:biood for AN as influential on AN-related dose metrics and the Pbrain:biood for CEO as
influential on CEO-related dose metrics. Table D-l provides estimates of PCs for AN and CEO.
Note that the critical PCs, the Pstomach:biood for AN and the Pbrain:biood for AN and CEO, as
predicted by the method of Poulin and Theil were within a factor of about 1.5 of the values
employed by Sweeney et al. (2003).
D-l
DRAFT - DO NOT CITE OR QUOTE
-------
Table D-l. Tissuerblood PCs
PCs for AN
PCs for CEO
Compa
Expb
Exp:comp ratio
Compa
Expb
Exp:comp ratio
Adipose tissue
0.30
0.94
3.11
Adipose tissue
0.22
0.79
3.52
Brain
1.12
1.34
1.19
Brain
0.99
1.40
1.41
Stomach
1.00
1.51
1.52
Stomach
0.89
0.27
0.31
Liver
1.02
1.51
1.48
Liver
0.94
0.27
0.31
Muscle
0.98
1.16
1.19
Muscle
0.83
1.84
1.97
Richly perfused
1.00
1.51
1.51
Richly perfused
0.97
0.27
0.28
"Computationally derived by Poulin and Theil (2002).
Experimentally obtained/reported by Sweeney et al. (2003).
To evaluate the influence of the PC in the model, the revised human PBPK model was
implemented with both sets of PCs (Table D-l, with Pbiood:air as measured for AN using human
blood and for CEO using rat blood). For three different exposure scenarios, the peak AN and
CEO did not change by more than 30% of initial value (Table D-2). This finding may be
explained by consistency between both estimates of PC and a general lack of sensitivity to the
PC (see Sweeney et al., 2003). Indeed, the only PCs with normalized sensitivity coefficients that
exceeded 0.2 were the Pst0mach:biood (AN), Pbrain:biood (AN), and the Pbrain:biood (CEO). Consistent
with these results, Table D-2 shows a greater impact of changing the PC in the brain than the
blood (an averaged effect of other PCs). All of the predicted concentration in brain tended to
decrease when the model was implemented with PC as per the method of Poulin and Theil
(2002).
D-2
DRAFT - DO NOT CITE OR QUOTE
-------
Table D-2. Impact of method of estimation of PCs on the peak AN and CEO
model predictions
Exposure
PC method"
Peak AN (jtg/L)
Peak CEO (jtg/L)
Blood
Brain
Blood
Brain
Inhalation
2 ppm, 8 h
S
4.87
11.4
0.998
1.23
PT
4.87
9.56
0.998
0.863
0.0%
-16.4%
0.0%
-29.6%
Inhalation
0.4 ppm, 1 wk
S
0.975
2.29
0.200
0.245
PT
0.975
1.91
0.200
0.173
0%
-16.4%
0%
-29.6%
Oral
6 pulses in 24 h,
0.2 mg/kg total
S
2.50
3.11
8.36
10.14
PT
2.68
2.77
8.64
7.44
7.3%
-10.9%
3.4%
-26.7%
aS = PCs as estimated by Sweeney et al. (2003) (listed in Appendix C); PT = PCs estimated using the method of
Poulin and Theil (2002). AN and CEO are from model simulations using other human parameters as listed in
Appendix C.
Parameters of metabolic clearance
Sweeney et al. (2003) developed a PBPK model of AN and CEO disposition in humans
based on human in vitro data and the rat model of Kedderis et al. (1996). Human in vivo
pharmacokinetic data were not available for model development; therefore, the authors proposed
a rodent-human parallelogram approach to consider uncertainties in the scaling of in vitro
metabolism data. Critical metabolic pathways included the oxidation of AN to CEO, the
conjugation of AN and CEO, and the hydrolysis of CEO.
With the exception of the latter, the scaling of in vitro rate constants for these pathways
was first investigated by Kedderis et al. (1996), who observed that the scaled rate constants (rat
model) derived from rat liver microsomes, did not match those constants fit from in vivo rat data.
Sweeney et al. (2003) employed both in vitro and in vivo data to derive an empirical correction
factor (CF) to account for uncertainties in scaling (rat). The scaling constant was assumed to be
constant across species; thus, the preexisting rat model, the empirical CF, and human in vitro
data were employed to specify a human PBPK model.
Because EPA altered a number of the rat metabolic parameters to accommodate the rate
constant for AN hydrolysis extrapolated from in vitro data and to obtain parameters in a way that
was computationally reproducible, while following the same approach as Sweeney et al. (2003)
for metabolic parameter extrapolation, the EPA obtained different values for humans. First, we
describe the derivation of the human Vmaxc for the oxidation of AN to CEO in the following
equations. The ratio of rates determined by human in vitro data and rat in vitro data is
considered (Kedderis et al. 1993c), as well as the allometrically scaled in vivo rate constant for
oxidation of AN in the rat.
D-3
DRAFT - DO NOT CITE OR QUOTE
-------
Vmaxc(human) VmaxC(rat) 0.3 . n _ \
(mg/h/kg07) = (mg/h/kg0 7)x| BWhuman | x[ [mgMSP /g liver]human jxf VLChuman j x VmaX|human
in vivo in vivo ^ BWrat J ^ [mgMSP / g liver]rat J [ VLCrat J [vmax,human
in vitro
22.1 = 7.1 x 5.422 x 1.4225 x 0.6425 x 0.626
Next, the value (and calculations) of the empirical CF = 5.785 is shown. This value represents
what is not accounted for in the in vitro to in vivo scaling in the rat.
CFin vivo i in vitro = (mg/h/kg01)
Wnax C
(rat) /(
Wnax C {rat) 103 ^
(mg I h I kg0 7) x [mgMSP / g liver]rat x — x VLCrat x BWrat0 3
• v k9
in vitro
in vivo
5.785 = 7.1 /(0.001164 x 40 x 1000 x 0.04 x 0.659)
The allometrically scaled in vivo rate constant for human pseudo first-order GSH
conjugation of AN and CEO was calculated in the same way (not shown).
The parallelogram method was applied differently to estimate the rate constant for
enzymatic hydrolysis by EH in humans. Because the EPA extrapolated the EH rate constant in
rats from in vitro to in vivo without use of an adjustment factor, the same was done for humans.
Kedderis and Batra (1993) determined a Vmax and Km EH-mediated hydrolysis of CEO using
liver microsome samples from six individual humans. The lowest estimated Km in the group was
600 |iM, so the EPA chose to describe the metabolism as first-order since in vivo concentrations
are expected to stay well below that value, using the ratio of Vmax/Km. The ratio of Vmax/Km was
first calculated for each individual since Vmax and Km tend to be statistically correlated due to the
way they are estimated, and an average value for the ratio was then determined to be 7.02 x 10"6
L/minute/mg MP. We can then apply the value of 56.9 mg MP/g liver from Lipscomb et al.
(2003), the liver fraction of 25.7 g/kg BW, and the standard value of 70 kg BW for a human.
The rate constant for a standard human is then l
-------
in-vitro-derived value came out to about one half of an in-vivo-fitted value. Thus, it appears that
a CF is appropriate to account for either or both of these factors. Sweeney et al. (2003) applied
the same approach to the enzyme-catalyzed GSH conjugation (cytosolic in origin). One concern
that had existed is that Sweeney et al. (2003) had been forced to assume that the same factor
holds for distinct enzymes (P450 and EH) that are membrane bound, because EH activity in the
rat was taken to be zero in the original model of Kedderis et al. (1996), but that assumption is no
longer required since the EPA was able to extrapolate the EH-mediated activity without
adjustment. The EPA must still assume that the factors for P450 and GST hold across species,
but, given that the in-vitro enzyme preparations and measurements of activity were made in the
same way for both animals and humans, this seems a reasonable assumption. Nevertheless, it
does represent a source of uncertainty.
The human PBPK model (with the EPA's revised parameters) can be tested by comparing
the data of Jakubowski et al. (1987) with mass-balance calculations from the PBPK model
(Table D-3). Unfortunately, there are some significant discrepancies between the data of
Jakubowski et al. (1987) and these simulations. The respiratory retention (nominally in the lung)
of AN from subjects exposed through a face mask ranged from 44 to 58%, with an average of
51.8%. The PBPK model predicts 99% uptake of AN inhaled to the pulmonary region.
Recognizing that -30% of inhaled air only enters the conducting airways (so-called "dead
space"), the PBPK model effectively predicts a retention of 69%, which is considerably higher
than measured. The respiration rates of the subjects in the Jakubowski et al. (1987) experiment
ranged from 366 to 625 L/hour, with an average of 508 L/hour, which corresponds to an alveolar
ventilation rate of 356 L/hour, while the PBPK model utilized a rate of 300 L/hour (for a 70-kg
person), but increasing the respiration rate in the model by this amount (while holding all other
parameters constant) only decreases the alveolar uptake fraction from 98.7 to 98.5%. A possible
explanation is that AN is subject to a considerable "wash-in/wash-out" effect, where some of the
material inhaled deposits temporarily in the conducting airways and then desorbs on expiration,
which is not accounted for in the model.
D-5
DRAFT - DO NOT CITE OR QUOTE
-------
Table D-3. PBPK mass balance predictions for an 8-hour human exposure
to 2 ppm AN
Parameter
Metric
Amount (mg)
Percent
Al
AN mass inhaled
10.389
AX
AN mass exhaled
0.133
1.3
ATi
AN final tissue mass (including blood)
0.725
7.0
AMt
AN metabolized by ...
5.102
49.1
AMo
AN oxidized
3.904
37.6
AMgt
AN enzymatically conjugated (in liver)
0.280
2.7
AMgn
AN non-enzymatically conjugated
0.918
8.8
Metabolic mass balance
(AMo + AMgt + AMgn)/AMt
100.00%
ABD
AN bound to Hb and sulfhydrils
4.429
42.6
Total AN mass balance
(AX + ATi + AMt + ABD)/AI
100.00%
CMo
CEO formed from AN oxidation3
5.080
CX
CEO mass exhaled
0.001
0.0
CTi
CEO final tissue mass
0.097
1.9
CEH
CEO hydrolyzed
1.128
22.2
CC1
CEO conjugated in liver
2.596
51.1
Ccti
CEO conjugated in other tissues
0.283
5.6
CBD
CEO bound to Hb and sulfhydrils
0.975
19.2
CEO mass balance
(Cti + CEH + CC1 + Ccti + CBD)/CMo
100.00%
aCMo = AMo X MWceo - MWffl.
The subjects of Jakubowski et al. (1987) were then exposed in a chamber for 8 hours to
"3
an average of 10.8 or 5.6 mg/m (5 or 2.6 ppm), and the total excretion of CEO in the urine was
found to be an average of 26.4 or 16.3% of the retained dose (portion not exhaled) at those
respective exposure levels. The PBPK model predicts that 38% of the retained AN is oxidized to
CEO and that 22% of that CEO is hydrolyzed, with these fractions being fairly constant at
<8 ppm. Assuming that all the hydrolysis product was excreted in the urine, that the acid-
extraction method of Jakubowski et al. (1987) would have reversed the hydrolysis, and that the
extraction/HPLC technique would have separated that product from the GSH conjugates and
other metabolites, the predicted urinary excretion of "CEO" would then be 8.4% of the retained
dose. Alternately, if it is assumed that all CEO conjugated with GSH was excreted in the urine
and that the method of Jakubowski et al. (1987) had cleaved the GSH conjugates, then the total
CEO predicted in the urine would be 79% of the CEO formed or 30% of the retained AN.
Therefore, the measured CEO excretion as a fraction of the retained dose is bracketed by the
predicted levels, depending on what is assumed about the assay method. These results indicate
that the model predictions of AN and CEO metabolism in humans subsequent to absorption are
at least in the right range.
Kedderis et al. (1996) suggest that the overestimation of CEO by the rat model at the
early time points (which also occurs in the EPA's revision) may be due to an intrahepatic first
D-6
DRAFT - DO NOT CITE OR QUOTE
-------
pass effects, as occurs with other epoxides formed in situ from their parent olefins (Filser and
Bolt, 1984). The introduction of such a structure ("privileged access" as coined by Kohn and
Melnick, 2000) would contribute to the efficient clearance of CEO; however, whether the needed
differential reduction of CEO levels in the early phase would also be accomplished is not clear.
The issue is that the model predicts a very rapid rise for CEO in blood to peak/plateau levels
(Figure 3-5a, top panels), while the data show a gradual rise over the first -10 minutes. It is not
evident that a privileged access model would account for this dynamic difference since the
current model allows for reaction of AN with GSH in stomach tissue and liver before reaching
the general circulation. Further, the same lack-of-fit occurs vs. the CEO i.v.-exposure data
shown in Figure 3-3, which could not be explained by a first-pass effect from GI absorption.
An alternate explanation is that the model currently assumes constant GSH levels, while
GSH may be depleted during the early part of the exposure. A model that includes this depletion
could predict that at very short times when GSH is near control levels, relatively little AN would
remain unconjugated for conversion to CEO, but, as the depletion occurs, more and more AN
would be available for oxidative conversion to CEO, hence the gradual rise. To accurately
calibrate parameters that accomplish intrahepatic clearance would require a description of GSH
kinetics.
Thus, the inability to fit the rat blood CEO levels over the entire time course and the lack
of understanding of the biological basis for those kinetics indicate a level of uncertainty in the
model that could be addressed through further research. While the model's overprediction of
AN absorption in humans indicates that the description of gas uptake could be improved, this
discrepancy is only 10-30% (much less than the apparent 3- to 10-fold overprediction of blood
and brain CEO levels in rats and [likely] in humans by the oral route). However the model does
predict AN and CEO blood and tissue levels from inhalation exposure in rats at the lowest
concentration measured (186 ppm) quite well (Figure 3-4a), indicating that the gas-uptake
description is adequate for prediction of rat dosimetry, and an adjustment of the description in
humans would lower the predicted uptake and hence the extrapolated risk. Therefore, the use of
the model to predict CEO levels after AN inhalation in humans should be adequately protective.
For oral exposures, however, the greater accuracy of the AN predictions vs. CEO (Figure 3-5b
vs. 3-5a), suggests greater certainty in the use of AN, although the mode of action suggests that
CEO is the active metabolite.
Sensitivity and Uncertainty Analysis
Sensitivity and uncertainty analyses were performed to investigate the dependence of
PBPK model predictions for AN and CEO on specific model parameters and to estimate the
overall uncertainty in those predictions, given estimates of uncertainty in specific model
parameters.
D-7
DRAFT - DO NOT CITE OR QUOTE
-------
The first step was to perform a sensitivity analysis on PBPK model predictions to identify
the degree to which model dose metric predictions depend on model parameters. For this
purpose, normalized sensitivity coefficients were calculated. For example, the sensitivity of a
model prediction of concentration (e.g., AN concentration in blood) to parameter P (e.g.,
metabolic Vmaxc for AN oxidation), the normalized sensitivity coefficient is:
S = (dC/dP) x (P/C)
The first term is the derivative of C with respect to P. (C is generally a function of time, so this
derivative will also depend on time.) This derivative is approximated by alternately increasing
and decreasing P by a small fraction, AP (taken to be 1% of P here) and calculating [C(P + AP) -
C(P - AP)] (2 x AP). The second term normalizes the sensitivity with respect to the value of C
at the normal value of P and with respect to P, essentially yielding the percent change in C given
a 1% change in P. These sensitivities typically vary between ±1, where values near 1 show high
sensitivity and values near 0 show low sensitivity.
Sensitivities were determined for four dose metrics, the blood and brain concentrations of
AN and CEO, under an inhalation and an oral exposure scenario for the human PBPK model.
For simplicity, the EPA considered continuous exposures to either 0.4 ppm AN by inhalation or
an oral absorption rate equal to 1 mg/kg-day and analyzed the values of these metrics at
24 hours, by which time, the human body is predicted to reach steady state. Under such
conditions, the AUC for each metric is just 24 hours times the steady-state concentration
(subsequent to reaching steady state), and hence, the sensitivity of the AUC is identical to the
sensitivity of the concentration, given that the coefficients are normalized as described above.
Since these concentrations are in the linear range of the model, they apply across the low-dose,
linear range of concentrations. The sensitivity coefficients for each parameter for which at least
one coefficient had absolute value greater than 0.1 are listed in Tables D-4 and D-5 for the
inhalation and oral scenarios, respectively.
D-8
DRAFT - DO NOT CITE OR QUOTE
-------
Table D-4. Sensitivity of AN and CEO metrics for continuous inhalation
exposure
Parameter
Dose metrics" (concentrations)
AN in blood
AN in brain
CEO in blood
CEO in brain
Qcc
-0.49
-0.57
0.51
0.64
Qpc
0.99
0.99
0.99
0.98
VLC
-0.01
-0.01
-0.64
-0.64
Vbrc
-0.01
-0.04
-0.04
-0.14
Vsc
-0.14
-0.10
-0.28
-0.28
Vabc
-0.07
-0.07
-0.12
-0.15
VvBC
-0.12
-0.08
-0.20
-0.20
Qlc
-0.76
-0.53
0.65
0.65
Qbrc
0.00
0.03
0.00
0.10
Qsc
-0.02
-0.01
-0.10
-0.10
gshl
0.00
0.00
0.27
0.38
GSHbr
-0.01
-0.04
-0.04
-0.14
GSHs
-0.13
-0.09
-0.27
-0.27
VmaxC
-0.09
-0.06
0.11
0.11
Km
0.09
0.06
-0.11
-0.11
kFC2
0.00
0.00
-0.86
-0.96
kEHC
0.00
0.00
-0.25
-0.25
kBC
-0.19
-0.16
-0.16
-0.16
kBC2
0.00
0.00
-0.13
-0.15
Pbr
0.00
0.97
0.00
0.00
PbR2
0.00
0.00
0.00
1.00
"Values are normalized sensitivity coefficients (i.e., (dC/dP) x (P/C), for sensitivity of concentration C to parameter
P) for steady state blood and brain concentrations predicted for humans during an exposure to 0.4 ppm AN.
Sensitivities are approximated by increasing and decreasing each parameter by 1% of its default value. Values of
0.2 or greater are indicated in bold and values between 0.1 and 0.2 in italics. Parameters for which all coefficients
are less than 0.1 are omitted.
D-9
DRAFT - DO NOT CITE OR QUOTE
-------
Table D-5. Sensitivity of AN and CEO metrics for continuous oral exposure
Parameter
Dose metrics" (concentrations)
AN in blood
AN in brain
CEO in blood
CEO in brain
BW
0.21
0.20
0.24
0.24
Qcc
0.30
0.36
0.27
0.39
VLC
-0.07
-0.07
-0.64
-0.64
Vbrc
-0.01
-0.04
-0.03
-0.13
Vsc
-0.10
-0.10
-0.19
-0.19
V VBC
-0.12
-0.12
-0.13
-0.13
Qlc
0.29
0.29
0.35
0.35
gshl
0.00
0.00
0.27
0.38
GSHbr
-0.01
-0.04
-0.03
-0.13
GSHs
-0.09
-0.09
-0.18
-0.18
VmaxC
-0.89
-0.89
0.11
0.11
Km
0.89
0.89
-0.11
-0.11
kFC2
0.00
0.00
-0.86
-0.96
kEHC
0.00
0.00
-0.25
-0.25
kBC
-0.17
-0.19
-0.01
-0.01
kBC2
0.00
0.00
-0.13
-0.15
Pbr
0.00
0.97
0.00
0.00
PbR2
0.00
0.00
0.00
1.00
"Values are normalized sensitivity coefficients (i.e., (dC/dP) x (P/C), for sensitivity of concentration C to parameter
P), for steady state blood and brain concentrations predicted for humans during a continuous oral infusion equal to
1 mg/kg-d. Sensitivities are approximated by increasing and decreasing each parameter by 1% of its default value.
Values of 0.2 or greater are indicated in bold, and values between 0.1 and 0.2 in italics. Parameters for which all
coefficients are less than 0.1 are omitted.
The EPA's results can be compared to those of Sweeney et al. (2003), derived for the
original model, with previous parameter values. For the inhalation scenario (Table D-4), the
EPA's results for AN are essentially identical to those of Sweeney et al. (2003), and those for
CEO show the same qualitative pattern and only have a few notable differences. The sensitivity
to the rate constant for CEO hydrolysis (kEHC in the EPA's model; VmaxC2 in Sweeney et al.,
2003) is -0.25 in the revised model and -0.5 in their previous model, indicating somewhat less
importance for this parameter now. On the other hand, the sensitivity to CEO-GSH conjugation
in the liver parameterized by kFc2 increased from -0.6-0.7 for Sweeney et al. (2003) to -0.9 with
the revised model, showing more significance now. Thus, there is a shift in the relative
importance of these two parameters and the rate constant for GSH conjugation in humans is an
especially important factor in the EPA's predictions, while the exact rate of CEO hydrolysis is
less influential, though still somewhat important.
D-10
DRAFT - DO NOT CITE OR QUOTE
-------
Important factors in comparing the EPA's results for an oral exposure to those of
Sweeney et al. (2003) are that the EPA held the dose in mg/kg-day constant, so that it varied in
proportion to BW when that parameter was varied, while Sweeney et al. (2003) held the dose in
mg/day constant, and they calculated sensitivities of the AUC and peak concentrations for only
the AN metrics, while the EPA used steady-state concentrations of AN and CEO. Thus, it is not
surprising that Sweeney et al. (2003) had a negative sensitivity to BW, since the same total dose
given to a larger body would result in a lower concentration, while the EPA's is positive,
reflecting that a proportionately higher dose is expected to yield higher-than-proportional
0 7
concentrations because the rates of metabolic elimination scale as BW
It is interesting to note that the liver volume (fraction), which effects the rates of first-
and second-order reactions in the model, and the rate constants for CEO removal but not CEO
production (Vmaxc) significantly affect the CEO metrics (absolute values from Sweeney et al.
(2003) are about 0.17-0.18 for sensitivity of CEO metrics to Vmax and Km). The EPA's lack of
significant dependence on blood flow rates to stomach (Qstc) and slowly-perfused tissues (Qsc)
is due to the fact that the EPA is analyzing a continuous infusion, as these do significantly affect
the peak concentration after a bolus exposure. Thus, these are only important for specific dosing
scenarios.
Likewise the rate of absorption (kA) from the stomach significantly affects the peak
concentration (verified but not shown for the revised model) but has a much smaller affect on the
AUC. These results for absorption are not surprising, since changes in kA do not alter the fact
that the entirety of an oral bolus is predicted to be absorbed. Whether or not there is an impact
on the steady-state concentrations depends on what one believes about oral absorption. In
particular, if one assumes 100% absorption, where the rate of absorption increases with the
amount in the GI tract, then at steady state, the rate of absorption must equal the rate of ingestion
and the exact rate of absorption is unimportant. It is only if orally ingested material may be
eliminated in the feces or otherwise transformed in the gut without absorption that the exact rate
constant for absorption would be important.
Finally, uncertainty in these steady-state model predictions was estimated using the
equation of Sweeney et al. (2003) for the approximate coefficient of variation (CV):
cvm = WJsJ^k)
where CVm is the estimated CV for metric "m," Smi is the sensitivity coefficient of metric "m" to
parameter p,, as tabulated above, and CV; is the CV of p;. For this calculation, the EPA decided
to use the individual parameter CV values as estimated and used by Sweeney et al. (2003, see
Table 3-5 in that paper). The CVs for physiological parameters (e.g., blood flow rates, tissue
fractions) would indeed be identical, and, while the EPA reestimated some metabolic parameters
D-ll
DRAFT - DO NOT CITE OR QUOTE
-------
and kA, the underlying in vitro data on which the human extrapolation is based are the same, and
hence, they are expected to be approximately the same.
The resulting CV values for the EPA's four metrics under the two exposure scenarios are
listed in Table D-6.
Table D-6. Estimated CVs for AN and CEO metrics
Exposure scenario
Dose metrics (concentrations)
AN in blood
AN in brain
CEO in blood
CEO in brain
0.4 ppm inhalation
0.620
0.656
0.893
1.15
1 mg/kg-d oral
0.790
0.873
0.711
1.01
The overall CVs in Table D-6 are interesting in themselves, in that they indicate the
overall level of confidence in the model's prediction of those dose metrics. But even more
interesting is the contribution of the individual parameters to each of these metrics, as listed in
Tables D-7 and D-8. What those contributions show is that the vast majority of the uncertainty
arises from a small number of parameters and that most of those parameters are physiological
values—alveolar ventilation (Qcc), cardiac output (Qpc), and blood flow to the liver (Vlc)—and
the brain:blood PC in the case of the brain metrics.
D-12
DRAFT - DO NOT CITE OR QUOTE
-------
Table D-7. Parameter contributions to overall CVs for inhalation exposure
Parameter
Dose metrics" (concentrations)
AN in blood
AN in brain
CEO in blood
CEO in brain
Qcc
9.3
11.5
5.0
4.6
Qpc
37.9
33.9
18.2
11.0
vLC
0.0
0.0
7.7
4.6
Vbrc
0.0
0.1
0.1
0.5
Vsc
0.7
0.3
1.4
0.9
Vabc
0.4
0.4
0.6
0.5
VvBC
1.1
0.5
1.5
0.9
Qlc
44.4
19.8
15.9
9.6
Qbrc
0.0
0.1
0.0
0.2
Qsc
0.0
0.0
0.4
0.2
gshl
0.0
0.0
1.1
1.3
GSHbr
0.0
0.2
0.1
0.7
GSHs
0.9
0.4
1.8
1.1
VmaxC
1.1
0.5
0.7
0.4
Km
0.4
0.2
0.3
0.2
kFC2
0.0
0.0
39.2
29.8
kEHC
0.0
0.0
3.4
2.1
kBC
2.9
1.8
1.0
0.6
kBC2
0.0
0.0
0.7
0.5
Pbr
0.0
30.0
0.0
0.0
PbR2
0.0
0.0
0.0
29.9
aValues are 100 x Smk2x CVk/S(Smi2x CVO- Values >5 (i.e., 5% contribution) are in bold.
D-13 DRAFT - DO NOT CITE OR QUOTE
-------
Table D-8. Parameter contributions to overall CVs for oral exposure
Parameter
Dose metrics" (concentrations)
AN in blood
AN in brain
CEO in blood
CEO in brain
BW
1.1
0.8
1.7
0.8
Qcc
2.2
2.5
2.1
2.3
VLC
0.1
0.1
12.1
6.0
Vbrc
0.0
0.1
0.1
0.5
Vsc
0.2
0.2
1.0
0.5
V VBC
0.7
0.5
0.9
0.5
Qlc
4.1
3.4
7.1
3.5
gshl
0.0
0.0
1.8
1.7
GSHbr
0.0
0.1
0.1
0.8
GSHs
0.3
0.2
1.3
0.6
VmaxC
66.1
54.1
1.2
0.6
Km
23.1
18.9
0.4
0.2
kFC2
0.0
0.0
61.9
38.6
kEHC
0.0
0.0
5.4
2.7
kBC
1.4
1.5
0.0
0.0
kBC2
0.0
0.0
1.0
0.7
Pbr
0.0
16.9
0.0
0.0
PbR2
0.0
0.0
0.0
38.7
"Values are 100 x Smk2x CVk/S(Smi2x CV;). Values >5 (i.e., 5% contribution) are in bold.
Since the PCs are expected to be similar in rats and humans, and the extrapolation really
depends on the rat:human ratio of those, the true resulting uncertainty is likely to be very small.
The nominal uncertainty in cardiac ventilation and blood flows is expected to be a composite of
uncertainty and inter-individual variability. And the only metabolic parameter to contribute
significantly to the uncertainty for inhalation exposure is the CEO-GSH conjugation rate. Thus,
a significant improvement in the overall uncertainty can be gained by further investigation of
only a small number of parameters, half of which are applicable to every PBPK model that might
be considered for gases.
The distribution of parameter influence for the oral simulation scenario, as shown in
Table D-8, is similar to that for inhalation in that most of the influence is distributed among a
few parameters, with the brain:blood PCs strongly influencing brain concentrations. Also similar
is that the CEO-GSH conjugation rate (kici) is a significant factor for the CEO metrics. Unlike
the inhalation case, alveolar ventilation (Qpc) has negligible influence, which is to be expected,
but also cardiac output (Qcc) has only small influence. This later case occurs because of the
strong first-pass effect for AN as it is absorbed through the liver, which also explains the very
high influence of the oxidation parameters, Vmaxc and Km, on the AN metrics.
D-14
DRAFT - DO NOT CITE OR QUOTE
-------
These results for oral exposure are quite similar to the results obtained by Sweeney et
al. (2003), except that their results indicate somewhat higher influence by blood flow to the liver
(i.e., 8-12% on AN metrics), liver GSH (6%), and a much higher influence of CEO hydrolysis
(kEHC here, parameterized by a Vmax and Km with influence of 12-30%) on CEO metrics. The
shift of influence from hydrolysis to GSH conjugation for CEO metrics between Sweeney et al.
(2003) and the EPA's results arises from the shift in the relative rates through those two
pathways, but in both, the rate of CEO removal is a strong determinant of the CEO metric.
D-15
DRAFT - DO NOT CITE OR QUOTE
------- |